environmental contaminants in food - office of technology

Environmental Contaminants in Food

December 1979

NTIS order #PB80-153265

Library of Congress Catalog Card Number 79-600207

For sale by the Superintendent of Documents, U.S. Government Printing OfficeWashington, D.C. 20402 Stock No. 052-003-00724-0

FOREWORD

This report presents the major findings of an OTA assessment of Federal andState efforts to deal with the environmental contamination of food. Undertaken atthe request of the House Committee on Interstate and Foreign Commerce, thestudy examines both regulatory approaches and monitoring strategies for copingwith contaminated food.

The assessment is concerned with chemical and radioactive contaminantsthat inadvertently find their way into the human food supply. To bring the scope ofinquiry within manageable bounds, we excluded naturally occurring toxins suchas fungal and microbial toxins.

The Office of Technology Assessment was assisted by two advisory panels ofscientists and representatives of public interest groups, agriculture, the chemicalindustry, fisheries, and State and foreign governments. The Food and Drug Ad-ministration, the Department of Agriculture, and the Environmental ProtectionAgency each designated staff members to attend panel meetings, provide back-ground information, and review draft reports, Background papers were commis-sioned concerning the scientific aspects of detecting and regulating environ-mental contaminants in food. The Congressional Research Service provided fiveanalyses of previous food contamination episodes, Reviews of the draft reportwere provided by the advisory panels, Federal agencies, and a number of inter-ested individuals not previously involved with the assessment.

Because this assessment addresses concerns of American citizens as well aspolicy makers and scientists, the summary of the report is also being published asa separate document. The summary provides the interested citizen with an in-formative and clear overview of this complex problem. Copies of the summary canhe obtained free of charge from the office of Technology Assessment, U.S. Con-gress, Washington, D.C. 20510.

JOHN H. GIBBONSDirector

///

.-

Tolerance Advisory Panel

James AndersonManager for Material SafetyAllied Chemical Corporation

Louise BurkeExecutive DirectorConsumer Council of Virginia FoundationConsumer Council of Virginia, Inc.

Kenneth CrumpDepartment of Mathematics and StatisticsLouisiana Tech University

Donald EppDepartment of Agricultural Economics and

Rural SociologyPennsylvania Stale [University

Frederick HalbertDairy FarmerDelton, Mich.

Joseph HighlandChairmanToxic Chemicals ProgramEnvironmental Defense Fund

Monitoring Advisory Panel

Raymond AllredDirectorEnvironmental Conservation ResearchContinental Oil Company

Richard AndersonChairmanStatistics DepartmentUniversity of Kentucky

John BourkeDirectorAnalytical DivisionFood Science DepartmentCornell University

Vaughan BowenWoods Hole Oceanographic Institute

Shirley BriggsExecutive DirectorRachel Carson Trust for the Living

Environment, Inc.

Leon ChesninDepartment of AgronomyUniversity of Nebraska

Robert JacksonAssistant State Health CommissionerDirector, Office of Health Protection and

Environmental ManagementVirginia State Health Department

Kazuo K, KimuraSchool of MedicineWright State University

Marvin LegatorDirectorDivision of Environmental ToxicologyDeportment of Preventive Medicine and

Community HealthUniversity of Texas Medical Branch

David LindsayTechnical SecretarySteering Group on Food SurveillanceMinistry of Agriculture, Fisheries, and

FoodUnited Kingdom

Roy Mart inDirectorScience and TechnologyNational Fisheries Institute, Inc.

William CooperProfessorDepartment of ZoologyCollege of Natural ScienceMichigan State University

John DaviesChairmanDepartment of Epidemiology and Public

HealthSchool of MedicineUniversitv of Miami

Blake EarlyEnvironmental Action, Inc.

Edward ElkinsDirectorChemistry DivisionWashington Research LaboratoryNational Food Processors’ Association

Ronald J. HingleAdministratorFood and Drug Control UnitBureau of Environmental ServicesLouisiana Office of Health Services and

Environmental Quality

Richard MerrillSchool of LawUniversity of Virginia

Joseph RosenDepartment of Food ScienceRutgers University

John RustProfessor EmeritusPharmacology/RadiologyUniversity of Chicago

Gary Van GelderCollege of Veterinary MedicineUniversity of Missouri

John Van RyzinRAND Corporation

George WhiteheadDeputy DirectorOffice for Consumer AffairsMichigan Department of Agriculture

S. Roy KoirtyohannEnvironmental Trace Substances CenterUniversity of Missouri

Charles KristerRetired ManagerBiochemicals DepartmentE. I. DuPont de Nemours and Company

John LaseterDirectorCenter for Bio-Organic StudiesUniversity of New or leans

John MartinMoss Landing Marine Laboratories

Rebecca SharitzSavannah River Ecology LaboratoriesUniversity of Georgia

Gerald WardDepartment of Animal ScienceColorado State University

Iv

Food and Renewable Resources Program Staff

Joyce C. Lashof, Assistant DirectorHealth and Life Sciences DivisionJ. B. Cordaro, Program Manager*

Walter Parham, Program Manager

Catherine Woteki, Project DirectorRobert Smith, Research Assistant

Chris Elf ring, OTA Congressional Fellow

Administrat ive Staff

Phyllis Balan Elizabeth Galloway Ogechee Koffler

Contractors

Lloyd W. Hazleton, Technical Advisor to Tolerance PanelRobert Huggett, Technical

Samuel

OTA Publishing Staff

Advisor to Monitoring PanelIker, Editor

John C. Holmes,

Kathie S. Boss

Food Advisory Committee

Martin E. Abel, Chairmen Lorne Greene

Publishing Officer

Joanne Heming

Robert 0. NesheimSenior Vice President Chairman of the Board Vice President, Science and

Schnittker Associates

Johanna Dwyer, Co-ChairmanDirectorFrances Stern Nutrition Center

David CallDean, College of Agricultural

and Life SciencesCornell University

Cy CarpenterPresidentMinnesota Farmers Union

Eliot ColemanDirectorCoolidge Center for the

Study of Agriculture

Almeta Edwards FlemingSocial Program CoordinatorFlorence County, S.C.

American Freedom FromHunger Foundation

Richard L. HallVice President, Science and

TechnologyMcCormick & Company, Inc.

Laura HeuserMember, Board of DirectorsAgricultural Council of America

Arnold MayerLegislative RepresentativeAmalgamated Meat Cutters and

Butcher Workmen of NorthAmerica

Max MilnerExecutive DirectorAmerican Institute of Nutrition

TechnologyThe Quaker oats Company

Kathleen O’ReillyDirectorConsumer Federation of America

R. Dennis RouseDean, School of AgricultureAuburn University

Lauren SethAgricultural Consultant

Thomas SporlederProfessor of Agricultural

EconomicsTexas A&M UniversitySylvan WittwerDirector and Assistant DeanCollege of Agriculture and

Natural ResourcesMichigan State University

*Resigned June 16, 1979, and continued to serve under contract to assessment completion.

ContentsChapter Page

I. Summary . . . . . . . . . . . . . . . . . . . `“ “ “ “ ● ● ● “ ● 00 G o “ ● “ o ● ● o ● o ● “ o ● 00 “ ● “ o “ “ “Health Impacts of Environmental Contamination. . . . . . . . . . . . . . . . . . . . . . .Economic Impacts of Contamination. . . . . . . . . . . . . . . . . . . ““ .“ . ““. ““. “ .“Major Problems in Identifying Environmental Contaminants . . . . . . . . . . . . .Problems of Regulating Environmental Contaminants. . . . . . . . . . . . . . . . . . .Findings and Conclusions . . . . . . . . . . . . . . . . . “ “ “ “ . “ “ “ . m “ “ “ o “ o “ . “ . “ “ “ “Congressional Options. . . . . . . . . . . . . . . ..” . . “ “ “ “ .0. “ o ‘. “ c o “ o “ “ o “ “ “ ““

Option l.--Maintain the Present System . ..........”””’ . . . .. .”.”””.Option2.—Amend the Food, Drug, and Cosmetic Act . . . . . . . . . . . . . . . . .Option 3 .—Establish an Investigatory Monitoring System. . . . . . . . . . . . . .Option 4.— Improve Federal Response to New Contamination Incidents . .

II. Environmental Contamination of Food.”~””.”. .“””.’””ooooQo ● +00 ”00”0How Food Becomes Contaminated . .........””””” ....”””..”.... .“””Magnitude of the problem . . . . . . . . . . . . . . . . . . . . .“.”””””.”..”. “...”

Evidence of Human Illness Resulting From Consumption ofContaminated Food. . . . . . . . . . . . . . . . . . ...””””””.””.” “o” ”””..

Polychlorinated Biphenyls. . . . . . . . . . . . . . . . . . “ ~ . “ “ m “ “ ~ 0. ~ 0. “ “ o ~ .Polybrominated Biphenyls . ..........”””” ..”””.””” .“.”.”.”””.”Mercury and Methylmercury . . . . . . . . . . . . . . . . . “ . “ “ ~ ~ .“ “ “ ~ “ . “ “ 0.Radioactivity . . . . . . . . . . . . . . . . . . . . “ “ “O 00 “ ““ ~ s o “ ~ 000 s Q “ o “ “ s “ o “

Number of Food Contamination Incidents. . . . . . . . . . . . . . . . “ . ““” ~ “ “ “.”Economic Impact . . . . . . . . . . . . . . . . ..””””.”” “0”””00.0””””. ““s”””

Costs of Food Condemned. . . . . . . . . . . . . . . ..”.”. “. ““””.””””.””Health Costs... . . . . . . . . .....,..””””.” 0“0.o’0”””””00 ““”””””..Distributional Effects and Costs . . . . . . . . . . . . . . . “ ““ “ . “ . “ “ “ “ “ .0 “ ““

Potential Food Contamination problems. . . . . . . . . . . . . . . .“ .“ “ . “. ‘“” ““ “ “Conclusions and Issues . . . . . . . . . . . . . . . “ . “ “ ~ “ . ~ ~ . ~ “ ~ . “ “ “ “ o “ “ o ~ o “ s “ “Chapter II References . . . . . . . . . . . . . . . . . . . “ “ “ ~ “ . . . “ 0.0. s o “ 0000. s “ “ .

111. Federal Laws, Regulations, and Programs” . . . . . . . . . . “ “ “ “ “ 00 ● o “ o “ ● 0000Federal Laws. . . . . . . . . . . . . . . . . “ “ .“” “ “ ““ “O . “ “ “ 000 “ “.” ~ o ‘ . “ 00 “ .- e oAction Levels and Tolerances. . . . . . . . . . . . . . “ ‘ “ “ “ “ “ ~ “ “ ~ “ “ “ ~ o “ o “ “ o “ “ ~

Criteria forgetting Action Levels and Tolerances . . . . . . . . . . . . . . . . . . . .Determining Action Levels and Tolerances . . . . . . . . . . . . . . “ . . “ . “ “ “ “ . .

Evaluation of the Scientific Data. . . . . . . . . . . . . . . . .“” . ““ “ ““ ~ ““” “ “ “Evaluation of Decisionmaking Approach . . . . . . . . . . . . . . ““ . .“ “.” . “ “

Federal Monitoring programs. . . . . . . . . . . . . . . . . “ . “ “ . “ “ 000 “ “ s m “ . “ “ “ “ ~Food and Drug Administration . . . . . . . . . . . . . . . “ “ . “ “ “ “ “ “ “ “ .0 “ o ‘ “ “ “ .U.S. Department of Agriculture . . . . . . . . . . . . . . . “ . “ “ ‘ “ “ “ “ “ “ “ o “ “ 0.00Environmental Protection Agency . . . . . . . . .......”..”.”.o ““” ”s”s.”sCritique of Federal Monitoring programs. . . . . . . . . . . . . . . . . . . . “ . . . “ . “

Chapter III References . . . . . . . . . . . . . . ““” ““”” ~ “ “ “ .’. o “ .0 “ . .“ s . “ o .“ “

V. Methods for Assessing Health Risks.....”“”””OCO””OOOO 0“00”00”0Possible Toxic Effects . . . . . . . . . . . . . . . . . . ..””.”””.””””.” .“”..”””””.Human Epidemiology. . . . . . . . . . . . . . . . ....”.”.o.o’s ““0”””””.”.”00Animal Tests and Other Methods for Toxicological Testing . . . . . . . . . . . . . .

The Carcinogen Bioassay. . . . . . . . . . . . . . . “ . ““ “ “ “ “ .0 “ “ 000. c “ .0.....Short-Term Tests for Mutagenesis and Potential Carcinogenesis. . . . . . . .

344556999

1010

151619

19192021222325252627292930

35353738393941424243444546

4 94950525255

595960626263

vii

Contents--continued

ChapterInteractions of Two or More Substances . . . . . . . . . . . . . . . . . . . . . . . . . . .Extrapolating From High Doses to Low Doses. . . . . . . . . . . . . . . . . . . . . . . .Chapter Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

VI.

VII.

VIII.

Ix.

Methods of Estimating and Applying Costs to Regulatory Decisionmaking . .Cost Effectiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Cost-Benefit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Application of Methods for Regulation, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Chapter VI References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Monitoring Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Strategies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Regulatory Monitoring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Investigatory Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..., , . . . . . .

Monitoring for Suspected Environmental Contaminants . . . . . . . . . . . . .Monitoring for Uncharacterized Environmental Contaminants . . . . . . . .

Specimen Banking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 . . . . . . . . . . . . . . .Quality Assurance. ,... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Chapter VII References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Monitoring Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Organic Environmental Contaminants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Analysis ., ..., . . . . . . . . . . . . . . . . , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Application to Investigatory Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Detecting and Quantifying Trace Metal Contaminants . . . . . . . . . . . . . . . . . .Analysis ....., . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Application to Investigatory Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . .Speciation of Trace Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Analysis of Foods for Radioactivity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Applications to Investigatory Monitoring. . . . . . . . . . . , . . . . . . . . . . . . . . .

Estimated Cost to Equip a Laboratory to Conduct Investigatory Monitoring. .Chapter VIII References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Congressional Options. .. .. .. .. .. .. .. ... ... ... ....$ . . . . . . . . . . . . . . .Option 1—Maintain the Present System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Option 2—Amend the Food, Drug, and Cosmetic Act. . . . . . . . . . . . . . . . . . . .

Option 2A--Simplify Administrative Procedures. . . . . . . . . . . . . . . . . . . . .Option 2B-Requiret reestablishment of Tolerances . . . . . . . . . . . . . . . . .Option 2C-Clarify the Role of Economic Criteria . . . . . . . . . . . . . . . . . . . .Option 2D-Establish Regional Tolerances . . . . . . . . . . . . . . . . . . . . . . . . .

Option 3-Establish an Investigatory Monitoring System . . . . . . . . . . . . . . . .Option 3A-Establisha National Monitoring System for Suspected

Environmental Contaminants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Option 3B-Establisha National Monitoring System for Uncharacterized

Environmental Contaminants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Option 4—Improve Federal Response to New Contamination Incidents . . . . .

Option 4A-Designate a Lead Agency . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Option 4B-Establish a Center to Collect and Analyze Toxic Substances

Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

AppendixesA. Substances Whose Production or Environmental Release Are Likely to Increase

in the Next Ten Years . . . . , . . , . . . . . . . . . . , , . . . , . , . . , , . . . . . , . . . , . , . . . . . .B. FDA Derivation of PCB Tolerance—Federal Register. . ., , . . . . . . . . . . . . . . . . . . .

Page656769

7373767778

81818181828485868889

93939396979798989999

102103105

109109110110110111111112

112

113113113

114

119125

.,.Vlll

Contents--continued

Appendix Page

C. Methods for Toxicologic Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137D. Review and Evaluation of Methods of Determining Risks From Chronic Low-Level

Carcinogenic Insult . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154E. Measuring Benefits and Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166F. Priority Setting of Toxic Substances for Guiding Monitoring Programs . . . . . . . . . . 177G. Approaches to Monitoring Organic Environmental Contaminants in Food . . . . . . . 187H. Analytical Systems for the Determination of Metals in Food and Water Supplies . 195I. Analysis of Foods for Radioactivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215J. Commissioned Papers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228K. Executive Agency Observers at Advisory Panel Meetings. . . . . . . . . . . . . . . . . . . . `229

LIST OF TABLES

Table No. Page1.

2.3.

4.5.6.7,8.9.

10.11.12.

13.14.

15.16.17.18.l9.

20.21,

A-].D-1.

G-1.G-2.

G-3.

H-1.H-2.

1617

171822263650

515151

5266

686974839598

101104123

164188

192199

201

Contents--continued

Table No. Page

H-3,

H-4.H-5.H-6,I-1.

Multielement Atomic Emission Determination Limits for Two Common PlasmaExcitation Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204Determination Limits for X-Ray Fluorescence Techniques. . . . . . . . . . . . . . . . . . 205List of Sample Preparation Facilities Required . . . . . . . . . . . . . . . . . . . . . . . . . . 209Summary of Cost-Productivity Estimates for Various Analytical Techniques . . . 210Costs for Space, Furnishings, Equipment, and Supplies .,,..............,.. 225

LIST Of FIGURES

Figure No. Pagel. Distribution of PCB Contamination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82. Distribution of States Responding to Questionnaire . . . . . . . . . . . . . . . . . . . . . . . 243. State Food Laws. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494. Scope of Investigation: Distribution of Materials . . . . . . . . . . . . . . . . . . . . . . . . . 535. Steps in the Ames Spot Test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656. A General Scheme for the Qualitative and Quantitative Analysis of the

Organic EPA Priority Pollutants in Semisolid Foods . . . . . . . . . . . . . . . . . . . . . . . 947. Comparison of the Gas Chromatographic Analysis of a Standard Mixture of

Polychlorinated Biphenyls on Two Types of Gas Chromatographic Columns . . . 96D-I, Illustration of Why the Dose-Response Curve Should Be Linear at Low Dose in

the Presence of Background Carcinogenesis ... , ... , ., . . . . . . . . . . . . . . .D-2. Typical Fit of Mantel-Bryan Curve to Experimental Data , . . . . . . . . . . . . . .D-3. Comparisons of “Safe” Doses Computed From the Mantel-Bryan Procedure

and From a Procedure Based on the Multistage Model . . . . . . . . . . . . . , . . .D-4a, Example of Data Exhibiting Downward Curvature . . . . . . . . . . . . . . . . . . . .D-4b, Example of Data Exhibiting Upward Curvature. . . . . . . . . . . . . . . ., , . . . . .E-1 .G-I .

G-2.

G-3.

H-1 .1-1.

Analysis of a Shift in Supply .-. , . . . . . . . . . . . ... , . . . . . . ... , ., , ., . . . . .A Simplified Diagram for the Qualitative and Quantitative Analysis of theOrganic EPA Priority Pollutants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... , .A General Scheme for the Qualitative and Quantitative Analysis of the EPAOrganic Pollutants in Semisolid Foods . . . ... , . . . . . . . . . . . . . . . . . . . , . . .

, . . 156. . . 158

. , , 161

. . . 162

. , . 162

.0. 170

. , . 189

, . , 189General Analytical Scheme To Detect and Monitor New Trace Organics andPriority Pollutants in Food and Water Samples . . . . . . . . . . . . . . . . . . . . . . . . . . 192Flow Diagram of the General Operational Protocol of a Monitoring Laboratory 197Comparison of Gamma Spectra Taken With a Sodium Iodide Detector and aGermanium Diode Detector.........,..,.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

— . . .—

\

Summary

I

— — — .—.J

.

Chapter 1

Summary

The environmental contamination of food is a nationwide problem. Anumber of recent incidents dramatically illustrate the potential health haz-ards and economic harm that can be caused by such contamination-animalfeeds in Michigan contaminated by polybrominated biphenyls (PBBs), theHudson River contaminated by polychlorinated biphenyls (PCBs), and Vir-ginia's James River contaminated by kepone.

These are some of the more serious of the 243 food contamination inci-dents identified in an OTA survey of the 50 States and 10 Federal agencies.These incidents have occurred in every region of the country. They have in-volved all categories of food. While the OTA survey clearly shows the nationalcharacter of such contamination, the true extent of the problem is stillunknown.

The latest major food contamination incident—one not included in theOTA survey-graphically points up the ominous dimensions of the problem.PCBs from a damaged transformer contaminated animal fats at a packingplant in Billings, Mont. The plant used the adulterated fats to produce meatand bone meal that were sold both to feed manufacturers and directly tofarmers. The contaminated feed spread through at least 10 States—pollutingpoultry, eggs, pork products, and a variety of processed foods (includingstrawberry cake). The result: contaminated food found in 17 States, and humdreds of thousands of pounds of food products seized or destroyed.

This assessment, undertaken at the request of the House Committee onInterstate and Foreign Commerce, examines the adequacy of current Federaland State efforts to deal with the environmental contamination of food. Inparticular, the study evaluates the effectiveness of 1) Federal and Statemonitoring systems in detecting contamination episodes before they reachcrisis proportions, and 2) Federal efforts to regulate contaminations. Thestudy explores alternative approaches to the problem and presents policy op-tions for Congress.

Environmental contaminants in food fallinto three categories—synthetic or naturalorganic chemicals, metals or their organicand inorganic derivatives, and natural or syn-thetic radioactive substances. Such contam-inants are regulated under the Federal Food,Drug, and Cosmetic Act. To regulate them un-der the law, the Food and Drug Administra-tion (FDA) defines environmental contami-nants as “added, poisonous, or deleterious”substances that cannot be avoided by good

manufacturing practices, and that may makefood injurious to health.

Unlike food additives, environmental con-taminants inadvertently find their way intothe human food supply (including sports fishand game). They can enter food directly or in-directly as a result of such human activitiesas agriculture, mining, industrial operations,or energy production. In no instance is theirpresence in food ever intended.

3

Four factors determine whether and howseriously the environmental contamination offood will affect human health: the toxicity ofthe contaminant, the amount of the substancein the food, the amount of the contaminatedfood eaten and the physiological vulnerabili-ty of the individual or individuals consumingthe food.

Based on other countries’ experiences,there is considerable evidence of human ill-ness caused by the consumption of food con-taining various organic chemicals andmetals. In such cases, the level of the con-taminant in food exceeded the levels usuallyfound in the U.S. food supply. The effects ofmercury poisoning are well-documented. Thebest known case involved the consumption ofmercury-contaminated fish from Japan’s Min-amata Bay. Some of the offspring of exposedmothers were born with birth defects, andmany victims suffered central nervous sys-tem damage.

Another incident in Japan stemmed fromthe inadvertent contamination of rice oil byPCBs. The consumption of food cooked withthis oil resulted in 1,291 cases of so-called“Yusho disease”—a condition marked bychloracne (a severe form of acne), eye dis-charges, skin discoloration, headaches, fa-

tigue, abdominal pains, and liver and men-strual disturbances,

No such mass-poisoning episodes have oc-curred in the United States. But there arestudies indicating that present levels of someenvironmental contaminants may cause phys-iological changes. For example, the acciden-tal contamination of animal feed in 1973 ex-posed most of the population of Michigan toPBB in dairy products and other foods. Evi-dence on what impact this exposure has hadon human health is conflicting, although somedisparities in white blood cell function havebeen noted in farm families, The long-termsignificance of these physiological changes isnot yet known.

The clinically obvious harmful health ef-fects of radiation are usually associated withmassive, high-level exposures. Past cases ofradioactive contaminated foods have in-volved relatively small amounts of radioac-tive substances with low dose rates. General-ly, the young are most sensitive to radiationexposure. However, since any amount of radi-ation is potentially harmful, prudent publicpolicy must assume that any unnecessary ex-posure to high-energy radiation should beavoided.

The economic impacts of a contaminationincident have traditionally been stated interms of the estimated dollar value of the re-sulting food loss. Only limited data on suchcosts are available. Dollar value estimates forcondemned food were available for less than30 percent of the contamination episodesnoted in the OTA survey. Thus, the real costof environmental contamination of food dur-ing the 1968-78 decade is at least severaltimes the $282 million reported to OTA.

The loss of food only partially reflects thetotal economic impact of environmental con-tamination, Health and “distributional” costsare also involved. The health costs include

medical expenses and lost workdays from ill-ness resulting from food contamination in-cidents. Since the health effects may not im-mediately be evident, the expected illnessesor deaths from an episode are usually esti-mated on the basis of available toxicity datafor a particular contaminant. In other words,estimated health costs are more likely to beprojections than actual figures based onknown cases of illness or death.

The “distributional” costs of environmen-tal contamination disclose the expenses orlosses incurred by affected businesses, indi-viduals, and government bodies. These mightinclude farmers, fishermen, food processors,

Ch. l—Summary ● 5

animal feed suppliers, chemical companies, fied, their actual dollar losses are not known.consumers, and local, State, and Federal To understand who is bearing the major eco-agencies. Although the individual organiza- nomic brunt of a contamination episode, ac-tions suffering such losses are usually identi- tual cost data are required.

MAJOR PROBLEMS IN IDENTIFYINGENVIRONMENTAL CONTAMINANTS

To determine whether an environmentalcontamination incident has occurred, it isnecessary to establish the presence of thecontaminant in food. In some instances, peo-ple or animals have become ill before the re-sponsible contaminant was identified. No oneknew or even suspected that the particularsubstance was present in food. This has beenthe pattern in many major contamination in-cidents—those involving PBBs, PCBs, andmercury.

Our regulatory monitoring system hasfailed to detect such environmental contami-nants as they entered the food supply. Thus,this assessment identifies and evaluatesother approaches for monitoring either foodor the environment for toxic substances thatmay harm human health. The ultimate objec-tive of monitoring is to prevent or minimizehuman exposure to environmental contami-nants in food.

The only sure way to prevent this kind ofcontamination is to make certain that toxic

substances are not released into the environ-ment. There are various Federal environmen-tal laws that are designed to limit such re-leases. But the laws and regulations are notlikely to prevent the deliberate or accidentalmisuse or disposal of the thousands of toxicsubstances manufactured in the UnitedStates.

The problem is compounded by disposaland handling practices that was accepted inthe past but are now recognized as posingserious environmental hazards—hazardsthat will persist for many years to come. Thetoxic chemical waste dump at the Love Canalnear Niagara Falls, N. Y., clearly illustratesthe threat. According to Environmental Pro-tection Agency (EPA) estimates, there are1,200 to 2,OOO of these abandoned chemicaland radioactive waste sites in the UnitedStates that pose an imminent danger to hu-man health and will cost as much as $50 bil-lion to clean up. As long as these substancesremain in the environment, the potential forfood contamination exists.

Once an environmental contaminant isfound in food, limits are established to controland restrict its presence. Such regulationsare set up and enforced by FDA for foodtraded in interstate commerce, and by Stateagencies for food produced and sold within aState. In either case, the aim is to limit thepublic’s exposure to a particular contami-nant.

Key factors involved in such regulation aretime and information. After a contaminant isidentified. authorities must have information

on its toxicity, the amount present in the food,and how much and what kinds of food arecontaminated. By monitoring the food supplyfor the contaminant, regulators can deter-mine the level and extent of the contamina-tion. With this information and necessarytoxicity data, they can establish regulatorylimits for the contaminant in food.

However, this kind of information general-ly takes time to generate—usually longerthan the public is willing to wait in the eventof a food contamination incident. As a result,

authorities are often pressed to set regula-tory limits before they have enough time to

develop information on the nature and extentof the contamination.

This assessment has focused on two cen-tral problems: regulating environmental con-taminants and identifying environmental con-taminants. Following are major findings andconclusions growing out of this assessment.

● FDA relies on action levels rather than tol-erances to regulate environmental con-taminants in food.Under the Food, Drug, and Cosmetic Act,

FDA is given authority to set tolerances forthe amount of an unavoidable contaminantpermissible in food. However, the proceduresrequired to set a tolerance are complex, cum-bersome, and time-consuming. Therefore,FDA relies on action levels, informal judg-ments about the level of a food contaminant towhich consumers may safely be exposed.

Action levels are administrative guidelinesthat can be developed and promulgated moreeasily and quickly than tolerances. Actionlevels are used when scientific data are in-complete. Public input is not required. Theyare used when new information is likely to beforthcoming that might alter the level. FDA isunder no constraints to review action levelsor to replace action levels with formal toler-ances. FDA has sometimes lowered or raisedaction levels as new data became available.FDA is now in the process of lowering thePCB tolerance.● No policy exists defining the relative

weights to be given to the evidence whensetting an action level or tolerance.

In setting an action level or tolerance, FDAtakes into account short- and long-term tox-icological data, available information on thelevels of the contaminant in food, the amountof contaminated food consumed by variouspopulation groups, the level that can be meas-ured, and the potential impact of various ac-tion levels or tolerances on the national foodsupply. Generally, the more informationabout a particular factor, the greater its in-

fluence. Because the amount and quality ofinformation available when FDA encountersan environmental food contamination prob-lem are inevitably unpredictable, it does notpredetermine the weighting of various fac-tors. However, FDA maintains that the publichealth factor outweighs all others in its con-siderations.

The Food, Drug, and Cosmetic Act does notspecify the role that the costs of a regulatorydecision should play in setting a tolerance oraction level. The Act does require that FDAtake into account the extent to which a sub-stance cannot be avoided in food production.FDA interprets this requirement as justifica-tion for weighing the costs of food condemnedagainst the health benefits derived from a tol-erance.

● To assess human risk from exposure tochemicals, FDA and EPA rely on already-existing animal studies and epidemiologi-cal evidence derived from previous hu-man exposures,

When a new environmental contaminant isdiscovered in food, regulatory agencies areunder intense pressure to act to protect thepublic. FDA and EPA (if the contaminant is apesticide) review the available literature onthe contaminant and calculate an action levelbased on that evidence. Rarely are newstudies commissioned—even when the dataare inadequate.

New human epidemiological studies andconventional 2-year animal studies are of lit-tle immediate help because so much time isrequired to generate results. However, toxi-cologists have developed a variety of teststhat can evaluate a substance’s possible toxiceffects in 90 days or less. Some of these short-term tests measure the potential of a sub-stance to produce mutations and possiblecancer,

Ch. I—Summary ● 7

Short-term tests could be used more widelyin screening environmental contaminants todetermine whether they are mutagens or po-tential carcinogens. Although the results ofsuch tests do not provide the data needed toset an action level or tolerance, they still canalert regulators to latent dangers that re-quire further investigation.

Conventional 2-year animal studies (whichusually entail an additional year for dataanalysis) would continue to serve an impor-tant role in the setting of tolerances. If datafrom a carcinogen bioassay are available atthe time an environmental contaminant is dis-covered in food, the information can provecrucial in reaching a regulatory decision. Ifdata were nonexistent or inadequate, a newlycommissioned carcinogen bioassay could beused to revise an initial action level.

Epidemiological studies would remain use-ful for confirming suspected chronic effectsof a toxic substance to which a populationhas unknowingly been exposed over a periodof time. They can also confirm retrospectivelyor refute the adequacy of regulatory actions.

No currently available toxicological testingmethods or statistical interpretation tech-niques are adequate for evaluating the com-bined effects of low-level exposure to toxicsubstances. Indeed, there are no satisfactorytechniques for testing the interactions ofmore than two substances.

● Current monitoring at both Federal andState levels is regulatory, designed to en-sure that substances in food do not exceedprescribed limits. Little effort is made todetect and identify substances in the foodsupply for which no action levels or toler-ances exist.

Technology now exists that would makepossible a national investigatory monitoringsystem to detect unregulated chemicals asthey enter the food chain. Such advancedtechnology is available in some Federal regu-latory monitoring laboratories and in a lim-ited number of State labs. It is not routinely

employed in Federal or State regulatory moni-toring.

The goal of food monitoring is to protectconsumers by determining short- and long-term trends in the levels of various chemicalsin food and the environment. Investigatorymonitoring could be designed to complementalready-existing regulatory monitoring. Eachof these approaches could be complementedby specimen banking-the regular collectionand storage of samples that could be lateranalyzed if a new contaminant is found infood. EPA and the National Bureau of Stand-ards are now working towards developingsuch a specimen-banking program.

However, food sampling may not be thebest approach to investigatory monitoring. Todiscover a substance as it enters the environ-ment and before it gets into the human foodsupply, it is necessary to monitor water, soil,air, river sediments, and nonfood organisms.

● Management of food contamination inci-dents is hindered by the complexity of thefood system, the rapidity with which foodis moved through the system, and failuresby State and Federal agencies to coordi-nate their information-gathering activ-ities.

Many food contamination incidents initial-ly fall under State jurisdiction. Technically,the Federal Government does not become in-volved unless requested by a State or untilcontaminated food enters interstate com-merce. This country’s food marketing systemis complex. Most food produced or processedwithin a particular State is distributed forconsumption in other States. Thus, most envi-ronmental contamination incidents are likelyto become interstate problems. Figure 1 illus-trates the extent of food contamination thatcan occur from a single source of contamina-tion, in this case PCB-contaminated animalfeed from a meatpacking plant in Billings,Mont.

The number of State and Federal agenciesinvolved complicates the generation and dis-semination of scientific information on the

51+-515 ~ - 79 _ 2

8 • Environmental Contaminants in Food

II

l II

II '

c:: ~ 't! ~ c E '0 ~ 01 2 a '0 c t1I

'---~ g u. W <.,) a:: :::l o (/)

Ch. /—Summary ● 9

toxicological and chemical properties of thecontaminant, the amount and type of foodcontaminated, and the concentration of thesubstance in food. At least three Federalagencies (EPA, FDA, and the U.S. Departmentof Agriculture (USDA)), each with differentresponsibilities, may provide technical assist-ance. At the State level, departments ofhealth, agriculture, and the environment mayshare accountability for regulating environ-mental contaminants in food.

tential exists for breakdowns in communica-tion. This was the case in the recent PCB con-tamination of animal feeds in 10 WesternStates. The Idaho Department of Agriculturedid not inform the Idaho Department ofHealth and Welfare of the PCB contamina-tion. USDA would report the results of its in-vestigations only to the Idaho Department ofAgriculture. EPA attempted to determine thesource of the PCBs by analyzing air and wa-ter samples, but failed to report its negative

In the absence of a clear authority to coor- results to the State.

dinate activities of various agencies, the po-

CONGRESSIONAL OPTIONS

There are four basic options for Congressto consider regarding the Federal response tothe environmental contamination of food.Each is discussed in greater detail in chapterIX. Congress can:

1. Allow the present system to continue bytaking no action. The present systemconsists of regulatory monitoring andthe establishment of action levels (andoccasionally tolerances) for environmen-tal contaminants in food.

2. Amend the Food, Drug, and CosmeticAct to specifically address the uniqueproblems posed by environmental con-tamination of food.

3. Establish a national investigatory moni-t o r i n g s y s t e m .

4. Improve the Federal response to newcontamination problems by designatinga lead agency or establishing a center toorchestrate the delivery of Federal as-sistance to affected States.

Option 1Maintain the present System

Pros: There are two principal advantagesin maintaining this system. No additional ap-propriations or legislation are required. Nochanges in existing regulations are neces-sary.

Cons: The time needed to identify an envi-ronmental contaminant in food and- take cor-rective action would not be shortened if thecurrent system were retained. Moreover, ac-tion levels and tolerances permit a certainlevel of contaminant to be present in food. Iftolerances or action levels are not reduced,little effort will be made to eliminate the con-taminant. There is no requirement for reviewof an action level once it is established. Thus,FDA is under no pressure to actively seek outnew data to verify the appropriateness of anexisting action level. Finally, States have noclearly defined authority to turn to when theysuspect environmental contamination of food.

Option 2Amend the Food, Drug,

and Cosmetic Act

An amendment to the Food, Drug, and Cos-metic Act could contain one or more of thefollowing changes. Each change is discussedin greater detail in chapter IX.

● Congress could amend the Food, Drug,and Cosmetic Act to simplify the admin-istrative procedure for setting toler-ances. The change could be modeledafter section 553 of the AdministrativeProcedures Act. This would encourageFDA to move from action levels to toler-

●

●

●

ances, thus bringing more public partici-pation into the process,Congress could amend the Food, Drug,and Cosmetic Act to require the estab-lishment of a tolerance within a speci-fied time after the setting of an actionlevel. This would encourage the FDA togather additional information on a con-taminant’s toxicity and the public’s ex-posure. It would result in a definitive tol-erance that FDA could enforce with lessconcern over legal challenge.Congress could clarify to what extenteconomic criteria can be used in settingtolerances for environmental contami-nants in food,FDA could be granted authority to set re-gional tolerances. This would provideFDA with flexibility to set differentlevels for different regions based on ex-pected levels of exposure, regional levelsof contamination, and local eating pat-terns.

Pros: Since its passage in 1938, the Food,Drug, and Cosmetic Act has been amendedseveral times to deal with new problems offood regulation. Congress has never directlyaddressed the environmental contaminationof food. There are several unique character-istics of this problem that could be clarifiedthrough an amendment dealing specificallywith the environmental contamination offood.

Cons: Even though the Food, Drug, and Cos-metic Act does not contain provisions on envi-ronmental contaminants, FDA has been ableto regulate them through interpretation ofsections 402 and 406.

Option 3Establish an Investigatory

Monitoring System

Congress could establish a national investi-gatory monitoring system based on monitor-ing for either suspected or uncharacterizedenvironmental contaminants. Some chemi-cals are not regulated by action levels or tol-erances but are suspected to be dangerous to

humans if consumed in food. Uncharacter-ized environmental contaminants are sub-stances that may have entered the food supply, but are not regulated or suspected foodcontaminants. A system that combines ele-ments of both approaches could also be setup. Because any of these monitoring ap-proaches would require some research anddevelopment before going into full operation,Congress could choose to establish a pilot pro-gram. Such a program would spur researchand development and assess the feasibilityand cost effectiveness of the various ap-proaches.

Pros: Investigatory monitoring would in-crease the probability of detecting unregu-lated substances in food. Present food-moni-toring efforts are not designed to detectunregulated environmental contaminants infood. The limited amount of investigatorymonitoring that does exist is primarily con-cerned with trace metals. To identify newcontaminants as they enter the food supply,more of this type of monitoring is needed.

Cons: The costs of setting up an investiga-tory monitoring program could be large, andthere is no certainty that the sampling planwould identify all environmental contami-nants before they enter the food chain. Fur-thermore, investigatory monitoring relies onsophisticated instrumentation that is gener-ally not found in Federal or State monitoringlaboratories,

Option 4improve Federai Response toNew Contamination Incidents

The Federal response to new contamina-tion problems has been hampered by the mul-tiplicity of agencies with regulatory or mon-itoring responsibilities for the environmentand for food. Congress could designate a leadagency or establish a center to orchestratedelivery of Federal assistance to affectedStates.

Pros: With a clearly delineated agency orcenter, States suspecting contamination of

Ch. /—Summary ● 11

food would have one reliable Federal sourcefor generating, evaluating, and disseminatingtechnical information. Response time mightbe shortened, duplication of effort reduced,and effective management of the incident en-hanced.

Cons: Better coordination among FDA,USDA, and EPA could accomplish the samegoals without the expense of establishing anew research center. Historically, the majorimpediment to timely Federal response tochemical contamination of food was lack ofawareness that food contamination had takenplace. When contamination became apparentand one or more Federal agencies werealerted, response was rapid. Furthermore,establishment of a lead agency or a new cen-

ter would not ensure that information wouldbe generated more quickly than is now thecase.

* * *

Options 2 through 4 are not mutually ex-clusive. If Congress wishes to put greater em-phasis on protecting consumers from contam-inated food, one or more could be chosen. Forexample, Congress could decide to simplifythe administrative procedures for settingtolerances (Option 2), require the setting of atolerance at some specified time after an ac-tion level is set (Option 2), establish a pilotprogram of investigatory monitoring for or-ganic chemicals (Option 3), and designateFDA the lead agency to deal with new con-tamination problems (Option 4).

Environmental Contaminationof Food

Chapter II

Chapter II

Environmental Contamination of Food

Maintaining an adequate, safe food supply has been a major goal of the Fed-eral Government since 1906, when the first Federal food and drug law was signedinto law. Historically, chemicals such as salt, sugar, and wood smoke have beenused to preserve foods. Modern food technology relies extensively on the use ofchemicals not only for preservation but also to produce appealing colors, flavors,aromas, and textures.

Most developed countries now have food laws designed to permit the use ofsuch chemicals in food under conditions judged to be safe. These chemicals arenot considered adulterants or contaminants and are classed as intentional addi-tives. Other chemicals may enter food as a result of their use in food production,handling, or processing. Such substances maybe legally permitted if they are un-avoidable under good manufacturing practices and if the amounts involved areconsidered safe. These chemicals are classed as incidental additives. The pres-ence of both these classes of chemicals in food is controlled by regulation.

Environmental contaminants include sub-stances from natural sources or from indus-try and agriculture. Many of the naturally oc-curring contaminants in food are of microbio-logical origin and consist of harmful bacteria,bacterial toxins, and fungal toxins. (Aflatox-in, a contaminant of peanuts and grains, is anexample of a fungal toxin or mycotoxin. ) Thesecond category of environmental contami-nants includes organic chemicals, metals andtheir complexes, and radionuclides. Onlythose environmental contaminants intro-duced into food as a result of human activitiessuch as agriculture, mining, and industry areconsidered in this assessment.

The environmental contamination of food isa result of our modern, high-technology soci-ety. We produce and consume large volumesof a wide variety of substances, some ofwhich are toxic. It is estimated that 70,000chemicals may currently be in commercialproduction in the United States and that 50 ofthese chemicals are manufactured in quanti-ties greater than 1.3 billion lbs per year.Seven percent of this country’s gross nationalproduct (GNP), $113 billion per year, is gener-ated by the manufacture and distribution of

chemicals (l). During the production, use, anddisposal of these substances, there are oppor-tunities for losses into the environment. Forexample, the Environmental Protection Agen-cy (EPA) estimates that there are more than30,000 chemical and radioactive waste dis-posal sites. Of these, 1,200 to 2,000 are con-sidered threats to human health (2).

Environmental contamination of food takestwo forms: long-term, low-level contaminationresulting from gradual diffusion of persistentchemicals through the environment, and rela-tively shorter term, higher level contamina-tion stemming from industrial accidents andwaste disposal.

An example of low-level contamination ispolychlorinated biphenyls (PCBs). This groupof substances was widely used in transform-ers and capacitors, as heat-transfer fluids,and as an additive in dyes, carbon paper, pes-ticides, and plastics (3). Although productionwas halted in 1977, PCBs remain an ubiqui-tous, low-level contaminant of many foods, es-pecially freshwater fish.

An example of the second type of contami-nation is polybrominated biphenyls (PBBs) in

15

16 . Environmental Contaminants in Food

dairy products and meat. PBBs, a fire retard- level contaminant in Michigan because theyant, were accidentally mixed into animal are very stable and resistant to decay. Ani-feed. Dairy cattle that were fed the contami- mals raised on farms affected by the originalnated feed produced contaminated milk. The feed contamination are now contaminated bydistinctions between the two types of food the PBB residues remaining in the pasturescontamination are not exclusive. For exam- and farm buildings.pie, PBBs have now become a long-term, low-

HOW FOOD BECOMES CONTAMINATED

gated railroad cars, trucks, ships, or storagebuildings used for transport or storage ofhuman food and animal feed are also sourcesof environmental contamination. The interi-ors are sprayed or fumigated with pesticides,and if not sufficiently aired, contamination ofthe food or feed occurs.

The manufacture of organic chemicals pro-duces sludges, gases, and liquid effluents ofvarying chemical complexities. The usualwaste disposal methods (sewage systems, in-cineration, landfill) are unable to prevent or-ganic residues from entering the environmentin spite of Federal laws and correspondingregulations governing disposal. The routes in-clude the atmosphere, soil, and surface orground water.

Chemicals contaminate foods through dif-ferent routes depending on the chemical andits physical properties, its use, and the sourceor mechanism of contamination.

Organic substances that have contami-nated food have been either industrial oragricultural chemicals. Pesticides are theonly agricultural chemicals known to be en-vironmental contaminants in food (see tables1-3). A pesticide becomes an environmentalcontaminant when it is present in foods forwhich the application or use of the substancehas not been approved. Livestock, poultry,and fish can be contaminated when applica-tion or manufacturing of pesticides occurs inthe vicinity or when residues are transportedthrough the environment. Improperly fumi-

i1

,

Table 1.— Reported Incidents of Food Contamination, 1968-78, by State and Class of Contaminant

S t a t e Pesticide Mercury PCB PBB Other Total

New York . . . . . . . • ., .Idaho. . . . . . . . . . . .South Carolina. . . . . . . . . . . . .Minnesota . . . . . . . . . . . . . . .Louisiana . . . . . . . . . . . . . .Colorado. . . . . . . . . . . . . . .Georgia. . . . . . . . . . . .Maryland. . . . . . . . .Texas . . . . . . . . . . . . . . .New Jersey. . . . . . . . . . . . . . . .

11

12

—151

1 — — 3311

172

15314

—1-biphenyl

—81

15

—4 —

——— —

2-petroleum111

— —— —

2 1-ß-methoxy napthaleneand tetraline

—

Kansas . . .Missouri . . . . . . . . . . . . . . . .New Mexico. . .Alaska. . . . . . . . . . .C a l i f o r n i a .Indiana .,

— (I)*—

1—3

—

— (l)*4

(:)*49

—4 —

—— —— ——

(l)*1

— — —— — —

3 well-documented 1 — —5 other incidents

131

M i c h i g a n .V i r g i n i a , . , . . .

1—

19

2—

1 — 171— —

56 1 4 88Totals ... . . .

“Several conservatively estimated as oneSOURCE Of fLce of Technology Assessment

8

Ch. //— Environmental Contamination of Food ● 17

Table 2.— Reported Incidents of Food Contamination, 1968-78, by State and Food

Fruit/ Game/meat/State Dairy Eggs Vege tab le F i sh / she l l f i sh Grain poultry

31

—19

——

312

———

2—

4—

127

——1

—1

—————1

——————3

—1

——

—1114

8

Table 3.— Number of Incidents of Environmental Contaminants of Food Reported by Federal Agencies, 1968-78

Agency Pesticides Mercury PCB’s Other Food affected

USDA/FSQS 39 — — — Chickens, turkeys. ducks. cattle, swine, lambs— 1 — — Swine— — 6 — Poultry— — — 1 (Phenol) Cattle

F D A . . . 21 — — —— 84 — —— — 3 —

Total. . . . 60 85 9 1

SOURCE Off Ice of Technology Assessment

Metals can be released into the environ-ment in several ways. The mining and refin-ing processes produce dust and gases whichenter the atmosphere. Metallic salts formedduring recovery and refining processes canescape as waste products into surface andground water. Sewage sludge used as fertiliz-er on agricultural land also poses a potentialfood contamination problem. Trace metalspresent in the sludge can be taken up bycrops grown on treated soil. Cadmium is thetrace metal in sludge that currently gener-ates the greatest concern.

Radioactivity in food stems from threesources: natural radioactivity. releases from

Fish. cheese, pastaFishFish, eggs, bakery products

operation of nuclear reactors and processingplants, and fallout from nuclear weaponstests. The primary route by which food be-comes contaminated is the deposition of air-borne material on vegetation or soil. The sub-sequent fate of the radionuclide is deter-mined by its chemical and physical natureand whether it is absorbed and metabolizedby plants or animals. Natural radioactivitymay become a concern when ores containingradioactive substances are mined and proc-essed. The products or wastes may concen-trate the radionuclides. Examples of this areuranium tailings, phosphate rock waste, orslags from phosphorus production. Radiummay enter the food chain when it dissolves in


ground water and is taken up through plantroots.

Nuclear reactors normally release radio-active noble gases that do not contaminatefoods. Reactors do contain large inventoriesof fission products, transuranics, and otheractivation products. Accidental releases cancontaminate vegetation by deposition of parti-cles on leaves and soil, or through water. Gas-eous releases would most likely involve thevolatile elements such as iodine and tritium,or those with volatile precursors, such asstrontium-90 and cesium-137. Aqueous re-leases would follow failure of the onsite ionexchange cleanup system. Any of the water-soluble elements could be involved. Table 4

summarizes the radionuclide contaminants ofsignificance for foods.

Nuclear waste-processing plants couldalso have either gaseous or aqueous releases.In this case, the fission products are agedbefore processing, and iodine and the gas-eous precursor radionuclides are not re-leased. Tritium and carbon-14 are the majorairborne products, while the waterborne ra-dionuclides are the same as for reactors.

Atmospheric nuclear weapons tests dis-tribute their fission products globally. Localdeposition depends on the size of the weaponand the conditions of firing (high altitude, sur-face, or underground).

Table 4.— Radionuclide Contaminants of Significance for Foods

Normally present in small amounts, Significant only whenenhanced

Normally present in small amountsMember of uranium series. Normally present, metabolized

somewhat Iike calciumAs radium-226, only a member of thorium seriesMembers of uranium seriesNormally present

Product of nuclear reactions

Product of 24’ Pu decay

Low energy, usually in form of water or organiccompounds

Low energy, usually in form of organic compounds

Short half-life, so important only for fresh foods, e.g., milkand leafy vegetables

Follows calcium somewhat in metabolismFollows potassium somewhat in metabolism

Most important are isotopes of zirconium, cerium, barium,rubidium, rhodium. Mostly surface contaminants

Follow stable elements in metabolism

Ch. //— Environmental Contain/nation of Food ● 19

MAGNITUDE OF THE PROBLEM

There is little information available on thenumber of food contamination incidents, theamount and costs of food lost through regula-tory actions, or the effects of consumption ofcontaminated food on health. To obtain infor-mation on the extent of the problem, OTA re-viewed the literature and sought informationfrom the States and Federal agencies.

Evidence of Human IllnessResulting From Consumption of

Contaminated Food

In evaluating the significance of environ-mental contaminants in food the key questionis whether consumption of contaminatedfoods poses a health risk. Measurable healtheffects depend on the toxicity of the sub-stance, the level at which it is present in food,the quantity of food consumed, and the vul-nerability of the individual or population. InJapan, foods contaminated with substancessuch as PCBs, mercury, and cadmium haveproduced human illness and death. No suchmass poisonings have occurred in the UnitedStates. However, in cases such as PBBswhere a large populace has been exposed,some physiological changes have been noted.But no conclusions can as yet be drawn on theultimate health effects.

It is known from limited surveys that theU.S. population is exposed to a wide varietyof chemical contaminants through food, air,and water. The long-term health effects andthe implications of possible interactionsamong these residues are unknown. A recentliterature review of over 600 publishedstudies (4) found that nonoccupationally ex-posed U.S. residents carry measurable resi-dues of 94 chemical contaminants. Twenty-six of these are organic substances, includingtwenty pesticides and pesticide metabolizes.The remainder are inorganic substances.

Americans also have been exposed to lowlevels of PCBs, PBBs, mercury, and ionizingradiation through their food. The followingsections briefly summarize current knowl-

edge and the extent of uncertainties on thehealth effects of these environmental con-taminants.

Polychlorinated Biphenyls

PCBs occur in food as the result of environ-mental contamination leading to accumula-tion in the food chain, direct contact withfood or animal feeds, or contact with food-packaging materials made from recycled pa-per containing PCBs (5). Several comprehen-sive literature reviews have been publishedin the last 5 years detailing the acute andchronic toxic effects of PCBs in animals andhumans (5-1 1).

Human illness has been caused by expo-sures to PCBs at much higher levels thanthose that occur in the United States. In theearly part of 1968 the accidental contamina-tion of edible rice-bran oil led to a poisoningepidemic among the Japanese families whoconsumed the oil. The disease became knownas Yusho or rice-oil disease. Its chief symp-toms were chloracne (a severe form of acne)and eye discharge; other symptoms includedskin discoloration, headaches, fatigue, ab-dominal pain, menstrual changes, and liverdisturbances. Babies born to mothers whoconsumed the rice oil were abnormally smalland had temporary skin discoloration. Thefirst symptoms of Yusho disease were regis-tered on June 7, 1968, and 1,291 cases hadbeen reported as of May 1975 (9).

Since the rice oil was also contaminatedwith polychlorinated dibenzofuran (PCDF), itis difficult to determine from the Yusho dataexactly what effect(s) exposure to PCBs alonecould have on humans. It has been calculatedthat the PCDF made the rice oil 2 to 3.5 timesmore toxic than would have been expectedfrom its PCB content alone (1 1). Careful rec-ords of the 1,291 Yusho patients have beenkept to determine possible long-term effects.At least 9 of 29 deaths that occurred as ofMay 1975 were attributed to cancer (malig-nant neoplasm), but a causal relationship be-

20 ● Environmental Contaminants in Food

tween PCBs and cancer cannot necessarilybe inferred because of the high concentrationof PCDF in the oil. The Yusho study, neverthe-less, had two important results: first, the in-formation established that PCBs can be trans-ferred from mother to fetus and from motherto child through breast feeding, and secondhighly chlorinated PCB compounds are ex-creted more slowly from the body than lesschlorinated ones (9).

More recent experiments in animals havedemonstrated a variety of toxic effects.Cancers have been produced in mice and ratsfed PCBs (6, 12). Monkeys fed levels of PCBsequivalent to the amounts consumed by Yu-sho patients developed similar reproductivedisorders (13-16). Young monkeys nursing onmothers consuming feed containing PCB de-veloped toxic effects and behavioral abnor-malities (15-1 7)0

Polybrominated Biphenyls

Practically every Michigan resident hasbeen exposed to PBB-contaminated food prod-ucts. It is estimated that some 2,000 farmfamilies who consumed products from theirown PBB-contaminated farms have receivedthe heaviest exposure (18).

Fries (19) studied the kinetics of PBB ab-sorption in dairy cattle and its elimination inmilk, If intake of contaminated milk alone isconsidered, those Michigan residents mostseverely exposed consumed from 5 to 15grams of PBB over the initial 230 days of theexposure. Those residents that coincidentallyconsumed contaminated meat and/or eggsmay have received higher total doses of PBB,but the number of such cases is probablysmall,

Geographically the residents of the lowerpeninsula, where the original accident oc-curred, were found to have the greatest levelsof exposure. In 1976, the Michigan Depart-ment of Public Health conducted a study onPBB concentrations in breast milk. It wasfound that 96 percent of the 53 womenselected from the lower peninsula and 43 per-cent of the 42 women selected from the upper

peninsula excreted PBB in their breast milk(20).

Low concentrations of PBBs also have beendetected in animal feed in Indiana and Illi-nois. Unconfirmed surveys of food throughoutthe country found extremely low levels belowthe Food and Drug Administration (FDA) ac-tion level in the following States (21):

State FoodAlabama. ., . . . . . . . . . . . ChickenIndiana. . . . . . . . . . . . . . . TurkeyIowa , ... , . . . . . . . . . . . . BeefMississippi. . . . . . . . . . . . ChickenNew York . . . . . ... , . . . . ChickenTexas . . . . . . . . . . . . . . . . ChickenWisconsin . . . . . . . . . . . . Duck

Wolff, et al. (22) reported that serum PBBwas higher for males than females. It wassuggested that the greater proportional bodyfat in women may account for this difference,but exposure may also be important. Malesmay consume more contaminated food orhave more direct contact with PBB thanfemales.

The same study found no consistent trendswith respect to age. It was observed, how-ever, that young males had greater concen-trations of serum PBB than young females.Young females had greater concentrationsthan older males, and older males hadgreater concentrations than older females. Itwas also found that very young children andindividuals who had lived on farms less than1 year had lower serum PBB levels than othergroups (22)0

Serum PBB concentration is related to theintensity of exposure. Most studies indicatethat consumers and residents of nonquaran-tined farms had significantly lower PBB lev-els than residents of quarantined farms; how-ever, families on quarantined farms stoppedconsuming meat and milk from their own ani-mals (20)0

In late 1974, the Michigan Department ofPublic Health conducted a survey to deter-mine if any adverse effects could be corre-lated with PBB levels in the body, A sample of165 exposed persons (quarantined farms)

Ch. Il—Environmental Contamination of Food ● 21

and 133 nonexposed (nonquarantined farms)was studied. Medical history interviews andphysical examinations were performed oneach subject and blood specimens weretaken, Blood PBB levels as high as 2.26 partsper million (ppm) were found in the exposedindividuals; about half exhibited levelsgreater than 0.02 ppm. Of the nonexposed in-dividuals, only two showed blood PBB levelsgreater than 0.02 ppm; 70 percent of theadults and 97 percent of the children exhib-ited levels of 0.0002 to 0.019 ppm. Compar-ison of a list of selected conditions and com-plaints revealed no significant differences inthe frequency of illness between the twogroups. Physical examinations and clinicallaboratory tests disclosed no effects attrib-utable to “chronic” PBB exposure (24).

The effect of PBB exposure on white bloodcell (lymphocyte) function of Michigan dairyfarmers who consumed contaminated farmproducts was examined by Bekesi, et al. (25).Forty-five members of Michigan farm fami-lies who had eaten PBB-contaminated foodfor periods of 3 months up to 4 years after theoriginal accident were compared for immuno-logical function to 46 Wisconsin farmers and79 New York residents. All of the exposed in-dividuals showed reduced lymphocyte func-tion, and 40 percent showed abnormal pro-duction of lymphocytes. There were also sig-nificant increases in lymphocytes with no de-tectable surface markers (“null” cells). How-ever, the short- and long-term health implica-tions of these differences are not now known.

Lillis (20) examined Michigan farmers andconsumers of dairy products and found thatthe effect of PBB on humans was mainly neu-rological in nature. He found marked fatigue,hypersomnia, and decreased capacity forphysical or mental work. Other symptoms in-cluded headache: dizziness: irritability; and

swelling of the joints with deformity, pain,and limitation of movement. Less severe gas-trointestinal and dermatological complaintswere also encountered.

Mercury and Methylmercury

Foods are the major source of human expo-sure to mercury. The mercury concentrationin food is dependent on the type of food, theenvironmental level of mercury in the areawhere the food is produced, and the use ofmercury-containing compounds in the agri-cultural and industrial production of the food.All living organisms have the ability to con-centrate mercury, Therefore, all animal andvegetable tissues contain at least traceamounts (26). Several recent reviews have ex-amined the health effects associated withconsumption of mercury (26-28). The resultsof these reviews indicate that the effects ofmethylmercury poisoning become detectablein the most sensitive adults at blood levels ofmercury of 20 to 50 µg/100 ml, hair levelsfrom 50 to 120 mg/kg, and body burdens be-tween 0.5 and 0.8 mg/kg body weight (26).

Since the Minamata Bay tragedy in Japan,the effects of chronic exposure to methylmer-cury have been well-documented. Mercuryreadily accumulates within the central nerv-ous system (29-3 1), and clearance of mercuryback into the bloodstream is slow (32). Conse-quently, the central nervous system is consi-dered to be the critical target in chronic mer-cury exposure. The clinical symptoms of cen-tral nervous system involvement includeheadache, vertigo, vasomotor disturbance,ataxia, and pain and numbness in the ex-tremities (30). The most prominent structuralchanges of the central nervous system result-ing from chronic mercury exposure are dif-fuse cellular degeneration (30).

In evaluating the teratogenic hazards ofmercury exposure to man, the placentaltransfer of mercury is particularly signifi-cant. Levels that are not toxic to pregnantwomen are sufficient to produce birth defectsin their offspring (33-35). Transfer of methyl-mercury across the human placenta resultsin slightly higher blood levels in the infant atbirth than in the mother (36). Table 5 com-pares fetal and maternal blood concentra-


Table 5. — Methylmercury Concentrations inNormally Exposed Populations

—Concentration (µg Hg/g)

Location Maternal blood Placenta Fetal blood

Japan ...- – ‘ - 0.017 0.072 0.020Sweden . . . . . . 0.006 — 0.008Tennessee. . . 0.009 0.021 0.011lowa. . . . . 0.001 0.002 0.001SOURCE Adapted from B J Koos and L D Longo ” Mercury To; IcIty In the

Pregnant W’oman, Fetus, and Newborn Infant “ A rner(can ,/ourrra/ of

Obstetrms and Gynecology 126(3) 390, 1976

tions in normally exposed populations inJapan, Sweden, and the United States,

In humans, the most widely reported fetalrisk associated with maternal exposure tomercury is brain damage. The placentaltransfer of mercury and its effects on thehuman fetus were first recognized in the1950’s with the well-known outbreak of con-genital Minamata disease in the towns ofMinamata and Niigata, Japan. By 1959, 23 in-fants suffering from mental retardation andmotor disturbances had been born to mothersexposed to methylmercury during their preg-nancies. The clinical symptoms of the infantsresembled those of severe cerebral palsy orcerebral dysfunction syndrome. They in-cluded disturbance of coordination, speech,and hearing; constriction of visual field; im-pairment of chewing and swallowing; en-hanced tendon reflex; pathological reflexes;involuntary movement; primitive reflexes; su-perficial sensation; salivation; and forcedlaughing (30). Only 1 of the 23 mothers exhib-ited any symptoms of mercury poisoning (32).

Radioactivity

Ionizing radiation (X-rays, gamma rays, orbeta particles with sufficient energy to stripelectrons from molecules and produce ions)can produce birth defects, mutations, andcancers (37). These adverse health effectsare usually associated with high dose levelsdelivered at high dose rates.

Such a combination is not ordinarily en-countered in food. Previous radioactive con-tamination of foods has involved relativelysmall quantities of radioactive elementswhich have delivered low dose rates (38).

In these situations, the effects of the radia-

tion exposure on health are extremely diffi-cult to evaluate. High dose rates (100 millionto 1 billion times background) are estimatedto produce 2,600 ionization events per secondin cells. Background radiation levels are esti-mated to produce less than one ionization inthe cell nucleus per day (37). Because cellshave the capacity to repair damage to theirgenetic material, repair of ionization damagemay occur at low radiation exposure. Higherexposures may overwhelm the cells’ repaircapacity. Whether any effects are observedin such cases depends on several factors.These include the dose delivered to the tis-sues, the nature of the emissions, and the me-tabolism of the cell. The following examplesillustrate these points:

Strontium-90 in food arouses most con-cern not only because of its long half-lifebut also because it behaves in the bodyin a manner somewhat similar to calci-um. The replacement of bone calciumwith strontium-go exposes tissues andcells covering the bone to radiation, Inaddition, bone marrow is subject to theionizing radiation from the strontium-go.Thus, cancer of the bone-forming andbone-covering tissue as well as leuke-mias of the bone marrow blood-formingcells can possibly result.Iodine is concentrated by the thyroidgland. Radioiodines produced in atmos-pheric nuclear detonations or releasedfrom nuclear power stations are alsotaken up and concentrated by the thy-roid, increasing the risk of thyroidcancer.Tritium, or radioactive hydrogen, com-bines chemically with oxygen to formwater. Tritium derived from food wouldbe widely distributed throughout thebody exposing all tissues to radiation.

The uncertainties surrounding the repaircapacities of cells and the irreversible natureof the possible health effects have led to theadoption in the United States of a prudentpolicy toward low-level ionizing radiation.Since any amount of radiation is potentiallyharmful, unnecessary exposure should beavoided.

Ch. //—Environmental Contamination 01 Food ● 23

Number of Food ContaminationIncidents

Questionnaires were mailed to the commis-sioners of health in each of the 50 States andthe District of Columbia as well as to Federalagencies. For the 10-year period 1968-78,each was asked to report on the number of in-cidents of environmental contamination offood that resulted in regulatory action. Thissurvey has limitations. Some States did notanswer all questions. The questions weresubject to interpretation and misunderstand-ing. The accuracy and completeness of theanswers were dependent on the respondent.The results presented are therefore prelimi-nary and do not necessarily represent com-plete and comprehensive information on allStates responding. Nonetheless, these dataare the first to be developed on the extent ofenvironmental contamination of food.

Responses were received from 32 States.Seven of the top ten agricultural States andsix of the top ten manufacturing States re-sponded to the questionnaire, The agricultur-al States in the top 10 were California, Texas,Minnesota, Nebraska, Kansas, Indiana, andMissouri. The manufacturing States in thetop 10 were California, New York, Michigan,New Jersey, Texas, and Indiana. Three ofthese States—California, Texas, and Indi-a n a —are in the top ten for both agriculturaland manufacturing production. A fairly rep-resentative distribution of States respondedfrom each region of the United States (figure2).

In the following discussions, an incident isdefined as a case in which a Federal or Stateagency has taken regulatory action againstcontaminated food, The Michigan PBB epi-sode is reported as one incident because thecontamination stemmed from one source andwas limited to one State. Mercury contami-nation is reported as separate incidents be-cause the sources differed (environmentalmercury v. industrial waste), the States in-volved are widely separated, and regulatoryactions were taken at different times, Eight-een States reported at least one environmen-

tal contaminant incident since 1968 for atotal of 88 incidents. All food categories wereinvolved and a variety of substances were im-plicated (see tables 1-3).

The data provided by States are comple-mented by the Federal responses. The twoFederal agencies responsible for regulatingthe Nation’s food supply reported the numberof environmental contamination incidentsthat they had identified since 1968. FDA had108 reported incidents, and the Food Safetyand Quality Service of the U.S. Department ofAgriculture (USDA) had 47 reported inci-dents (see table 3). The combined Federal andState total number of incidents comes to 243.

Neither State nor Federal responses indi-cated any significant radionuclide contam-ination episodes during the 1968-78 period.Extensive Government programs for monitor-ing radionuclides in food exist. Thus far, ra-dionuclide contamination of food has not beenfound to exceed the exposure limits recom-mended in the Federal Radiation Council Pro-tective Action Guides. In most cases, theamount of food contamination in the conti-nental United States has never even ap-proached these limits (39), While atmosphericnuclear testing is less a threat today thanbefore the signing of the 1963 Test Ban Trea-ty, radionuclide contamination of food is stilla concern of both Federal and State govern-ments.

The number of food contamination inci-dents reported to OTA does not represent thetotal number that has occurred in the UnitedStates, only those in which the Federal Gov-ernment and 18 State governments have ta-ken regulatory action. Many incidents nevercome to the attention of State or Federalauthorities. This is because local governmentofficials can and do handle environmentalcontaminant incidents by warning offendersor by condemning contaminated productswithout informing the appropriate State offi-cials. Also, the farmer whose livestock orpoultry has been environmentally contami-nated may negotiate directly with the firm re-sponsible for the contamination for financial

24 . En

viron

men

tal

Co

nta

min

an

ts in F

oo

du)

Figure 2.-Distribution of States Responding to Questionnaire

The United States would look like this if the sizes of States were proportional to their value of farm production. The shaded States responded to the questionnaire. The States are ranked by 1974 cash receipts.

~ . -.. HAWAII ()

ALASKA (~

.~~~

.~~ -~

SOURCE: Adapted from the The American Farmer. USDA/ERS, 1976, GPO Stock No. 001-Q19-OO325-1.

. ... VT~H t:I\ (Jd ~~o~~SS


reimbursement without reporting the con-tamination to Federal or State officials (40).

Economic Impact

The economic impact of an incident involv-ing the environmental contamination of foodincludes the cost of condemned food, healthcosts, and the corresponding distributionaleffects and costs. The magnitude of the eco-nomic impact is determined by:

the amount of food contaminated,the concentration of the contaminant infood,the chemical and toxicological charac-teristics of the contaminant, andthe corresponding regulatory actiontaken on the contaminated food.

The initial regulatory action taken by Fed-eral and State authorities may be the issu-ance of a warning or the establishment ofeither an action level or a tolerance. A moredetailed discussion of this regulatory action ispresented in chapter III. Action levels and tol-erances establish a permissible level for thecontaminant in food. Any food found to con-tain concentrations of the substances abovethis level is condemned and either destroyedor restricted from being marketed.

Costs of Food Condemned

In addition to the four factors listed above,the cost of condemned food is also affected byits position in the food production and mar-keting process at the time of condemnation,An action level or tolerance for a contam-inant is the most important of the five factors.If no action level or tolerance is set, no foodwould be condemned and thus there would beno costs incurred. The impact of such a reg-ulation will depend on the exact level of asubstance that is allowed to be present infood,

The chemical properties of a contaminantare also important because of the potentialfor long-term effects on the amount of food af-fected. Since many contaminants biologicallyand chemically degrade slowly, their pres-

ence in the environment can mean food con-tamination above the action level or tolerancefor many years after the source of the pollu-tion has been stopped. The James River in Vir-ginia, for example, is still closed to commer-cial fishing several years after kepone dis-charges into the river have been eliminated.The relative influence for each of these fac-tors on the final cost will vary in each con-tamination incident.

Estimates of the cost of food condemnedthrough regulatory action are most often ex-pressed in dollars. Consequently, this cost isusually (and incorrectly) cited as representa-tive of the total economic impact. Such costswere collected in OTA’s State and Federalsurveys. The data, however, only partially re-flect the total economic impact for environ-mental contamination of food in the UnitedStates. This is because the cost of condemnedfood is only one component of the total eco-nomic impact of an incident. In addition, fewof the incidents reported to OTA includeddata on the cost of food condemned. OTA esti-mates from the available data that the totalcost of condemned food as a result of environ-mental contamination in the United Statessince 1968 is over $282 million (table 6). T h eonly cost estimates used were those clearlystated for an incident by the reporting Statesor Federal agencies.

State Estimates. —Of the 18 States report-ing contamination incidents, only 6 provideddata on the economic impact in dollar terms.Of those six, Michigan represents 99 percentof the total cost ($255 million) while reportingonly 19 percent of the number of incidents inthe 18 States. Indeed, Michigan accounts for90 percent of the total costs reported in theUnited States while reporting only 7 percentof the incidents that occurred during the1968-78 period. It must be recognized, how-ever, that 84 percent of Michigan’s costs areattributed to the PBB incident. Many inci-dents reported by State and Federal agenciesare considerably smaller than the PBB epi-sode. Thus, the PBB episode is an indicationof how severe a contamination incident canbe.


Table 6.—Economic Impact of Food Contamination—————— ————— —

Reported incidents Total estimated cost ($)

StateIdaho. . . Dieldrin

PCPColorado. . . . Dieldrin

Maryland,Texas. . .Indiana .

Michigan.

Federal

MercuryMercuryMercuryDieldrinDieldrinMercuryPCBPCNBPBBPicloramChlordaneDDTToxapheneParathionDiazinonPentachlorophenolPCBDieldrin

USDA/FSQS PesticidesMercuryPCBPhenol

$ 100,0003,000

1003,700

23,00085,00025,027

250,00010,000,00030,000,000

100,000215,000,000

12,0002,5002,0002,000

32813,70028,468

150,00012,500

$255,813,323--

18,900,00063,000

7,450,000350— ——

26,413,350

Total United States ... . . . . . . . . . .

SOURCE Off Ice of Tec;n810gy Assessment

$282,226,673

Some States reported the amount of fooddestroyed without estimating the cost. Ken-tucky, for example, reported the destructionof 400,000 lbs of milk since 1968 because ofpesticide contamination. While such informa-tion can be converted into dollars, data onmarket position and price of product at timeof confiscation are not readily available.Many States were unable to provide any esti-mates on either the cost or the amount of foodcondemned as a result of reported contami-nation incidents. New York (with PCBs) andVirginia (with kepone) are two States thatcould not provide cost estimates for food con-demned as a result of environmental contami-nation. Virginia, however, has initiated astudy to determine the economic impact of thekepone incident.

Federal Estimates.—Of the two Federalagencies reporting information to OTA on en-vironmental contaminant incidents, USDA’s

Food Safety and Quality Service (FSQS) re-ported food condemnation cost estimates.These estimates, however, only cover live-stock and poultry—the food products overwhich FSQS has regulatory authority. FDA,which has regulatory authority over the re-maining food commodities, did not estimatecosts for reported environmental contamina-tion incidents (70 percent of the Federaltotal). Thus, a significant proportion of thetotal costs for environmental contaminationincidents requiring Federal action is un-known. Comparison of the two agency re-sponses with the State responses reveals lit-tle duplication in the reporting of incidents.

FSQS cost estimates were determined bythe number of animals or pounds destroyedmultiplied by the market value at the time ofconfiscation. Since most of these animalswere taken at the farm or wholesale level, themarket value was the farm or wholesaleprice. Most of the losses resulting from FDAactions would be based on a wholesale or re-tail price because the seized products had ad-vanced further in the marketing system.Therefore, their estimated costs would begreater than if they were seized at the pro-duction level (generally the case with FSQSseizures).

Summing up, the available data on the costof condemned food is limited; consequentlyOTA’s $282 million condemned-food estimateis likely to be a gross underestimation of theactual costs. The true cost would be impossi-ble to estimate from this limited sample.

Health Costs

Health costs are also an important compo-nent of economic impact. These costs are in-curred by the consumer whose health has orpotentially can be affected adversely by acontaminant present in food. These adverseeffects can cause illness and death, and therange of effects will vary depending on thetoxicity of the contaminant, the concentrationof the contaminant in food, and the amount offood consumed.

Ch II—Environmental Contamination of Food s 27

In this country, the concentration of con-taminants has been at levels that have notproduced immediate measurable and conclu-sive effects in exposed populations. Estimatesare therefore made for the potential long-term effects on exposed populations fromvarious contaminants in food.

Health costs can be estimated from suchprojected health effects. Costs would includehealth care costs for treating illness andburial expenses associated with death. Addi-tional costs would include estimated value ofproductive days or years lost from work dueto the projected illness or death associatedwith the contaminant in food. All of thesehealth-related costs, however, do not andcannot include the emotional and psycholog-ical impacts on those aff l icted and theirfriends and families.

Health costs are not available for previ-ous U.S. food contamination incidents. Ap-proaches and techniques for estimatinghealth costs are discussed in chapter VI.

Distributional Effects and Costs

Distributional effects and costs involve thevarious people, groups, and organizationswho are economically affected by an environ-mental contamination incident. Informationon the extent and distribution of such effectsand costs provides a clearer picture of thetotal economic impact on society. This infor-mation is usually couched in descriptiveterms. Those who are economically affectedare identified but the extent of the impact isseldom estimated in dollars. The exact distri-bution of costs from an incident through soci-ety is affected by the same five factors thatinfluence the cost of condemned food.

Many of the distributional effects andcosts for various types of environmental con-taminant incidents are discussed in the fol-lowing sections. The purpose of this discus-sion is not to identify all the distributionalcosts but rather to demonstrate the variety ofeffects and costs that can result from an inci-dent.

P r o d u c e r s . — Food producers are affectedeconomically in different ways by contamina-tion episodes. But all are affected directlywhen the food they produce is condemned.For example, food found contaminated at thefarm level is confiscated and destroyed. Thiswas the case for over 500 Michigan farmerswhose dairy herds were partially or entirelydestroyed (41). In such cases, farmers eitherreplace their livestock, plant a new crop, orgo out of business.

Farmers can be faced with severe econom-ic hardship, since they are not always reim-bursed financially for the animals or com-modities confiscated. While insurance pro-grams such as the Federal Crop InsuranceCorporation are available to cover naturalhazards which might destroy crops or live-stock, such Federal assistance is not avail-able to farmers for losses from environmentalcontamination. An injured farmer can obtaina loan at commercial rates or sue the respon-sible firm for compensation. But the loan andthe interest add to a farmer’s financial dif-ficulties, and suing for compensation can taketime that the farmer may not have.

The commercial fisher is faced with a dif-ferent situation. If a river, lake, or species offish is restricted because of environmentalcontamination, the fisher whose source of in-come depends on this species or waterwaymay have few employment alternatives. Thealternatives depend to some degree on the ex-tent of the contamination. If the only water-way available in a section of a State or awhole State is closed to commercial fishingbecause of the contamination, the fisher’ssource of employment is eliminated until therestriction is ended. Since the restriction canlast for years (depending on the chemical sta-bility of the contaminant), the fisher eitherwill have to move to other commercial fishingareas or seek other employment.

Food producers economically affected bythe condemnation of contaminated food arelikely to incur health costs. This is becausemany of the producers and their families reg-ularly eat the food that they produce or har-


vest. Consequently they are exposed to thecontaminated food at greater concentrationsthan the average consumer. This was thecase for several farm families in Michigan.

Firms Held Accountable for Environmen-tal Contamination.— In most instances blamefor a contamination incident can be estab-lished. Those accountable are subject to finesand lawsuits. Firms admitting responsibilityoften try to settle with producers out of courtif possible. Most of the compensation is forthe economic damages stemming from the de-struction of food or loss of employment. Com-pensation for people whose health has beenimpaired as a result of eating contaminatedfood would be sought through civil litigation.Such litigation, however, is rare in this coun-try, since the level of contamination in food isso low that demonstrating the necessarycause and effect is difficult.

Fines or compensation paid by the firmsheld accountable for the contamination are,in fact, poor indicators of the true costs in-curred by the producers. This is because thesettlement costs which are frequently negoti-ated or imposed bear little relationship to theactual costs incurred.

For example, compensation has been pro-vided by Michigan Chemical Corporation andFarm Bureau Services, Inc., to many of thefarmers whose livestock and poultry were de-stroyed following PBB contamination. Mich-igan Chemical and Farm Bureau Serviceshave together paid more than $40 million incompensation from a jointly established in-surance pool (42). In another case involvingPCB-contaminated fish meal sold to poultryproducers, Ralston Purina Company negoti-ated compensation for the 400,000 chickensdestroyed. The cost of the compensation hasnot been disclosed (43).

G o v e r n m e n t s . — Federal, State, and localgovernments also incur costs from an envi-ronmental contamination incident. Althoughthe Federal Government and most State gov-ernments have agencies with programs toregulate or control food safety problems,these programs usually are not funded to

handle the kind of long-term problems cre-ated by a PBB or kepone incident. The Michi-gan Department of Agriculture, for example,estimates it will spend $40 million to $60 mil-lion within the next 5 years to monitor andtest for PBBs in animals and animal byprod-ucts (44). This is money that could have beensaved or spent for other programs if PBB con-tamination had not occurred. In order to re-cover its expenses from the PBB incident, theState of Michigan filed a lawsuit against boththe Michigan Chemical Corporation andFarm Bureau Services, Inc., claiming morethan $100 million in damage (45).

Federal involvement is limited unless thecontaminated food is part of interstate com-merce, Many of these incidents are not con-sidered by the Federal Government to involveinterstate commerce, FDA may provide tech-nical assistance at the request of the Stategovernment when a contamination incident isregarded to be a local problem (43). Thesetechnical facilities and experts are availableto all States through the Federal and regionaloffices. Additional expenditures by the Fed-eral Government for contamination incidentsare limited. Additional State expenditures,however, can be substantial. Federal expend-itures are made when Federal regulationsare developed and promulgated for particu-lar contaminants in food such as PCB.

C o n s u m e r s . —Consumers can incur costsfrom an environmental contaminant incidentin several ways. The removal of food fromcommerce could increase prices for that foodproduct or other food products being sold.Thus, the consumer could pay more for foodas a result of an environmental contamina-tion incident. In order for this price increasein food to occur, however, a significantamount of a food product or food productswould have to be taken off the market. Suchprices of food might vary by State or regionand affect certain socioeconomic classes dif-ferently.

Health costs could increase as a result ofthe consumption of contaminated food. Thiswould not affect all consumers but ratherthose who received the most exposure and/or

Ch. II—Environmental Contamination of Food ● 29

those most susceptible to a contaminant, suchas children or senior citizens. While thesecosts would already be included in estimatedtotal health costs, the distributional effectscould indicate those consumers most likely tobe affected.

Indirect Costs. —Most of the costs men-tioned directly stem from an environmentalcontamination incident. However, indirect orsecondary costs can and do occur. For exam-

ple, a bait and tackle store on a lake that isclosed to commercial and sport fishing be-cause of an environmental contamination islikely to suffer economic hardship, Food proc-essors whose normal supply of food has beencondemned because of environmental con-tamination will also suffer economically un-less they find new sources of supply. Theseare just two examples of the many indirectcosts which might occur.

Because a limited number of substancesposing health problems already have beenidentified in food, concern exists that othertoxic substances are likely to contaminatefood in the future. This concern arises fromthe number of substances presently beingmanufactured, used, and disposed of in theUnited States, and the difficulties in prevent-ing them from entering the environment. Newsubstances developed to meet new needs or toreplace known toxic substances may createunexpected environmental problems if notproperly controlled. Byproducts of new tech-nologies such as synthetic fuels are also po-tential environmental contaminants. Theseare described in appendix A.

There are two methods of objectively as-sessing possible future contaminants: 1 ) bysampling the food supply for chemical con-taminants and ranking them according to po-

tential hazard and 2) by surveying the uni-verse of industrial chemicals and rankingthem according to their potential for enteringthe food supply in toxic amounts. These meth-ods are discussed in more detail in chapterVII, “Monitoring Strategies. ”

Of the three categories of environmentalcontaminants considered in this report, or-ganic chemicals probably pose the greatestpotential environmental and food contamina-tion problems. This conclusion is based on thenumber, volume, and toxicity of the organicsmanufactured and used in this country (40).Both trace metals and radioactive substancescontinue to warrant concern, but not as greata concern as organic substances. The extentof food contamination from these substancesdepends on our success in preventing themfrom entering the environment.

CONCLUSIONS AND ISSUES

Data presented here indicate that environ-mental contamination of food is a nationwideproblem of unknown magnitude. Long-term,low-level exposure to toxic substances in foodposes hea l th r i sks tha t a re d i f f i cu l t toevaluate given present techniques. Incidentsof high-level contamination of food that causehuman illness have not occurred in the UnitedStates, However, regulatory actions havebeen taken to restrict consumption of con-taminated food in cases where the potential

health risks were considered unacceptable.These episodes have resulted in economiclosses when contaminated food was removedfrom the market,

The following chapters analyze severalissues related to the regulation of environ-mental contaminants in food. These are:

Q Is our present regulatory system protect-ing the public health? (Chapters 111 andIV)


● Are methods used by the regulatoryagencies for estimating health impactsthe most appropriate ones? (Chapters IIIand V)

● Should economic impacts be an explicitpart of regulatory decisionmaking? If so,

how should economic impacts be evalu-ated? (Chapters III and VI)

c Should regulatory monitoring be capableof detecting substances as they enter thefood chain? (Chapters VII and VIII)

CHAPTER II REFERENCES

1. Storclz, William J. “C&EN’s Top Fifty Chemi-cal Products and Producers, Chemical andEngineering News 56(18):33, 1978,

2. Murray, Chris, “Chemical Waste Dispos-al A Costly Problem, ” Chemica~ and Engineer-ing News, p. 12, Mar, 12, 1979.

3. National Academy of Sciences, National Re-search Council, Commission on Natural Re-sources, Environmental Studies Board, Com-mittee on the Assessment of PolychlorinatedBiphenyls in the Environment. Polychlori-natea’ Biphenyls (ISBN O-309-02885-X), 1979,

4. Hammons, A. “Levels of Chemical Contam-inants in Nonoccupationally Exposed U.S.Residents,” draft manuscript prepared forHealth Effects Research Laboratory, Environ-mental Protection Agency, 1978.

5. U.S. Food and Drug Administration. “Poly -chlorinated Biphenyls (PCBS), Federal Regis-ter 42(63):17487, 1977,

6. Kimbrough, R, D. “The Toxicity of Polychlori-nated Polycyclic Compounds and RelatedChemicals, ” CRC Critical Reviews in Tox-icology 2:445, 1974,

7, Fishbein, L, “Toxicity of Chlorinated Biphen-yls, ” Ann. Rev. Pharmacol, 14:139, 1974.

8. Peakall, D. B. “PCBS and Their Environ-mental Effects, CRC Critical Reviews on En-vironmental Contaminants 5:469, 1975.

9. National Institute of Occupational Safety andHealth. Criteria for a Recommended Stanci-ard. Occupational Exposure to Polychlori-nated Biphenyls (PCBS), 1 9 7 7 (DHEW Publ.No, 77-225).

10. Kimbrough, R, D., et al. “Animal Toxicity, ”Environ, Health Persp. 24:173, 1978.

11, Cordle, F., et al. “Human Exposure to Poly-chlorinated Biphenyls and PolybrominatedBiphenyls, ” E n v i r o n . Health Persp. 24:157,1978.

12. Kimbrough, R. D,, and R. E. Linder. “Induc-tion of Adenofibrosis and Hepatomas of theLiver in BALB/oJ Mice by Polychlorinated Bi-

13,

14,

15.

16,

17.

18.

19.

20,

21.Toxic Substances, Special “Act;ons Group.“Assessment of the Hazards of Polybromi-nated Biphenyls (PBBs)” (draft), 1977.

22. Wolff, M. S., B. Aubrey, F. Camper, and N.Hames. “Relation of DDE and PBB Serum Lev-els in Farm Residents, Consumers, and Michi-

phenyls (Aroclor 1254), ” ]ournal of The Na-tional Cancer Institute 53:547, 1974.Allen, J. R., et al. “Residual Effects ofShort-Term, Low-Level Exposure of Non-Hu-man Primates to Polychlorinated Biphenyls, ”Toxicology a n d A p p l i e d P h a r m a c o l o g y30:440, 1974.McNulty, W. P. “Primate Study, ” Nation-al Conference on Polychlorinated Biphenyls,Chicago, 1975.Barsotti, D. A., et al. “Reproductive Dysfunc-tion in Rhesus Monkeys Exposed to Low Lev-e l s o f Polychlorinated 13iphenyls (Aroclor1248), ” Food and Cosmetics Toxicology 14:99,1976.Allen, J. R., and D. A. Barsotti. “The Ef-fects of Transplacental and Mammary Move-ment of PCBS on Infant Rhesus Monkey s,”Toxicology 6:331, 1976.Allen, J. R. “Response of the Non-Human Pri-mate to Polychlorinated Biphenyl Exposure, ”Federation Proceedings 34:1675, 1975.Meester, W. D. and D. J. McCoy, Sr. “HumanToxicology of Polybrominated Biphenyls, ”Clinical Toxicology 1o(4):474, 1977.Fries, G. F., G. S, Marrow, and R. M. Cook.“Distribution and Kinetics of PBB Residues inCattle, ” Environmental Health Perspectives23:43, 1978.Lillis, R., H. A. Anderson, J. A. Valcinkas, S.Freedman, and I. J. Selikoff. “Comparison ofFindings Among Residents of Michigan DairyFarms and Consumers of Produce PurchasedFrom These Farms,” Environmental HealthPerspectives 23:105, 1978.Environmental Protection Agencv, Office of


gan Chemical Corporation Employ ees, ” Envi-ronment] Health Perspectives 23: 177, 1978.

23. Cook, H. D., R. Helland, B. H, Vander Weele,and R. J, DeJong, “Histotoxic Effects of Poly-brominated Biphenyls in Michigan Dairy Cat-tle,”’ Environmental Research 15:82, 1978.

24. Kay, K. “Polybrominated Biphenyls (PBB) En-vironmental Contaminants in Michigan, 1973-1976, ” Environmental Research 13:74, 1977.

25, Bekesi, J. G., J. F. Holland, H. A. Anderson, A.S. Fischbein, W, Rem, M. S. Wolff, and I. J.Selikoff. “Lymphocyte Function of MichiganDairy Farmers Exposed to Polybrominated Bi-phenyls, ” Science 199:1207, 1978.

26, United Nations Environment Program and theWorld Health organization. Environmentall-iec~lth Criteriu 1: Mercury, Geneva, p. 23,1976,

27. National Academy of Sciences, National Re-search Council, An Assessment of Mercury inthe Environment, Washington, D, C., 1978.

28. U.S. Environmental Protection Agency, Ambi-en t Water Quulity Criteria: Mercury, 1979.

29. Lu, F. C. “’Mercury as a Food Contaminant, ”WHO ChronicJe 28:8, 1974.

30, D’Itri, F, M. “The Environmental MercuryProblem, ” CRC Press, p. 73, 1972.

31. Friberg, L., and J, Vestal. “Mercury in the En-vironment, ” Cl? Caress, 1972.

32. Gerstner, H, B,, and J. E. Huff. ‘*Clinical Tox-icology of Mercury, ” Journal of Toxicologyand Environmental Health 2:49 1, 1977.

33. Olson, F. C,, and E. J, Massaro. “Pharmacody -namics and Tera tologic Effects of Transpla-centally Transported Methyl Mercury in theNlouse, ’” Jnternationul Conference on HeuvyMet(lis in the Environment, pp. 32 and 184,1972.

34. Weiss, B., and R. H, Doherty. “Methylmer-cur~ Poisoning, “’ Teratology 12(3):311, 1976.

35. Rizzo, A. M., and A. Furst. “Mercury Terato-genesis in the Rat, ” Proceedings of the West-ern Pharmacology Society 15:52, 1972.

36. Evans, H. L., R. H. Garman, and B. Weiss.“Methylmercury: Exposure Duration and Re-gional Distr ibution as Determinants ofNeurotoxicity in Nonhuman Primates, ” Tox-icology and Applied Pharmacology 41: 15,1977.

37. National Academy of Sciences, ‘ AdvisoryCommittee on the Biological Effects of Ioniz-ing Radiations. Effects on Populations of Ex-posure to Low Levels of Ionizing Radiation,Washington, D. C., 1972.

38. National Academy of Sciences, Food Protec-tion Committee. 13adionuclides i n F o o d s ,Washington, D. C., 1973.

39. Dodge, Chris. “Contamination of Food by Ra-dioactive Nuclides, ” Congressional ResearchService, OTA Working Paper, August 1978.

40. Monitoring Advisory Panel, Sept. 21, 1978.41. Michigan Department of Agriculture, Infor-

42

43.

mation and Education Division. “A BriefChronology of the PBB Incident” (no date).Anonymous. “Two Firms Fined $4,000 Eachfor PBB Contamination of Cattle Feed inMichigan, ” Toxic Material News, p, 144, May24, 1978,Jaroslovsky, Rich. “Contamination of Fish-meal with PCBS Sparks FDA Study of Mishapat a Ralston Purina Plant, ” The Wall StreetJournal, p. 46, July 12, 1978.

44. Whitehead, George, Deputy Director for Con-

45

sumer Affairs, Michigan Department of Agri-culture, Jan. 16, 1978.McNally, JoAnn. “Polybrominated Biphenyls:Environmental Contamination of Food, ” Con-gressional Research Service, OTA WorkingPaper, Sept. 21, 1978.

Chapter III

Federal Laws, Regulations,and Programs

Chapter III

Federal Laws, Regulations,and Programs

FEDERAL LAWS

Congress has enacted several laws that not only regulate but also attempt tolimit or restrict the introduction of toxic substances into the environment. Table 7summarizes the Federal laws affecting toxic substances control.

Some of these laws give Federal agencies authority to prevent unsafe foodfrom reaching consumers. Most important in terms of this assessment is theFederal Food, Drug, and Cosmetic (FD&C) Act (l). Broadly speaking, this statuteprohibits the introduction of adulterated food into interstate commerce. TheFD&C Act allows the Food and Drug Administration (FDA) to establish tolerancesfor toxic substances whose occurrence in food cannot be avoided. The Poultry andPoultry Products Inspection Act (2) and the Federal Meat Inspection Act (3) givethe U.S. Department of Agriculture (USDA) authority to inspect meat, poultry, andtheir byproducts. The adulteration provisions of these Acts govern all environ-mental contaminants except for pesticides that may occur in meat and poultry.Under these laws, section 408 of the FD&C Act applies to such pesticide contami-nation. In practice, USDA uses the tolerances established under the FD&C Actand consults with FDA and the Environmental Protection Agency (EPA) to deter-mine an action level when no tolerance exists.

Following is a brief summary of the perti-nent provisions of the FD&C Act:

Section 402(a)(l) declares that any foodthat “bears or contains any poisonous ordeleterious substance which may render it in-jurious to health” is adulterated. The singleexception is if the substance is not, in thelanguage of the Act, “added” to the food. Insuch cases, the presence of the substancedoes not imply that the food is adulteratedunless it is present in sufficient quantity to“ordinarily render it injurious to health. ”

The FD&C Act recognizes that certain“added” toxic substances in foods requirespecial attention, Section 406 empowers FDAto establish tolerances for “added’” poisonoussubstances whose occurrence in food cannotbe avoided or whose use is “necessary” toproduce the food. Thus, Congress authorizedFDA to “license” the presence of certainpotentially toxic substances in food, seeming-

ly because of their economic utility or the in-ability of existing, commonly used productionmethods to eliminate them. The legislativehistory of section 406 is skimpy. But Con-gress’ principal objective apparently was topermit continued use of pesticides on rawagricultural commodities while giving FDA aneffective means of control—the power to de-clare illegal any food that contained anyamount of an added substance that exceedsFDA tolerance. Congress left the distinctionbetween “added”’ and other constituents un-defined, and did not attempt to clarify theconcepts of “necessary” or “unavoidable”’ insection 406(4).

Amendments to the 1938 FD&C Act dealwith specific categories within the broadclass of substances “added”’ to foods in-cluding pesticides, food additives, vitaminsand minerals, and animal drugs. Each amend-ment in effect establishes a system under

35


Table 7.— Federal Laws and Agencies Affecting Toxic Substances Control

Statute Year enacted Responsible agency Sources covered

Toxic Substances Control Act . .....1976 EPA

Clean Air Act. . . . . . . . . . . . . . . . . . 1970, amended 1977 EPAFederal Water Pollution Control Act

(now Clean Water Act) . ..........1972, amended 1977 EPASafe Drinking Water Act . ..........1974, amended 1977 EPAFederal Insecticide, Fungicide, and 1947, amended 1972, 1975, EPA

Rodenticide Act . . . . . . . . . . . . . . . . 1978Act of July 22, 1954, (codified as

Section 346(a) of the Food, Drug, 1954, amended 1972and Cosmetic Act) .

Resource Conservation andRecovery Act . . . . . . . . . . . . . . . . . .1976

Marine Protection, Research, andSanctuaries Act . . . . ..........1972

Food, Drug, and Cosmetic Act . .....1938

Food additives amendment . .....1958Color additives amendments ... 1960New drug amendments . . . . . . . . . . 1962New animal drug amendments. .. .1968Medical device amendments .. 1976

Federal Meat Inspection Act. . ......1967Poultry Products Inspection Act .. ..1957Egg Products Inspection Act . ......1970Fair Packaging and Labeling Act .. ..1976

Public Health Service Act ... . .....1944

Occupational Safety and Health Act .1970Federal Hazardous Substances Act. 1960

Consumer Product Safety Act 1972Poison Prevention Packaging Act .. 1970

Lead-Based Paint Poisoning 1973, amended 1976Prevention Act .,

Hazardous Materials Transportation 1975, amended 1976Act .,

Federal Railroad Safety Act . . . . . 1970

Ports and Waterways Safety Act 1972D a n g e r o u s C a r g o A c t . . . . . 1 9 5 2Federal Mine Safety and Health Act. .1977

CPSC = Consumer Product Safety CommlsslonDOT = Department of TransportationEPA = Environmental Protection AgencyFDA = Food and Drug AdmlnlstratlonHEW = Health Educatton and Welfare

EPA

EPA

EPAFDA

FDAFDAFDAFDAFDAUSDAUSDAUSDAFDA

FDA

OSHA, NIOSHCPSC

CPSCCPSC

CPSC, HEW, HUD

DOT (MaterialsTransportationBureau)

DOT (FederalRailroad Admin.)

DOT (Coast Guard)

Labor (Mine Safety

All new chemicals (other than foodadditives, drugs, pesticides, alcohol,and tobacco) and existing chemicalsnot covered by other toxicsubstances control laws

Hazardous air pollutants

Toxic water pollutantsDrinking water contaminantsPesticides

Tolerances for pesticide residues infood

Hazardous wastes

Ocean dumpingBasic coverage of food. drugs,and cosmetics

Food additivesColor additivesDrugsAnimal drugs and feed additivesMedical devicesFood, feed, and color additives;pest icicle residues in meat andpoultry products

Packaging and labeling of food anddrugs for man or animals, cos-metics, and medical devices

Sections relating to biologicalproducts

Workplace toxic chemicalsHazardous (including toxic)household products (equivalent inmany instances to consumerproducts)

Hazardous consumer productsPackaging of hazardous householdproducts

Use of lead paint: on toys or furniture,on cooking, drinking, and eatingutensils, in federally assistedhousing

Transportation of toxic substancesgenerally

Railroad safety

Shipment of toxic materials by water

Toxic substances and other harmfuland Health Admin.), physical agents in coal or otherNIOSH mines

HUD = Housing and Urban DevelopmentNIOSH = National Institute for Occupational Safety and HealthOSHA = Occupational Safety and Health AdministrationUSDA = United States Department of Agriculture

SOURCE Environmental Law Institute, An Andlys(s 01 ~asl federal Efforts To Contro/ Tox)c .Subslances Washington D C 1978

Ch. Ill—Federal Laws. Regulations, and Programs ● 37

which FDA is empowered to license, andthereby limit, the use (or in the case ofpesticide residues, the occurrence) of poten-tially toxic substances in or on food.

The Pesticide Chemicals Amendment of1954, now section 408 of the FD&C Act, pro-vides that a raw agricultural commodity shallbe deemed to be adulterated if it bears or con-tains any residue of a pesticide that does notconform to a tolerance established under sec-tion 408, Pesticides that are unintentionallypresent on commodities are usually consid-ered environmental contaminants by FDA,and are regulated under section 406.

At no point, in 1938 or subsequently, hasCongress specifically addressed the problemof environmental contaminants in food. FDA

could have regulated them all under the “mayrender injurious” language of section 402(a)(1). But this provision would not have givenFDA authority to determine administrativelywhat levels of a contaminant could be toler-ated. FDA would have been required to proveits claim of hazard each time it seized a con-taminated product.

Accordingly, since the early 1970’s FDAhas classified environmental contaminants as“added poisonous or deleterious substances’whose occurrence cannot e n t i r e l y b eavoided, thus avoiding the less rigorous “or-dinarily injurious” standard of section 402(a)(l). The tolerance-setting authority of section406 can then be applied to environmentalcontaminants in food.

Relying on section 406, FDA prescribes thelevel of a contaminant that, under section402(a)(2)(A), will render a food adulterated.Before FDA can ascertain this level, suffi-cient scientific data must be accumulated toanswer several questions implicitly posed bysections 402 and 406. FDA must be able todetermine that the environmental contami-nant in question is:

1.2.3.

4.

added,poisonous or deleterious,a substance unavoidable by good manu-facturing practice, andone which may make the food injuriousto health. -

Furthermore, an analytical method that canreliably detect, measure, and confirm theidentity of the contaminant in the food underscrutiny must be available (5).

To determine whether these requirementscan be met, FDA scientists explore and re-view the scientific literature, consult F D Afiles, and draw on information available inother agencies, in academia, or in private in-dustry. Then, based on the best scientific

data available (which are often incomplete),FDA will prescribe what level of contamina-tion will trigger enforcement action (5).

Regulatory procedures employed to controlenvironmental contaminants in food includethe establishment of action levels or toler-ances. A formal tolerance is a regulation hav-ing the force of law. Tolerances are adoptedthrough formal rulemaking procedures andspecify the level of a contaminant that willrender a food adulterated. If supported bysubstantial evidence in the rulemaking rec-ord, FDA’s tolerance cannot be questioned byany court. An action level is an in.formed judg-ment about the level of a food contaminant towhich consumers may safely be exposed. It isa statement of FDA’s professional judgmentand represents a commitment to initiate regu-latory enforcement action against any lots offood discovered containing excess levels. Es-sentially the same criteria are considered inestablishing tolerances and setting action lev-els. The principal differences between thetwo approaches lie in the procedures fortheir adoption, the strength of the scientificdata supporting them, and the differingweight they carry in court (4).


FDA will set a tolerance when the follow-ing conditions exist:

1. The substance cannot be avoided bygood manufacturing practice.

2. The tolerance established is sufficientfor the protection of the public health,taking into account the extent to whichthe presence of the substance cannot beavoided and the other ways in whichthe consumer may be affected by thesame or related poisonous or deleteri-ous substances.

3. No technological or other changes areforeseeable in the near future thatmight affect the appropriateness of thetolerance established (6).

To establish a tolerance, FDA first pub-lishes a proposal, accepts comments, andissues a “final’* regulation, Formal objectionscan be raised to this “final” tolerance. Suchobjections can, if they raise material issues offact, require a lengthy trial-type hearingbefore an FDA administrative law judge, whothen issues an initial decision based on theformal hearing record. That decision, in turn,can be appealed to the FDA Commissioner(who issued the original “final” tolerance).The Commissioner’s ultimate decision is sub-ject to review in a court of appeals (l).

Because the tolerance-setting procedure iscumbersome and time-consuming, FDA ini-tially relies on an action level when it regu-lates an environmental contaminant. An ac-tion level is an administrative guideline andthe functional, though not legal, equivalent ofa section 406 tolerance. It is establishedwhen “technological or other changes thatmight affect the appropriateness of the toler-ance are foreseeable in the near future” (6).

To set an action level, FDA simply an-nounces in the Federal Register that it isestablishing an action level for a contami-nant, and states that the data supporting thedesignated level are available for public in-spection. This announcement may briefly dis-cuss the pertinent factors that went into thedecision, but any discussion is not likely to be(nor is required to be) extensive. While theannouncement also notes that public com-

ments will be accepted, FDA makes no com-mitment to respond to any comments or, in-deed, to reconsider the action level withinany specified period (6). The process does notrequire a detailed public discussion of theselected levels, nor does it trigger a publicdebate about the correctness of FDA’s prem-ises or its balancing of relevant factors.

Finally, the Commissioner of FDA may ex-empt from regulatory action any contam-inated food if he determines “based upon allavailable scientific evidence, that the food issafe for consumption and that destruction ordiversion of the food involved would result ina substantial adverse impact on the nationalfood supply” (6). This has only happenedonce, and the action did not involve humanfood,

If the environmental contaminant is a pes-ticide for which no tolerance has been estab-lished by EPA, FDA relies on EPA to recom-mend an action level (as it did in the case ofkepone). In other respects, the proceduresand criteria for regulating pesticides as envi-ronmental contaminants are the same as forother contaminants.

Criteria for Setting Action Levelsand Tolerances

For the setting of action levels or toler-ances, neither the law nor regulations re-quire FDA or EPA to follow a standardizedset of toxicologic protocols to evaluate risk.No policy exists defining the relative weightto be given to evidence. The burden of provingthere is a health hazard lies with FDA. Whensetting an action level or tolerance, FDA con-siders the following types of data:

1.

2.

available acute and chronic toxicologi-cal data, including information on thebiological half-life of the substance andits metabolic fate;available data on the levels and inci-dence of the contaminant in the overallfood supply and specifically in the foodcommodity or commodities that arebeing considered for an action level ortolerance;

Ch. IIl—Federal Laws, Regulations, and Programs ● 39

3.

4.

5.

6.

7.

In

normal serving sizes of the concernedfood(s) and frequency of ingestion;susceptibil ity of certain populationgroups, such as infants and the aged, toadverse effects from anticipated dietaryexposure to the contaminant:the level at which available analyticaltechniques can detect , measure, andconfirm the identity of the contaminant;capability of manufacturers to monitortheir food production to ensure that theproducts comply with the action level ortolerance; andthe anticipated impact of various possi-ble levels of regulation on the nationalfood supply.

response to questions on how FDA eval-uates these data, Commissioner Donald Ken-nedy wrote:

Each factor is assessed individually (as-suming information on each is available) andthen collectively brought into balance by acomposite analysis in terms of the estimatedrisk to the public health versus both the ex-tent to which the substance is unavoidableand the quantity of food that would be unlaw-ful under levels being considered (5),

FDA has not fixed the weight to be given toeach of the above factors. Each will, to somedegree, influence the final decision; general-ly, the more information about a particularfactor, the greater its influence, This is onereason that FDA offers for not prescribing apredetermined quantifiable set of criteria foreach factor. The amount and quality of infor-mation available when FDA encounters anenvironmental food contamination problemare inevitably unpredictable, FDA maintainsthat because of this uncertainty, it is imprac-tical to state in advance the precise weight ofeach factor in the final determination. How-ever, FDA maintains that the public healthfactor outweighs all others in its considera-tions (5).

Determining Action Levelsand Tolerances

In general, EPA and FDA follow similarprocedures when evaluating the health risk

associated with consumption of a toxic sub-stance in food, Both agencies consider threeareas when evaluating the scientific informa-tion: 1) existing animal toxicity data, 2) ex-isting human toxicity data, and 3) exposuredata based on the level of the contaminant infood and the average consumption of thatfood. Both agencies also consider what effectan action level or tolerance will have on theavailability of food.

Evaluation of the Scientific Data

When a food contaminant is identified, thefirst step in establishing an action level ortolerance is to assemble and evaluate allavailable information on its toxicity. This in-formation comes from articles published inthe scientific literature, information providedby private industry, and data from other Gov-ernment agencies. From animal toxicity dataand whatever human toxicity data may beavailable, a no observed effect level (NOEL) iscalculated and expressed in milligrams perkilogram of body weight per day. NOEL is thelevel at which the substance had no observedeffects.

An acceptable daily intake (ADI) is thencalculated by dividing the NOEL by a safetyfactor. The term “acceptable” does not implyabsolute safety for all people in all cases, andthe term “safety factor” implies more than itseems. The safety factor reflects the uncer-tainty of translating animal data to humans,the variability of the human population, theinsufficiency of the data available, and theseverity and reversibility of toxic effects.

When 2-year chronic toxicity studies inanimals are available, the safety factor usedis 100. When threshold levels have been ob-served in humans, the safety factor employedis 10. If long-term studies are available andshow no irreversible effects, a much smallersafety factor might be selected. If evaluationof available toxicological data indicates thatno threshold exists, a very large safety factor(on the order of 1,000 or more) maybe used.F. ‘I’. Arnold, chairman of the Kepone ActionLevel Hearings, stated that “the determina-tion of an appropriate safety factor is an art


rather than a science and is dictated by thechemical in question, its toxicological proper-ties and surrounding circumstances” (7).

The next step is to calculate the maximumpermissible intake. The maximum permissi-ble intake is the product of ADI and theaverage weight of an adult (this figure variesfrom 60 to 70 kg). This figure is then com-pared to the maximum potential exposurethrough consumption of contaminated food.Tolerances are set so that the amount of thecontaminant consumed in food is less than orequal to the ADI. Mathematically, this can beexpressed as:

ADI x Average body weight

Tolerance = of consumer

Food factor x 1.5 k gW h e r e ADI = the NOEL divided by an appropr ia te

safety factorAverage body

weight ofconsumer = 60 kg for EPA calculations, and 70 kg for

FDA calculationsFood factor = percentage of the average daily diet

made up by the food in question1.5 kg = the average weight of food consumed in

a day

Appendix B provides a detailed example ofhow these concepts were applied by FDA inthe development of tolerances for polychlori-nated biphenyls (PCBs) in food,

These procedures have several limitations.The data on toxicity of an environmental con-taminant for humans and animals may oftenbe inadequate for setting a tolerance. Be-cause of time constraints and the necessity tomake a regulatory decision, action levels forenvironmental contaminants in food may bebased on incomplete, scanty toxicological in-formation. Once a level is set, however, nolaw requires FDA to collect new evidence onthe toxicity of the substance in question eventhough that issue may well continue to be ofcritical importance to the population receiv-ing the highest exposure and to food pro-ducers.

Because of the nature of the problem pre-sented by environmental contaminants infood and because so little data are usuallyavailable, EPA and FDA cannot have formal

requirements for toxicity data. Moreover, inmost instances there are no petitioners orsponsors of commercial uses of the materialto which FDA can look for the necessary addi-tional tests. Available industry data are fre-quently used because of the lack of publishedor publicly available information on the sub-stance. While known toxic effects of metabol-ic products of the substance in question areconsidered, unknown metabolizes cannot be.Finally, additive and synergistic effects be-tween the contaminant and other toxic sub-stances in food are not considered in the tol-erance-setting procedures.

Questions are frequently raised aboutsafety factors and the extrapolation from ani-mal data to humans. Comparisons betweenanimals and humans are based on milligramsof the toxic substance per kilogram of bodyweight, although many maintain that milli-gram of toxic substance per square centi-meter of body surface area is a more appro-priate comparison,

Little scientific evidence exists to supportsafety factors. Historically, a factor of 10 inextrapolating from animals to man and a fac-tor of 10 in extrapolating from the least-sensi-tive human to the most-sensitive have beenused although they have little theoretical orfactual basis. Many believe that the use ofdifferent safety factors for different toxico-logical effects (a greater safety factor for ir-reversible effects, a smaller factor for rever-sible effects) or the use of mathematical mod-els to extrapolate from animal to human riskwould be more appropriate. Finally, the safe-ty factor approach may not make allowancesfor vulnerable groups such as infants, exceptin those instances in which infants are con-sidered the primary population at risk. Indi-viduals with predisposing conditions or previ-ous exposure may not be adequately coveredby safety factors. In truth, it may be impos-sible to protect every individual with allergiesor predisposing physiological conditions.

The methods used for estimating dietaryexposure to toxic substances are limited bylack of sound data. FDA bases some tolerancedecisions on their Total Diet Studies, Because

Ch. III— Federal Laws, Regulations, and Programs ● 41

these are based on the diet of a teenage male,they do not reflect the dietary patterns ofvulnerable groups, nor do they reflect ethnicand regional preferences or vegetarian diets.EPA and FDA rely on USDA’s Food Consump-tion Survey, which was completed in 1965-66.It is believed that some shifts have occurredin consumption patterns since then, andUSDA is now conducting a new Food Con-sumption Survey.

Evaluation of Decisionmaking Approach

As discussed earlier, FDA regulates en-vironmental contaminants in food under sec-tion 406 of the FD&C Act. This section doesnot specifically address environmental con-taminants, but authorizes FDA to regulatefood for potentially toxic substances that are“added” and “unavoidable’” in the produc-tion of food. When setting an action level ortolerance, FDA considers the impacts on thenational food supply or, stated in anothermanner, the impacts on the availability offood to the American consumer. In quanti-fy ing th i s c r i te r ion , FDA es t imates theamount of food that would be banned fromcommerce because of the action level or toler-ance.

The final rule reducing PCB tolerances (8)illustrates FDA’s interpretation of section406 of the FD&C Act for environmental con-taminants in food. FDA clearly states that forPCBs “(i)t has had to decide, in effect, wherethe proper balance lies between providing anadequate degree of public health protectionand avoiding excessive losses of food toAmerican consumers. ”

FDA later goes on to state that:

(I)n establishing a tolerance for PCBs infish, FDA must take into account the amountof fish a given tolerance would remove fromcommerce. Sect ion 406 of the Act, however,neither requires nor authorizes FDA toweigh secondary economic impacts when itconsiders the level at which a toleranceshould he set. Consideration of such impactswould be inconsistent with the paramountconcern of section 406, which is protection ofthe public health, and would complicate the

decisionmaking process under section 406 ina way Congress did not intend. Obviously,consideration of the amount of food losscaused by a tolerance helps to ensure thatthe direct economic consequences of thetolerance (in this case, decreased sales andemployment in the commercial fishing in-dustry) will not be disproportionate to the in-creased degree of public health protectionaccomplished by the tolerance: but FDA con-siders secondary economic consequences,such as potential impact on the recreationalfishing industry, totally beyond the scope ofsection 406 (8).

The decisionmaking process used by FDAis a form of cost-effectiveness analysis—aprocedure to compare the change of healtheffects in biological terms with the change ofthe cost in dollars (a further analysis of thisprocedure appears in chapter VI). In the PCBdecision, FDA compared change in humanrisk data for 5 parts per million (ppm), 2 ppm,and 1 ppm levels in fish with the estimatedamount of food that would be condemned inorder to arrive at a 2-ppm tolerance level. InFDA’s judgment, a 2-ppm tolerance was aproper balance between “providing an ade-quate degree of public health and avoidingexcessive losses of food to American con-sumers. ” While other factors were consid-ered by FDA in its decision, the estimatedhuman risk data and loss of food are the twoprincipal factors weighed in the decision.

The language of section 406 provides FDAwith the flexibil ity to interpret the un-avoidability requirement as it sees fit. FDArecognizes that its regulatory decision willhave an economic impact, but FDA considersonly a component of the total economic im-pact in its decision-i. e.. food condemned.FDA also realizes that the amount of foodcondemned will have an effect on employ-ment and commercial sales associated withthe contaminated food product. It must berecognized that even for such a widespreadcontaminant as PCBs, an action level for anenvironmental contaminant is likely to have amore severe impact on local employment andcommercial sales than on the amount of foodavailable to the American consumer.


In addition, FDA’s attempt to measure theimpacts on availability of food by estimatingcost of food condemned has several flaws.First, FDA estimates the amount of food ex-pected to be condemned for one year from aproposed action level or tolerance. Forsubstances such as PCBs in freshwater fish,estimates for the amount of food expected tobe condemned should be for more than oneyear. This is because PCBs are ubiquitous inthe environment and degrade very slowly. Itis highly likely that freshwater fish will becontaminated with PCBs at levels above 2ppm and consequently restricted from com-merce for several years.

Second, estimating the cost of food bannedby the tolerance does not necessarily reflectthe impact on the availability of food. Whilethis might occur for small amounts of foodcondemned—a herd of cattle or a few hun-dred gallons of milk—this would not be the

case for incidents that condemn significantamounts of food. A more accurate estimate ofthe impact on the availability of food wouldbe to consider the percentage or amount of afood product condemned out of the totalamount of that particular product that is pro-duced in this country, Then an estimate of therelative importance of the affected food prod-uct to the American diet would need to bemade, FDA does not do this second step. Thistype of analysis would also attempt to estimatethe impact such a tolerance would have onthe supply and price of other foods availableto and consumed by the American consumer.

While the latter analysis may be theoreti-cally sound, given the time constraints for set-ting an action level for a newly discoveredcontaminant, it may not be practical for aninitial regulatory decision. The more thor-ough analysis is more applicable to tolerance-setting.

FEDERAL MONITORING PROGRAMS

On the Federal level, the responsibility ofmonitoring foods for environmental contami-nants falls mainly on two agencies, FDA andUSDA. USDA limits its monitoring activitiesto meat and poultry products, while FDA ana-lyzes samples of animal feeds, fruits, vege-tables, grain, eggs, milk, processed dairyproducts, and seafood. Most monitoring activ-ities of these two agencies could be classifiedas regulatory monitoring-analysis of foodsamples for known environmental contami-nants and for some suspected environmentalcontaminants for enforcement purposes.

Food and Drug Administration

FDA collects approximately 8,000 food andfeed samples a year and analyzes them for avariety of chemical residues, mainly pesti-cides. Domestic food commodities compriseabout 70 percent of this total, imported com-modities make up around 30 percent. In addi-tion, approximately 1,300 seafood samplesare collected for analysis annually. Most foodand feed samples are collected at their point

of origin or processing, and attempts aremade to collect the seafoods as close to thepoint of origin as possible (9).

Some agricultural products are analyzedfor the presence of trace metals—lead, zinc,and cadmium —as well as for synthetic or-ganic chemicals. All fish samples collectedfor determination of chlorinated pesticidesand PCBs are also analyzed for mercury.Some canned-tuna samples are analyzed forlead, cadmium, arsenic, selenium, and mer-cury. Because some containers have lead-soldered joints that may contaminate thefood, an unspecified number of canned-foodsamples are analyzed to determine lead con-tent (9).

FDA also determines the total dietary in-take and exposure trends of some known andsuspected environmental contaminants in itsTotal Diet Studies. The contaminants in-cluded in this program include some pesti-cides, PCBs, mercury, lead, cadmium, arse-nic, selenium, and zinc, Other organics andmetals are excluded (9). Because it involves

Ch. Ill—Federal Laws, Regulations, and Programs ● 43

not the analysis of individual raw food prod-ucts but rather combinations of preparedcooked foods which approximate a totaldietary intake, this program differs from theregulatory monitoring activities of FDA.

Monitoring for known environmental con-taminants is carried out through the use ofspecified “accepted” methods for extraction,cleanup, and identification. The procedures,when applied to samples in which regulatedsynthetic organic chemicals such as pesti-cides are to be determined, will often indicatethe presence of other compounds that arechemically similar to those under analysis.The FDA Compliance Program GuidanceManual (10) specifies that if, during theregulatory monitoring analyses, unidentifiedanalytical responses appear (thus indicatingan uncharacterized chemical) with “signifi-cant” intensity, data from the sample collec-tion and analysis should be transmitted to theBureau of Foods Laboratory in Washington,D. C., which presumably will identify the un-characterized chemicals.

FDA may select a chemical for furtherstudy based on production volume, toxic by-products, environmental stability, volubility,behavior, toxicity, uses, and methods of dis-posal. After an analytical method is devel-oped, samples that have the highest probabil-ity of being contaminated with the particular(selected) compound are collected. Thesesamples are often fish, since rivers, lakes,and estuaries receive chemicals not only fromdirect discharges from municipalities and in-dustry but also from erosion and runoff. Fur-ther research and monitoring activities on agiven chemical depend on the results of theseinitial analyses (9).

The present radionuclide-monitoring pro-gram is a joint undertaking by FDA, EPA, andthe States, This program monitors: 1) foodsgrown near eight selected nuclear power fa-ci l i t ies for trit ium, gamma emitters, andstrontium-90; 2) food samples from the totaldiet studies; 3) specified imported foods; and4) milk, fruit, vegetables, and water collectednear phosphate mines in Florida (11).

EPA monitors milk, air, water, and soil forradioactivity. The milk-monitoring program isa joint effort with State and local agencies,FDA advises and monitors the milk-samplingprogram, which is carried out by 63 State andlocal health agencies. Milk from each area issampled once a month by the State and/orlocal inspectors and submitted for analysis tothe EPA laboratory in Montgomery, Ala.

U.S. Department of Agriculture

USDA is responsible for evaluating thequality of meat and poultry products and pro-viding the consumer with products that meetthe criteria spelled out in the Meat andPoultry Inspection Acts. These criteria in-clude monitoring for environmental contami-nants. The majority of compounds evaluatedare those that are approved for use in agri-culture, either administered directly to foodanimals (such as growth promoters), or ap-plied to agricultural crops to which food ani-mals may eventually be exposed (such as pes-ticides).

Testing of meat and poultry products forresidues by USDA falls into broad monitoringand surveillance categories. The monitoringactivity, called the National Residue Monitor-ing Program, is designed to determine thefrequency at which tolerance-exceedingamounts of monitored compounds are occur-ring in the national meat supply. In effect, themonitoring program is designed to evaluatehow effectively users and/or manufacturersof the compounds are complying with thelaws or use restrictions (12).

Under the monitoring program, animal tis-sues are collected from slaughterhouses un-der Federal inspection throughout the UnitedStates at a rate that will detect violations ifthey are occurring in at least 1 percent of theanimal population [13). Based on statisticalcalculations, 300 samples per compound perspecies have to be collected annually to deter-mine a l-percent incidence with 95-percentassurance. In effect this means that the samesample of tissues may be analyzed for morethan one compound. This level of testing re-


quires sampling approximately 1 in 8,000head of livestock and 1 in 700,000 poultry.These collections are stratified according togeographic areas, with more samples beingcollected from areas where meat or poultryare slaughtered.

USDA does not monitor all compounds forwhich tolerances have been established.There are no suitable methods to analyzesome of them within existing regulatorymonitoring laboratory capabilities. Availableresources may also limit the number andvariety of compounds tested, The selection ofwhich compounds to monitor is based on fac-tors such as frequency and patterns of use,toxicity, previous testing results, and publicconcern. Major groups of compounds in themonitoring program include synthetic organicchemicals (mainly chlorinated hydrocarboninsecticides), trace metals, antibiotics, sulfon-amides, and certain hormones and drugsused for growth promotion and disease con-trol. Organophosphate pesticides were moni-tored for several years but this monitoringwas discontinued because residues of theparent compound were not found in animaltissue and suitable routine methods for de-tecting the metabolizes were not practical(12),

In 1978, about 150,000 tests were com-pleted on approximately 20,000 domesticsamples collected under the national residuemonitoring program. An additional 2,000samples were collected from imported prod-ucts (12). Many of these samples are analyzedfor potential contaminants other than tracemetals and synthetic organics. In 1977, for in-stance, around 2,300 out of 22,000 sampleswere analyzed for chlorinated hydrocarbonpesticides and 1,300 for trace metals. The re-mainder were analyzed for antibiotics, hor-mones, and drugs (13). Data generated by thisprogram are used not only for regulatoryfunctions but also to help pinpoint problems,assist in trend analysis, and indicate areasthat need more intensive sampling.

Surveillance samples are those collected toevaluate a problem. The area of samplingmay be as small as a single farm or as large

as a State or region, depending on the cir-cumstances. Indications of problems comefrom many sources: the National ResidueMonitoring Program, activities of USDA in-spectors, information from State or otherFederal officials, or from public newssources. Because these samples are biased(i.e., collected in response to a given need),they do not reflect the overall condition of thenational meat supply. In most cases the sam-ples are used to determine either the extentof a problem or to evaluate the acceptabilityof a herd or product. Any product found to bein violation may not be released into com-merce until subsequent samples show that itis in compliance. These followup samples areconsidered surveillance samples,

Analyses of samples collected in USDAprograms are performed in a manner similarto those in the FDA program—prescribedanalytical methods are used. USDA, like FDA,may try to identify an “uncharacterized” sub-stance when “unknown” peaks appear in theanalysis of a sample. Depending on the identi-ty of the compound, further investigation andsampling may be carried out,

Environmental Protection Agency

EPA has no mandate to regularly analyzefood commodities for chemical contamina-tion. Some of its programs designed to deter-mine the ecological impact of pollutants mayinclude certain types of foods for analysis.This is particularly true in the case ofseafood. Samples from aquatic food chainsare often selected for analysis to ascertainwhether a pollutant is concentrated as itmoves up the food chain.

In addition, the mobilization, degradation,and transfer of pollutants is often studied byEPA. For example, under the auspices ofEPA’s Chesapeake Bay program, trace metalsand synthetic organic chemicals are beingstudied in water, living organisms, and sedi-ments of the Chesapeake Bay by various Stateagencies and academic institutions. The re-sults of this and similar studies are not de-signed to protect the public from consumingcontaminated food and are seldom used as

Ch. Ill—Federal Laws, Regu/at/ens, and Programs ● 45

such. Rather, the results are used in protect-ing the environment from chemical insultsand helping EPA better regulate the introduc-tion of toxic substances into the environment.

Some data transfer between EPA and otherFederal agencies exists. For example, in theEPA-funded kepone studies in the Chesa-peake Bay region, kepone concentrations inedible fish, shellfish, and crabs are obtainedwhich assist the Commonwealth of Virginiaand FDA in their efforts to keep contaminatedsea food off the market. This is true eventhough the study is not designed specificallyfor this purpose.

Another program administered by the EPAis the Mussel Watch (14). It is designed toanalyze shellfish collected from strategiclocations in the marine coastal zone of theUnited States for selected organic chemicals,trace metals, and radionuclides. This pro-gram was started in 1976 and now involvesthe collection of oysters and/or mussels from29 stations on the west coast, 34 stations onthe east coast , and 26 stat ions on the gulfcoast , Collect ions from these s tat ions makeup a total of 107 samples,

The t race metals for which samples areanalyzed in this program include lead, cad-mium, silver, zinc, copper, and nickel. The ra-dionuclides measured in the samples includeplutonium-238 and -239, americum-241, ce-sium-137, curium-242 and -244, and lead-210.The synthetic organic chemicals included inthe analysis are the halogenated hydrocar-bons p,p’-DDE and p,p’-DDD—two of the prin-cipal breakdown products of the pest ic ideDDT—and PCBs. Samples are analyzed forthese synthetic organic chemicals as well asp e t r o l e u m h y d r o c a r b o n s , w h o s e p r e s e n c emay indicate oil pollution. Recently, a numbero f u n c h a r a c t e r i z e d s u b s t a n c e s h a v e b e e nfound by Mussel Watch scientists. They aren o w d e s i g n i n g t h e p r o g r a m t o e n c o m p a s ssystems to track and identify them (15).

Critique of Federal MonitoringPrograms

A chemical substance found in food be-comes the subject of Federal regulatory moni-

toring if it has caused a problem at some timein the past. In other words, a compound mayenter the food supply and be undetected foryears but categorized as a known contami-nant when discovered. An example would bekepone in seafood of the James River in Vir-ginia (16). This compound is an insecticidethat had been entering the aquatic food chainfor at least 7 years before discovery. Its pres-ence in seafood was determined after work-ers in the facility that manufactured it be-came ill from industrial exposure to this toxi-cant (17)0

As soon as the sick workers were discov-ered, samples of fish and oysters from the ad-jacent river were analyzed and shown to con-tain kepone. Within a few months, actionlevels were established for kepone in seafood.Because existing monitoring programs ana-lyze for known regulated environmental con-taminants, the compound was not discoveredbecause it was not sought. It is unlikely thatthe presence of kepone would be known andregulated today had the workers not becomeill. In this case, kepone was placed in theknown-environmental-contaminants categoryas a result of the illness of productionworkers, not chemical monitoring.

Another example of a toxic compound en-tering the food supply and going undetected isthe fire-retardant polybrominated biphenyls(PBBs), PBBs entered the food supply in Mich-igan in the late spring or early summer of1973 (18,19), Bags of a fire retardant contain-ing these compounds were shipped to farm-ers’ cooperative units in Michigan in place ofthe intended livestock feed supplement, Thefire retardant was unknowingly mixed withcattle feed and fed to herds throughout thesouthern part of the State. Even though farm-ers soon noted the symptoms of poisoning inthe cattle and reported the problem to Stateand Federal officials, it was almost a yearbefore the causative agents, PBBs, were iden-tified. During this year dairy products andbeef contaminated with these compoundswere consumed by the citizens of Michigan.Soon after the compound was identified infood products, action levels were establishedto protect the consumer. In this case the trig-


ger for designation of PBBs as known environ- check for known environmental contaminantsmental contaminants was sick cattle, not are insufficient to detect toxicants in food forchemical monitoring. which there have been no action or tolerance

From these two case studies it appears levels established.that monitoring programs designed mainly to

CHAPTER III REFERENCES

1. Federal Food, Drug and Cosmetic Act, 21U.S.C. 321 et seq.

2. Poultry and Poultry Products Inspection Act,21 U.S.C$ 451 et seq.

3, Federal Meat Inspection Act, 21 U.S.C. 601 etseq.

4. Merrill, R, A, FDA Re,gu~ation of Food Safety,unpublished manuscript, 1978.

5. Kennedy, D., Commissioner, Food and DrugAdministration, to J. B. Cordaro, Jan. 22,1979,

6.21 CFR 109.7. Arnold, F. T. Kepone Action Levels—Review,

Analysis and Recommendations, Office ofSpecial Pesticide Reviews, EnvironmentalProtection Agency, Mar. 9, 1977.

8. “Polychlorinated Biphenyls (PCBS), Reductionof Tolerances, ” Federal Register 44(127):38330, 1979.

9. Food and Drug Administration. Report in Re-sponse to the Office of Technology Assess-ment letter of June 5, 1978—’’Request for In-formation on Federal Capabilities for Envi-ronmental Mcmitoring and Toxicity Testing, ”1978.

10. Food and Drug Administration. ComplianceProgram Guidance Manual, Compliance Pro-gram 7305.007 -’’Pesticide and Metals inFish Programs” (FY ‘78), Oct. 1, 1977.

11. Food and Drug Administration, ComplianceProgram Guidance Manual, Compliance Pro-gram 7305.006-’” Radionuclides in Foods”(FY “78), Oct. 1, 1977.

12. Clark, G. U.S. Department of Agriculture,1979.

13. U.S. Department of Agriculture. The NationalResidue Program, USDA-Food Safety andQuality Service, May 1978.

14. Goldberg, E, D., V. T. Barrington, J. W. Har-vey, G. H. Martin, J, H. Parker, P. L. Rise-brough, R. W. Robertson, W. Schneider, E. S.and E. Gamble. “The Mussel Watch, ” Envi-ronmental Conservation, vol. 2, Summer1978,

15. Risebrough, R. W. Bodega Marine Labora-tory, University of California, Bodega Bay,Calif., personal communication, 1978.

16. Huggett, R. J., M. M. Nichols $ and M. E.Binder. “Kepone Contamination of the JamesRiver Estuary, ” Proceedings of Symposium onContaminants and Sediments: Fate andTransport, R, A. Baker (cd.), Honolulu,Hawaii, April 1979 (in press),

17. Food and Drug Administration. ComplianceProgram Evaluation (FY ‘77), Kepone andMirex Contamination (7320.79A).

18. McNally, J. M, “Pol~brominated Biphenvls:

19

Environ-mental Conta-mination of Food, ” spe-cial paper for OTA, Library of Congress, Con-gressional Research Service, 1978.Michigan Department of Agriculture, Infor-mation and Education Division. “A BriefChronology of the PBB Incident” (no date).

Chapter IV

State Laws, Regulations,and Programs

. .,=-

Chapter IV

State Laws, Regulations,and Programs

STATE LAWS

State food and drug laws, and the organizations that administer them, varywidely. Basic State food and drug statutes are based on the Federal food laws;however, not all States have adopted the model uniform State food, drug, and cos-metic bill of the Association of Food and Drug Officials. As shown in figure 3, 42States have adopted the model statute, which is almost identical to the 1938 Fed-eral Food Drug and Cosmetic Act. Consequently, these 42 States have the samelegislative authority for regulating food contaminants within their borders as theFederal Government does for food in interstate commerce. The 1906 Act, still re-tained by eight States, does not contain the tolerance-setting provisions of the1938 Act under which environmental contaminants are regulated (l).

Authority for regulating environmentalcontaminants in food rests with two or moreagencies in most States. Usually the Depart-ment of Agriculture and the Department ofHealth share responsibilities for food regula-tion. In some States a variety of other agen-cies and bodies are also involved in regula-tory or research activities: departments deal-ing with commerce, fish and game, consumerprotection, environmental improvement, pub-lic administration, conservation, along withuniversity divisions, independent laboratoryagencies, and various independent boardsand commissions (2).

The Food and Drug Administration (FDA)believes that this variability in laws and reg-ulatory organizations makes it more difficultfor the States to accomplish their goals:

The variability of organizational structurecomplicates the problems of many of the in-dividual State agencies in accomplishingtheir program goals because of overlappingresponsibilities and the lack of a clear delin-eation of responsibilities. For example, it isnot uncommon to find authority granted totwo agencies for some divided program seg-ments of a single program category (e.g.,

Figure 3.— State Food Laws

50r 42

40 —

30 —Inu%z

20 —

10 —

o —

f

8—

4FD&C Act 1906

25

FoodUniform FD&C Act Additives

20

2L

34

f Enrichment

I Pest ReqAmendment

(Food ) ColorAdditives

SOURCE Food and Drug Admtnlstratlon State Programs and Serwces m Foodand Drug Control, 1978

49

milk, shellfish). Frequently, two or more instances, the State agency has an unclear roledependent agencies of relatively equal rank as an advisor or consultant to the local gov-are charged with enforcement of portions of ernment. However, the local agency may notthe same general food and drug law. In still be legally bound to follow the advice and/orother States, there is no central State control direction that may be suggested by the Stateover the food and drug program. In these in- agency (1).

STATE Monitoring PROGRAMS

Non-Federal monitoring is limited almostexclusively to programs originated and car-ried out by individual States. In some casesthere is close coordination between State andFederal activities. Federal agencies may de-crease their monitoring in a given area ifState monitoring is considered sufficient. Forexample, FDA does not monitor seafood forkepone (3) even though the concentrations infinfish, crabs, or shellfish remain essentiallyunchanged since its discovery in 1975 (4).FDA feels that the ongoing monitoring pro-grams in Virginia are sufficient to protect theconsumer.

Analysis of the OTA State survey revealsthat State food-monitoring laboratories areequipped to analyze for those substances thatare regulated in foods through action levelsor tolerances. For instance, most had instru-mentation (atomic absorption spectrophotom-eters and gas chromatography) to analyze formercury and chlorinated hydrocarbon pesti-cides.

FDA published a study entitled “State Pro-grams and Services in Food and Drug Con-trol” in September 1978 (l). The publicationprovides a compilation of States’ analyticalcapabilities during the years 1975 and 1976,listing numbers and educational levels ofchemical analysts, types of analytical equip-ment, and expenditures for food inspection.

In 1974 and 1975, the States spent annual-ly about $64.9 million for food inspection andanalytical activities. This amounted to 72 per-cent of their total inspection and analyticalexpenditures (l). Table 8 shows programareas and expenditures for States. These

Table 8.—Program Areas and Expenditures forStates

Expenditures

Millions of Percent of totalProgram dollars (food and drug)

Food ‘. . . . . . . . . . . . . . . . . . $64.9 72.3Drugs, devices, cosmetics . . . 9.7 11.0Feed . . . . . . . . . . . . . . . . . . . . 6.4 7.0Weights and measures (food) . 5.5 6.0Pesticides ... , ... , . . . . . . . 3.3 3.7

Total . . . . . . . $89.8 100.0

SOURCE Food and Drug Admlnlstratlon S/ate Progrdms and Serv/ces /rJ foodand Drug Contro/, 1978

figures do not include the estimated $75 mil-lion expended by local governments.

The educational levels of the chief chemist,supervisory chemist, and chemist and labora-tory technicians are presented in table 9 (l).The salary ranges for the chemical personnelare shown in table 10 (l). Only 1 percent ofthe chemists working in State food and drugprograms earned more than $20,000 per yearin 1975 and 1976. Approximately 64 percentof them had annual salary ranges of $8,000 to$15,000.

The available analytical equipment andphysical facilities are listed in table 11 (l).These data confirm the OTA survey findings,since gas chromatography and atomic ab-sorption spectrophotometers rank first andthird, respectively, in numbers. A breakdownof food commodities analyzed, samples col-lected, and analyses performed is given intable 12 (l). The FDA document urges cautionin interpreting these data because “. . . someStates do not maintain comprehensive ana-lytical records on food analyses especially iffood is not the major laboratory workload.”

Ch. IV—State Laws, Regulations, and Programs ● 51

Table 9 .—Number of Employees by Category and Education Level in Various State Food and Drug Programs

Personnel ‘- Number of employees Advanced degree College degree Some college No college

Chief chemist. . . . . . . . . . 109 68 ‘ -4 0- - 1 0S u p e r v i s o r y c h e m i s t 232 62 167 3 0Chemist. . . . . ... . . . . . 633 66 561 6 0Laboratory technicians . . 833 0 127 375 332

SOURCE Food-and-Drug Adminlstratlon Slate Programs and Serv/ces (n food and Drug Conlro/ 1978

Table 10.—Salary Ranges of Various Chemical Personnel in State Food and Drug Programs

S a l a r y - r a n g e s

$ 6,000- 9,000- . . . . . . . . . . . . . . .$ 7,000-10,000 . . . . . . . . . . . . . . . .$ 8,000-12,000 . . . . . . . . . . . .$ 8.000-14,000 . . . . . . . . . . . . . .$10,000-15,000 . . . . . . . . . . . .$12,000 -18000. . . . . . . . . . . . . . . . .$16,000-20,000 . . . . . . . . . . ... . .over $20,000 . . . . . . . . . . . . . . . .

Number of chief chemists

o0022

1771

8

N u m ber of s u pervisory chemists

300

1714353

70

SOURCE Food and Drug Admlnlstratlon- Sfate Programs and Sew/ces In Food arrd Drug Cor?tro/ 1978 -

Table 11 .—Physical Facilities and Key Equipment in114a State Food and Drug Laboratories

Number of keyKey equipment items equipment itemsb

Spectrophotometers . . . . . . . . . . . . . - 269Flame photometers . . . . . . . 90Atomic absorption spectrophotometers . . 125S p e c t r o f l u o r i m e t e r s . 88Gas chromatography . . . . 348Polarographs. ... . 42Polarimeters . . . . . 54Mass spectrometer. ., . . 19Electrophoresis. . . . . ., . . 86Auto analyzers. ., ... . . . . . . . . . . 120Liquid chromatography . . . . . . ., . . . . . 18

Physical facilitiesAverage floor space (ft.2 ) . . . . . 10,245Average bench space (tin, ft.). . . . . . . . . 845A v e r a g e s t o r a g e ( f t . z ) 1,150

aTwo States not reporting lab equipment or spacebshows only total number of equipment Items reported by the s!ate a9encles

Does not show those labs that do not have one or more of the equipment itemslisted

SOURCF Food and Drug Admlnistratlon Stale Programs and Services In Foodand Drug Con Ire/ 1978

Number of chemists

o2670

153139

500

This probably accounts for the extensive ana-lytical activity reported in the “other food”category, rather than in the categorical pro-gram areas”’ (l). Even so, it is evident thatmilk and milk products are the most common-ly sampled and analyzed food commodities.

The large number of samples and analysesindicate that the States perform extensivemonitoring for regulated contaminants infood, But the low salary ranges, the lack of so-phisticated analytical equipment such asmass spectrometers, the time spent per anal-ysis, and the sample-type distribution indi-cate that State monitoring programs are asinadequate as Federal programs in detectingenvironmental contaminants in food forwhich no action or tolerance levels have beenestablished,


Table 12.—Total Number of Samples, Sample Determinations, and Man-Hours by Food Commodity CategoriesWith Number and Percentage of States Reporting Analytical Activity

Number of StatesFood commodity reporting analyticalcategories activities

Bakery products. ., . . . . . 35Soft drinks . . . . . . . . . . . . . 35Candy . . . . . . . . . . . 29Grade A milk (raw) . . . . . . 50Other milk products. 47Canned foods. . . 35Frozen foods . . . . . . . 27Seafood. . . . . . . . . . 34Shellfish . . . . . . . . . . . . 24Raw agricultural products 29Other foods. . . . . . 40

T o t a l s

Number of samplesPercent of States (thousands)

70 11.570 18.758 7.0

100 342.094 200,770 21.354 7.368 6.648 37.658 40,180 103,0

7958

Number ofdeterminations

(thousands)

23.364.618.1

1,157,1546.887027.634.4

106.389.6

326.3

2,481,1

Number of man-hours (thousands)

31.517,513.6

319.0185.648.519.333.267.6

112,7220.1

1,068.6

SOURCE Food and Drug Adnllnistrallon S/ate Programs and Serv/ces In Food and Drug Control 1978

FEDERAL/STATE LIAISONMany environmental contamination inci-

dents are initially State problems. Theoreti-cally, the Federal Government does not be-come involved until a contamination incidentis determined to be an interstate problem.Given the complexity of this country’s food-marketing system, most food produced orprocessed within a particular State is distrib-uted for consumption in other States. Thus,most environmental contamination incidentsare likely to become interstate concerns.Figure 4 reveals the extent of food contamina-tion that can occur from a single source ofcontamination, in this instance polychlori-nated biphenyls (PCBs) contaminated animalfeed from a meatpacking plant in Billings,Mont. (5). This widespread contamination offood occurred during an estimated time peri-od of 2 to 5 months.

The States and the Federal Government es-tablish liaison when contamination crossesState lines. Liaison is also established at therequest of States. States often require Fed-eral assistance in investigating a contamina-tion incident. The objective is to generate sci-entific information on the nature and extentof the contamination. This information wouldinclude the toxicological and chemical prop-erties of the contaminating substance, theamount and type of food contaminated, andthe concentration of the substances in food.

Such information is used by State and Fed-eral authorities to: 1) determine the appropri-ate Government response for protecting thepublic health and 2) inform the general publicabout the incident and explain the Govern-ment response. To contain an incident, scien-tific information needs to be accurate and im-mediately available. This is as true for an epi-sode involving a substance that has previ-ously contaminated food (1979 PCB contami-nation of animal feed in Billings, Mont. ) as fora substance which has not (1973 polybromi-nated biphenyl (PBB) contamination of animalfeed in Michigan).

The generation and dissemination of scien-tific information on an incident is hindered bythe number of State and Federal agencies in-volved. As already noted, three Federal agen-cies (the Environmental Protection Agency(EPA), FDA, and the U.S. Department of Agri-culture (USDA)), each with different respon-sibilities, can be involved along with variousState agencies. The PBB incident in Michiganand the PCB incident in Montana reflect thisparticular problem.

Generating Information

Before the needed scientific informationfor developing regulations is generated, the

Ch. IV—State Laws, Regulations and Programs ● 53

4I

.L_-

.54 . Env i ronmenta l Contaminants in Food

contaminating substance in food must beidentified. This identification process may re-quire lengthy investigative work which couldbe hindered by a multiplicity of involved Fed-eral and State agencies. The PBB incidentprovides an example of the extensive investi-gations sometimes necessary.

The Michigan Department of Agriculture(MDA) initially analyzed blood and feed sam-ples from a herd owned by Mr. Frederic Hal-bert (at his urging). These samples proved tobe negative because MDA was not analyzingfor PBB. Additional samples were analyzedby the USDA’s National Animal DiseaseCenter in Ames, Iowa. While the DiseaseCenter detected an unusually high peak of anunidentified substance, it did not determineits identity. The substance was eventuallyidentified by Dr. George Fries of USDA’sAgriculture Research Center at Beltsville,Md., who had previous experience in chemi-cal analysis of PBB (6). The discovery camenearly 8 months after adverse symptoms oc-curred in the herd at the Halbert farm.

The laboratories and agencies involved inidentifying PBB are part of the agriculturesystem in the United States. They were theobvious institutions to which Mr. Halbertwould go for assistance. If he had initiallycontacted the Michigan Department of PublicHealth (MDPH), it is likely that differentFederal laboratories and agencies wouldhave analyzed the samples. There was andstill is no systematic procedure at the Federallevel for assisting States in identifying apotential food contamination problem likePBB. Were it not for the perseverance of Mr.Halbert and the experience of Dr. Fries, theidentity of PBB might have taken longer andthe people of Michigan would have been ex-posed to PBB-contaminated food even longer.

Once a contaminant has been identified infood, the necessary scientific information canbe generated for either regulating or control-ling the contamination. In the PBB incident,FDA helped to develop and evaluate this in-formation and worked with MDA in control-ling the incident. FDA’s involvement wasbased on the fact that it has the Federal

authority for regulating contaminants pres-ent in food. Examples of actions taken byState and Federal authorities in such in-stances include removing food from the mar-ket, setting action levels or tolerances for thecontaminant in food, and disposing of the con-taminated food. Because little scientific in-formation on PBBs was available, it took timeto generate the information and establishfinal permissible levels (although contami-nated herds and food were identified asquickly as possible and removed from themarket). The PBB incident involved FDA,USDA, MDA, and MDPH. Even more Stateand Federal agencies were involved in thePCB contamination episode in the WesternUnited States.

The PCB contamination began in Montanawith animal feed and quickly spread to 16other States. Idaho was particularly affectedby the contamination of poultry and eggs.This incident involved all three relevant Fed-eral agencies— FDA, EPA, and USDA—aswell as the Idaho Departments of Agricultureand Health and Welfare, and district healthdepartments. USDA made the initial analysisof poultry samples which proved positive forPCB; FDA was involved with the removal ofPCB-contaminated food from the market; andEPA with the proper disposal of the contami-nated food. At least 5 days elapsed from thetime USDA was confident it had a PCB con-tamination incident to the time it notified FDAof its findings. It took FDA an additional 5days to begin its investigation of the contami-nation incident (7). Such delays would beunlikely if only one Federal agency were in-volved or communications between the twoagencies were better.

PCB is a substance whose chemical andtoxicological properties are fairly wellunderstood. It has contaminated food in thepast. Nevertheless, there was confusionamong the State agencies in Idaho as to theproper response to the contamination. Theconfusion resulted from two conditions. First,some of the State officials involved were notfamiliar with the chemical and toxicologicalproperties of PCB. PCB was a new food con-

Ch. IV—State Laws, Regulations. and Programs ● 55

taminant in Idaho, and the appropriate of-ficials had little experience with this type ofproblem. Second, the involvement of three dif-ferent Federal agencies obstructed eff icientcommunication between the State agenciesand the Federal Government.

While the Federal agencies had the mostexpert ise on PCBs, the sharing of that ex-perience was hindered by the fact that it wasavailable in and distributed by three sourcesinstead of one. Consequently, State agencieshad to go to different Federal sources, de-pending on what information they needed. Inaddition, the Federal agencies did not alwayscommunicate with the various State agencies.EPA, for example, took air and water samplesin the area surrounding Ritewood Farms todetermine whether PCBs that contaminatedRitewood’s eggs and poultry came from eitherof these two sources. This was before thePCB-contaminated animal feed was identi-fied. EPA, however. did not report their nega-tive findings to the State. In another instance,USDA initially would report its results only tothe Idaho Department of Agriculture, theState agency with which USDA has had along-standing association. The Idaho Depart-

ment of Health and Welfare, which is con-cerned with protecting the public health, atfirst was not informed by either USDA or theState Department of Agriculture of the PCBcontamination. Communication broke downat two levels, between the State and FederalGovernments and within the State govern-ment (8). The fact that there are several dif-ferent Federal and State agencies involvedwith different aspects of controlling andregulating a contamination incident furthercomplicates an already complicated problem.

The major environmental contamination in-cidents that occurred in Idaho and Michigancontinue to be major issues of concern amongthe residents of these States--a result oftheir fears over a potential health threat thatcannot be seen, smelled. or tasted. In Michi-gan, for instance, the PBB episode remains alive and controversial political issue. Conse-quently, it becomes imperative that the in-formation generated by the State and/or Fed-eral Government on an incident is accurateand appropriately applied. This objective ishindered by the variety of State and Federalagencies that become involved.

CHAPTER IV REFERENCES

1.

2.3.

4.

5.

U.S. Department of Health. Education, andWelfare, Food and Drug Administration.State Programs and Services in Food andDrug ~ontrol, 1978.OTA Survey of 31 States, 1978.Food and Drug Administration. ComplianceProgram Evaluation (FY ‘77), Kepone andhlirex Contamination ( 7320. 79A).Huggett, R. J., hi. hf. N icho l s , and M. E.Binder. “Kepone Contamination of the JamesRiver Estuary, ” Proceedings of Symposium onCon tc~minan ts and Sediments: F e t e andTransport. R. A. Baker (cd.), Honolulu, Ha-waii. April 1979 (in press).Gardner, Sherwin, Acting Commissioner ofFood and Drug Administration. Testimony be-

6.

7.

8.

fore Subcommittee on Oversight and Investi-gation, House Committee on Interstate andForeign Commerce, Sept. 28, 1979.Michigan Department of Agriculture, Infor-mation and Education Division. ‘‘A BriefChronology oft he PBB Incident ,“ (no dat e).Foreman, Carol Tucker, Assistant Secretaryfor Food and Consumer Services, U.S. Depart-ment of Agriculture. Testimony before theSubcommittee on Oversight and Investiga-tion, House Committee on Interstate and For-eign Commerce, Sept. 28, 1979.Gallagher, Ed, State Health Officer, Idaho De-partment of Health and Welfare, personalcommunication, Oct. 10, 1979.

Chapter V

Methods for AssessingHeaIth Risks

Chapter V

Methods for Assessing Health Risks

The availability and quality of data on the toxicity of contaminants in largepart determine the Food and Drug Administration’s (FDA) ability to protect publichealth. Chapter III reviewed the strengths and weaknesses of FDA’s proceduresfor setting action levels and tolerances. This chapter addresses four issues re-lated to the methods used for assessing health risks:

●

●

●

●

What is the role of human epidemiology in the setting of action levels andtolerances?What are the roles of animal and other toxicological tests?Are current testing procedures adequate for the assessment of interac-tions of toxic substances?Are methods for quantitative risk assessment sufficiently developed forapplication to environmental contaminants in food?

POSSIBLE TOXIC EFFECTSThe possible toxic effects of an environ-

mental contaminant depend on its chemicalnature, its concentration in food, and the typeof toxic action involved. If the substance ishighly toxic and/or is consumed in large quan-tity, acute toxic effects may occur and theonset of the symptoms would be rapid andsevere. If a small amount of a highly toxicsubstance is consumed, or if the substance isof low toxicity, the health effects may not beseen for many months or years (or may not beobserved at all). Potential effects of toxic en-vironmental contaminants in food include sys-temic toxicity, mutations, cancer, birth de-fects, and reproductive disorders.

Systemic toxicity involves changes in thestructure and function of organs and organsystems: weight change, structural altera-tions, and changes in organ system or wholeanimal function. Functional effects may in-clude changes in the lungs, liver, and kidneys,in cardiovascular function, in brain and nerv-ous system activity and behavior, and in re-sistance to disease (1). Systemic toxicity isstudied in whole animal tests.

Some environmental contaminants havebeen shown in experiments to cause point mu-

tations (which generally affect a single gene)or more extensive effects such as grosschanges in chromosome structure or changesin chromosome number. Such genetic effectsin humans often cannot be immediately de-tected. Indeed, damage to the human genepool may not become apparent for many gen-erations if the deleterious effect is due to a re-cessive gene (2). However, a chemical’s muta-genic potential can be evaluated indirectlyfrom various biological tests involving micro-organisms, mammalian cell cultures, insects,and whole mammals (l).

Some environmental contaminants in foodmay cause cancer. Direct cause-and-effectassociations between exposure to a specificchemical and human cancer are difficult toestablish because of the complex nature ofcancer and the vast number of potential car-cinogens to which humans are exposed. Insome cases, exposure to one toxic agent maytrigger a sequence of events leading tocancer. In others, carcinogenesis may dependon interactions of several factors, combina-tions of noxious agents, co-factors, andnatural or acquired metabolic peculiarities(2). The cancer-causing potential of a sub-

59

stance is evaluated through animal tests anda variety of short-term tests employing bacte-ria, insects, or cell cultures (1).

Environmental contaminants may producebirth defects other than inherited mutations.These are called “teratogenic effects. ” Mal-formations may include gross, histological,molecular, and sometimes behavioral anoma-lies. Human sensitivity to a teratogenic agentduring pregnancy is determined by: 1) thetime at which the insult is received duringgestation, 2) the dose of the compound,3) transfer of the teratogen to the developingfetus, 4) uterine factors, and 5) dietary fac-tors. The teratogen may work its effect byproducing mutations or chromosomal aberra-tions, interfering with cell division, alteringnucleic acid synthesis, inhibiting specificenzymes, or altering membrane characteris-tics. A seemingly small change may have far--reaching effects, since the fetus is undergoingrapid biochemical, structural, and physiologi-cal growth and change (1).

Environmental contaminants can haveother effects on the reproductive process inboth males and females. Toxic effects may in-clude reduced or altered sperm formation, in-hibition of ovulation, increased spontaneousabortion, fetal resorption, and increasednumbers of stillborn infants, Teratogenic andreproductive effects are evaluated throughanimal tests (l).

Given the range of possible adverse healthimpacts, it is clear that newly discovered en-vironmental contaminants must be subjectedto the best available toxicological testingtechniques so that any harmful effects can beuncovered. Furthermore, regulators musthave information on the possible toxic effectsof ingesting small amounts of a substance infood over an extended period of time, per-haps over a lifetime. It would also be desi-rable to know what effects other toxic sub-stances already present in our air, water,and food may have on the metabolism andtoxicity of a new contaminant,

HUMAN EPIDEMIOLOGY

The science of epidemiology seeks to deter-mine the distribution and causes of diseasesand injury in humans, It focuses on groupsrather than individuals (3).

There are several types of epidemiologicalstudies. Each provides different levels of in-formation on environmentally related diseaseor injury. The cohort-study method is the bestway to develop information on potential toxiceffects in a population exposed to an environ-mental contaminant in food. In a cohortstudy, individuals are classified into groupsaccording to the levels of exposure, includinga control group with no previous exposure tothe suspect substance. The groups are thenfollowed over a period of time and studied fordifferences in disease incidence (3). Such epi-demiological studies have an advantage overanimal tests. There are some agents known tocause disease in humans that do not producesimilar adverse effects in animals (benzene,for example) (4). Human epidemiological data

can also be used to directly estimate humanrisk, thus eliminating the need for extrapolat-ing from animal experiments.

Epidemiological evidence has sometimesweighed heavily in the setting of tolerances oraction levels. This was the case in FDA’s deci-sions on mercury and polychlorinated biphen-yls (PCBs). However, if epidemiological dataare lacking when an environmental contami-nant is discovered in food, the usefulness offurther epidemiological studies is restrictedby inherent limitations of the science.

Beyond the obvious ethical considerations,such studies are of little use when a rapidregulatory decision must be made. At the lowlevels that environmental contaminants usu-ally occur in food, pathological effects maynot occur for many years, Most cancers, forexample, have a latency period of at least 5 to10 years, As much as 40 years may elapse be-tween the onset of exposure and the develop-ment of the disease, Thus epidemiological

Ch. V—Methods for Assessing Health Risks ● 61

methods can only be used retrospectively toevaluate the health status of individuals whohave been through this 5- to 40-year period.The findings are compared to the health his-tories of “control” individuals who presum-ably have not experienced comparable ex-posures.

Such retrospective studies are vulnerableto scientific criticism: a) it is very hard to findan adequate control group (unlike animalswhose entire lives can be in a “controlled”environment) because of the variability inhuman susceptibility and personal behaviorpatterns, b) it is impossible to define potentialsynergisms and/or inhibitions by other sub-stances to which the target group may havebeen exposed, and c) it is difficult to quantifyprevious levels of exposure (5).

Retrospective studies suffer other handi-caps, They are extremely expensive, difficult,and time-consuming to carry out, A given typeof cancer, for example, may occur in the“normal” population at a rate of 1 case foreach 100,000 people. An exposure to a toxicsubstance that increases by a factor of 100the likelihood of that cancer occurring wouldraise the incidence to 1 case for each 1,000people. To ascertain in a statistically satis-factory and compelling way that such an in-crease had occurred, one would have to docomprehensive medical studies on at least3,000 people. Assuming an average cost of$200 for each person studied (which is a lowestimate considering the costs of physicians”time, laboratory analysis, and the technologi-cal assessments necessary to show a personfree of a given tumor or disease), the expensefor such a study would be $600,000. It wouldbe necessary under most circumstances tofind more than one new case of a disease toconvince a nonstatistician and even most

statisticians. This would double or triple thecosts (5). Furthermore, demonstrating thatthe incidence had increased would still leavethe vexing problem of correlating the in-crease with the exposure to the specificsuspect substance.

Such limitations make cohort epidemiologystudies inappropriate for determining the tox-ic potential of a given contaminant. More-over, in a regulatory scheme designed to pre-vent serious illness from developing in hu-mans, the time required to generate epidemi-ological data would preclude its use in initialregulatory decisions. The prospects for devel-oping a human epidemiological method thatwould meet such regulatory demands arepresently hard to imagine.

One area, however, where human epide-miological studies have and will continue tobe useful is in the confirmation of suspectedchronic-effect data when a population thathas inadvertently been exposed to a sub-stance or a hazard (such as radiation) over along period of time can be identified. In thisinstance, carefully designed studies canbe the “clincher’” which finally providesthorough documentation of the suspected tox-icity. Such studies can confirm or refute theadequacy of regulatory actions (in retrospect)(5), Again, however, these kinds of studies areextremely expensive and consume limited re-sources which might be better spent support-ing other types of evaluations using othertechniques.

People who fail to understand these limita-tions criticize reliance on animal tests whenepidemiological information is negative orlacking. This fundamental misunderstandingmay delay health-related regulatory actionand dilute its effectiveness in preventingillness.


ANIMAL TESTS AND OTHER METHODS FORTOXICOLOGICAL TESTING

Toxicologists have developed a number oftechniques to assess the toxicity of a com-pound. Some tests can be conducted in 90days or less. These include simple tests suchas 2-hour “range finding” to determine thedose of a substance that is lethal to 50 per-cent of the animals (LD50). More complextests include 90-day continuous exposure orpaired feeding studies, and short-term testsfor mutagenesis and potential carcinogene-sis. Tests requiring more than 90 days, suchas lifetime exposure studies, are generallyconsidered long-term. In addition to the timenecessary for exposure and data-gathering,analysis of the results may take up to an addi-tional year, depending on the complexity ofthe experiment, the number of animals used,and the amount of data collected (l).

Testing methods can be categorized by‘‘endpoint” as well as duration. Some ex-periments are based on expected results suchas functional changes, birth defects, or can-cer, By the use of an appropriate experi-mental design, several such endpoints can beassessed in the same experimental period (l).Appendix C describes the range of toxicologi-cal tests available for assessing the healthrisks associated with consumption of contam-inated food. Each section describes the testsfor assessing a given endpoint, and includes adiscussion of available long- and short-termtests,

Animal tests provide valuable informationfor the setting of action levels and tolerances.The animals serve as proxies for humans incases where data are lacking on the humanhealth effects of a contaminant and experi-ments involving humans would be unethical,Because of shorter animal lifespans, the ef-fects on several generations can be studied.Furthermore, animals can be raised in con-trolled environments, thus eliminating manyof the factors that complicate human epide-miological studies, In many cases, animalexperiments are the only means by whichneeded information can be obtained. Using a

carefully selected battery of 90-day tests,considerable data can be generated on a con-taminant’s potential toxic effects including itspotential for causing mutations and cancer.

Despite widespread scientific agreementon the usefulness of animal tests, several con-troversies exist. These include the appropri-ateness of the carcinogen bioassay tech-niques, the potential of short-term tests forevaluating mutagenesis and carcinogenesis,the lack of emphasis on testing for potentialtoxic interactions, and the usefulness ofmethods for extrapolating from high-dose ani-mal test results to low doses in humans.

The Carcinogen Bioassay

The carcinogen bioassay (6) is a chronictoxicity study requiring 2 years’ exposure tothe test substance and usually a third yearfor evaluation of the results. The carcinogenbioassay is used to determine the cancer-causing potential of a compound in males andfemales of two mammalian species, usuallythe rat and mouse. The test animals are ex-posed to the test substance throughout theirentire lifespan.

Different groups are exposed to differentlevels of the substance up to the maximumtolerated dose (MTD). The MTD is the highestdose given during a chronic study that predic-tively will not alter the animals’ normallongevity from effects other than cancer. Inpractice, the MTD is considered to be thehighest dose that causes no more than a 10-percent loss in weight compared to controlanimals. Throughout the study, animals areexamined weekly for signs of toxicity. Ani-mals may be killed at prearranged times andtheir tissues examined for signs of cancer. Atthe completion of the study, all remaininganimals are killed and detailed examinationsare made of their tissues.

Animal tests for carcinogenicity have beenquestioned because the results are general-ized to humans while the animals are fed


much larger doses of the suspected substancethan would be consumed by people. However,an earlier OTA report entitled Cancer Test-ing Technology and Saccharin (7) found that“animal tests are the best current methodsfor predicting the carcinogenic effect of sub-stances in humans. All substances demon-strated to be carcinogenic in animals are re-garded as potential human carcinogens; noclear distinctions exist between those thatcause cancer in laboratory animals and thosethat cause it in humans. The empirical ev-idence overwhelmingly suppor ts thishypothesis. ”

The report also found that the rationale forfeeding high doses was sound.

The rationale for feeding large doses of asubstance in animal tests is as follows. Asthe dose of a substance that causes cancer isincreased, the number of exposed animalsthat develop cancer also increases. To con-duct a valid experiment at high dose levels,only a small number of animals (perhaps sev-eral hundred) is required. However, to con-duct a valid experiment at low dose levels, avery large number of animals is required.(The smallest incidence rate detectable with10 animals is 10 percent or 1 animal. Todetect a l-percent incidence rate, severalhundred animals would be required.) An-other important variable is the strength ofthe carcinogen. The stronger the carcinogen,the greater will be the number of animalsgetting cancer at a particular dose (7).

If data from a carcinogen bioassay areavailable at the time an environmental con-taminant is discovered in food, the resultsprovide crucial information to guide theregulatory decision. However, if no data or in-adequate data exist, a newly commissionedcarcinogen bioassay would require 2 to 3years for completion. In this case, the car-cinogen bioassay could be used to revise aninitial action level.

Short-Term Tests for Mutagenesisand potential Carcinogenesis

Chemicals are evaluated for mutagenicitynot only to detect potential carcinogens but

also to detect the serious hazard posed by mu-tagens. In the long run, chemical compoundscausing mutations and teratogenic effectsmay create a greater burden on society by in-creasing the incidence of genetic disease andbirth defects than by causing cancer (2,11).

A number of short-term tests and batteriesof tests have been developed to assay chem-icals for mutagenicity (1,2). Many of the sametests are also viewed as screens for carcino-gens on the assumption that cancer arisesfollowing damage to the genetic material ofcells (2). The ability of these tests to detectcarcinogens is currently undergoing valida-tion (8). The Environmental Protection Agen-cy (EPA) currently uses some of these tests toidentify mutagens (9, 10). Presently there is nouniversal agreement on a “best” set of tests.

Environmental contaminants in food arescreened to identify those capable of causinggenetic damage, to determine the types ofdamage they cause, and to evaluate the risksthey pose to the general population and cer-tain subpopulations.

The major difficulty in designing tests todetect mutagens and potential carcinogens isthe need to reflect the metabolic capabilitiesof man as nearly as possible while remainingeconomically realistic. Whole animal testsmore nearly reflect responses of man and aretherefore useful in estimating risk to humans.Tests using lower organisms are less directlyapplicable to humans, but they are simplerand more economical to perform. As a conse-quence they are more popular for use in mu-tagen screening than animal tests. Risk esti-mates based on tests in lower organisms havebeen proposed (12) but are not accepted assufficient evidence for setting action levelsand tolerances.

Many chemicals are not directly muta-genic, but once ingested or absorbed are con-verted into mutagenically active derivatives.Activation therefore poses important practi-cal and conceptual problems that must bedealt with in evaluating the mutagenic poten-tial of environmental agents. The practicalproblem is one of designing tests to identify


mutagens that adequately mimic the metabol-ic capabilities of an intact animal. The con-ceptual problems focus on the fact that evi-dence of an agent’s mutagenicity is not suffi-cient to evaluate the actual risk posed to anindividual or a population by exposure to theagent (2).

The ability of an agent to cause geneticdamage in an intact animal depends not onlyon the mutagenic potential of the agent or itsmetabolizes but on a number of other factorsas well. The fate of any chemical substanceentering the body is determined by a dynamicprocess involving its absorption, distribution,biochemical alteration, and excretion. Manymutagenic agents are formed by the action ofintestinal bacteria. Consequently, the route ofexposure to an agent can play a significantrole in determining its activity. Enzymaticprocesses in the body inactivate as well ascreate mutagens. Many active mutagens mayexist in the body as intermediates in themetabolic pathways that process the parentcompounds. As a result, the active mutagenmay have a short lifespan and be distributedonly in those tissues that possess high levelsof activating enzymes (2).

There are four major factors determiningthe ability of an environmental agent to pro-duce a genetically adverse effect in an orga-nism, The first is the metabolic response ofthe organism to the agent. This response willdetermine the distribution, lifespan, and fateof the mutagen in the body tissue. The secondis whether the mutagen damages the geneticmaterial in cells. The third is the response tothe genetic damage by the DNA repair ma-chinery in the cell. Fourth is the type of lesionthe agent is capable of producing in the genet-ic material.

Short-term tests for detection of muta-genesis and potential carcinogenesis fall intofour categories:

●

●

●

Q.

procedures that can be carried out di-rectly in human populations,intact animal tests,tests employing cultured mammaliancells, andtests employing micro-organisms.

Several procedures that can be carried outdirectly in exposed human populations havebeen found to be indicators of the presence ofmutagenic agents. These procedures includeanalysis of white blood cells for chromosomeaberrations (cytogenetic analysis) (13), anal-ysis of blood and/or urine of the presence ofmutagenic agents (14), analysis of semen forabnormal sperm (15), and the detection of ex-tra chromosomes in sperm (16).

Intact animal tests usually use as testanimals either mice or fruit flies [Drosophilamelanogaster). The dominant lethal test de-tects lethal damage resulting primarily fromchromosomal aberrations by treating malemice with test agents and mating them. Theuteri of the pregnant females are examinedfor fetal death and resorption. Because theendpoint is detected in the offspring of thetreated animal, mutation in the germinal cells(the sperm) is detected by this method. Thespecific locus test involves the crossing of twogenetically distinct strains of mice with visi-ble markers, such as spotted coats. Mutationsare detected by the appearance of recessivetraits in the offspring. The utility of this testisrestricted because of the number of animalsneeded. Cytogenetic screening of white bloodcells or bone marrow cells from treatedanimals detects damage to chromosomes.The micronucleus test detects chromosomebreaks in bone marrow cells of treated ani-mals (17).

The use of cultured mammalian cells formutagenesis testing is a great simplificationover the use of intact animals. At the sametime, the cells retain much of the mamma-lian metabolism and genetic characteristics.Moreover, human cell cultures can be usedwhere exposure of people to toxic agentscould not be allowed. The major weakness ofcultured mammalian cells as test systems isthat they are isolated from the metabolic ac-tivity of the intact animal (2).

Several strategies have been developed totry to approximate the metabolic capabilitiesof the intact animal in mammalian cells, Thesimplest and most widely used activation sys-tem employs homogenates of mammalian


liver. Either the liver homogenate is incu-bated with the test substance before it is ap-plied to the cell culture, or the test substanceand the homogenate are included in the testplates. A major difficulty with the liver micro-some activation system is that it cannot beprepared with reproducible enzyme levelsfrom one batch to the next. Storage condi-tions also alter the activities of the prepara-tions, Furthermore, not all the activationmechanisms present in the intact animal arerepresented in the liver homogenate. Thegreatest advantage of the liver homogenatemethod is that it preserves the simplicity of invitro testing (18).

Methods for mutagen screening have beendeveloped using many micro-organisms, in-cluding fungi, bacteria, and viruses presentin bacteria. The most widely known system isthe “Ames test” which employs speciallybred strains of Salmonella bacteria. Figure 5illustrates the steps in the Ames spot test. TheAmes plate test allows quantitative compari-sons to be made of mutagenic potential (19).Numerous other tests have been devisedwhich are in varying stages of validation (20).Some of these are described in appendix C.

The great appeal of these techniques is thatthey are rapid, simple, and inexpensive. Noanimal colony is required, and the necessarytechnical skills are those required for conven-tional bacteriology.

These tests have been criticized becausethe cell structure of bacteria is very differentfrom the cell structure of mammals. Valida-tion studies now underway are attempting toevaluate 30 mutagenicity assay systems fortheir ability to predict chemical carcinogeni-city (8).

Bacteria also lack the enzyme systems thatare the principal mammalian mutagen-acti-vating enzymes. This criticism is partiallyovercome by the use of liver homogenates de-scribed above. Finally, each test is sensitiveto one particular type of genetic damage.Therefore, the best approach is to use a bat-tery of short-term tests designed to test forthe different types of genetic damage.

Interactions of Twoor More Substances

Tests designed to detect effects from inter-actions of two or more substances are of lim-

Figure 5. —Steps in the Ames Spot Test

Suspected carcinogenon filter paper

Rat-liver Testerextract bacteria

2

MUTAGENESIS is detected in the Ames test by mixing an assay” a dose of the chemical to be tested is placed on aextract of rat liver (which supplies mammalian metabolic disk of filter paper on the tester bacteria. After 2 days mostfunctions) with tester bacteria (which cannot grow because of the his - bacteria have died for lack of histidine (2), buta mutation makes them unable to manufacture histidine, a DNA damage caused by the chemical diffusing out fromnecessary nutrient) and plating the mixture on an agar the disk has given rise to mutations, some of which resultmedium so that a thin layer of bacteria covers the medium in reversion of the his- mutation. The histidine-makingevenly, as is shown on a microscopic scale (1). In this “spot revertant bacteria proliferate, forming visible colonies (3).

SOURCE Redraw n v, II h perm[sslon Devoret R Bacterldl Tests for Polen!lal Carcinogens Scfentlffc Arner/can 241 (2} 40 August 1979

ited use for regulatory actions. This does notmean that such information is not important.It rather reflects the rudimentary nature ofour present understanding of interactions,the complexities of the tests, and the difficul-ties in interpreting the results in a fashionmeaningful to regulators.

Six different types of interactions may oc-cur when two or more chemicals interact in abiological system. The effects produced maydiffer in magnitude from those caused by anyone of the chemicals alone.

●

●

●

●

●

●

If the interaction of two substances pro-duces an effect equal to the sum of theindividual effects, the response is calledsummation or addition.If the interaction of two substances pro-duces an effect greater than would beanticipated from the sum of the individ-ual effects, the response is known as po-tentiation or synergism.If the effect is less than the sum of thetwo would predict, the response is an-tagonism.When an inert substance, having noobserved effect at a given dose, en-hances the effect of another simultane-ously administered chemically, the ef-fect is usually referred to as activation.If an inert substance decreases the ef-fect of another chemical administered si-multaneously, the effect is called inhibi-tion.Finally, there may be no interaction andeach chemical would exert its own effectindependently of the other (2 1).

Once ingested, a chemical may exert an ac-tion locally in or on the stomach or intestines.While such effects possibly could be pro-duced by environmental contaminants, theyare not likely to be observed at the low levelsnormally encountered in food. The more ser-ious concern is the systemic effects that mayoccur after absorption from the gastrointes-tinal tract.

Following absorption from the gastrointes-tinal tract, additive chemicals may producetheir effect by acting on a target organ, byacting on different target organs or systems,

or by acting differently on different organs toproduce the same effect. Most of the interac-tions are at the biochemical level and themechanisms are still being studied.

The key consideration is whether presenttesting technologies are adequate to providedata that are useful in making regulatorydecisions. Techniques have been developed tostudy the interactions of drugs, other chemi-cals, and physical agents such as radiation.Such techniques should be applicable to thestudy of the effects of environmental contam-inants. However, their scope is usually lim-ited to the study of two component interac-tions because of the large number of test ani-mals required and the difficulties in inter-preting the results. For example, a relativelysimple test involving two agents in a factorialdesign would require about 100 animals. Amore recent study (22) of the combined ef-fects of cadmium, mercury, and lead used 50to 60 animals in each of 15 different treat-ment groups. The results indicated that a par-ticular combination of the three metals couldbe synergistic, antagonistic, or additive, de-pending on the relative doses employed (table13).

Ch. V— Methods for Assessing Health Risks ● 67

The testing for interactions among toxicsubstances is further complicated by the ne-cessity to limit the number of substancesstudied, Because of the large number of ani-mals required, and the difficulties in inter-preting results, the number of substancestested are confined to those that may be mostimportant, Deciding what is ‘‘important” tostudy out of the vast number of toxic sub-stances to which we are exposed is a difficultproblem, Perhaps a reasonable approachwould be to focus on those to which peopleare known to be exposed (see chapter VII).The choice could also be based on structure-activity relationship or known mechanism ofaction, Chemical selection for testing is beingreviewed and evaluated in an ongoing OTAstudy entitled “Assessment of Technologiesfor Determining the Carcinogenic Risk fromthe Environ merit.”

The study of the interactions of two ormore substances is an area in which far morebasic research is required before meaningfulinformation can be generated for regulatorydecisions. At the present time, no satisfactorymethods exist for testing the interactions ofmore than two chemicals.

Extrapolating From High DosestO Low Doses

Quantitative risk assessment. based onmathematical models, is often proposed as analternative to the current use of safety fac-tors (chapter 111) (24,25). This approach isnow most extensively employed in assessingcarcinogenic risks (26). The technique is alsobeing investigated as a means to evaluateother types of irreversible toxic effects (23).The following discussion of mathematical ex-trapolation of risk involves only its uses in de-termining risks from chronic low-level carcin-ogenic insults.

Mathematical models can be used to esti-mate the number of extra cancers that arelikely to be caused by the presence of a car-cinogen in the food supply. Models may beused to estimate a tolerance based on calcu-lations of a “safe dose’” for human consump-tion. They can also forecast the likely change

in the number of extra cancers that would ac-company some projected increase or de-crease in the level of human exposure occur-ring either as a result of regulatory action orinaction, The technique is therefore used tocalculate benefits to be derived from re-ducing human exposure to a substance.

The typ ica l carc inogen b ioassay usesaround 100 animals at each experimentaldose. If a particular experimental dose pro-duces evidence of a l i fetime increase incancer of 1 in 10, this increase can be meas-ured using 100 animals. But if the increasedcancer risk is less than 1 in 100, this increasewill often not be detected even with a 100-animal feeding experiment. The extra humanrisk resulting from environmental exposure isusually much smaller than 1 in 100 (perhapson the order of 1 in 100,000) for any givenchemical over a lifetime exposure. It wouldnot be practical to conduct an experimentwith enough animals to directly measure thissmall an increase in risk (30).

For these reasons, carcinogen bioassaysuse (in addition to a control dose of zero)several doses at which the projected extracancer incidence may be 1 in 10 or larger.These high-dose data are then used to esti-mate the extra risk at a very low dose wherethe extra risk may be no larger than 1 in1,000,000. These problems are often referredto collectively as the “low close extrapolationproblem” (30).

A low-dose extrapolation involves thechoice of a mathematical function to modelthe dose-response relationship and the choiceof statistical procedures to apply to the math-ematical function (24-29). The choice of math-ematical function is crucial to the outcome ofa low-dose risk estimate. If the assumed rela-tionship between tumor occurrence and dosedoes not apply in the low-dose regions towhich the extrapolation is being made, a seri-ous overestimate or underestimate of the“safe dose’” may result. Even though differentdose-response models may agree in the ob-servable response range, they could differ bymany orders of magnitude at low-dose levels.Furthermore, there is no experimental way to


predict the shape of the dose-response curveat very low doses,

Because of the possible disparity of dose-response functions when extrapolated to lowdoses, the dose-response function must be se-lected not only on the basis of how well it canbe made to fit experimental data but also onthe basis of known (or at least plausible) in-formation on the biological mechanismsthrough which a chemical induces or pro-motes cancer. This is a major source of un-certainty in extrapolation procedures, sincethe exact biological mechanisms throughwhich carcinogenesis may occur are un-known (30).

Several theories on the nature of the proc-ess by which cancer develops serve as thebases for different low-dose extrapolationmodels. Some “linear” models project thatthe relationship between the dose receivedand the cancer risk is a straight line at lowdoses and that there is no threshold. Othermodels level off at low doses and thereforepredict that a threshold dose exists belowwhich the agent has no carcinogenic effect(24-30).

Newer models take into account the effectsof metabolic activation and detoxificationupon carcinogenic dose response (26), These“kinetic” models encompass free toxic sub-stances, metabolizes, deactivators, and thepossible interactions of these substances,Only a “steady-state” situation is studied inthat the variation over time of the concentra-t ions of these agents is not considered. Ifdeactivation of the carcinogen is 100-percentefficient, the model predicts a threshold dosebelow which there is no carcinogenic r isk.However, in a naturally occurring process, itis likely that deactivation would be less than100-percent effective. A number of modifica-tions to the model allow for imperfect deac-tivation. These situations rule out a thresholdand would lead directly to a model for whichcarcinogenic response varies linearly withdose at low dose,

The mechanisms through which most car-cinogens produce cancer are not sufficientlyunderstood so that the shape of the dose-

response curve can be predicted with cer-tainty, As pointed out earlier, experiments ofsufficient size to permit direct experimentalinvestigations of the dose-response curve atlow dose cannot be conducted. There areplausible arguments that the dose-responsecurve is linear at low dose for many carcino-gens, On the other hand, no mechanism thatwould lead to a more cautious dose-responserelationship has been seriously proposed ex-cept for some dose-response relations inradiation-induced carcinogenesis.

In view of these uncertainties, it wouldseem reasonable to base estimates of addedcancer risk on a mathematical model that en-compasses low-dose linearity until the mecha-nisms through which the carcinogen operatesare understood sufficiently to conclusivelyrule out low-dose linearity,

Caution must be used in interpreting theresults of low-dose extrapolations, partic-ularly when they are applied to humans,Table 14 demonstrates that different modelsapplied to the same data sets yield differingestimates of virtually safe doses. Virtuallysafe doses based on the multistage model areidentical with those based on the one-hitmodel when there is only one experimentaldose. This is illustrated with Data Sets I and

Table 14.—Virtually Safe Doses ComputedFrom Three Different Data Sets and

Three Different Models

Virtually safe doses (lowerDose- 95-percent confidence bounds

Data response for dose) in ppb correspondingset model to extra risk of 1/1 ,000,000la ... Probit 14.2

SOURCE Adapted from K S Crurnp and M D Masterman Assessment of Carc Inoqenlc R(sk From PC Bs In Food OTA Working Paper 1979


II. The multistage model also yields virtuallythe same value as the one-hit model for DataSet III. However, as explained in appendix D,the multistage model can yield higher valuesfor the virtually safe dose than the one-hitmodel whenever the data exhibit upward cur-vature and are inconsistent with the one-hitmodel (30).

Caution also must be exercised when com-paring calculations of extra lifetime risks.Table 15 summarizes the results of an FDArisk assessment of PCBs (31) with a similarrisk assessment commissioned by OTA (30).Differences can arise in such calculationswhen different methods are used to extrapo-late from animals to man. The extrapolationcan be performed on the basis of milligramsper kilogram of body weight per day, partsper million of substance in the diet, milli-grams of substance per square centimeter ofbody surface area per day, or milligrams ofsubstance per kilogram of body weight perlifetime.

Table 15. — Extra Lifetime Risks of CancerAssociated With Consumption of PCBs in Food

Extra lifetime Upper limit of newDose (g/day) risk/100,000 cancers/year

FDA’9 2’ 4 4 211 4 9 ’ 72 34201 [ 9 8 47OTAb

3 3(! 013 4035 11

127f 5‘]BaCI,j r,n NC I birldssd 7 io j tc. td mallqn,+n Ifs fr,! mdlf,: .in{j f[malfs: Ba+fI! ,rl K !mh rrju qh ~ 1 q 751 S(I jff Y and hppa tor~lrular care {ncm]ds‘ Ba\Ifj c, n h [q h f~ f [ or SII MC rs ?f!lh ~If rc c r] t I If) of f I sh ~ per I(,S con tam I n ,itr?{j

* I t h PC. F3< I f trd(. rar> r c I,> I dt I I ~ h~, t j ,it 1 2 r r ~ [~ [ r.)[j Ba>e(j , ,r, FDA T r,ld I D IPt St (1 j Y 1 q ?(J

~’ Ba\P j ‘n F E! A T 01+ I D I( t 5 I hIj ~ 1 Q 75fBa~{, cj ,, ~1 ~+(.r~qf In tak ~ c f rf’flrlf ‘r n., II m IrTq mr)re than 24 I hs t par Lak cMlchiQan fish

1.

2,

3.

CHAPTER V

SRI International. “Assessment of h4ethodsfor Regulatin~ ‘Unavoidable” Contaminants inthe E’ood Supply, ” OTA Working Paper, 1978.Legator, hf. S. “Short ‘1’erm P r o c e d u r e s f o rDetermining Nlut:ig~;niclCar cinogenic Activ-ist},’” OTA ~lrorking Paper, 1979.Llachfahon, B., and T. F. Pugh. Epidemiolog}rPrinciples (]nfl Mcthmls, Little, Brow-n and

Differences also arise when varying ani-mal data serve as the basis for the risk as-sessment, And, of course, the choice of themodel also affects the final outcome. In thiscase, both extrapolations were based on mil-ligrams of PCB per kilogram of body weightper day. Both FDA and OTA applied a linearmodel: however, different assumptions weremade on the amount of PCBs that would be in-gested and on the size of the exposed popula-tion. FDA’s risk assessment therefore appliesto the 15 percent of the total U.S. populationexpected to consume the fish species knownto be most highly contaminated with PCBs.The OTA calculations are based on the aver-age daily intake of PCBs based on FDA's total

est imates of the PCB intakeconsumed more than 24 lbs ofsport fish per year.

Al though many differentmodels exist and can be used for low-doserisk assessment, it appears prudent at thistime to use a model that does not arbitrarilyrule out low-dose linearity. Models are stillbeing developed and when” appropriate couldbe used to guide regulatory decisions. Dose-response models for low-dose risk assessmentprovide a useful technique for assessing thepossible added risk attributable to environ-mental contaminants. Such models also mightwell be used in place of safety factors infuture procedures adopted for food contami-nant regulations. The major impediment tothe widespread use of the models in environ-mental food contamination considerations isthe lack of data on which to perform the anal-ysis.

REFERENCES

4.

5.

Company, Boston, Mass., 1970.Ray, V. “Are Benzene Effects Limited to theChromosomal Level?, ” Banbury Report 1—Assess ing Chemical A4utqgens: The Risk t~~Humans, Cold Spring Harbor Laboratory,1979.]ackson, R, S. “The Limitations of ApplyingIluman Epidemiology to the Regulator Pro-


6.

7,

80

9.

10.

11,

12.

13.

14.

15.

16.

17.

18.

cess for Toxic Substances, Particularly Car-cinogens,” OTA Working Paper, 1979.Sontag, J. M., N. P. Page, and U. Saffiotti.Guide~ines for Carcinogen Bioassay in Sma~lRodents, DHEW Publication No. (NIH) 76-801,Government Printing Office, Washington,D. C., 1976.Office of Technology Assessment. CancerTesting Technology and Saccharin, Govern-ment Printing Office, stock no. 0 5 2 - 0 0 3 -00471-2, Washington, D. C., 1977.U.S. Department of Health, Education, andWelfare. National Toxicology Program An-nual Plan Fiscal Year 1979,U.S. Environmental Protection Agency. FIFRATesting Guidelines, 1978.U.S. Environmental Protection Agency. TSCATest Standards, 1979.Neel, J. V. “Mutation and Disease in Hu-mans, ” Banbury Report l—Assessing Chemi-cal Mutagens: The Risk to Humans, ColdSpring Harbor Laboratory, 1979.Committee 17, Council of the EnvironmentalMutagen Society, “Environmental MutagenHazards,” Science 187:503, 1975.Kilian, D. J., and D. Picciano. “CytogeneticSurveillance of Industrial Populations, ”Chemical Mutagens, Principles and Methodsfor Their Detect ion, Vol. IV, A. Hollaender(cd.), Plenum Press, New York, 1976.Legator, M, S., L. Troung, and T. H, Connor.“Analysis of Body Fluid Including Alyklationof Macromolecules for Detection of Muta-genic Agents, ” Chemical Mutagens, Princi-ples and Methods for Their Detection, Vol. V,A, Hollaender and J, DeSerres [eds.), PlenumPress, New York, 1978.Heddle, J. A., and W. R. Bruce. “On the Use ofMultiple Assays for Mutagenicity, Especiallythe Micronucleus, Salmonella, and SpermAbnormality, Assays in Progress, ” GeneticToxicology, P. Scott, B. A, Bridges, and F. H.Sobels (eds.), 1977.Kapp, R. A. Mutation Research, in press,1979.Kilbey, B. J., M. Legator, W. Nichols, and C.Ramel. Handbook of Mutagenicity Test Pro-cedures. Elsevier, Amsterdam, 1977.Ames, B. N., W. E. Durston, E. Yamasaki, andF. O. Lee. “Carcinogens and Mutagens: ASimple Test Combining Liver Homogenatesfor Activation and Bacteria for Detection, ”Proceedings of the National Academy of

19.

20.

21.

22.

23.24.

25.

26.

27.

28.

29.

30.

31.

Sciences 70:2281, 1973.Ames, B. N. “A Bacterial System for Detect-ing Mutagens and Carcinogens, ” MutagenicEffects of Environmental Contaminants, H. E.Sutton and V. I. Harris (eds. ), AcademicPress, New York, 1972.Devoret, R. “Bacterial Tests for PotentialCarcinogens, ” Scientific American 241(2):40,1979.Fingl, E., and D. M. Woodbury. “General Prin-ciples, ” The Ph~lrmacological Basis of Thera-peutics, L. S. Goodman and A. Gilman (eds. ),5th edition, MacMillan Publishing Co., Inc.,New York, 1975.Schubert, J., E. J. Riley, and S. A. Tyler.“Combined Effects in Toxicology—A RapidSystematic Testing Procedure: Cadmium,Mercury, and Lead, ” Journal of Toxicologyand Environmental Health 4:763, 1978.Van Ryzin, J., personal communication, 1979.Interagency Regulatory Liaison Group, WorkGroup on Risk Assessment. “The ScientificBases for Identification of Potential Car-cinogens and Estimation of Risks,”’ JournalNational Cancer Institute 63:241, 1979.Scientific Committee of the Food Safety Coun-cil. “A Proposed System for Food Safety As-sessment ,“ Food and Cosmet. Toxicol. 16,Supplement 2, 1978.Cornfield, J. “Carcinogenic Risk Assess-merit, ” Science 198:693, 1977.Crump, K. S., D. G. Heel, C. H. Langley, and R.Pete. “Fundamental Carcinogenic Processesand Their Implications for Low-Dose Risk As-sessment, ” Cancer Research 36: 2973, 1976.Guess, H. A., K. S. Crump, and R. Pete. “Un-certainty Estimates for Low-Dose-Rate Ex-t r a p o l a t i o n s o f A n i m a l CarcinogenicityData, ” Cancer Research 37: 3475, 1977,Rai, K., and J. Van Ryzin. “Risk Assessment ofToxic Environmental Substances Using aGeneralized Multi-Hit Dose Response Mod-el, ” Energy and Hecdth, N. Breslow and A.Whittemore (eds, ) SIAM Press, Philadelphia,Pa., p. 94, 1979.Crump, K. S., and M. D. Masterman. “Assess-ment of Carcinogenic Risk from PCBS inFood, ” OTA Working Paper, 1979.U.S. Food and Drug Administration. An As-sessment of Risk Associated with the HumanConsumption of Some Species of Fish Con-t amina ted Wi th Polych~orinated Biphenyls(PCB’S), 1979.

Chapter VI

Methods of Estimatingand Applying Costs to

Regulatory Decisionmaking

*

54-515 0 - 79 - 6

Chapter VI

Methods of Estimating and ApplyingCosts to Regulatory Decisionmaking

In the previous chapter, various testing methods for assessing health riskswere evaluated for their usefulness in regulating environmental contaminants—i.e., setting an action level or tolerance. The primary issue involved in assessinghealth risks is not whether the potential risks from an environmental contaminantshould be evaluated for purposes of regulation but rather what testing methodsare most appropriate for assessing potential risks?

The situation is reversed, however, when the associated costs of an actionlevel or tolerance for an environmental contaminant are assessed. The primaryissue is whether the costs should be taken into account in the setting of atolerance or action level. This is a policy issue which is addressed in chapter IX,“Congressional Captions. ” In this chapter, the various approaches and techniquesfor estimating the costs and benefits of a proposed tolerance or action level areassessed, along with two common methods for applying cost and human risk datain the regulation of environmental contaminants. A more detailed discussion ofthe approaches and techniques for estimating the costs and benefits is provided inappendix E.

Two methods, the cost-effectiveness method and the cost-benefit method, areanalyzed for their strengths and weaknesses as regulatory decision-assistingtools. This analysis provides a basis for the discussion in chapter IX on the role ofeconomics in regulating environmental contaminants in food.

Cost-effectiveness is a regulatory decision-assisting tool that compares the estimated netcosts of the proposed regulation in dollarterms with estimated reduction of human riskexpressed in scientific terms. In the regula-tion of environmental contaminants, the cost-effectiveness method would compare the esti-mated change in costs of a proposed actionlevel or tolerance for a contaminant in foodwith the associated reduction in humanhealth risk—i.e., the benefits of the proposedtolerance or action level. The comparison ofnet costs and risk reduction is performed forseveral alternative levels in order to select atolerance or action level. The principal costof regulatory action is the cost of food held offthe market because it exceeds the proposedaction level or tolerance. There would also be

a reduction in medical costs as a result in thereduction of human risk. These costs wouldneed to be subtracted from the cost of re-stricted food.

The cost in lost food would also have asso-ciated distributional effects which wouldneed to be identified to fully weigh the impactof a regulation. These effects would includeidentifying the consequences for producers,change in the cost of food to consumers, andindirect impacts on such businesses as foodprocessors, animal feed manufacturers, orbait and tackle shops,

The cost of food losses and related effectswould be compared with the estimated reduc-tion of human risk for the contaminant at theproposed action level or tolerance. The hu-

73


man risk data could include such informationas the reduction in the estimated number ofnew cancer cases, the potential mutagenicand teratogenic effects, and other toxicologi-cal effects described in chapters II and V.The toxicological effects vary by substanceand are stated in scientific terms. The esti-mated net cost and reduced human risk atvarious proposed action levels or tolerancesare estimated and compared. The cost-effec-tiveness method then requires the regulatoryagency to make its own judgment on what isthe proper balance of net cost and reducedhuman risk when setting a tolerance or actionlevel.

The Food and Drug Administration (FDA),for example, followed a similar procedure inits decision to reduce the tolerance for poly-chlorinated biphenyls (PCBs). In deciding on a2 parts per million (ppm) tolerance for fish,FDA compared for 5 ppm, 2 ppm, and 1 ppmin fish the estimated reduction of human riskwith the estimated increase in the amountand cost of food expected to be taken off themarket. Table 16 shows the human risk fromcancer and cost data on commercial fish foreach of the three levels and the net changesin risk and cost for moving to a 2-ppm and 1-ppm tolerance. It is these net changes thatare evaluated in the cost-effectiveness meth-od, Though FDA reviewed other availabletoxicological data, it was FDA’s judgmentthat the reduction in risk (13.3 cancers) for al-ppm tolerance did not offset the increasedcost ($10, 3 million) and that a proper balancefor the net changes

Tolerance

5 parts per million (ppm)

2 p p m ,

1 ppm

was established at 2 ppm.

Thus, FDA has placed an implicit dollar valueon life. A more thorough application of thecost-effectiveness method would attempt toidentify all the net changes in the benefits(the reduction of all known risks) and allknown costs,

Judgment is an important aspect to thecost-effectiveness method. It allows the deci-sionmaking body, in this case FDA, flexibilityin interpreting health data and evaluating theimplications of that data on the setting of atolerance or action level. This flexibility canbe particularly important in the assessmentof insufficient or variable toxicological data.While an agency’s interpretation of toxicolog-ical data might differ from others and gener-ate disagreement about the final decision, thecost-effectiveness method does allow the com-plexity and shades of gray to be factored intothe decisionmaking process,

The health data used in the cost-effective-ness calculations are generated by the toxico-logical and epidemiological procedures re-viewed in the previous chapter. There arealso various approaches and techniques usedto estimate the costs of food likely to be takenoff or restricted from the market as a resultof a proposed tolerance or action level. Thecost estimates will vary depending on the ap-proaches or techniques employed. All theseapproaches require chemical analysis of astatistically significant number of food prod-ucts known or thought to be contaminated,With such data, the amount of contaminatedfood that exceeds the proposed action level ortolerance can be estimated,

Table 16.— Impact of PCB Tolerance

Lifetime risk of cancer for Reductionheavy consumers of fish in human risk

9.8 per 100,000 or 46.8 new cancersper year

Starting level7.2 per 100,000 or 343 new cancersper year

125 new cancers4.4 per 100,000 or 21 new cancers peryear

13,3 new cancers

One year cost (1974dollars) of commercial

fish (land value)

$ 06 million

$ 57 million

$16 million

Increasingcosts

Starting level

$51 million

$10.3 million

SOURCE Federa/ Reg/ster VOI 44 No 127 June 29 1979 p 38333

Ch. VI—Methods of Estimating and Apply/rig Costs to Regulatory Decisionmaking ● 75

FDA multiplies the estimated amount offood removed from commerce by the appro-priate price (production, wholesale, retail,etc. ) per unit (pound, bushel, ton, or animal,etc.). This method of estimating food costs isnot the most accurate way of estimating costsand does not attempt to measure the indirectcosts or distributional effects that result froman action level or tolerance.

There are two other approaches for esti-mating the costs from the reduction of thefood supply: the alternative-cost approachand the opportunity-cost approach. The alter-native-cost approach estimates the cost of thenext best alternative method or methods, ifavailable, for replacing food removed fromthe market.

For example, assume that 164,000 lbs oflake trout in Lake Superior were restrictedfrom commerce because of the 2-ppm toler-ance for PCBs. The alternative-cost approachwould require cost estimates of each avail-able method as well as various combinationsof methods for producing (and thus replacing)the entire 164,000 lbs of lake trout, theamount condemned by the tolerance. In es-tablishing the replacement cost , this ap-proach does not consider price or supplyshifts in the restricted food product nor shiftsin the supply of other food products. Suchshifts are likely to occur when large amountsof food in relation to the total supply aretaken off the market. Consequently, this ap-proach can lead in some instances to an over-statement of costs.

The opportunity-cost approach does in-clude shifts in supply. But what information isincluded in the analysis depends on whichtechniques are used. The budgeting tech-nique only includes shifts in supply for thecontaminated commodity. This techniqueuses data on the production of the contami-nated product to calculate the costs of pro-ducing and consequently replacing a particu-lar amount of that food product that has beenrestricted from commerce. It is assumed thatother commodities will be produced in the

same amounts and at the same price that pre-vailed before the establishment of the toler-ance or action level.

The modeling technique, however, includesnot only supply shifts but also resulting priceshifts in both contaminated food and otherfood products. With this technique, mathe-matical models of the relevant portion of theeconomy are employed to trace shifts in sup-ply, changes in the amount and price of var-ious commodities, and other factors affectedby the restriction on the use of a contam-inated product.

These models can, if properly programed,project changes in the production location ofparticular crops as well as changes in the useof various production factors in each regionof the country. Thus modeling can more real-istically describe likely reactions in food pro-duction to the change created by a tolerancelevel. The result is a better estimate of the po-tential cost of lost or restricted food.

In attempting to project replacement costs,the opportunity-cost approach (using eitherbudgeting or modeling techniques) more ac-curately estimates the true costs of a toler-ance or action level then either the alterna-tive-cost or the FDA approach.

The distributional effects of a tolerance oraction level can be incorporated in the cost-effectiveness method. But the opportunity-cost approach can better generate these dis-tributional effects. With either method theseeffects are most easily identified when thedata are initially being collected. The infor-mation is important in determining who willbear the brunt of the costs, both directly andindirectly. The distributional effects flowingfrom human health risks can also be identi-fied. The available risk data can be assem-bled to determine who bears most of the po-tential and actual risks. Thus the distribu-tional effects of the net cost and the reductionof risk from a tolerance or action level can becompared (l).


COST-BENEFITThe cost-benefit method compares the esti-

mated dollar costs of a proposed regulationwith the estimated dollar benefits of reducinghuman health risks. The cost-benefit methodof setting tolerance or action levels also re-quires the evaluation of several alternativelevels. The differences in benefits and costsas one moves from one proposed regulatorylevel to the next are then compared. The tol-erance or action level would be lowered untilthe costs (impact from food condemned) aregreater than the benefits (reduction of humanrisks). When this level is reached, the eco-nomically efficient tolerance or action levelwill be the next higher level at which the ben-efits exceed the costs.

By representing the costs and benefits fora proposed action level or tolerance in dol-lars, cost-benefit analysis attempts to makethe two sides of the ratio more comparable.This method recognizes, as does cost-effec-tiveness, that any proposed action level ortolerance will incur costs in terms of foodtaken off the market and benefits reflected inthe reduction of either actual or potentialhealth risks. As the level is lowered, costs in-crease and human health risks decrease.

Techniques of estimating costs are similarin the cost-benefit and cost-effectivenessmethods with the exception that the reducedmedical costs are included with the benefits.There are some ways of valuing benefits,though, which are unique to the cost-benefitmethod.

Though both methods rely on the availabletoxicological data, cost-benefit requires theconversion of all data into dollar values. Thiseconomic conversion is best accomplishedwhen the data are expressed as the numberof premature deaths per year avoided, thenumber of person-days lost to illness avoided,or the probability that some percentage of theexposed population would die prematurely orwould lose a specified number of days fromnormal activity due to illness.

Two approaches are available for convert-ing the risks into dollars: forgone earningsand willingness to pay. In the forgone-earn-ings approach the analyst places an expliciteconomic value on life in attempting to esti-mate the productivity lost as a result of illnessor premature death caused by a contaminantin food. This approach, however, does not in-clude associated health costs such as medicalexpenses incurred from illness. Includingsuch costs is required in order to more ac-curately represent the total reduction ofhealth costs—i.e., the benefits. The willing-ness-to-pay approach allows people affectedby the regulation (rather than the analyst) toestimate how much they would be willing topay to avoid a risk.

The forgone-earnings approach attempts toestimate the lost earnings of those individualswho are estimated to become ill or die prema-turely because of the contaminated food (l).The estimated dollar value of the benefit willbe determined by various techniques chosenor assumptions made by the analyst in con-verting the risk data to a dollar value. Dis-counting and earnings are two areas in whichthe analyst can influence the value of thebenefit.

Since many of the benefits of a toleranceare not likely to be realized until 10, 20, or 30years later, discounting is used to convertfuture dollar benefits into present dollars.Discounting is also used for estimating futurecosts. Discounting is required even if the fu-ture dollars are adjusted for the rate of infla-tion because a dollar spent now can be in-vested productively to yield a larger numberof real dollars—i.e., inflation adjusted—inthe future, For example, $100 invested at 5-percent interest becomes $105 in one year.Discounting is the reverse: $105 next yearhas a present value of $100 when the dis-count rate is 5 percent. This means that$10,000 worth of benefits that will occur 10years from now would actually have a pres-ent value of $6,135 if a 7-percent discountrate is used,

Ch. VI—Methods of Estimating and Applying Costs to Regulatory Decisionmaking ● 77

While most economists agree that futurecosts or benefits need to be discounted, theydo not agree on the value of the discount rate.The rate can vary from I percent to as highas 15 percent (z). The rate, however, has toremain constant for both the costs and thebenefits. Obviously, the value of the rate willaffect the estimate of benefits or costs, Thelower the discount rate, the greater the dollarestimate in present dollars. Consequently,more weight would be given to the benefitsand costs accrued in the future.

The estimate of the benefits also dependson whether gross or net earnings are used.Gross earning estimates include an individ-ual’s or a group’s total wages or salaries. Netearnings consist of total wages or salariesminus the individual’s or group’s consump-tion. Obviously the gross earnings estimate

will be greater than the net earnings esti-mate.

The willingness-to-pay approach is concep-tually a more correct approach in that it asksthe individual to place a dollar value on thereduction of associated risks from an environ-mental contaminant. This value is then usedfor placing a value on life itself. While con-ceptually correct, this approach does havesome inherent problems, such as: 1) the capa-bility of the questioned person to accuratelyunderstand the ramifications of the risk and2) the individual’s economic position. For ex-ample, an economically disadvantaged per-son might place a small value on risk not be-cause the person feels the risk is of little or noconcern but because that person cannot af-ford an increase in food prices. This ap-proach is affected by the assumptions madeby the public being surveyed.

APPLICATION OF METHODS FOR REGULATION

The use of the cost-benefit and cost-effec-tiveness methods for regulatory purposes isaffected by the following factors: 1) FDA’s in-terpretation of the Food, Drug, and CosmeticAct, 2) the approaches and techniques for es-timating the economic value of the costs andbenefits, and 3) the inherent difference be-tween the two methods.

As noted earlier, FDA interprets the Act asrequiring it to weigh only the impact of thecost and the amount of food removed fromcommerce in the setting of the proposed tol-erance or action level. FDA’s approach forestimating this cost can be applied in eitherthe cost-effectiveness or the cost-benefitmethod, but both of these methods can rely onalternative-cost or opportunity-cost tech-niques which more accurately estimate thecost incurred from condemned food. Theseother techniques also more accurately esti-mate the tolerance or action level’s impact onavailability y of food than the approach used byFDA.

Whichever technique is employed to esti-mate the costs, adequate information is

needed on the amount of food likely to be con-taminated. Such information was availablefor PCBs in fish, but it is not available forPCBs in milk, poultry, and eggs. This is be-cause contamination of milk, poultry, andeggs is likely to occur as a result of industrialaccidents. Such accidents are sporadic andtherefore difficult to predict (e.g., the July1979 PCB contamination of poultry and eggsin Idaho) (3). Consequently, the estimates forcosts incurred because of food removed fromcommerce cannot be determined for suchcontamination incidents, and thus neithermethod can be employed. Both methods couldbe used to set a tolerance or action level formercury and kepone in fish.

The approach and techniques used by thecost-effectiveness methods for generatingnecessary data can take considerable re-sources and time, 2 months to over a year togather the data just for the costs alone. Theamount of time needed depends on the ap-proaches and techniques used. The more ac-curate the information being generated, themore time and resources are required.


FDA, however, often has to make an initialdecision in the form of an action level in 2months or less for a newly identified environ-mental contaminant. A sufficient amount oftime is usually available for utilization of thevarious approaches or techniques if a follow-up decision is involved. For- example, 6 yearsexpired from the time an initial PCB actionlevel was proposed until a final tolerance forPCB was proposed this year. Either method ismore likely to be used in setting a formaltolerance than an initial action level for anenvironmental contaminant.

The substantive difference between thecost-benefit method and the cost-effective-ness method is that the cost-benefit methodplaces an explicit value on life by convertingthe health data to dollars while the cost-effec-tiveness method places an implicit value onlife by weighing the health data in its scien-tific form with the costs. As a result, a signifi-cant amount of judgment is exercised by theanalyst using the cost-benefit method whenselecting the different approaches and tech-niques for estimating the benefits. As dis-cussed earlier, the selection of these ap-proaches and techniques has a strong bear-ing on the outcome of the ratio and conse-quently the tolerance established by thismethod.

The cost-effectiveness method places agreater judgment burden on the agency andless on the analyst. While the analyst does af-fect the outcome, judgment is primarily exer-cised by the agency in weighing the net costwith the reduction in human risk (benefits).The agency exercises less judgment in the

1.

2.

CHAPTER VIEpp, Donald J. “Economic Analysis of Alter-native Action Levels in the Regulation of Envi-ronmental Contaminants of Food, ” OTAWorking Paper, April 197’9.Weinstein, Milton C., and William B. Stason.“Foundations of Cost-Effectiveness Analysisfor Health and Medical Practices, ” NewEngland Journal of A4edicine, Mar. 31, 1977.

cost-benefit method, which only requires acomparison of the numbers for each side ofthe ratio to establish the appropriate toler-ance.

The cost-effectiveness method has the po-tential to reveal more of an agency’s thinkingin the decision than cost-benefit does. Thiswas demonstrated with the earlier discussionof FDA’s setting of a 2-ppm tolerance forPCBs in fish. In addition, because it recog-nizes the uncertainties inherent in the esti-mates of the health risks, the cost-effective-ness method allows FDA the flexibility to ad-just the weight given to the benefits or thecosts in its decision. For these reasons, thecost-effectiveness method is the more appro-priate method at this time for weighing thecosts in the setting of a tolerance.

Neither method can evaluate all the infor-mation required by FDA in setting a toler-ance. For example, FDA requires for enforce-ment purposes analytical methodologies thatcan detect, measure, and confirm the identityof the contaminant at the level being pro-posed in food. This means that the tolerancecannot be set at a level below the availableanalytical capabilities for detecting the con-taminant in food. While the analytical capa-bility is an important factor in setting a toler-ance, it cannot be evaluated within the deci-sionmaking framework by either method.Consequently, these two methods should beviewed as decision-assisting aids that allowthe regulator the means to weigh many of therelevant costs and benefits in the setting of atolerance.

REFERENCES3. Foreman, Carol Tucker, Assistant Secretary

for Food and Consumer Services, U.S. Depart-ment of Agriculture. Testimony before theSubcommittee on Oversight and Investiga-tions, House Interstate and Foreign Com-merce Committee, Sept. 28, 1979.

\ Chapter VII

Monitoring Strategies

Chapter Vll

Monitoring Strategies

Monitoring involves the systematic collection and chemical analysis of foodsamples or other samples from the environment. The aim is to protect consumersby determining short- and long-term trends in the levels of various chemicals infood and the environment.

STRATEGIES

Monitoringachieve eitheridentify food

strategies can be shaped toof two objectives. The first is tolots that violate established

tolerances and action levels. The second is toidentify new environmental contaminants asthey enter the human food chain.

The first objective is met by regulatorymonitoring: the second through investigatorymonitoring. Each of these strategies could becomplemented by specimen banking. Neitheris incompatible or mutually exclusive.

Regulatory Monitoring

The Federal agencies responsible for limit-ing consumer exposure to contaminated foodnow conduct regulatory monitoring. Foodsamples are collected and analyzed for envi-ronmental contaminants for which actionlevels and tolerances have been established.Based on available information or agree-ments with States, not all samples are ana-lyzed for all regulated substances. Regulatorymonitoring employs standardized, acceptedanalytica1 techniques. Because the proce-dures are standard and can be verified byother laboratories, they generate data thatcan be presented in courts of law with littleprobability of being successfully contested.

Chapter III reviewed Federal monitoringprograms and chapter IV reviewed Statemonitoring programs. It is unlikely that thesemonitoring programs will detect new envi-ronmental contaminants, since both are re-stricted to searching for regulated chemicals.

Therefore, investigatory monitoring ap-proaches are vital.

Investigatory Monitoring

Investigatory monitoring attempts to detectunregulated chemicals as they enter the foodchain. This strategy involves the collectionand analysis of samples which may or maynot be foods, The analytical techniques em-ployed for the detection of unregulated chem-icals may or may not be accepted as standardmethods comparable to those used for regula-tory monitoring.

Analytical methods for investigatory moni-toring include broad-spectrum determina-tions that may sacrifice some quantitative in-formation (i. e., exactly how much of a givensubstance is present in a sample) for morequalitative information (i. e., better assess-ment of what or how many foreign sub-stances are in the sample). These analyticalmethods are not necessarily designed for usein litigation, They are designed primarily toindicate the presence of a potentially hazard-ous substance. If one is found, an acceptedanalytical method to detect the substancewould have to be developed—a method com-patible with instrumentation existing in regu-latory-monitoring laboratories.

Investigatory monitoring includes two dis-crete types of monitoring: monitoring for sus-pected environmental contaminants, andmonitoring for uncharacterized environmen-tal contaminants. Each of these (as well as

regulatory monitoring) can be complementedby specimen banking.

Monitoring for Suspected EnvironmentalContaminants

Some chemicals that are not regulated byaction levels or tolerances are suspected tobe dangerous to humans if consumed in foods.This group includes chemicals that may bepresent in food because of their use, toxicity,production volume, and persistence. Exam-ples of these chemicals can be found on theEnvironmental Protection Agency (EPA)/Natural Resources Defense Council (NRDC)priority pollutant list established in June1977. These substances may be called “sus-pected” or “potential” environmental con-taminants.

Monitoring for suspected environmentalcontaminants involves a different strategythan the one used in monitoring for regulatedcontaminants. Under the latter strategy,foods are analyzed for compounds with speci-fied action levels and tolerances to provide in-formation for regulatory enforcement. This isnot required for investigator y moni tor ing .Thus, the monitoring program for suspectedenvironmental contaminants is generally notas intensive (see chapter 111), Furthermore,the analytical methods for detecting sus-pected environmental contaminants may notbe as prescribed as those for regulated con-taminants.

Suspected environmental contaminantscould be identified by surveying the universeof industrial chemicals and ranking them ac-cording to their potential for entering the foodsupply in toxic amounts (l). Such an ap-proach has been recommended to FDA by aninternal study group established by formerCommissioner Donald Kennedy (7),

This method has been employed by theFood and Drug Administration (FDA) to devel-op a list of chemical contaminants in food,The criteria used in selecting the chemicalsincluded: occurrence in food, volume of pro-duction, associated impurities or byproducts,predicted environmental stability, pattern of

use, oil/water partition coefficients, bioac-cumulation potential, known toxicity, andmeans of disposal. The FDA list of chemicalcontaminants in foods is shown in table 17.

This approach is l imited by availableanalytical methods. Most of the chemicalsrecognized as food contaminants are thosethat are relatively easy to detect by gaschromatography or atomic absorption spec-trometry. Chemicals that cannot be easilydetected by these methods may, of course, re-main unidentified and unrecognized as foodcontaminants (l).

The factors used to identify a chemical’spotential for entering the food supply arebased on knowledge of the properties and en-vironmental behavior of other chemicalsalready known to be in food. This knowledge,in turn, is based on our information about theextent of contamination information thatdepends on our analytical capabilities. Thus,there is an inherent tendency to identify aspotential food contaminants those chemicalsthat are similar to chemicals already iden-tified in food. Such a tendency can only be off-set by the use of good scientific judgment orthe development of new data. This bias illus-trates a general weakness in all systems forsetting priorities: chemicals on which there isno information will automatically be givenlow priority unless some room is left for large-ly intuitive judgments (1). The scientific cri-teria and methods used in determining whatpriority various toxic substances receive inmonitoring programs are discussed in moredetail in appendix F.

Although these exercises in setting prior-ities suffer from many limitations (includinglack of data, poor choices of criteria on whichto set ranks, deficiencies in the scoring andranking systems, and deficiencies in scien-tific judgment), they still can serve a valuablefunction in guiding monitoring systems. Set-ting priorities is a prescreening exercise inwhich a compromise is struck between the ef-fort expended in preparing a priority list andthe effort that would be wasted in identifyingand quantifying all the chemicals present in asample,

Ch VII— Monitoring Strategies ● 8 3

Table 17. —Chemical Contaminants in Foodsa

up to 1up to 1 1up to 1up to 1 )01-100.1-10

0.02-008002-0.04

1.4

tr-6 7

03-2.0t r-O. 1

0.2-0.8unknownunknown 1

unknownunknownunknownunknown }

0.03-1620.03-1620.03 -16.2

Niagara River. N Y


Table 17.—Chemical Contaminants in Foodsa—(cont.)

Chemical contaminant Range in ppm Locations

White Lake, Mich

Ohio River, Ohio

Pine River. Mich

Houston ChannelHouston Channel

By applying appropriate criteria to theuniverse of industrial chemicals, it may bepossible to detect potential environmentalcontaminants in food that have not been iden-tified as significant by other methods. Al-though there is no truly independent way toverify the reliability y of priority lists, such listscould be generated for a pilot program de-signed to evaluate this approach vis-a-vis un-characterized monitoring.

Monitoring for UncharacterizedEnvironmental Contaminants

Uncharacterized environmental contami-nants are substances that may have enteredthe food supply but which have not beenclassified as regulated environmental con-taminants or suspected environmental con-taminants. Compounds may fall into this cate-gory because they are not known or sus-pected to occur in food. Because of a lack of

toxicity data on compounds, they may not berecognized as threats to human health. Thisclass of substances is similar to suspected en-vironmental contaminants in that there areno stipulated analytical methods to detectthem and no monitoring is mandated. Unchar-acterized environmental contaminants aredifferent from suspected contaminants inthat none have been placed on lists of poten-tially harmful substances.

Although validated analytical methods foridentifying uncharacterized environmentalcontaminants may be lacking, data on theirpresence or absence can be generated. Chem-ical analyses that are designed to show onlythe presence or absence of a compound in asample are called qualitative analyses. Inmany cases an accepted analytical methodfor one class of compound will yield quantita-tive results for those compounds and qualita-tive results for others, Therefore, the pres-

Ch. VII—— Monitoring Strategies ● 85

ence or absence of some suspected or unchar-acterized environmental contaminants maybe determined even though the chemist is notspecifically looking for them.

In setting up a system of monitoring for un-characterized environmental contaminants,some preliminary judgments would be madeabout the chemical nature of the target sub-stances. Classes of compounds to be moni-tored would be selected on the basis of theirstructural characteristics, their use, andtheir suspected toxicity. Trace metals, halo-genated hydrocarbons, or radioactive sub-stances are examples of such classes. Theclass of compound determines the type of ex-traction required to separate the substancefrom other constituents of food, as well as theinstrumentation needed to detect the pres-ence of the substance.

The available analytical methods bestsuited for the class or classes of compoundsunder consideration would be selected. Thequestion of whether analytical techniquesare sufficiently advanced to support unchar-acterized monitoring is explored in chapterVIII.

The establishment of an uncharacterizedmonitoring program would require the devel-opment of appropriate sampling and sample-handling guidelines. It would also be neces-sary to modify currently used analytical pro-cedures. For example, a typical program fororganic contaminants would involve prelimi-nary screening to establish baseline levels ofcontamination in samples of food and wateror selected indicator species over a givenperiod of time. This information would thenbe used to develop an appropriate samplingplan to determine changes or trends overtime. An increase in levels of an uncharac-terized substance indicates its entry into foodand water. This finding would trigger addi-t ional analyt ical efforts to characterize thenew compound. Preliminary information onthe subs t ance ’ s s t r uc tu r e wou ld be t r ans -mi t t ed t o t ox i co log i s t s f o r eva lua t i on andcomparison with available information onknown toxic compounds. If alarming trends

or changes were observed, corrective regula-tory actions could be taken.

Specimen Banking

It is difficult to detect environmental con-taminants unless one is specifically lookingfor them, Monitoring methods for identifyingsuspected and uncharacterized environmen-tal contaminants promise to partially allevi-ate the problems. Yet, even as new analyticalinstrumentation is developed and scientificknowledge expands, there will be new classesof compounds discovered in foods that havebeen present for years but were undetectablewith then-existing instrumentation. Informa-tion on how long the compounds have been infood, what kinds of foods are affected, andfrom what areas the foods were derivedwould be of great help to epidemiologists andpublic health officials who must decidewhether or not the chemicals have had (orwill have) an adverse impact on the public.

One approach to this problem is the collec-tion and storage of samples on a regular basisand in a manner that will protect their chem-ical integrity. In the future, when new instru-mentation is developed or a toxic compound isdiscovered in foods, samples can be with-drawn from storage and analyzed. This in ef-fect would be retrospective analysis. Investi-gators could go back in time, reconstructevents leading to a current situation, and esti-mate human exposures from the consumedfoods.

There are examples of this kind of retro-spective detective work. When high concen-trations of mercury were discovered in tunaand swordfish, pollution was widely heldresponsible, But analysis of museum speci-mens that had been stored for decades in-dicated that the mercury levels were prob-ably as high a century ago as now. Therefore,the mercury in fish may be from naturalsources and may have always been so. Thisdoes not mean that the metal is not a potentialhealth threat but rather that the potential ex-posure from eating the fish has not changedmuch over the years.


There is a problem in utilizing most exist-ing collections for retrospective chemicalanalysis. Since samples were not collectedfor use in chemical testing; they were notstored in a manner to maintain their chemicalintegrity. A 1975 survey of environmentalspecimen collections in the United States byOak Ridge National Laboratories concludedthat few of the existing collections were suit-able for retrospective chemical analyses (2).Therefore, a need exists for a national pro-gram to collect, store, and maintain environ-mental samples (including food) to allow ret-rospective investigations,

EPA and the National Bureau of Standardsare now working towards developing such aprogram by testing various methods of pre-serving samples for long periods of time with-out either adding unwanted chemicals or los-ing ones that are already in the sample. Anumber of scientists are encouraging thisprogram and similar efforts (3-5). If continuedfunding is made available for specimen bank-ing of environmental samples (includingfoods), future investigators will have an easi-er job of assessing what impact environmen-tal contaminants in foods may have.

SAMPLING

Sampling involves the systematic collectionof information from a portion of the environ-ment. Sampling is done in such a way that thecollected samples represent the whole interms of the information desired. In regula-tory monitoring the samples must be foodcommodities because the intent of the pro-gram is to determine the levels of regulatedsubstances in the food supply. This informa-tion is the basis for enforcement actions.

Food samples may not be the best indi-cators if the monitoring is meant to serve asan early warning system—in other words, todetect a substance soon after it enters the en-vironment and before it gets into foods. It maybe better to analyze nonfood samples such asriver sediments, water, or uneaten organsfrom food animals (the organs may concen-trate the substance to analytically detectablelevels before it can be seen in the flesh). Thefinding of an environmental contaminant innonfood samples would trigger the examina-tion of foods.

The following discussion outlines the pri-mary considerations in selecting samples forinvestigatory monitoring systems. Construct-ing a sampling plan for such systems involvesa number of decisions based on preliminaryinformation about the nature and extent ofenvironmental contamination. Such decisionsinclude the number, sites, frequency, andtypes of samples.

The number of samples to be taken de-pends on how much risk of being wrong weare willing to accept. In other words, to whatdegree of certainty do we want to know thatour food is free from environmental contami-nants? One-hundred percent certainty wouldrequire analysis of every food item. Accept-ance of a lesser degree of certainty allows theuse of less costly sampling approaches.

Before preliminary data are collected,there is no way to calculate the exact numberof samples needed to yield an answer ofspecified certainty. The most difficult factorto estimate is the variation specific to eachcontaminant and how it changes over timeand space. This kind of information wouldhave to be collected in pilot programs forsuspected and uncharacterized monitoringbefore a national sampling plan could bedeveloped. The number of samples taken willprobably be constrained by the money, man-power, and available laboratory resources.

The density and location of sampling sitesdepend on the socially acceptable level ofuncertainty, whether the contaminant stemsfrom a point or nonpoint source, and how it istransported in the ecosystem.

If one is dealing with widely distributednonpoint source environmental contaminantsthat are transported through water, the idealsampling locations would be rivermouths that

Ch. VII—Monitoring Strategies ● 8 7

are discharge points of major watersheds.one could very effectively monitor the in-dustrial portion of Michigan by samplingshellfish or fish from some two dozen majorrivers just as they enter the Great Lakes.Baseline levels in these foods could be deter-mined, but the origins of the contaminantswould be difficult to determine (6).

The other extreme is to monitor food prod-ucts on a production, site-specific basis. If theaim were to inventory all industries (includ-ing agriculture) that utilize and/or dischargea toxic substance, one could then routinelymonitor food products, fish, and game at eachidentified site. Theoretically, the kepone andpolybrominated biphenyls (PBB) s i tuat ionswould have been detected much earlier withs u c h a s y s t e m . T h e n u m b e r o f o p e r a t i o n s(large and small) that would need to be moni-tored is unknown, but the total appears to beso large that the costs would preclude consid-eration of this alternative (6).

A reasonable compromise approach is touse information generated under the ToxicSubstances Control Act on the types of chem-icals manufactured at various locat ions toguide in the development of a sampling planfor use in investigatory monitoring systems.The sampling plan would focus on some foodorganisms and some nonfood items. The dataderived from analyses of such samples wouldyield the greatest information about environ-mental contamination trends in the regionfrom which the samples were drawn.

The frequency of sampling depends on ther a t e s a t w h i c h t h e c o n t a m i n a n t m o v e sthrough, accumulates in, and decomposes outof the food product ion system being moni-tored, Different food production systems havedifferent genetic and environmental charac-teristics that determine the rates of materialdynamics or transfer. For a beef feedlot, arange of 50 to 180 days would include theperiod involving a single-batch process. Foran apple crop, a single sample per year wouldsuffice. The frequency of sampling may b edifferent for each type of production process.Once the species and the characteristics of

the production systems have been identified,the appropria te sampling frequency can bedetermined (6),

The selection of the types of samples to becollected is also cri t ical in identifying en-vironmental contaminants as they enter thefood chain. Although biological samples offermany advantages in monitoring systems, non-living samples may be preferable in some in-stances. An example might be bottom sedi-ments from rivers, lakes, or estuaries. Bottomsed imen t s a r e de r ived , f o r t he mos t pa r t ,f rom erosion of land and of ten br ing withthem to the aquat ic environment substancesthat are used on land such as herbicides orpest icides. Moreover , once they are in thea q u e o u s e n v i r o n m e n t , t h e y c a n “ s o r b ” o rconcentrate many substances found in indus-trial discharges. The contaminated sedimentsthen s e rve a s a mechan i sm to expose t heplants and animals that live in the waters to apart icular chemical . Contaminated sedimentsmay also be used to pinpoint the source of achemical once it has entered the river, lake,or estuary (6).

Other types of nonliving samples might in-c lude a i r , r i ve r wa t e r , d r i nk ing wa te r , o rrain. All have certain advantages and disad-vantages. For instance, the concentrations ofmany environmental contaminants in air andwater are very low, causing problems for theana ly t i ca l chemis t . W h e n a s u b s t a n c e i sfound in air or water, it is sometimes difficultto determine where i t entered the system.Howeve r , because we b rea the t he a i r anddrink the water as well as eat the food fromthese environments, air and water cannot beeliminated as potential samples (6).

The most appropriate biological samples inan invest igatory monitoring system shouldreflect key elements in the human food chain.Samples may include not only traditionalagriculture products but also fish, game,shellfish, crustaceans, and wild fruits andnuts, Many of these wild foods also accumu-late both point and nonpoint source environ-mental contaminants. Criteria for selection of

1– , 4—


the exact organisms should include the fol-lowing characteristics:

position in the food chain,lifespan,feeding behavior,understanding of the organism’s physiol-ogy and biochemistry,body fat content,mobility,availability for sampling, andutility to humans (6).

Once a sampling plan is developed andsamples co l l ec t ed and ana lyzed , t he da t amust be presented in a form useful to theregulator. The data analysis should be rapidand provide information on trends as well asspec i f i c concen t r a t i ons w i thou t s i gn i f i can tdistort ion or delet ion. Even with the best-designed computer retrieval system, the nec-essary data bank would become extremelylarge, complex, and expensive.

The supply of data must also be timely. Ifone wishes to regulate the level of an environ-mental contaminant in food when concentra-tions vary weekly, a monitoring system thatreports data with a 6-month delay is notworkable.

Finally, to develop information on exposuretrends and determine the effectiveness ofregulatory monitoring and enforcement, hu-man tissues, blood, and urine can be analyzedfor the presence of environmental contami-nants. Human monitoring can be performedfor either regulated, suspected, or uncharac-terized substances. A sampling plan to detecttrends in exposure among different popula-tion groups could be developed, but data gen-erated from human monitoring would be un-able in most cases to identify the source of ex-posure. At the present time, EPA houses theprincipal human monitoring program. Thisprogram is primarily concerned with pesti-cide residues.

QUALITY ASSURANCE

To ensure that data generated by any moni-toring strategy are as accurate as possible,schemes have been developed to pinpoint er-rors. These schemes are called quality assur-ance programs. Such programs are manda-tory in analytical laboratories because thepossibility y of errors always exist.

Errors arise from a number of sources. Forinstance, impure chemical reagents, dirtyglassware, or sample containers can impartcontaminants to the sample that may inter-fere with the analysis and result in falsereadings. Instruments are not always stableand may give false readings. Some samplesmay contain substances that interfere withanalyses, or a substance may be bound in asample in such a way that normal extractionmet hods will not extract it. Another factor isthe potential for human error in the labora-tory. These factors, singly or in combination,can lead to reported concentrations that arein error. Since commodities that violatestandards could be marketed and consumedif the results were erroneously low, human

health might be affected. If the results areerroneously high, undue economic hardshipmay be imposed on the food producer.

The Federal agencies that monitor foodsfor environmental contaminants are aware ofthese problems. Thus they use standardizedanalytical methods that have been tested andtherefore offer some assurance that the re-sults will be acceptable in court. When a vio-lative sample is found, the product or batch isreanalyzed whenever possible to assure thatresults are valid.

These two practices are part of a qualityassurance program. There are others as well.When a new chemical extraction or analysistechnique is tested, it is important to analyzea sample w i t h known composition t o checkthe validity of the technique. Also, duringroutine determinations samples of knowncomposition should be analyzed to check onthe other types of potential errors. Thesamples of known composition are called“reference material" and may come from

Ch. VII—Monitoring Strategies ● 89

var ious s o u r c e s including t h e N a t i o n a lBureau of Standards and EPA,

Avai lab le re fe rence mater ia l s do notalways satisfy the needs of current chemicalmonitoring programs, since such materials donot contain many known environmental con-taminants. Moreover, the contaminants maynot be stable under the storage method used.This is particularly true for synthetic organicchemicals. Another problem is that the typeof reference material-i.e., beef liver—maynot be similar enough to the food samples tobe analyzed-–i.e., fish—to be very helpful.

This points up an important gap in ourability to analyze accurately for environmen-tal contaminants in foods. The variety ofreference materials and the variety of com-pounds of known concentration in these mate-rials are insufficient to satisfy the needs ofthe analysts. More effort must be e x p e n d e dto correct this problem.

Collaborative studies that involve morethan one laboratory or group analyzing thesame sample by the same or different meth-ods are part of a quality assurance program.If all results are similar within acceptablelimits there is some assurance that the meth-od(s) are precise and perhaps accurate,

Often, more than one method can be usedto measure a given contaminant. Confidencein accuracy can be increased if the methodsagree. This is one of the reasons that a moni-toring laboratory should have several meth-ods available.

All of these aspects are important toassure that the proper answers are gener-ated by a monitoring laboratory. All are time-consuming and expensive. Therefore, anychemical monitoring program must allocateas much as 10 t o 20 percent of its time for thisincreased workload.

CHAPTER VII REFERENCES

Chapter VIII

Monitoring Instrumentation

.

Chapter VIII

Monitoring Instrumentation

Various instruments have been designed to detect and quantify organicchemicals, trace metals, and radioactivity in foods. This chapter focuses on notonly the technological state of this analytical art but also how such instrumenta-tion could be applied in an investigatory monitoring program.

ORGANIC ENVIRONMENTAL CONTAMINANTS

Analysis

The vast number and wide diversity of nat-ural and synthetic organic chemicals posedifficult analytical problems. Many methodsand instruments have been developed to de-tect and quantify specific environmental con-taminants in foods. There are generally threesteps required for each type of analysis: ex-traction, cleanup, and detection and quantifi-cation. Figure 6 is a flow chart depicting howthese steps would be applied to the analysisof the Environmental Protection Agency pri-ority pollutants in food.

Extraction usually involves mixing the foodsample with a selected solvent. In this step,the chemical is removed from the food anddissolved in the solvent. The time required forthis step depends on the physical and chemi-cal characteristics of the food sample and thesubstances. In some cases, the process maytake 24 hours or more. The initial extract con-tains not only the substance of interest butpossibly other organic compounds of similarvolubi l i ty that m u s t b e r e m o v e d b e f o r eanalysis .

The second step is known as the cleanup orisolation stage. The complexity and time re-quired for this procedure depends on the foodsample and the number of substances to beremoved. In many cases multiple cleanupsteps are necessary. The end products ofcleanup procedures are fractions containingdifferent classes of organic compounds.

The final step, detection and quantifica-tion, requires the use of highly sophisticated

instruments and techniques. Among the mostfrequently used methods are mass spectrome-try, gas chromatography, and liquid chroma-tog raphy . Tab l e 18 summar i ze s t he t e ch -niques available for qualitative and quantita-tive organic analysis. More information canbe found in appendix G .

The gas chromatography is an instrumentdesigned to separate, identify, and quantifyorganic compounds. In simplified terms itconsists of four components: an injectionport, a column, a detector, and a recorder.There are many types of injection chambers,columns, and detectors, but the principles ofoperation are similar.

Gas chromatography involves vaporizationof the sample to be analyzed. The gaseoussample then passes through a long tube orcolumn packed with a solid matrix which iscoated with an organic compound. This iscalled the “liquid phase. ” The rate at whichvarious organic compounds in the samplepass through this column is a function ofvarious chemical and physical properties,notably molecular weight. polarity, and geo-metric structure. The time of passage throughthe column is called the “retention time. ”Each compound has a characteristic reten-tion time in the column.

To determine when and how much of eachsubstance leaves the column, a detector isplaced at the end of the column. The retentiontime can be used to identify each compound,while the strength of the electrical signalfrom the detector indicates the quantity. The

93

94 Q Environmental Contaminants in Food

Figure 6.— A General Scheme for the Qualitative and Quantitative Analysis of the Organic EPA PriorityPollutants in Semisolid Foods

I Extraction procedure Alternatesuitable for particular

* methodsfood sample type ● w

t I

II

I Initial cleanup

bJ I

of extract- - 0 - - - - - - -

I

I Initial interpretationand quantification I

a Identification is based on the presence of characteristic ion fragments and associated chromatographic retention timedata. Absolute identification would require a detailed interpretation of the complete mass spectrum of the organic com-pound of interest.

SOURCE John L Ldseter Approaches to Monitoring Environmental Contaminants [n Food OTA Working Paper 1978

identity of the compound is confirmed by com- gerprint” are compared to a s tandard “f in-paring its retention time to that of a series of ge rp r in t ” t o de t e rmine how much o f e achstandard solutions (of known composition) in- substance is present in the sample, If a peakjected into the gas chromatography under the in the sample f ingerprint , for example, hass a m e conditions. T h e a r e a s u n d e r t h e the same retention time but twice the area asgraphed peaks of a sample readout or “fin- one in the standard fingerprint, it is assumed

Ch. VIII— Monitoring Instrumentation ● 95

Table 18.—Techniques Available for Qualitative and Quantitative Organic Analysis

ApproximateMethod detection limit, gm Specificity or common uses

Gas chromatographyR e t e n t i o n I n d i c e s . . ,E l e c t r o n c a p t u r e

F l a m e p h o t o m e t e r .N i t r o g e n / p h o s p h o r u s

Chemical methodsPyrolysis ... ... . . .C h e m i c a l r e a g e n t s . . ,E l e c t r o l y t i c s y s t e m s .

Instrumentationi n f r a r e d - g r a t i n gInf rared- in ter ferometerUltraviolet ... . . . . . . . .Proton magnetic resonance

Mass spectrometerBatch inlet .GC-MS mode . . . . . . . .Multiple-ion detection . . . .

Detects most compoundsHalides, conjugated carbonyls, nit riles, di- and

trisulfidesPhosphorus, sulfurNitrogen, phosphorus

Compound-type determinant ionClassical-functionality determinationSulfur, nitrogen, halogens

Compound-category-type identificationCompound-category-type identificationAromatics conjugated carbonylsExcellent for function, some molecular weight

data.

Best for complete identification,molecuIar weight structure, andfunction. Confirm any compound.

SOURCES Adapted f;om W McFadden Techniques of Comb!ned Gas Chromatography/Mass Spectrometry Wlleylntersc{ence New York N Y 1973, p 4, and “Trace Orqanlc Analysls ‘ Enwronmental Science and Technology 12757(1978)

that the identity of the sample peak is thesame as that of the known standard but thattwice as much is present.

Recent developments in analytical organicchemistry promise new, more sensitive tech-niques for monitoring of synthetic organicchemicals in foods. These developments in-clude high-resolution glass capillary columnsfor gas chromatography, and the linking ofgas chromatography to mass spectrometers.

The glass capillary column is much moreefficient in separating compounds than theolder “packed” columns (figure 7). The gaschromatograph’s packed column separatesthe individual components so that they exitthe column and enter the detector one at atime. Some of the early data on the pesticidefamily DDT-DDE-DDD were incorrect be-cause polychlorinated biphenyls (PCBs) hadsimilar retention times in packed columns. Asa result, reported concentrations of DDT,DDE, and DDD were often too high becausethe peak areas were really reflecting a com-bination of pesticides plus PCBs (l). Becauseof their superior resolution power, glass cap-illary columns avoid some of these problems.

Also because less cleanup is required, theiruse reduces the possibility that an importantenvironmental contaminant may be removedin the process of preparing a sample for anal-ysis.

Although the capillary column offers anumber of advantages, it is more expensiveand difficult to use than the standard packedcolumn. More highly trained personnel arerequired to operate capillary columns and in-terpret results. However, many chemists feelthat the advantages far outweigh any difficul-ties. Glass capillary columns are not nowwidely used for monitoring synthetic organicsin foods. Packed columns continue to play animportant role in the organic chemistry moni-toring laboratory, but their future use may berestricted to more specific analyses.

Gas chromatography are currently theprimary means by which laboratories moni-tor for organic chemicals. Some chemists,though, prefer mass spectrometers becauseof their high sensitivity and flexibility (z). Thecoupling of a gas chromatography to a massspectrometer introduces a new dimensionthat allows a chemist to identify compounds


Figure 7. —Comparison of the Gas ChromatographicAnalysis of a Standard Mixture of Polychlorinated

Biphenyls (Aroclor 1254) on Two Types of GasChromatographic Columns

A: Analysis Using a Conventional Packed Column

r I I 1 I 1 I I 1 10 6 12 18 24 3

Time (Minutes)

B: Analysis Using a High-Resolution GlassCapillary Column

I 1 1 1 I I I 1

0 10 20 30 40Time (Minutes)

SOURCE John L Laset II Approaches to Monltorlng Environmental Contamlnants Ir Foc)i OTA i’dork(ng Pdper 1978

classed as uncharacterized (3). Moreover, re-sults from a mass spectrometer can confirmthe identity of known chemicals whose pres-ence (as indicated by gas chromatographyalone) is in doubt because of sample contam-ination or other factors. Mass spectrometersare expensive and in addition require compu-terized data management systems.

Because of their physical and chemicalcharacteristics. many organic compoundscannot be analyzed by gas chromatography.

Therefore, instruments such as infraredspectrophotometers and high-pressure liquidchromatography are necessary. These instru-ments add many thousands of dollars to thecosts of setting up and equipping an up-to-date organic chemical-monitoring laboratory.

Application to InvestigatoryMonitoring

Although the instrumentation describedabove has been available for several years, ithas only recently been utilized for investiga-tory monitoring. Experimental projects incor-porating computer analyses are now understudy in at least three research facilities: theBodega Marine Laboratory, the University ofNew Orleans, and the Virginia Institute ofMarine Science, These projects employ high-resolution glass capillary gas chromato-graphs coupled to mass spectrometers andsophisticated computer analyses, The result-ing “fingerprints” or readouts from a sampleextract are highly complex and show thepresence of not only known or suspected envi-ronmental contaminants but also many or-ganic compounds that must be classified asuncharacterized.

The three research groups are now testingan approach to determine which compoundsdeserve further testing. The gas chromato-graphic fingerprints are computerized so thatinformation on retention times and peak in-tensities are stored. Subsequent samples arecollected and analyzed and the data are com-puterized in the same manner. By comparingresults from the same location over time (orwith samples taken from other locations), onecan determine whether peaks are increasingor decreasing, whether new peaks have ap-peared or old ones have disappeared, orwhether some peaks occur only in some loca-tions.

With appropriate computer programs or‘‘soft ware, ” the data bank could be asked ifthere is a peak that occurs in samples fromonly one area or location. If so, the inves-tigator can assume that the compound is an-thropogenic, or man-induced, The compoundcould then be identified and its source con-

Ch. VIII— Monitoring Instrumentation ● 97

trolled before the concentration increased tothe point that a major contamination episodeoccurred. This type of monitoring could havedetected kepone in the James River, Va., longbefore thousands of pounds of the pesticideentered the river.

Another question which could he asked ofthe data bank is whether there is a peak thatis increasing with time. In other words, doessampling over time indicate that a compoundis becoming more concentrated? Such a find-ing would flag the compound as one whichmerits attention, Had this type of monitoringprogram been in effect and this question beenasked earlier, widespread pollutants such asPCBs could have been detected long beforethey were,

Other information from the chemical labor-a tory car-r draw attention to an uncharacter-ized substance deserving attention. For ex-ample, gas chromatographic detectors exist

that respond mainly to organic compoundscontaining halogens such as chlorine or bro-mine. Because natural ly occurr ing halogen-containing organic chemicals are rare andare far outnumbered by manmade ones, anuncharacterized peak from a halogen-specif-ic detector may represent a manmade com-pound, Histor ical ly, halogen-containing or-ganics such as DDT, PCBs, polybrominatedbiphenyls (PBBs), and kepone have causedmajor pollution crises. Therefore, such infor-mation may be sufficient to focus attention ona given peak from a halogen-specific detec-tor.

Uncha rac t e r i zed subs t ances may o r maynot be dangerous to human heal th i f con-sumed. But no assessment of toxicity can bemade until the compounds are identified. Toperform the chemical and physical tests nec-essary to identify even one uncharacter izedpeak may cost from $10,000 to $100,000, andin some cases even more.

DETECTING AND QUANTIFYING TRACE METAL CONTAMINANTS

Trace metals pose many of the same prob-lems that plague the analyst monitoring fororganic chemicals. However, the number oftrace metals is much smaller. If only the totalconcentration of a metal is sought, more rig-orous extraction procedures can be em-ployed. But some metals, such as mercury,form organic complexes--for instance, meth-ylmercury--which give rise to a subset calledmetallo-organics. These metallo-organics arenot stable under harsh conditions and areoften changed during rigorous extraction pro-cedures.

A n a l y s i s

Analysis for trace metals usually requiresextraction and cleanup before the quantita-tive analysis can be performed. Extractionoften involves destruction of the sample’s or-ganic structure to release the metals from thesolid or liquid food. There are a few analyti-cal instruments such as X-ray emission spec-trographs that can accept solid or liquid sam-

ples, thus eliminating the extraction step (4).But these instruments cannot detect all met-als at environmental concentrations.

Depending on what instrumentation is usedin the final analysis, some cleanup of the ex-tract may be required. The cleanup step isnot nearly as frequent with trace metals,however, as with organic chemicals except ifa metal lo-organic complex is sought. Formetallo-organic compounds cleanup and sep-aration procedures may be similar to thosefor organic compounds. After extraction andpossibly cleanup, the sample is ready to beanalyzed by an instrument to determinewhich metals, and how much, are present.There are a variety of instruments availablefor the determinations.

Probably the most commonly used instru-ment is the atomic absorption spectrophotom-eter (AA). The cost of this instrument is nom-inal ($ 15,000 to $25,000), and i t can be oper-ated by a well-trained, motivated technician


with a high school chemistry and physicsbackground working under the supervision ofa chemist. One drawback of this instrumentis that only one metal can be analyzed at atime.

In contrast, more sophisticated instrumen-tation such as X-ray emission, proton-inducedX-ray emission, or plasma emission spectrom-eters can cost in excess of $100,000. Withthese instruments, 20 to 60 elements can bedetermined simultaneously. These more so-phisticated instruments require more highlytrained personnel to operate and maintain.Table 19 summarizes the techniques avail-able for qualitative and quantitative analysisof trace metals. More information can befound in appendix H.

Application to InvestigatoryMonitoring

It is important to note that most AA unitscan detect only the specific element (metal)that they are set up to analyze. By contrast,the gas chromatography may sometimes detectthe presence of organics other than thosebeing sought. As a result, metals are morelikely than synthetic organics to escape detec-tion simply because they are not programedfor analysis.

With the rapid development of lower costtechniques for detecting more than one ele-ment at a time, it would be possible to obtaindata on trace metals in foods that were pre-viously not sought and therefore not obtained.

This is analogous to uncharacterized organicmonitoring where many unsought compoundscan be found in the analysis of a food extract.Given the proper data systems, these “other”metals can be tracked, and perhaps pollutioncrises can be averted by detecting anomalousconcentrations early enough.

Speciation of Trace Metals

Often the active or functioning forms of ele-ments must be identified and measured. Theneed for the development of analytical meth-odology to identify and quantify the chemicalform of metals found in the environment—i.e., chemical speciation—is pointed up by thefollowing examples.

●

●

Although arsenic is toxic, the plus threeoxidation state, As (III), is more toxicthan the plus five state, As (V). The com-pound arsine, AsH3, is perhaps the mosttoxic chemical form of arsenic.

Alkyl (organic) mercury compounds posegreater propensities for bioaccumula-tion and the associated health effectsthan do the more common inorganicforms of mercury (but these also canvary greatly in toxicity).

Chemical forms or states are an importantdeterminant of a trace metal’s toxicity in bio-logical systems. The selection of elementalanalysis techniques capable of specificallymeasuring those chemical forms most impor-tant from a biological-effects viewpoint

Table 19.—Techniques Available for Qualitative and QuantitativeTrace Metal Analysis

Ch. Vlll—Monitoring Instrumentation ● 9 9

should be a primary objective of any environ-mental monitoring program.

The status of the present analytical tech-nology is inadequate for this purpose in mostroutine monitoring programs. As a result, thefunctioning forms of most elements cannot beadequately studied. Thus the determination

ANALYSIS OF FOODS

Analysis

Measuring radioactivity in foods is a phys-ical process. It is most efficient when theradioactivity from a relatively large samplecan be placed close to a detector. This meansthat direct measurements of radioactivity inbulk samples are only useful when levels ofcontamination are relatively high. Most meas-urements are preceded by preparation andpossibly chemical separation to reduce thebulk of the material and to improve the effi-ciency of the measurement.

Foods generally have a high water content.The primary method of reducing bulk isfreeze-drying or drying at room or at elevatedtemperatures. Most of the radionuclides of in-terest are not volatile under these conditions.Only elements such as tritium or iodine mayexperience losses. The dried materials can bereduced further by ashing at elevated tem-peratures, by cold-ashing with oxygen, or bywet-ashing with oxidizing acids. Dry-ashing isthe simplest process, but it is most likely tolead to loss of volatile elements. With careeven cesium, polonium, or lead can be re-tained. The other processes should not resultin losses of elements of interest, except foriodine, tritium, and carbon.

Another way to reduce bulk is to treateither the original sample or the dried orashed material with acids or other solventsand thus remove the desired elements fromthe bulk of the sample. This requires consid-erable testing beforehand to be certain thatthe process operates in the desired manner.The radionuclides of interest in ashed foodsshould be soluble in strong acids if ashing

of the important (threshold) concentrationlevels for various elements typically has beenrather crude. The most important current re-search need related to trace metal monitoringis the development of methods to measure thespecies rather than the total amount of tracemetals.

temperatures have not been excessive. Ifthere is concern that insoluble particulatemay be present, it is necessary to use moredrastic methods such as fusion to bring theentire sample into solution.

In addition to bulk reduction, radiochemi-cal separations are required to isolate thedesired radionuclide both from the remainingbulk constituents and from other radionu-clides that would interfere in the measure-ment. It is also necessary to convert the finalproduct to a form suitable for presentation tothe counter. This may involve electrodeposi-tion, precipitation, or other processes.

The actual mass of radionuclide that ismeasured is almost always extremely small.As a result, many of the normal chemical re-actions used in analytical chemistry to isolatean element are not appropriate. For instance,precipitates may not form. Therefore, it iscommon to add a few milligrams of a carriermaterial, preferably the inert form of thesame element. When such an inert form doesnot exist, it is frequently possible to use simi-lar elements as a carrier (for example, substi-tuting barium for radium). The inert formthen follows normal chemistry, carrying theradionuclide with it. It is also worth notingthat even when the separation technique doesnot depend on the mass of element present, acarrier may still be useful in preventing un-wanted coprecipitation or absorption onglassware.

In the chemical separations there is consid-erable use of classical analytical chemistrybased on precipitation. In most cases, precipi-tation will be the final collection step in

100 “ Environmental Contaminants in Food

preparing the desired radionuclide for count-ing. In other cases the techniques of ion ex-change, liquid extraction, distillation, andelectrolysis may be required to prepare thesample for counting.

In certain cases, the total amount of a sam-ple may be limited, and analysis for severalradionuclides may be required. Proceduresshould be available for the sequential analy-sis of single samples, even though separatesamples are used for routine work.

The measurement method used depends onthe type of radiation, the form of the sample,and (to some extent) the amount of radioac-tivity. The selected analytical procedureshould be designed so that the sample isbrought to a suitable form for the equipmentand conditions that exist.

The major emitted radiations are generallygrouped as alpha, beta, and gamma. Alpharadiation is characteristic of the natural andartificial radionuclides of high atomic weightand consists of energetic particles with verylow penetrating power. Its hazard is signifi-cant only within the body, where alpha-emit-ting nuclides can irradiate specific sensitivetissues. Beta radiation appears in both natu-ral and manmade radionuclides, and consistsof electrons possessing kinetic energy andmodest penetrating power. Gamma radiationis pure electromagnetic radiation and is ex-tremely penetrating. Thus, it is hazardous ex-ternally as well as when it is present in thebody. Gamma radiation can be directly meas-ured in foodstuffs, while alpha and beta emit-ters generally must be separated from thebulk constituents of the sample before theycan be measured.

It is possible to measure the total gamma,total beta, or even the total alpha activity in asample of food. Unfortunately, such data arevalueless in estimating human exposure. Theaccuracy of such estimates is very poor be-cause natural potassium usually interferes,and the chemical and radiation character-istics needed to evaluate possible hazards arenot known.

It is possible, however, to set a particulartotal activity level as a screening level for aspecific food, If the measured value is belowthe screening level, no analyses for individualradionuclides are performed. In such a case,the measurements should be considered in-ternal data and the numerical results shouldnot be published, Any report should merelylist the samples as having activities below thestated screening level. Table 20 summarizesthe typical limits of detection for radionuclidemeasurement.

Alpha Emitters.—The measurement ofalpha activity is best carried out on a verythin sample to avoid self-absorption of thealpha particles. Most metals are electro-de-posited onto small smooth discs of stainlesssteel, nickel, or platinum. Evaporation of puresolutions is used in some cases. The measure-ment of the total alpha activity can be con-ducted either in thin-window counters or byscintillation counting with zinc sulfide phos-phor. Both techniques are highly efficient, butthe scintillation method can give a considera-bly lower background with a consequentlylower limit of detection.

A somewhat less-sensitive technique is theliquid scintillation spectrometer described inthe next section.

Alpha spectroscopy provides two other al-ternatives: the Frisch grid ionization chamberor the silicon diode solid-state detector. TheFrisch grid unit can handle large area sam-ples but is slightly poorer in resolving closelyseparated energies. On the other hand, thesilicon diodes available are all quite smalland can only count samples of about 1-cm di-ameter with high efficiency.

Beta Emitters.—Beta counting is a littlemore flexible in the mass of material that canbe present at the time of counting. This is truefor higher energy beta emitters but carbon-14and tritium present a problem. Since each in-dividual beta emitter gives off particles witha range of energies from zero to a character-istic maximum, beta spectrometry is not pos-sible for food samples.

Ch. VIII—Monitoring Instrumentation c 101

Table 20.—Typical Limits of Detection for Radionuclide Measurement

CounterNuclide efficiency %

Amerlclum.241 25

Ceslum-144 22

Cesium-137 30

2

Cobalt-60, otheract ivatedp r o d u c t s 1

Trit ium 30

Iodine- 131 25

4

Phosphorus-32 45

Plutonium-239. -240 25

Radium-226 56

40

Strontium-90 45

Thor! urn-230, -232 25

U-Isotopic 25

Counterbackground cpm

0.001

0 4

0 4

006

005

5

0.4

0 3

0 3

0005

0 2

0.001

0.3

0001

0.005

Chemicalyield %

60

75

75

75

—75

80

80

60

85

90

85

80

70

75

10 min

0 3

6

4

30

100

15

5

25

3

0 6

1

1

2

0.3

0.7

SOURCE N H Harley Analysts of Foods for Radloactlvlty OTA Work Ing Paper 1979

LLD (dpm) per sample for various counting times

40 mln

015

3

2

14

50

6

2

13

1.5

0.3

0.6

0.5

1

0.1

0.3

400 mln

0.05

0 9

0 6

4 5

15

2

0.7

4

0.5

0.09

0.2

0.15

0,4

0.04

0.1

1,000 mln

003

0 6

0 4

3

10

1.5

0.5

2 5

0.3

0.06

0.1

0.1

0.2

0.03

0.07

The available counting equipment forquantitative measurement includes geigercounters, proportional counters, and scintil-lation counters. The geiger counter is rela-tively inexpensive and requires only simpleelectronics but is not popular and is generallynot available as a counting system. The thin-window proportional counter is used widelyand has both reasonably high efficiency andlow background. For low-level samples thebackground can be further reduced by anti-coincidence techniques. These add to thecomplexity and cost of the system but aresometimes necessary.

Scintillation counting can be performed intwo ways. Solid scintillators can be used forcounting chemical precipitates collected onfilter papers, and liquid scintillators can be

used whenever the sample can be made mis-cible with the scintillating solution itself. Thiscan even be done with solids by suspendingthem in a scintillating gel. The advantage ofliquid scintillators is their high efficiency,even for the low-energy emitters carbon-14and tritium. Scintillation systems for countingprecipitates are not commercially availableat present. There are, however, many liquidscintillation systems on the market, most ofthem with automatic sample changers. Theseuse high levels of activity and short countingtimes. The better systems also have a provi-sion for rather crude spectrometry. They candistinguish qualitatively and quantitativelyamong carbon-14, tritium, alpha emitters,and higher energy beta emitters.

Gamma Emi t t e r s .—Gamma rays are sopenetrating that the detector must have a


considerable mass to absorb enough energyto produce a response. Since energy absorp-tion is required for spectrometry, solid detec-tors are most useful, Sodium iodide is a popu-lar detector because crystals can be fabri-cated in large sizes and are transparent tothe scintillations produced by radiation. So-dium iodide has high efficiency but poor en-ergy resolution and is now applied to sampleswhere some separation has taken place. Theadvantages are that samples of food canoften be counted directly without chemicalpreparation. Milk is a good natural example,as the metabolism of the cow removes mostgamma emitters other than isotopes of ce-sium, iodine, and naturally radioactivepotassium.

More complex spectra can be resolvedwith the solid-state germanium diode detec-tor. Interferences from radionuclide impur-ities are greatly reduced compared to spectrafrom sodium iodide detectors. The efficiencyof the diode is low and, for many analyses, aspectrometer can only be used for one meas-urement a day, Another disadvantage is thatthe detector must be kept at liquid nitrogentemperature to maintain its detection capa-bility.

Diode spectrometers may also be used tomeasure the low-energy gamma-rays that ac-company alpha emission. This allows directmeasurement in some environmental sam-ples, but the levels in foods have not beenhigh enough for this technique.

General Requirements.—The choice of acounting procedure depends on the precisionrequired. The relative precision of a quantita-tive counting measurement, in turn, is in-versely proportional to the square root of thenumber of counts obtained. Thus any im-provement in precision must be obtained byincreasing the number of counts. This can bedone by using larger samples, by counting forlonger times, or by using counters with higherefficiency, A secondary improvement is possi-ble for low-activity samples by decreasing thebackground, Each of these improvements hassome drawback, and selection of the optimum

balance requires a weighing of cost, man-power, and quality.

Applications to lnvestigatoryMonitoring

The monitoring of foods for radioactivityshould not be considered a primary defenseagainst human exposure. The first indicationof hazards should always come from informa-tion on releases or from measurements of ra-dioactivity in air or water. Once the existenceof contamination has been established, foodscan be analyzed to evaluate the potential haz-ard to man.

Knowledge of the source of radioactivecontamination gives a good indication of thenuclides that can be expected in the sample.This information helps in planning the anal-ysis, since requesting a complete analysis forall radionuclides or even for all types ofradioactivity in a single sample would lead toa lengthy and expensive operation. Indeed,monitoring for suspected contaminants ismore applicable than the uncharacterizedmonitoring.

The general groups of nuclides that maybeencountered include those that occur natural-ly such as radioactive potassium and mem-bers of the uranium and thorium series, arti-ficial fission products such as transuranicelements, and other activation products thatresult from nuclear weapon explosions andnuclear reactor operations.

Fission products are a very complex mix-ture when they are formed, but the short-lived radionuclides die out rapidly and themixture becomes simpler within a few days.The transuranics (plutonium, americium, etc.formed by activation of the basic fissionablematerial) are of some interest because oftheir high toxicity when incorporated into thebody. Present evidence, however, indicatesthat their uptake through the gut is relativelysmall and that dietary intake is not a signifi-cant problem. The other activation productsare frequently elements that make up steel orother metal containers or structural ele-

Ch. VIII—Monitoring Instrumentation ● 103

ments. Radioactive manganese, chromium,cobalt, zinc, and iron are particularly com-mon and result from interactions of the mate-rials with neutrons released in the nuclearreaction, Contamination of foodstuffs withsingle nuclides is extremely unlikely, andmore than one member of any group willprobably be present in any sample.

In contrast to most other pollutants, the ef-fects of radiation are considered to have alinear response regardless of the level, Thusthere is no threshold and no absolutely safelimit. The analytical significance of this isthat the lower limits of detection for radioac-tive substances have been brought down tovery low levels. The simple yes-or-no testingfor acceptability that satisfies regulators formany other pollutants in foods cannot beused.

ESTIMATED COST TO EQUIP AINVESTIGATORY

Table 21 illustrates estimated costs, re-quired space, and estimated downtime for alaboratory designed to conduct investigatorymonitoring. In addition, the number of sam-ples the system would be able to analyze in ayear are estimated.

These figures do not reflect the total costsfor establishing a national system of inves-tigatory monitoring. Such a cost estimatewould require information on the total num-

Almost all radionuclides of interest incases of contaminating events now exist infoods in small but measurable quantities.Short-lived nuclides are the exception and thetransuranic elements are only present atlevels that require considerable effort inanalysis. Since most of the radionuclides arealready present in foods, measurementsmade for background information should pro-duce a numerical answer, not merely an in-dication that the amount is less than somepre-set value. This accumulation of back-ground data provides a valuable baseline forevaluating hazardous levels following a con-taminating event. The natural activity dataare equally valuable, since the amount of in-formation on food concentrations is presentlyinsufficient for valid comparisons with man-made radioactivity.

LABORATORYMONITORING

ber of samples to be collected and analyzed.Information of this sort could be obtainedthrough a pilot project to investigate thefeasibility of the two investigatory monitoringapproaches. This would determine the num-ber of laboratories, the number of people andtheir salaries, and costs of equipment mainte-nance, supplies, and training. Once the sys-tem is setup and running, some savings mightbe realized through automation.

54.5>5 0 - 79 - B


Table 21 .—Estimated Costs to Equip a Laboratory to Conduct Uncharacterized Monitoring

ApproximateEstimated

downRequired space Samples per year timeAnalytical instrumentation

Synthetic organics1. Small, high throughput, GC-MS-

data system with automated liquidInjection device.

2. Electron impact/chemicalIonlzatlon-equipped high-resolution mass spectrometerdata system with automatedi n j e c t i o n d e v i c e

3. Gas chromotographic flameionization detector, electroncapture detectors (additionalchromotographic systems may ber e q u i r e d ) .

4 Liquid chromotograph interfacedto mass spectrometer system

5. Central data management system

6. Cold storage and processingf a c i l i t i e s

Synthetic organics subtotal

capital cost

5,000-6,000 ft. ’ 200-300 for 20- 50%laboratory supporting uncharacterized environmental

facilities and equipment contaminants as welI as knownare required, environmental contaminants

$100,000

$200,000-$300.000

20- 50%

$15,000 20-50 %

$20,000 20- 50%

20- 50%$75,000-$1,000,000$500! 000

$1,585,000-$1,935,000

Trace metals1. Inductively coupled plasma $100,000 Approximately 2,500 ft.’

of laboratory equippedwith supporting facilities

and equipment arerequired for items 1-4

3,000-5,000 3-15%

Multielement atomic emissions y s t e m

2. Flame and furnace atomicabsorption spectrophotometer( s i n g l e e l e m e n t m o d e )

3. Electrochemical instrumentation4. Central data management system.

Trace meta ls subto ta l .

$25.000 3- 15°/0

3-15% 3-15%

Radionuclides1. Four position alpha spectrometer

and detectors, multichannela n a l y z e r a n d o u t p u t

1,000-4,000$21,000 Approximately 1,500 ft. ’laboratories equipped

with supportingfacilities and equipment

are required5 %2. Germanium diode gamma

spectrometer wit h detector,shield, electronics, PDP-11c o m p u t e r a n d o u t p u t

3. Four general purpose proportionalc o u n t e r s ,

4. Liquid scintillation spectrometer,automatic sample. or alpha-counting capabilities

Radionuclides subtotal

Total. . . . . . . . . . . . . . . . . . . . . . . .

$100,000

$10,000

$18.000

5 %

5%

$149,000$1,934,000.$2,309,000

SOURCES Adapted from J L Laseter Approaches to Monltonng Environmental Contaminants In Food OTA Work!ng Paper 1978 R K Skogerboe Analytical Systerns for the Determ!natlon of Metals {n Food an(j Water Supplles OTA Working Paper 1978 and N H Harley Analysls of Foods for Radloactlvlty OTAWork!ng Paper 1979

Ch. Vlll—Monitoring Instrumentation ● 105

1. Butler, P. A. Environmental Protection Agen- vironmental Contaminants in Food, OTA~~[)rking paper, 1978.cy, Gulf Breeze Environrnen tal Research Lab-

oratory, personal communication, 1978. 4, Sko,gerboe, R, K, “Analytical System for the2. Anonlrnous. “Trace Organic Analysis,”” En- Determination of Lletals in Food and Water

vironment~]f Science anci 7’ec~nology, 1 2 ( 7 ) : Supplies, ” OTA Working Paper, 1978.757, 1978. 5, Harley, N. Ii. “Analysis of Foods for Radio-

3. Laseter, J. L. “Approaches to Monitoring In- activity,”’ OTA W’orking Paper. 1979.

. .

Chapter IX

Options

.

Chapter IX4

Congressional Options

The present system of controlling environmental contaminants in food con-sists of two parts: regulatory procedures to set and enforce limits for environmen-tal contaminants, and monitoring procedures to detect lots of food in violation ofestablished limits. Each State has authority for regulating food grown and con-sumed within its boundaries. The Federal Government is responsible for regulat-ing food in interstate commerce.

Congress can choose to maintain this system. But if it wishes to put greateremphasis on protecting consumers from contaminated food, one or more of the op-tions discussed below could be adopted. None of these options (except for thefirst) are mutually exclusive.

The current regulatory approach to con-trolling environmental contaminants in foodinvolves the setting of action levels (and occa-sionally tolerances), coupled with regulatorymonitoring for known [and a few suspected)contaminants. Food containing an amount ofa contaminant that exceeds the action levelor tolerance can be identified through suchmonitoring. This food is then removed fromthe marketplace. Public exposure is thustheoretically limited to those foods containingquantities of contaminants that fall underprescribed action levels or tolerances.

Pros: There are two principal advantagesto maintaining this system. No additional ap-propriations or legislation are required. Nochanges in existing regulations are neces-sary.

Cons: There are a number of disadvan-tages in retaining the current system. Thetime needed to identify an environmental con-taminant in food and take corrective actionwould not be shortened, The people would

people against exposure, In other words, ac-tion levels and tolerances permit a certainlevel of contaminant to be present in food. Un-less action levels or tolerances are reduced,little effort will be made to eliminate the con-taminant, The threshold concept on which ac-tion levels and tolerances are based—thatthere are exposure levels to toxic substancesbelow which there are no effects on health—is being increasingly challenged (especiallywhen carcinogens are involved).

If high action levels or tolerances are es-tablished, exposure is not reduced. Loweredaction levels and tolerances may reduce pub-lic exposure. But even low limits are set onthe basis (among other things) of nationwideper capita consumption of a particular food.Contamination problems, however, may be lo-calized and further influenced by regionalfood consumption patterns. Thus, a local pop-ulation may be highly exposed to a toxic sub-stance although the tolerance, based on na-tional consumption, may be low.

cont inue to be exposed unt i l a contaminantMoreover, there is no requirement for re-

was detected and identif ied, and an act ionlevel put into effect,

view once an act ion level has been estab-lished. An agency is under no pressure to ac-

Moreover , a c t ion levels and tolerances tively seek out new data that might alter atend to institutionalize rather than protect prescribed level,

109


.

Tolerances and action levels have otherweaknesses. They are often not easily appliedwhen the environmental contaminant in-volved is a suspected carcinogen (such aspolychlorinated biphenyls (PCBs)). Further-more, the time required to perform a com-plete chemical analysis for contaminants innonprocessed foods such as fish makes its dif-

ficult to prevent some shipments from reach-ing the marketplace.

Finally, there are procedural problems inthe present system. States have no clearly de-fined authority to which they can turn whenthey suspect environmental contamination offood.

OPTION 2-AMEND THE FOOD, DRUG, AND COSMETIC ACT

The Food, Drug, and Cosmetic Act has beenamended several times since its passage in1938 to deal with new food regulatory prob-lems. Environmental contamination of food isnow a national problem which Congress hasnever directly addressed through legislation.Thus, Congress could choose to give regula-tory agencies more guidance by clarifying itsposition on environmental contaminants infood.

An amendment to the Food, Drug, and Cos-metic Act could include one or all of the fol-lowing points. None are mutually exclusive.

Option 2A-SimplifyAdministrative Procedures

Under current law and regulations, theFood and Drug Administration (FDA) sets anaction level for a contaminant soon after aninterstate shipment of contaminated food isdiscovered. FDA will then presumably launchthe elaborate rulemaking proceedings thatculminate in the establishment of a tolerance(under section 701(e) of the Food, Drug, andCosmetic Act). In reality, the costs and delaysinvolved in the complex rulemaking proce-dures now required for the adoption of toler-ances have discouraged FDA from movingfrom the first (action level) to the second (tol-erance) state of the process.

Congress could amend the Food, Drug, andCosmetic Act to simplify the administrativeprocedure through which tolerances are set.The changes could be modeled after section553 of the Administrative Procedures Act.This process involves publication of a toler-

ance proposal in the Federal Register, alongwith (as required by recent court rulings) therationale and factual data underlying theproposal. The public can then respond to theproposal with written comments. FDA mayalso hold legislative-style public hearings toallow presentation of oral arguments andevidence.

After considering all of the comments, FDApublishes a final rule (in this case, a toler-ance). This final rule includes explanation ofany changes from the original proposal andresponses to factual points raised by the pub-lic. Most agencies now use this model forrulemaking. It is a process that can be car-ried out expeditiously with modest investmentof an agency’s resources.

Pros: Adoption of this streamlined rule-making procedure would reduce the time andexpense now involved in setting a formal tol-erance. It may encourage FDA to move fromaction levels to tolerances, thus bringingmore public participation into the process.

Cons: Because action levels are adminis-trative guidelines, they can easily be changedwhen new scientific information becomesavailable. Even if FDA comes to use a simpli-fied procedure to set tolerances, it may stillbe slow to revise them in the light of newdata.

Option 2B--Require theEstablishment of Tolerances

Congress could amend the Food, Drug, andCosmetic Act to require the establishment of

Ch. IX—Congressional Options ● 111

a tolerance within a specif ic t ime after thesetting of an action level.

Pros: This change would encourage FDA togather additional information on a contami-nant’s toxicity and the public’s exposure. Itwould speed up a process. that now operatesunder no deadlines. And it would result in de-finitive tolerances that FDA could enforcewith less concern about judicial questioning.

Cons: This option, however, would substan-tially increase FDA’s workload unless toler-ance-setting procedures were simplified. In-deed, it has been the costs and delays of thecurrent rulemaking process that have de-terred FDA from moving from action levels totolerances. Thus, tolerances should not be re-quired without also simplifying present rule-making procedures,

Option 2C--Clarify the Role ofEconomic Criteria

Congress could amend the Food, Drug, andCosmetic Act to clarify to what extent eco-nomic criteria can be used in setting toler-ances. The Act does not specify that costs(the adverse economic effects) of a proposedtolerance be considered when setting a toler-ance. FDA, in practice, does weigh the cost offood lost when establishing a tolerance.

The Food, Drug, and Cosmetic Act could beamended to prohibit FDA from consideringcosts when setting a tolerance. Prohibitingany economic assessment would ensure thatpublic health would be the first priority in set-ting a tolerance.

Conversely, the Act could be amended torequire FDA to weigh the costs against thebenefits of a proposed tolerance. RequiringFDA to evaluate the economic consequencesof a tolerance would give FDA clear authorityto weigh such estimated effects together withthe potential health risks when establishing atolerance. Congress could require FDA togauge only the primary costs (as is now donewith food lost) or all associated costs (foodlost, employment impacts, distributional andindirect effects) for a proposed tolerance.The techniques available for estimating costs

require up to a year for generating the neces-sary data. Thus, weighing the costs is bestsuited for setting tolerances. not actionlevels.

The advantage in including costs in toler-ance-set t ing decisions is that adverse eco-nomic impacts are likely to be reduced. Thedisadvantage is that tolerance levels are like-ly to be higher than would be the case if costswere not considered.

Pros: By clearly defining to what extentcosts can enter into tolerance-setting deci-sions for environmental contaminants in food,Congress would eliminate ambiguities of in-terpretation and provide clear guidance toFDA.

Cons: FDA has interpreted the Food, Drug,and Cosmetic Act as allowing cost considera-tions, Legislation requiring economic assess-ment could limit FDA’s discretion to weighthe costs of food lost if it judges the situationwarrants such a treatment.

Option 2D--EstabIish RegionalTolerances

This option would give FDA the flexibilityto set different action levels or tolerances fordifferent regions, based on expected levels ofexposure, regional levels of contamination,and eating patterns.

Pros: Action levels and tolerances may notbe set low enough to protect those popula-tions that are most highly exposed, previouslyexposed, or most vulnerable. States may notexercise their authority to set toleranceswhich are more restrictive than the Federaltolerance because of budget limitations, in-adequate information, or political pressures.FDA can provide guidance to States and sug-gest more restrictive tolerances or warningsto the public, but FDA has no authority to in-tervene if the contaminated food does notenter into interstate commerce.

Cons: Regional tolerances would compli-cate monitoring and enforcement programs.Regional tolerances might also be viewed asFederal infringement on State authority.


OPTION 3-ESTABLISH AN INVESTIGATORYMONITORING SYSTEM

Environmental contaminants could be de-tected earlier in the food chain by improvingpresent environmental monitoring capabil-ities —establishing an investigatory monitor-ing system while maintaining current regu-latory monitoring programs.

Congress could set up a national investi-gatory monitoring system that monitors foreither suspected or uncharacterized environ-mental contaminants. A system combiningelements of both approaches could also be es-tablished. Since any of these monitoring ap-proaches would require some research anddevelopment before a fully operational sys-tem could be devised, Congress could chooseto create a pilot program. Such a programwould spur research and development andassess the feasibility and cost-effectivenessof the various approaches.

The investigatory monitoring systems dis-cussed in this assessment would call for dif-ferent sampling and quality control proce-dures. The development of these proceduresis as important as the development of moni-toring technology (some of it still in the ex-perimental stage). Indeed, there is no com-prehensive investigatory monitoring systemfor toxic substances in food and the environ-ment at any level of government.

Consequent ly , Congress might opt for apilot project to assess the capabilities and re-source requirements of various national mon-itoring systems instead of mandating a par-t i cu l a r mon i to r i ng app roach . Such a p i l o tproject would focus on the monitoring of or-ganic chemicals , inorganic, and radioact ivesubstances. I t would determine the technol-ogy, sampling, and quality control needs formonitoring these three toxic substance cate-gories in food, air, water, and soil. The broad-er purpose of the project would be to developa c o m p r e h e n s i v e m o n i t o r i n g p r o g r a m t h a tw o u l d m e e t t h e G o v e r n m e n t ’ s r e g u l a t o r yneeds, provide data to make cost-effect ive-ness assessments of the al ternative strate-

gies, and reduce public exposures to environ-mental contaminants as much as possible.

Option 3A-Establish a NationalMonitoring System for Suspected

Environmental Contaminants

Suspected environmental contaminantsare substances that are most likely to enterthe food supply and pose potential healthhazards, Lists of such substances could bedrawn up for organic chemicals, trace met-als, and radioactive substances in order thatthey be monitored in the food, Various cri-teria such as toxicity, volume of production,occurrence in the environment, persistence,and biodegradability could be used in puttingtogether the lists.

Pros: Present food monitoring efforts arenot designed to detect new environmentalcontaminants in food. The limited amount ofmonitoring for suspected contaminants thatdoes exist is primarily concerned with tracemetals. But far more of this type of monitoringis needed to anticipate new contaminants infood.

Cons: To draw up such lists successfully,considerable information is needed. Sub-stances for which there is little or no datawould automatically be given low priority. Inlarge part, the makeup of the lists would de-pend on scientific judgments. However, scien-tific judgments often vary (or even conflict).Thus, the reliability of the lists may be inquestion,

Priority lists of trace metals and radioac-tive substances, which are limited in numberand already well-investigated, would be morereliable than a list of organic compounds.There are thousands of such organic agentsmanufactured, And there is little toxicologi-cal or environmental information available ona great many of them.

If such lists were developed, the number ofsubstances monitored would necessarily be

Ch. IX—Congressional Options c 113

limited. Standards would be set up for deter-mining which substances would get priority.Of course, there would be no certainty thatunl is ted substances might not get into thefood supply and threaten human health. Thecost of such a monitoring system would de-pend on how many substances were beingtraced as well as the expenses of equippingand staffing laboratories.

Option 3B--Establish a NationalMonitoring System for

Uncharacterized EnvironmentalContaminants

Uncharacterized monitoring would be de-signed to detect substances that are: lackingtoxicity data for potential human risk, notknown to be present in food, and not evenknown to exist in the environment. This kindof monitoring is most needed for syntheticorganic chemicals. The purpose is to detectchanges in the levels of various syntheticorganics in environmental or food samplesover time,

Scientists would not necessarily know theidentity of individual substances they weremonitoring. But if the concentration of a par-ticular compound substantially increased,they could analyze it further to establish itsidentity, The literature would be reviewed onthe substances toxicological properties. Orperhaps the substance would undergo toxico-logical testing. Depending on what informa-

tion is developed, regulatory agencies couldthen take appropriate action.

This approach tries to create a mechanismfor quantitatively measuring uncharacter-ized substances in food. Proper guidelines arenecessary since the cost of quantitativelyidentifying one substance can range from$10,000 to $100,000. Moreover, this monitor-ing approach requires sophisticated equip-ment for analyzing food samples. Also, the in-formation generated by the analyses has tobe computerized. Computer technology makesit possible to correlate and interpret chemicaldata to provide a continuous surveillance ofthe levels in food.

Pros: The uncharacterized monitoring ap-proach is in the research stage at several lab-oratories in the country. If it is successfullydeveloped, this type of monitoring would re-duce the time during which the public is ex-posed to high concentrations of uncharacter-ized environmental contaminants in food. Ke-pone, for example, was polluting the JamesRiver, and people were eating kepone-con-taminated fish for several years before thechemicals presence was discovered. With anuncharacterized monitoring system, keponemay have been detected years earlier.

Cons: This combination of sophisticated in-struments, dependence on computers, andhighly trained personnel is expensive. Andthe “hardware” needed is generally notfound in Federal or State monitoring labora-tories.

OPTION 4--IMPR0VE FEDERAL RESPONSE TONEW CONTAMINATION INCIDENTS

All of the major food contamination inci- ance, Congress could choose to designate adents have been marked by confusion. This lead agency or establish a center for the col-stems from the involvement of three Federal lection and analysis of data.agencies—FDA, the U.S. Department of Agri-culture (USDA), and the Environmental Pro- Option 4A-Designate a Lead Agencytection Agency (EPA)—in the monitoring andregulation of environmental contaminants in The problem of conflicting Federal assist-food. To cut down on confusion and to im- ance efforts was dramatized by several re-prove del ivery of Federal technical assis t- cent incidents including the most recent PCB

114 ● Environmental Contaminants in Food *

contamination. This sort of situation is notunique, Official reactions following the poly-brominated biphenyl (PBB) episode in Michi-gan and the kepone incident in Virginia weresimilar.

The lead agency would serve as a clearing-house for all information coming from and go-ing to States. FDA would be the most likelycandidate for lead agency when food contam-ination is suspected.

Pros: With a clearly delineated lead agen-cy, States suspecting contamination of foodwould have one reliable source of technicalassistance. Conflicts of opinion would be set-tled internally, and public statements andtechnical assistance provided with less am-biguity.

Cons: Designation of a lead agency mightdecrease the amount of technical expertisemade available to the States. It would requirethe lead agency to develop new agreementswith the other two agencies, During a foodcontamination crisis the lead agency wouldneed to coordinate responses from the otheragencies.

Option 4B--Establish a Center toCollect and Analyze Toxic

Substances Data

Major delays in protecting the public fromenvironmental contaminants in food now re-sult from the time-consuming process of gen-erating sufficient data on a substance’s toxic-ity and dispersal in the food supply. Congresscould overcome this problem by setting up anew technical center.

Such a center would be able to rapidly as-semble technical teams skilled in the identifi-cation and analysis of organic, inorganic, andradioactive substances. A team would consistof a multidisciplinary group of experts, in-cluding chemists, toxicologists, food and ani-mal scientists, epidemiologists, biochemists,biostatisticians, medical doctors, and others.Its mission would be to identify the cause ofan actual or potential contamination incident,and assess the possible environmental and

human health impacts. It would have no regu-latory function.

The group would be able to mobilize within24 hours. It would be the lead Federal organi-zation that affected States could initially con-tact. In the wake of an episode, the teamcould be responsible for followup scientificresearch that would lay the groundwork forepidemiological studies of the exposed popu-lation. It would also be able to conduct arange of short-term toxicological tests to de-termine the mutagenicity and potential car-cinogenicity of uncharacterized substances.It would be able to examine an exposed popu-lation for any adverse health effects from in-gesting a particular contaminant in food.

The new technical center, in essence,would be similar to the Center for DiseaseControl (CDC). It would have the same capa-bilities in chemical epidemiology as CDC hasin infectious disease. The new center couldbe given responsibility for investigating alltoxic substance problems in the UnitedStates, not just those limited to food. If it wereplaced under CDC, it could build on work nowunderway in that organization’s Environ-mental Hazards Unit. Furthermore, the cen-ter would be able to capitalize on the long-standing relationships between CDC and theStates. Since the center’s sole mission wouldbe the development of information on environ-mental contaminants, it would not be linkedto regulatory decisions.

As an alternative, the center might be lo-cated within FDA’s National Center for Toxi-cological Research or another Federal labor-atory with existing analytical capabilities. Itcould then use the laboratory’s equipmentand trained personnel for investigative ana-lytical chemistry and toxicology.

It would be better not to locate the centerand the investigation teams in a regulatoryagency with direct responsibilities for food.This would ensure that there is no possibleconflict of interest between its job of factfind-ing and the agency’s responsibility for regu-lating, By distancing itself from the regula-tory process, the new center team would

Ch. IX—Congressional Options ● 115

strengthen its credibility with the media andthe public.

Pros: Accurate scientific information onthe nature and extent of food contaminationhas to be generated quickly when an incidentoccurs. Individual States often lack the scien-tific expertise to develop such data. The Fed-eral Government does have the necessary ca-pabilities, but the resources are scattered inseveral agencies with differing areas of au-thority. Such a dispersal of expertise hindersthe gathering of urgently needed informationfollowing a contamination episode. Jurisdic-tional problems crop up, and States find theyhave to deal with more than one agency. Theresulting delays slow the making of necessaryhealth, environmental, and regulatory deci-sions.

Reaction time would be shortened and dup-lication of effort reduced by a technical ortox i c subs t ances i nves t i ga t i on t eam wh ichcould react quickly in emergencies.

Cons: There is no assurance that a specialteam could generate information more quick-ly than is now the case. If the center were notpart of a regulatory agency, it might not onlygrow out of touch with the needs but a lsocould duplicate the investigatory work of theagency . Fu r the rmore , t he po t en t i a l f o r anadversary relationship exists.

It could be argued that better coordinationamong FDA, USDA, and EPA would accom-plish the same goals without the expense ofestablishing a new research center.

Appendixes

Appendix A

Substances Whose Production orEnvironmental Release Are Likelyto Increase in the Next 10 Years*

by Clement Associates, Inc.

OTA requested Clement Associates to develop a list identifying new chemical sub-stances likely to be manufactured in the near future and known chemicals likely to posean increased burden to the environment because of increased production, new applica-tions, or new technological developments, including new energy technology. For theseprojections, several factors were considered, including market trends, Federal regula-tory activity, available substitutes for recently banned or restricted chemical substances,and consumer needs.

During the development of the approach to this phase of the project, certain prob-lems and limitations became apparent. The nature of chemical substances under re-search and development but not yet introduced to the market is usually closely guardedproprietary information and therefore not available. In addition, there are no data sys-tems which bring together chemical information to facilitate the retrieval of necessarydata. An approach was developed to obtain a maximum amount of information in a lim-ited amount of time.

Sources, including personal contacts, were identified for information on new chemi-cals or chemicals whose production, use, and release to the environment -- -

crease sharply because of future needs. These sources of informationreference list.

Following arewere selected.

the list of chemicals and a brief indication of why

LIKELY TO HAVE SIGNIFICANTLY

were likely to in-are listed in the

these chemicals

CHEMICALSENVIRONMENTAL RELEASE IN THE NEAR FUTURE

Chemicals Whose Production and UseAre Likely to Increase Sharply

Because of Future Needs

SoIuble PolymersSoluble polymer “complet ing” agents are

being developed to remove toxic metals fromwaste waters or remove radioactive metals fromnuclear waste fluids. Polyethylene glycol and its

*Excerpt from OTA Working Paper entitled “Priority Settingof Toxic Substances for Guiding Monitoring Programs. ” A com-plete copy of the paper can be obtained through the NationalTechnical Information Service. (See app. 1.)

derivatives are the most versatile of the solublepolymers. Others are polyethyleneimine, polyvi-nyl sulfonic acid, polyacrylic acid with such che-lating groups as thiourea, 8-hydroxyquinoline,iminodiacetic acid and hydroxyciniline, and poly-mers based on acrylic acid. (C&EN, Mar. 27, 1978,p. 24)

The principal organic flocculants are polyac-rylamides, polyamides, and polyepichlorohydrins.Stiffening waste treatment regulations to reducesludge are making flocculants more attractive.The chemicals make up a $100 million per year

119

120 c Environmental Contaminants in Food

market that may double within the next 5 years.These polymers may also be used for enhanced oilrecovery if the price of oil would rise to makeenhanced recovery economical. (C&EN, Jan. 23,1978, p. 9)

Multifunctional AcryiatesThe use of radiation curing (ultraviolet or elec-

tron beam) is rapidly increasing. Radiation curingcontributes significantly to critical energy sav-ings and pollution abatement. Demand for the fol-lowing multifunctional acrylates used for ultra-violet inks and coatings is expected to increase:pentaerythritol triacrylate, 1,6-hexanedioldiacry-late, trimethylolpropane triacrylate, and tetraeth-ylene glycol. (C&En, Jan. 23, 1978, p. 12, CelaneseChemical Co. Advertisement)

Polyester ResinConcern about energy will augment the growth

of lightweight polyester resin in the United Statesthrough the 1980’s. The need for lighter weightmaterials in the automotive market, as well as thedemand for corrosion-resistant products in manyareas, will increase the demand for polyesterresin at an average of 9.2 percent a year through1987. Polybutylene terephthalate (PBT) and poly-ethylene terephthalate (PET) will be among thefastest growing polyester resins. PBT and PET areexpected to grow from a combined demand of 38million lbs in 1977 to 322 million lbs by 1987—anaverage annual growth rate of 24 percent. PBT iscurrently the most widely used thermoplastic pol-yester resin. However, it is expected that PET pro-duction will increase more rapidly, overtakingPBT production. The total polyester resin marketis expected to rise 140 percent in the next 10years. other uses and average annual projectedproduction increases from 1977 to 1987 includeunsaturated reinforced polyester (9. 1 percent),thermoset surface coatings (5 percent), culturedmarble and other unsaturated thermoses (4.0percent), strapping (mostly scrap, 28 percent),and fibers (4.0 percent). (C&EN Jan. 30, 1978, p.11)

ZeolitesZeolites (aluminosilicates) hold promise as de-

tergent builders, A 25-million-lb market in 1977has a prospect of a possible 400-million-lb marketin 1982. Zeolites containing detergents perform“equivalent or roughly equivalent” to the phos-phate-silicate formulations that now command 75percent of the current heavy-duty powder market.Environmental problems will continue to push

phosphate out of the heavy-duty home laundrymarket. Zeolite 4A is a cubic crystalline sodiumaluminosilicate with the formula

Textile ChemicalsThe market for textile chemicals is expected to

top the billion-dollar market by 1980. Examplesare:

Chemical UseBiphenyl . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dye batho-phenylphenol. . , . , . . . . . . . . . . . . . . additives1,2,4-trichlorobenzene. . . . . . . . . . . . . . .Methyl naphthalene. . . . . . . . . . . . . . . . . . . .Perchloroethylene ., , . . . . . . . . . . . . . . . . . .Alkyl aryl sulfonic acids . . . . . . . . . . . . . . . .

Ethoxylated alcohols , , , ., ., ., . . . . . . . . . .Quaternary ammonium compounds . . . . . . .Ethoxylated nonyl phenol, . . . . . . . . . . . . . . .Alcohols of alkyl aryl sulfonic acid . . . . . . . .Esters of phosphoric acid. . . . . . . . . . . . . . . .

Acrylic latex , . . . . . . . . . . . . . . . . . . . . . . . Finishing agentsStyrene-butadiene rubber latex . . . . . . . . . .Polyvinyl alcohol , ... , . . . . . . . . . . . . . . . . .Polyvinyl acetate , ., ., . . . . . . . . . . . . . . . . .Melamine formaldehyde . . . . . . . . . . . . . . . .Starches. . . . . . . . . . . . . . . . . . . . . . . . . . . .Polyvinylchloride . . . ., ... , ., . . . . . . . . . .Fluorochemicals . . . . . . . . . . . . . . . . . . . . . .Glyoxal (dimethylol, dihydroxyethylene urea)Carbamate (isobutyl, 2-methoxyethyl). . . . . .Tetrakis phosphonium sulfate ‘ammonia . .Cyclic phosphonates . . . . . . . . . . . . , .

Maleic anhydride. . . . . . . . . . . . . . . . . . . . . . PrintingPolyacrylic acids . . . . . . . . . . . . . . . . . . . . chemicalsAcrylates . . . . . . . . . . . . . . . . . . . . . . . . .Butadiene acrylonitrile . . . . . . . . . . . . . . . . .Hexamethylol melamine formaldehyde. . .Dioctyl sulfosuccinates . . . . . . . . . . . . . . .Ethoxylates . . . . . .

Pentachlorophenon . . . . . . Bacteriocideso-phenylphenol. . . , . . . . . . . . . . . . . . .Low (L and aromatic distillates. . . . . . . . . . .Urea . . . . . . . . . . . . . . . . . . . . . . . . . . .

Ethoxylated alkyl phenols . . . . . . . . [dispersingSodium lauryl sulfate. . . . . . . . . . . . agents/mulsifiers

((;&EN, May 29, 1978, p. 12)

PumicePumice has many similarities with asbestos.

They are both mineral oxides, low in density, heatresistant, nonflammable, chemically inert, andlow in cost, Rhodes (Division of Beatrice FoodsCo., Des Plains, Ill.) claims pumice may be the safeand economical alternative to asbestos (C&EN,May 22, 1978, p. 9). The manufacturer suggeststhat pumice be used in paints; chemicals (filtra-tion media and chemical carrier); leather buffing:compounders (powdered hand soaps and glasscleaners); metal and plastic finishing; and ap-

Append/x A—Substances Whose Production or Environmental Release Are Likely to Increase in the Next 10 Years ● 121

placations in the dental, rubber, glass, furniture,electronics, and pottery industries.

PolyvinylacetatesAcrylic and acetate resins compete in the mar-

ket as textile and binder emulsions for nonwovenfabrics. Currently, acrylics dominate both ofthese markets by two-thirds of the total, Polyvinyl-acetates make up only 14 percent of the market,with the remainder going to other resins. Newtechnical developments may change the statusquo. By 1987, acetates could surpass acrylics inmany quality paint and textile markets. (C&EN,Mar, 20, 1978, p. 11)

Hydrazine and Its DerivativesOriginally used in rocket fuels, today hydrazine

and its derivatives are commercially used in her-bicides, pesticides, blowing agents for plastics,and water treatments. Its consumption is growingby 15 percent per year according to Olin Chemi-cals’ advertisement in C&EN.

Olefinic Thermoplastic ElastomersOlefinic thermoplastic elastomers are also

called TPO rubbers and were introduced in 1973,having a market volume of 1.5 million lbs. It is an-ticipated that the demand for TPO rubber willreach 44 million lbs by 1980 and that uses in auto-mobiles will account for more than half of it.Mechanical goods and wire and cable uses ac-count for 12.7 and 18,2 percent of TPO demand,respectively. Olefinic thermoplastic elastomersare also used by carmakers in electrostaticallypaintable body filler panels, air deflectors, andstone shields. Recently they have been used assound deadening material on diesel-powered vehi-cles, jacketing and insulating material for wireand cable coating, and many custom-molded andextruded mechanical goods. (C&EN, Oct. 23, 1978,p. 9)

SiliconesNew markets for silicones as PCB replacements

in electrical transformers, in brake fluids, and aselastomers are developing. Methylchloride isused as an intermediate in the production of sili-cones, Other methylating agents could be used toreplace methyl chloride in most applications, butthe potential substitutes (e.g., dimethyl sulfate,methyl bromide, methyl iodide) are much more ex-pensive and have toxicity and handling problemswhich make them less desirable.

ChlorobrominationBromine chloride has been shown to be more ef-

fective as a disinfectant than chlorine in field tri-als. It is a heavy red liquid and 12 times more sol-uble in water than chlorine. Its vapor pressure isone-third as great. Bromine chloride completelyhydrolyzes almost immediately to hypobromousacid and reacts with ammonia in sewage to formbromoamines. It is believed that chlorobromina-tion is the best alternative to current chlorinationpractices.

PolyethyleneterephthalateDuPont is continuing a vigorous effort to devel-

op a market for molded plastics made from con-ventional forms of thermoplastic polyethylene ter-ephthalate. (See also Polyester resin, ) This resinhas a multibillion-pound demand as polyesterfibers and films and is finding a large market inplastic beverage bottles. DuPont’s trade name forPET is Rynite. Rynite’s prime targets will bemetals replacement, especially in automobiles.

Chemicals Produced or Released Dueto Energy Development Technology

Coal Liquefication andGasification Program

The conversion of coal to liquid to gaseous hy-drocarbons for fuel technologies will result in therelease of many chemicals to the environment.The contaminants may enter the environment viatwo pathways: 1) emission into the atmospherewith the consequent potential for long-rangetransport and 2) direct discharge via runoff andleaching into the aquatic and terrestrial domainwhere impact might be expected to be more local-ized. Contaminants from coal combustion can beclassified into three groups: 1) organic chemicals,2) inorganic chemicals, and 3) radionuclides.

Organic ContaminantsOrganic contaminants from coal-derived proc-

into several categories as


Hetrocyclics . . . . . . . . . . .

Simple aromatichydrocarbons ., ., . . . .

Phenols. . . . . . . . . . . . . . .

Polycyclic aromatichydrocarbons . .

Sulfur compounds

Organometallics. ,

Naphthyl cyanides

. . . . . .

. . . . . .

., ...,

,. ...,

Trapped organic

PyridinesQuinolines

BenzeneTolueneXylenesPhenolsCresolsXylanols

Benzo(a)pyreneDibenzofluoreneDibenzoanthraceneBenzoanthracenesBenzo(a)anthroneDimethylbenzoanthraceneChryseneMethylchrysenesBenzocarbozolesIdenopyrenesCarbozolesPyrenesBiphenylAcenapthaleneAcenapthalyneFluoreneAlkylanthracenesAlkylphenanthracenesAnthracenePeryleneBenzoperyleneCoroneneThiophenesMercaptansCarbon disulfideMethyl thipheneNickel carbonylTetraethyl leadNapthyl cyanideAmmonium thiocyanate

compounds and aromaticunits in coals that were isolated, separated, andidentified by gas chromatography and mass spec-trometry are shown in table A-1.

Inorganic ChemicaIsCategoryAcids . . . . . . . . . .

Chromium salts . .

Sulfur compounds

Trace elements. . .

.. .,,,

. . . . . .

. . . . . .

, , . . ! .

ExamplesSulfuric acidNitric acidHydrochloric acidChromium chlorideChromium sulfideHydrogen sulfideCarbon disulfideAntimony ManganeseArsenic MercuryBarium MolybdenumBeryllium NickelBismuth SeleniumCadmium SilverChromium TelluriumCobalt ThalliumCopper TinFluorine UraniumGallium VanadiumLead Zinc

Fine particulate , . . . . . .

Gasses . . . . . . . . . . . . . . .

RadionuciidesEmissions from a

Sulfur particulateRespirable coal dustsTarsootsCarbon monoxideSulfur dioxideSulfur trioxideSulfur tetraoxideNitrous oxide

100-MW electricity genera-tion powerplant that burns coal at a rate of ap-proximately 100 tons per hour are estimated tocontain 1 ppm of uranium and 2 ppm of thorium, Aplant of this size is expected to release 1 percentof its fly ash to the atmosphere. Under these con-ditions thorium-228 and -232, radium-224, andlead-212 each contribute approximately 5 x 10-’Ci per year; uranium-234 and -238, thorium-230and -234, radium-226, lead-210, polonium-210,and bismuth-2 10 each contribute approximately 8x 10-:~ Ci per year; uranium-235 and praseo-

dymium-231 both contribute approximately 3.5 x10-4 Ci per year; and radon-220 and -222 togetheraccount for approximately 1.2 Ci per year. (ERDAreport 77-64, August 1976)

Radioactive emissions of bituminous coals havea high uranium content (75 ppm).

Solar Heating and Coolingof Buildings

Water contamination can occur in solar-heateddomestic hot water systems at heat exchanger in-terfaces. Serious health consequences could beexpected if the contaminated water is ingested.Water contamination could result from the heatexchange fluids themselves, or in water-basedsystems from such additives as:

Corrosion inhibitors —Chromates, berates, ni-trates, nitrites, sulfates, sulfites, arsenates, ben-zoate salts, various triazoles, silicates, and phos-phate compounds.

Freeze protestants—Glycols.Heat transfer fluids —Paraffins, aromatic and

other synthetic hydrocarbons.Bacteriacides— Chlorinated phenols.Solar collectors used in heating and cooling

systems utilize organic chemical compounds as in-sulators that can emit highly toxic substancesunder overheat or fire conditions. Fumes usuallyconsist of simple starches and phenolic com-pounds: ammonia, hydrochloric acid, hydrofluoricacid, toluene diisocyanate (TOI), and hydrogencyanide.

Appendix A—Substances Whose Production or Environmental Release Are Likely to Increase in the Next 10 Years ● 123

Table A-1 .—Trapped Organic Compounds and Aromatic Units in Coal

4546474849505152.

5354

55565758596061.

6263

646566676869

707172737475

76

77

78

798081

82

B = branched T = identlflcatlon tentative ~ = identlflcatlon uncertain

Fuels From Biomass

Thermochemical biomass conversion can pro-duce gases, tars, oils, and unconverted residue(char) and ash, depending on the particular con-version process. Thermochemical reactions gen-erated sulfur-containing (H2S, COS, CS2, SOX) andnitrogen-containing (HCN, NOX, NH 3) gases.

83848586

8788

899091

92

93

94

9596

97

9899.

100101

102103104105106

107

108

Water can be affected by the residuals producedfrom thermochemical conversion. Low-molecularweight oils, phenols, leachates from char and ashresidues, and scrubber solution runoff may enterwater bodies by direct discharge or by percola-t ion to subsurface waters from evaporat ionpoints.


REFERENCE LIST FOR SUBSTANCES WHOSE PRODUCTION ORENVIRONMENTAL RELEASE ARE LIKELY TO INCREASE

IN THE NEXT

Chemical Marketing and Economics Abstracts.The Division of Chemical Marketing and Econom-ics of the American Chemical Society (ACS) pre-sents papers at ACS national meetings on sub-jects related to the responses of the chemical in-dustry to economic changes as well as responsesof the financial community to changes in thechemical industry. Abstracts of these papers arepublished by ACS.

A Study of Industrial Data on Candidates forTesting. This document, published by the U.S.Environmental Protection Agency, Office of ToxicSubstances (EPA Contract No. 68-01-4109, No-vember 1976) contains market forecasts for 10major classes of chemicals (109 individual chemi-cals) with an annual production greater than orequal to 1 million lbs. Chemicals that are used ex-clusively as drugs or pesticides are not included.The market forecasts include: 1) a discussion ofproduction and trade statistics, 2) consumptionpatterns, whenever possible, 3) growth trends, 4)a brief summary of current uses as well as poten-tial new applications, and 5) growth trends in end-market consumption. A discussion of possible sub-stitutes for some of the chemicals is also included.

Environmental Development Plans. The Envi-ronmental Development Plans (EDPs) published bythe U.S. Department of Energy (March 1978) wereconceived and prepared as basic documents toassist in planning and managing environmentalprograms of energy technology development. Ap-proximately 30 EDPs covering major developingenergy technologies were prepared.

A Review of Current Information on Some Eco-logical and Health-Related Aspects to the Releaseof Trace Metals Into the Environment AssociatedWith the Combustion of Coal. This document byMerrill Heit is a technical report (HASL-320) from

10 YEARS

the Health and Safety Laboratory, Energy Re-search and Development Administration, NewYork, N.Y. 10014, reviewing the literature on oneclass of pollutants. Information on the environ-mental levels, ecological effects, and potentialtoxicity to man of 35 elements that may be re-leased into the environment by coal combustion orgasification is presented.

Fossil Energy Update. This monthly journalcompiled by the Department of Energy lists ab-stracts of current scientific and technical reports,journal articles, conference proceedings, theses,and monographs on all aspects of fossil energy, in-cluding factors involving the environment, health,and safety.

Chemical Engineering News. The Chemical En-gineering News is a weekly publication of theAmerican Chemical Society that contains rele-vant information in such sections as “ChemicalWorld, “ “This Week,” “ Business Concentrates, ”and “Science/Technology,” and in profiles onselected chemicals.

Chemical Marketing Reporter. This weekly,published by Schnell Publishing Company, Inc.,contains information on chemical market reportsand profiles on selected chemicals.

Trapped Organic Compounds and AromaticUnits in Coal. This article, published by TyoichiHayatsu et al. in Fuel 57:541 (1978), contains adetailed analysis of the organic constituents ofthree coals: a lignite, a bituminous, and an an-thracite. Organic compounds trapped in the coalmatrix, residuals, and products of the originalcoalification process were described.

Personal Communications With Several Orga-nizations, Including the Chemical DevelopmentAssociation and Chemical Marketing ResearchAssociation.

Appendix B

FridayJune 29, 1979

Part XII

Department ofHealth, Education,and WelfareFood and Drug Administration

Polychlorinated Biphenyls (PCB’s);Reduction of Tolerances

125


30330 Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations

DEPARTMENT OF HEALTH,EDUCATION, AND WELFARE

Food and Drug Administration

21 CFR Part 109

[Docket No. 77N-0080]

Polychlorinated Biphenyls (PCB’s);Reduction of Tolerances

AGENCY: Food and Drug Administration.ACTION: Final Rule.

SUMMARY : The Food and DrugAdministration [FDA) is reducing thetolerances for unavoidable residues ofthe industrial chemicals polychlorinatedbiphenyls (PCB’s) in several classes offood. Specifically, the agency is reducingthe tolerances in milk and dairyproducts from 2.5 parts per million (ppm)to 1.5 ppm (fat basis), in poultry from 5ppm to 3 ppm (fat basis), in eggs from 0.5ppm to 0.3 ppm, and in fish and shellfishfrom 5 ppm to 2 ppm.DATES: Effective August 28, 1979;objections on or before July 30, 1979.ADDRESS: Written objections to theHearing Clerk (HFA-305), Food andDrug Administration, Rm. 4-85, 5000Fishers Lane, Rockville, MD 20857.FOR FURTHER INFORMATION CONTACT:Howard N. Pippin, Bureau of Foods(HFF-312), Food and DrugAdministration, Department of Health,Education, and Welfare, 200 C St. SW.,Washington, DC 20204,202-24$3092.SUPPLEMENTARY INFORMATION : In theFederal Register of April 1,1977 [42 FR17487), FDA proposed to reduce thetemporary tolerances for unavoidableresidues of PCB’s in several classes offood. The agency received over 100comments on the proposal frominterested individuals, consumer groups,businesses, trade associations, Stategovernment agencies, and others. Theagency has considered these commentsand is now issuing a final order reducingthe PCB tolerances as originallyproposed. Following a brief discussionof the background, this document willrespond to the comments the agencyreceived and explain the agency’sreasons for adopting the reducedtolerance levels.

I. BackgroundPCB’s are a class of toxic industrial

chemicals that have become persistentand ubiquitous environmentalcontaminants as a result of pastwidespread, uncontrolled industrial use,As explained in the preamble to FDA’sproposal initiating this rulemaking

proceeding (see 42 FR 17489), one resultof PCB contamination of theenvironment has been contamination ofcertain foods. In the Federal Register ofJuly 6,1973 (38 FR 18098), FDA issuedregulations to deal with the problem ofPCB contamination of food. Amongthose regulations was one establishingtemporary tolerances for unavoidablePCB residues in various categories offood. Those original tolerances are nowcodified in $109.30 (21 CFR 109,30). Theorder FDA is issuing in this documentreduces certain of those tolerances.

FDA’s authority to issue tolerances forunavoidable food contaminants isderived from sections 402(a)(2)(A) and408 of the Federal Food, Drug, andCosmetic Act (the act) (21 U.S.C.342(a)(2)(A) and 346). Section402(a)(2)(A) deems food adulterated,and thus prohibited from interstatecommerce, if it contains “any addedpoisonous or added deleterioussubstance” that is unsafe within themeaning of section 4080 Section 408deems any added poisonous ordeleterious substance to be unsafeunless its presence in the food isrequired in the production thereof orcannot be avoided by goodmanufacturing practice. Section 408 alsoauthorizes the agency to promulgateregulations limiting the quantity of sucha required or unavoidable substancethat can be present legally in food. Suchlimits, called tolerances, are to be set byFDA at the level found necessary toprotect the public health, taking intoaccount the extent to which the .substance is required or unavoidableand the other ways the consumer maybe affected by the same or otherpoisonous or deleterious substances,Once a regulation establishing atolerance has been promulgated for aparticular poisonous or deleterioussubstance, food containing thatsubstance in an amount exceeding thetolerance is deemed adulterated undersection 402(a) (2)( A].

One of the primary purposes ofsection 408 of the act is to enable FDAto deal effectively with environmentalcontaminants such as PCB’s, Thesesubstances often enter food as a resultof events beyond the reasonable controlof the food manufacturer or processorand, once in the food, usually cannot beremoved by good manufacturingpractice. For example, in the case ofPCB’s, some species of fish have becomecontaminated to varying degrees as aresult of the dumping of PCB-containingindustrial waste into the nation’s waters(see the Federal Register of March 18,1972 (37 FR 5705)]. Once thecontamination occurs, there is little that

can be done to remove the PCB’s fromthe water or from the fish; their presenceis, in that sense, unavoidable. Becausethe initial contamination of fish withPCB’s cannot be avoided (nor the PCB’sprocessed out), the only way to avoidPCB’s in fish is to remove fish fromcommerce if it contains PCB’s above agiven tolerance level. The degree ofavoidance accomplished by this methodis, of course, a function of the level atwhich the tolerance is set, In this way, itis theoretically possible to avoid PCB’sin fish absolutely by removing fromcommerce all fish that contain anyamount of PCB’s.

Section 408 of the act authorizes FDAto make a practical judgment in dealingwith such environmental contaminants:Based on an assessment of the degree towhich the contaminant poses a threat toconsumers, the agency can decide totolerate the contaminant’s presence infood up to a level the agency considersappropriate to protect the public health,taking into account, among other factors,the extent to which the presence of thecontaminant is unavoidable, In makingthis judgment, the agency’s paramountconcern is protection of the publichealth: The tolerance cannot be setabove the level. the agency findsnecessary to protect the public healthadequately. But in determining whattolerance level provides an adequatedegree of public health protection, FDAis required by section A08 to considerthe extent of unavoidability—in the caseof PCB contamination of fish, theamount of PCB-contaminated fish thatmust be disposed of to reduce humanexposure to PCB’s to a tolerable level,As a practical matter, of course, atolerance, if it is to be enforceable,cannot be set below the level at whichthe contaminant can be reliablymeasured for enforcement purposes byavailable analytical methods.

The toxicological data available onPCB's make it clear that, in an idealsituation, it would be preferable not tohave PCB’s in food at any level. Asdiscussed more fully below, the data donot permit the identification of any levelof PCB exposure that can be said toprovide an absolute assurance of safety,It is equally clear, however, that thereduction of PCB exposure from foodsources to zero, or to a levelapproaching zero, would requireelimination of large amounts of food,especially fish. Hence, in deciding theappropriate levels for PCB tolerancesunder section 408, FDA has had to makesome extraordinarily difficult judgments.It has had to decide, in effect, where theproper balance lies between providingan adequate degree of public health

Append/x B— FDA Derivation of PCB Tolerances— Federal Register ● 127

Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations 38331

protection and avoiding excessive lossesof food to American consumers.

The comments received on theproposal reveal that by far the mostcontroversial aspect of this rulemakingproceeding is the balancing judgmentFDA made in proposing to reduce thePCB tolerance in fish from 5 to 2 ppm.Some comments argued that theproposed reduction would cause anexcessive loss of food and significantadverse economic impact withoutproviding any significant increase inpublic health protection. Othercomments argued the converse, i.e., thatthe proposed reduction to 2 ppm wouldnot adequately protect the public healthand that the tolerance should bereduced to 1 ppm (the lowest level atwhich PCB residues in fish can bereliably measured fop enforcementpurposes), despite the additional lossesof food that a reduction to 1 ppm wouldcause. In each case, the commentsbolstered their arguments by contendingthat FDA has either overestimated orunderestimated the toxicity of PCB’s andthe impact of the proposed reduction ofthe tolerance in terms of food loss andadverse economic consequences.

The comments criticizing the proposedreduction of the fish tolerance highlightthe difficulty of the judgment FDA mustsometimes make in establishingtolerances. Not only must FDA make aqualitative judgment about the properbalance between adequate public healthprotection and excessive loss of food, italso must often make that judgment onthe basis of data that are incomplete, oreven in dispute, and that can easily leadreasonable people to differingconclusions. As the comments illustrate,it is nearly always possible to conductadditional studies and investigations torefine further the knowledge of asubstance’s toxicological profile, theincidence and-degree of humanexposure to it, and the impact a giventolerance reduction will have on thefood supply. As an agency whose firstresponsibility is to protect the publichealth, however, FDA must act on thebasis of the information available to it,even when the information isincomplete. Neither the agency nor thepublic can afford to wait until everyuncertainty is resolved. See Ethyl Corp.v. Environmental Protection Agency, 541F. 2d 1, 24-29 (D.C, Cir,) (en hmc), cert.denied, 426 U.S. 941 (1976].

In the case of PCB’s, even thoughthere are obvious shortcomings in theavailable data, which are discussedbelow, FDA considers the data toprovide a more than adequate basis forthe exercise of its judgment in reducingthe PCB tolerances. There would be no

advantage in delaying this actionbecause it will take years to resolvecertain of the shortcomings in the dataon PCB’s, if they can be resolved at all.For example, no chronic toxicity studieshave been performed on the specific,chemically distinct composition of PCB’sfound in fish residues. Even if suchstudies were begun immediately, itwould be 3 to 4 years before resultscould be available. That plainly is toolong to wait to take action necessary forthe protection of the public health.

Because of the emphasis thecomments placed on the proposedreduction of the fish tolerance, thisdocument reviews the basis on whichthe reduction was proposed andexplains why, after considering thecomments, the agency has decided topromulgate the reduction as proposed.After discussing the fish tolerance andthe major points raised about it in thecomments, this document responds tothe remaining comments received onother aspects of the proposal.

11. The Tolerance for Fish and Shellfish

In the preamble to the April 1, 1977proposal, the agency discussed newtoxicity data that had become availableafter the original PCB tolerances werepromulgated in 1973 [42 FR 17488-9). Incontrast to the data underlying theoriginal tolerances, which consistedprimarily of data from retrospectivestudies of humans in Japan who wereexposed to high doses of PCB’s andshowed acute toxic effects from theexposure (42’FR 17487+), the newtoxicity data consist primarily of animalstudies showing an association betweenPCB exposure and serious subchronicand chronic toxicities, including adversereproductive effects, tumor production,and, possibly, carcinogenicity, as wellas effects on numerous biochemicalsystems [42 FR 17488-9). Although thedata do not fully resolve such importantquestions as the carcinogenicity ofPCB’s, they lead to the conclusion thatneither “no effect” nor “allowable dailyintake” levels for PCB’s can beestablished with any confidence andthat, from a toxicological point of view,human exposure to PCB’s should bereduced.

The preamble to the proposal alsodiscussed data FDA had gathered onhuman exposure to PCB’s, especiallyfrom dietary sources (42 FR 17489-90),These data show that the currentincidence of PCB contamination of foodhas declined significantly in comparisonto that on which the original PCBtolerances were based (see 37 FR 5705),Indeed, the new data show that fish arethe only food group in which detectable

levels of PCB contamination are nowroutinely found.

Based on the declining incidence ofPCB contamination, which means thatPCB’s are avoidable in food to a greaterdegree now than they were earlier, aswell as the new toxicity data suggestingchronic toxic effects, FDA decided thePCB tolerances should be reduced.

In the preamble to the proposal, theagency analyzed the new toxicity andexposure data as they bore specificallyon the tolerance for PCB’s in fish (42 FR17492-3). The agency concluded thatreduction of the tolerance from 5 to 2ppm was necessary to protect the publichealth adequately, even though thatreduction would result in the estimatedloss of a minor percentage of marine fish(approximately 0.2 percent) and up to 25percent of freshwater fish shippedinterstate (the loss of marine andfreshwater fish having a combinedlanded value of approximately $8million per year). The agency concludedthat the increment of public healthprotection afforded at least theoreticallyby a further reduction of the tolerance to1 ppm did not justify such a reduction inlight of the substantially greater loss offood that would result (a combinedlanded value, marine and freshwater. ofapproximately $18 million per year).

As noted, a large majority of thecomments on the proposal dealt withsome aspect of the agency’s proposal toreduce the fish tolerance to 2 ppm. Someof the comments agreed that theproposal struck a proper balancebetween the need to protect the publichealth and the need to avoid excessiveloss of food. Other comments arguedthat the tolerance should be reduced to 1ppm in light of the new toxicity data onPCB’s, despite any additional loss offood that might result.

Most of the comments on the fishtolerance, however, were submitted bymembers of the fishing industry, bytrade associations, and by agencies ofState governments involved incommercial fishing matters, who arguedthat reduction of the tolerance to 2 ppmis not justified. Some of these commentscontended that the health hazardpresented by occasional consumption offish containing 5 ppm PCB’s is notsignificant and that any reduction in riskto consumers accomplished by reducingthe tolerance to 2 ppm would be minor.These comments also argued that anysuch risk reduction would beoutweighed by the resulting adverseeconomic consequences, which someargued would be far in excess of thosecited by the agency in its proposal. Insupport of the latter argument, some ofthese comments estimated the impact a



2 ppm tolerance would have not only onthe commercial fish catch, but also onemployment and income in the fishingand related industries and onrecreational fishing. Arguing that theStates would curtail recreational fishingin certain areas if the tolerance were reduced to 2 ppm, the commentsprojected large losses of sales amongthose supplying boats, licenses, tackle,and bait to sport fishers.

Due to the large volume of commentschallenging the proposed reduction ofthe fish tolerance, the agency hascarefully reassessed the justification forlowering the tolerance from 5 to 2 ppm.It has reviewed the toxicological dataand has attempted to estimate inquantitative terms the degree to whichlowering the tolerance would reducerisk to consumers. In addition, it has re-examined the question of how muchadditional loss of fish would occur as aresult of the proposed reduction. Basedon its reassessment, the agencyconcludes that reduction of thetolerance for PCB’s in fish to 2 ppmstrikes the proper balance between theneed to protect the public health and theneed to avoid unnecessary loss of food.Hence, the reduced tolerance is beingpromulgated as proposed.

A. Risk Reduction

As noted earlier in this preamble, theproposal to reduce the fish tolerancewas based in part on new toxicity datashowing a relation between PCBexposure and an increased incidence ofvarious subchronic and chronic toxiceffects, including adverse reproductiveeffects, tumor production, and, possibly,carcinogenicity (42 FR 17487-9). Theproposal itself noted certain factors thatcomplicate the evaluation of PCBtoxicity (e.g., varying degrees of toxicityamong the several forms of PCB’s, thepresence of toxic impurities such aschlorinated dibenzofurans incommercial preparations of PCB’s, thedifferences in chemical compositionbetween commercial PCB’s and PCBresidues in fish, and varyingsusceptibilities of different animalspecies to the toxic effects of PCB’s);these complicating factors were alsopointed out in some of the commentsreceived on the proposal,

Notwithstanding these factors,however, there is little genuine disputeover the fact that exposure to PCB’smust be considered to pose a risk ofserious, chronic toxic effects in humans.The toxicological judgment that flowsfrom this fact—i.e., that a reduction inhuman exposure to PCB’s will reducethis risk-was an important part of theagency’s rationale for proposing to

reduce the fish tolerance. Nothing in thecomments and nothing discoveredduring FDA’s reassessment of thetoxicity data alters the validity of thatfundamental judgment. The agencytherefore concludes that it is importantas a matter of public health protection tominimize human exposure to PCB's.

The real question raised by thecomments is whether the degree of riskreduction accomplished by lowering thefish tolerance to 2 ppm is sufficient tojustify the increased loss of food that thelower tolerance will cause. This is anextremely difficult question because it isnot now possible for toxicologists toquantify precisely, on the basis oftoxicity data derived from animalstudies, the risks posed to humans.Using classical toxicological methods,the most that can be done reliably is tomake qualitative judgments about risks:A statistically significant increasedincidence of adverse effects in animalsis good evidence of a risk to humans,and, generally, the greater the incidenceof effects in animals, the greater the riskto humans (Ref. 43). Having identifiedthe risk of a chronic toxic effect fromexposure to a substance, classicaltoxicological principles lead to theconclusion that reduction in exposurewill reduce the risk (Ref. 44). Again,there is no evidence that theseprinciples do not apply to PCB’s.

Scientists have recently developedmethods, incorporating mathematicalextrapolation models, for makingquantitative estimates of risks tohumans based on toxicity data fromanimal studies. These risk assessmentmethods do not purport to quantifyprecisely the expected human risk, butrather attempt to estimate inquantitative terms an upper limit on therisk to humans that can be expectedfrom a given level of exposure to a toxicsubstance, assuming humans are nomore susceptible to the effects of thesubstance than are the most susceptiblemembers of the animal species forwhich toxicity data are available. Theserisk assessments can be useful as ameans of comparing risks at variousexposure levels and illustrating thetoxicological judgment that a reductionin exposure will reduce risk. Because ofall the problems inherent inextrapolating from animal data to theexpected human experience, however,the numbers produced by a riskassessment must be interpretedcautiously: They are estimates of upperlimits on risk and, though potentiallyuseful for comparative purposes, cannotbe said to quantify actual human riskprecisely. These assessments attempt toavoid underestimating human risk, but

even that cannot be guaranteed. TheWork Group on Risk Assessment of theInteragency Regulatory Liaison Group(IRLG) has recently prepared a reportthat discussers many of the principlesinvolved in risk assessment.

As part of its review of thetoxicological justification for reducingthe fish tolerance, the agency hasperformed a risk assessment aimed atcomparing the estimated risksassociated with PCB exposure at thevarious levels of exposure that wouldresult from different tolerance levelsThe written report on this riskassessment has been made a part of therecord of this proceeding as Reference45.

As explained in that report, the riskassessment involved the use of the mostrecent available data on the incidence ofPCB contamination of fish to calculatethe level of exposure to PCB’s that couldbe expected to result from tolerancelevels of 5,2, and 1 ppm (Table 4, Ref.45). These calculations were based onthe assumption that under a giventolerance level, no fish containing PCB’sin an amount above that level would beconsumed, It is true, of course, that anFDA tolerance level directly affects onlyfish shipped in interstate commerce, butStates often adopt FDA’s tolerancelevels for application to intrastate andrecreational fishing. Thus, even if theexposure calculations used in the riskassessment (Table 4, Ref. 45) dosomewhat overstate the absoluteamounts of exposure reduction, theynevertheless demonstrate that areduction of the PCB tolerance for fishwould result in a significant reduction ofPCB exposure [e.g., for heavy consumersof the affected species, reduction of thefish tolerance from 5 to 2 ppm reducesexposure from an estimated 20.1micrograms (µg) per day to an estimated14.9 µg per day). Such significantreductions in PCB exposure from fishare especially important in terms of riskreduction because fish are the only foodgroup in which detectable levels ofPCB’s are still regularly found.

Based on the calculations of exposureat various tolerance levels and toxicitydata from animal studies, the agencyused a linear extrapolation method toestimate the upper limits on certain risksposed by exposure to PCB’s, Thisanalysis resulted in estimates ofsignificant potential risk to humans whoconsume PCB-contaminated fish on acontinuing basis, especially fishcontaminated at or above the 5 ppmlevel (Tables 6 and 7, Ref. 45). Forexample, using the total malignancydata from the National Cancer Institute(NCI] Bioassay (Ref. 19), it is estimated

Appendix 6— FDA Derivation of PCB Tolerances—Federal Register . 1 2 9


that the upper limit on the lifetime riskof cancer for heavy consumers of fishmost affected by the tolerances is 9.8incidence of cancer per 100,000 of thepopulation, assuming the tolerance is 5ppm; 7.2 per 100,000, assuming thetolerance is 2 ppm; and 4.4 per 100,000,assuming the tolerance is 1 ppm (Table6, Ref. 45). Stated another way, it isestimated that the upper limit on thenumber of new cancers per year amongheavy consumers of fish most affectedby the tolerances is 46,8, assuming atolerance of 5 ppm; 34.3, assuming atolerance of 2 ppm; and 21, assuming atolerance of 1 ppm (Table 7, Ref. 45).

As explained in the report (Ref. 45),the utility of this risk assessment forevaluating actual risk to humans fromexposure to PCB’s is extremely limited.This is due both to difficulties inherentin making such extrapolations fromanimals to humans and, perhaps moreimportantly in this instance, to gaps anduncertainties in the data available forthis particular risk assessment. Forexample, the toxicity studies on whichthe risk assessment is based usedcommercial preparations of PCB’s,which are chemically different from thePCB residues found in fish and whichcontain small amounts of highly toxicimpurities (e.g., dibenzofurans) notknown to be present in fish residues,Also, in making the exposure estimatesrequired for the risk assessment, it wasnecessary to use existing data on thenumerical distribution of PCB levels infish and rely on the assumption that theeffect of a given tolerance level is toremove from commerce all fishcontaining PCB’s exceeding thetolerances. It is possible that neither theassumption nor the data precisely reflectwhat actually occurs.

For these reasons and othersdiscussed in the report (Ref. 45), the riskassessment does not provide a basis forprecise quantification of the amount ofrisk reduction accomplished by reducingthe fish tolerance, Despite thelimitations inherent in the riskassessment, however, the agencyregards it as illustrative of the basicvalidity of the toxicological rationale forreducing the tolerance for PCB’s in fish:Reduction of the tolerance will result ina significant reduction in risk amongthose who consume PCB-contaminatedfish. FDA considers this risk reductionto be of significant public health value,even though it cannot be preciselyquantified.

B. Loss Of Food

In the preamble to the proposal, theagency estimated that the loss of foodfrom commercial channels resulting

from a 2 ppm tolerance for PCB’s in fishwould be approximately $8 million inlanded value, compared toapproximately $1 million for the 5 ppmtolerance and $18 million for a 1 ppmtolerance, The estimated $8 million lossresulting from a 2 ppm toleranceencompassed a negligible percentage ofthe marine-fish catch (about 0.2 percent)and about 25 percent of the freshwatercatch (42 FR 17492),

The agency arrived at these figures byassuming that all fish containing PCB’sabove the tolerance would be removedfrom both interstate and intrastatecommerce (“Economic ImpactAssessment for Proposed Reduction ofTemporary Tolerances forPolychlorinated Biphenyls in Food,” Ref.39). There are several difficultiesinherent in this assumption. On the onehand, it may tend to overstate the lossbecause (a) some states may not applyFDA’s reduced tolerance to intrastatefish, (b) some violative fish will be partof nonviolative lots, and (c) someviolative lots may enter commerceundetected. On the other hand, it maytend to understate the loss because oncethe violative percentage of a givenspecies reaches a certain level,commercial fishers may stop fishing thatspecies altogether. Some of thecomments cited these difficulties insupport of arguments that FDA hadeither overestimated or underestimatedthe amount of fish that would be lost asa result of a 2 ppm tolerance. Despite itsacknowledged limitations, adoption ofthe assumption is a necessary andreasonable method for dealing with theuncertainties inherent in predicting theimpact of a tolerance reduction. None ofthe comments suggested an alternativemethod for estimating the amount of fishthat would be removed from commerceas a result of the proposed tolerancereduction.

Because of the comments it receivedquestioning the justification for theproposed reduction in the fish tolerance,the agency has re-examined itsprojections of the food loss expected toresult from such a reduction. Theprojections made in the preamble to theproposal were based on data obtainedin 1974 on the levels of PCB’s incommercial fish, primarily from theGreat Lakes (Ref. 39), In making thoseoriginal projections, the agency wasforced to rely on the assumption thatPCB levels in freshwater fish nationwidewere as high as those found in the GreatLakes. FDA now has more recent andmore representative data on PCB levelsin commercial fish, which it obtainedthrough a nationwide sampling programconducted in 1978 and 1979, Based on

these more recent data, the value of thefish projected to be lost at tolerancelevels of 5 ppm and 2 ppm issubstantially less than was projected inthe proposal. The loss projected under a1 ppm tolerance would remain about thesame (Table B, “Regulatory Analysis forFinal Regulation for Reduction ofTemporary Tolerances forPolychlorinated Biphenyls in Food,” Ref.46). Specifically, the amount ofcommercial fish now projected to be lostas a result of a 2 ppm tolerance is about$5.7 million (expressed in 1974 dollars)compared to the previously estimated $8million; the current estimated loss of fishunder a 5 ppm tolerance is about $0.6million (compared to the previouslyestimated $1.1 million). Under a 1 ppmtolerance, however, the projected fishloss, using the new sampling data onPCB levels, is about $16 million(compared to the previously estimated$18 million). The percentage of thefreshwater fish catch now estimated tobe lost under a 2 ppm tolerance is 14percent (compared to the 25 percent thathad been estimated from the 1974 data);under a 1 ppm tolerance, the currentlyestimated loss of freshwater fish is 35percent (compared to the previouslyestimated 43 percent) (Ref. 46),

As noted earlier in this preamble,many of the comments argued that theimpact of the proposed tolerancereduction must be measured not only bythe amount of the resulting fish loss butalso by other economic impacts, such aspotential unemployment and loss ofincome in the fishing industry andpostulated disruption of the recreationalfishing industry (e.g., reductions in boat,tackle, and bait sales]. The commentsprovided figures ranging into thehundreds of millions of dollars on thetotal economic value of these industriesand, without offering any furtheranalysis, contended that the impact onthem would be “severe” or “major.” Thepredicted impact on recreational fishingwas premised on the possibility thatState governments would severelycurtail recreational fishing if thetolerance were reduced to 2 ppm.

In establishing a tolerance for PCB’sin fish, FDA must take into account theamount of fish a given tolerance wouldremove from commerce. Section 406 ofthe act, however, neither requires norauthorizes FDA to weigh secondaryeconomic impacts when it considers thelevel at which a tolerance should be set.Consideration of such impacts would beinconsistent with the paramountconcern of section 406, which isprotection of the public health, andwould complicate the decisionmakingprocess under section 406 in a way


Federal Register / Vol. 44, No. 127 / Friday, June 29, 1979 / Rules and Regulations.

. Obviously,Congress did not intendconsideration of the amount of food losscaused by a tolerance helps to ensurethat the direct economic consequencesof the tolerance (in this case, decreasedsales and employment in the commercialfishing industry) will not bedisproportionate to the increased degreeof public health protection accomplishedby the tolerance; but the agencyconsiders secondary economicconsequences, such as potential impacton the recreational fishing industry,totally beyond the scope of section 408.

None of this should suggest that theagency is unaware of, or unconcernedabout, the economic consequences of itsactions. It is keenly aware that actions ittakes to protect the public health canhave adverse economic consequences,both direct and indirect, and that theseconsequences can sometimes be feltwith particular severity in certainnarrow segments of the economy. Forexample, some of the comments on theproposal argue that the impact of a 2ppm PCB tolerance for fish will beespecially severe for small-scale,freshwater fishers who specialize incertain species that happen to beheavily contaminated. The agencyacknowledged this possibility in thepreamble to the proposal (42 FR 17492).

in the present case, however, theagency has reason to believe that theclaims of adverse economic impact areexaggerated. Based on the 1078/1979data on PCB levels in freshwater fish, a2 ppm tolerance will remove fromcommerce about $5.7 million worth ofcommercial fish. Although it is possiblethat fishing for certain heavilycontaminated freshwater species maycease entirely in locations where PCBcontamination is concentrated, at leastsome affected fishers-both commercialand sport-can be expected to adjust tothe reduced tolerance by increasingtheir catch of other species ortransferring their activities to other, lesscontaminated locations within theircurrent area of operation.

In evaluating claims of economicimpact, it is theoretically andpragmatically sound to take into accountthe motives and opportunities foradaptive behavior by affectedindividuals and fins. If the publicdemand for commercially caught fishremains stable or increases, and if theattractions of sport fishing remainstrong, it can be expected that somefishing activity will shift to species thatare not contaminated above thetolerance. Over time, the shifts willbecome easier as the levels of PCBcontamination decline because moreand more species will have average PCB

levels well below 2 ppm. Over the longterm, the adjustments will help tominimize the net economic impact of thetolerance reduction on both individualfishers and the overall commercialfreshwater fishing industry.

None of the comments attempted toquantify in dollar terms the impact ofthe tolerance reduction on therecreational fishing industry, but severalpostulated a “severe” or “major” impactpremised on voluntary decisions byindividuals not to fish and mandatorycurtailments of recreational fishing byState authorities. FDA is in no betterposition than were those submitting thecomments to make precise predictionsabout the future behavior of individualsand State agencies. However, theagency considers the premisesunderlying the projections of “major” or“severe” impact to be somewhatspeculative and of questionable validity.AS noted to the extent that the behaviorof individual recreational fishers isaffected by the tolerance reduction atall, they, as much as commercial fishers,can be expected to adjust to thetolerance by shifting their activities tothe less contaminated species andlocations. Also, even if State agenciesdecide that some curtailment ofrecreational fishing is necessary in lightof the reduced tolerance, it is reasonableto expect that their actions will betailored by species and location. In thepast, the most common response ofState agencies to FDA’s PCB tolerancefor fish has not been the mandatorycurtailment of recreational fishing.Instead, they have issued warningsconcerning particular species andlocations and made suggestionsregarding both limitations onconsumption of particular species andmethods of preparing and cooking fishthat minimize the amount of PCB’sactually consumed from contaminatedfish. Thus, there is little reason tobelieve that a 2 ppm tolerance will leadto widespread, mandatory curtailmentof recreational fishing and the resultingdrastic economic impact the commentspostulate.

C. Conclusion

Based on the data now before it, theagency concludes that a reduction of thefish tolerance from 5 to 2 ppm will resultin a meaningful decrease in the riskexperienced by consumers fromexposure to PCB’s. Some reduction ofthe tolerance is clearly in order becausethe toxic effects associated withexposure to PCB’s are serious andirreversible; and, due to declining levelsof PCB contamination the current 5 ppmtolerance permits contamination that

can fairly be termed “avoidable’’ -evenamong the-more highly contaminatedcommercial species most likely to beaffected by a reduced tolerance, only aminor percentage (about 1.5 percent)contain PCB’s at levels as high as 5 ppm(Table A, Ref. 46). The agency’sjudgment is that the balance betweenpublic health protection and loss of foodis properly struck by a 2 ppm tolerance.As noted, as 2 ppm tolerance effects ameaningful decrease in risk toconsumers while still excluding fromcommerce only a relatively smallamount of food (about $5.7 millionlanded value in 1974 dollars).

Several comments argued that anadequate degree of public healthprotection can be provided only bylowering the fish tolerance to 1 ppm, thelowest level at which PCB’s can bereliably measured in fish forenforcement purposes. Indeed, as onewould expect, the risk assessmentperformed by the agency, and discussedabove, indicates that the estimated risksthat might be experienced by consumersof contaminated fish would be reducedeven further by a reduction of thetolerance to 1 ppm (Tables 6 and 7, Ref.45). Based on the evidence now beforeit, however, the agency does notconsider a reduction to 1 ppm necessaryor appropriate in light of the policy ofsection 408 of the act.

The risk assessment the agency madeincorporated several conservativeassumptions that were designed toavoid understatement of the human risk.Thus, it is expected that the actual riskexperienced by consumers of the 12more heavily contaminated speciescovered by the risk assessment is lessthan that estimated. Moreover, theaverage consumer, who eats fish from avariety of freshwater and marinesources, will actually experience a farlower level of PCB exposure and acorrespondingly lower degree of riskthan those whose fish consumption isconcentrated among the more heavilycontaminated (predominantlyfreshwater) species. For these reasons,notwithstanding the quantified riskestimates produced by the riskassessment, the agency reaffirms theconclusion it expressed in the preambleto the proposal: The 2 ppm toleranceprovides an adequate degree ofprotection for all but those whoconsume above-average amounts offreshwater fish taken fromcontaminated waters (42 FR 17493).

In the agency’s judgment, theadditional increment of public healthprotection that might be provided byreducing the tolerance to 1 ppm does notjustify the additional loss of food that



would resuIt. First, as discussed above,the agency estimates that under atolerance of 1 ppm, approximately $16million worth of the commercial fishcatch would be vioIative and thus,presumably, removed from commerce.This is nearly triple the $5.7 milIionworth estimated to be violative under a2 ppm tolerance. It is far more likelyunder a 1 ppm tolerance than under a 2ppm tolerance that the more heavilycontaminated species of freshwater fishwould be violative in percentages highenough to put an end to theircommercial exploitation and, possibly,force some segments of the freshwaterfishing industry to cease operationscompletely. Thus, the actual loss of foodresulting from the 1 ppm tolerance couldgreatly exceed even the $16 millionlanded value (1974 dollars) estimatedabove.

Second, for the average consumer,current exposure to PCB’s in fish is at atolerably low level, when considered inlight of the criteria of section 406 of theact, without a 1 ppm tolerance, Theaverage consumer eats a modest amountof fish from a variety of sources, bothfreshwater and marine, most of whichyield fish with PCB levels below 1 ppm.Because their exposure is thus low tobegin with, they are adequatelyprotected by a 2 ppm tolerance, whichensures that they will not be exposed tothe unusually high levels of PCB’s foundin some species of fish. The slightadditional protection these averageconsumers might gain from a 1 ppmtolerance does not justify thesignificantly greater impact such atolerance would have on the availabilityof food. On the other hand, atypicalheavy consumers (e.g., the Great Lakessport fisher who catches and consumeslarge quantities of the contaminatedspecies) would likely not be adequatelyprotected by even a 1 ppm tolerancebecause of the amount of fish they eatand because those fish are seldomaffected by FDA tolerances (eitherbecause they are sport fish or are fromintrastate commercial channels and, ineither case, are outside FDA’sjurisdiction). Protection of theseconsumers depends on actions by Stateauthorities.

Finally, though the new toxicity dataon PCB's clearly support the need toreduce exposure to this contaminant, theuncertainties in the data (discussedabove) cast some doubt on the degree towhich consumers are at risk fromextremely low levels of PCB exposure,and therefore weigh against loweringthe tolerance to 1 ppm. If, for example,more definitive and incriminating dataon the reproductive risks posed by

PCB’s are forthcoming, the agency mightconsider establishing a 1 ppm tolerancedespite the effect that would have on theavailability of food.

For these reasons, the agencyconcludes that at this time a 1 ppmtolerance would not strike the properbalance between protection of thepublic health and the need to avoidexcessive loss of food.

Though FDA considers 2 ppm to bethe appropriate tolerance level for PCB’sin fish under the criteria imposed bysection 406 of the act, the agency isconcerned about the health of certaingroups that may not be adequatelyprotected by a 2 ppm, or even a 1 ppm,tolerance. As noted, sport fishers andothers who consume abnormally largeamounts of the more highlycontaminated species may be at riskfrom PCB’s regardless of any toleranceFDA establishes. (The agency’s riskassessment, using data from a study ofLake Michigan sport fish eaters,estimated that the upper limit on thelifetime risk of cancer for heavy eatersof sport fish from Lake Michigan isabout 12 to 14 times greater than thecorresponding risk for heavy eaters ofthose commercial fish most affected bya PCB tolerance, even assuming thetolerance remained at 5 ppm (Table 6,Ref. 45).] Those individuals, whose highexposures to PCB’s tend to result fromlocalized conditions and fishingpractices beyond the control of FDA,should take steps to reduce theirexposure to PCB’s. FDA urges State andlocal health officials to evaluate thesituation in their own localities anddetermine what steps, if any, they cantake to address these special situations.In the past, some State and localagencies have made FDA’s tolerancelevel for PCB’ s applicable to fish inintrastate commerce and have issuedadvisories to sport fishers warning thatconsumption of certain species of fishshould be minimized and suggestingother ways in which PCB exposurecould be reduced. These agencies shouldreview their past actions in light of thecurrent state of knowledge about PCB’sand make the changes or take theadditional steps that may now beappropriate, FDA will cooperate withthese agencies, as it has in the past, byproviding technical advice andassistance. FDA is sending letters to thegovernors of States most affected byPCB’s in fish, discussing the agency'sconcerns about aspects of the PCBproblem that may require an up-to-datereview in their States.

The agency is advising that Statehealth departments be particularlyconcerned about women of childbearing

age, especially pregnant and lactatingwomen, who may have consumed, or areconsuming, higher than normal amountsof PCB-contaminated fish. Data thatwere discussed in the preamble to theproposal (42 FR 17468-9) suggest anassociation between PCB exposure andreproductive disjunction in rats andmonkeys. They also show acute toxiceffects in the nursing offspring ofmaternal monkeys that had beenexposed to toxic levels of PCB’s. Datagathered by FDA since it issued theproposal in 1977, and discussed in thereport on FDA’s risk assessment onPCB’s (Ref. 45), establish more clearlythe link between PCB exposure andadverse reproductive effects in therhesus monkey. They also confirm theearlier data showing acute toxic effectsin the nursing offspring of PCB-exposedmaternal monkeys. As explained in therisk assessment report (Ref. 45), it is notpossible at this time to determine withconfidence the significance of these datain terms of human risk. There have beenno reports of human reproductiveabnormalities or overt toxic effects innursing human infants that can beattributed to PCB’s. That fact is of onlylimited significance, however, becauseepidemiological studies adequate todetect such adverse effects in humanshave not been conducted.

An additional reason for concern inthis area is that PCB’s ingested byhuman mothers are found, and to someextent are concentrated, in humanbreast milk [see the discussions in thepreamble to the proposal and in the riskassessment report (Ref. 45)). In a recentnationwide survey, consisting of 1,038samples of human breast milk collectedin 44 States, the mean concentration ofPCB’s was estimated to be in the rangeof 1.00 to 1.10 ppm (on a fat basis) (Ref.45). Though the data are scanty, it isreasonable to assume that amongwomen who consume above-averageamounts of PCB-contaminated fish, orwho are exposed to PCB's from othersources, the levels of PCB’s in breastmilk are significantly higher, As noted itis not now possible to determine thesignificance of these facts in terms ofincreased risk to the nursing infant.

In sum, although the agency concludesthat a 2 ppm tolerance for PCB’sadequately protects most consumers,women of childbearing age, especiallypregnant and lactating women, areamong those who should be careful toavoid abnormally high exposure toPCB’s in fish, They can avoid suchexposure by minimizing consumption ofboth commercial and noncommercialfish from waters known to becontaminated with PCB's and avoiding



entirely those species of sport fishknown to contain high levels of PCB’s(e.g., coho and chinook salmon from theGreat Lakes, and freshwater trout,striped bass, and catfish from somelocations). State and local governmentshave the important role of advisingconsumers about conditions inparticular localities.

The agency is aware that its decisionto set the fish tolerance for PCB’s at 2ppm, rather than leaving it at 5 ppm orreducing it further to 1 ppm, is inherentlyjudgmental in character. Section 406 ofthe act provides no formula forbalancing public health protectionagainst loss of food, and, hence, there isno way for the agency’s decisions undersection 400 to be arrived atmechanically or quantitatively or toappear clear-cut in every case. In thiscase, for example, forceful argumentshave been made in the comments insupport of both a 5 ppm and a 1 ppmtolerance, but those arguments all reflectthe subjective” judgments of those whomade them. In the end, the agency hasbeen mandated by the Congress to makeits own informed judgment about whatis necessary to protect the public health.It has done that herein setting the fishtolerance at 2 ppm.

The statute provides an opportunityfor a public hearing on the agency’sorder lowering the PCB tolerance forfish. Such a hearing would providepersons adversely affected by the orderan opportunity to present any additionalevidence they may have bearing on thematters that influenced the agency’sjudgment. As always, the agency isprepared to reevaluate its position inlight of evidence adduced at a hearing.

D. Other Comments on the Fish andShellfish Tolerance

In addition to the points addressedabove, the comments raised severalother points relating to the tolerance forfish and shellfish:

1. One comment recommended thatFDA review its entire mechanism forhandling recurrent problems ofenvironmental contaminants in fish. Thecomment stated that the PCB toleranceshould remain at 5 ppm for marine fishbecause the levels in those fish are lowenough that a reduction to 2 ppm wouldhave no increased protective effect, butwould result in economic problems thatare unnecessary for species with onlyoccasional high PCB levels. Thecomment stated further that tolerancesshould be set for freshwater fish basedon their individual place in the market—their tonnage, distribution patterns, andconsumption patterns. When suchfactors combine to present a risk, it was

argued, the tolerance should be appliedselectively to both the species and thebody of water.

The individualized approach toestablishing and enforcing tolerances forenvironmental contaminants suggestedby this comment is not feasible becausethe necessary species-by-species,location-by-location data on PCBoccurrence do not exist. Furthermore,many lots of fish, as currently packagedand shipped, do not bear the water-of-origin information required for therecommended regulatory approach.These limitations make it necessary forthe agency to establish tolerances forfish on a generic basis. The result is auniform regulatory approach for allspecies, which provides clear and fairrules for all segments of the fishingindustry and is necessary to ensure thatuncertainties and limitations in data willnot result in increasing human exposureto PCB’s. To the extent that certainspecies only occasionally have PCBlevels above 2 ppm, the economicimpact of the reduced tolerance will beslight.

2. One comment stated that any FDAregulatory action regarding PCB’s in fishshould apply to sport fish as well ascommercial fish.

FDA’s regulatory authority extendsonly to foods shipped in interstatecommerce and clearly does not extendto fish caught and consumed byindividual sport fishers. FDA cooperateswith the State agencies who haveauthority over sport fishing by sharingdata and views regarding toxicological,analytical, and compliance matters, butFDA has no direct control over theregulatory approaches adopted by theStates. As noted, however, the agencyurges State and local health officials tolook closely at the PCB problem in theirareas and take whatever steps they findnecessary to address those aspects ofthe PCB problem, such as the exposureof sport fishers, that are beyond FDA’sauthority.

3. One comment requestedreconsideration of the proposal toreduce the fish tolerance on the groundthat overall ingestion of PCB’s isreportedly declining. Because levels inother foods have already decreasedconsiderably, it was argued, there is lessneed to lower the fish tolerance,

The agency is aware that PCB levelsin foods other than fish have declinedand that overall PCB intake is lowerthan it was in 1973, when the originaltemporary tolerances were established.However, as discussed in the preambleto the proposal and in section II of thispreamble, toxicological considerationsnow make it desirable to reduce dietary

exposure to PCB’s even further.Reduction of the tolerance for PCB’s infish will bean especially effective steptoward accomplishing that goal, becausefish are the one remaining significantsource of dietary exposure to PCB’s.

4. One comment contended that fishproducts are being subjected to anentirely different regulatory standardthan are poultry products, with noreasonable basis for the differenttreatment. The comment stated that theemphasis in establishing 2 ppm as thetolerance for fish appears to have beensafety to the consumer despite aconsiderable economic impact. Yet, itargued, the higher level of 3 ppm forpoultry is based on economicconsiderations relating to feedcontamination, apparently withoutpublic health considerations. Thecomment went on to state that theaverage per capita consumption of fishis 19 grams (g) per day compared to 63 gper day for poultry products. Accordingto the comment, this means that underthe proposed tolerances, and assumingmaximum permissible levels in all foods,the average person will receive fivetimes as much PCB’s from poultry asfrom fish.

The agency does not agree with thecomment’s contention that theconsiderations involved in establishingthe tolerances for poultry and fish resultin different or conflicting regulatoryapproaches for these products, First, the3 ppm tolerance for poultry is based onPCB residues in the fat of the bird, not inall the edible tissue as it is for fish,Poultry generally averages about 10percent fat; hence, the 3 ppm toleranceis comparable to a level of about 0.3ppm for the entire edible portion. Thus,even taking into account the higheraverage level of chicken consumptionand assuming all foods containmaximum permissible amounts of PCB’s,poultry will actually be regulated at alevel that will result in a substantiallylower intake of PCB’s from poultry thanfrom fish. Second, data show thatdetectable PCB residues occur soinfrequently in poultry that exposure toPCB’s from that source is already at aninsignificant level. Hence, furtherreduction of that tolerance would notsignificantly reduce dietary exposure toPCB’s and would not enhance protectionof the public health, Fish data, on theother hand, show frequent occurrence ofPCB residues at significant levels, sothat reduction of the tolerance willresult in increased protection forconsumers of fish.

As explained in the preamble to theproposal (42 FR 17491-2), the agencyselected 3 ppm (fat basis) as the



tolerance for PCB’s in poultry to allowfor the regular use of poultry feedcontaminated up to, but not exceeding,the 0.2 ppm tolerance for PCB’s inpoultry feed. (0.2 ppm is the lowestfeasible tolerance for PCB’s in poultryfeed because of limitations on analyticalcap ability.) The 3 ppm level took intoaccount the biomagnification of PCB’s inpoultry that results from regular feedingwith poultry feed contaminated up to,but not above, 0.2 ppm. The agencyreasoned that it would be inconsistentto set tolerances on two products atlevels such that the use of one productthat complies with the applicabletolerance causes the second product tobe illegal and, thus, that it would beinappropriate to do so in the absence ofother overriding considerations [e.g.,safety). For the reasons stated in thepreceding paragraph, the 3 ppmtolerance for poultry adequatelyprotects the public health and is thusconsistent as a matter of public healthprotection with FDA’s other tolerancesfor PCB’s.

5. One comment stated that anydecision to lower the fish tolerancemade in reliance on the regulation ofpoint source discharges andmanufacture of PCB’s should not fail toconsider the fact that PCB levels incontaminated waters are not expectedto decline for many years.

The agency is aware that, despiteEnvironmental Protection Agency(EPA) antipollution activities and theresulting gradual decline in PCB levelsin at least some contaminated waters,there will continue to be a significantoccurrence of PCB’s in fish for at leastthe next several years because of thestability and persistence of the PCB’snow contaminating the environment.That fact was taken into account indeciding to reduce the tolerances.

6. One comment stated that, becausepollution of water with PCB’s isexpected to continue, PCB levels in fishwill continue to rise, and susceptiblefish should be harvested now before theincreased contamination makes them allinedible.

Although the levels of PCB’s in waterscurrently contaminated may notdecrease substantially in the nearfuture, the agency does not expect thoselevels to increase, nor does it expect thelevels of PCB’s in fish to increase. Bettercontrol of PCB levels should result fromefforts by the EPA and industry tocontrol discharge of additional PCB’sinto the environment. Hence, even if itwere possible to harvest whole speciesof fish now, that step would not havethe effect of preventing increased futureexposure to PCB’s. Finally, FDA has no

authority to regulate the pace at whichparticular species of fish are exploitedcommercially.

7. Several comments stated that thedecision to reduce the tolerance for fishshould be reconsidered and the current5 ppm level reaffirmed because PCB’sare being steadily eliminated from theenvironment and may be expected todisappear as a significant problemwithin the next decade.

The agency does not agree that PCB’scan be expected to be an insignificantproblem within 10 years. AlthoughEPA’s continuing activities haveresulted in a significant decrease in theamount of PCB’s being introduced intothe environment especially into water,the stability and persistence of thesechemicals and the likelihood that someamount of additional contamination willcontinue to occur from waste disposalsites ensures that PCB contaminationwill remain a problem for theforeseeable future. Moreover, that PCBlevels are declining (i.e., that PCB’s arebecoming more avoidable) is a reason toconsider lowering the tolerance. not ajustification for leaving it unchanged.

8. One comment argued that thedecision to reduce the fish toleranceshould be reconsidered because bylowering the fish tolerance, therebypreventing consumption ofcontaminated fish, some might be led tobelieve that the problem of exposure toPCB’s had been solved. Thismisconception could in turn reduce thepressure to attack the real problem—pollution, However, the commentargued, if the environmentalcontamination itself is viewed as the“real” PCB problem of importance,changing the fish tolerance is almostirrelevant, given the small quantity ofPCB’s affected.

PCB contamination of theenvironment is itself an important partof the PCB problem because, asdiscussed in the preamble to theproposal (42 FR 17469-90), some humanexposure to PCB’s comes from the airand water, though the amount isprobably minimal. EPA is addressingthat part of the problem. However, FDAdisagrees with the view that exposure toPCB’s from dietary sources isinsignificant in comparison to theamount of exposure from the air andwater. The agency has based theproposed tolerance reductions on itsconclusion that dietary exposures toPCB’s pose significant risks toconsumers, which can be reduced byreducing exposure. That there is someexposure to PCB’s from other sources isnot a good reason for withholding action

that can significantly reduce dietaryexposure.

9. Two comments requested FDA tohold a public hearing before finalizingreduction of the fish tolerance.

The agency does not consider a publichearing on the fish tolerance to benecessary or appropriate at this time.Tolerances are established undersection 406 of the act under the formalrulemaking procedures set forth insection 701(e) of the act (21 U.S.C.a71(e]]. Under those procedures, anyperson adversely affected by this ordermay file objection within 30 days andrequest an evidentiary hearing on theissues raised by those objections. Theopportunity for a hearing ensures thatall genuine, material issues relating tothe PCB tolerances will be fully aired.Holding a hearing before issuing thisorder would only duplicate theopportunity for a public hearing alreadyavailable in formal rulemaking andunnecessarily delay the proceedings.

10. One comment stated that the 5ppm tolerance for PCB’s in fish shouldbe retained but requested that FDAprovide guidance to State agenciesregarding use or implementation of the 2ppm tolerance if it is adopted.

As noted, FDA provides data andviews to the States on a range of mattersrelated to implementation of tolerancesfor PCB’s in food and will continue to doso.

11. One comment asked whetherprocedures other than reducing the fishtolerance have been evaluated asalternative means of reducing intake ofPCB-contaminated fish.

The agency has considered the use ofgeneral public warnings and/or labelingas ways to limit consumption ofcontaminated fish. Such approacheshave been rejected, except as they applyto certain heavy consumers ofcontaminated sport fish (discussedabove]. A public warning about fishgenerally, or even about particularspecies of fish, would not be effective inprotecting the general public fromcommercial fish because, assuming nochanges are made in labeling,consumers have no way to determinethe species or waters of origin of mostcommercially prepared fish products. Inaddition, general public warnings mightunduly discourage consumption of fish,most of which is safe to eat andnutritious. Similarly, the requirement ofwarning labels on fish products in lieu ofa tolerance, even on a species-specificbasis, is not a sufficiently preciseregulatory approach because not all fishfrom even the most heavilycontaminated species contain levels ofPCB’s above the 2 ppm tolerance level.



Thus, as with general public warnings,warning statements on labels are likelyto discourage consumption of safe fish.

12. Some comments contended thatmost Americans would probably preferto be warned of the potential dangerfrom PCB residues and retain the optionof eating freshwater fish, rather than bedeprived of any choice in the matter byhaving the fish removed from themarket.

The agency acknowledges that somepeople would probably prefer to be leftwith the choice of whether to consumefish contaminated with PCB’s above the2 ppm level. As noted, however, theconsumer of commercially marketed fishgenerally lacks the information onwater-of-origin, size, and sometimeseven species that is needed to controlhis or her intake of PCB’s. Under thesecircumstances, there is no genuineopportunity to exercise informed choice.Moreover, the agency believes that as ageneral matter it is obligated undersection 408 of the act to exercise itsscientific judgment and determine whatlevel of exposure, and thus whattolerance level, will provide anadequate degree of public healthprotection,

13. One comment referred to theagency’s decision not to reduce thetemporary tolerances for infant andjunior foods and for animal feeds on theground that the current tolerances are“at the lowest level at which PCB’s canbe reliably determined for enforcementpurposes” and argued that lowering thetolerance for fish would probably createmuch greater economic hardship thanwould developing and using moresensitive analytical methods so thatother tolerances could be lowered,

The agency acknowledges that thefish losses resulting from a 2 ppm fishtolerance would probably be greaterthan the costs of developing and usingthe more sensitive enforcement analysesthat would be necessary for a reductionof the other tolerances. The occurrenceof PCB residues in infant and juniorfoods and animal feeds is now soinfrequent, however, that those foods donot contribute significantly to dietaryexposure to PCB’s. Thus, spending theresources to develop more sensitivemethodology and thereafter reducing thetolerances for these foods would notsignificantly increase the protection ofconsumers, and it still would benecessary to reduce the fish tolerance,Because PCB’s do occur consistently atsignificant levels in some fish, thereduction of the fish tolerance canprovide increased protection forconsumers.

14. Some comments included requestsfor compensation for commercial fishersand processors whose livelihoods aredestroyed by reduction of the fishtolerance. One comment asked, in effect,that the effective date of the tolerancereduction be delayed for 10 years sofishers would have time to adjusteconomically.

For reasons discussed earlier in thisdocument, the agency considers itunlikely that the reduction of the fishtolerance to 2 ppm will have the direconsequences on which these commentsare premised. Moreover, FDA hasneither the authority nor the resourcesto provide compensation for economiclosses that might be suffered as a resultof regulatory actions it takes. Theproposed 10-year postponement of theeffective date would be inconsistentwith the agency’s conclusion that areduction of the tolerance is necessary toprotect the public health.

15. A number of comments wereconcerned that if the 2 ppm tolerance forfish is adopted, FDA will close certainwaters to fishing or prohibit fishing ofcertain affected species in certainwaters. They requested that morestudies be carried out beforedetermining whether such steps shouldbe taken.

The concern underlying thesecomments is misdirected, FDA does nothave authority either to close waters tofishing or to prohibit harvesting orpossession of fish. Any actions to closewaters to fishing would have to beinstituted by State agencies.

16. One comment suggested that if the2 ppm tolerance is adopted, the countiesaffected should be allowed to conductmore comprehensive testing of residuelevels in the fish before any ban orimpoundment of fish in interstatecommerce is imposed.

In enforcing the fish tolerance, FDAwill sample and analyze individual lotsof fish in interstate commerce and takeregulatory action against lots, or theshippers of lots, that exceed thetolerance, There is nothing to prohibitany interested party, including local andState authorities, from conductingcomprehensive testing of fish beforeshipment in interstate commerce andfrom withholding from commerce fishthat exceed the tolerance.

17, One comment suggested that theproposed 2 ppm tolerance for PCB’s infish is inadequate for protection ofpublic health, The comment stated thatthe tolerance levels must be based onthe “no-effect” level observed in-themost sensitive animal species for whichtoxicological data are available, and itsuggested that the rhesus monkey is

more sensitive to PCB’s than the dog orrat.

This comment is based on anapparent misunderstanding of thetoxicological rationale underlying the 2ppm fish tolerance. In evaluating thesafety of substances in food, FDAordinarily attempts to determine the“no-effect” level for the substance, i.e.,the highest level of exposure at whichno adverse effect is observed inappropriate animal studies, It then usesappropriate safety factors to extrapolatethe results of the animal studies to thehuman situation and determine safelevels of human exposure. In this case,however, the reduction of the fishtolerance is not based on any “no-affect” level, It is based instead on abody of data that associate PCBexposure with several serious chroniceffects but that do not permit theestablishment of “no-effect” levels forthose effects, Thus, the comment’sargument that one species is moresensitive to PCB's than another and thatthe tolerance should be based on the“no-effect” level observed in the mostsensitive animal species is not relevantto the toxicological rationale the agencyrelies on for reducing the PCB toleranceto 2 ppm.

18. One comment disagreed with theproposal to establish a 2 ppm tolerancefor fish instead of a 1 ppm tolerance.One ppm is the lowest level of PCBresidues in fish for which there isanalytical methodology suitable forenforcement purposes. The commentstated that toxicological information,especially that suggesting thecarcinogenicity if PCB’s coupled withthe presence of PCB residues in humanmilk, requires the lowest possibletolerance.

With respect to the carcinogenicpotential of PCB’s, NCI has concludedthat PCB’s (specifically, Aroclor 1254,the commercial PCB most similarchemically to the PCB residues in fish)are not carcinogenic in Fischer 344 ratsunder the conditions of the bioassay(Ref. 47). After thoroughly reviewingNCI’s report, the Data Evaluation/RiskAssessment Subgroup of theClearinghouse on EnvironmentalCarcinogens accepted the report’sconclusion that PCB’s were notdemonstrated to be carcinogenic in thatstudy, but suggested that PCB’s mightact as a tumor promoter. For the reasonsdiscussed in the preamble to theproposal (42 FR 17489), FDA considersthe question of the carcinogenicity of thePCB’s unresolved. For the purposes of itsrisk assessment on PCB’s (Ref. 44),however, the agency treated the variousPCB’s as though they were carcinogenic,

Appendix B— FDA Derivation of PCB Tolerances—Federal Register . 135


and it considers the carcinogenicity ofPCB’s to be a matter worthy of furtherserious inquiry.

The agency has long been concernedwith the exposure of nursing infants toPCB’s in human breast milk. This, too, isan area in which more must be learnedbefore definitive statements can bemade about the incremental risks posedby this particular avenue of exposure.For reasons discussed earlier in thisdocument, however, the agencyconsiders a 2 ppm tolerance adequate toprotect all but those who consumeabove-average amounts of the moreheavily contaminated species.

111. Response to Comments on OtherAspects of PCB’s

Following are the agency’s responsesto the comments that did not specificallyaddress the reduction of the fishtolerance:

1. Some comments recommended thatthe government regulate PCB’s only inthe environment rather than in foodproducts. Another comment suggestedthat the limits on PCB’s in foods not bereduced until PCB levels have beenreduced in the environment, where thefoods are produced.

EPA has the authority to controlenvironmental pollution and has alreadytaken important steps to prevent furtherpollution by PCB’s. Some environmentalcontamination with PCB’s alreadyexists, however, and will undoubtedlypersist for some years. FDA would befailing in its duty to protect the publichealth if it withheld the actionsnecessary to minimize human exposureto PCB’s from dietary sources until thelong-term problem of environmentalcontamination has been solved.

2. One comment asserted that theprimary toxicological basis upon whichFDA established the temporarytolerances for PCB’s in 1973 (38 FR18096) consisted of two long-termsfeeding studies in rats and dogs thatwere performed by the same testinglaboratory and that demonstrated a “no-effect” level for PCB’s at 10 ppm. Thiscomment also suggested that these sametwo studies serve as the primary basisfor the current proposal to reduce thoseoriginal temporary tolerances: Thecomment stated that discrepancies andinconsistencies have recently beenfound in these two feeding studies, aswell as in other unrelated studies fromthe same testing laboratory, whichwould indicate that toxic effects mightactually have been produced in bothrats and dogs, at dietary levels as low as1 ppm. The comment requests that FDAextend its audit of the testing laboratoryin question to include a review of the

data obtained in the two toxicity tests ofPCB’s in rats and dogs. The commentsuggested that FDA reconsider theproposed temporary tolerances on thebasis of a reevaluation of the data fromthe two long-term studies and, if judgednecessary, propose new tolerances orreopen the matter for public comment.

Though the data from the two long-term toxicity studies of PCB’s in rats anddogs referred to in the comment wereconsidered, the human toxicologicaldata formed the primary basis fordeveloping the original temporarytolerances for PCB’s. This fact wasstated in the July 6, 1973 documentestablishing the tolerances and in thepreamble to the April 1,1977 proposal toreduce some of the tolerances.

The agency is aware that doubt hasbeen cast on the validity of the twolong-term toxicity tests of PCB’s in ratsand dogs referred to in the comment, aswell as on the validity of numerousunrelated toxicity tests performed by thelaboratory facility in question (Ref. 48).Therefore, the results from these twotests are no longer considered worthy ofreliance and, as explained earlier in thisdocument, these studies played no partin the agency’s decision to lower thePCB tolerances.

3. One comment asserted that theagency’s statement in the preamble tothe proposal that it was unaware of anyconsumers who had suffered deleteriouseffects caused by PCB ingestion (42 FR17491) is misleading, in that thestatement actually reflects a lack ofknowledge rather than awareness of theresults of properly designedepidemiological studies.

In making this statement the agencyrelied on, the results of anepidemiological study carried out withsport fishers in Michigan that failed toestablish a correlation in humansbetween the ingestion of PCB’s and theoccurrence of deleterious effects (Ref.40). The study is discussed in thepreamble to the proposal (42 FR 17492-3). The only purpose of the statementwas to cite an instance in whichrelatively high exposure to PCB’s in fishhad not resulted in overt, acute toxiceffects, such as occurred in the Yushoincident in Japan (42 FR 17488). TheMichigan example was intended toillustrate the observation the agencymade in the preamble to the proposalthat the amount of PCB’s inenvironmental samples required tocause Yusho-type effects is not known.This study has no direct bearing on theagency’s conclusion that the chroniceffects of PCB’s require a reduction ofthe tolerances,

4. One comment opposed reduction ofthe temporary tolerance for PCB’s ineggs on the ground that there are nosubstantial data that suggest that thecurrent temporary tolerance is notsufficient to protect consumers of eggs.The comment contends that lacking suchevidence, there is not justification forreducing the tolerance.

The agency acknowledges that thedata indicate that eggs do not contributemeasurably to dietary PCB exposureand that reduction of the egg tolerancewill not significantly affect PCB intakes.However, tolerances established undersection 408 of the act are intended topermit only those residues that areunavoidable. Because the available dataindicate that residues above theanalytical limits in eggs are avoidableand because no evidence was presentedto the contrary, it is appropriate toreduce the temporary tolerance forPCB’s in eggs as proposed.

5. One comment requested that thetemporary tolerances for PCB’s inanimal feed be reduced, but it did notpresent a rationale to justify reduction.

The presence of PCB’s in animal feedis of concern because PCB’s transfer andaccumulate in human food productsderived from animals that consumecontaminated feed. The tolerance forfinished animal feed is currently set at0.2 ppm—the lowest level at whichavailable analytical methodology canmeasure PCB’s in animal feed forenforcement purposes. It would serve nouseful purpose to reduce the tolerancebelow this level in the absence ofanalytical methodology for enforcing areduced tolerance. Moreover, in light ofthe rare occurrence of PCB’s in animalfeeds, the agency considers the 0.2 ppmlevel to provide an adequate degree ofpublic health protection, For thesereasons, the agency declines to reducethe tolerance for PCB’s in finishedanimal feed.

IV. “Temporary” Status of theTolerances

As currently codified in S 109.30, thetolerances for PCB’s are designated as“temporary,” The term “temporary” wasused to reflect the fact that thetolerances are subject to revision asnew data become available. In thepreamble to the proposal for reducingthe tolerances, the agency stated that itwould retain the “temporary”designation because of the possibilitythat further downward revisions of thetolerances might be necessary (42 FR17493). The agency has nowreconsidered this use of the term“temporary” and has decided toabandon it. The term has never had any



legal significance as applied totolerances established under section 406of the act, and its use is not provided forin FDA’s procedural regulationsgoverning tolerance setting in Part 109(21 CFR Part 109). When circumstancesare changing so rapidly that a particulartolerance level is likely to be renderedinappropriate in the near future, theagency establishes an action level ratherthan a tolerance (see $ 109.6(c) (21 CFR109.6(c))). In the case of PCB’s, however,the agency has concluded that formaltolerances are appropriate. The term“temporary” is being abandoned toavoid the suggestion that the legal statusof the PCB tolerances is something otherthan that of a formal section 406tolerance.

Any FDA tolerance, just like any otherregulation, is “temporary” in the sensethat it is subject to reevaluation and, ifnecessary, revision as new data becomeavailable. The agency will continue tomonitor the PCB problem and, ifappropriate in light of changingcircumstances or new data, will proposerevisions in the PCB tolerances.

V. Analytical Methodology

Section 109.30(b) has been revised torefer to FDA’s updated compilation ofanalytical methodology for PCB’s“Analytical Methodology forPolychlorinated Biphenyls, June 1979.”There have been improvements in theanalytical methodology for measuringPCB residues since 1973, and most of therevised procedures have now beenpublished in scientific journals. A copyof each procedure or a reference to theappropriate journal is provided in theupdated compilation. As stated in3 109.30(b), the compilation is availablefrom the Hearing Clerk, Food and DrugAdministration, Room 4-65, 5600 FishersLane, Rockville, MD 20857.

VI. References

The preamble to the proposal cited 42references the agency relied on indeveloping the proposal and stated thatthose reference documents had beenplaced on file with the Hearing Clerk,FDA (42 FR 17493-4). The followingadditional references, which are cited inthe foregoing preamble, have also beenplaced on file with the Hearing Clerk,FDA, Rm. 4-85, 5800 Fishers Lane,Rockville, MD 20857, and may be seenbetween 9 a.m. and 4 p.m., Mondaythrough Friday.

43. DuBois, K. P. and E. M. K. Ceiling,“Textbook of Toxicology,” OxfordUniversity Press, 1959, pp. 24-28.

44. Ariens, E. J., A, M. Simonis, and J.Offermeier, ‘*Introduction to General

Toxicology,” Academic Press, 1976, pp.124-31.

45. An Assessment of Risk Associated withthe Human Consumption of Some Speciesof Fish Contaminated with PolychlorinatedBiphenyls (PCB’S), 1979, FDA document.

46. Regulatory Analysis for Final Regulationfor Reduction of Temporary Tolerances forPolychlorinated Biphenyls in Food, 1979,FDA document,

47. National Cancer Institute CarcinogenesisTechnical Report Series No. 38, 1978.

48. Letter from Donald Kennedy,Commissioner of Food and Drugs, tovarious clients of Industrial Bio-TestLaboratories, Inc., June 1977.

Therefore, under the Federal Food,Drug, and Cosmetic Act (sees. 306,402[a), 406, 701[a), 701[e), 52 Stat. 1045-1046 as amended, 1049 as amended,1055, 70 Stat. 919 as amended (21 U,S,C.336, 342(a], 346, 371(a], 371[e))) andunder authority delegated to theCommissioner of Food and Drugs (21CFR 5,1), Part 109 is amended in $109,30by revising the section heading andparagraphs (a)(l), (2), (3), (4), and (7] and(b) to read as follows:

$109.30 Toterences for polychlorinatedbiphenyls (PCS’S).

(a) ● ● ●

(1) 1.5 parts per million in milk (fatbasis).

(2) 1.5 parts per million inmanufactured dairy products [fat basis).

(3) 3 parts per million in poultry (fatbasis).

(4) 0.3 part per million in eggs.● * * * *

(7) 2 parts per million in fish andshellfish (edible portion), The edibleportion of fish excludes head, scales,viscera; and inedible bones.●

(b) For determining compliance withthe tolerances established in thissection, a compilation entitled“Analytical Methodology forPolychlorinated Biphenyls, June 1979" isavailable from the Hearing Clerk, Foodand Drug Administration, Department ofHealth, Education, and Welfare, Room

20857,● * * * *

Any person who will be adverselyaffected by the foregoing regulation mayat any time on or before July 30, 1979,submit to the Hearing Clerk (HFA-305),Food and Drug Administration, Rm. 4-65, 5600 Fishers Lane, Rockville, MD20857, written objections thereto andmay make a written request for a publichearing on the stated objections. Eachobjection shall be separately numberedand each numbered objection shallspecify with particularity the provisionof the regulation to which objection is

made. Each numbered objection onwhich a hearing is requested shallspecifically so state; failure to request ahearing for any particular objectionshall constitute a waiver of the right to ahearing on that objection. Eachnumbered objection for which a hearingis requested shall include a detaileddescription and analysis of the specificfactual information intended to bepresented in support of the objection inthe event that a hearing is held; failureto include such a description andanalysis for any particular objectionshall constitute a waiver of the right to ahearing on the objection. Four copies ofall documents shall be submitted andshall be identified with the HearingClerk docket number found in bracketsin the heading of this regulation,Received objections may be seen in theabove office between the hours of 9 a.m.and 4 p.m., Monday through Friday.

Effective date. This regulation willbecome effective for foods initiallyintroduced into interstate commerceafter August 28, 1979 except as to anyprovisions that may be stayed by thefiling of proper objections. Notice of thefiling of objections or lack thereof willbe announced in the Federal Register,[Sees. 306, 402(a), 408, 701(a), 701(e), 52 Stall1045-1046 as amended, 1049 as amended,1055, 70 Stat. 919 (21 U.S.C. 336, 342(a), 346,371(a], 371(e)])

Dated: June 26, 1979.Donald Kennedy,Commissioner of Food and Drugs.

[FS Dec. 79-z4wM Fded %2s-7s s.45 am]

BILLINQ COI)E 411 O-O3-M

Appendix C

Methods for Toxicologic Testing *by SRI International

METHODS FOR TOXICOLOGIC TESTING

137


scheme, experiments are designed and data arecollected based on expected results such as func-tional systemic changes, teratogenicity, or car-cinogenicity. By the use of an appropriate experi-mental design, several endpoints can be assessedin the same experimental period such as is donein FDA’s three-generation studies (2).

For ease of presentation, this appendix hasbeen subdivided by endpoint into sections on Sys-temic Toxicity, Carcinogenicity, Mutagenicity,Teratology and Effects on Reproduction, Metabo-lism, and Structure-Activity Relationships. Thisappendix is not intended as an exhaustive surveyof all testing methods used, but is meant to give anoverview of those methods most commonly usedtoday by toxicologists.

LocaI and Systemic Toxicity

Some of the fastest and simplest methods fordetermining the toxicities of substances involvethe observation of changes in the structure andfunction of organs and organ systems. Thesemethods generally involve absolute and relativeweight changes, gross and microscopic structuralalterations, and primary and secondary tests fororgan, system, or whole animal function. With ad-vances in the chemical, physiological, and behav-ioral sciences, modifications for testing systemictoxicity have been proposed that make these pro-cedures more sophisticated and relatively compli-cated. Several good texts are available which re-view systemic toxicity [3,4).

Range FindingThe classic determinations of toxicity involve

percent lethal or effective dose, concentration, ortime. These tests may employ any route of expo-sure, the ones chosen usually being based on fac-tors such as chemical and physical properties ofthe agent and potential routes of exposure fromthe environment. The results obtained from thesedeterminations are usually specific for the spe-cies, sex, age, and condition of the organism, andfor the route of exposure and environmental con-ditions before, during, and after exposure. Theendpoints of these tests may be either structuralor functional changes, but they are usually lim-ited to gross effects such as death or narcosis.These tests are primarily used to determine rela-tive toxicities of various agents and for rangefinding for maximum tolerated dosage prelimi-nary to beginning a subacute study. They are notusually used to directly evaluate the hazard.

In general, these tests will employ young adultrats and another mammalian nonrodent species.

Selection of this other species " . . should con-sider such factors as comparative metabolism ofthe chemical and species sensitivity to the toxiceffects of the test substance . . . . “(l) The route ofadministration chosen is that most nearly iden-tical to the potential human exposure. Doses areusually chosen to give results in the 20- to 80-percent lethal or effective range and are usuallyseparated by 0.5 log units (5). Many modificationsof this basic procedure are accepted.

IrritationThe irritation potential of substances is tested

by observation of the reflex behavior of the ani-mal and by direct observation of the site of con-tact with the agent, Attempts to quantitate reflexbehavioral responses, i.e., eye rubbing, regurgita-tion, or shallow breathing, have met with littlesuccess. The simplest protocols for evaluating ir-ritation involve the skin and eyes. Semiquantita-tive systems for scoring skin and eye irritationhave been proposed by several authors (6-12) andinvolve placing the suspected irritant in contactwith the skin or eye of New Zealand White rab-bits. Protocols for skin irritation involve contactwith both intact and abraded skin to differentiatethe agent’s ability to penetrate the skin barrier,and occlusion of the contact site to maximize theresponse. The severity of erythema and edema isscored as the endpoint. Eye irritation studies arecarried out without washing and with washing atvarious intervals to determine the effectiveness ofremoval of the agent to reducing the adverse ef-fect, Opacity, area affected, iris reaction to light,hemorrhage, swelling, and discharge are scored.Results from these tests can vary greatly depend-ing on the method of application of the substance,whether dry or premoistened, etc.

Other potential sites of irritation such as thesensory nerves, respiratory system, urinary sys-tem, and gastrointestinal tract are usually evalu-ated secondarily or through necropsy. Secondaryeffects include shallow breathing, regurgitation,agitation on urination, and eye rubbing, whichare broadly categorized as reflex behavior, andblood in feces, urine, and sputum, or nasal dis-charge, which are more indicative of the primaryirritant effect. These results are not quantifiableby present methodologies. Primary evaluationthrough necropsy is also a qualitative procedureand has not undergone the extent of standardiza-tion and validation that the skin and eye testshave. Other methods for testing irritation such asresistance/compliance tests of the pulmonary sys-tem, direct observation by scope of the esophagus

Appendix C—Methods for Toxicologic Testing ● 739

and gastrointestinal tract, and roentgenographicexamination with and without radio-opaque dyes,have not received wide use as testing techniquesfor Government regulatory purposes.

SensitizationSome substances, although not necessarily pri-

mary irritants, elicit an irritant-type responseafter repeated contact with the organism. Testsfor this sensitization potential involve exposingthe animal to an agent at doses below those neces-sary to produce signs of primary irritation, wait-ing an appropriate interval, and then challengingthe animal with the substance again at a differentsite (11). If the response on retest is substantiallyhigher than the initial response, the agent can beclassified as a sensitizer. Various test methodsand modifications have been proposed for testingsensitizing potential (11, 13-21). Perhaps the mostcommon direct test for sensitization is the guineapig maximization test. In this test, the agent ispresented in Freund’s complete adjuvant whichincreases the response. Studies on the mechanismof sensitization have shown that the agent or ametabolize of it (antigen) may induce the lym-phocytes of the body to form a complex molecule(antibody) which reacts with the antigen to forman antigen-antibody complex. This reaction maybe with circulating free antibody or with lympho-cyte-bound antibody, T h e f o r m a t i o n o f t h eantigen-antibody complex induces the productionand release of histamine and other compoundswhich cause the erythema and edema at the siteof antigen attack, or may cause anaphylaxis if theantigen reaches the blood stream (22,23). Theproblem with testing methods based on thismechanism, such as immunoelectrophoresis ,radioimmunoassay, ring test, hemagglutinationtests, or microphage migration, is that they do notmeasure the actual adverse effect [dermatitis orshock) but measure an indicator response. Themethods are valuable, however, in demonstratingthe presence of antibody capable of producingthese health effects. Several comparative testshave demonstrated that these in vitro techniquesare often more sensitive indicators of the hazardthan the classic in vivo ones (22,23).

Structural EffectsThe basic determination of structural effects

on organs and systems begins with the determina-tion of absolute and relative weight changes.Decreases in absolute body weight or rates ofweight gain for a test with a substance incorpo-rated into the food or water may show either that

the agent is unpalatable or that it is interferingwith the energy balance, the central nervous sys-tem (CNS) regulation of food or water consump-tion, or the motivation of the animal. Althoughsubstances administered by other routes of expo-sure may also interfere with palatability of food,through direct or indirect effects on the sensorynerves, this is less common. Increases in weightmay be caused by a proliferating tumor mass. Theevaluations of structural changes can be obtainedfrom animals exposed during subacute experi-ments.

In general, subchronic or subacute experi-ments are designed to last approximately 10 per-cent of the animals’ lifespan (90 days for rats). Im-mediately preceding and during the experimentalperiod, observations on animals should includerate of growth, food and water consumption, de-meanor, and reflex behavior; blood, urine, andfeces should be collected. During the experimen-tal period, tissue biopsies may be taken for obser-vation of structural changes. These techniquesmay be unrel iable , however , i f a s t ructuralchange is localized and not included in the biopsymaterial, and such manipulation is often notallowed by regulatory testing guidelines. If biop-sies are done, additional animals are required tomaintain the statistical validity of the experiment.At the end of the experimental period, the animalsare sacrificed and the organs are inspected forgross changes, removed and weighed, and pre-served for histologic treatment and microscopicexamination (5). The specific organs and tissuesremoved and examined will depend somewhat onthe expected action of the agent administered(usually perceived from preliminary testing) butshould include at least the brain, liver, kidneys,spleen, heart, testes (and epididymis) or ovaries(and uterus), thyroid, and adrenals.

Changes in organ weights may signal a func-tional change in this organ or in other organs; forexample, an increase in heart weight could be dueto a decrease in oxygen diffusion from the lungs,an increase in adrenal weight could signal ablockage of steroid synthesis within it, etc. Theweights of organs can be directly compared withthose from control animals; however, this often in-troduces an artifact since experimental and con-trol body weights are usually different. It is com-mon practice therefore to determine the relativeweights of the organs in relation to the total bodyweight of the animal. Recently it has been pro-posed that the relative weights should be taken asa function of the animal’s brain weight, the postu-lation being that the brain’s growth curve devi-


ates the least of any tissue in the body. While nor-malization based on this procedure would tend toemphasize changes more than other techniquescurrently in use, it is not yet widely accepted.Tissue dry weight, after desiccation or ashing,has also been used as a tool for determiningmechanisms of growth and metabolic balances(24). This method has a major drawback, how-ever, because it removes the organ from furtherstudies such as microscopic examination.

After gross observation and weighing, theorgans and tissues are preserved for histologicalpreparation and microscopic examination. Themost common methods for the preparation andstaining of individual tissues involve fixing with10-percent buffered formalin solution, embeddingin parafin, and staining with hemotoxylin/eosin,Many pathologists prefer other fixing and embed-ding media, and certain tissues require differentprocedures. There are also special stains forhighlighting different cellular components. Thereis no one best method for preparation and obser-vation of the tissues. The most valuable procedurefrom the pathologists’ viewpoint is to prepare thetissues in a number of ways, which allows com-parison of various aspects such as specific cellu-lar components, nuclei, cell membranes, etc.(25-27).

Special consideration can also be given to tech-niques in histochemistry and electron micro-scopy. These methods are not used routinely intoxicological evaluation and depend on a knowl-edge of the mode of action of the toxic agent. Theycan, however, indicate changes in cellular metab-olism or structure before those changes becomemanifest by the conventional histological proce-dures, and therefore they may be more suitablefor observing changes from agents whose toxici-ties are low or develop slowly. The equipmentnecessary for these techniques is generally moreexpensive than that needed for the more conven-tional microtechnique methods. They are alsomore time consuming and less standardized thanconventional methods. The histochemical meth-ods, although they might be more appropriatelyclassified as tests of organ function, are becomingmore widely accepted with investigators studyingmechanisms of toxic: action.

Functional EffectsFrequently, changes in organ or system func-

tion are observable before any change in struc-ture becomes apparent. The test methods dis-cussed below have generally been adapted fromhuman to animal use, and results are ordinarily

compared with animal control values and are notnecessarily comparable between species, Meth-ods for evaluation of pulmonary function (28), car-diovascular function (29), and brain and neuralactivity (30) have been modified for human andanimal use. These methods include testing ven-tilator flow, resistance, compliance, and gas dif-fusion capacity for the pulmonary system; elec-trical activity of the heart, and blood flow andpressure for the cardiovascular system; electricalactivity of the brain (field and single unit) andmuscles; perception threshold, reflexes, andchronaxy for neural function. The significance ofchanges in brain activity as a determinant of tox-icity is under question at present, however.

Generally, in these evaluations, each animalserves as its own control. Baseline data for eachprocedure is determined prior to administrationof the agent, and any changes in function arenoted during and after administration, since it isimportant to determine whether the agent causesreversible or irreversible changes in function.

Various other methods of testing for organ orsystem function rely on both primary and second-ary parameters. For example, liver function maybe assessed by dye clearance studies (primary) orby analysis of serum enzyme concentrations (sec-ondary). Most of the secondary procedures arenow automated and available through variousclinical laboratories at a reasonable cost, Manyinvestigators, however, still prefer to perform thetests manually, and standard procedures arewell-defined and available in several texts(31-33). Various modifications of these tests forspecific animal systems have been developed andpublished, Tests usually considered appropriateinclude total and differential blood counts, serumenzyme and ion analysis, urinalysis (especially formetabolizes of the agent), and liver and kidneyfunction tests (dye clearance). The value of thesetests is that abnormal results will often precedeobvious structural damage of the organ system inquestion and will be apparent at lower doselevels.

The study of hematologic effects encompasseschanges in the bone marrow as well as those inthe cells of the circulating blood. Observationsare made of the cells and of their absolute andrelative numbers. Specific tests such as dye dilu-tion for blood volume, specific gravity, sedimenta-tion rates, osmotic fragility, hematocrits, or clot-ting time, are not routinely performed but may beindicated. Serial bone marrow biopsies may alsobe performed for hematologic effects; the results

Appendix C—Methods for Toxicologic Test/rig ● 141

give both structural and functional information,but these techniques are also not widely used.

The functions of the liver may be tested forbiliary obstruction (icterus index, alkaline phos-phatase), liver damage (thymol turbidity, plasmaprotein ratios, cholesterol ratio, glucose level,transaminase level, and cholinesterase level), ex-cretory function (bromsulphalein clearance, bili-rubin tolerance), and metabolic function (glucosetolerance, galactose clearance), The most com-mon tests used in toxicology are the serum alka-line phosphatase and serum transaminases, andin some cases a dye clearance (bromsulphalein) orglucose tolerance.

The kidneys are responsible for excretion ofcertain substances, e.g., urea, and for concentra-tion and dilution of urine. Tests for excretion in-volve dyes like phenolsulphonphthalein and alsomeasure such substances as urea and creatinine,Concentration and dilution tests involve measure-ment of urine specific gravity after fasting forvarious periods. These tests are generally notused in toxicology screening studies unless thereis reason to believe the toxicant acts on thekidneys.

The evaluation of these tests may proceed withor without modification of the tissue metabolism.That is, promotors and inhibitors of enzyme sys-tems, e.g., SKF-525A for mixed function oxidase,may be used to enhance the susceptibility of aparticular organ or system to damage from a toxi-cant. This in effect maximizes the response sothat the toxic action can be more readily ob-served.

Many other specific tests are available forevaluating various organs and systems such assperm motility, specific gravity of cerebrospinalfluid. calcium-phosphorus ratios for the skeletalsystem along with tensile strength and compac-tion, epinephrine sensitivity of heart muscle,acetylcholine test of lungs, metabolism of excisedtissue, work and strain measurements of the vari-ous muscle systems, etc.

Behavioral EffectsA recent addition to the field of toxicology has

been behavioral testing. Testing methods havebeen devised for everything from simple percep-tion to complex tasks involving perception, learn-ing, judgment, motivation, and motor activity. Thevalue of the behavioral methods lies in the abilityof the nervous system to respond to toxic agentsa t doses much lower than those necessary to pro-duce “classic” signs of toxicity in the organism.Therefore, these methods are a potential sensitive

indicator of hazard and can be used as an “earlywarning system. ” Several good reviews of behav-ioral toxicology are available (34-37).

Some of these methods rely on newer methodsof analysis such as contingent negative variation(CNV). Some are really an application of preexist-ing principles such as dorsal evoked potentialsand neuromuscular transmission time that havebeen widely used by experimenters in the field ofneurophysiology. Most of these techniques haveonly recently been turned to the evaluation of tox-icity.

At the present time, standardization and vali-dation of behavioral techniques has not been ac-complished. The question often raised by regula-tory agencies is how do you relate an observed be-havioral decrement to an adverse health effect,especially if there is no concurrent structuralchange apparent in the nervous system. Becauseof these factors, behavioral studies are often der-ogated by these agencies when setting exposurelimits for toxicants. Current research is beingconducted, however, under Government con-tracts to answer some of these questions.

Comparison of Short- and Long-TermMethods for Systemic Toxicity

Most of the procedures noted in this section areequally applicable to short- and long-term testing,obvious exceptions being irritation and sensitiza-tion tests. The value of long-term testing for sys-temic toxicity lies in the ability to use low dosesthat do not produce detectable adverse effects ina short time period to see whether bioaccumula-tion and cumulative effects occur. Predictions ofthe effect of bioaccumulation can be made know-ing the effects of short-term high doses, but finalevaluation of the toxicity depends on the long-term effects observed, As will be pointed out inthe metabolism section, toxicants can be poten-tiated or inhibited by the metabolism and relativeaccumulation of the toxic moiety. Without defi-nitely knowing the various metabolic reactions,rates, and probabilities, it is impossible to ac-curately predict toxic effects. High short-termdoses may induce a toxic reaction, such as deathfrom pulmonary edema, that might mask long-term, low-level exposure effects such as livercancer, The differences in short-term and long-term tests for systemic toxicity include the num-ber of interim measurements allowed, the abilityto ascertain the types of effects which mightdevelop only over a long period and the progres-sion or time course of toxic manifestations, and

142 w Environmental Contaminants in Food

the ability to evaluate mechanisms of bioaccumu-lation or adaptation in the organism.

Mutagenicity

Rapid identification of a food contaminant as apossible mutagen is necessary to reduce the po-tential genetic risk to humans who might contactthe contaminant. Mutagenic effects on humansoften cannot be directly detected, and deleteriouseffects on the human gene pool may not becomeapparent for many generations if, for instance,the deleterious effect is due to a recessive gene.Heritable genetic damage in humans may resultfrom any of several types of effects on the geneticmaterial, The two major classes of effects arepoint mutations, which generally affect a singlegene or part of a gene, and more extensive chro-mosomal effects such as gross changes in struc-ture or changes in number.

Only a few tests are available that directlyevaluate genetic effects of exposure of mammalsto chemicals: however, the potential of a chemicalto produce heritable genetic alterations in mancan be evaluated indirectly from its effects on ge-netic material in various biologic test systems, in-cluding micro-organisms, mammalian cell cul-tures, insects, and intact mammals.

For substances that cannot feasibly be elimi-nated from the human environment, it is not suffi-cient to identify the existence of a genetic hazard;quantitative assessment of the risk involved isnecessary for appropriate regulatory activity,such as establishing action levels or tolerancesfor food contaminants.

Mutagenicity testing is also used to prescreenchemicals as an indicator of carcinogenic poten-tial and, less frequently, other toxic effects suchas teratogenicity. This application is based on em-pirical demonstration or correlation between mu-tagenicity and carcinogenicity of chemicals (38)and does not depend on the assumption that thesame mechanism is involved in both types of ef-fect.

Approaches to TestingAs in all toxicological tests, mutagenicity tests

may produce false negatives (a negative resultwhen the substance is actually mutagenic) andfalse positives (a positive result when it is notmutagenic), and correlation between the resultsfrom two test systems may be poor. Ideally, a mu-tagenicity test system should be sensitive enoughto detect any chemical that may cause heritablegenetic damage and its results should be repro-

ducible. Finally, the test results should be quan-titatively applicable to mutagenesis in humans.Since no single test can fulfill these requirementsand none is reliable enough to stand alone as anindicator of mutagenic potential, mutagenicity‘testing should include a variety of systems se-lected to show whether the test substance or itsmetabolize produce any of a range of genetic ef-fects. The test battery approach includes systemsthat will detect several types of gene mutations,chromosomal aberrations, and DNA repair; thus,this approach offers the greatest reliability fordetermining mutagenic potential, Tests that eval-uate effects in intact mammals are essential forpredicting mutagenicity in humans.

Since screening large numbers of chemicals bythe test battery approach (39,40) may be prohibi-tively costly, a hierarchical approach to mutage-nicity testing, known as tier testing (41), has beensuggested. Tier 1 consists of relatively inexpen-sive short-term prescreening tests. These usemicro-organisms or other in vitro systems to de-termine priorities for indepth testing. Substancesthat produce positive results in these tests, aswell as those that are negative but are structur-ally similar to known mutagens or to which thereis a substantial risk of exposure for humans dur-ing or preceding their reproductive years, shouldcontinue into Tier 2.

Tier 2 tests are usually designed to detect sub-stances that are not mutagenic in vitro but aremetabolized to an active form in the intact mam-mal. Tests used at this level may include the domi-nant lethal test, in vivo cytogenetic tests, the host-mediated assay, and body-fluid analysis. Sub-stances that are negative in Tiers 1 and 2 are gen-erally considered safe for use and are given verylow priority for further testing.

Only substances for which it is important to as-sess risk are subjected to Tier 3 testing, designedto permit quantitative evaluation of mutagenic po-tential. Tests used at this level include multigen-eration mammalian studies, such as the heritabletranslocation test, X chromosome loss test, andspecific loci test in mice.

Tier testing may represent an efficient use ofresources in large-scale mutagenicity testing, butthe use of prescreening tests carries serious dis-advantages in determining mutagenic potential, Iftest chemicals are prescreened by a single micro-bial test, the proportion of false negatives may beunacceptably high, and potentially hazardous oruseful substances may escape further testing.The use of two or three tests at this level, in-cluding both micro-organisms and mammalian

Appendix C—Methods for Toxicologic Testing . 143. . . , . . . . . .—- >--

cell cultures with and without activation by mam-malian enzyme systems, may substantially in-crease the reliability of prescreening (96). Never-theless, such cell systems may not approximatemetabolic events in the intact mammal closelyenough to reveal the mutagenic action of somesubstances that are potential human mutagens.

Whatever the testing approach and test sys-tems selected, mutagenicity tests should include apositive control as well as negative (untreatedand solvent) controls. The positive control sub-stance, a known mutagen in animal systems thatis selected for its structural similarity to the testchemical, serves to demonstrate the sensitivity ofthe test organism and the efficacy of the metabol-ic activation system used.

Current Test SystemsChromosomal effects of many substances have

been demonstrated in plants such as Vicia fabaand Tradescantia (42), and the latter organismhas also been used in detection of somatic muta-tion (43). While the genetic events involved (alter-ations in DNA) are the same as those in mammali-an cells, their relevance to human mutagenesishas been questioned because of the major phylo-genetic and physiologic differences betweenplants and animals; thus, a negative result inplants does not indicate that a substance is not amutagen in mammalian systems.

Of the many bacterial species that have beenused to detect point mutations, the most exten-sively employed are the Salmonella typhimuriummutants developed by Ames (44,45). The Amestest uses a series of histidine-requiring mutantstrains that revert to histidine-independence byspecific mechanisms, either base-pair substitu-tions or frameshift mutations, The original strainshave undergone several further modificationsthat increase their sensitivity to mutagens by in-terfering with DNA repair or modifying the cellwall to enhance the penetration of chemicals intothe cell. Bacteria treated with the test chemicalare plated on selective media or cultured in liquidsuspension to determine the number of revert-ants. A reproducible mutation rate twice thespontaneous (control) rate is usually consideredevidence of mutagenic activity.

Because microbial cell systems do not possessthe metabolic capabilities of mammals, they willnot detect chemicals that exert a mutagenic effectthrough metabolic intermediates, Several activat-ing systems have been developed for use with invitro test systems to duplicate the effects of mam-malian metabolism. The most extensively used

means of metabolic activation is the addition ofmicrosomal mixed-function oxidase enzymes, typi-cally from rodent liver homogenates, to metabo-lize the test chemical in vitro, This activating sys-tem is added to the culture medium as part of theAmes testing procedure with S. typhimurium, andit has provided evidence for the mutagenicity ofmany substances that have no direct mutageniceffect on these bacteria (51). Microsomal enzymeactivation is also used with other microbial testsystems (46-48). The major drawbacks of this sys-tem are that it would not detect chemicals metab-olized to mutagenic intermediates by mechanismsother than liver microsomal enzymes, e.g., sub-stances metabolized by the intestinal flora, and itis possible that the in vitro metabolism of thesubstance does not adequately mimic its metabo-lism in the intact organism because of competingreactions. Another drawback is that a standard-ized in vitro activation system has not been de-vised to date.

Another widely used bacterial system is themultipurpose strain of Escherichia coli developedby Mohn and coworkers (49). This strain can beused to measure reverse mutations restoring theability of the bacteria to synthesize the nutrientsarginine and niacin. Forward mutation rates intwo genes controlling galactose metabolism canalso be scored in this strain of E. coli. Use of thistest organism permits the detection of severaltypes of mutation in a single experiment.

Eukaryotic micro-organisms that are used todetect the ability of chemicals to produce pointmutations include haploid strains of the yeastsSaccharomyces cerevisiae and Schizosaccharo-myces pombe (47,5o) and of the ascomycete Neu-rospora crassa (51). A diploid strain of S. cerevi-siae permits detection of chromosomal damageexpressed as mitotic recombination that producesphenotypic color changes (52).

Whole-animal activation mechanisms can cir-cumvent this problem but they are generally muchless sensitive than tests using in vitro activation.In these systems, rodents are exposed to the testchemical by an appropriate route, and the effectof rodent metabolizes on microbial genetic mark-ers is determined by either body-fluid analysis orhost-mediated assay procedures,

In the body-fluid analysis (53-56), the micro-organisms are treated with the urine, blood, orhomogenized tissues of the exposed animals. Cau-tion is necessary in interpreting negative resultsof these studies, in the absence of supplementarypharmacologic data, since even if the chemical ismetabolized to a mutagen, other factors such as


t issue-specif ic act ivat ion and detoxif icat ionmechanisms and the half-life of the compound andits metabolizes may affect test results. In the host-mediated assay (57,58), the micro-organisms areexposed to mammalian metabolic products of thetest substance by being introduced into the perito-neal cavity, circulatory system, or testes of thehost mammal. The host is treated with the testsubstance, and after an appropriate incubationperiod, the indicator organism is removed and ex-amined for mutations.

Genetic damage in micro-organisms can also beassessed indirectly through the use of DNA re-pair-deficient strains of bacteria (44,59). Thesetests organisms and otherwise identical strainsthat have normal ability to repair DNA aretreated with the test substance. Toxic action ofthe test substance produces zones in which bac-terial growth is inhibited, and the difference insize between the inhibition zones in repair-defi-cient and normal strains indicates the extent towhich this toxicity is due to damage to the DNA.

A number of mammalian cell systems in culturehave been developed for detecting point muta-tions, including cell lines derived from mouse lym-phomas, Chinese hamster ovaries and embryos,and human fibroblasts and lymphoblasts (60-63).In addition, gross chromosomal changes such asbreaks, gaps, and rearrangements can be micro-scopically observed in these cells. Stable re-arrangements, such as translocations and inver-sions, are considered evidence of her i tablechanges. The induction of only gaps and breaks isnot regarded as evidence of mutagenicity, be-cause these aberrations often occur as a result ofgeneral cytotoxicity and thus may be present onlyin moribund cells. Like microbial systems, in vitromammalian cell tests can be used in conjunctionwith activation by mammalian enzymes or withwhole-animal activation to permit detection ofmutagenic effects by metabolic products of a testsubstance.

Mammalian cells, including human white bloodcells, are also used to detect chemical damage toDNA by measuring unscheduled DNA synthesis(63). This indirect indicator of genetic damage isevaluated by measuring the uptake of radioactivethymidine for repair of damaged DNA duringthose stages of cell growth when DNA synthesisdoes not normally occur. A similar test withmouse spermatocytes exposed in vitro or in vivodemonstrates effects on DNA in germinal cells(64),

Sister chromatid exchange, a reciprocal ex-change of segments at homologous loci. measured

by autoradiographic methods, has also been usedto examine a variety of chemicals (65,66). Sisterchromatid exchange in various cell systems hasbeen demonstrated following exposure to knownmutagens, but additional work is needed to definethe extent to which results are correlated withmore traditional mutagenicity tests,

The fruit fly Drosophila melanogaster is used ina comprehensive and extensively characterizedmutagenicity test system which can detect alltypes of mutagenic activity at a fraction of thetime and cost of in vivo mammalian testing (67).The large number of genetic markers and knownchromosomal aberrations make it possible toassay a chemical for many types of mutagenic ac-tivity in a single test. The sex-linked recessivelethal test (68) in Drosophila is a very efficientmutagenicity assay, since about 20 percent of theinsect’s genetic material is located in the Xchromosome. Recessive lethal changes caused bypoint or chromosomal mutation can be mappedand in most cases the nature of the change caus-ing the mutation can be determined. This test alsopermits a quantitative assessment of mutagenicactivity. Because of the large size of Drosophilachromosomes, this organism can also be usedreadily to assess meiotic and mitotic recombina-tion, dominant lethality, translocations, and dele-tions. Some indirect mutagens that required meta-bolic activation have been shown to be mutagenicin Drosophila, indicating that these insects have amicrosomal mixed-function oxidase system (69).However, to determine whether Drosophila testresults are useful for risk assessment, more in-formation is needed on how their metabolism offoreign chemicals compares to that in humans. Afew other species of insects have also proven use-ful in mutagenicity testing, including ‘severalspecies of the parasitic wasp, Habrobracon (70).

Clearly the most accurate predictions of muta-genic potential in humans can be drawn fromtests that determine direct genotypic and pheno-typic effects in mammals exposed to the testchemical by routes relevant to human exposures,Several direct mammalian tests exist, but thesehave the disadvantage of detecting only a few ofthe possible types of genetic damage. Chromo-somal damage occurring in vivo can be detectedin several different cell types, such as bone mar-row cells and circulating lymphocytes, and thepresence of micronuclei in red blood cells (7 I).Cytogenetic tests using mammalian lymphocytesand the micronucleus test offer the advantage ofpermitting direct comparison with effects on hu-mans resulting from accidental exposures; how-


ever, these tests demonstrate only effects on so-matic cells and do not provide direct evidence ofheritability.

Cytogenetic changes in mammals can also beevaluated in germinal tissue from the testes. Inthe direct spermatocyte test (72), male mice areexposed to the test substance. After sufficienttime for the treated spermatogonia to reach thespermatocyte stage, they are examined for cyto-genetic abnormalities. This test allows for the ac-tual observation of induced cytologic changes inpremeiotic male germ cells, but it does not permitdetection of effects in postmeiotic cells or theirtransmission to the offspring.

Effects on offspring can be evaluated in mice bythe heritable translocation test (73) and the Xchromosome loss test (74). In the heritable trans-location test, F, male offspring of treated mice aremated to determine sterility, and indication opossible translocation heterozygosity. Chromo-somal effects are then confirmed by cytogeneticanalysis of the germinal cells of the male of offspr-ing. The X chromosome loss test permits thedetection of chromosome loss resulting from non-disjunction in the female, since, unlike somaticchromosome aneuploids, animals of XO genotypeare usually viable, Aneuploidy for the X chromo-some can be detected by genetic markers and con-firmed by cytologic observations.

The dominant lethal assay, usually performedin the rat or mouse, uses fetal loss as an indicatorof induced chromosomal mutations in male ger-minal cells (75). The death of the zygote is assum-ed to result from chromosomal abnormalities inthe sperm of male mice exposed to the test chem-ical. This test is relatively easy to perform and itsresults have been positively correlated with muta-genicity in other animal systems. Preimplantationloss alone is not used as an indication of muta-genicity since it has been found to occur for rea-sons other than chromosomal changes in thesperm. Disadvantages of this test are its relativeinsensitivity and difficulty in clearly distinguish-ing weakly positive results.

Only one test is available at present that candetect heritable gene mutations induced in mam-malian germ cells. In the specific locus assay inmice, forward mutations at seven loci, affectingcharacteristics such as coat and eye color aremated with mice homozygous for recessive allelesat these loci (75). Because such a small number ofloci are involved, this test required the scoring of20,000 to 30,000 offspring at each dose level toproduce reliable results and is therefore verycostly and time consuming.

Carcinogenicity

In the event of massive or long-term environ-mental contamination of food destined for humanconsumption, one of the decisions to be made iswhether the contaminant appears to pose a sig-nificant carcinogenic risk. With this in mind FDAsubmits the candidate compound to the ChemicalSelection Working Group at the National CancerInstitute for consideration under the carcinogenbioassay screening program (76,77).

Prerequisites for a CarcinogenicityStudy

Once the compound has been selected, it isscreened using a chronic or lifetime exposureregimen (76,78,79). However, before the long-termstudy is undertaken, specific toxicologic profilesmust be obtained. Young healthy adult animals ofeach sex and strain to be used in the long-termstudies should be used in the preliminary studies.The animals should be of uniform age and weightand should be tested using the same formulationand route of exposure to be used in the long-termstudies. The first is an acute study designed togain additional information on the acute toxicity,assuming there is a paucity of data on this aspectof toxicity, and to determine the lethality of thetest compound. The duration of this test shouldnot exceed 24 hours and should include at leastthree dose levels determined by a geometric pro-gression. one of the dose levels selected shouldrepresent the highest dose to be used in subse-quent studies. Throughout the investigation, allrelevant clinical signs should be recorded. Ne-cropsies should be performed on a random selec-tion of animals of each sex and strain, and any ab-normal histopathologic changes should be noted.

After the 24-hour study, a 14-day investigationshould be initiated in an effort to ascertain thedoses necessary for the subchronic study, thenext prerequisite investigation for the chronicstudy. This toxicologic study requires five doselevels, with the highest one, estimated from the24-hour acute study, producing no more than 10-percent lethality. The other dose levels shouldrepresent geometric decrements of the highestdose. Animals should be treated with the testsubstance for no more than 14 days, held another24 hours, and then sacrificed for necropsy.Throughout the study, the animals should be ob-served for clinical signs of toxicity. Other toxicitydata, such as those derived from organ functiontests and metabolism studies, are also necessary.


The next toxicity study involves the administra-tion of the test substance for 90 days and is usedas a predictor of the maximum tolerated dose(MTD). This can be defined as the highest dosegiven during a chronic study that can be pre-dicted to not alter the animals’ normal longevityfrom effects other than carcinogenicity, In prac-tice, MTD is considered to be the highest dose thatcauses no more than a 10-percent decrement inweight compared to controls. Five dose levels arerequired in this study, with a minimum of 10 ani-mals of each sex and strain in each dose group.The highest dose level used should be the lowestconcentration that produced any detectable un-toward toxic effects in the 14-day study. The re-maining dose levels should be determined as inthe 14-day study, If the selected dose levels do notproduce a discernible no-effect level, the studyshould be repeated with lower doses.

Carcinogenic BioassayThe chronic study (76,78,79) represents the

essence of the carcinogenicity bioassay. It is usedto determine the carcinogenicity of a compound inmales and females of two mammalian species,usually the rat and mouse. The species selectedare tested throughout their entire lifespan. Eachtest group should consist of a statistically repre-sentative number of animals. The highest selecteddose should represent MTD and the remainingdose levels should be adjusted accordingly. Thereshould be at least one control group, in which theanimals receive only the vehicle used for adminis-tration of the test material, If no vehicle is used,this control group should be untreated but iden-tical in every way to the experimental groups. Inaddition to the concurrent control group, a colonyor historical group should be used for the com-parison of longevity, spontaneous diseases, andspontaneous tumor incidence. The historical con-trol may also be used for statistical comparisons.In some studies, a positive control group that hasbeen treated with a compound structurally simi-lar to the test compound and known to be carcino-genic in the test species may be indicated. How-ever, because of the added risk of handling aknown carcinogen, a positive control group is sel-dom used.

Throughout the study, animals must be ob-served for signs of toxicity. Every animal shouldbe examined carefully each week, Animals shouldbe weighed and food consumption measured. Insome cases, it is desirable to evaluate tissuedistribution and concentration of the substanceor its metabolizes.

The animals in any one test group should besacrificed at an adjusted or prearranged date.However, a group can be terminated earlier ifthere has been high cumulative mortality. Mori-bund animals should be sacrificed immediatelyupon discovery to lessen the likelihood of unob-served deaths and subsequent autolysis or canni-balism. Control groups should be sacrificed ac-cording to the original or adjusted sacrifice date;the later date is preferred. Because of the strongdependency on histopathologic results and toavoid possible criticisms of the study, necropsiesand histopathologic examinations should followstandard procedures required by regulatoryguidelines.

The major drawback to the carcinogenic bio-assay procedure is the time and cost required tocomplete and analyze such a study. If the study isdone properly, however, the results should beconclusive and for the most part indisputable, al-though there still remains the question of extrap-olation of results to the human population.

Short-Term Testing as a Prediction ofCarcinogenicity

Evaluation of carcinogenicity has generally re-lied on the results of long-term animal studies, Touse this kind of testing approach for every sub-stance suspected of being carcinogenic would becost-prohibi t ive and certainly impractical interms of the overall time required to test all sus-pected chemicals. Therefore, short-term tests arebeing developed to identify carcinogenic sub-stances. There has been much criticism concern-ing the comparison of short-term testing resultsfrom different laboratories because of the vary-ing conditions and refinements in techniquespracticed among testing facilities. The protocolsfor these short-term tests, especially those involv-ing mammalian enzyme activation systems, havenot been standardized or validated through inter-laboratory comparative testing procedures; thus,comparisons of the data from one laboratory tothe next have often produced conflicting conclu-sions. In this light, a study conducted in one lab-oratory that compares several short-term testingsystems has added importance in clarifying therelative usefulness of the compared systems,

In a recent study by Dr. Ian Purchase (80), 120organic chemicals (50 known carcinogens and 62noncarcinogens, based on published experimen-tal data) were evaluated for activity in six short-term test systems. These systems included: 1) mu-tation of Salmonella typhimurium (45), 2) celltransformation (81), 3) degranulation of endoplas-

Appendix C—Methods for Toxicologic Testing ● 147.—

mic reticulum (82), 4) sebaceous gland suppres-sion (83), 5) tetrazolium reduction (84), and 6) le-sion formation after subcutaneous implant (85).Four additional tests used in a preliminary studywere found to be insufficiently accurate or sen-sitive to justify a full evaluation. The tests re-jected were transplacental blastomagenesis (86),piperidine alkylation (87], iodine test (88), and theacridine test (89).

Although there were considerable variationsbetween tests in their ability to predict car-cinogenicity, two tests were quite accurate indistinguishing between the known carcinogensand the noncarcinogens. These were the celltransformation test and the bacterial mutationtest, which had accuracies of 94 and 93 percent,respectively. The use of cell transformation andbacterial mutation together provided an advan-tage over the use of either alone, predicting 99.19percent of carcinogens. Not surprisingly, the in-clusion of the other four tests in a screening bat-tery with these two resulted in an improved abili-ty to detect carcinogens (99.97 percent), butgreatly decreased the accuracy and discrimina-tory value of the battery. It is important to notethat all tests generated both false positives andfalse negatives: the percentages for both can bereadily calculated by subtracting the positivepredictability values from 100 percent.

A description of each of the tests, with compar-ative percent accuracies for predicting carcino-genicity as determined by Purchase et al. (80) isas follows:

● Bacterial mutation. The procedures usedwere those of Ames, in which four strains ofS. typhimurium (TA 1535, TA 1538, TA 98, TA100) were tested with each compound in anassay medium containing a metabolic activa-tion system composed of rat liver postmito-chondrial supernatant (S-9 fraction) and co-factors. The overall accuracy of the test inpredicting the carcinogenicity of the com-pounds in this study (80) was 91 percent forcarcinogens and 94 percent for noncarcino-gens. These figures agree with the previouslypublished value of 90 percent by McCann etal. (90) but are considerably higher thanthose published by Heddle and Bruce (91)who found 65 and 81 percent, respectively.

● Cell transformation. The procedures usedwere those of Styles (81) and involved threetypes of mammalian cells—human diploidlung f ibroblasts (WI-38) , human l iver-derived cel ls (Chang), and baby Syrianhamster kidney cells (BHK 21/cl 13). In all

●

●

●

●

assays, the cells were used with and withoutmetabolic activation with S-9 fraction asdescribed previously. Without activation,very few carcinogens transformed hamsteror human cells in the period of study. Withactivation, all cell lines detected carcinogenswith an accuracy of 88 percent or better.Furthermore, by the use of both the hamstercells and either of the human cells, theoverall accuracy was improved to 94 percent(91 percent for carcinogens and 97 percentfor noncarcinogens).Degranulation. The procedure used is that ofWilliams and Rabin (82) and the test is com-monly called the Rabin test. The test meas-ures the loss of ribosomes (degranulation]from isolated rat liver endoplasmic reticu-lum following incubation with the test com-pound. The overall predictive value was 71percent for both carcinogens and noncar-cinogens.Sebaceous gland test. The procedures usedwere those of Bock and Mund (83) where testchemicals were applied directly to the skin ofmice and a depression in the ratio of seba-ceous gland to hair follicles indicates a posi-tive response. The overall predictive value ofthe test was 65 percent (67 percent for car-cinogens and 64 percent for noncarcino-gens).Tetrazolium reduction. The procedures usedwere based on those described by Iversenand Evensen (84). Test solutions were ap-plied directly to the skin of mice and the skinsamples were incubated in tetrazolium redsolution. An increase in the in situ biologicreduction of the colorless tetrazolium to acolored formazan compound indicates a posi-tive response. The overall predictive value ofthis test was 57 percent (40 percent for car-cinogens and 71 percent for noncarcino-gens).Subcutaneous implant. The procedures usedwere essentially - those of - L o n g s t a f f a n dWestwood (85] and involved the subcutane-ous implantation of a filter disc overlaid witha gelatinous suspension of the test c o m-pounds into mice. After 3 months, the sur-rounding tissues were scored for lesions. Theoverall predictive value for this test was 68percent (37 percent for carcinogens and 95percent for noncarcinogens).

While the work of Purchase and associates (80)does present a very salient assessment of themore pertinent in vitro assays, it also points out


the shortcomings of this type of approach. Short-term tests do not use the induction of cancer as anendpoint, but each has a parameter, such as in-duction of a point mutation, that varies with thecarcinogenicity or noncarcinogenicity of the testsubstance. Accordingly, the authors feel thatthese test parameters should not be given agreater weight than that of any other arbitraryresponse, regardless of how biologically signifi-cant any of these tests might appear to be withrespect to the theories of the chemical inductionof cancer. Also, however much generalized datamight be generated to support the predictive ac-curacy of the given test, this accuracy should notbe assumed to apply uniformly to compounds ofevery chemical class,

Teratology and Effects onReproduction

An investigation of a teratogenic agent involvesthe study of congenital malformations other thanthose that are inherited. Teratogens themselvesact as triggers for malformation induction. Mal-formations may include gross, histological, mo-lecular, and behavioral anomalies,

The sensitivity of an animal to a teratogen is de-termined by: 1 ) the period in which the insult is re-ceived during the gestation period (this includesbefore germ layer formation and during embry-ogenesis or organogenesis); 2) the dose and theroute of administration of the compound; 3) pla-cental transfer of the suspect teratogen, includingits lipid volubility, protein binding ability, andmetabolism : and 4) uterine and dietary factors.

A toxicologic profile of a suspect teratogen re-quires an evaluation of potential hazards to re-production and, particularly, to developmentalprocesses that respond to environmental insultthrough mutation, chromosomal aberrations, mi-totic interference, altered nucleic acid synthesis,enzyme inhibition, and altered membrane charac-teristics. This is particularly important in the con-ceptus, embryo, and the neonate where the bio-chemical, morphologic, and physiologic proper-ties change rapidly. Thus risk assessment mustnot only address the teratogenicity of a contami-nant but also its effect on reproduction.

Classical ApproachFor an adequate teratologic assessment (79,92,

93), exposure to the toxicant should parallel asclosely as possible that expected in the humanpopulation. The pharmacologic activity of thecompound as well as its acute and chronic toxici-

ty is also a consideration. For teratogenic studies,the toxicant is usually administered daily on thespecific days of gestation representing the periodof greatest sensitivity. Administration of the testsubstance should begin at or before implantationand should continue throughout the period of ma-jor organogenesis.

The selection of an animal species for evalua-tion of teratogenicity is an important considera-tion. Test protocols currently in use recommendat least two mammalian species, the first being arodent (e. g., mouse. rat, hamster) and a nonrodentmammalian species (e.g., rabbit). One speciesshould be the same as that used in the test forreproductive effects.

In conducting the teratogenic investigation, atleast three dose levels should be used, along withconcurrent control groups. These control groupsshould consist of untreated animals, animalstreated with the vehicle of administration only,and, possibly, animals treated with an agent thatis known to cause the effect that is being in-vestigated. The use of historical or colony con-trols may also be helpful in evaluating the data.Both test and control animals must be young,mature, prima gravida females of uniform age,size, and parity. The control groups should behandled and maintained like the test groups. Thehighest dose level to be considered should pro-duce signs of embryo or fetotoxicity as suggestedby fetal growth retardation and more significant-ly by maternal or fetal mortality: however, mater-nal mortality should not exceed about 10 percent.The other dose levels can be obtained in a de-creasing logarithmic fashion to a suspected no-ob-servable-adverse-effect level. In the classicalteratology study, treatment may be by gavage sothat the administered dose is accurately known. Ithas been shown, however, that use of this route ofadministration may induce anomalies in progenythat are a consequence not of the test compoundbut of the stress of dose administration (94).

Prior to initiating the study, it is important todetermine whether sires and dams have success-fully mated, This includes examinations for thepresence of plugs or evidence of sperm in vaginalsmears. Throughout the study, females should beobserved for behavioral changes, food and waterconsumption, body weight, vaginal bleeding in-dicating possible abortion, and spontaneousdeaths. Females showing signs of aborting or de-livering prematurely should be sacrificed.

Fetuses should be obtained 24 hours before an-ticipated parturition by cesarian section, Damsshould be sacrificed and a complete necropsy per-


test material for at least 120 days. At this pointthey are bred to produce the F, generation.

The types of data to be collected includegrowth and time of delivery for each weanling aswell as overt s igns of toxici ty , the generalbehavior and condition of the mothers, measure-ments of spermatogenesis in all Ff generationmales used to produce the F: generation, littersize, number of stillborn/live births, and any phys-ical or behavioral anomalies.

A statistically valid number of animals (malesand females) obtained from the F, generation andused to produce the F~ generation should be sacri-ficed and examined at the appropriate time, withspecial emphasis on the histopathologic state ofthe reproductive system. In addition, an adequatenumber of weanlings of each sex from each doselevel, including controls, should also be sacrificedand used for histopathologic analyses.

Data derived from the above should be eval-uated for the existence of a relationship betweenexposure and the incidence and severity of ef-fects on reproduction and behavior, tumors, andmortality. The no-observable-adverse-effect levelshould also be determined.

Compared to a teratology investigation, whichrequires about 3 months to complete the exposureand to analyze the results, a three-generationreproduction study represents a considerablylonger expenditure of time, approaching 1.5 yearsbetween the initiation of the study and the com-pletion of the analysis of the gathered data. How-ever, an elementary profile on teratogenicity andreproductive performance can be obtained withina period of a year or less as a part of a continuingl o n g - t e r m t o x i c i t y s t u d y b y a d d i n g t h e a p p r o p r i -a t e n u m b e r o f a n i m a l s a t t h e b e g i n n i n g o f t h elong-term study (2).

Metabolism

Metabolic assessment studies are not directlyused for the assignment of risks, tolerances, or ac-tion levels, These studies are used to determineparameters, such as absorption, distribution,storage, and excretion, that may affect the per-formance of materials in biologic systems, thusenabling the researcher to design testing proto-cols that measure the overall effect of exposure tothe material rather than just measuring a part ofthe biologic response. These protocols can thentake into account such factors as tissue concen-tration, length of time in contact with specificorgans or tissues, ease of reactivity, and the pro-duction of significant (or nonsignificant) changes


in overall body concentrations that may alter theoutcome of a specific testing regimen.

Metabolic study systems are usually centeredaround the concept that the circulatory system isthe major means of transportation of the material,regardless of the route of exposure, Although spe-cific materials may be readily metabolized at ornear the site of entry into the body (e.g., the lungs,skin, and intestines), the majority of materials aretransported unchanged to the liver and then dis-tributed via the circulatory system to other tis-sues and organs. At the same time, some of thematerial may be excreted unchanged in the urine,feces, and air or converted to various metabolizesthat are excreted or bound in the tissues. The pro-portions of the metabolic products that are ex-creted or bound depends on the chemical natureof the original compound, the dose, route of ad-ministration, species, strain, sex, diet, and envi-ronmental factors.

The objective of the metabolic study is to math-ematically evaluate the rates and relative im-portance of these processes in limiting the con-centration of materials in the tissues of the body.For this purpose, the body is usually visualized asa group of pharmacokinetic compartments, a sim-plistic view that surprisingly approximates fairlyaccurately the complex, interdependent proc-esses that actually occur in the body (95). Thesecompartmental models allow the researcher touse measurements of blood concentration as in-direct estimates of tissue concentrations and todetermine the length of time the material remainsunaltered (i. e., the biologic half-life). Rates of ab-sorption from one compartment to another canalso be readily measured, thus giving the re-searcher information on the rate constants of dif-fusion into tissues and on the rates of distributionfrom the blood to various tissues. Clearance orelimination rates can also be determined for theexcretion of the material from the body via thekidneys, gastrointestinal tract, lungs, saliva, andperspiration.

The rate of metabolism of materials depends onmany factors, among the most important of whichare the physiochemical characteristics of themolecule itself. Polar compounds are usually ex-creted very rapidly from biologic systems largelyunchanged because of the chemical activity ofthese compounds, whereas nonpolar compoundssuch as lipids usually remain in the body longerbecause they must be metabolized to polar com-pounds before excretion occurs.

Another important factor in the metabolism ofa compound is its structural resemblance to natu-

rally occurring substances in the body. Foreigncompounds that closely resemble normal bodyconstituents are frequently metabolized by thesame specific enzyme systems that metabolizetheir normally occurring analogues. Most foreigncompounds, however, have no endogenous coun-terpart and must be metabolized by relativelynonspecific enzyme systems. These nonspecificenzymes catalyze many different types of reac-tions leading to a diversity of metabolic products.In general, the nonspecific enzyme reactions canbe categorized into two types, The first includesthe conversion of one functional group intoanother (oxidation of alcohol to aldehyde), thesplitting of neutral compounds to fragments hav-ing polar groups (hydrolysis of esters and amides),or the introduction of polar groups into nonpolarcompounds (hydroxylation). The second type in-cludes the conjugation of the created polar groupwith glucuronate, sulfate, glutathione, or methylgroups to form a soluble, excretable product. Theproduct formed in the first reaction may be eithermore or less toxic than the parent compound ormay possess a different type of toxicity. Often, itis this product that actually causes the toxic ef-fects, including cancer, mutations, cellular necro-sis, hypersensitivity, fetotoxicity, and blood dys-crasias. A portion of the chemically reactivemetabolize formed becomes bound to tissue mac-romolecules, such as cellular proteins, DNA,RNA, glycogens, or lipids. In this way, it disruptsthe normal function of the macromolecule, caus-ing adverse biologic effects. In contrast, the ma-jority of the second type of reaction products areusually either nontoxic or considerably less toxicthan the parent compound,

The metabolism of a foreign material is con-trolled by enzymes, and any factor which affectsthese enzymes also affects the metabolism of thecompound and consequently its toxicity. The me-tabolism of a compound may be inhibited or stimu-lated by the presence of competing substrates.Pretreatment with drugs, steroids, food additives,pesticides, polycyclic hydrocarbons, polycyclicamines, and normal constituents of food can re-sult in an increase in the activity of enzymes thatmetabolize foreign compounds. This increase inenzyme activity differs according to the inducingchemical, but is mediated through an increasedrate of synthesis of enzyme protein, a decreasedrate of turnover of enzyme, or an activation of en-zyme, possibly by changes of structure or con-formation. Because of this phenomenon, chronicadministration of a foreign material may enhance


the activity of the enzymes that catalyze itsmetabolism.

Several types of biologic phenomena are readi-ly studied using a metabolic test system. These in-clude activation, antagonism, synergism, and po-tentiation. Activation has been briefly touched onabove with the description of the process of tox-ification: it involves the rendering of an inactivemolecule into an active molecule, usually throughthe removal or substitution of a neutralizing fac-tor attached to the molecule, but it may also in-volve the direct addition of a constituent to themolecule. One of the most common examples ofactivation, the conversion of a precarcinogen to acarcinogen, is the addition of oxygen molecules topolycyclic aromatic hydrocarbons via the micro-somal NADPH-dependent cytochrome P-450mixed function oxidase enzyme system. This con-version to the oxide produces a carcinogenicagent of considerable potency, whereas the par-ent material does not possess a direct carcino-

APPENDIX

1.2.

3,

4.

5.

6.

7.

8.

9.

10.

11,

12.

13.

14.

15.16.

41 CI+’R 51206 (1976).Food and Drug Administration Advisory Commi[-tee on Protocols for Safety Evaluations: Panel onReproduction Report on Reproduction Studies inthe Safety Evaluation of F’ood Adciit ives and Pest i-icicle Residues ( 1970). Tox, App). Pharm. 16:264.A40dern Trends in Toxic~Jlogy ( 1968). E. Eloyiandand R. Coulding, ecis,, But terworths, London.Methods in ‘1’oxicolog}’ ( 1970). G, E. Pagett, cd.,Davis, Philadelphia.Frawlev, J. P. ( 1955). Fooff Drug Cosmetic L(IW J.1 o: 703.Sm}th, 1{. I+’,, Jr,, an(i (;. P . Carpenter (1944) . J.Jnd. Hyg. (Jnd Toxicd. 26:269.Friedenwt~ld, J. S., W. l“, Hughes, Jr., and H. Her-mann ( 1944). Arch. of Oph th. 31 :279.Carpenter , C. P., and H. F, Smyth, Jr, (1946). A m .J. Ophth. 29: 1363.Draize, J. H., G. Woodard, and H. O. Calvery[ 1944),]. Phc~rm. find Expt. “rherc~peu. 82:377.I)raizc, J. 11., and E. A. Kelly ( 1952). Proc. of theSci. Sec. of Toilet G(MN]S Assoc. 17:1.Ilraize, J. H. ( 1955). Foo{j Drug Cosmetics Low ].10:722.Draize, J. 11. ( 1965). Assoc. of Food and DrugOf ficicds ~)f the U. S., Topeka. Kans., 46.Lansteiner, K., and J. Jacobs (1935). ]. Exp. Med.61: 643.Lansteiner , K,, and J. Jacobs (1936). ]. Exp. Meal,64:625.Buehler, E. V. (1!365). Arch. Derm, 91 :171.Baer, R. L., S. Rosenthal, and C, J. Sims [1956). J.Invest. Derm. 27:249.

genie potential per se. The determination of meta-bolic activation can permit the experienced re-searcher to predict the kind of adverse effect thatis likely to be elicited from the parent molecule,thus enabling the researcher to better design ex-periments to observe these effects.

From a properly designed and well-carried-outmetabolic study, the researcher can gain valuableinsight into the potential toxic actions and bio-logic effects to be expected from a foreign com-pound. If a compound is not absorbed or if thecompound is destroyed or rapidly eliminated fromthe body there is little likelihood of pronouncedtoxic effects. If the principle metabolic productsare polar, conjugation (and thus elimination)rapidly occurs and again there is little potentialfor pronounced toxicity. On the other hand, if me-tabolism produces an activation product, toxicityis enhanced and biologic effects may be pro-nounced.

REFERENCES

17.

18.

19.20.214

22.

23.

24.

25.

26.

27.

28.

29.

30.

31,

32.

Llarzulli, F. N., and Ef. 1. klaibiich ( 1970). J. Soc.Cosmet. Chem. 21 :695.h4aguire. H. C., Jr, ( 1973). ]. Soc. Cosmet. Chern.24: 151.Levine, B. B. ( 1960). ~. Exp. A4efi. 112: 1131.Salvin, S. B. (1965). Fed. Proc. 24:40.Turk, J. L. (1964). Int. Arch. Allergy 24:191.Bio]ogy of the Immune Response ( 1970). P, Abro-noff and hf. LaVie, eds., hlcGraw-Hill, N.Y.Textbook of Immunop(lthologv ( 1976). H. G. hlLLl-ler-Eberhard, cd., Grune and St rat ton, N. Y..Medicul Physiology ( 1968). V. B. hlountcastle, cd.,hfosby, St. Louis.A Textbook of Histology ( 1964). W. Bloom and D.W. Fa~cet t, eds., Saunders, Philadelphia.Pfltholog}r (1971). L$’. A. D. Anderson, cd., Mosbv,St. Louis,Nelson, A. A. [ 1955). Foo(j Drug Cosmetic L~Jw j.10:732,Comroe. J. H., Jr,, R. Il. F“orster II, A. B. Dubois, W.A. Briscoe, and E. Garleson ( 1962). ‘rhe Lung: (lin-ic{d Physiology (lntl Pulm~~n(] r-y Func ti(~n Tests,Yearbook hfedical Publ., (;hicago.Armst rong, hl. L. ( 1974). Ii~[~ctr[)cf)r-( ii(~gr{lms,Yearbook hledical Publ., Chicago.Current Practice of Clinical El[~ct~()(>rlc[?l~h~)l()g-r(lphy ( 1977). D, W’. Klass and D. D. Daly, eds.,Raven.Goodale, R. 11. ( 1965). Clinicul In terpre ta t ion o.fLahor~ltory Tests. Davis, Philadelphia.Henry, R. J. ( 1974). (linic(d (;hernistry: Principlesan{j Techniques, Harper hledical.


33.

34.

35.

36.

37.

38.

39.

40,41.42.

43.

44.

’45.

46.

47.

48.

q:] .

50,

51.

52.53.

54.55.

56.

57,

58,

Amuino , J. S, (1976). Clinical Chemistry: Prin-ciples and Procedures, Little.Xintaras, C., and B. L. Johnson (1976). Essays inToxicology, W, J. Hayes, cd., 7:155, AcademicPress, New York.Ekel, G. J., and W. H. Teichner (1976), A nAnalysis and Critique of Behavioral Toxicology inthe U. S. S. R., 13 HEW (NIOSH) Publication No.77-60, Cincinnati.Behavioral Toxicology (1974). C. Xintaras, B. L.Johnson, and 1, DeGroot, eds. DHEW (NIOSH)Publication No. 74-126, Cincinnati.Weiss, B., J. Brozek, H. Hanson, R. C. Leaf, N. K.Mello, and J. M. Spyker ( 1975). Principles for Evu]-uu ting (Jhemicais in the Environment t, 198 Na-tional Academy of Sciences, Washington, D.C.Alc(; arm, J., and B. h!. Ames ( 1976). Ann. fV. Y.Ac{](I. Sci. 271 :5.Bridges, B. A. ( 1973). Environ. Heo]th Perspect6:22 1,Bridges, B. A. ( 1974). Mu(. H[?s, 26:335.b’lamm, W. G. ( 1974). h’fut. Res. 26:329.Ehrenberg, L. ( 1973). In: Chernicul Mu ta~ens:Princi{j]es (In(i M[~lh(xls for Their Detection, v o l .2. A. }Iollaender, d., Plenum Press, New York.Underbrink, A. E., L. A, Schaierer, and A. H.Sparrow (1973), In: Chemical Mutagens: Princi-ples and Methods for Their Detection, vol. 2, A.Hollaender, cd., Plenum Press, New York.Ames. B. N.. F. 1). Lee, and W. E. Durston (1973).Pr(m. lV(It, Ac(Id. Sci., U.S.A. 70:782.Ames, B, N., J. hlc(hnn, and E. Yamasaki ( 1975).Mut. ~[?S. 31 :347.I.[]ishes, B. A. , and 11. E’. S[ich ( 1973). Biochern.Biophys, Res. [lmlmun. 52:827.Brusick, D. J., :]nd V. W“. Nfavcr ( 1 9 7 3 ) . Env.He({lth Persp[’c[. 6:83.ong, ‘1”-hf, []nd 1{. V. ifal]in~ ( 1 9 7 5 ) . M u t . lies.31 : 195.hlohn. {;.. J. Ell[?nlwrger, :]nd D. hlc~re~or ( 1974).MU (. 1{[]s. 25: 187.Mortimer, R. K., and T. R. Manney (1971). I nChemicul Mutagens: Principles and Methods forTheir Detec~tion, vol. 1, A. Hollaender, cd., PlenumPress, New York.De%rnes. ~’. J., a n d If. V. hl:]lling ( 1 9 7 1 ) . I n :[;hcmi[;(il MU tflgens, Principles (]n(i Meth(xls forTheir Detecti{jn, vol. 1, A. 1 Iollaender, cd., PlenumPress, New York.Z i m m e r m a n l+’. K. ( 1975). Mut RCS. 3 1:71.

Gabri(ige, Ll, C., A. L)eNunzio, and M. S. Legator( 1969).N(ltur[?68:221,Siebert, D. ( 1973). Niut. Res. 17::~07,Durston, M’. ~;., fln(~ Il. N. Ames (1!174). Proc. Nut.Ac(I(I, Sci. USA 71 :737.Commoner, B.. A . J . frith:lv;ithil. and J . H e n r y( 1974). N(lt urc 249:850.Brew en, J. (;., P . Nettesheim, i~nd K . P . J o n e s( 1970), Mu(. 11(:s, 10:645.I,eg[}tor, hf. S., (l nd I{. ~’. hi;} II ing [ 1971 ). In: Chemi[;(li Jlfu t(]gens: Prin(; iples (ln(l Metho(ls for Their

59.

60.

61.

62.

63.

64.

65.66.

67.

68.69.

70,

71.72,

7:3.74.

75.

76.

77.

78.

Detection, vol. 1, A. Hollaender, cd., PlenumPress, New York,Slater, E. E., M. D. Anderson, and H. S. Rosen-kranz (19i’1). Cuncer Res. 31 :97o,Chu, E. H. Y. ( 1971). In: Chemicul Mutagens: Prin-ciples and Methods for Their Detection, vol. 1, A.Hollaencier, ed., Plenum Press, New York.Kao, F-T, and T. T. Puck (1974). Methods Cell BioJ8:23.Cline, D., and J-A F. S. Spector (1975). Mut. Res.31: 17.Regan, J. D., and R. B. Setlow ( 1973). In: ChemicalMutugens: Principles and Methods for Their De-tection, vol. 2, A. Hollaender, cd., Plenum Press,New York.Sega, G, A., J. G. O w e n s , and R. B. Cumming(1!276). Mut. ~t?S. 36: 193.‘1’aylor, J. H. ( 1958). Genetics 43:5 15.Allen, J. W., and S. A. Latt ( 1976). Nature260:449.Abrahamson, S., and E. B. Lewis (1971). In: Chem-ical Mutagens: Principles (]n(f Methods for TheirDetection. vol. 1, A. Hollaender, eci., P l e n u mPress, New York.Vogel. E., and El. Leigh ( 1975), Mut. Hes. 29:383.Casida, J, E. ( 1969). In: Micr-osomes (Infl Drug Oxi-du tion, J. R. Fouts and G. J. hlannering. eds., Aca-demic Press, New York.Smi th , R. 11., and R. ~;, von Borstel ( 1973). In:(jhemic~d Mut(],gens: Principles und Methods forTheir Detection. vol. 2, A. Ilollaender, ed., plenumpress, New York.!+:hmid, W. ( 1975). Mut. R[?s. 31 :9.[,eoll:lrd, A. ( 1 g73), In: Chemical Mutugens: Prin-ciples un[j Meth(~(js for Their Deter; tion, vol. 2, A.}Iollaender, cd., Plenum Press, New York.I,eonard, A. ( 1975). Mut. HCS. 3 I :291.Russell, 1,. B. ( 1976). [n: (11[’mi~;~ll Mut{lgens:Pr inciples (ln(j Meth(xis ff~r Their fletection, vol.4, A. Ilollaencier, cd., Plenum Press, New York.Batem~~n A. J. :~rld S. S. ~PStein ( 1 9 7 1 ) . In:Chemic~lf Mut~lgens: Principles (In(i Metho(is forTheir Detection, vol. 1, A. Ilollaender, cd., PlenumPress, New York.Sontag, J. N1., hf. P, Page, and U. Saffiotti ( 1976).I n : Cui(ielines for Curcin(~gen Bimlss[]y in Sm(dlHo(jen ts. U.S. Department of Health, Educationand Welfare, Public Health Service, National In-sti tutes of [iealth. National Technical In forma-t ion Service PB-264 061, Springfield, Va.Saffiotti, [J., :]nd 11. Autrup ( 1978). In: In Vitro(1] rc;inogenesis; Guide to t h e Liter(lture. A(l-v(lnces (ln~l Lahor(ltory Procedures, U.S. Depart-ment of I Ieal t h, Education and Welfare, Public}iealth Service, National Institutes of Health. Na-tional Technical Information Service PB-281 737,Springfield, Va.(~arcinogenicit y ( 1973). In: The Testing of Chem-ic(ils f~~r t~cll.[;inf~g(~nic;ity, Mu t(lgenic;ity, (Ind ‘I’eru -togenicity, Nlinistry of Ilealth a n d W e l f a r e ,(;anada .

Append/x C—Methods for Toxicologic Test/rig ● 153

79.80.

81.82.

83.

84.

85.

86.

87.

88.

89.

43 (;F’R 37336 ( 1978).Purchase, I. F. H., E. Longstaff, J. Ashby, J. A.Stvies, D. Anderson, P. A. Lefevre, and F, R.Westwo(d ( 1978). Br. /. (hncer (in press).St Vlcs, J. A. ( 1977), Br. ], Cancer 36:558.VL’illiams, I). J., and B. R. Ral)in ( 1971). Nclture232: 102.Bock; h’. 11., and R. Mund ( 1958), (llncer Heseurch18:887.Iverscn, (), }1.. and A. Evensen ( 1962). In: F:XJX>ri-men t~~~ Skin (l] rein ogencsis in Mice, NorwegianIJniversit y Press.I,ongst{]ff. E,, and F. R. L1’estwood ( 1978). [Jnpub-lishe(i.DiPaol:), J. A., R. L. Nelson, P. J. Donovan, and C,11, Ev:)ns ( 197:3). Archs. P(]thol. 95::380.Epstein, J., R, W’. Rosenth:]l, and R. J. F;ss [ 1 9 5 5 ) .Anfll. fllcrn. 27: 1435.Szcnt-GVorg yi, A., 1. Is(:nberg, and S . IJ. Baird( 1960). f%(~c. N(lt. Ac~Ici. Sci. USA. 16: 1444.Szent-Gvorg}’i, A., :]n(i J. hlcLau8hlin (1!361 ). Prf)c.Nflt. Ac[I~l, Sf;i. 11. S.A. 47: 1397.

90,

91.

92.

93.

94.95.

96.

McCann, J., and El. N . Ames ( 1976). Proc. Nat.Acud. Sci. U.S.A. 73:950.IIeddle, J. A., and W . R. Bruce ( 1977). In: Originsof Humun Cflncer. 13(x)k C. Human Risk Assess-ment. H. H. Hiatt, J. D. VL’atson, and J. A, W“insten,eds. Cold Spring Harbor 14abor a torv.‘reratogenicity (1973). In: The ‘1’[?s ting of Chem-iculs for C(1 rcinogcn ici ( 1, M u t(lgf’n ici ty, (ln(i Terf I-tr)genicity, Ministr>r of Ile:]lth an(i L’1’elfarc,Canada.Courtney, K. D., and hi. ( ;h[?rnoff ( 1974). In: ‘~rf]in-ing M(lnuol for 7’er[l t{jl~)g}’. office o f R e s e a r c hand Development, U.S. Environmental ProtectionAgency, Research ‘1’ri:]n~le Park, N.C.Green, E. L. ( 1962). Gcnctif;s. N.Y. 47: 1085.Compartment ts, Pools (~n(i Sj)fJces i n MwliccdPhysiol[~g~ ( 1967). P.-Ii. E. [3ergner ;ind f;. C. ],ush-baugh, eds. U.S. Atomic F;nergy (X)mmission, Divi-sion of ‘1’echnicnl In forma t ion, 0:] k Rid8e, ‘1’enn.Purchase, I. ( 1976). N(It ure 264:624.

Review and Evaluation

Appendix D

Methods ofDetermining Risks From Chronic Low-

Level Carcinogenic Insult *by Kenny S. Crump and Marjory D. Masterman

BASIC PRINCIPLES AND CONSIDERATIONS

To aid in determining a proper regulatory ac-tion regarding a carcinogen that is present inman’s environment, whether it be a feed additive,industrial pollutant, or otherwise, it is helpful tohave some knowledge about the number of extracancers that are likely to be caused by the pres-ence of the carcinogen in the environment. It isalso helpful to have some knowledge of the likelychange in number of extra cancers that would ac-company some projected increase or decrease inthe level of human exposure occurring either as aresult of regulatory action or inaction. This kindof information is usually impossible to obtain di-rectly from human data. For this reason it is oftennecessary to use data from animal feeding experi-ments to estimate human risk. This procedure in-volves two difficult steps: 1) relating the animalrisk at high doses to doses very near to zero and 2)relating the animal risk to risk in humans.

Typically, animal experiments use on the orderof 100 animals at each experimental dose. If aparticular experimental dose causes a lifetime in-crease in cancer risk of 1/10, this increase can bemeasured with a small degree of accuracy using100 animals. But if the increased cancer risk isless than 1/100 this increase will often not even bedetectable by an animal feeding experiment. Forexample, if the true risk is 1/100 it would requirethat over 400 animals be tested at that dose in

order to be 99-percent sure of detecting any car-cinogenic response at all (i. e., for there to be aprobability of 0.99 that at least one animal getscancer). If background or spontaneous carcino-genesis is present even larger numbers of animalswill be required, On the other hand, the extra hu-man risk that we may want to estimate resultingfrom environmental exposure is usually (andhopefully) smaller than 1/100 for any given chem-ical, perhaps on the order of 1/1,000,000. It isclear that it would not be practical to conduct anexperiment with enough animals to measure di-rectly an increase in risk this small.

For these reasons the procedure has been de-veloped of conducting lifetime animal feeding ex-periments using, in addition to a control dose ofzero, several doses at which the projected extracancer risk may be 1/10 or larger, These high-dose data are then used to estimate the extra riskat a dose where the extra risk may be no largerthan, say, 1/1,000,000. An equally important vari-ant to this problem is the calculation of the so-called “safe” dose, that is, a dose for which thereis some measure of statistical assurance thatthe extra risk at that dose is no more than, say,1/1,000,000 These problems are often referred tocollectively in the literature as the ‘‘low-dose ex-trapolation problem. ”

Performing a low-dose extrapolation involves the choice of statistical procedures to apply to thethe choice of a mathematical function to model mathematical function. The choice for this mathe-the dose-carcinogenic response relationship and matical function turns out to be extremely crucial

to the outcome of low-dose risk estimation. If the*Excerpt from OTA Working Paper entitled “Assessment of

Carcinogenic Risks From PCBS in Food.”’ A complete copy ofassumed relationship between tumor occurrence

the paper can be obtained from the National Technical Infor- and dose does not apply in the regions to whichmation Service. (See app. J.) the extrapolation is being made, a serious over-

154

Appendix D—Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult ● 155

estimate of the “safe” dose may result (Manteland Bryan, 1961, p. 458). Chand and Heel (1974)compared five standard dose-response modelsand observed that they could differ by manyorders of magnitude at low dose levels for whichextra risks are on the order of 1/100,000,000.

It might be supposed that it should be possibleto discriminate among the various potential dose-response functions on the basis of experimentaldata but, unfortunately, two different dose-re-sponse functions can often fit experimental dataequally well but still differ by several orders ofmagnitude at very low doses. Moreover, even if aparticular dose-response function were to give a

significantly better fit to data than several othersthis would still not furnish assurance that thisfunction would necessarily correlate in any waywith the true dose response at very low doseswhere it is not feasible to measure the true extrarisk directly. As a consequence of the great dis-parity of dose-response functions at low doses it isimperative that the dose-response function be se-lected, neither arbitrarily nor solely on the basisof how well it can be made to fit experimentaldata, but, insofar as is possible, it should reflectknown or at least plausible information regardingthe biological mechanisms through which a chem-ical induces or promotes cancer.

WHAT SHAPE SHOULD BE EXPECTED FOR THE DOSE-RESPONSECURVE AT LOW DOSES?

Tumors of so many different types arise in sucha diversity of different tissues, their etiology is sol i t t le understood, and the agents that causetumors affect a subject in such diverse ways, thatit might seem that no general conclusions can bedrawn. However, for a certain broad class of “di-rectly acting” chemical carcinogens the range ofuncertainty associated with the shape of the dose-response curve at low doses can be greatly nar-rowed. As used in this paper, the term “directlyacting carcinogen” encompasses (Guess, Crump,and Pete, 1977) carcinogenic agents for whicheither the agent itself or a metabolize acts directlyat the cellular level and produces a heritablechange that eventually leads to the formation of atumor. Carcinogens that are carcinogenic by rea-son of their mutagenicity should fall into the cate-gory of “directly acting carcinogens. ” According-ly, carcinogens that test positively using the Amesmutagenicity screening test for carcinogenicityare very likely to be directly acting (see McCannand Ames, 1976). In a recent study (McCann,Choi, Yamasaki, and Ames, 1975) in which about300 carcinogens and noncarcinogens were testedusing the Ames test, 90 percent (157 out of 175) ofthe carcinogens were mutagenic including almostall of the known human carcinogens. This indi-cates that the class of directly acting carcinogensmay encompass most of the known carcinogens,

A partial solution to the low-dose extrapolationproblem for the case of directly acting chemicalcarcinogens has been given in Peto (1977), Crump,Heel, Langley and Peto (1976), and Guess, et al.(1977). The key result is that, at least as long asbackground carcinogenesis is present, we shouldexpect the dose-response curve not to be absolute-

ly flat at zero dose. What this means is simply thatwhen risk is plotted against dose response on or-dinary linear scales, the tangent line to the dose-response curve at zero dose should have a posi-tive slope. When a dose-response function hasthis property we will say it is linear at low dose.This simple property can have far-reaching con-sequences on low-dose extrapolation. For exam-ple, consider the two potential dose-responsefunctions:

Both of the curves give a risk of 1/10 at a dose of d= 1 and are practically indistinguishable athigher doses. However, at a dose of d = 1/1,000:1) predicts a risk of 1/100,000 and 2) predicts arisk of 1/10,000,000, a difference of two orders ofmagnitude. We note also that 1 ) has a tangent linewith a positive slope at d = O whereas 2) does not.

One explanation of why the dose-responsefunction should be linear at low dose when back-ground is present may be found in Crump, et al.(1976) and Peto (1977) and will be briefly outlinedhere. When background carcinogenesis is pres-ent, the cellular mechanism through which thetest agent produces cancer should already be op-erative in producing background tumors. Whenthis is true the effect of the test agent is to add toan already ongoing process. The result of thisadditive effect is illustrated in figure D-1. Thedose-response curve is for all tumors producedthrough the mechanism through which the testagent acts. Background carcinogenesis is allowedfor in the figure by an effective background dosedo. We see that, in this case, the added riskcaused by a dose d of the test agent should be ex-


pected to increase approximately linearly near d= O (i.e., the tangent line at d = O will have a posi-tive slope). Implicitly assumed by the way figureD-1 is drawn is the fact that an added dose of acarcinogen acting through this mechanism doesnot produce a smaller risk, If background carcino-genesis is allowed for as in figure D-1 by positingan effective background dose do that is estimatedfrom the data, then the wide range of risks ob-tained using different models effectively disap-pears (Pete, 1977). We note that the existence of atangent line with a positive slope at zero dosedoes not, in itself, imply any lower bound for extrarisk at low doses since the slope of the tangentline could possibly be very small.

The evidence given above for a positive slope tothe dose-response function at zero dose appliesparticularly to the case in which background car-cinogenesis is operative. This does not imply thatwe expect the dose-response curve not to be lin-ear at low dose in the absence of background car-cinogenesis. For example, the multistage modelsof cancer (Armitage and Doll, 1961) are a fairlybroad class of models in which it is assumed that

Figure D-1 .—Illustration of Why the Dose” ResponseCurve Should Be Linear at Low Dose in the Presenceof Background Carcinogenesis. dO Is the Effective

Background Dose and d Is the Dose of theCarcinogen of Interest

at d =0

a number of events are required to occur at thecellular level to initiate cancer. Although all mod-els in this class are linear at low dose providedbackground carcinogenesis is present, a sizablesubclass of them are linear at low dose in theabsence of background carcinogenesis (Crump, etal., 1976).

Watson (1977) has recently proposed a morespecialized model for cancer induction and pro-motion based on reversible epigenetic cellularchanges. Watson concludes that his model sup-ports the low-dose-linearity hypothesis and states“as suggested by a different argument of Crump,et al. . . . it is reasonable to assess the risk due toan additional carcinogen at low constant dosageby a linear relation. ”

The evidence for low-dose linearity given aboveapplies mainly to directly acting carcinogens. Anindirectly acting carcinogen might be one thatcauses some gross physiological change such assuppression of ovulation that could predisposethe subject to cancer. For such carcinogens theshape of the dose-response curve at low dose ishighly speculative. There could possibly be athreshold dose below which the agent has no car-cinogenic effect at all on an individual, Howevereven if a threshold mechanism is operative, thereis likely to be considerable variation in individualthresholds in a large population. Consequently thedose-response curve for the entire populationcould still exhibit a linear trend at risks as low as1/1,000,000 or lower.

The effects of metabolic activation and detox-ification on carcinogenic dose response have beenrecently considered by Cornfield (1977) through akinetic model that encompasses free toxic sub-stance, metabolize, deactivator, and the interac-tions of these substances. Only a steady-state sit-uation is studied in that variation over time of theconcentrations of these agents is not considered.The model predicts a threshold dose below whichthere is no carcinogenic risk under the assump-tion that the deactivator is 100-percent efficientin deactivating the carcinogen. However, in a nat-urally occurring process it is likely that deactiva-tion would not be perfect and would be less than 100-percent effective in always combining with100 percent of the carcinogen before an amountof the active metabolize reaches a cancer targetsite, Any of a number of modifications to themodel to allow for nonperfect deactivation wouldrule out a threshold and would lead directly to amodel for which carcinogenic response varieslinearly with dose at low doses, Cornfield’s ownmodification of perfect deactivation, that of al-


lowing the deactivating reaction to be reversible,leads, as Cornfield points out, to a model which islinear at low dose. This occurs regardless of howslowly the reverse reaction takes place, as long asthe possibility is not eliminated entirely. Further-more, even in the extremely unlikely case ofperfect deactivation, an otherwise realistic modelshould still imply low-dose linearity since thetheoretical time required for perfect deactivationwould not be zero and would likely be infinite.

For most, perhaps all, carcinogens, the mecha-nisms through which cancer is produced are notsufficiently understood so that the shape of thecarcinogenic-response curve can be theoreticallypredicted with certainty. As pointed out earlier,neither can experiments of sufficient size be con-ducted that would permit direct experimental in-vestigation of the dose-response curve at lowdose. We have noted that there are plausible ar-guments that the dose-response curve is linear atlow dose for many carcinogens. On the otherhand, this author knows of no serious proposal of

a mechanism that would lead to a more conserva-tive dose-response relationship such as the riskvarying approximately as the square root of doseat low dose. In view of these uncertainties itwould seem reasonable to base estimates ofadded risk of cancer on a mathematical modelthat encompasses low-dose linearity unless, ofcourse, the mechanism through which the car-cinogen operates is sufficiently understood sothat low-dose linearity can be conclusively ruledout. Once the principle of low-dose linearity is ac-cepted the problem of estimation of risks at lowdose is nearly solved. This is because the dis-agreement between the upper statistical con-fidence bounds on risk at low doses based on amodel that incorporates low-dose linearity andone that does not is typically several orders ofmagnitude whereas the corresponding disagree-ment between two reasonable models both ofwhich incorporate low-dose linearity is usuallymuch less than this.

Mantel= Bryan

The Mantel-Bryan procedure as originally pro-posed (Mantel and Bryan, 1961) and “improved”(Mantel, Bohidar, Brown, Ciminera, and Tukey,1975) is for the purpose of conservatively choos-ing a “safe” dose of a carcinogen, a “safe” dosebeing defined as one for which it can be expectedthat, with a given level of statistical assurance(e.g., 99 percent), the true dose producing a pre-assigned “safe” level (e. g., 1/1,000,000) of riskwill lie above the “safe” dose. In the Mantel-Bryan procedure the mathematical model usedfor the dose-response model is the probit function:

where d represents the dose of the carcinogenand P(d) represents the probability of a cancerousresponse in an animal subjected to a dose d. Theparameters in the model are an intercept parame-ter a, a probit slope parameter b, and C, whichrepresents the probability of a response in un-treated animals. The parameter b is not estimatedfrom the data but rather is arbitrarily set equal to1. This choice is stated as being conservative(Mantel and Bryan, 1961), the argument for thisbeing that typical dose data exhibit a probit slopein the experimentally observable region above 1 -percent incidence that is greater than one, InMantel and Schneiderman (1975) it was observed

that a set of DES data (Gass, Coats, and Graham,1964, C3H females) exhibited a probit slope ofone-half, but the general use of a probit slope of b= 1 was still suggested.

With the probit slope parameter fixed at b = 1the remaining parameters, a and C, are estimatedfrom the data and then adjusted so as to producea higher level of risk at a given dose that corre-sponds to an upper 99-percent statistical limit onthe true risk at a given dose. The safe dose is thendetermined to be the one producing a given lowrisk (e. g., 1/1,000,000) based on the adjusted val-ues of a and C.

As pointed out in Mantel, et al. (1975), theMantel-Bryan procedure rewards larger and bet-ter experiments in that the more evidence there isof safety, the higher the calculated safe dose willbe. However this advantage should be shared byany extrapolation method that uses reasonablestatistical procedures.

Some have considered the Mantel-Bryan proce-dure to be too conservative (Federal Register, vol.42, 1977, p. 10419) in that it involves three con-servative choices (99-percent statistical assur-ance, lifetime risk of 1/1,000,000, and probit slopeset equal to 1) and that any one of these assump-tions alone could provide adequate protection tothe public. The first two of these choices are regu-latory decisions that would have to be made withany extrapolation procedure. However, the arbi-


trary selection for the slope parameter seems to . response relationship does not apply at the lowbe peculiar to the Mantel-Bryan procedure. To in-vestigate its effect upon the extrapolation proce-dure a typical fit is presented in figure D-2 (fromCrump, 1977b) of the Mantel-Bryan probit curve(equation 1) to experimental carcinogenicity datawhen the probit slope parameter is fixed at b = 1.As can be readily seen the probit curve typicallyprovides a very poor fit, curving downward evenwhen the trend of the data is toward an increas-ingly upward curvature. This typically bad fit todata of the probit curve raises serious questionsregarding the validity of statistical procedures as-sociated with the Mantel-Bryan method (see Sals-burg, 1977; and Crump, 1977b).

On the other hand, the Mantel-Bryan proce-dure utilizing the choice b = 1 was put forth asconservative procedure and it gives that appear-ance in figure D-1 since the probit curve appearsto lie far above the trend of the data at the lowestdoses. However as mentioned earlier and alsopointed out by Mantel (Mantel and Bryan, 1961, p.458) a procedure may, while appearing conserva-tive at experimental dose levels, at the same timeseriously overestimate the “safe” dose (i. e., beseriously anticonservative) if the assumed dose-

Figure D-2.—Typical Fit of Mantel” Bryan Curveto Experimental Data (From Crump, 1977b)

—

o

o 0

Mantel Bryanmaximum likelihoodcurve

\

“safe” dose curve

2Dose

4 6

risk levels to which extrapolation- is ‘being made.Thus, before the degree of conservation can beevaluated for any procedure, the properties of thedose-response curve at very low doses must beevaluated,

As described earlier, there are strong argu-ments that indicate the dose-response curveshould be “linear at low dose” particularly fordirectly acting carcinogens in the presence ofbackground carcinogenesis. This has led Peto(1974) to recommend extrapolation proceduresusing only dose-response functions from a classcontaining only dose-response functions that arelinear at low dose, At the very least, however, itwould seem prudent not to go to the opposite ex-treme and use a dose-response function that rulesout linearity at low dose by assumption, However,the Mantel-Bryan procedure, through its use ofthe probit curve (equation 1) rules out linearity atlow dose in favor of a “flatness property” (seeHartley and Sielken, 1977; Mantel, 1977; andCrump, 1977b) at low dose which is anticonserva-tive to the extreme. This property implies thatmathematical derivatives of all orders of theprobit curve approach zero (through positive val-ues) as the dose approaches zero. This unusualproperty is more often discussed within the con-text of mathematical oddities rather than in con-nection with a scientific investigation. It impliesthat if the true dose-response curve comes froman extremely broad class of functions known asanalytic functions and which pervade scientificapplications of mathematics, then the probitcurve will eventually underestimate the true riskat low doses. Furthermore, at low enough doses,the degree to which the risk will be underesti-mated will be arbitrarily large (i.e., the ratio ofthe true risk to the probit estimate will grow ar-bitrarily large).

It was emphasized earlier that it is importantwhen extrapolating to low doses for the assumeddose-response function to incorporate known orat least plausible facts about the mechanisms ofcarcinogenesis. In neither the original paper(Mantel and Bryan, 1961) nor in the paper outlin-ing the improved version is biological justificationgiven for the selection of a curve having the abovedescribed “flatness property.” It should be men-tioned at this point that the incorporation of back-ground carcinogenesis into the Mantel-Bryanprobit model (equation 1) using the parameter Cimplies that the mechanism through which thetest carcinogen produces cancer is independentof the mechanisms through which all of the back-


ground cancers are produced (Crump, et al.,1976). In keeping with the discussion in the lastsection it would seem more proper to incorporatebackground into the probit model by positing aneffective background dose do which adds to thedose d of the test carcinogen. If background is in-corporated in this way the probit curve no longerhas the “flatness property” and becomes linearat low dose (Guess, et al., 1977). In fact. withbackground incorporated in this way, the probitcurve assumes a shape similar to the one-hit mod-el, sometimes referred to as the most conservativeof all procedures (e.g., Mantel, 1977).

Even though the “flatness property” impliesthe probit curve should at suitably low doses beanticonservative to the extreme, the “flatness”property holds only for doses approaching zeroand the feature of arbitrarily fixing the probitslope at 1 mitigates the anticonservativeness im-plied by the “flatness” property at any given lowdose (although the property itself will hold for allchoices of the parameters a, b, C), The cancerrisks that are typically extrapolated to are in therisk ranges 1/10,000 to 1/100,000,000. We will ex-amine the outcome of Mantel-Bryan extrapola-tions to these risk levels in a later section whenwe compare them with extrapolations based onthe multistage model.

Linear Extrapolation

The technique for linear extrapolation wasrecommended by Heel, Gaylor, Kirschstein, Safi-otti, and Schneiderman (1975) for use on an in-terim basis until better procedures could be de-veloped. The procedure is straightforward and isbased on an assumed linear relationship betweendose and response at low dose. The procedureutilizes only the data for the group of controlanimals and a single other dose group, usuallyeither the highest dose that elicits no response orelse the lowest dose that elicits some response. Inthe case there are no cancers in the control ani-mals the “safe” dose, based on a maximum risk of1/1,000,000 and 99-percent statistical assurance,is calculated as follows: An upper 99-percent con-fidence bound is calculated for the cancer risk inthe dose group of animals. From this risk and doseone extrapolates back toward zero dose and zerorisk using a straight line relationship. The dosecorresponding to a risk of 1/1,000,000 on thisstraight line is taken to be the “safe” dose. Ifthere are cancers in the control animals this pro-cedure is modified to allow for the statisticaltreatment of the response in the control group

while retaining the straight line relationship.When data at other experimental doses are avail-able this method of linear extrapolation has theobvious shortcoming of not fully utilizing theavailable data.

A linear dose-response curve is linear at lowdose but the converse is not necessarily true. Acurve can be linear at low dose and still have ahigh degree of nonlinearity at higher doses.

Linear extrapolation is viewed by some as avery conservative procedure. For example, com-ments were made during the decision on whichextrapolation procedure to incorporate into theSOM document to the effect that linear extrapola-tion is the most conservative of all procedures.Crump, et al. (1976) examined the extent of theconservatism of a linear dose-response functionwhen compared with a multistage dose-responsemodel (Armitage and Doll, 1961). The multistagemodel assumes that a cell must go through a num-ber of different stages before cancer is initiatedin that cell and the model can encompass a highdegree of nonlinearity. It was determined that themaximum possible degree of conservatism of alinear model relative to a multistage model de-pended rather heavily on the incidence at the ex-perimental dose relative to the background inci-dence. (This is consistent with the general rela-tionship between background carcinogenesis andlinearity at low dose as discussed earlier. ) For ex-ample, when the incidence at the experimentaldose is four times the incidence at zero dose theextra incidence at low dose derived from the lin-ear dose response differs from the incidence de-rived from the multistage model by, at most, lessthan a factor of 2.5 regardless of the number ofstages in the multistage process. Thus, whenbackground carcinogenesis is present, the lineardose-response curve is not overly conservativerelative to the multistage dose-response curve. Infact the linear dose-response curve is anticon-servative when compared to the one-stage or one-hit models.

Linear extrapolation has long been proposedfor use in radiation carcinogenesis (see Brown,1976, for a review of the relevant reports). TheBIER (1972) report on radiation risks from the Na-tional Academy of Science recommended linearextrapolation as a “best estimate” approach asopposed to a conservative approach. Brown re-viewed arguments both for and against linearityand concluded that “linear extrapolation ofhuman data from high dose of low LET radiationcannot be said to overestimate the risk at low


doses. In fact, there is some doubt as to whetherthe risk is not underestimated. ”

Certainly much remains to be learned aboutboth radiation and chemical carcinogenesis. How-ever, if both radiation and chemicals cause can-cer through similar mechanisms then it should beexpected that there would also be similarities be-tween the respective carcinogenesis dose-re-sponse functions. Direct damage to DNA by thecarcinogenic agent has been implicated as onecancer-initiating mechanism for both radiationand chemicals (Brown, 1976; and McCann andAmes, 1976). Thus, the findings related to thepotential linearity of the dose-response functionfor radiation has implications for chemical car-cinogens as well, particularly for “directly act-ing” carcinogens.

Extrapolation Methods Basedon the Multistage Model

Two methods of low-dose extrapolation whichare alternatives to the Mantel-Bryan or linearprocedures have recently been proposed inde-pendently by Guess, Crump, and Deal (Guess andCrump, 1976, 1978; and Crump, Guess, and Deal,1977) and Hartley and Sielkin (1977). Both ofthese methods utilize a multistage dose responsefunction of the form:

where q~, ql, . . ., q~ are all non-negative param-eters to be estimated from the data. This dose-re-sponse function is general enough to yield a con-siderably wide range of responses at low dose. Onthe one hand, if q, >0 and q,= O for i~ 2 the dose-response function (equation 2) becomes the one-stage model which yields risks at low doses com-parable to what would be obtained with linear ex-trapolation, On the other hand, the model can pro-duce risks even as low as the probit curve (equa-tion 1) down to any fixed positive low dose. Thusthis model is capable of fitting both highly linearand highly nonlinear dose-response relations.Since the model has the property of “linear at lowdose” if q,> O and does not have this property ifql = O, use of this model does away with having tomake the arbitrary but crucial decision of havingto either assume linear at low dose as in linear ex-trapolation. Thus “safe” doses computed usingthis model should provide a more realistic meas-ure of the true uncertainty of low-dose extrapola-tion than would “safe” doses based on either anassumed linear curve shape or an assumed highlynonlinear curve shape.

The dose-response relation (equation 2) con-tains all of the Armitage and Doll (1961) multi-stage dose-response models as special cases butalso contains curves that are much flatter at lowdose than any of the multistage curves.

The two extrapolation procedures based onequation 2 have some features that are different.When computing “most likely” estimates the pro-cedure of Guess, Crump, and Deal uses an infinitedimensional maximization procedure so that it isnot necessary to specify a value of k, the degree ofthe polynomial in equation 2. However the twomethods differ chiefly in the way the statisticalconfidence intervals are computed. There havenot yet been sufficient comparative calculationsmade to determine how safe doses may differusing the two approaches, Mantel (1977) hasmade a critical review of the statistical procedureused by Hartley and Sielkin for calculating the“safe’ dose.

Both the Hartley and Sielkin and the Guess,Crump, and Deal (as extended by Daffer et al.,1979) procedures can utilize times at whichcancer is detected in the experimental animalsrather than just the dichotomous information ofwhether or not an animal contracted cancer be-fore it died of some other cause or before the ter-mination of the experiment, The utilization of timedata in low-dose extrapolation is important for atleast two reasons: 1) the age at which cancers oc-cur should be important in assessing the magni-tude of the harmful effect of a carcinogen on man(e.g., cancers that occur early in life should beviewed as more serious than those that occur inextreme old age); and 2) in many animal carcino-genicity experiments the response data at thehighest doses lies below the trend of the lowerdose data, This sometimes appears to be due tothe fact that at the highest doses some of the ani-mals are being poisoned by the chemical beforethey have a chance to develop cancer, When thisoccurs the high-dose data is often just deletedfrom the analysis. However if the times at whichthe animals die are properly used the high-dosedata might not appear anomalous. More researchneeds to be done on the proper utilization of ani-mal time of death data to assess the harmful ef-fects of chemicals to man.

Comparisons Between Mantel-Bryanand Multistage Extrapolations

To compare low-dose extrapolations using theMantel-Bryan probit model (equation 1) with


those using the multistage model (equation 2) wepresent figure D-3 based on the same data asfigure D-2. In this figure, the Mantel-Bryan “safe”dose is plotted on a log-log scale as well as boththe multistage “most likely” curve and the “safe”dose based on (equation 2) and computed as out-lined in Crump, et al. (1977). A 99-percent statisti-cal assurance was used for both “safe” dosecurves. We note that the Mantel-Bryan “safe”dose lies above the multistage safe dose curve atvalues of added risk below 5 x 10-4, The Mantel-Bryan “safe” dose curve lies above the multistage“safe” dose curve by a factor greater than 20 foran added risk of 10~ and by a factor greater than300 for an added risk of 10~. Because of the“flatness” property of the Mantel-Bryan probitfunction (equation 1) described ear l ier , theMantel-Bryan “safe”’ dose curve will lie above themultistage safe dose curve by arbitrarily largefactors at extreme low doses. Guess, et al. (1977)have compared the Mantel-Bryan “safe” dosecurves to the multistage “safe” dose curves andfound this to be a typical situation. Thus it is clearthat if the true dose-response curve could besimilar to the multistage dose-response function(equation 1) then the Mantel-Bryan procedurecould not be justifiably called conservative (see

Figure D-3.—Comparisons of “Safe” DosesComputed From the Mantel” Bryan Procedure

and From a Procedure Based On the MultistageModel (From Crump, 1977b)

r

Added R i sk

also Crump, 1977 for further discussion of thispoint.) On the other hand, we have seen that thereare quite plausible arguments for the true dose-response curve to have the same shape at lowdose (linear) as the estimated multistage curve.

We note that both the multistage “safe” dosecurve and “most likely” curve have a slope 1 infigure D-3 as plotted on the log-log scales which isequivalent to the curves being linear at low dose.The fact that the “most likely” curve has slope 1is due to the fact that with this particular data setthe linear coefficient q, in equation 2 will alwaysbe linear at low dose regardless of whether or notthe linear coefficient q, is estimated to be positive.This property should be shared by any valid sta-tistical procedure based on a dose-response func-tion that does not rule out linearity at low dose byassumption as Mantel-Bryan does. Just as it is notpossible to prove statistically the existence of athreshold, it is likewise not possible to rule out thepossibility that the true dose-response curve islinear at low dose on the basis of statisticalanalysis. (See Guess, et al. (1977) for a thoroughdiscussion of this important point.) The Mantel-Bryan obtains “safe’” dose estimates which areconsiderably higher than those obtained using themultistage model because it assumes away linear-ity at low dose, an assumption that we have seenis probably unwarranted for that majority of car-cinogens which are classified as ‘‘directly acting”carcinogens.

Since extrapolation based on a model such asthe multistage model (equation 2) must always belinear at low dose, the question arises as to howdifferent the result will be from simple linear ex-trapolation. For some data the difference will beminimal. For example, for the data upon whichfigure D-2 is based, “safe doses” computed usingthe multistage model are almost identical with*’safe’” doses based on linear extrapolation. Forsome data sets, however, the difference could beconsiderable. For example, with the Gass, et al.(1964) DES using the C3H female mice, the “safe”dose based upon linear extrapolation is lowerthan the “safe” dose based on the multistagemodel by a factor of about five and there areother data sets where this difference is greaterthan an order of magnitude.

The Gamma Multihit CarcinogenesisDose-Response Model

Rai and Van Ryzin (1978) have proposed basingrisk estimation on the gamma multihit model:


where P(d) is the lifetime probability of cancer ina tissue when subjected to a constant dose rate, d,of the carcinogen. This model is obtained by as-suming that cancer due to the carcinogen occursrandomly according to a Poisson distribution. Themanner in which background carcinogenesis is in-corporated into the model is equivalent to assum-ing that the event “cancer occurs due to the ac-tion of the carcinogen” is independent of theevent “cancer occurs spontaneously, ” This as-sumption would not apply, for example, to proc-esses in which the effect of the carcinogen is tospeed up the rate at which the “spontaneous”events occur which lead to the background can-cers. At low dose rates, the response is approxi-mately given by:

Consequently, this dose response model is linearat low dose rates when and only when k = 1. Raiand Van Ryzin calculate confidence intervals foradded risk at a given dose and for the dose pro-ducing a fixed added risk using asymptotic max-imum likelihood theory. Although in the theoreti-cal development k must be an integer, in the appli-cations k is allowed to assume any positive value.

Lower statistical confidence limits on the doseproducing a given low amount of extra risk (“vir-tually safe dose” or “VSD”) computed using thisprocedure can be compared with those computedfrom the multistage model (equation 2) by consid-ering two general classes of data.

If the data exhibit a general downward curva-ture as illustrated in figure D-4a, lower con-fidence limits on a VSD computed from the gammamultihit model (equation 3) should generally beless than or equal to corresponding lower con-fidence limits computed from the multistagemodel (equation 2). In certain instances the gam-ma multihit lower limits could be much less thanthe corresponding multistage lower limits. Thiscould occur when the data is consistent with k <1in the gamma multihit model (presumably cor-responding to a fraction of a hit).

On the other hand, if the data exhibit a generalupward curvature as illustrated in figure D-4b thereverse situation will hold; gamma multihit lowerconfidence limits on VSDs will generally begreater than or equal to corresponding multistagelimits, Gamma multihit limits will not in generalshare the low-dose linearity of multistage limits

by orders of magnitude at low doses. The reasonthese large differences can occur is as follows.The multistage family of models contains mem-bers which are simultaneously linear at low doseand exhibit upward curvature at moderate doses.For example, the particular multistage model:

is linear at low dose since q, >0 and still can ex-hibit upward curvature at moderate doses sinceqz >0. This means there are dose-response curvesin the multistage class which are both linear atlow dose and can adequately describe data of thetype exemplified by figure D-4b. On the otherhand, this will generally not be true of gammamultihit models. All dose-response curves in the

Figure D-4a.— Example of Data ExhibitingDownward Curvature

%

x

Dose

Figure D-4b.—Example of Data ExhibitingUpward Curvature


gamma multihit class that are linear at low dosemust exhibit downward curvature and conse-quently would generally not be consistent withthe data in figure D-4b. When this is true gammamultihit lower confidence limits on VSDs will besublinear (e.g., quadratic) at low dose and conse-quently much larger than the multistage confi-dence limits.

Confidence limits based on the gamma multihitmodel will be approximately correct wheneverthis model is the correct model, A similar state-ment could be made for the probit model, multi-stage model, or any other model to which reason-able statistical methods are applied. However,there may be considerable uncertainty as to whatthe true model may be in a particular situation.The multistage model not only reflects some rea-sonable assumptions regarding the carcinogenicprocess which dovetail nicely with epidemiologi-cal data for many cancers (Pete, 1977 b), but italso reflects some of the uncertainty with regardto the true model by virtue of encompassing arelatively large class of dose-response functions,For example, as noted earlier, the multistageclass contains dose-response functions which arelinear at low dose and also exhibit upward curva-ture at moderate doses. On the other hand, thegamma multihit class is more restrictive at thispoint in that it does not permit such behavior. Isthis extra restrictiveness of the gamma multihitmodel justified? To help answer this question,consider the following modification to this model.Suppose that the hits (phenomenological eventswhich are required to occur in a tissue in orderthat a cancer appear) can possibly occur sponta-neously in the absence of the carcinogen. Part ofthe effect of the carcinogen could then be to speedup the rates at which the spontaneous hits are oc-curring. For example, if one of the “hits” is in in-correct base substitution in DNA during mitosis,the carcinogen could speed up the rate at whichthese “hits” are occurring in individual cells byproviding more aberrant bases for use in such anincorrect substitution. With this modification tothe model, the gamma multihit lower confidencelimits on VSDs will no longer be sublinear at lowdose and should be at least as small as corre-sponding limits calculated from the multistagemodel (equation 2). Thus, in order to obtainsublinear lower confidence limits on VSDs withthe gamma multihit model a modification such asthe one described above must be ruled out, not onthe basis of data, but by assumption.

By way of summary, confidence limits based onthe multistage model will always be linear at lowdose. Confidence limits based upon the gamma

multihit model may be either “super linear” (cor-responding to k < 1) or “sublinear” (correspond-ing to k > 1). Superlinearity is achieved by makingthe model too broad in that a fraction of a hit isallowed which has no biological basis. On theother hand, sublinearity is achieved by makingthe model possibly too restrictive in that modelswhich are reasonable from a biological viewpointare ruled out by assumption.

In addition to questions as to the appropriate-ness of the gamma model, there is also a very sig-nificant difficulty with the statistical procedureapplied to the model in the Food Safety Council(1978) report. This difficulty is illustrated in tableD-1. Here are listed lower confidence limits forVSDs for the gamma multihit model which weretaken from table D-1 in Food Safety Council(1978). For each of the 14 data sets a goodness-of-fit test was conducted for compatibility of thedata to the subclass of gamma models for whichk =1. As indicated by table D-1, the subclass ofmodels with k = 1 provided an adequate fit in 10out of the 14 data sets. For these 10 data sets arelisted the VSD predicted by the best fitting gammamodel with k = 1. The significant point is that in 8of these 10 cases, the lower confidence limit onthe VSD is greater than the VSD from the modelwith k = 1, greater in some cases by enormous fac-tors, e.g., a factor of 3,600 for aflatoxin B,, a fac-tor of 1,000 for diethyl-nitrosamine, and a factorof 18,000 for sodium saccharin. Put another way,in the case of the sodium saccharin data, there isa member in the gamma multihit class which fitsthe data quite adequately and for which the VSDis less by a factor of 18,000 than the lower con-fidence limit for the VSD reported in Food SafetyCouncil (1978). This obviously implies there is aserious problem with the computation of thelower confidence limits for the VSD. These lowerconfidence limits are much too large, even basedon assuming the gamma multihit model is the cor-rect model, for these 8 data sets. This problemalso exists for the 4 data sets which could not befit adequately by a model with k =1. This is well-illustrated by the graph in Food Safety Council(1978) for the NTA data. In this graph, the lower97.5-percent confidence limit for the VSD corre-sponding to a risk of 1O-G falls almost directly ontop of a data point for which the measured risk is1/91. Put another way, even though the measuredrisk at this dose is 1/91, the statistical procedureused indicates, with 97.5-percent assurance, thatthe risk is no greater than 1/1,000,000.

These difficulties can be overcome by using adifferent statistical procedure. However, whenthis is done, the lower confidence limits will be


Table D-1.—Comparison of Lower 97.5-Percent Confidence Limits on Virtually Safe Doses From GammaMultihit Model Given in to Virtually Safe Doses From Best Fitting Gamma Multihit Model With k = 1

—— —— —. — -—

No. ofexperimental

No. Substance dose levels——— —.1 ‘“ ‘- ‘ -- ‘“234

5678

9

1011121314

.0013

.035

.000029

.0120

.011

.7625.3

p <.005(9 d.f .)

.000032

.020

.00020

significant lack of fit

.00016

.000042significant lack of fit

many times smaller than those reported in FoodSafety Council (1978). For data sets such as the 10referred to in table D-1 which are compatible withk= 1, lower confidence limits for VSDs mustessentially be calculated from a model with k <1and consequently be smaller than those calcu-lated from the one-hit model. This will lead, inmany cases, to super-small confidence limits, asillustrated by the lower confidence limit for vinylchloride, For those data sets, such as botulinumtoxin-type A, which are incompatible with k = 1because of their strong upward curvature, alower confidence limit for the VSD would still belarger than one computed from the one-hit model.However, the theoretical objections raised earlierto the mathematical form of the gamma modelwould still apply.

It may not be generally realized that it is un-common to find data sets which are incompatible

with k = 1 by reason of strong upward curvature.The four data sets in table D-1 which fall into thiscategory all involved a large number of animalsand a number of different experimental doselevels. These four data sets contained from 330animals distributed over 11 dose levels to 720animals distributed over 6 dose levels. Most car-cinogenicity bioassays are smaller than these, Forexample, the standard bioassay for a given sexand strain in the National Cancer Institute pro-gram seems to be 200 animals distributed over 4dose levels. For experiments of this size it is moredifficult to rule out k= 1. Consequently it wouldseem that, for most of the data sets that are likelyto result from a large screening program, appro-priate statistical procedures applied to the gam-ma multihit model would yield lower confidencelimits for VSDs which are even smaller than thosecalculated from the one-hit model.

Append/x D—Review and Evaluation of Methods of Determining Risks From Chronic Low-Level Carcinogenic Insult “ 165

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

APPENDIX D REFERENCESArmitage, P. and R, Doll. Stochastic models forcarcinogenesis. I n P r o c e e d i n g s of the F o u r t hBerkeley Symposium on Mathemuticul Statisticsand Probability, vol. 4 (Berkeley and Los Angeles,Calif.: University of California Press, 1961), pp.19-38.BIER. The effects on populations of exposures tolow levels of ionizing radiation, National Academyof Sciences. National Research Council (Washing-ton, D. C., 1972).Brown, J. M. Linearity vs. nonlinearity of dose re-s p o n s e f o r r a d i a t i o n carcinogenesis. HeulthPhysics 31: 231-245, 1976.Chand, N., and D. G. Heel. A comparison of modelsfor determining safe levels of environmentalagents. In Reliability and Biometry, F. Prochan andR. J. Serfling (eds. ) (Philadelphia, Pa.: SIAM, 1974),pp. 382-401.Cornfield, Jerome. Carcinogenic risk assessment.Science 198:693-699, 1977.Crump, K. S., D. G. Heel, C. H. Langley, and R.Pete. Fundamental carcinogenic processes andtheir implication for low dose risk assessment.Cancer I-lesearch 36:2973-2979, 1976.Crump, K, S., H. A. Guess, and K. L. Deal. Con-fidence intervals and tests of hypotheses concern-ing dose response relations inferred from animalcarcinogenicity data. Biometrics 33:437-451. 1977.Crump, K. S, Response to open query: Theoreticalproblems in the modified Mantel-Bryan procedure.Biometrics 33:752-755, 1977.Da ffer, P. Z., K. S. Crump, and M. D. Masterman.Asymptotic theory for analyzing dose responsesurvival data with application to the low-dose ex-trapolation problem (submitted), 1979.Food Safety Council. Proposed system for foodsafety assessment. Food and Cosmetic Toxicology(December), 1978.Gass. G. H.. D. Coats, and N. Graham. Car-cinogenic dose response curve to oral ciiethylstil-bestrol. j. Nat]. Cuncer Inst. 33:971-977, 1964.Guess, H. A. and K. S. Crump. Low-dose rate ex-trapolation of data from animal carcinogenicityexperiments—analysis of a new statistical tech-nique. Ma thematicul Biosciences 30:15-36, 1976.Guess, H. A. and K. S. Crump. Maximum likelihoodestimation of dose-response functions subject toabsolutely monotonic cons t ra ints. Annals of Sta tis -tics 6:101-1 11, 1978.Guess, H. A., K. S. Crump, and R. Pete. Uncertaintyestimates for low-dose rate extrapolations of ani-mal carcinogenicity data. Cancer Research37:3475-3483, 1977.

15,

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28,

Hartley, H. O., and R. L. Sielken. Estimation of“safe doses” in carcinogenic experiments. Bio-metrics 33:1-30, 1977.Heel, D. G., D. W. Gaylor, R. L. Kirschstein, U. Saf-fiotti, and M. A. Schneiderman. Estimation of risksof irreversible, delayed toxicity. J. o.f Toxicologyand Environmental Health 1:1 33-151, 1975.Mantel, N. and W. R. Bryan. Safety testing of car-cinogenic agents. J. Nat]. Cancer Inst. 27:455-470,1961.Mantel, N., and M. Schneiderman. Estimating“safe” levels, a hazardous undertaking. CancerResearch 35:1379-1386, 1975.Mantel, N., N, R. Bahidar, C. C. Brown, J. L. Cimi-nera, and J. W. Tukey. An improved Mantel-Bryanprocedure for the “safety” testing of carcinogens.Cancer Research 35:865-872, 1975.Mantel, N. Aspects of the Hartley-Sielken ap-proach for setting “safe doses’” of carcinogens. Inorigins of Human Cancer Book C, Human Risk As-sessment, H. H, Hiatt, J. D. Watson, and J. A.Winsten (eds. ) (Cold Springs Harbor, N. Y.: ColdSprings Harbor Laboratory, 1977), pp. 1397-1401.McCann, J., E. Choi, E. Yamasaki, and B. N. Ames.Detection of carcinogens as mutagens in the Sal-monella/microsome test. Assay of 300 chemicals.Proc. Nat. Acad. Sci. USA 72:5135-5139, 1975.McCann, J . , and B. N. Ames. Detec t ion ofcarcinogens as mutagens in the Sa]monella/micro-some test. Assay of 300 chemicals: Discussion.Proc. Nat. A cad. Sci. USA 73:950-954,1976.Pete, R. Paper delivered at the NIEHS WrightsvilleBeach conference on Low-Dose Risk Estimation,1974.Pete, R. The carcinogenic effects of chronicexposure to very low levels of toxic substances.Presented at an NIEHS conference on cancer riskestimation, Pinehurst, N. C., Mar. 10-12, 1976. En-vironment tal Health Perspectives, 1977.Pe te , R . Ep idemio logy , mu l t i s t age mode l s , andshort-term mutagenicity y tests. In Origins o.f HumanCancer, Book C, Human Risk Assessment, H. H.Hiatt, J. D. Watson, and J. A. Winsten (eds. ) (ColdSprings Harbor, N. Y.: Cold Springs Harbor Lab-oratory, 1977 b), pp. 1397-1401.Rai, K. and J. Van Ryzin. Risk assessment of toxicsubstances based on a generalized multihit doseresponse model. Proceedings of SIMS Conference,Alta, Utah, June 26-30, 1978.Salsburg, D. S. Open query: Theoretical problemsin the modified Mantel-Bryan procedure. Bio-metrics 33:4 19-421, 1977.Watson, G. S. Age incidence curves for cancer.Proc. Nat. Acad. Sci. USA 74:1341-1342, 1977.

Appendix

Measuring Benefits and Costsby Donald S. Epp

tolerance level

The benefits to be evaluated are defined as thereduction in hazards to human health. This defini-tion follows directly from the concern of the Food,Drug, and Cosmetic Act and the activities of FDA.Any effects on the health of animals or plants, ex-cept as they become food and thus affect thehealth of humans, is not considered in this report.This is not to say that economic analysis is incapa-ble of considering other values associated withanimal and plant life. Rather, it is to limit discus-sion to those aspects most germane to the decisionunder consideration.

The measurement of human health and the de-termination of a relationship between a particu-lar contaminant in food and subsequent humanhealth impairment must be made in biological,medical, and physical science terms. For exam-ple, the conclusion that daily consumption of foodcontaining x parts per million (ppm) of a given

*Excerpt from OTA WoPi ing Paper entitled “Priority Settingof Toxic Substances for Guiding Monitoring Programs. A com-plete copy of the paper can be obtained from the NationalTechnical Information Service. (See app. J.)

substance over a period of 3 years or longer willresult in a 50-percent loss of function of the armsand hands is a medical and biological science con-clusion. Until a determination of the health conse-quences is made, very little can be done towardcomparing benefits and costs of regulation,

Health hazards from exposure to food contam-inants may be stated in various ways dependingon the health effects of the substance and on thestate of knowledge about those effects. The haz-ard from a substance that increases mortality inthe exposed population could be stated as thenumber of premature deaths per year. If illnessrequiring time-off from regular activity resultsfrom a certain exposure level, the hazard to thepopulation might be stated as the number of per-son-days lost to illness per year. When it is notcertain that exposure will cause a particular con-sequence, but that the likelihood is increased,probabilities may be attached to the outcomes.The hazard from exposure would then be statedthat some percentage of the exposed populationwould die prematurely or would lose a specifiednumber of days from normal activity due to ill-ness, This method of stating hazard is encoun-tered with cancer risks due to smoking tobacco.

166

Appendix E—Measuring Benefits and Costs ● 167

In many instances the substance in questionmay not have been studied for health effects or itmay have been developed so recently that long-term effects of chronic exposure are uncertain. Itmay, therefore, be impossible to calculate proba-bilities, let alone specify the numbers of people af-fected, This does not mean the matter is of no con-cern. Based on the chemical structure of the sub-stance, scientists may believe that the long-runconsequences may be very serious. In such casesa precise probability cannot be attached to a spe-cific event occurring, It cannot even be said withcertainty that a health hazard exists. What canbe said is that a serious health risk may be pres-ent and the consequences of that risk are so gravethat society may wish to avoid whatever risk mayresult. While this type of statement of the healthhazard is more difficult to use in a benefit-costanalysis, it is not impossible.

No matter how the hazard is stated, the firsttask in economic analysis is to convert the hazardstatement into units of measurement that are com-parable with the statements of costs to be devel-oped later. Many (although not all) of the costs as-sociated with those decisions are measured inmonetary units—dollars. Likewise, at least someof the most important benefits to be achieved fromrestricting exposure to health risks are also meas-ured in dollars. The most fruitful way to proceed,then, is to assemble the appropriate monetarymeasures and make conversion into monetaryterms wherever possible for those effects not someasured.

Cash Cost Approach

One of the simplest ways to measure the cost ofillness and at least part of the cost of prematuredeath is to add up the expenditures made fortreatment or burial and other out-of-pocket costs,These costs can be obtained in a straightforwardmanner and summed for the number of individu-als that will be affected at the tolerance levelbeing evaluated,

Even though the procedure does not involvecomplex calculation, this approach does involvedecisions and judgments about the appropriatecosts to us. Medical service costs vary greatly be-tween procedures and between locat ions, I twould be advisable, therefore, to obtain medicalcosts for the specific types of procedures re-quired and for the part of the country where af-fected individuals would likely reside, In somecases where the contaminated food would likelybe consumed nationwide, a weighted average cost

of the appropriate medical service could suffice. . .Where the contaminated food is consumed in par-ticular regions of the country, e.g., catfish, thehealth effects would likely be concentrated inthose regions also and the appropriate treatmentcosts should relate to the same regions.

While the cash cost approach is relatively sim-ple to calculate, it ignores some of the true coststo society from an untimely death or an illness.One of the most obvious omissions is the failure toaccount for an individual’s contribution to eco-nomic output, had sickness or early death not in-tervened. This weakness is corrected with the for-gone-earnings approach.

Forgone-Earnings Approach

Generally, the forgone-earnings approach usesthe discounted value of the future stream of earn-ings as the appropriate estimate for the cost of anuntimely death, and thus, the appropriate esti-mate of the value of a life saved from a prematuredeath. While it is clear that if an individual hadnot died or become ill in a particular year hewould have continued to be productive for a num-ber of years , it is not immediately obviouswhether the correct measure to use is gross earn-ings or net earnings.

Those who support the use of gross earningsargue that what is of interest to society is lost pro-duction, and an individual’s contribution to thatproduction is measured by his or her earningsstream. Thus, the loss to society is the discountedpresent value of the stream of annual earningsweighted by the probability that the individualwill be alive and well enough to earn that year’sincome. This approach is best suited to the anal-ysis of illness rather than death and has beenused to study the loss to society from mental ill-ness, syphilis, and illness of all kinds collectively,

Those who prefer to use net earnings arguethat, in the case of death, society may lose the pro-duction of the decedent, but it also no longer hasto supply goods for the decedent’s consumption.Death releases resources to the rest of society fortheir use. Thus, the appropriate measure of lossto society is the discounted present value to thestream of differences between an individual’s ex-pected income in each year and that individual’sconsumption during the same period. Each annualnet earnings would be weighted by the probabilityof the individual surviving to that year. With thismethod of calculating, the cost to society of adeath could be negative (or a benefit) if the indi-vidual consumes more than he or she produces.

5U-515 o - 79 - 12


Retired and unemployed individuals are likely tohave such negative life valuations.

The possibility of a negative value for a life hasled to several criticisms of the approach. Someindividuals produce products that are not mar-keted, but are none-the-less, real and valuable,The largest single example is the output of house-wives. Other objections have been raised to theconclusion that the unemployed and retired are ofno value. It must be admitted, however, that theselatter objections appear to confuse the economicvalue of an individual with an ethical value of ahuman life, It is clear that the forgone-earningsapproach examines only one portion of an individ-ual’s contribution to society—the contribution tothose things measured in the net national product(NNP). If the goal of society is the maximization ofNNP, then assessing the costs or benefits of ahealth-impacting project according to the for-gone-earnings approach can be rationalized. If so-ciety holds other goals as well, then the use offorgone earnings is merely an expedient approx-imation.

Willingness-to-Pay Approach

The most telling argument against the previ-ously discussed evaluation techniques is that theyare conceptually incorrect for use in benefit-costanalysis. They are directed to find the answers toquestions like “What is my life worth to other peo-ple like my heirs and society in general?” The con-ceptually correct question to ask is “How muchwould I pay to avoid a small probability of mydeath or illness?” Approaches that answer thisquestion are consistent with other measures of so-cietal welfare since they estimate the aggregateconsumer surplus involved in the reduction inmortality or morbidity rates.

What has come to be known as the “willing-ness-t o-pa y’” approach has received extensivetreatment in recent economic literature. A reviewof much of the theoretical literature and the dis-cussion of some technical points related to a per-son’s willingness to pay for small changes in mor-tality rate may be found in Epp, et al. (ch. 4), Thatreview will not be repeated here, but rather, theremainder of this section will be addressed to re-viewing techniques for determining an individ-ual’s willingness to pay. Two approaches havetheoretical validity: compensation for risk-takingand questionnaires of willingness to pay.

Compensation for risk-taking as a technique forestimating the loss in consumer surplus due to in-creased risk of death, illness, or injury, has the

advantage of being observable to the market-place. If an individual agrees to undertake a haz-ardous occupation which increases his probabil-ity of death in exchange for a given sum of money,that sum can be used as an estimate of his willing-ness to pay for safety, Thaler and Rosen useddata that measures the relative riskiness of jobsto estimate the tradeoff between wages and risk.Thirty-seven broadly defined job classificationsshown to be actuarially riskier than the averageoccupation were matched against a cross-section-al earnings survey. The results showed that in-dividuals in the risky occupations received in-creased compensation of about $200 per year(1967 dollars) for jobs where the risk of death was1 in 1,000 greater than the average.

This estimate may be conservative when ap-plied to the general population. The occupationssurveyed in the Thaler and Rosen study are ap-proximately five times riskier than the averageU.S. occupation. People who take these jobs havedifferent reservation prices for risk than individ-uals who pick less risky jobs. The derived esti-mate of compensation required to offset greaterrisk is, therefore, an extremely conservative ap-proximation of aggregate willingness to be com-pensated. A further rationale for believing this es-timate to be conservative is that the riskiness ofthese jobs can in some measure be affected by theindividual employee. He or she is to some degreein control of the personal risk level within the con-text of the risk level of the occupation. If the in-dividual is not able to affect his or her personalrisk exposure, it is quite likely that they wouldneed to be compensated to a greater extent toundertake the unwanted risk.

The other technique for assessing willingnessto pay is the questionnaire method. A representa-tive sample of individuals are asked how muchthey would be willing to pay to achieve a specifiedchange in a particular condition. For the problemof environmental contamination of food, the ques-tion would most likely be directed toward the will-ingness to pay for a reduction in the probability ofsuffering specific health effects, such as speechimpairment and nerve sensations from exposureto methylmercury. Questionnaire design rangesfrom direct questions about dollar amounts for aspecific risk change to a series of questions aboutitems related to the specific risk change of inter-est from which the analyst can infer an answer tothe “what would you pay?” question,

While the willingness-to-pay approach is ask-ing the conceptually correct question, it has seri-


ous difficulties in application, because the itemsof interest are going to be consumed in equalamounts by everyone and no one can be excludedfrom consuming the item if it is provided for any-one. For example, a reduction in the risk of lungdisorders due to reduced sulfur dioxide levels inthe air will be available to everyone in the arearegardless of their preferences or payment, If it isprovided for one person, it is impossible to ex-clude others. The communal consumption aspectof this large class of similar goods and servicesprovides an incentive for people to misrepresenttheir true preferences when asked, For instance,if a person believes he will be required to paysome amount in proportion to his answer yet thereis a high probability the good will be providedwhether or not he contributes and if it is provided,he will be able to enjoy it, then it is rational forhim to respond that he is unwilling to pay any-thing for it, even if he knows he really would paysome amount, if necessary. On the other hand, ifan individual knows that someone else will payfor the good if it is found to be valued highly willfind it rational to grossly overstate its value tohim. Thus, the incentive to be a free rider makes itdifficult to accurately measure the value to soci-ety of an improvement in food safety.

In spite of the difficulties associated with em-pirical measurement of willingness to pay, sever-al studies have been done and have reported somesuccess with various methods of presenting thequestion to the respondents. Randall, et al. (1974)used pictures of the powerplant at Fruitland, N.Mex., at various levels of smoke emission to elicitresponses from residents and tourists as to theamount they would pay (in the form of a sales taxincrease) to move from a less preferred view tothe respondent’s preferred view. The techniquewas used in a similar fashion to examine ques-tions about a powerplant location in southernUtah (Brookshire, et al. ) and about reclamation ofstripmined land in Kentucky (Randall, et al.,1978). A modification of the technique was usedby Mann, et al. to determine willingness to pay forsmall changes in human mortality. Their studyfound that people are able to comprehend riskchanges of the order of IO-6. Although the risk pre-miums stated in the Mann, et al. study were sub-stantially different from those’ found by Thalerand Rosen, the study indicated that individualswere able to answer the type of question posed,and that with improvements in the survey instru-ment, the willingness-to-pay questionnaire may bea workable approach.

COSTSThe costs associated with establishing a partic-

ular tolerance level are the loss of social welfareresulting from the reduction in supply becausesome otherwise useful food products can nolonger be sold for consumption. The reductions inconsumer and producer surplus, which are dis-cussed in detail in this section, account for theeconomic value of resources, such as labor andproduction capital, which may no longer be em-ployed in producing the food product being con-sidered, It is possible, however, that some re-sources may not have any alternative employ-ment, In such case the tolerance level may causethese resources to be unemployed. While this lackof alternative employment is not strictly an eco-nomic cost, it is clearly a consequence that a deci-sionmaker may wish to consider. Therefore, othereffects (some might wish to call them costs) of thedecision are examined in the following section en-titled “Distributional Effects. ”

The starting point for assessing the costs ofmoving to a particular tolerance level is the phys-ical product that will be affected. The results ofthe monitoring program and various special test

procedures can tell what proportion of variousfood products from any specified area will notmeet the tolerance level. The task of the economicanalysis is to evaluate the impact of such a reduc-tion in the supply of that food product.

When analyzing the effects of a tolerance levelthat removes some food products from the market,there are two major alternative methods. Thesetwo are what have been called the alternativecost method and the opportunity cost method.Either of these methods may employ models of theaffected industry, but as will be shown in the fol-lowing sections, the opportunity cost method re-quires more sophisticated models and delivers amore comprehensive description of likely effects.

Alternative Cost Approach

With the alternative cost method, the analystexamines the additional cost necessary to achievea given objective using the next best alternativemethod to the one under study. Applying this tothe PCB contamination of lake trout in Lake Supe-rior, for example, would require that the analyst


examine the next best method for producing164,000 lbs of lake trout that did not exceed 2ppm of PCB to replace that amount from Lake Su-perior which does exceed 2 ppm of PCB. This ap-proach requires that there be some alternativemethod available which will achieve the samelevel of output in order that the costs of the twoalternative ways of producing the output can becompared.

If there are several alternative methods avail-able for producing a “replacement” amount ofproduction, the alternative cost approach re-quires not only that the cost of total replacementwith each method be examined, but that combina-tions of methods be considered also. To properlyevaluate the alternative cost figure, the analystmust know the costs of various amounts of prod-uct from each alternative method. To continuewith the PCB contamination of lake trout example,the analyst must consider not only the cost of164,000 additional lbs of lake trout with less than2 ppm of PCB from each of the other Great Lakes(and any other large sources), but the costs ofvarious increments of production, such as 10,000lbs. from each source. The least cost combinationof increments would be found by starting with theleast costly way of obtaining 10,000 additional lbsand adding to that the next least costly way of get-ting 10,000 more lbs until a total of 164,000 lbsare accounted for.

Opportunity COSt Approach

With the opportunity cost method the analystexamines the additional cost necessary to achievea new market equilibrium amount of a food prod-uct which reflects the likely higher price and,therefore, lower consumption of consumers. Sincerestricting the supply of a food product from pre-vious sources means a shift to higher cost alter-native sources, we can use our knowledge of mar-kets and people’s preferences for the food prod-uct to estimate their adjustment to the restriction,This gives a different estimate of costs due to re-striction than does the alternative cost methodwhich assumes no change in the amount of thefood product consumed.

A simplified partial analysis of the differencebetween the alternative cost method and the op-portunity cost method is presented with the use offigure E-1. The decision to establish a tolerancelevel such that some portion of the product cur-rently offered for sale may no longer be sold isshown by the shift in the supply curve. Curve Sorepresents the industry supply curve prior to the

change in tolerance level, while curve Sl repre-sents supply under the higher cost next best alter-native method of production which does not vio-late the new tolerance level. The alternative costmethod examines the increase in cost of obtainingthe initial equilibrium quantity, Q,. This increasein cost is represented by the area ABP,PO. The op-portunity cost method, on the other hand, recog-nizes that a new equilibrium price and quantitywill emerge after the shift in supply. Quantity willshift to QI and price to Pz. Comparing the newequilibrium with the old, the analyst using the op-portunity cost method observes that the consumersurplus has been reduced by an amount equal toarea ACPZPO in figure E-1. This reduction in well-being is clearly less than the one calculated withthe alternative cost method. Because the opportu-nity cost method recognizes changes in produc-tion and consumption, it is the preferred method.The alternative cost method tends to overstate thecost of more restrictive tolerance levels.

Figure E-1 .—Analysis of a Shift in Supply

I Io! i 1

Q1 Qo

The above diagrammatic analysis presents theessential features of the budgeting approach tothe opportunity cost method. This is discussed inmore detail below. Another approach, modeling,carr ies the above analysis further to showchanges in the market for factors of production aswell as other product markets, While this methodcannot employ graphic analysis due to the com-plexity of relationships, the mathematical model-ing techniques described below permit a muchmore comprehensive analysis of likely effects of achange in tolerance level.


As was noted above, the opportunity cost ap-proach measures effects of a change in tolerancelevel as the change in consumer surplus in theeconomy. There is abundant literature discussingthe problems of estimating consumer surplus;some authors even question the value of the con-cept in many empirical contexts. For this exer-cise, however, it seems appropriate that consum-er surplus be used. The ultimate consumer of thecontaminated food products should provide thebasis for evaluation. This becomes particularlycrucial where ramifications of the productionshift may include a variety of products, not justthe contaminated product. Consumer surplus be-comes the common denominator allowing compar-ison among a variety of production effects. Inpractice, the opportunity cost approach may em-ploy either of two methods: budgeting or model-ing.

BudgetingWith budgeting the analyst acknowledges that

production shifts are likely for the contaminatedproduct. This technique uses data on inputs usedin the production of a product to calculate thecosts of producing a particular amount of thatproduct, assuming that other commodities will beproduced in the same amounts and at the sameprices as with the status quo. The budgeting ap-proach is a more limited and restricted approachthan the modeling one, but for some circum-stances may prove advantageous.

ModelingThe second and more complete method of ap-

plying the opportunity cost approach is throughthe use of production models. With this method,mathematical models of the relevant portion ofthe economy are employed to trace the shifts insupply curves and the changes in the amount andprice of various commodities and factors whichresult from the restriction on the use of a contami-nated product. If these models are specified to in-clude geographic areas and the various alterna-tive production activities which take place ineach of these areas, they are able to projectchanges in the location of production of particu-lar crops and changes in the use of various pro-duction factors in each region of the country. Thismore realistically describes the likely reactions tothe change in tolerance level and permits the cal-culation of a more accurate estimate of thechange in consumer welfare resulting from theaction level decision.

Comparison of Approaches

The alternative cost and opportunity cost ap-proaches are not equally adapted to handling allproblems—each has several advantages and dis-advantages. The alternative cost approach hasthe advantage of using data that is more easily ob-tained and not requiring extensive developmentof mathematical models prior to the analysis.Since the alternative cost approach compares thestatus quo with the next best alternative, subjectto the change in tolerance level, the only addi-tional data needed beyond present productiontechniques is an estimate of how the same amountof product could be produced under the best al-ternative available with the new tolerance level.This data can frequently be obtained from ex-perts in the production of the food under con-sideration.

The major disadvantage of the alternative costapproach is that it is conceptually erroneous. O u rknowledge of economic adjustments to changingproduction conditions recognizes that adjust-ments will be made in the enterprise combina-tions and in factor combinations for producing aparticular food product. If a significant portion ofthe total production is removed from use or a spe-cific location is no longer capable of producing asafe food product, the ensuing production adjust-ments will not be limited to factors of production,but may also include changing the foods producedor the location of production. The amount con-sumed of the food for which a new tolerance levelis being established will probably adjust tochanges in the cost of obtaining the food in a safeform (within tolerance levels). The alternativecost approach ignores all of this knowledge by as-suming that the food in question, using the nextbest alternative, will be produced at the samelevel as before.

The conceptual error leads to a second error—the overstatement of the costs associated with thetolerance level, Because this approach ignoresthe adjustments in the amount of a food productproduced, it leads the analyst to a cost figure thatis greater than the one that would actually result.As production costs increase it is likely that thequantity produced will decrease. Thus, the alter-native cost approach has the analyst multiply ahigher cost per unit by more units than would ac-tually be produced. This leads to an erroneouscalculation of the change in consumer surplusand to the overstatement of the cost of settingtolerance at a particular level.


The major advantage of the opportunity costapproach is conceptual: it allows the analyst t osee the adjustments that the economy is likely tomake in response to changing production costs asa new tolerance level removes some of the prod-uct from the market. Where agricultural productsare involved, the use of an econometric model ofthe agricultural sector facilitates the analysis ofchanging comparative advantage and the result-ing production pattern, This gives a more accu-rate indication of the ultimate cost of making thetolerance level decision,

The use of models can also facilitate distribu-tional analysis if those models have spatial (re-gional) variables introduced into them. Throughthe use of regional supply and demand models,the analyst is able to estimate not only the marketequilibrium supply and demand adjustments, butalso the regional production adjustments in re-sponse to the overall market changes. It is thuspossible to note changes in the regional locationof production and to estimate the impact of thesechanges on various income and social groups thatare distributed differently in the various regionsof the country. Thus, the models that are used forthe opportunity cost approach can facilitate amore detailed and sophisticated analysis of the ef-fects of a change in the tolerance level. The rea-sons for examining distributional effects as wellas a brief description of some ways these may bedone are included in the next section of thispaper.

The major disadvantage of the opportunity costapproach is that it requires a very large amountof data. Production costs, outputs, factor require-ments, factor prices, and a variety of other bits ofinformation are required for each alternat ivefood production activity in each region includingalternative ways of producing the food under con-sideration. Even if the supply equations and activ-ities are limited to those most likely to enter intothe solution, the requirements are formidable formost agricultural and fishery commodities. Thedemand side also requires regional considerationwith specification of the demand for each of thecommodities included in the supply side of themodel. Careful attention must be given to the in-clusion of complements and substitutes so that the

- model will give a reasonable approximation of ac-tual market adjustments.

A second disadvantage of the opportunity costapproach is that the models developed for analyz-ing production and market shifts are usuallyshort-run and static, This means that thesemodels must be revised periodically in order to in-

clude new developments in factor prices, productprices, and production technology. Thus, the mod-els are expensive to maintain. They are also ex-pensive to create in the first place. Most modelsrequire a great deal of prior research on the tech-nical relationships in production and specifica-tion of the factors related to consumption. Whilemuch of this work has been done for agriculturalcommodities, it is recognized that the material isfrequently inexact and often the models are out ofdate, It would be necessary, therefore, to under-take some rather expensive research in order toincorporate the opportunity cost approach for acommodity that did not already have substantialprior work. This is likely to be the case for sea-foods, freshwater fish, and certain agriculturalproducts produced in a few local areas.

Obviously, the detailed knowledge of produc-tion relationships and therefore the expense is re-duced if one uses the budgeting method ratherthan the modeling method in the opportunity costapproach, Budgets usually involve a partial anal-ysis of the adjustments and therefore do not re-quire the development of production relationshipsfor commodities not closely related to the com-modity under consideration, Even so, the budget-ing approach requires more information than thealternative cost approach because of the consid-eration of production changes. It is unlikely thatthe budgeting approach can be used for a region-alization analysis that involves more than a veryfew regions. Thus, the lesser cost produces lessinformation.

Since the two approaches, alternative cost andopportunity cost, differ in the amount of data thatthey require and in the cost of acquiring and proc-essing the data, it is necessary for an analyst tochoose between these two methods. Severalpoints should be considered when making a selec-tion of the most appropriate analytical method.

Demand ElasticityIf the product for which the tolerance level de-

cision is being made has a very inelastic demand, ’it may be appropriate to use the alternative costapproach. With a very inelastic demand, it is like-ly that the assumption of producing the samequantity regardless of cost is a reasonably closeapproximation to reality, Thus, the disadvantagesenumerated above for the alternative cost ap-proach are of less significance for many agricul-

Appendix E—Measuring Benefits and Costs . 173

tural products, such as certain vegetables, sugar,and cereals for human consumption, as well asfor fish.

Information AvailabilitySince the opportunity cost approach requires a

great deal of information about production rela-tionships, not only for the commodity under con-sideration but for alternative commodities, it ishelpful to have these relationships previously de-veloped. Much of this information is so time con-suming to develop that it would be virtually impos-sible to create a research program that could giveresults in time to be useful for any action leveldecision if a great deal of groundwork had notbeen laid prior to that research effort. If no in-formation is available concerning production ofthe product in question or the alternative productthat might be produced or consumed, it is likelythat the al ternat ive cost approach would beadopted. This approach requires less data andthe material that it requires probably can be pro-duced in time to be useful in making a tolerancelevel decision.

Complexity of InterrelationshipsIf a product is produced in many areas with a

large variety of production alternatives, both withregard to factors of production and to alternativeproducts, one needs to use an opportunity cost ap-proach. In such a situation the interrelationshipsare so complex that it is difficult to judge fromobservation of the data the likely combinationthat will result from a tolerance level decision.Under these conditions it is very desirable to usethe opportunity cost approach if at all possible.

Availability y of Mathematical ModelsThe stage of development of the mathematical

programing models for a particular product orsector of the economy is an important considera-tion in choosing a method. If programing models

are fairly well-developed for most of the impor-tant alternative commodities as well as the com-modity under consideration, it may be possible tomodify the existing work relatively inexpensivelyto obtain the information needed for the tolerancelevel decision, For example, a great deal of workhas been done with the feed grain-food grain sec-tors of American agriculture. Less work has beendone on the livestock sectors, although there aresome models that incorporate feed grains andlivestock and some models include the dairy sub-sector. If the decision to be made involves food orfood grains, serious consideration should be givento using the opportunity cost approach and someof the models that have been developed, withwhatever modification seems appropriate. On theother hand, there has been very little modelingthat includes specialty crops, such as specificfruits and vegetables, into a general agriculturalmodel. It would be expensive and of dubious valueto develop a programing model for a tolerancelevel decision involving those crops.

. It is readily apparent that the key assumptionof the alternative cost approach—no change inthe quantity of product produced after restrictingthe amount of a particular contaminant that maybe present—is not realistic in most cases and canlead to substantial error in estimating the socialcosts of a decision. It is the opinion of the authorthat the alternative cost approach is appropriateonly in those cases where an alternative source ofthe product is available at virtually the sameeffective price or where alternative productiontechniques are available at no increase in perunit cost of production and which would involvenot more than negligible shifts in the location ofproduction. These are very restrictive conditions.For most tolerance level decisions involving majoragricultural products these conditions do not holdand the opportunity cost approach is strongly ad-vised.

DISTRIBUTION EFFECTS

The previous two sections have outlined consid- jor interest here is the analysis of how a decisionerations and methods for determining the benefits impacts on various groups within society. Evenand costs of a change in tolerance level for a par- though a decision may produce an improvement inticular contaminant in a specific food product. welfare for the whole of society, there may beFrom an economic perspective and a society-as-a- particular groups for which welfare is reduced.whole viewpoint, these sections cover the econom- Such distribution effects may be due to the bene-ic analysis of welfare changes. There are, how- fits and costs being shared differently by dif-ever, related areas of concern which may be cru- ferent income groups or by different regions ofcial to the social acceptability of a decision and the country, There may also be concern aboutfor which economic analysis can be useful. Of ma- how the consequences of the decision affect the


environment in various locations or its effect ofesthetic considerations. Each of these points is re-viewed briefly in this section.

The analysis of benefits and costs may showgreater addition to benefits from a particular tol-erance level change than the addition to costs;thus, a desirable decision, The consequences ofthat decision may not, however, be desirable forall affected parties, It is, therefore, important insome cases to examine the distribution of benefitsand costs among social and economic groups.

For example, the decision to set a very restric-tive tolerance level on a particular contaminantin fish may mean that fishing in certain bodies ofwater is no longer possible. That may have a veryslight effect on consumers, since they can easilyswitch to fish from other areas. The fishermenwho lose their employment may, however, havemuch lower incomes than the average consumerand may have few or no alternative sources of em-ployment. In this case the benefits in healthhazard avoided by a large number of people maygreatly exceed the costs of idled resources in fish-ing and thus, the decision is appropriate. It is wellfor a decisionmaker to realize, however, that themajor portion of the costs will be borne by peopleof much below average incomes with few alterna-tives to mitigate the losses.

Another decision may involve restriction ofcontaminant levels in a food eaten primarily by aspecific ethnic or racial minority whose membershave low incomes, The producers of this food mayhave several alternative products they can pro-duce and the consumers may also find substitutesreadily available. In this case the costs are slightand a low-income minority receives substantialhealth benefits.

Most of the data needed to describe the distri-bution of effects among income groups is avail-able from the analysis of benefits and costs. Insome cases it is not clear which income or socialgroup is affected, but the analyst can usually de-termine this at the time he collects the data on ef-fects. Thus, it is important that benefit-cost ana-lysts be alert for information about the incidenceof effects as they collect their data. What may berelatively easy to determine while the originalsource is being contacted may be time-consumingand difficult to find out at a later point in theanalysis. The presentation of distributive impactswould probably be in narrative form rather thana quantitative analysis. The important point is toinclude this additional information for the deci-sionmaker so that the fullest possible knowledgeof impacts is available when the decision is made.

The interregional differences in the impact ofbenefits and costs may also be of interest to thedecisionmaker. Frequently the alternative sourceof product to replace that which is contaminatedabove the new tolerance level will be in a differ-ent area. Where a particular stream or valleyproduces contaminated food, the production mayshift only a few miles to a nearby river or valley.Then the impact on the economy of a particularcommunity may be slight. In other cases, a wide-spread area, such as a large estuary or part of aState, may be contaminated and production willshift to a completely different part of the country,Interregional differences may arise from prod-ucts which are produced in one area, but con-sumed in another, For example, most of the winterhead lettuce eaten in the United States is grown inthe Imperial Valley of California, yet only a minus-cule portion of the crop is consumed there. If a tol-erance level for some contaminant made it impos-sible to sell Imperial Valley lettuce, almost all ofthe cost would be borne by that small part of Cali-fornia while the health hazard reduction would beshared throughout the rest of the country.

The data for constructing the regional impactsof a tolerance level decision can be found mosteasily in mathematical programing analyses un-der the opportunity cost approach. Not all pro-graming models have regional specifications, butmost agricultural models do, Careful review ofbudgeting results by experts on the productionand marketing of the foods in question can also re-veal some of the regional impacts or can call at-tention to the likely importance of regional differ-ences. Further regional analysis can be per-formed if desired. The alternative cost approachdoes not provide much data on regional impacts.Here again, however, a careful review of the re-sults by a knowledgeable analyst can suggest thelikelihood of regional shifts that warrant furtheranalysis with finer detail on regional specifica-tion.

The fact that a particular decision will severelyharm a particular region or social group does notnecessarily imply that the decision should beavoided. If the society-wide analysis of benefitsand costs suggests a net gain from the decision,there is good reason to go ahead. The analysis ofthe distribution of effects may serve to point outthe need for special programs to assist thoseharmed adjust to the decision. For example, re-gional impact analysis may point out the need fora retraining program for unemployed workers ina particular region far enough in advance that the


program can be ready to go when the tolerancelevel decision takes effect.

The analysis of impacts on various regions andsocial groups can provide the starting point foranalysis of the sociological and psychological im-pacts of the tolerance level decision. In some, per-haps many, cases the effect on a particular com-munity or group of people may have serious con-sequences for social structure, organizationalstability, and even the mental health of the af-fected people. Such a possibility was pointed outand briefly discussed by O’Mara and Reynolds(P P. 61-66) in their analysis of restricting fishingin the James River of Virginia due to kepone con-tamination. Since the present report is directedtoward economic analysis, the techniques of soci-ological analysis will not be discussed here. Someof the analyses likely to be used are briefly out-lined in Epp, et al. (pp. 134-138).

Another area of potential concern is the impactof a tolerance level decision on the environmentor on the esthetic properties of a region. It is pos-sible that a tolerance level decision may cause

production to shift to an area where environmen-tal damage would increase.

For example, prohibiting the sale of milk fromone area may cause production to concentrate inanother, leading to increased wastes from dairyfarm holding pens and milk-processing plants.What may have been an acceptable load on theenvironment previously may become excessive.Analysts should be alert to the possible environ-mental and esthetic impacts of decisions. In somecases these effects can be measured in monetaryterms and compared with benefits and costs. Inother cases, the effects must remain in biologicalterms, but described in such a way as to makethem useful in a tolerance level decision. The useof environmental food chain models and impactmatrices has been suggested by the author withrespect to pesticide decisions (Epp, et al., chs. 5and 6) and the interested reader is referred tothat discussion for greater detail, Not all toler-ance level decisions will involve a significant en-vironmental effect, but the procedure of analysisshould include a notation to check for such effectsand an analytical protocol for use when needed.

ALTERNATIVE TOLERANCE LEVELS

The sections on benefits, costs, and distributioneffects have described how an analysis might beconducted of a particular decision to change thetolerance level from one point to another. For ex-ample, one decision might move from no restric-tion to prohibiting the sale of a particular foodwith 5 ppm or more of a specific contaminant. Ifbenefits exceed costs and distributional as wellas other effects are not overwhelming, then themove to a 5-ppm tolerance is efficient. That doesnot say that 5 ppm is the most efficient tolerancelevel. There may be some other level, say 2 ppm,

that would produce even greater net benefits. Theproper use of economic methods requires thatconsideration be given to the likely changes inbenefits and costs if the tolerance level werechanged a little either way. If it seems possiblethat net benefits would increase significantly bymoving to a different tolerance level, then a de-tailed analysis as described above should bemade of the change from the originally proposedtolerance level (5 ppm in the example) to the newlevel (2 ppm).


LITERATURE CITED IN APPENDIX E

Brookshire, D. S,, B. C. Ives, and W. D, Schulze, 1976,“The Valuation of Aesthetic Preferences,”’ }ournulof Environment ta] Economics a n d A4anagement3:325-346.

Epp, D. J., et al., 1977, Identification una’ Specificationof Inputs for Benefit-Cost A40deiing of Pesticide Use,EPA-600/5-77-01 2, U. S. Environmental ProtectionAgency. Washington, D,C.

George, P. S., and G. A. King, 1971, “Consumer Demandfor Food Commodities in the United States with Pro-jections for 1980, ” Giannini Foundation MonographNumber 26, California Agricultural ExperimentStation, Berkeley.

I,eW,is, p F., 1977. “Economic Impact Assessment forProposed Reduction of Temporary Tolerances forPolychlorinated 13iphenyls in F’ood. ”” Unpublishedmanuscript, U.S. Food and Drug Administration,Washington, D.C,

hlann. S. H., et al., 1 9 7 6 , “The Determination of theEconomic Effects of Pesticide Use on Human Health:A Gaming Approach to Willingness-to-Pay, ” VVork-ing Paper Series No. 32. ‘rhe Center for the Study ofEnvironmental Policy. The Pennsylvania State Uni-versity, University Park, Pa.

LlcNally, J. M.. 1978a, “Poly(:h(llorirl[~ted” Biphenyls:Environmental Contamination of k’oad, ” Congres-sion:i 1 Research Service, Washington, D. C., OTAWrorking Pi]per.

McNally, J. M., 1978b, “Polybrominated Biphenyls: En-vironmental Contamina t ion of Food, ” CongressionalResearch Service, Washington, D. C., OTA WorkingPaper.

0’Mara, G. K,, and R. R. Reynolds, 1976, Evaluation ofEconomic and Social Conseffuences of RestrictingFishing Due to Kepone Pol]ution in the ~arnes River,Virginiu, U. S. Environmental Protection Agency,Washington, D,C.

Randall, A. B., et al,, 1974, “Benefits of Abating Aes-thetic Environmental Damage from the Four Cor-ners Power Plant. ” New Nlexico State UniversityAgricultural Experiment Station Bulletin, LasCruces, N. Mex.

Randall, A., et al., 1978, “Reclaiming Coal SurfaceMines in Central Appalachia: A Case Study of Bene-fits and Costs,”’ L(~nd Economics 54:472-489.

Reichelderfer, K. H., et al., 1979, “’Pesticide Policy Eval-uation i?lo(iels, ” NRED Working Paper (unnum-bered draft), Natural Resource Economics Division,Economics, Statistics, and Cooperatives Service,U.S. Department of Agriculture, VVashington, D.C.

Thaler, R.. and S, Rosen, 1973, The V(llue of Saving aLife: Evidence from the I.(lhor Market. National Bu-rea u of Economic Rcsca rch, Con ference on Incomeand Weal t h.

‘1’racer Jitco, Inc., 1978, “Risk Assessment of Polybro-mina ted 13iphenyls, ‘“ 0’1’A Ll}orking Paper,

Appendix F

Priority Setting of Toxic Substancesfor Guiding Monitoring Programs *

by Clement Associates, Inc.

INTRODUCTORY REMARKS ON THE ESTABLISHMENTOF PRIORITIES

There are many purposes for which the FederalGovernment must set priorities among toxic chem-icals in the environment. The universe of environ-mental contaminants is very large, and resourcesfor research, testing, monitoring, and regulationare limited. Some statutes explicitly require Gov-ernment agencies to set priorities for testing orregulation, and in many other cases agencies maybe called on to explain their decisions to tacklesome problems before others. Accordingly it isvery desirable to develop rational systems for set-ting priorities.

Establishment of priorities involves preparing alist of problems, ordered according to their per-ceived *’importance” or urgency, before attempt-ing to solve them. It is by definition a preliminarystep and should be designed to maximize the effi-ciency of the problem-solving process, Although itis desirable to use as much relevant informationas possible in setting priorities, it is also possibleto use too much information. If too much time isspent in overelaborate priority-setting exercisesthe gain in efficiency achieved by tackling themost urgent problems first is offset by a delay intaking any action at all. The first task in designingan efficient priority-setting system is thus to makea reasonable compromise between the effort ex-pended on this preliminary step and the gain in ef-ficiency achieved by doing it well.

Priority-setting exercises are usually limitedprimarily by the unavailability of data. (If exten-sive data are available, then it is usually obviouswhich problems should be tackled first. ) For thisreason there is usually little to be gained by devel-oping elaborate multifactorial schemes to rank

*Excerpt from OTA Working Paper entitled “Priority Settingof Toxic Substances for Guiding Monitoring Programs. ” A com-plete copy of the paper can be obtained from the National“rechnical Information Service. (See app. j.)

problems in numerical order. Good systems for es-tablishing priorities will inevitably involve a sub-stantial element of sound scientific judgment inweighing and interpreting incomplete and unsys-tematic data. The second task in designing an effi-cient priority-setting system is thus to make a rea-sonable compromise between the use of objectivecriteria and scientific judgment.

This report is concerned specifically with theestablishment of priorities for monitoring toxicchemicals in food. There are two general methodsby which this might be done: 1) directly, by sam-pling the food supply for chemical contaminantsand ranking them according to potential hazard;and 2) indirectly, by surveying the universe of in-dustrial chemicals and ranking them according totheir potential for entering the food supply in tox-ic amounts. The first method has obvious advan-tages and in fact has been pursued by FDA. Themethod is, however, limited by the available ana-lytical methodology: specifically, most of thechemicals recognized to date as food contami-nants are those relatively easily detected by gaschromatography or atomic absorption spectrom-etry. The principal question addressed in thisreport, therefore, is whether the second methodcan complement the first by identifying potential-ly significant contaminants that have not yet beendetected in foods. Although this report indicatesthat it can do so, it should be recognized that thesecond method also is limited to some extent byavailable analytical methodology. The factors wecan use to identify the potential of a chemical forentering the food supply are selected on the basisof a knowledge of the properties and environmen-tal behavior of other chemicals already known todo so. This in turn is based on our knowledge ofthe extent of contamination and hence on our ana-lytical capabilities. Thus, there is an inherentbias towards identifying as potential food con-

177


taminants those chemicals that are similar to in all systems for setting priorities: chemicals onchemicals already identified in food. This bias which there is no information at all will automati-can only be offset by the use of good scientific cally be given low priority, unless some room isjudgment. This point illustrates a general problem left for largely intuitive judgments,

CRITERIA AND METHODS USED IN PREVIOUS EFFORTSTO ESTABLISH PRIORITIES AMONG TOXIC CHEMICALS

IN THE ENVIRONMENT

Survey of Toxic ChemicalPriority Lists

Thirty-two priority lists of toxic chemicals com-piled during the last 5 years have been identified.Most of these lists (28 out of 32) were preparedfor the Federal Government. For each list, a de-scription sheet indicating the purpose of the listand the criteria used in selecting chemicals wasprepared. In addition, the name of the list, theFederal agency or other institution for which thelist was compiled, the date of completion, the con-tract or document number, type of substances,the number of chemical substances, the means bywhich the chemical was identified (for example,CAS number or chemical name), whether the listis machine readable, and relevant commentswere included.

Methods

Chemical SelectionCompiling a comprehensive list. The first step

in priority-setting efforts is usually compilation ofa comprehensive list defining the universe ofchemicals under consideration. Since chemicalcompounds and mixtures are often referred to byseveral different names, each chemical compoundon the comprehensive list should be uniquely iden-tified. This can be done by specifying for eachchemical on the list the IUPAC chemical name,synonyms, molecular structure, and/or the Chem-ical Abstracts Service (CAS) registry number. Theuse of CAS numbers facilitates computerization.

Most often, the source of chemicals for thiscomprehensive list is previously existing lists.Errors of omission are likely with this method. Achemical not considered previously will not beconsidered by someone using existing lists as asource, The more comprehensive the source lists,the less likely chemicals are to be omitted fromconsideration. An alternative method for select-ing chemicals for consideration is reliance on

recommendations of experts. The likelihood oferrors of omission when this method is used willdepend on the extent of knowledge and experi-ence of the experts. These methods can be com-bined by supplementing source lists with chemi-cals of concern to experts.

Reducing the size of the comprehensive list. I tis almost always necessary to reduce the size ofthe comprehensive list before ranking. Since theresources needed for compiling and evaluatingthe necessary data for setting priorities are usu-ally limited, a “narrowing” or “truncating” stepis performed. Errors are very likely to occur atthis step because, for practical reasons, decisionsat this stage are usually based on incomplete in-formation, In a multistage priority selectionscheme, such as the one by which the TSCA Inter-agency Testing Committee selected chemicals torecommend to the EPA Administrator for testing,additional information is obtained at each stageof reduction. In that effort, an initial listing of3,650 substances was reduced to a master file of1,700 substances, then to a preliminary list of 330substances from which approximately 100 chemi-cals were selected for dossier preparation. Twen-ty-one chemicals or groups of chemicals from thatlist have been designated for recommendation bythe committee. In this type of process, the earlieststages may be the most subjective and the mostlikely to result in errors of omission.

Grouping. Individual chemicals are often com-bined in groups or classes. This is usually done inan attempt to reduce a list of chemicals to a moremanageable set, There are, however, other rea-sons for grouping. Some monitoring or analyticaltechniques do not distinguish between closely re-lated chemicals. Some classes of chemicals, suchas fluorocarbons, may be considered as a groupfor regulatory purposes. Such chemicals as PCBsare not manufactured or used as individual chem-ical compounds and are, therefore, nearly alwaysconsidered as a group. Groups can be based onchemical and physical properties, on chemical

Appendix F—Priority Setting of TOXIC Substances for Guiding Monitoring Programs ● 179

structure, on uses (e.g., flame retardants, fluores-cent brighteners), or on some combination ofthese factors. Sometimes a single compound orseveral compounds are selected as representativeof the group.

An advantage of grouping chemicals with simi-lar structure or functional groups is that it some-times allows chemical, physical, and biologicalproperties of individual compounds to be pre-dicted. Members of a chemical class may behavein a qualitatively similar fashion. In addition,some properties increase or decrease systemati-cally along homologous series. Attempts to derivestructure-activity correlations are much morelikely to be successful for chemical and physicalproperties than for biological properties. Becauseof some unique properties of biological systems,such as enzyme specificity, it is possible for close-ly related chemical compounds, even such stereo-isomers as aldrin, dieldrin, and endrin, to havevery different biological properties.

Some weaknesses associated with groupingare:

●

●

●

Members of the group may vary widely withrespect to the factor being scored, making itdifficult or meaningless to assign a singlescore to the group.No single compound can truly represent anentire class.It is often difficult to assign compounds withseveral different functional groups to asingle chemical class. This difficulty cansometimes be overcome by computerizedsubstructure search systems that allow allchemical compounds with a specific func-tional group to be retrieved regardless of thepresence or absence of other functionalgroups.

At one stage of the National Science Founda-tion (NSF) Workshop Panel, multistage effortgrouping was done in an attempt to reduce the listof chemicals to a more manageable set. Fourteenchemical classes were defined, and a representa-tive chemical was identified for each. Weak-nesses the panel found to be associated with thisprocedure included:

● The classes were not completely valid eitherchemically or functionally.

● The physical and biological properties ofcompounds in a given class varied widely.

● No single compound was truly representa-tive of a class.

The class concept was abandoned in the nextstage of the effort and selected members wereconsidered as individual chemicals.

Determining Factors To BeConsidered in Setting Priorities

The purpose for which the list is to be usedmust be clarified. Once the problem the prioritylist is intended to solve is understood, factors rele-vant to that problem can be determined. Criteriabased on those factors can then be formulated forassigning chemicals a place on the list. Generalcriteria must often be defined in terms of specificparameters for which information is available.Those parameters for which quantitative dataare readily available are the easiest to use forscoring and ranking. On the other hand, data arefrequently lacking on the factors of greatestrelevance to the specific problem. In such casesfactors of less relevance may have to be selected.The choice of factors to be used in setting prior-ities often involves a compromise between the fac-tors that are most relevant and the factors onwhich sufficient information is available.

Assembling Data on Each ChemicalSound decisions on assignment of priorities

require reliable data on each chemical to beranked. The available information may, however,be massive, contradictory, unverifiable, scanty,or absent. It is usually difficult or impossible toobtain existing information that has been classi-fied as proprietary or confidential.

When information is needed on a large numberof chemicals and only a limited amount of time ormoney is available for assembling data, second-ary sources are usually relied on. Certain types ofinformation (e. g., annual production volume, LD~l~,octanol-water partition coefficients) are oftentabulated in secondary sources. Other types of in-formation (e.g., fate and transport in the environ-ment, metabolic pathways) may be obtainableonly from primary reports of research studies.These individual studies require considerabletime and effort to locate, obtain, and review. In amultistage screening process, primary sourcesmay be used only in the last stage, when the num-ber of chemicals has been reduced to a small por-tion of the comprehensive initial list. In a very lim-ited effort, primary sources may not be used atall.

Assigning PrioritiesTwo general methods used for assigning prior-

ities to toxic chemicals are the ranking of individ-ual chemicals [or classes) and the assigning of in-dividual chemicals (or classes) to priority catego-ries. Only 6 of the 32 lists surveyed are actuallyranked. In most of the other lists, chemicals were


either assigned to a single “high priority” cate-gory or distributed among several priority levelcategories.

Some numerical index or score is nearly alwaysused for ranking and often used for assigningchemicals to priority level categories. Whenchemicals are to be ranked solely by parametersfor which comparable quantitative information isavailable (e.g., oral LDfO in the rat or annual pro-duction volume) these data can be used directly.When the criteria for ranking cannot be definedin terms of parameters for which comparablequantitative information is available for all chem-icals under consideration, chemicals are usuallyassigned scores. Scoring is also usually usedwhen quantitative and nonquantitative factorsare jointly considered for ranking.

The scoring process involves assigning numeri-cal values to one or more parameters to producean index or series of indices for each chemical. In-dices can then be weighted and combined to pro-duce a single score for each chemical,

Scoring can be totally subjective, totally objec-tive, or somewhere in between, Subjective scoringinvolves the use of experts to assign priorities.The experts may be provided with information oneach chemical or may simply be asked for theiropinion. The credibility of subjective scoring willdepend on the perceived expertness of panelmembers. Delphi techniques can be used to im-prove the validity or rankings, Scoring systemsbased on the Delphi method rely on the develop-ment of consensus among experts acting inde-pendently or as part of a panel. In the NSF Work-shop Panel effort to select organic compoundshazardous to the environment, panel membersidentified chemicals of concern on the basis oftheir own experience and refined their judgmentsby considering available data for production, im-ports, uses, disposal, and toxicity. Through suc-cessive examinations by panel members, individu-ally and together, in progressively greater detail,a priority list was developed. A weakness of suchsubjective scoring methods is their probable lackof reproducibility, which exists because differentexperts would be likely to reach different conclu-sions, (However, we know of no investigation ofthe reproducibility of such results in which differ-ent groups of experts were asked to assign scoresto the same chemical s.)

Where scores can be assigned to specific val-ues of parameters for which quantitative informa-tion is available, objective scoring may be prefer-able. Often, however, modifying factors may makethe use of some degree of subjective judgment de-sirable in the assignment of scores even where

quantitative information is available. For exam-ple, the octanol-water partition coefficient can beused as an index of bioaccumulation, but a moresophisticated scoring system would consider suchother factors as chemical reactivity, metabolicfate, and vapor pressure of the compound.

The human health hazard posed by a chemicalsubstance is a function of exposure to the sub-stance and the inherent toxicity of the substanceitself. Scores for various exposure and toxicityfactors are usually combined for hazard-rankingpurposes, No consensus has yet been developed,however, as to the optimum algorithm for hazardranking in this way.

Linear weighting schemes are efficient andeasy to use, and they have produced reasonableresults. These schemes rely on the assignment ofscores to a number of factors. Weights are as-signed to each factor, and the weighted factorscores are summed to produce an overall scorefor the chemical. The choice of weight is usually asubjective judgment, involving consideration ofboth the importance of each factor and the relia-bility of the information used in assigning thescore.

Multiplication of individual factor scores pro-vides a wider spread of overall scores, Multiplica-tion is frequently used with actual unscored quan-titative data. This is especially likely where someof the data are in the form of fractions. For exam-ple, fraction of production lost would be multi-plied by total production or fraction of dose ab-sorbed would be multiplied by total administereddose. Multiplication cannot be used unless somecorrection is applied to allow for data gaps. In thedevelopment of one list, an algorithm for toxicitywas applied which compensated for types of tox-icities for which data were unavailable. Thescores actually assigned to a chemical for variouscategories of toxicity were summed [T(total)]. Themaximum toxicity possible for each of these cate-gories for which data were available were alsosummed [T(possible)]. The ratio of these sums[T(total) ÷ T(possible)] was used as the toxicityfactor in the hazard-ranking algorithm. The haz-ard-ranking algorithm used for ranking manufac-tured organic air pollutants was:

Append/x F— Priority Setting of TOXIC Substances for Guiding Monitoring Programs ● 181

In some cases the scores assigned to certainfactors are logarithms of data or are selected sub-jectively in such a way as to approximate loga-rithms. In such cases, addition of the scoresachieves the same result as multiplication of theactual data. Weighting of the factors before addi-tion then corresponds to changing the base of thelogarithm.

Criteria

The two general criteria used for assessmentsof health hazard are exposure and toxicity. Previ-ous efforts to set priorities among toxic chemicalsin the environment have been reviewed, and spe-cific criteria used to assess the extent of exposureto these chemicals or the toxicity of these chemi-cals have been considered.

ExposureThe three aspects of exposure that are of great-

est importance in hazard ranking are the numberof people exposed, the frequency of exposure, andthe quantities of the chemical to which they areexposed. Where direct information is available onthese exposure factors, e.g., from monitoring ofthe ambient environment, of food, or of human tis-sue, it is obviously desirable to use this informa-tion. In most cases, however, we do not have di-rect information regarding the amount of a chemi-cal to which members of a population are ex-posed, so that we must rely on indirect informa-tion, Several factors have been used, alone or incombination, as surrogates for direct measures ofexposure.

Production.—The most commonly used index ofexposure is the annual production volume. Sincethis information is quantitative, it is easy to usefor scoring or ranking. In most cases, productionvolume does provide a rough indication of theamount of a chemical substance potentially avail-able for release into the environment. However,the fraction actually released may vary over sev-eral orders of magnitude, from less than 0.001 forwell-contained industrial intermediates to 1 forchemicals which are used dispersively.

For organic compounds produced by three ormore manufacturers and produced in volumesgreater than 5,000 lbs, production information iscompiled in the annual reports of the Interna-tional Trade Commission and is easily available.production information can sometimes also b efound in other literature sources or obtained frommanufacturers or trade associations, but it is noteasy to acquire and will be missing for a large

number of chemicals. Production data for chem-icals produced by less than three manufacturersare considered proprietary. In the near future,the EPA Toxic Inventory List, being preparedunder section 8 of the Toxic Substance ControlAct, may be a convenient source of production in-formation, but at this time it is not clear how muchof this information will be released to the public.

In addition to annual production volume, dataon production capacity, transport volumes, im-ports, exports, net sales, or levels of consumptionmay sometimes be available. A major problemwith using any of those figures as indices of en-vironmental exposure is that these figures arebased on commercial production and sales. Theamounts of a chemical used as a captive inter-mediate are not included although such a chemi-cal may be an occupational health hazard or aneighborhood pollutant in the vicinity of the plant.Very toxic and highly reactive chemicals areoften manufactured at the same plant in whichthey are used. Manufacturing byproducts maynever enter commerce but because they arewaste products they are very likely to be releasedto the environment in plant emissions and in ef-fluents. Toxic chemicals formed in the environ-ment (byproducts of microbial degradation orphotochemical reaction) also will not be reflectedin production figures.

For naturally occurring chemical substances,such as heavy metals, the most relevant param-eter is not production, but transfer from one com-partment of the environment to another or con-version from one chemical or physical form toanother. Such transfers are measures of the in-crease in the amount of the substance convertedinto a form in which it is potentially available forhuman exposure.

Occurrence in the environment. Actual meas-urements of chemicals occurring in the environ-ment are available for only a small number of sub-stances. Federal agencies tend to monitor on aregular basis only when required to by law. Airpollution measurements are available for carbonmonoxide, total hydrocarbons, nitrogen oxides,sulfur oxides, total suspended particulate, andphotochemical oxidants. Water pollution monitor-ing data are available primarily for pesticides,PCBs, and heavy metals. Monitoring informationin Government data banks is often not retrievablein readily usable form. There may be a timelag ofseveral years before a Federal agency summa-rizes, analyzes, and reports monitoring data. Forqualitative or quantitative data on occurrence inthe environment of chemicals not included in


monitoring programs, it is necessary to search theliterature for individual studies. Since thesestudies will usually report data for a small num-ber of samplings at a limited number of locations(often one), they are difficult to use as a basis forthe comparisons needed for ranking chemicals.The number of reported incidents of environmen-tal occurrence will often be a function of interestin the compound rather than of actual frequencyof occurrence.

An advantage of using this type of information,when available, is that it provides direct evidencethat the chemical is an environmental contami-nant.

Use. Patterns of use can be very useful indica-tors of potential for human exposure. A greaterexposure risk is usually associated with disper-sive uses than contained uses. Contained uses arethose in which the chemical is not systematicallyreleased to the environment as a direct result ofuse (e.g., PCBs in transformers). In dispersiveuses, chemicals are released to the environmentas a direct consequence of use (e.g., pesticides,aerosols, and volatile solvents). The size of thepopulation at risk is usually greater for consumeruses than for commercial or industrial uses. Theclass of the use (e. g., food additive, pesticide, ordrug) will often determine which Federal agencyhas regulatory responsibility for a chemical. Theuse pattern will be a major determinant of thetype, frequency, and amount of human contactwith a chemical substance.

Unfortunately, accurate, complete, and up-to-date use information is very difficult to acquire.Many industrial uses are considered “proprie-tary. ” Uses can change rapidly so that informa-tion compiled in secondary sources is frequentlyout of date. One problem encountered in using useinformation is the difficulty of obtaining quantita-tive data. Since some uses are much more likely tolead to human exposure than others, a quantita-tive breakdown of uses would be particularly val-uable. Even when use data are available, it is im-portant to realize that not all human exposure tochemicals occurs during use. Exposure can alsooccur during production, formulation, and dis-posal or after environmental transformation.

Release. The rate at which a chemical sub-stance is released into the environment can be auseful parameter in ranking toxic chemicals.Chemicals that are found in effluents or emissionsare most likely to become environmental pollut-ants. Information on routes of entry of chemicalsinto the environment can in certain cases beessential to determining population exposure. In

1971 the sole U.S. producer of PCBs voluntarilyrestricted sales to those for closed-system ap-plications. They assumed that this would mini-mize risk to human health and the environment.They overlooked the fact that most environmentalrelease of PCBs occurs, not through use, butrather through disposal.

The quantity of a chemical released into the en-vironment is usually unknown so that scores as-signed for this factor must be based on estimates.The “release rate” is an interesting index re-cently developed by the NSF Workshop Panel toSelect Organic Compounds Hazardous to the Envi-ronment. The release rate (R) was defined as fol-lows:

where P = overall annual U.S. productionI = annual quantity importedL = fraction of the production lost at plant site during

manufacturing, conversion, and formulationD = fraction of the material which goes to noninter-

mediate dispersive uses

Estimates of L and D are usually semisubjec-tive, based on knowledge of the manufacturingprocess and on the likely breakdown of uses.

Environmental transport. A universal or widelydispersed contaminant will usually be givenhigher priority than a contaminant found only atspecific locations or a contaminant whose entryinto the environment results from a unique eventor sporadic events.

For a persistent chemical, the distinction be-tween a local, widely dispersed, or universal con-taminant may be a function of time. Unexpectedlywide transport has been observed in recent yearsfor chemicals such as DDE and PCBs, PBBs en-tered the environment as a result of a single acci-dental occurrence and contaminated a small num-ber of Michigan farms. Since these chemicals arepersistent and have entered the soil, water, andfood supply, they will continue to disperse.

Substances that appear to be relatively innocu-ous or inert at their point of release may be trans-ported to segments of the environment where theyhave significant adverse effects. Useful modelsfor the reliable prediction of the movement ofchemicals in the environment would be extremelyvaluable for setting priorities among toxic chem-icals for purposes of monitoring, control, or regu-lation. Such information would also be valuable indetermining which segments of the environmentshould be monitored or controlled. Unfortunately,although models are available for problems suchas plume dispersion for stack emissions, there areno generally accepted, easy to apply, widely used

Appendix F—Priority Setting of Toxic Substances for Guiding Monitoring Programs ● 183

models for environmental transport. Moreover,realistic models of environmental transport areusually extremely complex, and it is impractica-ble to use them in priority-setting exercises.

Persistence. Chemicals can be transformed anddegraded in the environment through chemical,photochemical, or microbial processes. The rel-ative importance of these processes will be in-fluenced by the compartment of the environmentwith which we are primarily concerned. Reac-tions leading to simpler compounds which arepart of a series leading ultimately to the free ele-ments or elemental oxides are considered to rep-resent “degradation,” Other reactions lead to‘‘secondary pollutants.

Persistent chemicals are often given priorityover easily degraded or transformed chemicals inhazard assessment systems, on the assumptionthat persistent chemicals are more likely tospread through the environment and will continueto accumulate as long as they are released. Whenpersistence is used as a criterion for hazardassessment, it is important to be aware, however,that there are cases in which degradation prod-ucts or secondary pollutants are more toxic thanthe chemical substances initially released,

Biodegradability. Although all living things canplay a role in transformation of chemicals in theenvironment, the lead role is attributed to micro-organisms. The role of micro-organisms in trans-forming chemicals is extremely important in thelithosphere and also important in the hydro-sphere. Micro-organisms are frequently efficientat metabolizing natural chemicals in their envi-ronment. Industrial chemicals are inherently lesslikely to be rapidly degraded by micro-organisms.Parameters often used as indices of biodegrad-ability are the biochemical oxygen demand (BOD)or chemical oxygen demand (COD), Estimates ofbiodegradability can sometimes be based onchemical structure.

Chemical reactivity. The assessment of thechemical reactivity of a compound is often basedon known physical and chemical properties suchas:

●

●

●

●

●

●

b●

physical state,vapor pressure,solubilities,auto-oxidation propensity,redox potential,acid-base characteristics,rate of hydrolysis, andrate of photochemical degradation.

One parameter which can-be used for persist-ence in the environment is half-life, the time re-

quired for removal of half of the molecules of agiven compound from the environment or from aspecific compartment of the environment. Half-life in the atmosphere, for example, would be afunction of such properties as photoreactivity,reactivity towards active forms of oxygen, watervolubility, volatility, and adsorption to fine par-t i cu la te .

Bioaccumulation. Bioaccumulation can be de-fined as the long-term presence of a xenobioticsubstance in living tissue at concentrations sig-nificantly higher than those in the surrounding en-vironment. Chemicals that bioaccumulate areoften given priority over other chemicals in haz-ard-assessment schemes. Concentrations of chem-icals that bioaccumulate are likely to increasealong a food chain, a process referred to as “bio-magnification. ” For such chemicals , res iduelevels tend to be higher in herbivorous animalsthan in plants, higher in carnivorous animals thanin herbivores, and highest in carnivores that eatother carnivores. Since humans may eat orga-nisms in any of these categories and since chemi-cals which bioaccumulate will usually be storedin human tissues, if such chemicals have any toxicproperties they may pose a higher risk to humanhealth than would be predicted on the basis of en-vironmental concentrations alone.

Since bioaccumulation is often associated witha high lipid volubility relative to water volubility,the partition coefficient can be used as an indi-cator of the tendency to bioaccumulate. The parti-tion coefficient is a measure of the distribution ofa solute between two immiscible liquid phases inwhich it is soluble. For a single molecular species,the partition coefficient is a constant and does notdepend on relative volumes of solutions used. Themost commonly used coefficient indicates relativevolubility in organic and aqueous phases, provid-ing information on the hydrophobic or hydrophilicnature of the compound.

Although absorption and accumulation of toxicpollutants are related to their partition coeffi-cients, partition coefficient alone does not deter-mine bioaccumulation. Other characteristics ofthe chemical (reactivity, vapor pressure) and ofthe organism (metabolic pathways, nutritionalstatus, fat content of body tissue) will also in-fluence the extent of bioaccumulation. Active ac-cumulation of a substance that resembles anotherchemical is frequently predictable from the prop-erties and structures of both,

For inorganic compounds relative volubility isnot the primary factor in determining extent ofstorage in body tissues. Passive accumulation

r .- II - “ 1 - 1 ‘,

184 w Environmental Contaminants in Food

through formation of complexes with organicligands, especially those containing sulfhydryland amine groups, can also occur. The syntheticcompounds most likely to accumulate in this wayare organometallic chemicals. The tendency of achemical to accumulate in this way can be pre-dicted from knowledge of the likelihood of thechemical to form complexes with natural bases.

Population at risk. The size or type of popula-tion exposed may be considered as a factor in set-ting priorities for toxic chemicals. Types of popu-lations at risk could be:

● the general human population,Q population of a geographical region,“ population of a specific neighborhood,● population in the immediate vicinity of an in-

dustrial plant,● people whose diets contain a high amount of

particular foods,● occupational groups, and● highly susceptible groups.Population groups can be further classified ac-

cording to whether their exposure is voluntary orinvoluntary. Priority considerations should begiven to chemicals to which there is extensive in-voluntary public exposure or to which susceptiblesegments of the population are exposed. Someranking schemes consider certain segments of thepopulation more ‘‘valuable’ than others (e.g., theyoung more “valuable" than the old). This ap-proach is highly subjective and ethically ques-tionable.

ToxicityThe toxicity of a chemical is the second deter-

minant of hazard. The difficulties encountered inestimating the biological activity of a chemicalare as severe as those encountered in estimatingexposure. Here we find ourselves at ‘‘the fron-tiers of scientific knowledge” in attempting toevaluate the significance of laboratory data forhuman experience.

Most priority-setting efforts have scored chemi-cals on a series of toxic effects, There are, how-ever, some general indices of toxicity, such as thethreshold limit values for atmospheric exposurein the workplace established by the U.S. Depart-ment of Labor and the American Council of Gov-ernmental and Industrial Hygienists, tolerancelevels for pesticides in food set by EPA, or “actionlevels’” for food contaminants set by FDA. Theseindices are based on reviews of a range of toxicitydata and represent attempts to identify thresholdlevels for hazard or for lack of assurance of safe-ty. Although the reliability of such indices as

precise measures of degrees of hazard is debat-able, where available they are useful in priority-setting exercises because they represent an at-tempt to weigh all toxicity data available at thetime when they were established.

Metabolic factors.1. Uptake and absorption.—Even if the amounts

2.

3.

to which people ‘may be exposed could bedetermined, we still would not know theamount that could be expected to enter thehuman body and cause biological damages.The term “penetrability” is sometimes usedto represent that portion of the chemical towhich the person is exposed that is expectedto be absorbed into the body, This factor in-cludes consideration of both the routes of ex-posure and the fraction of the chemical ab-sorbed after exposure by each route,Biotransformation. -The metabolic conver-sion of a chemical within the human bodyshould be considered, as far as possible, inassessing potential health effects. A chemi-cal can be detoxified, activated, or convertedto a compound with different biological prop-erties. The metabolic pathway can be dose-dependent.Elimination. —T h e t u r n o v e r r a t e i n t h ehuman body can be an important factor inassessing biological effects. If, however, achemical has been scored for bioaccumula-tion as an exposure factor, it may not be nec-essary for it to be scored for turnover inhumans, as these factors are closely related,

Absorption, bioconversion, and elimination arecomplex factors and are difficult to handle forranking purposes, Information is not readilyavailable and is rarely compiled in easy-to-usesecondary sources. For many chemicals, no meta-bolic information is available. When metabolicstudies are found, they are often reported inqualitative rather than quantitative form. Relat-ing metabolic information to adverse health ef-fects often requires a high level of expertise intoxicology. Despite all these problems, a toxico-logical ranking system which takes into accountthese metabolic factors is more sophisticatedthan and usually superior to a system which doesnot,

Acute toxicity. The biological effect of a chemi-cal depends on the quantity of the chemical withwhich the organism must deal, Chemicals havingadverse effects at low dosage would be consid-ered more toxic than chemicals having similar ef-fects only at much higher dose levels. It is difficultto perform this ranking at low effect levels. The

Append/x F—Priority Setting of TOXIC Substances for Guiding Monitoring Programs ● 185

effects of low dosages are often subtle. Direct ex-perimental estimation of the level affecting 1 per-cent of the population may require several hun-dred animals to obtain adequate statistical pre-cision. For these reasons, lethal dosage levels areusually used in ranking chemicals for acute tox-ic i t y.

A common parameter of acute toxicity is the

animals would die. To be comparable, resultsshould be based on animals of the same species,strain, sex, and age. The same route of adminis-t ration should be used.

Other commonly used parameters of acute tox-icity based on lethal doses are:LC-,(; =

LD, ,, =LC,,, =

the concentration which is lethal to halfthe test population. The duration of ex-posure should be specified. LC is usu-ally used for concentration in air, butcan also be used for concentration inthe ambient water to which aquatic or-ganisms are exposedthe lowest reported lethal dosethe lowest reported lethal concentra-tion

Available data on these parameters for 16,500different chemicals are summarized in the Regis-try of Toxic Effects of Chemical Substances p u b -lished by NIOSH. The major objection to acutelethality as a basis for ranking environmentalcontaminants is that the acute lethal doses arevery much higher than those levels one would ex-pect to find in the environment. Furthermore,values for acute lethality can vary widely be-tween species. Use of this parameter can be justi-fied, however, on the basis that the LD-),, is at leasta crude index of biological activity. It is reason-able to consider a chemical w i t h a very low LD,(, tobe a relatively toxic chemical. on the other hand,a chemical with a high LD,,, is not necessarilysafe, since long-term exposure a t low concentra-tions may have such adverse effects as carcino-genicity or damage to the reproductive system.

Data on acute toxic effects other than lethalityare less frequently compiled, are more difficult touse for ranking purposes since they arc not usual-ly reported in numerical form, and may not be in-dicative of effects occurring at the low dosagelevels ordinarily found in the environment. Never-theless, nonlethal acute effects can provide usefulinformation about target organs and should beconsidered as part of the overall picture whensubjective judgments are made.

Carcinogenicity. In principle, it would be usefulto categorize chemicals as strongly carcinogenic,

moderately carcinogenic, weakly carcinogenic,co-carcinogenic, or having no neoplastic effects.There is usually not enough information avail-able, however, to do this. Dose-response curvesare rarely determined for carcinogens. Negativeresults in carcinogenicity tests are usually not ac-cepted as proof of noncarcinogenicity. Compari-sons between carcinogenic “potency’ of variouschemicals have been made on the basis of per-cent age of a test population developing malignanttumors, the number or types of species in whichpositive results have been reported, the types oftumors observed, the lowest dosage causing ma-lignancy, or the lag time between administrationof chemical and observation of tumors.

Chemicals are sometimes classified with re-spect to the degree of certainty of their carcinoge-nicity (e. g., known carcinogen or suspected car-cinogen). EPA has recently attempted to order theNIOSH list of suspected carcinogens according tothe relative degree of concern that might be war-ranted regarding possible human carcinogenicpotential. A four-digit code was used. The firstdigit represented the species in which carcino-genic response was reported. The second digitdesignated the number of different species forwhich a carcinogenic response was reported. Thethird digit was assigned on the basis of route ofadministration. The last digit was a count of thenumber of different species-route combinations.

The most complete source of data and refer-ences related to chemical carcinogenesis is thePublic Health Service’s Survey of C o m p o u n d sWhich Have Been Tested for Carcinogenic Activ-ity. A master index of the series is maintained ontape at the National Cancer Institute headquar-ters.

Evaluations of carcinogenic risk are made bythe International Agency for Research on Cancerunder the auspices of the World Health Organiza-tion. They do not use a formal ranking system.

Mutagenicity, A mutation is defined as any her-itable change in genetic material. The results ofmutations are usually undesirable and may in-clude abortion, congenital anomaly, genetic dis-ease, lowered resistance to disease, decreasedlifespan, infertility, mental retardation, senility,or cancer.

The reliability and accuracy of the predictionof mutagenicity in man from the ability of a chem-ical to produce mutations in experimental testssystems is controversial. Many types of tests arecurrently available for use as indicators of muta-genic potential. Some (e. g., the Ames test, thedominant lethal test) are actual tests for mutage-


nicity. Others test for a variety of events relatingto, or at least correlated with, mutagenesis. Somesuch tests are for chromosome aberrations, DNAdamage or repair, increased rate of sister chro-matid exchange, or other phenomena consideredto be relevant to mutagenic potential. Becauseeven those tests that actually demonstrate muta-genicity are highly specific for detecting a singletype of mutation, results of a battery of tests aredesirable for an evaluation of mutagenic potentialof a chemical substance,

Structure-activity relationships have met withlimited success in prediction of mutagenicity.Many chemical mutagens (but not all) act as elec-trophiles permitting a prediction of mutagenic po-tential of compounds that might form a reactiveelectrophilic species.

Many difficulties are encountered in extrapo-lating to man results obtained in microbial, mam-malian cell culture, or test animal systems. Poten-tial for mutagenic effects in man should be as-sessed only after results of testing in various ex-perimental systems are considered in conjunctionwith other relevant factors such as chemicalstructure, biochemical activity, and metabolismin man.

Teratogenicity. Certain chemicals, when ad-ministered to a pregnant female, can interfere ata critical stage of embryogenesis or organogene-sis and can cause embryonic death or malforma-tion, A chemical with such properties is a tera-togen. Testing for teratogenicity should ideally in-volve administration of the test substance only atthe period critical for organogenesis of the systemsensitive to the teratogen. Sensitive periods of or-ganogenesis are characteris t ic of part icularchemical teratogens.

At the present time, only tests in pregnantmammals are generally accepted by scientistsand regulatory agencies as valid tests for terato-genicity, because the maternal-placental embry-onic relationship cannot be duplicated in sub-mammalian test systems. Other test systemssometimes used as indicators of teratogenic po-tential include bacteria and other unicellularorganisms, somatic cells in culture, tissue culture,organ culture, intact invertebrate embryos, cul-

tured mammalian embryos, and incubating chickembryos. Of these systems, chicks and to a lesserextent fish are the most widely used,

It is difficult to rank chemicals on the basis ofteratogenic potency. The type of dose-responsedata needed for quantification of teratogenic riskis seldom available. Interspecies extrapolation ispoorly understood. There is an enormous range ofvariability of teratological end points. Chemicalstructure has been found to be of little use inpredicting teratogenicity.

Other toxic effects.1, Effects observed in humans, —Reports of tox-

2.

. .ic effects in humans can be based on epi-demiological or clinical observations. Thereare severe problems in interpreting this typeof information. Humans are rarely exposedto only a single chemical substance, It isusually difficult to find control groups withzero exposure to an environmental contami-nant. Results based on occupational expo-sure may underestimate risk, because sus-ceptible segments of the population (infants,children, the aged, the chronically ill) are ex-cluded from the workplace. When workersbecome sick they often stop working and areno longer part of the occupational cohort.Another problem in interpreting human datais that the doses are often unknown.

Chronic toxicity, —Biological effects of achemical administered in low doses over anextended period of time may be quite dif-ferent from acute effects. Chronic exposuretest protocols are more relevant than areacute protocols to the types of exposure dueto environmental contaminants.

Because of the wide range of end pointsobserved during chronic exposure tests,comparisons are difficult to make. Compar-ing one type of effect with another (for exam-ple, comparing impaired fertility with liverdamage) on the basis of severity is highlysubjective. Ranking chemicals on the basis ofchronic toxicity therefore requires a highlevel of scientific expertise.

Appendix G

Approaches to Monitoring OrganicEnvironmental Contaminants in Food*

by John L. Laseter, Ph.D.

INTRODUCTION

Trace quantities of potentially toxic organiccompounds are frequently found in the environ-ment. These compounds sometimes possess prop-erties that may have teratogenic, mutagenic, andeven carcinogenic effects on humans and ani-mals. Some of the compounds, such as pesticides,have been intentionally released into the environ-ment (1), while others have found their way intothe environment by accident, through careless-ness, or as byproducts of industrial processes.Many of these compounds are subject to break-down in the environment as a result of both physi-ochemical and biological processes, such aschemical weathering, photodecomposition, me-tabolism, and biodegradation by micro-organisms(1,2). In many cases, it is not uncommon for thebreakdown products to have greater health ef-fects on humans and animals, than the parentcompounds.

Because of the potentially toxic properties of somany of these compounds and their breakdownproducts, several pieces of Federal legislationhave been enacted in recent years to monitor,evaluate, and control the amounts of the pollut-ants in order to protect our health and environ-ment. As a result, monitoring ] (screening) pro-grams have been established for many types of or-ganic pollutants in the environment. All of theseprograms rely heavily on the ability of analysts tocorrectly identify and quantify these compoundsat the parts per million (ppm) and parts per billion(ppb) levels in a variety of sample matrices (3). Ino r d e r t o s u c c e s s f u l l y o b t a i n u s e f u l i n f o r m a t i o n a tt h e s e c o n c e n t r a t i o n l e v e l s , i t i s e s s e n t i a l t h a tm e t h o d s w h i c h a r e s e n s i t i v e a n d s e l e c t i v e b e d e -v e l o p e d a n d u s e d . W h a t e v e r m e t h o d i s u l t i m a t e l y

* ~~t)r purp{)st’s of this (i[)fwmont. fl)()[l r(?pr(w}nts ,ill st)ll(i,sf)m IS(JI i(i, ;I n(i IICIII i[i forms of f{)()[i pro(iu{ts. including I)l)t t I[v Iw,1 t [’r, ( t )Ils u m[ x ] I]v m:] n.

“I I1(! t[’rms ‘‘m{ Im t I ) r I n:” :] n(i ‘‘ S( rm’n i n~” ;I rc usmi in t (>r-( il(i nq(’,~l)ii I n t il is (iIJ{ L]m(’n t.

selected, it is imperative that it provide une-quivocal results.

Most methods for monitoring and for the analy-sis for trace levels of organic pollutants in the en-vironment necessarily consist of several steps <

These include sample collection and storage, sam-ple workup, and component identification andquantification.

The collection and storage of samples is an im-portant phase of an analytical method if meaning-ful interpretation of the data obtained is to beachieved. The sample selected for analysis mustbe representative of the whole system being ex-amined, and must be free of contamination due toimproper collection and handling techniques.Once collected, the samples must be stored underconditions that will reduce or eliminate changesin their composition.

Another important phase in an analyt icalmethod is the sample workup. During this stage,the sample usually undergoes an extraction proc-ess whereby the compounds of interest are re-moved from the sample matrix. Organic solvents,and in some cases inert gases, are usually em-ployed in the extraction process. Because the ex-traction process is seldom very selective, manyorganic compounds in addition to those of interestare a lso extracted. In order to reduce theamounts of these other compounds that may inter-fere with the analysis, a sample cleanup proce-dure usually follows the extraction process.

The last and most difficult phase of an analyti-cal method is the identification and quantificationstage. The identification process is often accom-plished by comparison of the physical and chemi-cal properties of the unknown compound againstthe same properties of an authentic standardcompound. For complete unknowns, the identifi-cation process can be very difficult. The quantifi-cation process can only be accomplished after theunknown compound has been identified and usu-

187


ally involves comparison of the detector responsefor the compound of interest against the detectorresponse for known quantities of an authenticstandard.

In the analysis for trace levels2 of organic com-pounds in the environment, it is often very diffi-cult to obtain accurate and reliable qualitative aswell as quantitative results. This is evident by thecountless examples in the literature of errors inboth qualitative identifications and quantitativeestimations of trace quantities of many organiccompounds in the environment (5).

Of the many methods currently available forthe qualitative identification and quantification oforganic compounds, few are sensitive and spe-cific enough for meaningful trace analysis. TableG-1 summarizes the techniques for organic analy-sis and some of their advantages and disadvan-tages (4,6).

The most common techniques in use for traceorganic analysis are gas chromatography withthe use of selective detectors (7) such as the elec-tron capture (EC). Hall electrolytic conductivity,and flame photometric detectors; high perform-ance liquid chromatography (HPLC) (8); and com-bined gas chromatography-mass spectrometry(GC-MS). Selective detectors for gas chromatog-raphy are necessary because environmentalsamples are often very complex and the selectivi-ty of the detectors simplifies the analysis byallowing only certain compound classes to be de-tected at any one time. The most powerful of thesetechniques is the GC-MS technique (4,6,9-1 1),since it not only provides qualitative informationof nanogram quantities of single compounds pres-ent in the sample, but also provides informationwhich can be used for quantification of individualcomponents in the sample.

A more recent technique involves the combina-tion of liquid chromatography and mass spectrom-

E’Or purposes Of this II[)fument, ‘‘~ rare levels” is df?finfxi [ls aronccn tr(] t ion helt)w lhf’ IOU p{] rts-pcr-m ill ion level. I)ctf?rtiona t these rf)nrcnt r{) t ions is import;lnt he(tius[! m:inv org:]nirsarc biologir{i 1 Ii’ ~] (’ t ivc even :] t t hf: pa rts-pf!r-t rl I lion If?vel.

etry (12-14). This technique expands the area oftrace organic analysis to the identification andquantification of compounds that are not suitablefor analysis by gas chromatographic techniques.

Table G-1 .—Techniques Available for Qualitativeand Quantitative Organic Analysis

detection Specificity orcommon uses

Detects mostcompounds

Halides, conjugatedcarbonyls. nit riles. di-and trisulfides

Phosphorus, sulfur

Nitrogen, phosphorus

Compound typedetermination

Classical functionalitydetermination

Sulfur, nitrogen,halogens

Best for completeidenti f icat ion,molecular weightstructure, andfunction Confirm anycompound

Append/x G—Approaches to Monitoring Organic Environmental Contaminants in Food ● 189

ANALYSIS FOR EPA PRIORITY POLLUTANTSIN FOODS AND WATER

The recent EPA/NRDC consent decree estab- food samples. An estimate of the space, man-lished an analytical procedure for the analysis of129 priority pollutants (chemical indicators of or-ganic pollution) in industrial waste water (1 5). Ofthese, 15 pollutants are metal, with the remainderbe ing ind iv idua l o rganic compounds and com-pound c lasses . The procedure for the analysis ofthe organics is strictly a GC-MS computer methoddesigned to provide qualitative as well as quan-t i ta t ive informat ion about the presence of thepriority pollutants in waste waters.

B a s i c a l l y , t h e p r o c e d u r e r e q u i r e s t h a t f o u rsepara te analyses be performed on the sample .These are an analys is for the more vola t i le or -

power, and cost associated with the analysis ofthe present 114 organic EPA priority pollutants infood and water samples is given in table G-2. Itshould be pointed out that considerable researchand development effort must be expended toadopt the methods proposed to foods in general.

Figure G-1 .—A Simplified Diagram for theQualitative and Quantitative Analysis of

the Organic EPA Priority Pollutants

involve liquid-liquid extractions, and subsequent NOTES


Table G-2.—Estimated Space, Manpower, and Cost Associated With the Analysis ofPriority Pollutants in Tissue and Water Samples

According to EPA Analytical Protocola

Category Response and/or cost

A. Instrumentation . . . . . . . . . . . GC-MS data system with automated Iiquid injectiondevice.

-$200,000 eachB. Space. . . . . . . . . . . . . . . . . . . . 1,500 ft2 equipped with typicalb laboratory facilities and

furniture including adequate air-conditioning.C. Downtime for instrumentation 30% (this figure can vary from 20 to 50% as a function of

staff experience and logistics support).D. Minimum assignable . . . . . . . Ph. D or equivalent - 1

manpower M.S. or equivalent - 1B.S. or equivalent - 2

Total 4C

Estimated operational cost per routine sample -$2,000 to $2,500 (Scheme One as illustrated infigure G-l).

Estimated operational cost per routine sample -$2,500 to $3,000 (Scheme Two as illustratedin figure G-2, as a function of cleanup difficulty).

a It , ~ a~~umed t-hat the laboratory ha~-~ther ~n~~ing actlvlt Ies and further that samples would be analyzed on Iy once and In

succession The ultimate size and cost of such a monitoring faclllty IS based on processing only a few hundred samples peryear No capital costs are calculated Into the operational cost per sample

bTyplcal laboratory facllltles and furniture Include laboratory benches tables with sinks shelves for rea9ents and equl Pment.vented hoods, refrigerators, laboratory balances and scales, pH meters, hot plates, laboratory glassware, c hem Ical, etc ThisIncludes special toxic chemical and carcinogen processing fac I I ltles Laboratory space of the type desc rlbed dbOW cost onthe order of $lOO/ft Z

CManpower requ ,rements are not Included for maintenance and 1091Stlc suPPort

SCREENING FOR UNSPECIFIED POTENTIALLY TOXICCOMPOUNDS AND CHEMICAL CLASSES IN FOOD AND WATER

There is a possibility that other compounds orclasses of compounds not included in the EPApriority pollutants list will find their way into theenvironment. For this reason, it is essential thatsome form of monitoring system be establishedthat will look for the appearance of particularcompounds or classes of compounds in food andwater over a period of time. A recent NationalResearch Council report on environmental moni-toring (20) recommends the establishment of newmonitoring programs to anticipate pollution prob-lems and to discover environmental pollutants intheir early stages of development so that ap-propriate corrective measures can be imple-mented before the problem becomes unmanage-able, or worse, irreversible. A typical monitoringprogram is the EPA Mussel Watch Program (2 I),the EPA National Pesticide Monitoring Program,and the National Pesticide Monitoring Networkfor Birds operated by the U.S. Fish and WildlifeService.

In establishing a monitoring program, the typeof compounds to be monitored would be selectedas candidate compounds on the basis of theirchemical class, their use, and their suspected tox-icity, These candidate compounds might include

steroids, phenols, amines, halogenated organics,polynuclear aromatic hydrocarbons, and anynewly appea r ing o rgan i c t ha t i l l u s t r a t e s amarked increase in concentration over a period oftime.

The monitoring program design would involvethe establishment of appropriate sampling andsample-handling guidelines, together with themodification of currently used analytical pro-cedures to include the new compounds or classesof compounds. A typical monitoring programwould involve the use of high-resolution gaschromatographic techniques using the universalflame ionization detector (FID) capable of detect-ing traces of known candidate compounds, andcomputer techniques that allow for rapid com-parisons of samples to establish trends. With theuse of high-resolution gas chromatography (em-ploying glass capillary columns”), one can readilyseparate in an environmentally derived sample,

Appendix G—Approaches to Monitoring Organ/c Environmental Contaminants in Food . 191

several hundred organic species, the vast majori-ty of which have not been characterized.

The analytical program would involve a prelim-inary screening to establish baseline levels of thecandidate compounds in samples of foods andwater, or selected indicator species over a givenperiod of time, followed by periodic screening ofsimilar samples to determine changes or trendswith time. If alarming trends are observed, cor-rect ive act ions and measures can be imple-mented.

Of the techniques available for analysis, themethods best suited for the class or classes ofcompounds under consideration would be se-lected for routine monitoring. These would in-clude high-resolution gas chromatography orhigh-performance liquid chromatography andsupporting computer methods which would in-clude the use of internal and external standards,

and use of pattern recognition techniques. 5 T h emonitoring would be set at a concentration levelbelow the actual legal accepted 1evel (for exam-ple, 0.01 ppm or one-tenth of the action level if it isknown),

Should the screening result in an observation ofan increase in levels of either unknown or se-lected compounds or classes of compounds, sug-gesting the entry into the food or water of newmaterials, additional analytical efforts would beemployed to attempt characterization of the newcompound(s) observed. Preliminary informationwould be transmitted to associated toxicologistsfor evaluation and comparison with informationavailable on known toxic compounds.

LABORATORY REQUIREMENTS

Because organic analysis at the submicrogram(ppm and ppb) levels is a complicated and difficultprocess requiring sophisticated instrumentationand expertise, the laboratory selected to performanalyses for trace organics in foods and watermust be equipped with state-of-the art technologyand experienced personnel. The best setting forsuch a laboratory is in a location where it canestablish ties with the R&D community where ex-perts in a variety of disciplines can be found whocan serve and participate as consultants at vari-ous levels. It should be pointed out that analyticaltechniques may require modifications to upgradethe technology and a substantial expenditure offunds to keep the program operating at maximumefficiency and information output.

The typical nationwide monitoring programwould have several regional centers, each well-equipped with the appropriate instrumentationand personnel to perform the analyses and con-duct the program. In addition each center wouldhave a review board composed of senior scientificpersonnel in such fields as toxicology, analyticalchemistry, environmental chemistry, etc., to as-sess and interpret the data developed. To facili-tate the handling of all the data acquired, each

center would be equipped with data-archiving fa-cilities for both GC, LC, and GC-MS data and data-processing methods that would allow for rapid re-trieval, comparison, and evaluation of these data,Some of these data-handling techniques are nowavai lable and others are under development(23-26] for such a central data management sys-tem (see table G-3). Figure G-3 illustrates a pro-posed analytical scheme to monitor for the ap-pearance of unknown organics in addition to thecurrent EPA priority pollutants in food and watersamples.

In support of the nationwide monitoring pro-gram would be a quality control laboratory whichwould coordinate intercenter calibrations, andwould spot-check and confirm selected data de-veloped by the regional laboratories. Of necessi-ty, the quality control laboratory would be betterequipped than the regional centers, so that itcould resolve the problems and issues that can oc-cur during routine analyses. A minimum of 10 per-cent of all samples analyzed by the regional cen-ters would be confirmed by the quality controllaboratory, Additionally, a library of samplesused in the actual analytical studies would bestored for future reference.


Table G-3.—Estimated Space, Manpower, and CostAssociated With the Monitoring of UnknownPotentially Toxic Compounds in Addition to

the EPA Priority Pollutants

Category

A Instrumentation

B Space

C Downtime forInstrumentat ion

D Min imumassignablemanpower

E Estimatedoperational costper routinesampIe

a

b

c

d

(?

Response and/or cost

Small, high throughputGC-MS data-system withautomated Iiquid injected device-$50,000.El/Cl-equipped high-resolutionMS-data system with automated in-jection device -$200,000 to$300,000GC-FID-EC system $15,000 (addi-tional chromatographic systemsmay be required).LC Interfaced into MS system-$20,000Central data management system

Figure G-3.—General Analytical Scheme To Detectand Monitor New Trace Organics and Priority

Pollutants in Food and Water Samples

I

I I

I r,

I I z i ~ratl ,n ~

NOTES

1 A vdrlel} of f r,+c I Ionat 1011 .in[j I>oldl ion pro( (>(1 (J rett ion to I Ia u I(i I I(I u Id ex t rdc t Ion

Appendix G—Approaches to Monitoring Organic Environmental Contaminants in Food ● 193

DEVELOPMENTS THAT WILL IMPACTLOW CONCENTRATION ORGANIC MONITORING

Of all the developments in technology for chem- (FT-IR) detectors; and 3) developments in com-ical analysis, those which will have a greater im- puter system software capable of handling mas-pact on monitoring programs are the following: sive volumes of chromatographic as well as mass1 ) developments in mass spectrometry instrumen- spectrometric data of the type obtained in a moni-tation, such as pulsed positive-negative chemical toring program. Developments in other methodsionization and detection techniques; 2) develop- such as electrochemical techniques and plasmaments in selective detectors for gas chromatogra- chromatography show some promise for trace or-phy and liquid chromatography, such as atomic ganic monitoring, but will only be useful if theyabsorption (AA) and atomic fluorescence (AF) can be coupled to GC or LC systems.spectrometry, and Fourier transform-infrared

ACKNOWLEDGEMENTS

The technical assistance and cooperation of 1. FDA and EPA and Mr. R. Finnigan, of FinniganR. DeLeon and the help of D, Trernbley in prepar- Corpora t ion, for their helpful technical discus-ing this manuscript is gratefully acknowledged. sions.The author also mishes to thank the personnel of

SELECTED REFERENCES

6.

7.

8.

9.1 (),

11.4’2, 3 ( 1978).

15. U.S. Envi ronmen t{]) Pr[)tert i[)n AgencP’, ‘‘Samplingand A II;] lYIS is Procedu r [?s f( )r S r re[?ni n~ I ndus t r i:] IEfflut?nts for Prioritl’ Pollutants,”” ( ;in(innnti, ohio”(Revised, April 1977).

16. 1. R. Poli tzer, S. Gi t hens, B, J, L)(Iw tk, and J. 1,. 1,ase-tf;r, ]. (l]r~)m(lt[)~r. Sci., 13, 378 ( 1 975).

17. B, J . Dowtpr, 1., E, (;ret;n, :]nd J. L . L{]set[?r, AIIIII(;hem., 48, 947( 1976].

18. I. R. Poli tzer, B. J. Ilowtl, :ind J. 1,. I,[]s[?t(!r, (linic[rl(;hf?m., 22, 1775( 1976).

19. I). L\’. Kuehl, An(l). f;hem., 50, 182 ( 1978).20. “‘Environmental hioni torirl~, ~’(llllm(l IL’, N:) t ion:]l

A(:adem; of Scien(:c?s, [\’[lshing ton. ]]. (;., pp. 6 - 8( 1977).

2 1 . E. I). (j[)ldhf:rg, ~’ ‘l’ . 130wen, J. bf’. F’a rrington. (;,Htlrvey, J . 11. hlartin, p l,. p:] rkcr, ~. P\’. Rise-

t)rough, W“, Robertson, E. S(:hnci(i[?r, an(i E. (;aml)l(~.Fl]iiron, CI)ns(>rv. 5, 1 ( 1978).

22. ‘1’. L. Iserlhour, H. R. Kow[llski, and P. (;, Jurs, (;li(;(Jritic(ll }ICJ in An~lJ (jh[~m, (July 1974).

23. E. B. Ov[?rton, (1. F’, Steele, and J. 1,, Las(?ter, ‘ ‘Im-proved Data Prwx?ssing Software fur (;1:{ss (;iipil-

lar~ Separ~]ti(Jn o f (;omplf?x Erliir(]rlnl[?rlt:l! S;lm-pies, ” Pittsl)ur~h (h)nf[?r~?nc[? F’t;h. 27-hli~r. :3, 1978,(;level:~n(i, ohio.” 22,

24. 1. I{. Suffet :ln(i E. R. (jl;]s(?r, /, ( ;h r~)rn~l t{),gr. .${i,, 1 (;,12 ( 1978),

2 5 . 1 . t!’. h:]r[~sok.. ln~lut(ri(~l R ‘/1, ,20, 113 (April 1978).~ [j. J-:. 13. ( )v(lr t on, (;, F. S I (?(?1 [?. i 111( i J. 1,. 1,:] s[; t [: r, J. H i v 1)

H(~t{)l~l [](~n (;lll~)r]lflff)gz{]]~llk (sul)mitt[”(i 1 978].


GLOSSARY OF ABBREVIATIONS AND TERMS USED

AbbreviationsAA—atomic absorptionAF—atomic fluorescenceCI—chemical ionizationDS—data systemEC—electron captureEI—electron impactFID—flame ionization detectorFT-IR-Fourier transform infraredGC—gas chromatography GC-MS--combined gas chromatography-mass

spectrometryHPLC--high performance liquid chromatographyLC—liquid chromatographyLC-MS--combined liquid chromatography-mass

spectrometryMS—mass spectrometryTLC--thin-layer chromatography UV—ultraviolet

TermsAtomic Absorption Spectrometry—a form of spectro-

chemical analysis usually applied to the determina-tion of the elements. The sample is heated to a rela-tively high temperature to cause dissociation of thechemical compounds into atoms. A source of radia-tion characteristic of the element to be determined ispassed through the sample. If the sample containsthe element, absorption of the radiation by the sam-ple atoms occurs and the amount of absorption canbe measured for quantitative determinations.

Atomic Fluorescence Spectrometry—a form of spec-trochemical analysis usually applied to many organiccompounds and some inorganic compounds whichemit radiation energy after they have first absorbedradiant energy of a particular frequency. The sim-plest form of AF is the fluorescence provided by amonoatomic vapor, such as sodium.

Chemical Ionization—a method for ionizing samples formass spectrometric analysis. The ionization of thesample results from the reaction between the samplemolecules and low-velocity reagent ions which re-sults in the transfer of a charged species other thanan electron.

Electron Capture Detector —a very sensitive and selec-tive detector for gas chromatography. This detectorresponds to the presence of a variety of compoundscontaining atoms with an affinity for electrons, suchas the halogens--chlorine, bromine, and fluorine—and o t her atoms, such as oxygen, and sometimes.even sulfur.

Electron Impact—a method for ionizing samples formass spectrometric analysis. The ionization of the

sample results from the bombardment under highvacuum of the sample molecules by a beam of elec-trons, usually at an energy of 70 electron volts (eV).

F l a m e I o n i z a t i o n D e t e c t o r — a u n i v e r s a l d e t e c t o r f o rgas chromatography. This detector is fairly sensitiveand very linear, and responds well to most organiccompounds.

Fourier Transform-Infrared Spectrometry—this is thestate-of-the art of infrared spectrometry. This tech-nique employs minicomputers and Fourier transformmethods in the acquisition of infrared spectra. Ad-vantages of this technique include the making ofmeasurements in a fraction of the time required forthe more conventional methods, and increased sen-sitivity .

Gas Chromatography —one of the most widely usedanalytical techniques for trace organic analysis. Thetechnique is simple and very rapid to use, is extreme-ly sensitive, allowing the use of minute amounts ofsamples, and can be very useful for preliminaryscreening of environmental samples.

Gas Chromatography-Mass Spectrometry—a verypowerful analytical technique that combines thefeatures of gas chromatography and mass spectrom-etry.

High-Performance Liquid Chromatography—high-res-olution, high-speed, and high-sensitivity liquid chro-matography.

Liquid Chromatography —a separation technique thatallows the partition of the sample between twophases, a liquid and a solid, or two liquid phases.

Liquid Chromatography-Mass Spectrometry—a re-cently developed analytical technique that combinesthe features of liquid chromatography and massspectrometry.

Mass Spectrometry—a very powerful tool for provid-ing the structural identity of complex organic mole-cules. Mass spectra furnish information about the

radical ion, which is usually produced by electronbombardment in the ion source of the mass spectrom-etry.

Thin-Layer Chromatography—a form of liquid-solidchromatography conducted on the surface of spe-cially prepared plates. This technique is generallyvery quick and simple to use and is effective in per-forming separations of small amounts of sample.

Ultraviolet Spectrometry—a form of absorption spec-trometry generally suited for analysis of compoundsthat are capable of absorbing ultraviolet radiation.These include aromatic compounds, conjugated ke-tones, and other conjugated compounds, such aspolyolefins.

Appendix H

Analytical Systems for theDetermination of Metals in

Food and Water Suppliesby R. K. Skogerboe, Ph. D.

INTRODUCTION

The analysis of food products or water suppliesfor toxic or potentially toxic elements, e.g., heavymetals, is the central problem addressed in thisreport. An evaluation of this problem in terms ofthe capabilities and limitations of applicable ana-lytical techniques is the primary thrust of thisreport. Thus, recommendations regarding the se-lection of the best analytical methods and tech-niques for the determination of toxic elements arepresented below. Before doing so, it is appropri-ate to delineate the criteria used in the develop-ment of the recommendations and to discuss phil-osophical rationales on which these judgmentalguidelines are based.

The Analytical Process

Any chemical analysis can be divided into threesequential steps: 1) collection of the sample(s);2) chemical and/or physical preparation of thesample(s) for analysis: and 3) measurement of theconcentrations of the target constituents in thesample(s). Although these steps are interdepend-ent and should not be considered otherwise, thepresent discussion will focus on sample prepara-tion and measurement. Sampling should be dis-cussed in the broader context of the overall prob-lem of monitoring. The selection of an appropriateanalytical technique must be based on the type ofinformation desired and the purposes of collect-ing that information. In the present context, thismay be delineated in fairly general terms.

The Monitoring Question

With rare exceptions, the central question as-sociated with a monitoring program is:

● Are there one or more chemical entities pres-ent in the target material (e.g., food or water)

in sufficient quantities to cause deleteriouseffects on the consumer population?

This question may be considered qualitative inthat it actually requires only a yes or no answer.Given the knowledge that each of the chemicalsbeing monitored must be present at or above somethreshold’ concentration before they individuallyor collectively produce observable effects on therecipient population, the answer is no if all arebelow their respective threshold effect levels andyes if one or more is above. As a result, it is quitecommon to use analysis approaches capable ofdetermining or detecting the chemicals only atlevels down to, but not below, their respectivethreshold levels. Although this practice can oftenbe justified on an economic basis, it is subject tochallenge for scientific reasons.

The astute recipient of the answer to the cen-tral question will immediately raise other ques-tions regardless of whether the answer is positiveor negative. If informed that all constituents ofconcern are below their individual or collectivethreshold levels, two questions are obvious:

● What degree of confidence can be assignedto the results?

195

196 c Environmental Contaminants in Food

● Are the concentrations of any chemicals in-creasing significantly over time for a partic-ular monitor and/or a particular collectionsite?

When advised that one or more constituentsare present above threshold levels, these sametwo questions are raised and a third becomes per-tinent, i.e.,

● What is the cause or source of the observedcontamination?

These questions are clearly quantitative innature. Decisions regarding possible impacts onconsumer populations, or the prevention thereof,should not be based on less than quantitative anddefensible information. Adherence to this philoso-phy is complicated by the facts that reliable desig-nations of threshold effect levels are often lackingand that two or more contaminants may act syner-gistically or antagonistically.

There is one general criterion that may be de-fined on the basis of the above discussion. Allanalyses should rely on analytical methods thatare capable of determining all target contami-nants at concentration levels below the thresholdeffect levels. While methods capable of makingthese measurements at concentrations 100 to1,000 times below these levels would clearly bedesirable, the use of methods providing measure-ment capabilities 10 to 100 times below the levelsmay prove more practical on an economic basis.

This discussion provides a basis for delineatinga general protocol for the operation of the analyti-cal laboratory responsible for answering theabove questions. A brief discussion of this is pre-sented here to provide a general basis for suc-ceeding topics.

OPERATIONAL PROTOCOL FOR A MONITORING LABORATORY

A general flow diagram of the laboratory oper-ation is presented in figure H-1. The analysis se-quence given is depicted in the context of the mon-itoring questions, the checks required to validatethe results, and the regulatory actions likely toprevail. Examination of this protocol indicates theprobable need for inclusion of: 1) a quality-assur-ance program as a means of validating results;2) complementary analytical methods to ensure

that all required analyses can be completed andto provide comparative information relevant tovalidation of results; and 3) a data storage-re-trieval system consistent with the requirements ofthe monitoring questions, the quality assuranceprogram, and regulatory actions. The general util-ity of this protocol diagram and the actions it por-trays will be expanded upon below.

CRITERIA FOR SELECTION OF ANALYSIS METHODS

The selection of an appropriate analysis ap-proach must be based on the information require-ments. These may be formulated on the basis ofthe answers to two questions:

. What are the chemical entities that must bedetermined?

c What are the concentration levels (thresholdlevels) of primary concern for these targetentities?

Although answering the first question shouldnot be particularly difficult in most instances, theissue for toxic elements has often been obscuredby the use of expedient analysis approacheswhich may or may not provide the information re-quired. Historically, most analysis methods usedfor the determination of elements simply measurethe total amount of an element present withoutdifferentiating between the various chemical

states of the element that may be present. Suchmeasurements may, in fact, be relatively nonspe-cific indicators of potential or actual deleteriouseffects on biological systems. In numerous in-stances, the identification and measurement ofthe active or functioning forms of the elements isactually needed. The following examples testify tothe general importance of this statement.

1.

2.

3.

Although-arsenic is toxic, the plus three oxi-dation state (As(III)) is clearly more toxicthan the plus five state (As(V)); the compoundarsine (AsH,) is perhaps the most toxic chem-ical form of arsenic.Although chromium is classified as a nutri-tionally essential element, Cr(III) is toxicwhile Cr(VI) is relatively innocuous.A measurement of the level of vitamin Bl~ inanimal tissue is ordinarily more useful than a

A p p e n d i x H — A n a l y t i c a l S y s t e m s f o r t h e D e t e r m i n a t i o n o f M e t a l s i n F o o d a n d W a t e r S u p p l i e s ● 1 9 7

+ 7 ’+7

element (in all of its chemical forms) can be ra-tionalized. For example, if the analysis indicatesthat the total concentration is below the thresholdeffect level for any one or combination of the par-ticular chemical forms for an element, the needfor measurement of the concentrations of thechemical forms is negated. The total concentra-t ion measurement consequent ly serves as ascreening device indicative of the possible needfor chemical form measurements.

Once the elements and the concentrat ionranges of interest for each have been defined, thecriteria which must be invoked in the selection of

ty, selectivity, reliability, scope, sample prepara-tion requirements, and time-cost considerations.It would be folly to select an approach incapableof providing measurements a t or below the re-quired concentration (threshold effect) levels.Thus, selection of a sufficiently sensitive methodis of paramount importance. Although it is com-mon to discuss sensitivity in the context of theterm detection limit, it is also usually impossibleto obtain a sufficiently accurate concentrationmeasurement when said concentration is barelydetectable. Thus, sensitivity should be consideredin association with the term determination limit,i.e., the lowest concentration at which a suffi-ciently reliable concentration measurement canbe carried out.

The scope of an analytical approach is definedon several bases. The ability to determine reliablylarge numbers of individual chemical constituentseach of which may be present in a variety of sam-ple types within a broad concentration range isthe primary connotation, Thus, universal applica-bility (utility) is an alternate terminology implyingthe ideal case. It can also be argued that an ana-lytical method which requires simple samplepreparation operations offers greater scope thanone requiring more complex preparative steps:the ideal case involves direct analysis of thesamples without prior chemical treatment.

Analytical selectivity is usually interpretedsynonymously with specificity. The analysis ap-proach used must provide an unequivocal meansof identifying each chemical constituent of inter-est irrespective of the compositional characteris-tics of the sample material being analyzed. In ad-dition, the measurement of the concentration ofa n y constituent of interest should provide accu-rate results independent of the variations in thec o n c e n t r a t i o n s of other constituents present inthe sample materials. Analysis methods that donot satisfy these conditions lack selectivity (or


specificity): such methods are subject to what areoften termed interference or matrix effects, Inessence, reliability (accuracy) is the criterionwhich defines a sufficiently selective analysis;confidence in the results and the decision(s) basedon them is explicit.

Minimization of the time-cost commitmentsmust be an objective considered in the selection ofan analysis approach. It is unfortunate that thisconsideration often leads to one of two extremestances. Management often opts for the adoptionof analytical technology of limited utility on thebasis of lower initial capital costs and/or the factthat such technologies frequently can be appliedby analysts with lower levels of expertise. Such

selections are often economic errors when consid-ered on a longer term basis, At the other extreme,management may invest in expensive facilitiesand expertise and then require that they be usedto carry out analyses that can be accomplishedmore economically via another approach. This,too, can prove to be a false economy.

Summary

This background discussion has been pre-sented as a preface to the evaluations presentedbelow; the intent has been to provide a commonbasis for comments, evaluations, and suggestionswhich follow.

PREMISES FOR SELECTION OF LABORATORY FACILITIES

In considering the operation of a laboratorydedicated to monitoring food and water suppliesfor toxic elements, several premises may be es-tablished. These are essential in defining the nec-essary facilities and require discussion. The fol-lowing list is not necessarily all inclusive nor is itset down in order of importance.

1.

2.

3.

4.

The majority of analysis requests will re-quire the determination of several elementsper sample.To be most effective, the laboratory must beable to comply with these requests in a rea-sonable (short ) time.The analyses carried out should be accom-plished at reasonable costs.The analytical results must be sufficientlyaccurate to avoid challenge of the integrityof any decisions based on them.

The combination of these premises clearly im-plies that the ideal laboratory facility would beone capable of accurately analyzing for all con-stituents requested at concentration levels downto and below their respective threshold effect lev-els in a short period of time. The use of a systemcapable of simultaneously measuring all constitu-ents of interest in each sample is definitely im-plied. The time and cost premises further implythe desirability of utilizing analysis techniqueswhich do not require extensive sample prepara-tion operations: the ability to directly analyze

samples in an “as received” form may be consid-ered ideal. Sample preparation operations arealso primary sources of contamination or loss ofthe analytical constituents. The direct analysis,minimal preparations, capability is desirablefrom the accuracy standpoint as well. Finally, theaccuracy requirement indicates the need for ahighly specific analysis approach which is notsubject to significant interference problems andthe maintenance of a quality assurance program.No single analytical technique will necessarilysatisfy all of these requirements for the elementsand/or sample types of interest; a combination oftechniques will be required. Properly selected,the techniques used may be complementary interms of providing the range of elemental anal-yses required and in terms of providing redun-dant analyses for some elements. The latter willbe useful for accuracy validation purposes (seediscussion below).

In effect, these premises lead to the defensibleconclusions that: 1) the laboratory facil i t iesshould be primarily comprised of multicomponentanalysis systems; 2) more than one such systemwill likely be required: 3) the selection of systemsthat will provide some redundant analytical in-formation is desirable; and 4) the inclusion ofsecondary analytical systems to be used in sup-portive capacities may be essential. The followingevaluation is predicated on these bases.

Appendix H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 199

EVALUATION OF ANALYTICAL TECHNIQUES FOR THEDETERMINATION OF

The qualitative identification and quantitativedetermination of toxic or potentially toxic ele-ments in food products and water can be based onseveral different analytical techniques for whichcommercially available instrumentation exists.The primary techniques to be considered include:

1.2.

3.

4.

5.

6.7.

neutron activation analysis (NAA);molecular absorption and fluorescence spec-trophotometry:solids (spark source) mass spectrometry(SMS);atomic absorption (AAS) and atomic emis-sion spectrophotometry (AES);electrochemical techniques—anodic stripping voltammetry (ASV),—differential pulse polarography (DPP);plasma emission spectrometry (PES);X-ray techniques—X-ray emission spectrometry (XES),—proton-induced X-ray emission spectrom-

etry (PIXE).Each of these are considered below.

Neutron Activation Analysis

This technique is perhaps the most sensitive ofthose available when all elements are considered.Absolute determination limits, expressed as nano-grams (1 x 10-(J gm) that must be present in thesample for quantitative determination, are sum-marized in table H-1. Such determination limitsshould be converted to actual concentrations tolend them perspective for the present evaluation.For the analysis of food products or water, a 10-gm or 10-ml sample generally represents a rea-sonable upper limit on the sample size that can beactivated for analysis. Thus, taking 0.5 ng as thedetermination limit for arsenic as an example(table H-1) this element can be determined at ap-proximately 0.5 ng/10 gm = 0,05 ng/gm (rig/ml forwater) or 0.05 parts per billion by NAA. Evalua-tion of the other elements of interest on this prox-imate basis indicates that most could readily bedetermined at required concentration levels viathis technique. While this is an encouraging con-clusion, there are other factors which detractfrom it.

TOXIC ELEMENTS

Table H-1 .—Determination Limits for NeutronActivation Analysesa

—

Element

Ag ... . . . .As . . . . . . . . .Be . . . . . . . . . . . . .BiCd : : : : : : : : : : : :Co . . . . . . . .Cr. .,C uFe . . . . . .Hg . .Mn . . . . . . .Mo . . . . .NiPb : : : : : : : : : : : :Sb . . . . . . . . . .Se . . . . . . . . . .Sn ... .Te. . .TI ., .V . . . . . . . . . . .Zn. . . . . . . . . . . . . .

Concentrationvalue, ppbc

0.0050.052.5

252,50.25

500.05

2,5000.50.00352.5

5000.25

25025

2.5—0.055

aData taken from R- K Skogerboe and G H Morrison, Trace Analys!s E ssentlal Aspects, In Treatise on Analy~lcal Chernls(ry (1 M Kolthoff and P J Elvlng, eds ), New York Wiley and Sons, 1971, pp 5842-5843

bNanograms of element that must be present In sample activated to Permit a

quantitative determlnatlon, detection Ilmlts are approximately a factor of 5lower Interference free measurement condt!lons are assumed

cAss\, mlng a 10 gm samp Ie IS act Ivated, thus concen trat Ion values = absolute

values – 10 These values should be Increased by a factor of 100 to 1 000 !Ocompensate for the 10ss of neutron flux I f a neutron accelerator were used I nstea[j of a nuclear reactor

To achieve the sensitivity required, the use of anuclear reactor providing a high flux of thermalneutrons is required. The acquisition cost of thereactor is several million dollars; the operationalcosts are also comparatively high. Considerablereduction of these can be achieved by replace-ment of the reactor with a neutron accelerator.The neutron fluxes available with such accelera-tors are, however, about a factor of 100 to 1,000less than those typical of a reactor. The analyticalsensitivity available is reduced in proportion [seefootnote to table H-1).

If chemical separations are carried out afteractivation of the samples, proponents of NAAargue that analytical interferences can be virtu-ally eliminated. This argument is often specioussimply because the use of chemical separations is


often a primary source of errors. Since such sepa-rations, even when free of errors, are time con-suming and add to the expense; activation ana-lysts usually prefer what is called the purely in-strumental approach.

Two types of interferences are common to thisapproach. Direct interferences occur when twoor more sample constituents emit radiation (gam-ma rays or beta particles) of nearly the same en-ergy. If the emitting species undergo radioactivedecay at significantly different rates, correctionfor these interferences can be based on measure-ment of the radiation at different times. Other-wise, the interfering species must be chemicallyseparated prior to the measurement step. Indirectinterferences due to contributions to the samplespectrum from Bremsstrahlung and Compton in-teractions are also common, Correction for thesemust be obtained by subtraction techniques. Thenet effect is, however, a reduction in the analyt-ical sensitivity and/or an increase in the uncer-tainties associated with the measurements, Whilethese instrumental approaches are widely used inNAA to avoid the need for chemical separations,such avoidance still restricts the potential scopeof the analyses. They also tend to lengthen thetime required to obtain analysis results. On thesebases, NAA does not appear to be the best choicefor a laboratory facility: capital costs, operationalcosts, and operator training requirements are pri-mary weighting factors influencing this negativejudgment.

Molecular Absorption andFluorescence Spectrophotometry

These techniques have been used extensivelyfor elemental analysis. They generally rely on car-rying out a reaction of the element of interest witha reagent (or series of reagents) to form a productwhich has properties required for the absorptionof light. Identification of the element incorporatedinto the light-absorbing product is based on thewavelength of light absorbed while measurementof the concentration relies on the extent of ab-sorption; hence the term molecular absorption,Some light-absorbing species regain a more stableenergy configuration by release of the light en-ergy absorbed as light (fluorescence or phosphor-escence). Measurement of the wavelength atwhich this occurs and the intensity of the lightemitted is used to identify the species responsibleand the amount present,

The use of these techniques requires that spe-cific or semispecific chemical reactions be used

for the formation of the absorptive and/or fluores-cent products, While such reactions are generallyavailable for the elements of interest, the analysisof any one sample for several constituents wouldnecessarily have to rely on carrying out severalindividual reactions. Even then, there are onlylimited instances for which a reaction will occurfor only one constituent (a specific reaction). Mostreactions involving a particular reagent set tendto occur for each of several elements having simi-lar properties and, as a result, their absorption offluorescence spectra tend to be quite similar.Such spectra are subject to some degree of wave-length coincidence such that spectral interfer-ences are not uncommon; the measurements areless specific than desirable, While other featuresof these techniques could be discussed, theirchemical reaction requirements coupled with thespecificity problem are deleterious from the mul-ticomponent analysis standpoint. The techniquesshould not be classified as essential for the pres-ent elemental analysis purposes.

Solids Mass Spectrometry

This analytical technique offers sensitivitycompetitive with that characteristic of NAA (seetable H-1) for a wide range of elements. To utilizeit, nonconducting samples must be rendered elec-trically conducting. The constituents of an aque-ous sample would ordinarily be analyzed after thewater was evaporated away; the residue wouldbe mixed with a conductor for analysis. Solidmaterials such as foods are also typically non-conductive; by mixing them with a conductor suchas graphite or silver powder they become conduc-tive. Food products must also be oxidized (by wetor dry oxidation techniques) to destroy the organ-ic constituents which cause serious spectral in-terferences in the analysis step.2 These samplepreparation steps are quite extensive and can ob-viously be the sources of serious errors unlesscarried out with caution.

In practice, the analyses must be based on twosample mixtures to obtain a complete analysissubject to less interference problems. Elementalsubset A may be determined based, for example,on mixing the sample residue or ash with high-purity graphite. However, the carbon polymers ofwhich graphite is composed are observed in themass spectrum and preclude the possibility ofanalyzing for those elements that would normally

See the following” reference for [i more complete discussionof the prohlems tind r;]p{~bilities: C. A. Evans and G. H. Mor-rison, Anal. (;hem., W, 869 ( 1969).

Append/x H—Analytical Systems for the Determination of Metals in Food and Water Supplies ● 201

be measured at these mass positions. The analysisfor elemental subset B would subsequently bebased on the use of high-purity silver, gold, cop-per, or aluminum powder as the conducting ma-trix. Again, the use of any of these choices toachieve sample conductivity results in spectral in-terferences which necessitate analyses based ontwo or more subsets of elements. The conductivematerials used must be high purity, i.e., 6-9s or99.9999-percent pure. This requirement limits thepossibilities and affects their costs, Finally, itshould be noted that solid mass spectrometricanalyses require unusually long times eventhough it is possible to obtain analyses for -40 to60 elements on each sample. Although the samplepreparations required are time consuming, theanalysis itself is also rather slow. Given samplesready for analysis, a well-organized SMS labora-tory would be hard-pressed to analyze more than5 to 10 samples per man-day. These factors, cou-pled with high acquisition costs (~$250,000) andhigh operational and instrument maintenancecosts, place solids mass spectrometry in a nega-tive posit ion relative to other possibilities,

Atomic Absorption andEmission Spectrophotometry

Recent instrument sales figures indicate thatonly gas chromatography is more widely usedthan atomic absorption spectrophotometry. Whena flame is used as the energy source required toproduce the gaseous atomic populations from thesample dissolved in aqueous solution, atomic ab-sorption offers favorable analysis capabilities fora reasonably impressive array of elements (seetable H-2]. Atomic emission from the same atompopulations in the flame is totally complementaryand supplementary to atomic absorption. Factorswhich affect absorption also tend to affect emis-sion; the ultimate sensitivity achieved with eithermeasurement approach is limited by the ability toproduce the atomic populations. As a result, itcan be shown that atomic absorption is generallymost favorable (on the basis of sensitivity) for thedetermination of those elements requiring moreenergy to produce atomic emission, i.e., an excita-tion potential above approximately 4.5 electronvolts (eV). Elements with lower excitation poten-

table H-2.—Determination Limits for Flame Atomic Absorption and Emission and Furnace Atomic Absorption’

Elements

AgA s

B eB iC dc oC rCuFeHgMnMoN iP bSbSe.S nT eT IV .Z n

Flame methods b

Atomic Atomicabsorpt ion emission

002050010200050030020010.021 0001020020.050505010 50101001

0.0450.5220.10020050.12.00.02050.1052

102

2000.10.050.2

Flame AAanalysis of

tissue digestsc

0.250.120.050.30.20.10.2

100.120.20.555151101

Furnaceatomic

absorption

0000040.02000020.0010.000020.0010.0010.0010.00060.020.00010,0080.0020.0010.0060.020.010050.00060020.00002

Furnace AAanalysis of

t issue digests c

0.00040 20 0 0 20.010.00020.010.010.010.0060.20.0010.080.020.010.060 20.010 50.060 20.0002

‘Data I rom J D VJI n~ fordner el al APPI SPec trosc Rev/e ws 712J 147 (1973) A II va I ~jes g lven In P 9 m I for aqueo~ls sol ~~ tlonstlDeflne{j for ~)e~t flafllf> con(jltlons for each element values In l,q ml ~ppm~r Based on [1 I SSOIVI n q 10 q m o f h(r]locjlcal I I s$ue I wet welg ht I per 100 m I of dc Id va Iues I n Ig g m (ppm) wet wmg h Id~ef, nP(f for ~lDt, n),)” f”, “ac e ~ ~ncj, 1,Ons , n ea~ h case and “se of a 25,, I sample lnjec I Ion values In ,,g m I (ppm)


tials are best determined by atomic emissionwhen flames are the energy media. These gener-alizations require the use of an instrument that iswell-designed for both absorption and emissionmeasurements (two of the three major U.S. atomicabsorption manufacturers supply such instru-ments).

In spite of the combined capabilities offered byflame emission and absorption, they are frequent-ly inadequate for the types of analyses in questionunless preconcentration and/or separation proce-dures (e.g., solvent extraction) are used. These in-adequacies may be due to a lack of sufficient sen-sitivity and/or to the occurrence of interferenceeffects. These, coupled with the difficulties at-tendant to separation/preconcentration proce-dures, have led to the development of what isoften called nonflame atomic absorption. The firstsignificant development in this area involved thereduction of mercury ion in aqueous solution toatomic mercury (Hg) such that it could be carrieddirectly to the gas phase into the optical path ofan AA instrument for measurement. The meas-urement actually involves separation of the Hgfrom the sample under conditions less subject tointerferences. If carried out under appropriateconditions so that the Hg arrives at the measure-ment cell rapidly, an effective preconcentration isalso realized. In effect, the measurement sensi-tivity is determined by the concentration of atomsdelivered to the measurement cell per unit time.The nonflame methods are all designed to take ad-vantage of this thereby enhancing the ability toanalyze at lower concentrations; methodologicaldevelopments for this purpose have taken twogeneral tacts, i.e., chemical generation and fur-nace vaporization.

Beyond the Hg method described above, thehydride generation method for the determinationof arsenic, selenium, germanium, lead, tellurium,tin, antimony, and bismuth has received wide at-tention. The hydrides of these elements are rapid-ly formed by reaction in acid media with sodiumborohydride. The metal hydrides, being gases atambient temperatures, are readily transportedvia carrier gas to a flame or a heated ( - 800° C)quartz cell for atomization and measurement. Thesuccess of this general approach has led to thecommercial availability of hydride generator ac-cessories for atomic absorption. It has also beenadopted in commercial autoanalyzer systems. De-tection limits of less than 1 ng metal/ml (1 ppb)solution are readily obtained.

Furnaces, fabricated from graphite or tung-sten, that can be temperature ramped by resist-

ance heating have been developed to high levelsof refinement. Liquid samples are delivered to afurnace, which is resident in the optical path ofan AA unit, and a programed heating cycle is ini-tiated, In this cycle, the liquid is first evaporateda t ~120° C; the salt residue remaining is “ashed”to convert it to a “common” chemical form at-3000 to 50 0

0 C; and the ash is rapidly vaporizedand atomized at -2,0000 C for the absorptionmeasurement. The use of such systems providesimprovements in the analytical sensitivity formost elements (see table H-2). The extent of im-provement is limited by the amount of sample thatcan be placed in the furnace ( ~25 to 50 µ1) butgenerally amounts to a factor of 10 to 100 whencompared with flame AA capabilities.

These capabilities combined with the generalsimplicity of operation and lower instrument costshave been largely responsible for the widespreadacceptance of atomic absorption. During the prin-cipal time of AA development, the primary com-peting techniques were flame, arc, and spark-emission spectrometry. The last two were ratherquickly abandoned by the analytical spectrometrycommunity because they were “notably subject tointerference problems associated with the vapori-zation-atomization system. It is rather ironic thatthe same community, in less than 10 years, re-verted to the use of furnace AAS systems whichare subject to the same interference problems forthe same reasons. It is also ironic that an exten-sive fraction of the furnace AA research reportedin the past 5 to 10 years has dealt with the studyof interference effects and means for their elimi-nation. A large percentage of these studies arereaching the same conclusions and developing thesame compensatory methods that resulted fromarc and spark spectrometry interference effectstudies before the advent of atomic absorption.Nevertheless, atomic absorption analysis is well-established and here to stay. The furnace meth-ods, in particular, clearly offer the required sen-sitivities for a wide range of analysis problems.

A primary historic limitation has been thatatomic absorption has been basically a single ele-ment analysis technique; analyses for several ele-ments in a sample are carried out sequentially intime. The emergence of plasma emission spectro-metric systems which allow the simultaneousanalysis of several elements (to be discussedbelow) has forced the atomic absorption commu-nity to the development of multicomponent anal-ysis systems. Only one commercial instrument of-fering this capability for more than two elementsis available at this date, This unit is, in fact, a se-


quential analysis system since the elements of in-terest are not measured simultaneously. Its usesaves analysis time but not as much as it would inthe simultaneous mode. At least one other manu-facturer will introduce a truly simultaneous multi-element AA analysis system in the next 1 to 2years. Prototypes of such a unit have been devel-oped and tested at Colorado State University.3 Thetests have shown that sets of 5 to 10 elements canbe determined simultaneously without sacrifice ofanalytical sensitivity. The availability of such in-strumentation will advance the state of the art foratomic absorption.

The inclusion of atomic absorption in an ele-mental analysis laboratory may be consideredworthwhile as a secondary facility at least. If saidinstrument offers the simultaneous measurementcapability, the inclusion may be justified on a bet-ter economic basis.

Electrochemical Techniques

These techniques are perhaps the most classi-cal of those considered herein, And yet, electro-chemistry has been reborn in the past two dec-ades largely through the development of whatmay be classified as pulse or differential tech-niques. The electrochemical techniques of pri-mary interest in the present context rely on themeasurement of the Faradaic current (FC) pro-duced or used during oxidation or reduction reac-tions involving the element to be determined; thepotentials at which FC changes occur depend onthe elements (ions) involved in the reactions. Suchreactions are almost universally accompanied bynon-Faradaic processes involving other sampleconstituents which also result in the productionor utilization of current. In classical direct cur-rent polarography or voltammetry, the ability tomeasure a low Faradaic current (low analyte con-centration) is limited by the magnitude of the non-Faradaic current changes occurring simultane-ously. The revitalization of electrochemistry hasbeen largely based on the fact that when the po-tential required to induce a redox reaction is re-moved, the non-Faradaic current decays morerapidly than the Faradaic current . Thus, by“pulsing” the potential up to the reaction level re-quired, shifting it back to below the reaction level,and waiting an appropriate time period for thenon-Faradaic current to decay; the Faradaic tonon-Faradaic current ratio can be significantly

‘See for example: F, S. Chuang, D. F. S. Natusch, and K. R.0’Keefe, Anal, Chem. 50, 525 [1978).

improved. A variety of these differential pulse ap-proaches have been developed to achieve analyti-cal measurements at concentrations 10 to 1,000times lower than previously possible. The tech-niques which appear to offer the most significantcapabilities are differential pulse polarography(DPP) and differential pulse anodic stripping volt-ammetry (DPASV). At least one commercial in-strument offering both of these measurement cap-abilities is available.

The electrochemical literature indicates that-16 to 20 elements can be determined by DPP,DPASV, or cathodic stripping voltammetry (CSV).This is true, but all such elements cannot be de-termined under a single set of experimental con-ditions. Some must be determined using a goldworking electrode (for example) while others re-quire the use of a mercury electrode. Some deter-minations require the use of a specific supportingelectrolyte solution while others do not. Finally,the redox potentials of some entities are suffi-ciently similar that they cannot be determined inthe presence of each other without the use of cor-rection techniques. In some cases, prior separa-tions are required.

The electrochemical analysis literature indi-cates that Cd, Cu, Pb, Zn, Tl, and Fe can usuallybe determined simultaneously in a single support-ing electrolyte solution using mercury as theworking electrode material, These elements canbe determined at levels of 1 ng/ml or less in theelectrolyte solution by DPP or DPASV. As a result,these techniques are quite widely used for theanalysis of the above elements. Although Cu andZn interfere with each other via formation of theCu:Zn intermetallic in the mercury electrode andthe reduction of Fe(III) to Fe(II) interferes with theCu determination, these can be corrected via useof expedient instrumental procedures. Other ele-ments frequently determined are As(III) byDPASV or DPP and Se(H) by CSV. Again, analysesat or below the part per billion concentrationlevel are common.

The electrochemical methods detect only theelectroactive species, e.g., the ions, This state-ment must be qualified in terms of the time-scaleof the measurement step. To illustrate, consideran electroactive metal ion (M+ 11) which may bepresent in the sample solution primarily as a com-plexed or molecular species designated, for exam-ple, by MXn. To measure the total M+ n plus MX n

concentration, the following reaction must occureither prior to or during the measurement period:

MXn-, M + n + nX -

If the duration of the measurement is short, theabove reaction must go to a reproducible state of


completion within that time period; in such casesthe complexed/molecular entity may be classed aslabile. Thus, there is a growing investigative ef-fort in the electrochemical community involvingthe use of this conceptual approach as a meansfor the identification of the chemical forms ofmetals in natural systems. The general thrust ofthese efforts involves either the use of chemical orelectrochemical means for systematically shiftingthe above type of reactions toward the formationof an electroactive species for identification pur-poses. A typical electrochemically induced shiftexperiment, for example, might involve ASV. Byelectrodepositing from the sample solution forsuccessively longer periods and measuring theamount of M + I’ reduced during each period, thetime required to drive the above type of reactionto completion might be deduced. This may be in-dicative of the “lability” of the complexed/molec-ular form, i.e., its thermodynamic or kinetic stabil-ity, Extensive research must be completed to es-tablish the potential utility of such approaches.However, these possibilities combined with thehigh sensitivities that can be achieved with mod-ern electrochemical techniques suggest that suchfacilities should prove to be valuable laboratoryfacilities. (See further discussion below.)

Plasma Emission Spectrometry

It has been suggested that atomic emissionspectrometry offers what may approximate theideal approach to multielement analyses. Indeed,the use of the radio frequency inductively coupledplasma (ICP) as the atomic excitation source incombination with a direct-reading emission spec-trometer permits the simultaneous determinationof numerous elements at low concentration lev-els. 4 Similarly, the use of a direct current plasma(DCP) excitation source for direct-reading emis-sion spectrometry shows comparable promise.The evolution of such systems over the past dec-ade has brought emission spectrometric analysisback to the forefront of analytical capabilities.Such plasmas are principally used to excite ana-lytical constituents delivered to them via solutionnebulization (aerosol production) systems. Theconversion of solid samples to aqueous solutionsfor analysis (e.g., by wet oxidation) results inelimination of many of the major compositionaldifferences between samples such that interfer-

4 A good introductory review has been presented by V. A.Fassel and R. N. Kniseley, Anal. Chem, 46, 11 10A (1974). Seealso, %ience, 202, 183( 1978).

ence effects due to matrix differences may beeliminated or reduced. As a result, a single set ofoperational conditions may be used for the simul-taneous determination of 20 to 60 elements,

Quantitative determination limits for the ICP-and DCP-Multielement Atomic Emission AnalysisSystems (MAES) are listed in table H-3. Examina-tion of these data indicates that these systems areadequate for the simultaneous determination of amajor fraction of the elements of interest at levelscommensurate with the anticipated threshold ef-fect concentrations. This is one reason why theICP-MAES and DCP-MAES manufacturers haveenjoyed significant annual sales improvementsover the past 5 years.

Some elements of high concern, because oftheir toxicities or propensities for bioaccumula-tion, cannot be determined at low enough concen-trations by direct solution analysis, e.g., Hg, As,and Se. However, the hydride generation ap-proaches used to solve this problem when AAS isthe analysis method, are equally applicable toMAES. In fact, all hydride-forming elements can

Table H-3.—Multielement Atomic EmissionDetermination Limits for Two Common

Plasma Excitation Systemsa

ICP-MAES DCP-MAESFor tissue Tissue

In solution, digest in solution, digest~lglml ~lg/gm (ppm)b ~~g/ml g/gin (ppm)b

0.01 0.1 — —0.3 3 0.2 20.02 0.2 0.02 0.20.002 0.02 0.05 0.50.2 2 — —0.01 0.1 0.05 0.50.01 0.1 0.08 0.80.02 0.2 0.02 0.20.005 0.05 0.01 0.10,01 01 0.01 0.10.1 1 0.01 0.1001 01 0.05 0.50.03 0.3 0.05 0.50.05 0.5 0.05 0.50.1 1 0.05 0.51 10 — —0.2 2 0.2 2.01 10 1 100.4 4 — —1 10 — —0,2 2 0.5 50,02 0.2 0.02 0.20.01 0.1 0.02 0,2

aData for the Icp trom R K Wlncje et al S~ectrochlm Acta, 326 327 ( 1977)Data for the DCP from R K Skog;rboe H E Taylor and G W Johnson, Spec-trochlm Acta, In press

bFor 10 ~m O( biological Ilssue (wet wt ) dissolved In too ml of Solutlon


be determined at required concentrations on a si-multaneous basis. (Tissue detection limits of 0.01µg/gm (ppm wet weight) can be achieved. )

Although the measurement of atomic emissionoffers a high degree of qualitative specificity,there are two common types of interferenceswhich affect the quantitative specificities of suchsystems. Both plasma types are highly efficientexcitation media. As a result, extremely intenseradiation from common elements such as Ca andMg is delivered to the spectrometer-measurementsystem, The result is the observation of stray lightinterference effects for some other elements. Themagnitudes of these effects depend on: the designcharacteristics of the spectrometer used, the con-centrations of the elements from which the straylight originates, the concentrations of the ele-ments subject to the interferences, and the typesof approaches used to alleviate the effects. An ex-pedient means of correcting for these effects maybe based on determination of the concentrationsof the causative elements. To illustrate, let Cm

represent the measured concentration of a partic-ular element (analyte) subject to interference dueto a concomitant constituent present at concen-tration C(. The true (corrected) concentration (C1)of the analyte may be determined from:

where a is a correction coefficient determined bysimple experimental procedures.

The other type of interference effect involvesinterelemental processes in which the presence ofone constituent changes the extent of excitationof another in the plasma, While such interelementeffects are less common, the above correctionprocedure can also be used for compensation.Commercial ICP- and DCP-MAES systems are rou-tinely equipped with minicomputer or microproc-essor systems for control, data acquisition, anddata correction purposes. The use of the types ofcorrective procedures described above is, thus,easily automated.

In effect, the general capabilities of plasma-MAES are such that it should be considered a pri-mary facility.

X-Ray Emission Techniques

The bombardment of samples with X-rays toproduce X-ray fluorescence (XRF) has remainedin wide usage. In fact, the development of the lithi-um drifted germanium, Ge(Li), or silicon, Si(Li), de-tectors has advanced the status of this approachby the reduction of spectral interference prob-lems and a general increase in the sensitivity

available. Such energy-dispersive detection sys-tems have rather rapidly replaced the convention-al wavelength-dispersion units. The production ofX-ray fluorescence (emission) by X-ray bombard-ment and measurement with an energy dispersivedetector offers reasonable analysis capabilitiesfor several elements of interest (see table H-4).

In 1970, the potential of heavy, charged parti-cles for X-ray excitation was recognized, and im-proved capabilities have been demonstrated. Thecapability improvements of accelerator (particle)beams are due to: 1) the high particle fluxes avail-able, Z) the relatively low background radiationassociated with the excitation process: 3) the factthat the excitation cross-sections of many ele-ments for particles are higher than for photons orelectrons, and 4) a single charged particle can in-duce emission of several X-ray photons as it pene-


trates the sample. As a result, proton induced X-ray emission analysis (PIXE) has rapidly emergedas a sensitive analysis approach which may besubject to fewer interference problems than X-ray induced emission. Thus, the X-ray source isreplaced by a van de Graaff accelerator to pro-duce proton beams in the 2.5 to 3.0 Mev energyrange.

A potentially significant capability associatedwith PIXE analysis is that of direct analysis of bio-logical tissue sections or blood. Recent publica-tions5 6 have shown that Cl, K, Fe, Cu, Zn, Pb, Se,Br, Sr, and S could be directly measured in a 30µm thick section of human kidney; that the sameelements plus Mn, Ni, Hg, Rb, and Zr could be de-termined in a thick section of carp muscle; andthat several elements can be directly measured intissue sections of liver, kidney, lung, and bone.The analysis of liquid- or wet-digested materialsmay be based on evaporation of the liquid phaseto leave a residue deposit on an appropriate ana-

‘lJ. L. Campbell et al. “Acivances in X-ray Analysis, ” vol. 17(C. L. Grant et al,, eris.), Plenum Press, New York, 1974, pp.457-466.

‘P. S. ong et al., [bid., vol. 16 (L. S. Birks et al., ecis. ), 1973,pp. 124-133.

PREPARATION OF

lytical substrate, Ions in solution may also be pre-concentrated for analysis using filters impreg-nated with ion exchange resins. ’ Determinationcapabilities for XRF and PIXE are listed in tableH-4,

The general capabilities of these techniquesare such that they can be used in a laboratory ofthe type considered. It should be emphasized,however, that quantitation of the measurementsis subject to difficulties particularly for directmeasurements . Proponents of the techniquesargue that these problems can be readily over-come; others (cynics??) argue that this will re-quire extensive development efforts. The truthappears to be intermediate between these ex-tremes. The ultimate decision to include the X-raycapabilities in the laboratory facility should prob-ably be based on the essentiality (desirability) ofbeing able to analyze solid samples (e.g., tissuethin sections) directly for several elements of po-tential interest. This capability may be consid-ered by some to be advantageous simply from thesemiquantitative screening standpoint.

S. L. Law and W. J, Campbell, “Advances in X-ray Anal-ysis, ” vol. 17, Plenum Press, New York, 1974, pp. 457-466.

(BIOLOGICAL TISSUE) FOR ANALYSIS

Many techniques require that solid samples beconverted to solution form for analysis. The litera-ture on methodology for sample decomposition isimmense, The procedures cited find both wet-(acid digestion) and dry-(ashing) oxidation meth-ods extensively used. Dry-ashing methods are usu-ally implicated when problems with recovery orlosses of analytical constituents are reported. Incomparing wet- and dry-ashing methods, the pau-city of data specific to real-life samples makes itinappropriate to s tate categorical ly that onemethod is superior to another. Some generaliza-tions can, however, be made on the basis of proce-dural differences. Wet oxidation has the advan-tage of requiring a minimum of apparatus and isless prone to volatilization and retention lossesthan dry-ashing. Wet-ashing may suffer in thatrelatively large amounts of reagents having signif-icant levels of contamination may be required andcontact with glassware may account for a higherrisk of contamination than dry-ashing. Dry-ashingrequires few, if any, reagents and handling oflarger samples presents less problems. The risk of

volatilization, convection, and retention losses ishigher, however, unless ashing conditions arecarefully controlled, These risks have led to thefairly widespread adoption of low temperatureashing (LTA) based on reaction of the sample ofthe oxygen free radical generated via a radio fre-quency field. While it is true that materials can beconverted to oxides under these conditions, it isalso apparent that several elements may still belost by volatilization at typical operational tem-peratures of 100° to 150° C. These are largely ele-ments with a tendency to form volatile chlorides,oxychlorides, or hydrides, Thus, the extent oftheir loss may be particularly influenced by thehalogen (chloride) content of each sample, Theuse of programed-temperature ashing furnaceshas been shown to be effective in preventing orreducing volatilization losses. Raising the temper-ature at a rate permitting slow charring and oxi-dation of the sample is to be highly recommended.

Nitric acid is a universally used wet-oxidant,The azeotrope boils at 120° C, a factor which as-sists in its removal after oxidation but also limits


its effectiveness in completing the oxidation proc-ess. The most effective medium for wet oxidationis a mixture of nitric and perchloric acid; the aze-otrope boils at 180° C to force oxidation. Extra-ordinary care to avoid explosions and fume hoodsis required. The precaution of keeping sufficientnitric acid present until all easily oxidized materi-al is gone (cessation of brown fumes) is particu-larly important in wet digestion of tissues havinglipid (fat) contents. Because the acids used arenever absolutely free of metal contamination, theobjective of wet oxidation should always be tocomplete the oxidation with the smallest possibleamount of acid. This minimizes the blank problem.The use of reflux digestion apparatus equippedwith condensers is focused on this objective aswell as that of minimizing losses. Recent systemswhich rely on microwave ovens for heating theacid-sample mixture also show considerablepromise. 8 9 The microwave system heats the solu-

“A. Abu-Samr~, ]. S. Alorris, ~n[j S. R. Koirtyohann, Anal.Chem., 47, 1475 (1975].

‘IU. S. Patent No. 4,080.168 (Mar. 21, 1978).

tion rapidly and prevents bumping and frothing.Recovery studies run on a range of elements sug-gest that loss problems are minimal. Bovine liveris notorious for being difficult to wet digest. Ther e p o r t10 that 2 gm (wet weight) or 1 gm (dryweight) of beef liver can be digested in 10 ml of ni-tric/perchloric acid in 2 to 3 minutes indicates anattractive capability that will surely guaranteeextensive use of the microwave system.

For any of the analysis techniques which re-quire sample dissolution, the approximate upperlimit of “dissolved salts” that can be tolerated is 2percent (wt/vol). Thus, for a wet biological tissuewhich yields 10 percent ash, this upper limitwould be equivalent to dissolving 20 gm of thetissue (wet wt) per 100 ml of analytical solution.Working with half that weight per 100 ml wouldhave practical advantages.

“’Report to the U.S. FDA (Contract No. 223-75-2268] Der. 2,1976.

ASSURANCE OF ANALYTICAL ACCURACY

Quality assurance in an analysis laboratory re-fers specifically to the question: Are the analyt-ical results valid (accurate or reliable]? Obtainingthe answer is complicated by the fact that allanalyses are subject to random (indeterminate)errors and may also be subject to systematic (de-terminate) errors. Errors are cumulative; thosewhich characterize the accuracy of any analysisresult may include the composite contributions ofthe random and systematic errors inherent in theanalysis method(s) used and those characteristicof each analyst, Laboratory, or set of equipmentinvolved in the analysis. To maintain quality as-surance, an operating laboratory must base itsprogram on some combination of the possible stra-tegic approaches. These include:

1. Recycling of submitted samples to obtaincross check analyses,

2. Recycling of certified or secondary standardreference materials to estimate analyticalaccuracy,

3. Spike-recovery studies to estimate accuracy,4. Participation in collaborative test programs,

and5. Comparison of results obtained by independ-

ent analysis methods.Each of these provides specific types of relevant

information but all should be included in the pro-gram. Moreover, any of the above tests should becarried out incognito. Otherwise, the analystsmay be tempted to devote inordinate attention tothe check analyses.

All of these methods add to the expense of laboperation. It is necessary, however, for the labpersonnel to be acutely aware of the ways inwhich they or their analytical methods can failand (ideally) know when they have failed. Other-wise, their products—the analysis results—willbe subject to challenge. Experience with the legalprocess suggests that the agreement between re-sults obtained by two or more independent analyt-ical methods is a primary indicator of success (ac-curate analyses). Thus, the planned inclusion ofanalysis redundancy (5 above in particular) in thelaboratory operation is to be recommended; a typ-ical quality assurance program would involvecheck analyses amounting to 5 to 10 percent ofthe total load.

‘ I An expanded discussion of quality assurance programs andthe necessity thereof is given in the paper by R, K. Skogerboeand S. R. Koirtyohann, 4’Accuracy Assurance in the Analysis ofEnvironmental Samples.’” NBS Special Bulletin 422 ( 1976).


LABORATORY INSTRUMENTATION ANDFACILITIES REQUIREMENTS

Consideration of the above discussion indicatesthat the majority of the analytical requirementscan be satisfied by plasma emission spectrometryor by atomic absorption spectrometry with heavyreliance on furnace atomization systems. The in-clusion of more than one measurement techniquein the laboratory can also be justified on variousbases. The advantages, limitations, and tradeoffsinvolved will be considered below. Regardless ofwhich instrumentation facilities are chosen asprimary; the space requirements, ancillary facil-ities requirements, and personnel requirementsare comparable. Thus, discussion of these prior toconsideration of the analytical instrumentationfacilities is appropriate,

Personnel, Space, and AncillaryFacility Requirements

The general design of the laboratory should in-clude four physically separated types of space:1) office space for personnel, 2) a sample receiv-ing and storage room, 3) a sample preparationlaboratory, and 4) the analysis laboratory. Work-ing desk space should be included in the last threetypes of space in addition to the facilities dis-cussed below. All lab facilities will require tem-perature control to plus or minus 5° F and humidi-ty control (less than 50 percent).

It is recommended that computer capabilitiesshould be a primary ancillary facility included inthe laboratory operation. This is based on severalrationales including:

1.

2.

3.

several types of instrumentation likely to bepresent are most effectively used under acomputer control-data acquisition mode ofoperation:personnel requirements, time commitments,and human errors can be reduced in a com-puter-oriented operation; anda computer system may be essential as adata management, quality assurance evalua-tion, warning assessment, and report prepa-ration tool in a laboratory operation of thesize likely for the present program.

Although the computer configuration selected willbe dependent on the laboratory purpose(s), theanalytical facilities installed, the size of the oper-ation, and other subsidiary factors, it is likely acomputer acquired with one of the instrumentsdescribed below could be adapted for use in an in-

tralaboratory, interactive (time-shared) mode andtied to a larger computer system (external to thelaboratory) which would perform those data man-agement functions that need not be carried out ona real-time (fast response) basis. Systems whichuse internal, instrumentation-coupled, computersfor control and data acquisition purposes andtransmit the data to central management comput-ers are presently in operation at the USGS WaterResources Laboratory in Denver and the EPA Lab-oratories in Cincinnati. Although these operationsmay not be the best model examples, the concep-tual approach which they embody is recom-mended for the present operation. Further com-ments relating to this are inserted in appropriatesections which follow.

Sample Receiving and Storage Room—The re-ceiving operation will necessarily include facili-ties for logging in samples and the associated ana-lytical work requests. A computer terminalshould provide the most effective capability. Thestorage facilities should include: shelving or cabi-nets for those samples that can be stored underambient conditions, a cold room (4°C) for storageof water samples, and a freezer for storage of cer-tain biological samples. The size of the room andstorage facilities required will depend on the an-ticipated sample load and variations in the sam-ple submission rate. Since such data have notbeen supplied, size estimates and costs are not in-cluded in this report,

Sample Preparation Laboratory—The design ofthis laboratory will be sample type and sampleload dependent. If the sample types to be ana-lyzed include those which must be ground in thedry form such that atmospheric contaminationcan occur, the grinding facilities should be physi-cally isolated from the other sample preparationoperations, In addition, the sample preparationlab should basically be a clean room12 operationor it should be equipped with laminar flow, filterhoods in which certain sample preparation opera-tions can be carried out. Both wet- and dry-oxida-

(;lean room or clean hood environments are classified onthe b[]sis of (x)ntroliing the concentrations of particles (dust) inair which rnn cont:lminate the samples by f;~llout. A class-100(’le;~n room, for t:xtlmple, must ht]ve no more than 100 particlesin the 0.5- to 5- micrometer diameter size range per cubic footof air. A cI:lss- 1,000 room permits 1,000 pa rticu]a tes per cubicfoot in the ahove r:tn,gc. ‘1’o meet these specifications, air filtra-tion is re(~uire(i. S(;e th[! 1963 Fe(teral Stand[~rd No. 209:1,


tion facilities should be included in the laboratoryplan to permit handling of various sample types inthe most expedi t ious manner . A l i s t of samplepreparation facilities likely to be required is givenwith their approximate costs in table H-5. Accessto a computer terminal to permit convenient entryof sample preparation data is also recommended.

Analysis Laboratory —The instruments se-lected will influence the facilities and size re-quirements of this laboratory. It is assumed thatthe prepared samples will be delivered to this labin closed containers and will be handled in waysthat will minimize the probability of contamina-tion via atmospheric contact or fallout. Thus, al-though rigorous clean room operation has notbeen recommended for this lab, it will be neces-sary that all analytical operations which can re-sult in simple vaporization be carried out underventilation conditions where the discharge is ex-ternal to the lab.

Personnel Requirements and Performance—Stipulation of exact personnel requirements willagain be dependent on the sample load and thetypes of samples received. The sample loggingoperation can be handled by an individual with nochemical training but with secretarial or key-punch skills: this individual might also performsome of the simpler sample preparation opera-tions, e.g., weighing and grinding. The prepara-tion laboratory should be staffed by individualswith B.S. degree training in chemistry with 1 to 3years of wet chemical experience.13 One such indi-vidual can typically prepare 15 to 30 biologicaltissue samples per day by wet- or dry-oxidationtechniques depending on the complexi ty of thep r e p a r a t i v e s teps involved . T h e a n a l y s i s l a bshould also be staffed by chemists with at leastB.S. degree training plus 2 to 4 years of experi-ence.14 The experience should preferably be in the

general area of instrumental trace analysis deal-ing particularly with the types of instrumentationto be used. A working familiarity with basic elec-tronics and minicomputers would be desirable.The availability of major repair capabilitiesthrough instrument maintenance contractsshould be insured.

For the lab personnel, it may be assumed thatthe average analysis production hours per man-day will range between 5 and 6. The remainingtime will be utilized for preventive maintenanceof equipment, recordkeeping, cleanup, etc. Al-though there are nominally 260 working days perannum, this reduces to about 200 days when vaca-tions, holidays, sick leave, and refresher trainingtime are taken into account. The number of per-sonnel required and the sample throughput capa-bilities of the lab must be defined on this basis.

Table H-5.— List of Sample PreparationFacilities Required

Preparation facility Approximate cost, $a

$2.5008.000 automatic3,000 manual4.000

4.0008.000

6,000

8,000

4,5002,000

$5,000-10,000

ANALYTICAL INSTRUMENTATION

The previous discussion has presented the gen- H-6. Although the data presented in that summaryeral analytical features, capabilities, and limita- have been based on personal experience and dis-tions of the various analytical techniques consid- cussions with other practicing analysts, they mustered applicable to the present problem. A summa- be considered as estimates only. This applies inry of the cost, space, and throughput features of particular to the cost per analysis because it willthese types of instrumentation is given in table be strongly influenced by the applicable salary


Table H-6.—Summary of Cost-Productivity Estimates for Various Analytical Techniques

Approximate - Space - —————Estimated Approximate Data Estimated

acquisition requirements no. analyses cost per system downInstrument facility cost, $a sq. ft.b per man-dayc analysis, $d requirement e time, % f

ICP-MAES. ,-:- . . . . . . . . . . . . . . ; .-:——————

100,000 200-400 1,500-2,000 0.1-0.3 Yes” 3-5%DCP-MAES . . . . . . . . . 60,000 200-400 1,200-1,500 0.1-0.3 Yes* 3-5%Flame & furnace AAS

Single e lement modeg. . 25,000 200-400 100-200 (flame) 0.7- 1.0 Desirable 3-5%50-100 furnace 1.0- 1.5 Desirable 3-5%

Multielement modeh . . . . . . . 40,000 200-400 400-800 (flame) 0.3-0.5 Yes’ 3-5%200-400 (furnace) 0.5-0.8

Electrochemical i . ... . . 25,000 200-400 50-100 1.0- 1.5 Yes* 3-50/0N. A.A.J. ... . 40-50,000 k 200-400 200-700 ? Yes* ?X - r a y f l u o r e s c e n c e . . . 60-100,000 200-400 200-700 7 Yes* 3-50/0PIXEJ . . . . . . . . . . . . . . ., 300,0001 300-500 400-800 7 Yes* ?

—aFOr ~o.~~e-ment ~“~1~ SIs Capab;l llles lncludl~g ;strument mstallatlonbin addltlon 10 sample preparation and office space but allowlng for adjacent working sPace.Csamples Previously prepared for analysls, data management system available, 5 hours on the Instrument per 8.hour day, one anal Ysls defined as determination of oneelement per sample, approximately 10 percent of total analytical load I nvolvlng quail ty assurance assumed, all analyses to Involve simultaneous multlcomponentdeterminations of 10 to 15 elements on each sample

dcost per element per Sample, :jamples Previously prepared for analysis, approx 5-year instrument depredation assumed, see footnote C aboveel nstrument control, data acquls.ltlon and data management system considered essential for all facllltles, . Ind!cates Incluslon of a minicomputer or microprocessor ofat least 8K memory In acquwtlon costs

fAn Upper Ilmlt estimate Intenderj to Include preventive mmlenance9AnalYSeS based on rjeternllnatlon of one element at a time Cost estimate Includes hollow cathode lamps for 30 elements and for autOmatlC sample lnjeCtlOn but nOtfor a computerlmlcroprocessor system

hAnalySes based on determination of 5 t. 7 elements at a tfme cost estimate Includes hollow cathode lamps for 30 elements, an auton’ratlc sample lnJE!CtOr, and a dataacquislt!on and control system

‘Llmlted principally to analyses for As, Cd Cu, Fe, Pb, Se, Tl, and Zn, other metal analyses possible but not widely practicedIsame type of readout facllltles could be used for botf’r tecflnlques Radlatlon protection required particularly for NAA and PIXE systems‘Does not Include reactor costsIReduce by -2513000 If surplus van de Graaff accelerator IS available

structure, to actual types and numbers of anal-yses to be performed, the efficiencies of the em-ployees involved, and the final laboratory designconfiguration, The instrument acquisition esti-mates include initial installation and assume thatthe experience of the operating analyst will mini-mize startup time requirements. The estimatednumber of analyses per 8-hour day assume sam-ples have been previously prepared for the anal-ysis; that the instrument operation will rely oncomputer-controlled data acquisition and man-agement; that the salary of the operation wouldapproximate $20,000 to $24,000 per annum; andthat the instrument would depreciate completelyover a 5-year period. These estimates do not allowfor laboratory refitting or remodeling costs or foroverhead costs above the direct instrument oper-ational costs. Such factors have been excludedbecause the costs involved are highly contingenton the extent of lab refitting required, the locationof the lab, and the wide variations in overheadcharges. It should also be noted that the overheadcosts will be essentially the same regardless ofthe lab instrumentation facilities selected. Thefootnotes to table H-6 provide further qualifyinginformation.

The decision to exclude estimates of the costsof preparing the samples for the analyses has

been based on several factors including: 1) lack ofinformation regarding sample types and numbers,2) the influences of the preparation proceduresactually selected on costs, and 3) the fact that thesample prep costs will be essentially the same foreach of the analytical techniques considered.

Comparisons of the estimates presented indi-cate that the plasma emission techniques offersuperior economic advantages which are comple-mentary to the sensitivity advantages previouslydiscussed. Although more analyses per day canbe carried out with the plasma techniques, thepresent estimates have been based on the expec-tation that it will be necessary to analyze eachsample via two different sample introductionmethods to achieve the required sensitivity. Thus,several elements may be determined by directnebulization of the aqueous sample solutions. Asecond set of elements, e.g., As, Bi, Hg, Sb, Se, andSn, may have to be determined using the hydridegeneration method to achieve the required sensi-tivity (low-threshold effect levels). A similar situa-tion is anticipated for atomic absorption spectro-photometry; direct flame measurements will beadequate for some elements while furnace meas-urements will be required for others. Analysis inthe single-element mode for AAS is clearly morecostly. The inclusion of the simultaneous multiele-


ment AAS estimates is based on present experi-ence at Colorado State University. Conversion ofsingle-element AAS units to the multielement cap-ability can be anticipated in the near future. Thelong-term cost advantages are obvious.

The restricted capabilities of the electrochemi-cal system tend to remove it from competition ex-cept as a supplementary or specialized capability.Primary arguments for the use of NAA or the X-ray techniques are based on the ability to directlyanalyze tissue (solid) samples without dissolution.Unless this capability is important, the cost differ-entials evident in table H-6 argue against theiruse.

In view of these data, the use of the ICP-MAESsystem as the primary instrument appears quite

rational, The inclusion of an AAS system and anelectrochemical (EC) system as support (e.g.,cross-check) techniques should be recommended.The cost estimates given for the AAS and EU unitsin table H-6 were based on the expectation thateach might be the primary lab facility. When theirroles are reduced to a secondary (support) level,less sophisticated (high versatility and produc-tion) units may be acceptable. The cost of an ade-quate AAS facility could thus be reduced by-$8,000 to $10,000 and that for the electrochemi-cal unit by -$5,000 to $10,000. It should also benoted that these lower cost units could be inter-faced to the ICP-MAES computer system prefer-ably using inexpensive ($300 to $500) microproc-essors as buffers.

ESTIMATED ANNUAL SAMPLE

The above estimates of sample throughput cap-abilities may be used to approximate annual ana-lytical productivities. These are based on a totalof 260 working days per annum minus 60 days forvacations, holidays, downtime, quality assurancetime, etc., leaving 200 effective 8-hour days. Theprojection below is based on the assumption thatthe primary load will involve the analysis of bio-logical (organic) tissues; such analyses probablyrepresent the most rigorous time/cost require-ments,

Using the estimate of 15 to 30 samples per dayas the load that can be handled by a sample prep-aration technician, the annual preparative capa-bility ranges from 3,000 to 6,000 samples. Analy-sis of these samples for 30 elements (as an upperlimit example) using the plasma emission capabili-ty involves performing 90,000 to 180,000 anal-yses. Applying the lowest analytical throughput(1,500 analyses/day; table H-6) suggests that 60 to120 days would be required for these analysesonce the samples were prepared. In essence. thisemphasizes a fairly universal observation, i.e.,preparation of samples for analysis is often thefactor which limits laboratory productivity. Sev-eral inferences may be drawn from the abovesample preparation and analysis estimates:

1. The analytical capability proposed is suffi-

2

cient to keep 2 to 4 sample preparation per-sonnel busy if biological tissues comprise theprincipal workload of the lab.The sample preparation methods used needto be upgraded in terms of throughput perunit time.

3.

THROUGHPUT LOAD

The analytical facility could be used for addi-tional types of analyses which do not requiresophisticated (time consuming) sample prep-aration operations, e.g., water samples.

Comments on the last two possibilities are ap-propriate.

Primary factors which affect the sample prepa-ration time/cost requirements include: the ele-ments to be determined and the types of samples.The determination of toxic elements in naturaland effluent waters is perhaps the simplest case.The water samples must be filtered and appropri-ate preservatives added; these operations mustbe carried out as soon after sample collection aspossible and preferably in the field so essentiallyno lab prep operations are required, Such anal-yses could occupy the additional time available onthe instrumentation. The preparation of animaltissues (particularly liver) is at the other extreme.These contain varying amounts of fatty materials(lipids) which are difficult to decompose by wetoxidation techniques thereby requiring that rigor-ous conditions be used. Although dry oxidation es-sentially circumvents this problem, the fact thatseveral elements may be lost by volatilizationforces the use of slower oxidative methods, e.g.,temperature-programed furnaces, low-tempera-ture ashing units, or sealed [high-pressure) bombsystems. The estimates given above have beenpredicated on the use of wet oxidation (reflux)techniques in common use; these appear to be lesssusceptible to problems than the dry techniques.It has also been assumed that complete destruc-


tion of lipids present is required. An examinationof the literature, however, has not produced con-vincing evidence that this is essential. The pointraised is simply: Can the most significant fraction(e.g., >90 percent) of the elements of interest be“extracted” from the tissue matrices via wet di-gestion without complete destruction of the fat,lipids, or cellulose present? If so, the sample prep-aration times can be prominently reduced. Final-ly, it should be noted that the wet ashing of bio-logical materials in microwave ovens shows prom-

ise in alleviating the digestion time problem.15

Spike-recovery studies carried out when foodproducts were digested via this method are veryencouraging. The ability to reduce sample diges-tion times by a factor of two or more appears to bea reasonable estimate. Factors of this naturemust be considered in the final planning stage forthe present program.

Ad[;] At)u-%mra, J. S. hlorris, and S. R. Koirtyohann, An:il.Chem., 47. 1475 ( 1975). Set! a]so reports on L?SPH contrnct No.223-75-2268 and U.S. P{~tent 4,080,1 t38.

ANALYTICAL ACCURACY COMPARISONS

As emphasized above, the validity (accuracy) ofthe measurements strongly influences the integri-ty of any decision based on the results, Any meas-urement is subject to errors which may be influ-enced by several factors including the measure-ment methods used and the analyst responsible.This is the primary reason for stressing the impor-tance of a quality assurance program. The gen-eral degree of accuracy that can be achieved withanalytical techniques discussed above can be an-ticipated to be essentially the same for all tech-niques. This first approximation expectation isbased on the following. First, each of the tech-niques discussed essentially requires the samesample preparation procedures. Since these arelikely to prove to be significant (primary?) sources

of analytical errors, the errors which may accrueduring the preparative steps will be essentiallythe same for all techniques. Second, the analystsinvolved can be responsible for the cause or pre-vention of errors depending on the expertise andcaution they exercise. The analyst that tends touse poor technique or judgment when applyinganalysis method A will, in all probability, do thesame for method B; the errors for which he is re-sponsible will be comparable in both cases. Final-ly, the measurement accuracies of each techniquediscussed above are generally similar. As a re-sult, there are no clear-cut, easily defensible rea-sons for suggesting preference for one approachover another on the basis of improved accuracy.

RESEARCH NEEDS ASSOCIATED WITH THE PRESENT PROGRAM

In the section discussing the criteria involved inthe selection of an analytical method, the generalimportance of being able to identify and measurethe chemical forms or oxidation states of severalelements was emphasized. This emphasis wasbased on the fact that all segments of the ecologi-cal system are affected to varying extents by theelements (e.g., metals) present. Indeed, the state-ment that “the life processes of every living cellare conditioned by the types and amounts of met-als present” would be accepted by a majority ofscientists. Some metals are essential to the healthof living systems and, yet, they may also be insidi-ous pollutants because of their generally nonbio-degradable nature. Only a few metals are com-pletely nontoxic at any concentration level; mostcause deleterious effects at some exposure level.The ultimate definition of: 1) what constitutes adeleterious effect?; 2) what actually causes it?;

and 3) what are the operative threshold-effectconcentration levels? cannot be considered trivialproblems. It is unfortunate that simply measuringthe total amount of a particular element presentin an ecological system may, in fact, be only agross indicator of potential or actual deleteriouseffects, What is often needed is the identificationof the active or functioning forms of the elementsin question. Although numerous examples whichsupport this statement can be cited, the fact re-mains that chemical form is often extremely im-portant.

For this reason, the development and refine-ment of analytical methods and techniques for theidentification of chemical form (chemical specia-tion), at the trace to ultratrace concentrationlevels so often of interest, has received considera-ble attention in the past decade, Although this is


clearly an important area of research, progresshas been slow for several reasons. These include:

1.

2.

3.

4.

5.

The analytical techniques most suitable forthe determination of chemical form are alsogenerally those which lack sufficient sensi-tivity for use at the low concentrations char-acteristic of many elements in biological sys-tems.Biological systems are inherently very com-plex mixtures of a wide variety of chemicalconstituents. It is a truism that the complexi-ty of the identification problem is dramat-ically enhanced by the compositional com-plexity of the target system.While it can be argued that chemical iden-tification techniques are necessary for thedelineation of what chemical forms are bio-logically or toxicologically important, it ap-pears equally valid to argue that such knowl-edge on an a priori basis is highly beneficialas a n aid i n focusing the development t work.The analytical chemical community is oftenonly partially aware of what types of chemi-cal speciation measurements are consideredmost important by the medical, toxicology,and ecology communities. Similarly, the lat-ter are often only peripherally aware of themost promising emergent analytical technol-ogy.The scientific communities involved in thischemical speciation question have oftenbeen forced by the pressing circumstancesso frequently associated with deleterious ef-fects to expend their efforts on the use of lessthan satisfactory approaches leaving lesstime for the required types of development.

Although other contributing factors could becited, the case in point is simply that the chemicalspeciation capabilities so badly needed are avail-able in only limited instances. The development ofadequate speciation technology should be a cen-tral thrust of research efforts involving collabora-tive efforts between the toxicology, biological ef-fects, and analytical chemical communities. Gov-ernmental and private funding agencies shouldstrongly encourage these efforts, Because propo-gation of chemical speciation developments willultimately be essential to the success of the over-all monitoring program, the laboratory facilitiesavailable at the outset should take advantage ofpresent capabilities. Moreover, the prime stanceof the program should be that of providing feed-back to the relevant scientific communities as onemeans of focusing speciation research efforts on

/

those elements which may appear to be more im-portant from the effects standpoint.

Technology presently exists for the differentia-tion between the organic (alkylated) and inorgan-ic forms of the toxic elements: arsenic, lead, mer-cury, and selenium. These rely on the fact that theinorganic forms can be easily converted to gase-ous forms (mercury vapor and the hydrides) byreaction with borohydride while the organicforms must first be photodecomposed by exposureto ultraviolet (UV) radiation, Thus, analysis of thegaseous reaction products before and after UVirradiation provides measurements of the rele-vant organic and inorganic concentrations. Thischemical differentiation approach has been inter-faced with atomic absorption and emission spec-trometry as the measurement tools; inclusion ofthis capability in the laboratory is recommended.

Similarly, the ability to differentiate betweenthe possible oxidation states for some metals isimportant, e.g., As(III) versus AS(V) or Cr(III) ver-sus Cr(VI), Such differentiation for As, Cr, andsome other metals can be accomplished by use ofselective oxidation-reduction (redox) reactions.The redox reactions may be controlled by properselection of chemical reagents or by judicious useof electrochemical (polarographic or voltammet-ric) principles. Electrochemical measurementscan also be used to differentiate between electro-active and nonelectroactive (chemically bound)forms of some elements and between “labile” and‘‘nonlabile”’ bound forms of some elements. In thiscontext, labile and nonlabile refer to the thermo-dynamic and/or kinetic stabilities of the metal-ligand systems in question. Although this type ofdifferentiation may be only semiquantitative orsemiempirical in nature, such measurementshave been shown in some instances to be perti-nent in experimentally defining bioavailability,transfer mechanisms, transfer rates. etc. It is cer-tain that research must still be done to further de-lineate the diagnostic potentials of such types ofdifferential measurements, Present knowledgeseems sufficient, however, to support the inclu-sion of polarographic and voltammetric instru-mentation in the laboratory facilities for “labil-ity’” and/or oxidation state measurements.

A limited number of other possibilities exist forchemical speciation measurements. These arelargely very specialized in terms of applicabilityto certain types of samples and will not be dis-cussed herein. It is certain that the need for chem-ical speciation exists and that its importance willbecome even more apparent as more definitive


monitoring programs evolve. It is also certain that tion problems will involve the application of someno single analytical technique for speciation or combination of chemical and physical principles,any other purpose will be a universal panacea. The research programs most cognizant of this areThe ultimate solutions to the majority of specia- most likely to be successful.

GLOSSARY OF ABBREVIATIONS

AAS—atomic absorption spectrophotometryAES—atomic emission spectrometryASV—anodic stripping voltammetrycc—cubic centimeterscm--square centimetersDCP—direct current plasmaDPP—differential pulse polarographyFC—Faradaic or analytical currentGe(Li)—designates a germanium crystal which has

lithium drifted into it to give it uniquely definedproperties

g i n - g r a mICP—inductively coupled (radiofrequency) plasmaMAES—multielement atomic emission spectrometryMev--million electron voltsml—milliliterNAA—neutron activation analysis

Ng—nanograms—one billionth of a gramPES—plasma emission spectrometryPIXE—proton induced X-ray emission (spectrometry)ppb—parts per billion ( 1 x 10-’) gms/gm on a weight

basis or 1 x 10-j µg/ml on a weight per unit volumebasis)

ppm—parts per million ( 1 x 10-6; gm/gm or 1 µg/ml as forppb)

Si(Li)--a silicon crystal which has lithium drifted into itto provide unique physical properties

SMS—solids mass spectrometryµg—one millionth of a gramµl —one millionth of a literµm—one millionth of a meterXES—X-ray emission spectrometryXRF—X-ray fluorescence (emission spectrometry)

Appendix I

Analysis of Foods for Radioactivity*by Naomi H. Harley

The analysis of foods for radioactivity shouldnot be considered as a primary defense againsthuman intake. The first indication should alwayscome from information on releases or from meas-urements of radioactivity in the airborne or wa-terborne releases. Once the existence of contami-nation has been established then the foods can beanalyzed to evaluate potential hazard to man.

In contrast to most other pollutants the effectsof radiation are considered to have a linear re-sponse regardless of the level, thus, there is nothreshold and no absolutely safe limit. Instead itis necessary to set some lower level below whichthe radioactivity in foods is no longer of interestas compared with other sources of radiation orother hazards of life. The analytical significanceof this is that the lower limits of detection for ra-dioactive substances have been brought down tovery low levels and the simple yes or no testing foracceptability that satisfies regulations for manyother pollutants in foods cannot be used.

The radionuclides of interest in the case of con-taminating events are almost all present now infoods in measurable quantities. Short-lived nu-clides are the exception and the transuranic ele-ments are only present at levels that require con-

The most useful classifications of a radionu-clide are those based on the characteristics of theradiation emitted and by the identity of the chemi-cal element. The former is both a guide to the na-ture of the hazard involved and to the measure-ment required. The chemical species (e.g., ele-ment, oxidation state) regulates the metabolicpathways in the biosphere as well as the nature ofany chemical separations required in the meas-urement procedure.

* f{k( [:rpi from [III () I A 11 ~~rhllw I)(lp[’r l:rl II tlfxj “‘Arl;ili \is [II

1 II{)(Is for R:i[iro;i(ll\it} .’” A (l)mpl{’tc rIIpV of the p(iper (:ln 1)(){ )t) t ii I rl~x 1 [rt )m t ht; Nii t I I JIII i 1 1 I;( h rl] ~ i~ I I n [( )rm; i t I( )rl S(I rv I( t’. ( SC(?

i i[)~). J . )

siderable effort in analysis. Since most of the ra-dionuclides are already present, measurementsmade for background information should producea numerical answer, not merely an indication thatthe amount is less than some pre-set value. Theaccumulation of background data provides a valu-able baseline for evaluating excursions followinga contamination event, The natural activity dataare equally valuable since the amount of informa-tion on food concentrations is presently insuffi-cient for valid comparisons with manmade radio-activity.

This report will describe the requirements andconsiderations for establishing a system to pro-duce acceptably accurate measurements of radio-nuclides in foodstuffs. The basic concepts will bedescribed, but to maintain the necessary brevity,the detailed procedures that might be used will begiven only by reference.

During the preparation of this report, FDA hasproposed certain recommendations for State andlocal agencies on Accidental Radioactive Contam-ination of Human Food and Animal Feeds. Thismaterial appeared in the Federal Register for De-cember 15, 1978, page 58790, and is interestingbackground material for this topic.

RADIONUCLIDES

The emitted radiations are generally groupedas alpha (cY), beta (ß), and gamma (~). Alpha radia-tion is characteristic of the natural and artificialradionuclides of high atomic weight and consistsof energetic particles with very low penetratingpower. Its hazard is significant only within thebody, where alpha-emitting nuclides can irradiatespecific sensitive tissues. Beta radiation appearsin both heavy and light natural and manmade ra-dionuclides, and consists of electrons possessingkinetic energy and having modest penetratingpower. Gamma radiation is pure electromagneticradiation and is extremely penetrating. Thus, itcan be a hazard externally as well as when it is

215


present in the body. For the present purpose, weare only concerned that the penetrating nature ofgamma radiation allows its direct measurementin foodstuffs, while alpha and beta emitters gen-erally must be separated from the bulk constitu-ents of the sample before measurement is possi-ble.

There are other processes in radioactive disin-tegration that produce emissions. Alpha emissionis usually accompanied by low-energy gammarays that may be used for measurement. X-rayscan be produced by electron capture and somegamma emitters decay by internal conversion, aprocess where a fraction of the gamma rays areconverted to monoenergetic electrons. Theseprocesses do not really modify our measurementconcepts but the decay modes of the significantradionuclides must be known for their accuratemeasurement.

The physical half-life of the radionuclide tendsto control its persistence in the environment. Forexample iodine-l 31, with a half-life of about 8days, is important for only a few weeks, whilecesium-137, with a half-life of about 30 years, maybe a problem for centuries.

A third classification that has some value is thesource that produced the radionuclides. Knowl-edge of the source of radioactive contaminationgives a good indication of the nuclides that are tobe expected in the sample. This is of considerableassistance in planning the analysis since request-ing a complete analysis for all radionuclides oreven for all types of radioactivity in a single sam-ple would lead to a lengthy and expensive opera-

tion. Many radionuclides are not potent healthhazards, particularly those that are not metabo-lized by the body. The general groups of nuclidesto be expected include the natural activities, spe-cifically radioactive potassium and members ofthe uranium and thorium series, and the artificialfission products, transuranic elements, and otheract ivat ion products that resul t f rom nuclearweapon explosions and nuclear reactor opera-tions.

Fission products are a very complex mixture atthe time of formation but the short-lived radionu-clides die out rapidly and the mixture becomessimpler within a few days, The transuranics (plu-tonium, americium, etc., formed by activation ofthe basic fissionable material) are of some inter-est because of their high toxicity when incorpo-rated into the body but present evidence indicatesthat their uptake through the gut is relativelysmall and that dietary intake is not a significantproblem. This should be true particularly if therelative hazard of other radioactive contaminantsprobably present in the same sample is taken intoaccount, The other activation products are fre-quently elements that make up steel or othermetal containers or structural elements. Radioac-tive manganese, chromium, cobalt, zinc, and ironare particularly common and result from inter-actions of the materials with neutrons released inthe nuclear reaction. It is worth pointing out thatcontamination of foodstuffs with single nuclides isextremely unlikely, and that more than one mem-ber of any group will probably be present in anysample.

DISTRIBUTION OF RADIONUCLIDES

The source of the radionuclides involved willgenerally control their distribution in the environ-ment and their consequent transfer to the foodchain. The sources considered here will includenatural radioactivity, releases from operation ofnuclear reactors and processing plants, and fall-out from nuclear weapons tests.

Natural activity may be of concern when it isenhanced by man’s intervention, say by mining tobring material to the surface and processing theore to yield either products or wastes that mayconcentrate the radionuclides. Good examplesare radium in uranium tailings, in phosphate rockwaste, or in slags from phosphorus production,Radium may enter the food chain by dissolving inground water and transferring through plantroots.

Nuclear reactors in normal operation releasechiefly the radioactive noble gases that are not ofinterest in considering foods. Reactors do containlarge inventories of fission products, transura-nics, and other activation products, however, andaccidental releases can contaminate vegetationby deposition or through the water pathway. Gas-eous releases would most likely involve the vola-tile elements such as iodine and tritium or thosewith volatile precursors, such as strontium-90and cesium-137, Aqueous releases would followfailure of the onsite ion exchange cleanup systemand any water-soluble elements could be in-volved.

Processing plants could also have either gas-eous or aqueous releases, but only fuel reproc-essing is likely to be a significant contributor. In

A p p e n d i x I — A n a l y s i s o f F o o d s f o r R a d i o a c t i v i t y ● 2 1 7

th is case , the f i ss ion products are aged beforeprocessing and iodine and the gaseous precursorradionuclides are not released. Tritium and car-bon-14 are the major airborne products, while thewaterborne radionuclides are the same as forreactors.

Atmospheric nuclear weapons tests distributetheir fission products, transuranics, and other ac-tivation products globally, with local deposition

being more or less, depending on the size of theweapon and the conditions of firing (high altitude,surface, underground).

In summary, the deposition of airborne materi-al on vegetation or on soil is the route by whichfoodstuffs become contaminated and the subse-quent behavior of the radionuclide is controlledby its chemical nature, including solubility andplant or animal metabolism.

CONTAMINATION OF FOODSTUFFS

Contaminat ion of foods can occur ei therthrough atmospheric deposition or by transferwith water. In the first case it is possible for theradioactive material to be in the form of insolubleparticulates rather than in a more available formwhere it will follow the chemistry of the elementsinvolved. A knowledge of the pathway is not abso-lutely necessary but it does assist in deciding onthe proper preliminary treatment of the sample ofevaluating exposure. For example, surface con-tamination may have a different significance thanthe same material present in a plant through rootuptake.

Since pathway information is not always avail-able i t is generally considered proper to measureradioactivity in samples that have been preparedas if for eating, so as to approximate the true ex-pected intake. This will usually result in strippingoff or washing off of a considerable fraction ofsurface contamination. Cooking is generally notpart of the preparation, as the mode of cookingand the use o f j u ices, cooking water, and the likecannot be predicted,

Milk is often recommended as an indicator foodfor studying radioactive contamination. It hasmany advantages:

1. It is available locally at most desired sam-pling locations.

2. It is marketed rapidly, so that short-lived ra-dionuclides. such as iodine-l31, can be eval-uated.

3. It is a major diet component in the UnitedStates, both directly and as an ingredient ofprepared foods.

4. There is a lot of background informationavailable on previous contaminating events.

It is worth noting that some of these “advan-tages’ are the factors that contribute t o the roleof milk as a source of human exposure.

Milk is a poor indicator of many contaminants.The natural activities, the transuranics. and theactivation products have relatively low concen-trations in milk. The first two are low because ofpoor biological transfer and the last because theirpathways are almost entirely through the aquaticor marine food chain. Thus, the best approach isto know what is in the environment through othermonitoring systems, and to design the food anal-ysis program to fit the circumstances.

Other monitoring data are also necessary infixing the geographical extent of a contaminatingevent. As a general rule, nuclear tests are globalin radionuclide distribution, with enhanced levelsnear the test site. Releases from other nuclear op-erations tend to be more local in their effects andthe food-monitoring plan can be modified to suit.

SAMPLING

The mechanics of obtaining representative foodsamples will not be considered here since the pro-cedures are common to all types of food analysis.There are certain points that must be consideredhowever. The first is whether the m e a s u r e m e n t sare being made to determine human intake or thesource of the contamination. In the latter case theearly approach used by FDA(1) for radioactivity isapprop r i a t e. There, each sample was identified

as to its place of origin. For evaluating intake i t ispossible to collect total diet samples for a partic-ular population group and to measure the radio-activity in this composite diet. This approach hasbeen described by FDA(2) and by EPA(3) and isnormally applied in institutions where mass feed-ing is carried out. A more elaborate procedure isto simulate a total diet by measuring a number ofcomponent food classes selected on the basis of


statistical information regarding consumption.This approach was originally used by the AtomicEnergy Commission(4) and was applied originallyto three major cities. This approach does allowidentification of specific food types that are con-taminated but requires much greater effort andcost in analysis.

The use of indicator foods as an intermediatetype of monitoring is widespread. As mentionedpreviously, most of the systems depend on thesampling of milk which is available in most partsof the country either with specific information asto place of origin or the general area of the milk-shed. Part of the reason for using milk is the easein sampling but it also has significance as a pri-mary food for the youngest and most susceptiblepopulation group and it also does tend to pick upseveral of the fission product nuclides of dietarysignificance such as radiostrontium, radiocesium,

and radioiodine. Most other foods give limitedgeographical or seasonal coverage and are lesssatisfactory.

The preservation of samples in the field duringtransport and in the laboratory awaiting analysisis only an esthetic matter in the case of radionu-elides. Decomposition processes do not changethe radioactivity, and sample contamination byradioactivity is unlikely. Thus, freezing, formalde-hyde addition, or any other method that will main-tain the sample is adequate. Any additive shouldbe checked to assure it does not contain signifi-cant amounts of the radionuclide sought.

For the purpose of this report we will assumethat an adequate quantity of a representativesample is available for analysis and that anotherportion is available for storage, either permanent-ly or until the analytical results are accepted assatisfactory.

SAMPLE PREPARATION

The measurement of radioactivity is a physicalprocess and it is most efficient when the radioac-tivity from a relatively large sample can be placedclose to the detector. This means that direct meas-urements of bulk samples are only useful at rela-tively high levels of contamination and that mostmeasurements are preceded by preparation andpossibly chemical separation to reduce the bulk ofthe material and to improve the efficiency of themeasurement.

As mentioned previously, sample preparationmay include removal of inedible portions of thefoods or those portions not generally eaten. Forexample, citrus rinds, apple cores, outer leaves ofleafy vegetables, and aboveground portions ofroot vegetables would normally be discarded. Thegeneral goal is to prepare the foods as if for cook-ing or consumption.

Foods generally have a high water content anda primary method of bulk reduction is drying atroom temperature, at elevated temperatures, orfreeze drying, Most of the radionuclides of inter-est are not volatile under these conditions andlosses must be considered only for elements suchas tritium and iodine. The dried material can bereduced further by ashing at elevated tempera-tures, by cold ashing with activated oxygen, or bywet ashing with oxidizing acids. The sample dry-ashing process is most likely to lead to loss ofvolatile elements but with care even cesium, polo-nium, and lead can be retained, The other proc-

esses should not lead to losses of elements of in-terest with the exception of iodine and tritium,mentioned above, and of carbon.

Another approach that is essentially one of re-ducing bulk is to extract either the original sam-ple or the dried or ashed material with acids orother solvents and thus remove the desired ele-ments from the bulk of the sample. This requiresconsiderable testing beforehand to be certainthat the process operates in the desired manner.

All of these procedures reduce the bulk of thesample and in the case of extraction may alsoseparate the desired constituents from some ofthe remaining inert material. They do not, how-ever, separate the radionuclides of interest com-pletely from the other radionuclides present inthe sample. Such separations will be covered inthe next section, but they may not be necessary ifthe measurement technique can provide bothqualitative and quantitative information. In mostcases this limits the possibilities to gamma spec-trometry on the prepared sample.

If the samples are going to be subjected tochemical analysis, the sample preparation mustinclude the dissolution of the dried or ashed mate-rial. This has already been done, of course, inpreparing the wet ashed or extracted solutions.The radionuclides of interest in ashed foodsshould be soluble in strong acids if ashing tem-peratures have not been excessive, and this treat-ment is usually accepted. If there is concern that

Appendix l—Analysis of Foods for Radioactivity “ 219

insoluble particulate may be present, it is neces-sary to use more drastic methods such as fusion tobring the sample into solution. This should not benecessary, however, if human hazard is the prob-lem under consideration.

At the time of preparation, if not before, thebasis of measurement must be established. De-pending on the use of the data, wet weight, dryweight, ash weight, volume, or even numericalcount (e. g., eggs) has to be determined. Frequent-ly, this may have to be done in the field, but for-tunately, relatively crude measures are adequate,

Since unforeseen questions often arise, it is rec-ommended that as many of these quantities bemeasured as is possible.

The requirements for measurement of radionu-clides in foods are such that sample preparationis best handled by the group responsible for therest of the analysis. The chief difference fromother types of analytical work is the initial sam-ple. The usual range is from 1 to 20 kg, and thereduction of this amount of material is a special-ized problem.

RADIOCHEMICAL SEPARATIONS

Radiochemicallate the desired

separations are required to iso-radionuclide both from the re-

maining bulk constituents and from other radionu-clides which would interfere in the measurement.In addition, it is necessary to convert the finalproduct to a form suitable for presentation to thecounter. This may involve elect redeposition, pre-cipitation, or other processes. There are a num-ber of manuals giving the details of specific radio-chemical procedures and these details will not berepeated here. There are a few generalities how-ever that may be of interest.

The actual mass of radionuclide that is meas-ured is almost always vanishingly small. Thismeans that many of the normal chemical reac-tions used in analytical chemistry will not takeplace; for instance precipitates will not form. Forthis reason it is common to add a few milligramsof carrier material which is preferably the inertform of the same element. Where this does not ex-ist it is frequently possible to use similar elementsas carrier, such as the substitution of barium forradium. The inert form then follows normal chem-istry, carrying the radionuclide with it. It is alsoworth noting that even when the separation tech-nique does not depend on the mass of elementpresent, a carrier may still be useful in prevent-ing unwanted coprecipitation or absorption onglassware.

Because of the high degree of purification re-quired in radiochemistry, it has been customaryto make a number of repeated separations eitheridentical or different to insure purity. To carrythese out in a reasonable time it is better to lose asmall amount of the nuclide sought while remov-ing a large fraction of the undesired materialrather than retaining all of the nuclide and muchof the undesired material. Fortunately, it is possi-

ble to measure these losses in analysis and tomake a correction at the end. One approach is tomeasure the carrier at the end of the analysiseither by gravimetric or instrumental analysis. Ifthere was none of the carrier substance presentin the initial sample, the fraction recovered willbe equal to the fraction of desired radionucliderecovered. Where carriers are not available orwhere they are present in variable amounts in theoriginal sample it is frequently possible to useradioactive tracers, that is radioactive isotopes ofthe same element or a similar element. The frac-tion of the amount added that is left at the end ofthe analysis can be used to determine the recov-ery of the radionuclide sought.

In the chemical separations there is consider-able use of classical analytical chemistry basedon precipitation and in most cases it will at leastappear as a final collection step to put the desiredradionuclide in a condition suitable for counting.In addition the general techniques of ion ex-change, liquid extraction, distillation, and elec-trolysis are used.

Since many of the separations in radiochemicalanalysis are carried out in small volumes the cen-trifuge is widely applied, It has become commonpractice to redissolve and reprecipitate ratherthan to wash precipitates carefully. This may berepeated several times very rapidly and will gen-erally give good decontamination (separationfrom other radionuclides). As noted above, wherethe element sought and the contaminant havesimilar properties it is preferable to use two dif-ferent precipitations, a precipitation followed byan extraction, or any two widely different steps toachieve good decontamination.

One process that is frequently of value is calledscavenging. This term is usually applied to a pre-


cipitation carried out to remove contaminating ra-dionuclides after the bulk matrix of the samplehas been removed. The process involves the addi-tion of a group carrier, precipitation, and discard-ing of the precipitate. A typical example would bean addition of iron carrier followed by hydroxideprecipitation to remove rare earths and otherheavy metals in a determination of an alkali oralkaline earth. Frequently the scavenging proce-dure is repeated to improve decontamination, butit is only rarely that the scavenge precipitate isredissolved and reprecipitated to recover any ofthe desired constituent that may have been ab-sorbed.

After suitable radiochemical separations havebeen made, it is necessary to collect and mountthe sample for counting. This is most often doneby precipitation or by elect redeposition. In thesection on measurement, the requirements for

sample preparation are discussed: for example,the sample area and mass should be reproducibleand the sample should be mounted on a sampleholder identical with that used for the counterstandard. Since it is frequently necessary to de-termine the weight of precipitate for recovery de-termination, this factor must also be considered.

The selection of a mounting technique is usual-ly a compromise between convenience and thecounting requirements. Besides consideration ofthe type and energy of emission, practical matterssuch as the counter size and sample mounts avail-able must be weighed,

In certain cases, the total amount of sampleavailable may be limited and analysis for severalradionuclides may be required. Proceduresshould be on hand for the sequential analysis ofsingle samples, even though separate samples areused for routine work.

MEASUREMENT

The method of measurement to be selected de-pends on the type of radiation, the form of thesample, and to some extent on the amount of ra-dioactivity. It is necessary that the complete ana-lytical procedure be designed so that the sampleis brought to a suitable form for the equipmentand conditions that exist.

It is possible to measure the total gamma, totalbeta, or even the total alpha activity on a sampleof food, Unfortunately, such data are valueless inestimating human exposure. The accuracy of thedetermination is very poor, natural potassiumusually interferes, and the chemical and radia-tion characteristics needed to evaluate possiblehazard are not known. It is possible, however, toset a particular total activity level as a screeninglevel for a specific food. If the measured value isbelow the screening level, no analyses for individ-ual radionuclides are performed. In such a case,the measurements should be considered as inter-nal data only and the numerical results should notbe published. Any report should merely list thesamples as having activities below the statedscreening level.

Qualitative identification of radionuclides onoriginal samples of foods is limited to gamma emit-ters at relatively high levels. The sensitivity canbe increased if some bulk reduction, as describedunder “Sample Preparation, ” is carried out. Theidentification of alpha and beta emitters dependson radiochemical separation for element identifi-cation and the measurement of energy or half-life

on the separated material for radionuclide identi-fication. The latter step may be omitted if otherconsiderations limit the possibility to a single ra-dionuclide.

The equipment available for quantitative meas-urement of radioactivity is sufficiently sensitivefor all foreseeable cases. Instruments for detec-tion of the three major emissions are describedhere, and their applications are shown in table 20(chapter VIII) in terms of the detection limits forvarious radionuclides.

It is worth pointing out that a number of the in-struments described are not commercially avail-able. They have been in the past, but the low de-mand has removed them from the market, Thenecessary electronic components are availablebut the mechanical assemblies for the detectorsare not, and the larger laboratories tend to buildtheir own systems.

Alpha Emitters. The measurement of alpha ac-tivity is best carried out on a very thin sample toavoid self-absorption of the alpha particles. Thisis even more true for spectrometry since degrada-tion of the original alpha energy will give a spec-trum with poor resolution. The measurement ofthe total alpha activity can be carried out eitherin thin-window counters or by scintillation count-ing with zinc sulfide phosphor. Both techniqueshave high efficiency but the scintillation methodcan give a considerably lower background with aconsequently lower limit of detection. Unfortu-nately, neither system is readily available as a

A p p e n d i x I — A n a l y s i s o f F o o d s f o r R a d i o a c t i v i t y . 2 2 1

complete unit ami must be assembled from commef(~ially available components. A somewhat less sensitive technique is to use the liquid scintillation spectrometer rlescribed in the next section.

Alpha spectroscopy also has two possibilities. the Frisch grid ionization chamher or the silicon diode solid-sta te detector. The Frisch grid unit can handle large area samples but is slightly poorer in resolving closely sepa ra ted energies. On the other hancL th(~ silicon diodes available are all quite small and can only count samples of about 1-em diameter with high efficiency. This is not a serious limitation, since the Frisch grid systems are not pres(mtiv Hvailable in this country. This m(~,lT1S that the form of the sample presented to the counter is limited. i\lost metclis are electrodeposited onto small smooth discs of stainless steel, nickel, or pl,lIinum, vvith evaporation of pure solutions being used in some cases.

A specializ!~d t(~chniqlle is available for measuring raclium-226. by means of its gaseous daughter product radon-222. If the sample can be put into solution and stored for about 3 weeks the r;!(jon-222 \vill be close to equilibrium (equal activity) vvith the raclium-226. The gaseous radon can then be transferred to a scintillation counter or an ionization chamber for measurement of its "I .. h" ""i;";i,, 'rh;" ;" "nih"," n""";"l;",,,rl n~r~ ih", oqJII<1 <1LllVlly. 1111" 1<> I rtlllt"1 "PC\,IClLIE,Gll. rtllll lilt"

equipment is not listed in this report since radium-226 can be determined by other procedures. For exampl(~. sepa ra tion and c()un ting as the sulfate or chromate using barium as carrier is quite adequate.

Beta Emitters. Beta counting is a little more flexible in the mass of material that can be present at the time of counting. This is true for higher energy beta emitters but carbon-14 and tritium present a problem. Since each individual beta emitter gives off particles with a range of energies from zero to a characteristic maximum. beta spectrometry is not possible for food samples. Some quaiitative information may be obtained from absorption measurements. which allow an estimate of the characteristic maximum energy of single emitters or simple mixtures (5).

The available counting equipment for quantitative meaSllf!:ment includes geiger counters. proportional counters. and scintillation counters. The geiger counter is relatively inexpensive and requires only simple electronics but is not popular and is generally not available as a counting system. The thin-window proportional counter is used widely Hnd has both reasonably high efficiency and low background. For low-level samples the background can be further reduced by antico-

incidence techniques. These add to the complexity and cost of the system but are sometimes necessary.

Scintillation counting can be performed in two ways. Solid scintilla tors can be used for counting chemical precipi ta tes collected on fil ter papers anci liquid scintilla tors can be used whenever the sample cnn be made miscible with the scintillating solution itself. This can even be cione \vith solids by suspending them in a scintilla ting gel. The advantage of liquid scintilla tors is their high efficiency, even for the low-energy emi tters carbon-14 and tritium. Scintillation systems for counting precipitates are not commercially available at present. There are. hovvever, many liquid scintillation systems on the market. most of them with automatic sample changers. They enjoy a good demand in medical and biological studies with tritium and radiocarbon. These use high levels of activity and short counting time. The better systems also have a provision for rn ther crude spectrometry. They can distinguish qualitatively aiHi quantitatively among carbon-14. tritium, alpha emi tters. and higher energy beta emi t ters.

Gamma Emitters. Gamma rays are so penetrating that the detector must have a considerable mass to absorb enough to produce a response. For spe(~tr()m8try, complete energy RIJs{)rpti{)n is re ... qui red, so solid detectors are most useful. Sodium iodide is a popular detector, since crystals can be fabricated in large sizes and the crystal is transparent to the scintillations produced hy radiation. Sodium iodide is used for most of the gamma counting today, and is still useful in spectrometry. It has high efficiency but poor energy resolution and is now applied to samples where some separa tion has taken place. i\,1ilk is a good na turnl example, as the metabolism of the cow removes most gamma emitters other than isotopes of cesium and iodine.

More complex spectra can be resolved with the solid-state germanium diode detector. A comparison of a sodium iodide with a germanium diode detector spectrum is shown in figure 1-1. It is possible to resolve the sodium iodide spectrum of a sample containing several radionuclides but this requires a sizable computer and very careful work to allow for drift in instrument characteristics and for the possibility of other radionuelides than those sought being present. The diode detector gives much better resolution and present spectrometer systems frequently include sufficient computer capacity to resolve the resulting spectra and produce quantitative data. At the same time, interferences from radionuelide im-


Figure I-1 .—Comparison of Gamma Spectra Taken With a Sodium Iodide Detector (Upper Curve) anda Germanium Diode Detector (Lower Curve). Note that the main single peak in the sodium iodide spectrum is

resolved into 9 peaks for 6 radionuclides with the germanium diode spectrum “

o

purities are greatly reduced compared to spectrafrom sodium iodide detectors. The efficiency ofthe diode is low and, for many analyses, a spec-trometer can only be used for one measurement aday. Another disadvantage is that the detectormust be kept at liquid nitrogen temperature tomaintain its detection capability. Newer diodeshave been developed that do not require storageat liquid nitrogen temperatures; however, theymust be cooled during measurement. Since thesystem is in use most of the time, this may not be asignificant advantage.

Diode spectrometers may also be used to meas-ure the low-energy gamma-rays that accompanyalpha emission. This allows direct measurementin some environmental samples, but the levels in

foods have not been high enough for this tech-nique.

General Requirements. The choice of a count-ing procedure depends on the precision required.In turn, the relative precision of a quantitativecounting measurement is inversely proportionalto the square root of the number of counts ob-tained. Thus, any improvement in precision mustbe obtained by increasing the number of counts.This can be done by using larger samples, bycounting for longer times, or by using counterswith higher efficiency. A secondary improvementis possible for low-activity samples by decreasingthe background. Each of these improvements hassome drawback, and selection of the optimum bal-ance requires a degree of experience to weigh

Appendix l—Analysis of Foods for Radioactivity “ 223

cost, manpower, and quality. For example, han-dling larger samples increases the effort in sam-ple preparation and radiochemistry while longercounting times require more counters, and in-creased counter efficiency or lower backgroundis expensive.

Counting-room operation requires the mainte-nance of detailed records on standardization,background, and sample measurement. This in-formation can frequently allow the recovery ofbad data or the correction of calculational errors.

In addition, maintenance of control charts(5) willsignal when instrument problems arise.

Experience with modern nuclear instrumenta-tion has been good, and downtime of the order of 5percent is common. This requires that service beavailable immediately or the next day following abreakdown. The increasing complexity of measur-ing systems tends to preclude in-house servicingin most cases, but some diagnostic capability andcompetence in minor repairs is very valuable.

CONTROLS

An analytical laboratory carrying out measure-ments of radioactivity in foodstuffs will require aprogram to document the validity of the measure-ments, This is necessary in legal cases and is alsohighly desirable when data are being presentedfor use in decisionmaking. A suitable program ofquality control should be carried on in addition tothe necessary calibrations and standardizations.

Calibration is the determination of the relation-ship between a desired quantity and the responseof a particular instrument. In the case of radioac-tive materials this may mean that the instrumentmust be calibrated with each radionuclide to bemeasured unless there is evidence that the instru-ment response is independent of the energy orother characteristics of the radiation. Alterna-tively a complete response v. energy calibrationmay be substituted. These calibrations, to have alegal standing, should probably be traceable tothe U.S. National Bureau of Standards or compar-able authority. This turns out to be a requirementthat is far from trivial in the effort required.

Fortunately, a complete calibration is not re-quired at frequent intervals and, depending onexperience, may not be needed oftener than everyfew months. In the meantime of course it is neces-sary to have assurance of the proper operation ofmeasuring equipment but this can be done withsimple standards or even with samples that arereproducible over a period of time. For example, asimple counter standard may be run every morn-ing before starting operations just to be sure thatthe counter is working properly. Similar stand-ards should also be available for checking spec-trometer energy response.

An ideal quality control program should in-clude the checking of the complete procedurefrom sample preparation through measurement.

This requires that standard samples be availablefor testing the full procedure. Additional controlswould include running of blind duplicate samplesto test reproducibility of analyses, and blanksamples to check on the possibility of laboratoryor reagent contamination. These three types ofsamples should make up at least 10 percent of thelaboratory output if the quality of measurement isto be followed closely. It is also necessary that thedata be published with the ordinary laboratoryresults and that any deviations from the expectedresults should be used to make suitable correc-tions in laboratory operations.

There is always some danger that a laboratorymay drift out of control in spite of an adequate in-ternal quality control program, This is possiblesince an internal program may emphasize consist-ency while the absolute values drift. Therefore,some part of the quality control effort must bedevoted to intercomparisons with establishedgroups, such as IAEA or EPA. This should be doneon an annual basis, at least.

It must be noted that standard samples are fre-quently not available, and recourse must be hadto “spiking,” which is the addition of a knownamount of a radionuclide to a sample that is rela-tively free of that nuclide. Recovery of the spike isnot necessarily a good test of a chemical proce-dure, since a natural sample may be more diffi-cult to analyze.

A final component of a good quality control sys-tem is a careful, responsible review of all the dataproduced. This means checking the arithmetic,knowing the characteristics of the equipment and,hopefully, having a sixth sense that recognizesthat certain results do not look right. This isespecially true when relying on automatic count-ers or on computer processing of the data.


STAFF AND FACILITIES

A minimum facility could be designed around astaff of six, including a B.S. or M.S. senior chemistwith experience, a B.S. junior chemist and threetechnicians for chemistry, sample preparation,and counting-maintenance, A- secretary-adminis-trative assistant could handle reports, local pur-chasing, and similar duties. With this small staff,considerable versatility and flexibility would berequired.

The measurement of radioactivity in foods re-quires fairly extensive facilities to handle thevaried analyses that might be called for, In addi-tion to a modest office space, separate roomswould be necessary for sample preparation, wetchemistry, measuring equipment, and for mainte-nance support and storage. The first two lab-oratory rooms would require hoods, chemicalbenches, and laboratory safety devices such asshowers and eye fountains. Since the samples tobe measured are normal foods there are no spe-cial requirements for radiation safety, If the sam-ples were radioactive enough to be a personnelhazard in the laboratory they certainly would notrequire this type of precise measurement.

If the laboratory is to be in continuous opera-tion it is most likely that the principal chemicaland measurement systems will have to be avail-able at least in duplicate. This and similar consid-erations lead to the conclusion that a certainminimum size and sample throughput are not nec-essary if the radioactivity laboratory is part of alarger operation which can furnish support. Onecontinuing problem is the need for electronicmaintenance of radiation measuring equipment.This is rather specialized and must be consideredeither when staffing the laboratory or in locatingit where such services are readily available.

It is not always possible to purchase the idealcounting equipment for a particular purpose. Thetotal demand for many systems is small and com-mercial instrument makers are not interested.Large laboratories can make some of their ownequipment but this is not possible for the groupdescribed here. The construction of alpha andbeta radiation detectors requires services of afirst-class machinist plus sufficient electronicknow-how to transfer the detector signal to theavailable commercial equipment. Where such ca-

pability already exists in the overall organization,it may be fruitful to copy advanced noncommer-cial instruments, but staffing specifically for thisfunction is probably not economical. Thus, mostsmall laboratories must operate with the less-than-optimum commercial equipment,

If significant uses are to be made of gammaspectrometry there will be a need for at leastmodest computing facilities. These can be part ofthe purchased spectrometer system but this is anexpensive method if suitable computer time isavailable otherwise. The former approach is usedhere, and it might be noted that the spectrometercomputer has capacity for other work.

Additional space is required not only for stock-ing necessary reagents and materials but also forstorage of incoming samples and for storage ofresidual material from samples that have beenrun. As a general rule it is desirable to take alarger sample than would be required and to setaside a portion of this for possible contingenciesor if the data are later brought into question.

The output of a group this size should run be-tween 1,000 and 4,000 samples ] a year, dependingon the difficulty and the activity level, One limit-ing factor is that, at background levels, eachcounter can only turn out one or two samples perday. Following a contaminating event, it should bepossible to process smaller samples and to countfor shorter times, both of which would allow high-er output. It might be possible to add 3 more lab-oratory staff plus 50 percent more space andequipment dollars and essentially double the out-put. Further increases would probabl y run intoother bottlenecks due to the need for added sup-port.

Table I-1 details the costs for space modifica-tion and furnishings ($140,000), for equipment($200,000), and for supplies ($20,000). These costsare based on a building in place, lights and parti-tions in place, and utility stubs in each room, It isalso assumed that building services are provided.

IA sample is defined as a ~x]mp]ete gamm{i spectral an[]lysisor a single raciionuclicie an[]lysis that requires preparation plusr:idi(x:hemist rv.

Appendix I— Analysis of Foods for Radioactivity ● 225

Table I-1. —Costs for Space, Furnishings, Equipment, and Supplies

Cost cost

Space modification and furnishings

Chemical laboratory—20 x 30 ft.Base cabinets $ 8,500Wall cabinetsStorage cabinetsS i n k s ( 2 )Hoods (3)Services and Installation .

Preparation area— 10x 30 ft.Base CabinetsWall cabinetsStorage cabinetsSinkHoodBenchesServices and Installation

Counting area— 10x 20 ft. (air-conditioned)BenchesRackBase cabinetsWal l cabinetsDesk, etc.Services and installation

Storage area— 10x 20 ft.Acid storageSolvent storage ,.CabinetsServices and Installation

Office area —library 10x 20 ft.Desks, etc (4)C a b i n e t s , b o o k c a s e sServices and Installation

Equipment

Chemistry areaHot plates (2)Stirring plates (6)pH meterDemineral izerCentrifuges (3)Platinum ware(12 items)

2,5001,8001,2007,000

40,000

3.8001,8001.500

6002,500

60025,000

1,2001,0001,700

600500

12,000

800600

3,0001,500

2,0003,0001,000

600900350500

1,80012,000

APPENDIX I REFERENCES

1.

2.

3.

E. P . I.aug, et al. ‘radioactivity in F’oods, ”’ journalA(~A(; 44, 52.3-30 ( 1961), :+Iso 4’Strontium-90 An~~l-yses of I [uman an(i A n i m a l F o o d s , H(l(liologic(]] 4.HefJ](h fljt{] 1, 36-40 (hfay 1960).‘1’een-Ager Diet Survey, Food and Drug Administra-tion, ll{~(lif)]f)gicfl) Fff?fI)th ll{!(fl 4, 18-22 ( 1963)R. E. Simpson, E. J. Barat ta, and (;. h’. Jelinek, “Radio- 5.nucli[ies in F’(Nxis: Nloni (orin~ Program. ” Rcl(fio (i{)n

Data un(~ Heports 15, 647-56 ( 1974). (rI’his is a laterreport thti t summa rizes much of the EPA program. )J. H. Harley and Joseph Rivera, Summ[lr} of Av(lil-a b l e h] ta on the Stron fium-{~[~ Con tent of F(mds (]n(lTotal Diets in the United St{ltcs, LJ, S. AEC Repor tHASL-90 (August 1960).J. 11. Ilarley IEciitor). EML Procmiures M(lnu(ll, U . S .I)OE Report F{ ASL-300 ( 1972, revised annually).


GENERAL REFERENCES

Radiochemical Methods

National Academy of Sciences, Nuclear Science Series(NAS-NS 3001 ff), Office of ‘1’echnical Services,Washington (1960 and later), A Series of Mono-graphs on Individual Elements and Techniques.

WHO-FAO-IAEA, Methods of Radiochemical Analysis,World Health Organization, Geneva ( 1966).

Johns (Editor), National Environmental Research Cen-ter, Las Vegas, Handbook of Radiochemical Analyti-cal Methods, EPA-680/4-75-001, February 1975.

Harley (Editor), Environmental Measurements Labora-tory, DOE, EML Procedures Manual, HASL-300,1972 (Updated Annually).

International Atomic Energy Agency, Reference Meth-ods for Marine Radioactivity Studies (Sr, Cs, Ce, Co,Zn), Technical Reports Series No. 118 (1970).

Ibid. (Transuranics, Ru, Zr, Ag, I)Technical Reports Series No. 169 (1975)

Measurement of Radioactivity

National Council on Radiation Protection and Meas-urements, A Manual of Radioactivity Measure-ments Procedures, Report 58 ( 1977).

National Council on Radiation Protection and Meas-urements, Environmental Radiation Measurements,Report 50 (1976).

International Commission on Radiation Units and Meas-urements, Measurement of Low-Level Radioactiv-ity, Report 22 ( 1972).

International Atomic Energy Agency, Measurement ofShort-Range Radiations, Technical Reports SeriesNo. 150 ( 1973).

International Atomic Energy Agency, Metrology andRadionuclides (Symposium), IAEA, Vienna ( 1960).

Crouthamel (Editor), Revised by Adams and Dams, Ap-plied Gamma-Ray Spectrometry, Pergamon, NewYork (1970).

Data Compilations

Lederer and Shirley, Table of Isotopes—7th Edition,John Wiley, New York (1978).

Health, Scintillation Spectrometry, Gamma-Ray Spec-t r u m Catolog ( 3 r d E d i t i o n ) , USAEC R e p o r tANCR-lOOO ( 1974).

Chanda and Deal, Catolog of Semiconductor Alpha-Par-ticle Spectra, USAEC Report IN-1261, March 1970,

Niday and Gunnink, Computerized Quantitative Anal-ysis of Gamma-Ray Spectrometry, USAEC ReportUCRL-51061, 4 Volumes (Includes Decay SchemeData Library).

Quality Assurance

Kanipe, Handbook for Analytical Quality Control in Ra-dioanalytical Laboratories, EPA Report 6 0 0 / 7 -77-088 (April 1977),

Ziegler and Hunt, Quality Control for EnvironmentalMeasurements using Gamma-Ray Spectrometry,EPA Report 600/7-77-144 (I)ecember 1977).

See also:Metrology Needs in the Measurement of Environmental

Radioactivity Environment International 1, 7 f f(1978),

GLOSSARY

Activation Products—The radionuclides formed ineither fissionable material or in surrounding mate-rial during a nuclear reaction, usually by capture ofneutrons,

Carrier—A stable element added in radiochemicalanalysis to provide sufficient mass that a radionu-clide can follow various separation steps.

Detec tor—Any device that transforms the radiationemitted by a radionuclide into a signal that can behandled electronically. The most usual signal is anelectrical pulse which can be recorded.

Disintegration-The spontaneous process by which aradionuclide gives off energy in the form of nuclearradiation.

Fissionable Material—Any of the heavy radionuclideswhich can undergo fission, or splitting into two ormore lighter atoms. The fission process is accompa-nied by a large release of energy.

Fission Products—The radionuclides formed when fis-sionable material splits into two or more atoms.

Geiger Counter— A detector based on gas ionizationwhich converts any ionization within the detectorinto a large electrical pulse.

Germanium Diode Detector—A gamma-ray detectorcomposed of very pure germanium, usually acti-vated with lithium.

Indicator Food—A food that is measured to give anestimate of the total dietary intake of man.

Ionization Chamber—A radiation detector based ongas ionization which converts any ionization withinthe chamber into an equivalent electrical pulse.

Pathway—The route by which a radionuclide is trans-ferred from the source through the environment toman.

Picocurie—See units of radioactivity. Equal to 2.2 dis-integrations per minute.

Proportional Counter—A detector based on gas ioniza-tion which converts any ionization within the detec-tor into an amplified electrical pulse proportional tothe amount of ionization.

Radionuclide—Any atomic species which is unstableand gives off nuclear radiation to attain stability.

Scintillation Detector—A detector which converts nu-clear radiation to a pulse of light. This, in turn, canbe converted to an electrical pulse with a photo-tube.

Sodium Iodide Detector—A scintillation gamma-ray de-tector made up of a single crystal of sodium iodideactivated with silver.

Spectrometry—The measurement of the energy of radi-ation emitted during decay of a radionuclide or mix-ture of radionuclides. Quantitative as well as qual-itative information may usually be derived.

A p p e n d i x I — A n a l y s i s o f F o o d s f o r R a d i o a c t i v i t y . 2 2 7

Tracer—A radionuclide added in radiochemical anal-ysis to follow the distribution of the desired constit-uent in various separation steps.

Transuranic Element—An artificial element having ahigher atomic number than uranium, formed by ac-tivation, usually with neutrons.

Units of Radioactivity—There are various ways of ex-pressing the disintegration rate of a radionuclide.In this report the disintegrations per minute (dpm)unit is used. In other reports, the curie and its sub-multiple (millicurie, microcurie, picocurie) areused and the new International Standard nomen-clature is the bequerel, equal to one disintegrationper second.

Commissioned

Appendix

Papers

1, Clement Associates, Inc., Washington, D. C.,“Priority Setting of Toxic Substances forGuiding Monitoring Programs. ”

2. Congressional Research Service, Science Pol-icv Research Division,●

●

●

Christopher H. Dodge, “Contamination ofFood by Radioactive Nuclides. ’Sarah L. Hartman, “Nitrate Food Contami-nation: A Case Study. ”Jo-Ann M. McNally, “Mercury: Environmen-tal Contamination of Food, ” “Polybromi-nated Biphenyls: Environmental Contamina-tion of Food, ‘“Polychlorinated Biphenyls:Environmental Contamination of Food. ”

3. Dr. William E. Cooper and R. V, Farace, De-partment of Zoology, Michigan State Universi-ty, “Toxic Substances in Food InformationSystem: Design and Management.”’

4. Dr. Kenny S. Crump and Marjory D. Master-man, Science Research Systems. Ruston, La.,

5

6.

7.

8,

90

“Assessment of Carcinogenic Risks FromPCBS in Food. ”Dr. Donald Epp, Department of AgriculturalEconomics and Rural Sociology, PennsylvaniaState University, “Determining the EconomicImpact of Alternative Action Levels. ”Dr. Naomi H, Harlev, Institute of Environmen-tal Sciences, New “York University MedicalCenter, “Analysis of Foods for Radioactivity. ”Dr. John L. Laseter, Environmental Affairs,Inc., New Orleans, La. “Approaches to Moni-tor ing Organic Envi ronmenta l Contaminantsin Food’ (appendix G).Dr. Frank Lu, School of Medicine, Universityof Miami, “Environmental Contaminants inFood: International Activities. ”Dr. Rodney K. Skogerboe, Department ofChemistry, Colorado State University, “Ana-lytical Systems for the Determination of Met-als in Food and Water Supplies” (appendix H),

10. SRI International, Menlo Park, Calif., “Assess-ment of Methods for Regulating ‘Unavoidable’Contaminants in the Food Supply.”’

11, Tracer J i tco, Inc. , Rockville, Md. , “CaseStudies on Consumer Risk from EnvironmentalContaminants in Food. ”

,

228

Appendix K

Executive Agency Observers atAdvisory Panel Meetings

U.S. Environmental Protection Agency

Fred Arnold* Marty Halper*

Rufus Morrison Janice Ryan

U.S. Department of Health, Education, and WelfareFood and Drug Administration

John Wessel* Charles Jelinek Pasquale Lombardo

Bart Puma Sidney Williams

U.S. Department of AgricultureFood Safety and Quality Service

Grace Clark* Richard Ellis

William Leese John Spaulding

*Official agency liaison to the assessment,

229