dietary exposures to food contaminants

Environmental Research Section A 84, 170}185 (2000)doi:10.1006/enrs.2000.4027, available online at http://www.idealibrary.com on

Dietary Exposures to Food Contaminants across the United States1

Charlotte P. Dougherty,* Sarah Henricks Holtz,* Joseph C. Reinert,- Lily Panyacosit,?Daniel A. Axelrad,- and Tracey J. Woodruff-

*Industrial Economics, Inc., 2067 Massachusetts Avenue, Cambridge, Massachusetts 02140; -U.S. Environmental Protection Agency,Of=ce of Policy, Economics and Innovation, Washington, DC 20460; and ?School of Public Health, University of California, Berkeley,

California 94720

Received March 16, 1999

Food consumption is an important route ofhuman exposure to pesticides and industrialpollutants. Average dietary exposures to 37 pollu-tants were calculated for the whole United Statespopulation and for children under age 12 yearsby combining contaminant data with food con-sumption data and summing across food types.Pollutant exposures were compared to benchmarkconcentrations, which are based on standard tox-icological references, for cancer and noncancerhealth effects. Average food ingestion exposures forthe whole population exceeded benchmark concen-trations for arsenic, chlordane, DDT, dieldrin,dioxins, and polychlorinated biphenyls, whennondetects were assumed to be equal to zero. Foreach of these pollutants, exposure through Ashconsumption accounts for a large percentage offood exposures. Exposure data for childhood agegroups indicated that benchmark concentrationsfor the six identiAed pollutants are exceeded bythe time age 12 years is reached. The methodsused in this analysis could underestimate risksfrom childhood exposure, as children have a longertime to develop tumors and they may be moresusceptible to carcinogens; therefore, there maybe several additional contaminants of concern. Inaddition, several additional pollutants exceededbenchmark levels when nondetects were assumedto be equal to one half the detection limit. Uncer-tainties in exposure levels may be large, pri-marily because of numerous samples withcontaminant levels below detection limits. ( 2000

Academic Press

Key Words: consumption; contaminant; cumula-tive; data base; exposure; food.

1 The views expressed in this paper are those of the authors anddo not necessarily re8ect those of the USEPA.

170

0013-9351/00 $35.00Copyright ( 2000 by Academic PressAll rights of reproduction in any form reserved.

INTRODUCTION

Food consumption represents an important path-way for exposure to contaminants from a variety ofsources, including pesticide application and con-tamination of water from industrial sources. Recentstudies have indicated that exposures to contami-nants in food may pose a public health risk (NationalResearch Council, 1993; MacIntosh et al., 1996). Forexample, MacIntosh et al. (1996) found that someportion of the adult population may be exposed toindividual contaminants in food at concentrationsabove thresholds of concern. Reports from the Na-tional Research Council of the National Academy ofSciences (NRC) and the Environmental WorkingGroup have also found that pesticide exposures tochildren could be high enough to cause immediateadverse health outcomes (National Research Coun-cil, 1993; Wiles et al., 1998).

These reports have focused either on a small groupof contaminants and speci7c types of foods, such aspesticides in fruits and vegetables (National Re-search Council, 1993; Wiles et al., 1998), or on a sub-set of the United States population, such as adults(MacIntosh et al., 1996) or children (National Re-search Council, 1993; Wiles et al., 1998). A morecomprehensive assessment of food contaminant ex-posures, which would include both pesticides andindustrial contaminants, as well as both childrenand adults, will help identify which populations aremost at risk from contaminant exposure in foods andwhich contaminants have the greatest public healthsigni7cance.

To better assess the national distribution of expo-sures to a broad array of contaminants in food, datawere collected on 37 contaminants in foods andcombined with estimates of consumption from die-tary pro7les for demographic groups across the

FIG. 1. Overview of analytical methods for food contamination exposures.

DIETARY EXPOSURES TO FOOD CONTAMINANTS 171

United States. The analysis considers exposures toindividual contaminants from multiple foods andmultiple contaminants in combination. This in-formation provides a screening-level assessment foridentifying contaminants of highest concern. Thisarticle presents an overview of the analysis of expo-sures from ingestion of food contaminants, includinga description of the contaminants, foods, and popula-tion subgroups analyzed, the sources of data used,the analytical approach taken, and estimates ofaverage food ingestion exposures for the whole popu-lation and for children under age 12 years.

METHODS

This analysis estimates dietary exposures to cer-tain pesticides and industrial chemicals through se-lected foods considered representative of the diet ofthe American population. The analysis included (1)creating a food consumption database that providesinformation on consumption patterns by demo-graphic characteristics, (2) creating a food con-taminant database by obtaining and compilingcontaminant data, (3) combining contaminant andconsumption information to estimate exposures from

2 Abbreviations used: CSFII, Continuing Survey of Food In-takes by Individuals; DDT, dichlorodiphenyltrichloroethane;FDA, United States Food and Drug Administration; MARCIS,Microbiology and Residue Computer Information System; NFPA,National Food Processors Association; NSI, National SedimentInventory; PCB, polychlorinated biphenyls; PDP, Pesticide DataProgram; RfD, Reference Dose; TAS, Technical Assessment Sys-tems, Inc.; TDS, Total Diet Study; USDA, United States Depart-ment of Agriculture; USEPA, United States EnvironmentalProtection Agency.

individual food types, (4) estimating total dietary ex-posures by summing across all food types, and (5)comparing exposures to benchmark concentrations todetermine potential public health impacts. Figure 1provides an overview of each of these steps.

Foods selected for the analysis were those thatcomprised at least 1% of the average diet either forthe entire United States population or for certainage groups based on the 1977}1978 Nationwide FoodConsumption Survey, posed the greatest cancer riskfrom contaminants according to the National Re-search Council’s ‘‘Regulating Pesticides in Food;The Delaney Paradox’’ (National Research Council,1987), or were evaluated by the U.S. Department ofAgriculture (USDA)2 in its Pesticide Data Program

172 DOUGHERTY ET AL.

(PDP) (U.S. Department of Agriculture, 1994). Dueto a lack of contaminant data, sugar beets, canesugar, coconut oil, soybean oil, and soybeanswere dropped from the analysis, even though theymet the 7rst criterion listed above. The foods se-lected for the analysis are listed in Table 1.

The analysis estimates the dietary intake of 37 con-taminants, including 30 pesticides and 7 industrialchemicals. These chemicals, listed in Table 2, are

TABLData Sources Used for the Fo

Food Pesticides

Fruits, vegetables and grainsApples, processed TDSApples, raw PDP/TDSBananas PDP/TDSBroccoli PDPCabbage TDSCarrots, processed PDP/TDSCarrots, raw PDP/TDSCelery, processed PDP/TDSCelery, raw PDP/TDSCorn NFPA/TDSGrapefruit PDP/TDSGrapes PDP/TDSGreen beans PDP/TDSLettuce PDP/TDSOats TDS/NFPAOrange juice TDS/NFPAOranges PDP/NFPAPeaches PDP/TDSPears TDS/NFPAPeas, green TDS/NFPAPotatoes, white PDP/TDSRice TDSSpinach TDSTomatoes, processed NFPA/TDSTomatoes, raw TDSWheat 8our TDS/NFPA

Meat and PoultryBeef MARCIS/TDSChicken MARCIS/TDSPork MARCIS/TDS

FishFish, freshwater NSIFish, saltwater NSIShell7sh NSI

Dairy and Egg ProductsMilk TDSEggs TDS

Note. For pesticides, the sample data were derived primarily from ththose pesticides not included in the 7rst. Sources are as follows: T1992}1993; MARCIS, Microbiology and Residue Computer Inform1988}1993; NFPA, National Food Processors Association, 1987}191989}1991; USEPA, USEPA Dioxin Report, 1989}1991.

considered to be most important for the analysis ofcontaminant exposure through food ingestion andhave available data. Note that some of the con-taminants have been completely banned from use inthe United States but, because of their persistence,continue to have measurable residues in the foodsupply. Selection of these contaminants was basedupon the following factors and primary sources: (1)toxicity of the contaminant, based on toxicity

E 1od Contamination Database

Industrial chemicals Dioxin

TDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDSTDS

TDS USEPA/USDATDS USEPATDS USEPA

NSI NSINSI NSINSI NSI

TDS USEPATDS USEPA

e 7rst database listed. Data were taken from the second database forDS, Total Diet Study, 1988}1993; PDP, Pesticide Data Program,ation System, 1990}1995; NSI, National Sediment Inventory,

92; USEPA/USDA, USEPA/USDA joint study on dioxins in beef,

TABLE 2Contaminants Included in the Food Analysis

Pesticides Industrial chemicals

2,4-D Dieldrin ArsenicAcephate Endosulfans CadmiumAlachlor Heptachlorb DioxinsAtrazine Hexachlorobenzene ManganeseAzinphos-methyl Imazalil MercuryCaptan Iprodione PCBsCarbaryl Lindane SeleniumChlordane MalathionChlorothalonil MethamidophosChloropropham MethoxychlorChlorpyrifos O-PhenylphenolCyanazine PermethrinDDT a SimazineDiazinon ThiabendazoleDiphenylamine Toxaphene

Note. Contaminants in italics have been banned for all uses.a Exposures to DDT assessed in this study represent the sum of

the isomers o,p@-DDT, p,p@-DDT, o,p@-DDE, p,p@-DDE, o,p@-DDDand p,p@-DDD.

b Heptachlor is almost completely banned and may be used onlyto control 7re ants inside buried, pad-mounted electrical trans-formers and in underground cable television and telephone cableboxes.


information in USEPA’s Integrated Risk Informa-tion System (IRIS) and information on chemical car-cinogenicity from USEPA’s Of7ce of Prevention,Pesticides and Toxic Substances and a National Re-search Council study (National Research Council,1987); (2) frequency of detection in foods as deter-mined by the Food and Drug Administration’s (FDA)Total Diet Study (U.S. Food and Drug Administra-tion, 1994), USDA’s National Residue Program, andUSEPA’s National Sediment Inventory (U.S. Envir-onmental Protection Agency, 1992); (3) prevalence inthe environment, based on information on thequantity of chemicals applied to crops annually froma National Research Council study (National Re-search Council, 1987); and (4) availability of con-taminant data.

Other sources, including the National Human Ex-posure Survey (NHEXAS) (Technical AssessmentSystems, 1994) and USDA’s Pesticide Data Program(U.S. Department of Agriculture, 1995a), were alsoused in evaluating contaminants for inclusion in theanalysis.

Food Consumption Database

Food consumption levels for the foods included inthe analysis were compiled from USDA’s ContinuingSurvey of Food Intakes by Individuals (CSFII)

(Technical Assessment Systems, 1995). From 1989to 1991, the CSFII collected food consumption dataannually from 12,000 individuals considered repre-sentative of the United States population over anonconsecutive 3-day sampling period. Survey par-ticipants recorded the weight of each food eaten ateach meal during the sampling period. Consumptionwas reported in a standardized manner using bodyweights (grams of food ingested per kilogram of bodyweight). Average daily consumption values wereavailable for consumers only (individuals that con-sumed the food at least once during the samplingperiod) and for all individuals.

Food consumption data were summarized for ninepopulations, strati7ed by age and gender: childrenless than 1 year of age, children 1 to 5 years of age,males and females 6 to 11 years of age, males andfemales 12 to 19 years of age, males and females 20to 64 years of age, and adults greater than 64 yearsof age. Food consumption values were calculated foreach of the nine population subgroups and for eachof the 34 food types included in the analysis, as longas there were more than three individuals in thesubpopulation who consumed the food. Means andStandard deviation for each subpopulation were cal-culated by averaging 3-day individual averages torepresent population averages and variability. Thisanalysis calculated average food ingestion exposuresusing average daily consumption for all individualsand for children less than 12 years of age.

CSFII data include consumed quantities of pre-pared foods such as pizza and spaghetti sauce, whichwere converted into quantities of raw agriculturalcommodities. Standardized recipes were used to cal-culate the quantity of the raw agricultural commod-ity in each prepared food (Technical AssessmentSystems, 1995). Total consumption of each raw com-modity was estimated by summing the amount in allfoods.

Food Contamination Database

Residue data for the selected contaminants wereobtained from various food contaminant data sour-ces, including the USDA Pesticide Data Program,National Food Processors Association (NFPA)(Chemical Information Services, 1995), U.S. Foodand Drug Administration Total Diet Study (TDS),Microbiology and Residue Computer InformationSystem (MARCIS) (U.S. Department of Agriculture,1995b), USEPA Dioxin Report (U.S. EnvironmentalProtection Agency, 1994), USEPA/USDA Dioxin inBeef Study (Winters et al., 1994), and the NationalSediment Inventory (NSI). Of the seven food


contaminant data sources used, the NSI, MARCIS,and USEPA Dioxin Reports used raw samples, where-as the others used prepared and/or cooked samples.

Years selected for each data source were based onavailability as well as sample size (i.e., more years ofdata were used in databases in which the number ofsamples per year is small), but were generally repre-sentative of the early 1990s. As shown in Table 1,residue data for most vegetables were taken fromthe PDP database. If PDP data were not availablefor a particular food type, the NFPA data were used.If NFPA data also were not available, TDS data wereused. The MARCIS database supplied most of theresidue data for beef, chicken, and pork, while theUSEPA/USDA Dioxin in Beef Study and the USEPADioxin Report provided supplemental dioxin data onbeef, pork, chicken, milk, and eggs. TDS data wereused for industrial chemicals in all food types except7sh. Residue data for 7sh were obtained from theNational Sediment Inventory database.

USDA Pesticide Data Program. The USDA Pes-ticide Data Program started in 1991 as a monitoringprogram to collect residue data on selected fruits andvegetables. Fresh fruits and vegetables evaluated byPDP are collected as close to the consumer as pos-sible and prepared as for consumption, i.e., washedand peeled. The commodities are selected to accu-rately represent national distribution levels and pes-ticide use for a speci7c crop by growing region andseason. Of the 37 selected industrial chemicals andpesticides, the PDP database contains informationon 23.

PDP data from 1992 and 1993 were used in thisanalysis. In 1992, the USDA collected PDP samplesfrom six states representing 40% of the UnitedStates population, and in 1993, the USDA collectedsamples from nine states representing 50% of thepopulation.

The National Food Processors Association. TheNFPA monitors the illegal or unnecessary presenceof pesticide residues in processed food. The NFPAdatabase consists of residue data, from both randomand targeted sampling, of raw foods purchased spe-ci7cally for processing and processed foods, such ascanned and frozen foods. The database contains resi-due data on 28 of the 37 contaminants. The databasehas small sample sizes and less sensitive detectiontechniques than PDP, resulting in higher detectionlimits. Data were obtained from the years1987}1992.

FDA Total Diet Study. The FDA’s Total DietStudy collects residue data for approximately 300

pesticides, radionuclides, and industrial chemicalsin 261 foods that represent 3500 typically consumedfoods. Sampling procedures consist of purchasingthe 261 foods in three cities in each of four regions ofthe country. The samples from the three cities with-in each region are combined into one ‘‘market bask-et.’’ Samples are collected four times per year. Thefoods sampled include processed foods (bottled, can-ned, and frozen), fresh foods including fruits andvegetables, baby foods, dairy products, fresh meats,cereals, peanut butter, and prepared foods such aspizza. The three samples of like foods are combinedinto one sample, prepared as for consumption, andanalyzed for contaminants. Data from the TDSdatabase, which includes 28 of the 37 contaminantsused in this analysis, were from 1988 to 1993. Animportant limitation of the study is its small samplesizes.

The Microbiology and Residue Computer Informa-tion System. MARCIS is a USDA-sponsored monit-oring program, the purpose of which is to ensurethat USDA-inspected products are safe for humanconsumption. MARCIS contains data from tests onmeat, poultry, and egg products for any of 100 com-pounds in eight classes of animal drugs and pestici-des, including nine contaminants in this study:chlordane, chlorpyrifos, DDT, dieldrin, heptachlor,hexachlorobenzene, lindane, methoxychlor, andPCBs. Contaminant testing occurs in raw fat sam-ples. Thus, the residue data do not re8ect con-taminant levels in cooked or processed foods, ascooking is likely to decrease the percentage of fat andthus decrease concentrations of lipophilic con-taminants. Processing and cooking of foods may alsobreak down contaminants and reduce concentra-tions. Data from the years 1990 to 1995 were used inthis analysis.

USEPA Dioxin Report and USEPA/USDA Dioxinin Beef Study. The USEPA Dioxin Report presentsaverage dioxin levels in United States meat anddairy products, including pork, chicken, milk, andeggs, based on data from the years 1989 to 1991. Inaddition, the 1994 USEPA/USDA Dioxin in BeefStudy analyzed the concentration of dioxins in theback fat of United States beef animals using a stat-istically based sampling survey that accounted for99.9% of all beef animals slaughtered in the UnitedStates. Data for all foods are presented as dioxintoxic equivalents, which were calculated using theinternational toxic equivalency factors for chlorin-ated dibenzo-p-dioxin (CDDs) and chlorinated diben-zofurans (CDFs). Tests were performed on rawfoods, resulting in possible overestimation of total


dioxin exposure. An additional limitation of theUSEPA Dioxin Report is the small sample sizes thatwere used to develop average contaminant levels.

National Sediment Inventory. The NationalSediment Inventory was implemented in the early1980s to monitor 96 compounds, including PCBs,endosulfans, dioxins, and DDT in 7sh and shell7sh.The National Sediment Inventory compiled informa-tion on 7sh tissue contaminants from several datasources, including STORET (Storage and Retrievalfor Water Quality Data), EMAP (EnvironmentalMonitoring and Assessment Program), ODES(Ocean Data Evaluation System), and DMTS(Dredged Material Tracking System). Sampling datawere collected on a national and a regional scale, inboth freshwater and saltwater bodies. Some NSIsampling procedures, such as targeted sampling inareas thought to have high contamination levels andsampling in raw 7sh 7llets, may overestimate con-taminant concentrations and therefore exposurelevels. Data used were for the years 1988 to 1993.

Estimating Exposure

The analysis estimates exposure to speci7c con-taminants via individual foods by combining data oncontaminant concentrations in foods with data onconsumption of those foods. For each contaminantand food evaluated, the following equation was usedto calculate average exposure levels for a particularpopulation subgroup:

Average Daily Exposure (lg/kg body weight]day)

"Consumption (g/kg body weight]day)

]Contaminant Concentration (lg/g).

Having calculated exposures from each food, theestimates for each contaminant were summed acrossfood types, providing estimates of each populationsubgroup’s total average daily exposure to each con-taminant through the food pathway. There werea number of nondetects in the contaminant data. Toaccount for the possibility that there were concentra-tions below the detection limit, we used two alterna-tive assumptions. We 7rst assumed that actualconcentrations for nondetect samples were equal tohalf of the detection limit. However, this wouldoverestimate the concentrations if there were trulyno contamination of the sampled foods. Therefore,we also calculated exposures with the alternateassumption that actual concentrations for all non-

detect samples were equal to zero. These two as-sumptions provide a range of exposure estimates foreach contaminant with nondetects. Separate expo-sure analyses were conducted for children (less than12 years old) and for the whole population.

Comparison to Benchmark Concentrations

To screen for the potential public health signi7-cance of estimated exposures, exposure values werecompared to benchmark concentrations for each con-taminant. A benchmark concentration representsa daily concentration below which there is a highprobability of no adverse health effect. This is differ-ent than a benchmark dose, which is a statisticallyderived value used in setting a Reference Dose fornoncancer health effects. The benchmark concentra-tions for carcinogenic effects were derived usingUSEPA cancer slope factors and represent exposureconcentrations at which lifetime cancer risk is one inone million. This level is de7ned as a public healthprotective concentration in the Congressional HouseReport to the Food Quality Protection Act of 1996(104th Congress, 1996). In addition, the one in amillion cancer risk has been used in other regulatoryprograms, such as those for air toxics, as a de mini-mus concentration and thus an appropriate level forscreening (1990, Clean Air Act Amendments; Cal-dwell et al., 1998). The benchmark concentrationsfor noncarcinogenic effects are USEPA ReferenceDoses. The Reference Dose is an estimate, with anuncertainty spanning perhaps an order of magni-tude, of a daily exposure to the human population(including sensitive subgroups) that is likely to bewithout appreciable risk of deleterious effects duringa lifetime (U.S. Environmental Protection Agency,1990).

Of the 37 contaminants studied, 20 have availablecancer benchmark concentrations, 34 have availablenoncancer benchmarks, and 17 have both cancerand noncancer benchmark concentration values(Table 3).

Exposure levels were compared to benchmarks bycalculating hazard ratios. Hazard ratios are cal-culated by dividing the average daily exposures bythe benchmark concentrations. Hazard ratiosgreater than one indicate that the average exposurelevel exceeds the benchmark concentration.

Calculation of cancer hazard ratios for children’sexposures involved a step in which childhood expo-sures were converted to equivalent whole-life expo-sures. A lifetime average daily dose (LADD) fromchildhood exposure was calculated by summing to-gether the estimated intake levels over the 7rst 12

TABLE 3National Average Exposures and Benchmark Concentrations for Contaminants Included in the Food

Ingestion Analysisa

Exposure per capita Exposure per capita Cancer benchmarkNumber of samples nondetects"0 nondetects"0.5 DL Oral RfD concentrationb

Chemical name (and detects) (lg/kg]day) (lg/kg]day) (lg/kg]day) (lg/kg]day)

2,4 D 3,965(88) 0.001 0.009 10 ;

Acephate 6,366(961) 0.02 0.04 4 0.115Alachlor 169(0) 0 0.02 10 0.0125c

Arsenice 1,421(136) 0.2 0.2 0.3 0.001Atrazine 169(0) 0 0.02 35 0.0045c

Azinphos-methyl 9,548(581) 0.02 0.1 1.5d?

Cadmium& 1,577(751) 0.2 0.2 1 ;

Captan 6,901(573) 0.03 0.06 130 0.29Carbaryl 8,850(388) 0.02 0.06 100 0.044Chlordane 26,598(1340) 0.01 0.07 0.06 0.001Chlorpyrifos 35,591(510) 0.8 0.8 3 ;

Chlorothalonil 8,281(1079) 0.007 0.03 15 0.09c

Chloropropham 3,864(744) 0.004 0.04 200 ;

Cyanazine 76(0) 0 0.03 2c 0.001c

DDT g 26,098(1927) 0.04 0.1 0.5 0.003Diazinon 11,709(535) 0.003 0.02 0.9c

;

Dieldrin 26,608(1155) 0.003 0.03 0.05 0.0000625Dioxinsh 142(93) 1e-06 2e-06 ; 6.7e-09c

Diphenylamine 2,962(347) 0.1 0.2 25 ;

Endosulfansi 12,933(1130) 0.02 0.05 6 ;

Heptachlor 18,377(84) 0.0001 0.03 0.5 0.00022Hexachlorobenzene j 27,142(104) 0.0001 0.04 0.8 0.001Imazalil 4,317(831) 0.009 0.05 13 ;

Iprodione 10,592(1441) 0.09 0.1 40 ;

Lindane j 27,614(83) 0.001 0.02 0.3 0.00077c

Malathion 5,986(92) 0.04 0.07 20 ;

Manganesek 980(886) 20 20 140 ;

Mercury l 6,057(5047) 0.04 0.08 0.3 c;

Methamidophos 8,820(578) 0.01 0.02 0.05 ;

Methoxychlor 2,9180(93) 0.005 0.07 5 ;

O-phenylphenol 2,095(12) 0.0002 0.01 ; 0.515c

PCBsm 27,626(1523) 0.07 0.1 0.02 0.00013Permethrin 6,591(1136) 0.03 0.07 50 0.056Seleniumn 1,081(473) 1 1 5 ;

Simazine 231(1) 0.0002 0.03 5 0.008c

Thiabendazole 10,938(1749) 0.2 0.3 100d;

Toxaphene 4,009(13) 0.0004 0.2 ; 0.001

Note. Contaminants in italics have been cancelled for all uses. Source: Oral RfDs and CSFs obtained primarily from USEPA’sIntegrated Risk Information System (45) Quarter, 1995) except where indicated otherwise.

a Numbers in boldface represent whole-population average estimated exposures that exceed the benchmark concentrations for carcino-genic effects or for both carcinogenic and noncarcinogenic effects.

b Benchmark concentration for carcinogenic effects equals 10~6 divided by the cancer slope factor and represents the exposureconcentration at which lifetime cancer risk is one in one million.

c Toxicity values obtained from USEPA’s Health Effects Assessment Summary Table, 1994.d Toxicity values obtained from USEPA’s Of7ce of Pesticide Programs RfD Tracking Report, 4/14/95.e Listed as Arsenic (inorganic).f Listed as Cadmium (food).g Listed as DDT (p,p@-Dichlorodiphenyltricloroethane).h Listed as Tetrachlorodibenzo-p-dioxin, 2,3,7,8 (TCDD).i CAS 115-29-7.j Listed as Hexachlorocyclohexane, gamma.k Listed as Manganese (food).l Listed as Mercury (inorganic).m PCBs (CAS 1336-36-3) used for Slope Factor, Aroclor 1254 (CAS 11097-69-1) used for Reference Dose.n Listed as Selenium (and compounds).



years of life and dividing the result by 70 years. Thisapproach assumes zero exposure from ages 12 to 70to convert childhood exposures into lifetime termsfor direct comparison with the cancer benchmarksthat are also based on lifetime exposure.

RESULTS

Analysis of National Average ContaminantExposures

Table 3 presents national average daily food inges-tion exposures to the 37 contaminants included inthe analysis. These exposures are estimated on aper capita basis and therefore are derived fromconsumption values for all individuals surveyed, in-cluding those who did not consume particular foodsduring the survey period. For each contaminant,exposures are presented separately assuming thatnondetects in the food contamination database areequal to zero and one half of the detection limit.Values in bold face in Table 3 indicate exposuresthat exceed either cancer or noncancer benchmarkconcentrations.

Figures 2 and 3 present hazard ratios for carcino-genic and noncarcinogenic effects, respectively, un-der two assumptions: (1) nondetects are equal to zeroand (2) nondetects are equal to onehalf of the detec-tion limit. Figure 2 indicates that average exposuresto six contaminants exceed benchmark concentra-tions for their carcinogenic effects when nondetects

FIG. 2. Cancer hazard ratios for ave

are assumed to be zero. These contaminants arearsenic, chlordane, DDT, dieldrin, dioxins, andPCBs. Figure 3 shows that, assuming nondetects areequal to zero, the only contaminant that exceedsbenchmark concentrations for noncarcinogeniceffects is PCBs.

Figures 2 and 3 and Table 3 also show that aver-age exposure estimates are highly dependent uponassumptions about the value of nondetects. Whennondetects are assigned a value of one half of thedetection limit, exposure values are 2 times as highin most cases and in several cases 10 times as highas when they are assigned a value of zero. However,changing the nondetect value from zero to one half ofthe detection limit does not signi7cantly change theexposure levels for eight of the contaminants;arsenic, cadmium, chlorpyrifos, iprodione, manga-nese, PCBs, selenium, and thiabendazole.

The value assigned to nondetects also has an effecton the number of contaminants exceeding bench-mark concentrations. When nondetects are assigneda value of zero, 6 of the 37 contaminants exceedbenchmark concentrations. When nondetects areassigned values of one half of the detection limit,however, the number of contaminants exceedingbenchmark concentrations is 16. Five of the addi-tional 10 contaminants exceeding benchmark con-centrations have very small samples sizes and/ora small number of samples with detected contamina-tion. It is unlikely that average exposure levels to

rage national exposure (per capita).

FIG. 3. Noncancer hazard ratios for average national exposure (per capita).


these contaminants;alachlor, atrazine, cyanazine,simazine, and toxaphene;actually exceed bench-mark concentrations, since less than 1% of thesamples for each of these contaminants are abovedetection limits.

These results re8ect a large number of nondetectsamples in the food contamination database. Thepercentage of detects for different contaminantsranges from 0% for alachlor, atrazine, and cy-anazine to over 90% for manganese, though manycontaminants are detected in less than half thesamples.

Contribution of Individual Foods to NationalExposure. Exposures to contaminants in indi-vidual foods were analyzed to identify those foodswith the largest contributions to average nationalcontaminant exposures. To avoid highlighting expo-sures that exceed benchmark concentrations solelydue to positive values assigned to nondetects, theanalysis of the contribution of individual foodsfocused on the scenario under which nondetectswere assumed to equal zero. As indicated by Figs. 2and 3, average national exposure under this scenarioexceeds benchmark concentrations for the followingsix contaminants: arsenic, chlordane, DDT, dieldrin,dioxins, and PCBs.

Figure 4 presents the contributions of individualfoods that contribute more than 5% to total exposureto the six contaminants, assuming that nondetectsare equal to zero. As illustrated in Fig. 4, exposurelevels from 7sh account for a large fraction of totalexposure for each of the six contaminants analyzed.Exposures from saltwater 7sh contribute the largestpercentage to total exposure of all foods analyzed forthree contaminants (chlordane, 67%; dioxins, 64%;PCBs, 60%). Exposures from freshwater 7sh con-tribute the largest percentage for two contaminants(DDT, 75% dieldrin, 52%). Exposures from shell7shcontribute the largest percentage for one con-taminant (arsenic 44%). No other food type com-prises more than 5% of national exposures to any ofthe six contaminants analyzed in detail, with theexceptions of beef, which accounts for between 6 and9% of total exposure to DDT, dieldrin, and dioxins,and rice, which accounts for nearly 11% of totalexposure to arsenic. After 7sh, the largest contribu-tors to exposure are other meat and animal products,including beef, chicken, pork, and milk. The onlynonanimal products contributing more than 1% ofexposure to any of the six contaminants are rice,potatoes, wheat 8our, and spinach. All other foodsanalyzed in the study, including oats, eggs, andother fruits and vegetables, contribute negligible

FIG. 4. Contribution of individual foods to total average national exposures.


amounts to national average contaminant exposuresfor these chemicals.

National Childhood Average ContaminantExposures

This analysis examines national childhood aver-age food ingestion exposure for three age groups: allchildren under 1 year of age, all children ages 1 to5 years, and all children ages 6 to 11 years. Forcomparison to cancer benchmarks, exposures to chil-dren less than 12 years old were converted intolifetime equivalent doses, as described underMethods. Assuming that nondetects are equal tozero, childhood food ingestion exposures exceed theone in one million cancer benchmark concentrationfor the same six contaminants identi7ed above forthe whole population (DDT, chlordane, dioxins, ar-senic, dieldrin, and PCBs). Hazard ratios (exposuredivided by benchmark concentration) for exposuresto these contaminants in the 7rst 12 years of liferange from 4 to 127 (see Fig. 5). When nondetects areassumed to equal half of the detection limit, a total of13 different food contaminants have childhood expo-sures that exceed the one in one million cancer riskbenchmark.

Only one contaminant, PCBs, exceeds noncancerbenchmarks for the childhood age groups less than

12 years when nondetects are assigned a value ofzero. The number of contaminants exceeding non-cancer benchmarks increases when nondetects areset equal to one half of the detection limit. Exposureto chlordane, for example, is greater than the non-cancer benchmark for all child age groups above1 year. For children ages 1 to 5 years, exposures toarsenic, chlordane, dieldrin, and methamidophosare greater than the noncancer benchmarks, and forchildren ages 6 to 11 years, arsenic is greater thanthe noncancer benchmark.

To further explore the impact of cumulative expo-sures, noncancer hazard ratios for each age groupwere aggregated for all contaminants, since multiplecontaminants with exposure levels slightly belowthe benchmark may be associated with potentialhealth concerns when combined in aggregate. Thiswas done as a screening exercise to ascertain thepotential magnitude of cumulative exposures whenconsidering multiple contaminants and accountingfor differences in potency and hazard of the variouscontaminants. Figure 6 shows that children ages 1 to5 years have the highest aggregate hazard ratios fornoncarcinogenic effects. Children ages 6 to 11 yearshave the second highest aggregate hazard ratios fornoncarcinogenic effects, whereas children less than1 year of age the lowest aggregate hazard ratios andtherefore the lowest exposure levels compared with

FIG. 5. Cancer hazard ratio for childhood exposures up to age 12 years.


benchmark values. These relationships are the samewhether the nondetects are set equal to zero or toone half of the detection limit. Setting nondetectsequal to one half the detection limits does, however,more than double the estimate of children’s aggre-gate hazard from pesticide and industrial chemicalexposures through food ingestion. Similar resultswere found for aggregate cancer hazard ratios. How-ever, as with individual contaminant comparisons tocancer benchmarks, the aggregate noncancer haz-

FIG. 6. Aggregated noncancer hazard rations

ard ratios for children overestimates risk becausethe benchmarks assume that children would be ex-posed at the estimated levels for a lifetime.

Contribution of Individual Foods to ChildhoodExposure

Children consume more food per unit body weightand thus more contaminants per unit body weightthan adults. Children also consume different foods

for childhood exposures up to age 12 years.


and different amounts of foods than the generalpopulation. Infants, for example, have high arsenicexposure relative to other contaminants, due largelyto a high consumption of rice and high arsenic con-centrations in rice. Infants, however, have muchlower exposures to other contaminants than allother subgroups, because they do not consume fresh-water 7sh or shell7sh and only a small amount ofsaltwater 7sh per unit body weight.

The contribution of individual foods to total expo-sures was analyzed for all children ages 1 to 5 years.This analysis was performed for the six con-taminants identi7ed in the previous analysis. Re-sults are presented in Fig. 7. Similar to the resultsfor adults, the analysis shows that 7sh, meat, andother animal products are the largest contributors toexposure for children ages 1 to 5 years. In addition,rice (for arsenic) and milk (for DDT) are largercontributors to children’s exposure than to adults’exposure.

DISCUSSION

This analysis represents an important step in as-sessing and characterizing the potential hazards as-sociated with contamination of food and suggestsadditional steps to be taken to further re7ne ourunderstanding of this potential problem. Initialscreening of the data found that estimated exposureto a number of contaminants in the average diet ofadults and children exceed benchmark concentra-

FIG. 7. Contribution of individual foods to tota

tions for cancer and noncancer effects, though theseresults should be interpreted cautiously due to lim-itations of the available data.

This study provides insights on the magnitude ofpotential exposures from food contamination. Theresults of the analysis have identi7ed certain con-taminants of particular concern, including arsenic,chlordane, DDT, dieldrin, dioxins, and PCBs. Theone in one million benchmark for lifetime cancer riskfor each of these contaminants is exceeded by thetime age 12 years is reached. Exposures from age 12to 70 years pose additional risks above those result-ing from exposure in the 7rst 12 years of life. Theassessment of cancer hazard ratios for childhoodexposure applied a standard exposure metric, thelifetime average daily dose. However, this approachmay understate cancer risks, as childhood exposuresmay have greater probability of producing tumorsthan exposures in adulthood. Risks may be greaterfor children due to potential enhanced potency ofexposures in childhood as well as the increased num-ber of remaining years of life for tumors to form(McConnell, 1992; National Research Council,1993). Therefore, the risks of several pollutants withchildhood exposure hazard ratios somewhat lessthan one in Fig. 5 may actually exceed the cancerbenchmark level of one in one million. In the case inwhich nondetects are assumed to be equal to zero,these pollutants include toxaphene, heptachlor, lin-dane, carbaryl, and permethrin. For several of thesecontaminants, exposures are largely attributable to

l average exposure for children ages 1}5 years.


residues on produce. For example, childhood car-baryl exposure is associated primarily with applesand grapes.

Most of these pollutants were identi7ed becausethey exceeded the cancer benchmark, which repres-ents a one in a million risk level, rather than thenoncancer benchmark. Exposures greater than theone in a million cancer risk screening level canresult in large numbers of people potentially at risk.However, the one in a million cancer benchmarktends to be a more conservative screen than thenoncancer benchmark. For example, a study byGaylor (1989) compared the lower 95% con7dencelevel of the dose which results in 1% of test animalsexhibiting effects for both teratogenic and carcino-genic effects (this risk metric is often referred to asthe LED

01). He found that the LED

01for teratogenic

effects was similar to or lower than the LED01

forcarcinogenic effects for four of nine chemicals evalu-ated, indicating that the dose associated with a 1%effect level was similar or lower for teratogeniceffects than for carcinogenic effects. This indicatesthat the benchmark for noncancer effects may notbe as conservative a screen as that for the cancereffects.

It is important to note that some of the identi-7ed pesticides have been banned for all uses inthe United States. DDT, for example, has beenbanned for over two decades. However, becausemany of these compounds, such as DDT, are highlypersistent, they continue to show up at signi-7cant levels in the food supply, indicating the im-portance of considering persistence when assessingpotential risks.

The results indicate that exposures to contami-nants identi7ed in the analysis are largely driven bycontamination in 7sh. However, there are some ca-veats to these results. One is that the 7sh sampleswere collected and tested in raw tissues. The totalamount of contaminants actually consumed in 7shmay be less than levels present before cooking be-cause lipids and lipophilic compounds are partiallyremoved during cooking and processing. Previousstudies indicate that the cooking and processing of7sh decreases contaminant levels of at least 7ve ofthe six contaminants analyzed in detail;chlordane,DDT, dieldrin, dioxins, and PCBs;with averagelosses of these contaminants ranging from 20 to 46%(Voiland et al., 1991; Sherer and Price, 1993; Zabikand Zabik, 1995; Zabik et al., 1995). However, evenwith a 46% decrease in 7sh contaminant concentra-tions, estimated national average exposure levels forthe 7ve contaminants would still exceed benchmarkconcentrations.

In addition, the testing procedures for the 7sh inthe National Sediment Inventory include targetedtesting for contaminants expected to be present in7sh at the sampling sites or for sites that are ex-pected to have contaminants present. Overall, thispotential bias is likely to overestimate contaminantconcentrations and resulting exposures. To assessthis potential bias, contaminant data in NSI werecompared to two other data sources;FDA’s TotalDiet Study and USEPA’s National Study of Chem-ical Residues in Fish (NSCRF) (U.S. EnvironmentalProtection Agency, 1992). For freshwater 7sh, it ap-pears that contaminant levels measured in the TDSand NSCRF databases are lower than those mea-sured in NSI, although the comparisons are some-what uncertain because of the small sample sizes inTDS and NSCRF. For shell7sh and saltwater 7shthe difference in contaminant levels is mixed. Forsaltwater 7sh, there are lower contaminant levels ofarsenic, cadmium, and hexachlorobenzene in TDSand NSCRF than in NSI, but higher levels of DDT,dieldrin, and PCBs. Similarly for shell7sh, there arelower levels of arsenic, cadmium, mercury, andPCBs in TDS and NSCRF than in NSI, but higherlevels of DDT.

Evaluating the contribution of different foodsources to overall exposure 7nds that, even if thefreshwater 7sh contaminant levels in NSI areoverestimated, other foods still contribute signi7-cantly to exposure. Signi7cant decreases in the con-tribution of freshwater 7sh to exposure levels wouldresult in somewhere between 20 and 70% reductionin exposure for chlordane, DDT, dieldrin, dioxins,and PCBs. However, the exposure levels for each ofthese contaminants are at least an order of magni-tude greater than the cancer benchmarks and in thecase of dioxins and PCBs, 1000 times greater thanthe cancer benchmarks.

Arsenic is a food contaminant that has been iden-ti7ed through our screening analysis as potentiallyrepresenting a health risk. It is important to notethat the measurements of arsenic in food used in thisstudy are for total arsenic, which is made up ofinorganic and organic arsenic, with inorganic ar-senic typically comprising less than half the total.Inorganic arsenic is more toxic than organic arsenic;the results of this study could overestimate potentialrisks because the RfD and cancer benchmark used inthe analysis are for inorganic arsenic rather thantotal arsenic. A recent analysis of arsenic data fromTao and Bolger (1998) by the National ResearchCouncil (National Research Council, 1999) assessedthe daily intake of inorganic arsenic for various ageand gender groups. The assessment was based on


estimates of the percentage of inorganic arsenic inthe measured concentrations of total arsenic fromthe FDA Total Diet Study for 1991}1997. The studyestimated a range of 0.066}0.34 lg/kg/day of intakeof inorganic arsenic with an average of about0.14 lg/kg/day. The national average exposure esti-mate derived in our study for total arsenic, with noadjustment for inorganic arsenic, was 0.2 lg/kg/day,based primarily on measured arsenic concentrationsfrom the National Sediment Inventory. Our study’sestimates for total arsenic are similar to the NRCestimates for inorganic arsenic, indicating that ourestimates based on total arsenic are unlikely tosigni7cantly overestimate exposure to inorganicarsenic.

Our study’s exposure estimates represent in-formation from the early 1990s, and levels of con-taminants, particularly some of the persistentcompounds such as DDT and PCBs, have likely de-creased some since then. Most studies evaluatingtrends in persistent compounds have focused ontrends prior to 1990 or through the mid-1990s(Robinson et al., 1990; Fensterheim, 1993; Noren,1993; Papke et al., 1994; Becher et al., 1995; Boppet al., 1998). A study of PCBs in the human dietfound a 2- to 10-fold decline in PCB contamination ofshell7sh and 7sh from the early 1970s to the late1980s and similar decreases in the diet and adiposetissue of the public (Fensterheim, 1993). However,Fensterheim (1993) also notes that the rate of de-cline will slow, since most of the dramatic declinesare associated with major regulatory initiatives inthe 1970s and 1980s. Similar results would be ex-pected for other persistent compounds such as DDTand dioxins. A study of sediment in the HudsonRiver basin, which would contribute to 7sh concen-trations, also found declines from the 1960s to themid-1980s to mid-1990s in DDT, dioxins, and PCBsdue to regulatory measures, though there were a fewareas which saw increases in sediment concentra-tions (Bopp et al., 1998). Decreases in dioxins inblood and breast milk are also noted in several Euro-pean studies (Noren, 1993; Papke et al., 1994;Becher et al., 1995). Overall, these indicate thatthere has been a substantial decrease in 7sh con-taminant levels from the 1970s due to regulatorycontrols, but the rate of decline may be slower and insome cases nonexistent in the 1990s.

An important consideration in this analysis is thelarge number of nondetects. The percentage of non-detects for the contaminants ranged from 10 to100%, with an average of 86%. The analysis wasperformed assuming that nondetects were equal toeither zero or one half of the detection limit to assess

the impact of different nondetect values on the expo-sure outcomes. For some contaminants, assumingthat nondetects are equal to one half of the detectionlimit dominates the overall contaminant exposure.In these cases, assuming a value of one half of thedetection limit is likely to overstate the contaminantlevels. A more re7ned analysis of nondetects in thecase of pesticides currently in use could consider thefood source and whether pesticides were applied tothat crop to determine whether one half the detec-tion limit is an appropriate assumption. Futurework will consider the distribution of the detects inassessing the likely values below the detection limitand other methods for determining appropriatevalues for the nondetects. Despite the presence ofmany nondetects, exposures to six contaminants ex-ceed benchmarks even when nondetects are setequal to zero. In addition to the number of non-detects, the sample size is fairly small for some of thecontaminants (e.g., dioxins), and thus results forthese contaminants may be less reliable than resultsfor other contaminants.

Another source of uncertainty in this analysisarises from the fact that one food type may be com-prised of several different forms of a particular food.For example, average freshwater 7sh contaminationlevels were calculated by averaging concentrationsin different freshwater 7sh species sampled. How-ever, contamination levels are species dependent.For lipophilic contaminants, species with a higherfat content will have higher contaminant levels.Therefore, if the actual consumption of freshwater7sh is composed of 7sh species that, on average, havea higher fat content than the 7sh sampled, the pre-liminary analysis will understate exposures tolipophilic contaminants in freshwater 7sh. Whilethere are inconsistencies in the composition of otherfoods in the consumption and contaminationdatabases, it is particularly important for the three7sh categories, as 7sh comprises a large proportionof the average national exposures to certain con-taminants.

This analysis estimates average contaminant ex-posure across both consumers and nonconsumersof the individual food items. Only 4, 12, and 30% ofthe population are consumers of freshwater 7sh,shell7sh, and saltwater 7sh, respectively. Thus, thecontaminant exposure level from 7sh for the averageconsumer is much greater than the average for thewhole population. In addition, this analysis examinesonly the mean level of contaminant exposure and doesnot assess exposures for those populations that con-sume higher levels of particular foods or foods withmuch higher than average levels of contaminants.


There are also limitations associated with the foodconsumption data due to the short sampling period.USDA collected food consumption data over 3 days,which were scheduled in different seasons to accountfor the potential for seasonal variation in consump-tion. Because of the short duration of sampling,however, the consumption data may not accuratelyrepresent average daily consumption over a longerperiod of time. Better characterization of the con-sumption of foods that drive the analysis (e.g., 7shand meats) are needed to reduce the uncertaintyassociated with the consumption data. This includesbetter information on both the proportion of indi-viduals that consume each food and the quantityconsumed by these consumers.

This analysis has focused on the combined expo-sures to food contaminants that result from theirpresence in several different types of foods. Underthe Food Quality Protection Act (FQPA) of 1996, theU.S. Environmental Protection Agency is currentlydeveloping methods to extend the scope of exposuresconsidered in analysis of pesticides (U.S. Environ-mental Protection Agency, Of7ce of Pesticide Pro-grams, 2000). FQPA requires that USEPA considerthe aggregate exposures to pesticides from multipleroutes and pathways;combining food ingestion ex-posures with drinking water exposures and oral,inhalation, and dermal exposures that may resultfrom residential pesticide application. FQPA alsorequires that USEPA consider the cumulative risksto human health that may result from exposures tomultiple pesticides with common mechanisms oftoxicity. When fully implemented, these new ap-proaches will provide a broader perspective on thepotential of various pesticides to pose substantialrisks to human health.

Despite the limitations associated with the analy-sis, the results point to potentially high exposures tocontaminants in food and represent an importantstep toward better characterization of these expo-sures. The analysis has identi7ed research prioritiesfor improving the data needed to better understandfood contaminant exposures. The small number ofsamples for shell7sh and saltwater 7sh combinedwith the potentially high contribution of these foodsto total exposures for some contaminants (arsenicand PCBs in shell7sh; arsenic, chlordane, dieldrin,dioxin, and PCBs in saltwater 7sh) suggest thattesting of these contaminants in shell7sh and salt-water 7sh should be a priority for future monitoringactivities. In addition, the food contaminationdatabase contains a small number of samples ofcertain contaminants with potentially high expo-sures, including arsenic, dioxins, carbaryl, permeth-

rin, simazine, and toxaphene. Additional researchefforts should focus on collecting data for these con-taminants. Results also indicate that detectionlimits for many samples exceed benchmark concen-trations. Lower detection limits are needed to betteridentify the frequency with which contaminant con-centrations are at levels of public health concern.

ACKNOWLEDGMENTS

This work was funded by the U.S. Environmental ProtectionAgency, Of7ce of Policy. The authors acknowledge the researchsupport provided by Todd Lookingbill, Anne Petersen, and Eliza-beth Snell (formerly of Industrial Economics, Inc.); the assistancein compiling the food consumption data provided by KathrynFleming of Technical Assessment Systems, Inc. and BarbaraPetersen of Novigen Sciences (formerly of Technical AssessmentSystems, Inc.); and the technical review provided by John Evansof the Harvard School of Public Health.

REFERENCES

Becher, G., Skaare, J. U., Polder, A., Sletten, B., Rossland, O. J.,Hansen, H. K., and Ptashekas, J. (1995). PCDDs, PCDFs, andPCBs in human milk from different parts of Norway andLithuania. J. Toxicol. Environ. Health 46, 133}148.

Bopp, R. F., Chillrud, S. N., Shuster, E. L., Simpson, H. J., andEstabrooks, F. D. (1998). Trends in chlorinated hydrocarbonlevels in Hudson River basin sediments. Environ. Health Per-spect. 106(Suppl. 4), 1075}1081.

Caldwell, J., Woodruff, T., Morello-Frosch, R., and Axelrad, D.(1998). Application of hazard identi7cation information for pol-lutants modeled in EPA’s Cumulative Exposure Project. Toxi-col. Indust. Health 14, 429}454.

Chemical Information Services. (1995). National Food ProcessorsAssociation Search Results 1987}1992.

104th Congress. (1996). Food Quality Protection Act of 1996.Fensterheim, R. J. (1993). Documenting temporal trends of

polychlorinated biphenyls in the environment. Regul. Toxicol.Pharmacol. 18, 181}201.

Gaylor, D. (1989). Comparison of teratogenic and carcinogenicrisks. Regul. Toxicol. Pharmacol. 10, 138}143.

MacIntosh, D. L., Spengler, J. D., Ozkaynak, H., Tsai, L., andRyan, P. B. (1996). Dietary exposures to selected metals andpesticides. Environ. Health Perspect. 104, 202}209.

McConnell, E. (1992). Comparative responses in carcinogenesisbioassays as a function of age at 7rst exposure. In ‘‘Similaritiesand Differences between Children and Adults: Implications forRisk Assessment’’ (C. H. P. S. Guzelian and S. S. Olin, Eds.),pp. 66}78. ILSI Press, Washington, DC.

National Research Council. (1987). ‘‘Regulating Pesticides inFood;The Delaney Paradox.’’ Natl. Acad. Sci.}Natl. Res. Coun-cil, Washington, DC.

National Research Council. (1993). ‘‘Pesticides in the Diets ofInfants and Children.’’ Natl. Acad. Sci.}Natl. Res. Council,Washington, DC.

National Research Council. (1999). ‘‘Arsenic in Drinking Water.’’Natl. Acad. Sci.}Natl. Res. Council, Washington, DC.


Noren, K. (1993). Contemporary and retrospective investiga-tions of human milk in the trend studies of organochlorinecontaminants in Sweden. Sci. Total Environ. 139+140,347}355.

Papke, O., Ball, M., and Lis, A. (1994). PCDD/PCDF in humans,a 1993-update of background data. Chemosphere 29, 2355}2360.

Robinson, P. E., Mack, G. A., Remmers, J., Levy, R., andMohadjer, L. (1990). Trends of PCB, hexachlorobenzene, andbeta-benzene hexachloride levels in the adipose tissue of theU.S. population. Environ. Res. 53, 175}192.

Sherer, R. A., and Price, P. S. (1993). The effect of cooking pro-cesses on PCB levels in edible 7sh tissue. Qual. Assur. 2,396}407.

Tao, S., and Bolger, P. (1998). ‘‘Dietary Intakes of Arsenic inthe United States.’’ Presented at the Third InternationalConference on Arsenic Exposure and Health Effects, SanDiego, CA.

Technical Assessment Systems, Inc. (1994). ‘‘Development of aDietary Exposure Potential Matrix.’’ Prepared for U. S. Envir-onmental Protection Agency.

Technical Assessment Systems, Inc. (1995). ‘‘Continuing Surveyof Food Intakes by Individuals 1989}1991 (Summary Tables).’’Prepared for Industrial Economics, Inc.

U.S. Department of Agriculture. (1994). Pesticide Data Program,Summary of 1992 Data, Appendix 4, Agricultural MarketingService.

U.S. Department of Agriculture. (1995a). 1992}1993 PesticideData Program Databases.

U.S. Department of Agriculture. (1995b). Microbiology andResidue Computer Information System 1990}1995, ResidueEvaluation and Planning Division.

U.S. Environmental Protection Agency. (1990). ‘‘InterimMethods for Development of Inhalation Reference Concen-trations.’’ Of7ce of Research and Development, Washington,DC.

U.S. Environmental Protection Agency. (1992). ‘‘National Studyof Chemical Residues in Fish.’’ Of7ce of Science and Techno-logy, Washington, DC.

U.S. Environmental Protection Agency. (1994). ‘‘Estimating Ex-posure to Dioxin-Like Compounds.’’ Volume II: ‘‘Properties,Sources, Occurrence and Background Exposures.’’ Of7ce ofResearch and Development, Washington, DC.

U.S. Environmental Protection Agency, Of7ce of Pesticide Pro-grams. (2000). Science Policy Issues & Guidance Documents.http://www.epa.gov/oppfead1/trac/science/ (accessed June 2,2000).

U.S. Food and Drug Administration. (1994). Total Diet Study,1982}1991; 1992}1994.

Voiland, M. P., Gall, K. L., Lisk, D. J., and MacNeill, D. B. (1991).Effectiveness of recommended fat-trimming procedures on thereduction of PCB and mirex levels in brown trout from LakeOntario. J. Great Lakes Res. 17, 454}460.

Wiles, R., Davies, K., and Campbell, C. (1998). ‘‘Overexposed:Organophosphate Insecticides in Children’s Food.’’ Environ-mental Working Group, Washington, DC.

Winters, D., Cleverly, D., Meier, K., Dupuy, A., Byrne, C.,Deyrup, C., Ellis, R., Ferrario, J., Harless, R., Leese, W., Lor-ber, M., McDaniel, D., Schaum, J., and Walcott, J. (1994). ‘‘AStatistical Survey of Dioxin-Like Compounds in United StatesBeef: A Progress Report.’’ Presented at the Meeting of the 14thAnnual International Symposium on Chlorinated Dioxin andFurans, Kyoto, Japan.

Zabik, M. E., and Zabik, M. J. (1995). Tetrachlorodibenzo-p-dioxinresidue reduction by cooking/processing of 7sh 7llets harvestedfrom the Great Lakes. Bull. Environ. Contam. Toxicol. 55,264}269.

Zabik, M. E., Zabik, M. J., Booren, A. M., Daubenmire, S., Pascall,M. A., Welch, R., and Humphrey, H. (1995). Pesticides and totalpolychlorinated biphenyls residues in raw and cooked walleyeand white bass harvested from the Great Lakes. Bull. Environ.Contam. Toxicol. 54, 396}402.

dietary exposures to food contaminants

Health & Medicine

ingestion of food contaminants

food contamination exposures

immediate food exposures

multiple contaminants

risks of contaminants

introduction food consumption

exposure data

contaminant exposure