envhazards

Environmental Health Risk Assessment

Guidelines for assessing humanhealth risks from

environmental hazards

D e p a r t m e n t o f H e a l t h a n d A g e i n g a n d e n H e a l t h C o u n c i l


Guidelines for assessing humanhealth risks from

environmental hazards

June 2002

enHealth Risk Assess 31/7/02 11:11 am Page A

IBSN: 0 642 82091 0Publications Approval number: 3096

© Commonwealth of Australia 2002

This work is copyright. Apart from any use as permitted under the Copyright Act 1968, no part may bereproduced by any process without prior written permission from the Commonwealth available fromInformation Services. Requests and inquiries concerning reproduction and rights should be addresses tothe Manager, Copyright Services, Information Services, GPO Box 1920, Canberra ACT 2601, or by e-mail [email protected]

Crown copyright material is reproduced with the permission of the Controller of HMSO and theQueen’s Printer for Scotland.

To obtain further copies of this publication you can contact the following:

Department of Health and AgeingPopulation Health DivisionPublication Distribution OfficerMDP 129GPO Box 9848CANBERRA ACT 2601

Email: [email protected]: 1800 020 103

Note the document may be accessed electronically from:

www.health.gov.au/pubhlth/strateg/envhlth/risk/

The enHealth Council, a subcommittee of the National Public Health Partnership, brings together topEnvironmental Health officials at the Federal and State/Territory level along with representation fromthe Australian Institute of Environmental Health, the environment and public health sectors, theIndigenous community and the wider community. The Council has the responsibility for providingnational leadership, implementation of the National Environmental Health Strategy, forgingpartnerships with key players, and the development and coordination of advice on environmental healthmatters at a national level. The advice development process is strongly based on collaboration andconsultation.

DisclaimerThis document has been prepared in good faith exercising due care and attention. However, norepresentation or warranty, expressed or implied, is made as to the relevance, accuracy, completeness orfitness for purpose of this document in respect of any particular user’s circumstances. Users of thisdocument should satisfy themselves concerning its application to, and where necessary, seek expertadvice about, their situation. The Department of Health and Ageing, the enHealth Council and theDepartment of Human Services shall not be liable to the purchaser or any other person or entity withrespect to any liability, loss or damage caused or alleged to have been caused directly or indirectly by thispublication.

The document does not necessarily represent the views of the Department of Health and Ageing, theenHealth Council, the Department of Human Services nor the authors.

Designed by: Looking Glass Press, Unit 2, 10 Kennedy Street, Kingston, ACT, 2604Printed by: Goanna Print Pty Ltd, 151–155 Gladstone Street, Fyshwick, ACT, 2609

enHealth Risk Assess 31/7/02 11:11 am Page B

Guidelines for Assessing Human Health Risks from Environmental Hazards

i

List of figures vii

List of tables vii

Objectives viii

Audience viii

Acknowledgments ix

List of participants x

Summary xi

Abbreviations xvii

Glossary xix

1 Background to Risk Assessment 1

1.1 When to Undertake Risk Assessment 2

1.2 Types of Risk Assessment 2

1.3 Assessing Risk Assessment Methods 3

1.4 Models of Risk Assessment 3

1.5 Bayesian Tools for Risk Assessment 4

1.6 Australian Models of Risk Assessment 4

1.7 Risk Assessment and Health Impact Assessment 7

1.8 Principles of Risk Assessment 7

1.9 Dealing with Uncertainty and Variability 8

1.10 Risk Assessment and the Precautionary Principle 8

1.11 Risk Assessment and Particular Population Groups 9

2 An Australian Framework for Risk Assessment 17

2.1 Context of Risk Assessment 17

2.2 Community Consultation and Involvement 17

2.3 Risk Perception and Risk Communication 19

3 Issue Identification 25

3.1 Introduction 25

3.2 Identification of Environmental Health Hazards 25

3.3 Environmental Sampling and Analysis 26

3.4 Putting the Hazards into their Environmental Health Context 26

Contents


ii

3.5 Identification of Potential Interactions between Agents 26

3.6 Stating why Risk assessment is Needed 27

3.7 Limitations and Uncertainties 28

4 Hazard Assessment— Part 1: Hazard Identification—Toxicology 31

4.1 Introduction 31

4.2 Toxicity Testing—Major in vivo Study Types 32

4.3 Important Issues in Toxicity Testing and Assessment 33

4.4 Assessment of the Quality of the Data Characterising the Hazard 35

4.5 Analysis and Evaluation of Toxicity Studies 36

4.6 Analysis and Evaluation of Major Study Parameters 36

4.7 Evaluation of the Weight-of-Evidence and Consideration of the Toxicology Database in toto 46

4.8 Methods for the Hazard Identification of Carcinogens 47

4.9 The Hazard Identification Report: Structure and Format 48

5 Hazard Assessment—Part 2: Hazard Identification—Epidemiology 51

5.1 Introduction 51

5.2 Bias and Confounding: Key Concepts in Environmental Epidemiology 51

5.3 Types of Epidemiological Study—An Overview 52

5.4 Assessing the Relationship between a Possible Cause and an Outcome 56

5.5 The Strengths and Limitations of Observational Epidemiology versus Experimental Toxicology 59

5.6 Critical Evaluation of Published Research 62

5.7 Evaluation of Meta-analyses 66

5.8 Common Omissions and Errors in Published Research 67

5.9 Undertaking Health Studies 69

5.10 Nature of the Health Study 70

5.11 Ensuring the Quality of a Health Study 72

5.12 Contents of a Health Study Protocol 72


iii

6 Hazard Assessment—Part 3 : Dose–Response Assessment 75

6.1 Introduction 75

6.2 Methodologies 76

6.3 Threshold Approaches 77

6.4 Non-Threshold Approaches 78

6.5 Threshold vs Non-Threshold Approaches 78

6.6 Mechanistically Derived Models 79

6.7 Benchmark Dose Approach 80

6.8 Inter- and Intra-Species Considerations 82

6.9 Mixtures 83

6.10 Hormesis 85

7 Hazard Assessment—Part 4 : Hazard Assessment Reports 87

7.1 Introduction 87

7.2 Sources of Toxicological and Tolerable Intake Data 88

8 Exposure Assessment 91

8.1 Introduction 91

8.2 Issues in Exposure Assessments 91

8.3 Environmental Distribution 92

8.4 Environmental Persistence 93

8.5 Environmental Sampling and Analysis 93

8.6 Meteorological Data 103

8.7 Content of Environmental Sampling and Analysis Reports 103

8.8 Modelling Exposures 107

8.9 Use of Point Estimates and Probability Distributions 107

8.10 Environmental Monitoring 116

8.11 Choice of Tissue 117

8.12 Choice of a Test 118

8.13 Biomarkers 120

8.14 Health Monitoring 121

8.15 Exposure Assessment of Volatile Agents 121


iv

8.16 Default Values for Exposure Assessments 122

8.17 Sources of Exposure Assessment Data 124

8.18 Appraising Exposure Assessments 124

8.19 Exposure Assessment Reports 125

9 Risk Characterisation 127

9.1 Introduction 127

9.2 Key Principles in Environmental Health Risk Characterisation 127

9.3 Quantitative and Qualitative Risk Characterisation 128

9.4 Risk Conclusions 129

9.5 Uncertainty 129

9.6 Exposure Durations and Exceedances of Acceptable Daily Intakes (ADIs) 132

10 Appraisal of Assessments 135


10.2 General Appraisal 135

10.3 Specific Appraisal 136

11 Setting Environmental Health Criteria 141

11.1 Principles for Setting Criteria 141

11.2 Determination of NO(A)ELs, ADIs (RfD) and TDIs for Humans 143

11.3 Determination of Risk-based Environmental Health Criteria 143

Appendices 153

Appendix 1: Environmental Health Risk Assessment for Contaminated Sites 154


1.2 Identifying the Issues 154

Appendix 2: Environmental Health Risk Assessment for Air Pollutants 160



2.3 Hazard Identification 161

2.4 Exposure Defaults 163

2.5 Risk Characterisation 164


v

Appendix 3: Environmental Health Risk Assessment for Food 169




3.4 Chemical Risk Assessment 169

3.5 Microbiological Risk Assessment 169

3.6 Analytical Methodologies 171

3.7 Assessment of Summary Statistic Data 171

3.8 Censored Data 171

3.9 Exposure Assessment 171


Appendix 4: Environmental Health Risk Assessment for Water 173




4.4 Multiple Barrier Approach to Reduce Contamination and Health Risks 179

4.5 Monitoring Methodologies 179

4.6 Assessment of Summary Statistics and Presentation of Data 180

4.7 Censored Data and Levels of Reporting 180

4.8 Dose–Response Assessment 180

4.9 Exposure Assessment 180


Appendix 5: Australian Models of Risk Assessment 183

5.1 Chemical Risk Assessment 183

5.2 Occupational Risk Assessment 184

5.3 Exposure Standards (NOHSC) 188

5.4 Use of Toxicological/Exposure Database 188

5.5 Factors Considered in Standard Setting 188

5.6 Role of the Chemicals Unit of the TGA in Public Health Risk Assessment 189

5.7 Standards Australia Model of Risk Management 193


vi

Appendix 6: International Models of Risk Assessment 194

6.1 Risk Assessment in Canada 194

6.2 Risk Assessment in the United States 194

6.3 Risk Assessment in the United Kingdom 195

6.4 Terminologies Used in Risk Assessment 196

Appendix 7: WHO/IPCS Conceptual Framework for Cancer Risk Assessment 198


7.2 Postulated Mode of Action (Theory of the Case) 198

7.3 Key Events 198

7.4 Dose–Response Relationship 198

7.5 Temporal Association 198

7.6 Strength, Consistency and Specificity of Association of Tumour Response with Key Events 198

7.7 Biological Plausibility and Coherence 199

7.8 Other Modes of Action 199

7.9 Assessment of Postulated Mode of Action 199

7.10 Uncertainties, Inconsistencies and Data Gaps 199

Appendix 8: Microbiological Risk Assessment 200


8.2 Definitions 200

8.3 General Principles 200

8.4 Microbiological Risk Assessment—Paradigms and Frameworks 201

8.5 Microbiological Risk Assessment Bibliography 202

References 204

Index 222

enHealth Council Membership and Terms of Reference 227Membership 227

Terms of Reference 227


vii

List of figuresFigure 1: Risk Assessment Model 5

Figure 2: Relationship of risk assessment and risk management 17

Figure 3: Assessing the relationship between a possible cause and an outcome when an association is observed 56

Figure 4: Hypothetical curve for an animal carcinogenicity study 75

Figure 5: Different dose–response curves for different effects from a hypothetical substance 76

Figure 6: Graphical illustration of the benchmark dose 81

Figure 7: Components of exposure assessment 92

Figure 8: Principles of the Monte Carlo method 108

Figure 9: Decision tree for the development of risk-based environmental health criteria 146

Figure10: Decision tree for cancer risk management 150

List of tablesTable 1: Factors influencing the absorption of chemicals 11

Table 2: Factors influencing the distribution of chemicals in the body 12

Table 3: Factors influencing the rate of metabolism of chemicals 12

Table 4: Factors influencing the elimination of chemicals from the body 12

Table 5: Study designs in environmental epidemiology that use the individual as the unit of analysis 54

Table 6: Applications of different observational study designs 55

Table 7: Advantages and disadvantages of different observational study designs 55

Table 8: Guidelines for the assessment of causation 57

Table 9: Relative ability of different types of study to ‘prove’ causation 58

Table 10: Checklist for evaluating published research 68

Table 11: Models used in risk extrapolation 80

Table 12: Toxic equivalency factors (TEFs) for human and mammalian risk assessment 83

Table 13: Useful vs not useful graphics 101

Table 14: Some key variables for which probability distributions might be needed 113

Table 15: Substances likely to be suitable for biological monitoring 119

Table 16: Example of an uncertainty table for exposure assessment 130


viii

ObjectivesThis document provides a national approach to environmental health risk assessment.

Risk assessments are being undertaken for a wide variety of projects by governments and industry.Environmental health agencies need to be able to assess their content and approach against abenchmark. The document presents a general environmental health risk assessment methodologyapplicable to a range of environmental health hazards. The focus is on chemical hazards in the firstinstance but the core methodology can also be applied to physical (e.g. radiation, noise) andmicrobiological hazards. The core methodology is intended to be able to accommodate specialised‘modules’ that will deal with issues such as physical and microbiological hazards and mixtures as theybecome available. The links to risk management and community consultation/risk communication willbe identified.

Due to the complexity and scale of the environmental health risk assessment process a concise‘cookbook’ is not practicable. Similarly, the situation-specific issues are often sufficiently complex and‘situation-specific’ that a manageable and complete algorithm for decision-making cannot be drafted; thedocument provides a series of guidelines and checklists to assist the decision-making process. Wherepossible, the document is prescriptive about certain aspects of risk assessment. Having specificrequirements for the content of investigations and having them presented in uniform, coherent andlogically developed reports will enable more efficient, accurate, timely and transparent decision-makingand a greater consistency of environmental health decision-making across Australia.

AudienceThe document is primarily intended to be used by environmental health agencies reviewing riskassessments, by people preparing risk assessments for environmental health agencies and by thoseregulatory agencies reviewing risk assessments. It is also intended to be of assistance to a broaderaudience seeking information about processes of environmental risk assessment in Australia.

Risk assessors should have a basic grounding in epidemiology and toxicology.


ix

AcknowledgmentsThe document has been developed from a document prepared for the National Environment ProtectionCouncil ‘Health Risk Assessment Methodology’ which was in turn based on work published in theProceedings of four National Workshops for the Assessment and Management of Contaminated Sites.The assistance and contribution of the National Environment Protection Council Risk AssessmentTaskforce is gratefully acknowledged in relation to the Appendix on the risk assessment of air pollutants.

The development of this document has been made possible by the financial support of theEnvironmental Health Section of the Department of Health and Ageing.

The draft document was prepared by Dr Andrew Langley. Mr Jack Dempsey, Dr Les Davies and Dr Jim Fitzgerald prepared the Hazard Identification section. Dr Roscoe Taylor reviewed and extensivelyupdated the Epidemiology section. Dr Andrea Hinwood and Mr Leo Heiskanen prepared Appendix 2:The Risk Assessment of Air Pollutants. Dr Steven Batt prepared the material on Occupational RiskAssessment. Mr Sam Mangas and Ms Belinda Tassone contributed to the text. Dr David Cunliffe andProfessor Steve Hrudey developed Appendix 3: The Risk Assessment of Water Contaminants.Mr Patrick Le Map developed Appendix 8: Microbiological Risk Assessment. Professor John McNeil,Ms Jane Heyworth, Dr Ted Maynard, Dr Peter Di Marco, Ms Judy Goode, Dr Anne Neller, Dr JimFitzgerald, Mr Len Turczynowicz, Mr Sam Mangas, Mr Geoff Morgan, Dr David Simon, Dr KevinBuckett, Dr Wayne Clapton and Dr Angela McLean have all provided valuable information and advice.Ms Mariann Lloyd-Smith, Dr Darryl Luscombe and Dr Anne Geschke provided considerable assistancewith the community consultation case studies. Ms Bronwyn Landy, Miss Susie Cameron and GlendaAdams provided indispensable assistance in the scientific editing. Sandra Sowerby provided her skills in thepreparation of figures.

Professor Steve Hrudey, Dr Andrew Langley, Dr Hugh Davis and Dr Les Davies provided extensivewritten editorial advice augmenting the comments made by people at the two workshops that reviewedthe document.

Many people provided valuable comment as part of the public consultation process and theircontributions are much appreciated.


x

List of participantsThe following people participated in workshops to review the document:

Dr Steven Batt National Industrial Chemicals Notification and Assessment Scheme

Dr John Beard Northern Rivers Institute of Health and Research, NSW

Dr Kevin Buckett Commonwealth Department of Health and Ageing

Dr Peter Burns Australian Radiation Protection and Nuclear Safety Agency

Mr Phil Callan Office of NHMRC, Commonwealth Department of Health and Ageing

Dr Scott Cameron Department of Human Services, SA

Dr Steve Corbett NSW Health Department

Dr Hugh Davis Health Canada

Mr Jack Dempsey Therapeutic Goods Administration

Dr Peter Di Marco Health Department of Western Australia

Dr Roger Drew SHE Pacific

Dr Jim Fitzgerald Department of Human Services, SA

Dr Ann Geschke Department of Human Services, Victoria

Mr Leo Heiskanen Commonwealth Department of Health and Ageing

Dr Andrea Hinwood Department of Environmental Protection, WA

Professor Steve Hrudey University of Alberta, Canada

Dr Mark Jacobs Department of Health and Human Services, Tasmania

Dr Bruce Kennedy National Environment Protection Council Service Corporation

Dr Andrew Langley Department of Human Services, SA

Ms Mariann Lloyd-Smith National Toxics Network

Dr Darryl Luscombe Greenpeace Australia

Professor John McNeil Monash University, Victoria

Mr Sam Mangas Department of Human Services, SA

Professor Michael Moore National Research Centre for Environmental Toxicology

Mr Geoff Morgan NSW Health Department

Dr Anne Neller University of the Sunshine Coast, Queensland

Dr Gerard Neville Queensland Health

Dr Brian Priestly Therapeutic Goods Administration

Dr Deborah Read Environmental Risk Management Authority, NZ

Dr Roscoe Taylor Queensland Health

Dr Luba Tomaska Australia New Zealand Food Authority


xi

Summary

Risk AssessmentIntuitive risk assessment and risk management have been fundamental for human survival and evolution.Those who appreciated risk were more likely to survive and reproduce whereas those who could notwere more likely to perish from environmental hazards (Thomas and Hrudey, 1997).

A specific quantitative concept of ‘risk’ has only been appreciated in relatively recent times. Estimates ofrisk based on probability only developed in the late seventeenth century.

In considering risk as either a prediction or expectation, there are various elements:

• a hazard (the source of danger);

• an uncertainty of occurrence and outcomes (expressed by probability distributions);

• possible adverse health outcomes;

• a target;

• a time frame; and

• the importance of the risk for people affected by it (Thomas and Hrudey, 1997).

Risk assessment provides a systematic approach for characterising the nature and magnitude of the risksassociated with environmental health hazards. All activities, processes and products have some degree ofrisk. The ultimate aim of risk assessment is to provide the best possible scientific, social and practicalinformation about the risks, so that these can be discussed more broadly and the best decisions made asto what to do about them.

Risk assessment has been used in various forms for many years in Australia although it may not alwayshave been called ‘risk assessment’.

The use of risk assessment as a tool in the decision-making process has become increasingly importantover the last two decades as it has become evident that situations cannot be judged simply as either ‘safe’or ‘unsafe’.

Risk assessment takes into account factors relevant to the situation such as the current or proposedhuman activities, physico-chemical and bioavailability characteristics of chemical hazard(s), the infectivedoses of microbiological agents, and the opportunities for exposure to the agent.

Generic risk assessments and assessments of potential risk may be made in situations such asdetermining environmental standards for additives or contaminants in soil, air, water and food or todetermine whether particular products such as pesticides and pharmaceuticals can be used. Situation-specific risk assessments can be undertaken where there is an actual or potential environmental hazardsuch as contaminated land or industrial emissions from a proposed factory and should take into accountfactors relevant to those particular circumstances.

Health risk assessment is intended ‘to provide complete information to risk managers, specificallypolicymakers and regulators, so that the best possible decisions are made’ (Paustenbach, 1989, p. 28).Good risk assessment is dependent upon a high degree of scientific skill and objectivity and should bedistinguished from the risk management process which selects options in response to health riskassessments and which incorporates ‘scientific, social, economic and political information’ and which


xii

‘requires value judgements e.g. on the tolerability of risk and reasonableness of costs’(ANZECC/NHMRC 1992, p. iii). Risk assessment should provide a ‘credible, objective, realistic andbalanced analysis’ (US EPA, 1992).

Risk assessment gathers and organises information and may enable:

• risks at a point in time (including baseline risks) and changes in risk over time to be estimated andallows judgement as to whether action is necessary;

• Health Guidance Values to be established for public health hazards;

• assessments of new or exotic risks;

• a comparison of the potential health impacts of various environmental health interventions (thus enabling cost-effectiveness estimates);

• the identification and comparison of different factors that affect the nature and magnitude of the risk;

• risk-based standards setting for regulatory exposure limits and clean-up standards;

• prioritising issues according to their levels of risks;

• questionable theories, methods and data to be challenged and addressed by providing a clearlydocumented and open process (Covello and Merkhofer, 1993);

• better appreciation of the tradeoffs and opportunity costs which occur when addressing one sourceof risk;

• consistent, transparent appraisal and recording of public health risks; and

• risk based policy making.

Risk assessment may not always provide a compelling or definitive outcome and will often be limited bythe data available.

Risk assessors should appreciate that the community may see risk assessment as an excuse for pollutingbehaviours.

A preliminary situation-specific risk assessment can be undertaken by choosing to apply environmentalhealth criteria, which are derived using risk assessment techniques and can be applied generically to arange of situations. Where the level of a hazard exceeds the risk-based environmental health criteria,more detailed situation-specific health risk assessment may be used to determine the nature of actionrequired to address the risks. The action may range from informing the community to requiring large-scale remediation measures.

While this document is about environmental health risk assessment, it recognises that environmentalhealth risk assessment is complemented by the process of ecological risk assessment.


xiii

Environmental Health Risk Assessment MethodologyThere are several models of risk assessments and various definitions for the relevant terms. Thisdocument uses a model developed by and for environmental health agencies and uses a modelcompatible with WHO models. It is comprised of:

• issue identification;

• hazard identification;

• dose–response assessment;

• exposure assessment for the relevant population; and

• risk characterisation.

These five stages are linked and dependent on the preceding stages when they are part of a riskassessment.

Issue Identification identifies issues for which risk assessment is useful and establishes a context for therisk assessment by a process of identifying the concerns that the risk assessment needs to address. Thedetermination of the ‘problems’ is necessary to establish a context for the risk assessment and assists theprocess of risk management.

Issue Identification involves determining:

• what is causing the identified problem;

• why the problem is a problem;

• how the problem was initially identified;

• what types of (adverse) health effects might be caused by the problem;

• how quickly and for what duration the problems might be experienced; and

• what the public perceptions of the hazard are (Health Canada, 1999, p. 12).

Hazard Assessment comprises Hazard Identification and Dose–response Assessment

Hazard Identification involves determining:

• what types of (adverse) health effects might be caused by the problem; and

• how quickly the problems might be experienced. (Health Canada, 1999, p. 12)

Dose–response Assessment evaluates both qualitative and quantitative toxicity information to estimate‘the incidence of adverse effects occurring in humans at different exposure levels’. (US EPA, 1989, p. 1.6)Where available, human and animal evidence will be assessed as part of this process.

Exposure Assessment involves the determination of the frequency, magnitude, extent, duration andcharacter of exposures to a hazard. Estimates can be made for past, present and likely future exposures.There is also the identification of exposed populations and particularly sensitive sub-populations, andpotential exposure pathways. Environmental monitoring and predictive models can be used todetermine the levels of exposure at particular points on the exposure pathways. The contaminant intakesfrom the various pathways under a range of scenarios can then be estimated (US EPA, 1989).


xiv

Given this information, Risk Characterisation details the nature and potential incidence of effects forthe exposure conditions described in the exposure assessment. An integral part of this stage is toevaluate the uncertainties and assumptions in the risk assessment process. The nature and magnitude ofthe uncertainties should be clearly detailed so that they can be taken into account in the riskmanagement of a situation. The uncertainties may be addressed by gathering further information, or byincorporating safety factors.

The process of risk assessment should enable consistent application of methodology to be made by thespecialists undertaking the process. Expert professional judgement can be an integral part of theprocess. Situation-specific risk assessments should not lead to significant variations in the estimatedrisks of similar situations (Langley and El Saadi, 1991).

The situation-specific process is a multi-disciplinary task and requires considerable expertise. Peopleinvolved in specific components of the environmental health risk assessment process should beadequately qualified and experienced and have a broad understanding of health risk assessment andmanagement and the practical realities of environmental health practice. Professional skills that may beused include environmental health, engineering, history, chemistry, planning, statistics, occupationalhygiene, occupational and public health medicine, toxicology, health science, communication, sociology,psychology, economics and epidemiology. While it is unlikely that one person will have the breadth ofskill to undertake all components of the health risk assessment, there must be a single personcoordinating and taking responsibility for the assessment.

In many instances, situation-specific health risk assessments may not be necessary as the nature andmagnitude of the risks will be quite apparent, there may be no population at risk, or decisions on riskmanagement may be made on other grounds. In such cases, the significant resources required for a detailedrisk assessment would be better directed to risk management steps (ANZECC/NHMRC, 1992, p. 20).

The level of risk can be described either qualitatively (i.e. by putting risks into categories such as ‘high’,‘medium’ or ‘low’) or quantitatively (with a numerical estimate or probability density distribution).Current risk assessment methods do not enable accurate quantitative estimates of risk for low levels ofexposure to environmental hazards. Numerical estimates of risk will rarely be feasible because oflimitations in toxicological and exposure data which will be reflected in the uncertainty assessment, butquantification may be possible for some components such as exposure assessment. Clearly definedqualitative categories can enable reliable and effective risk management decisions.

It should be recognised that, as a consequence of testing limitations (for example, not every square metreof a contaminated site nor every item of food in the marketplace will be tested), situation-specific healthrisk assessment is a screening process where there may be low rates of false negatives and false positives.‘Risk assessment is based on probabilities rather than absolutes and this should be reflected in decision-making’ (ibid, p. 34). Uncertainty is usually caused by inadequate knowledge but can also relate to:

• parameter uncertainty (measurement errors, random errors, systematic errors);

• multiple uncertainty (errors arising from the incorrect models or reality); and

• decision–rule uncertainty (not knowing how to interpret predictions) (Finkel, 1990)

Variability occurs when a single number is used to describe something that actually has multiple orvariable values such as bodyweight or susceptibility to adverse affects, or something that varies over timesuch as the population of an area. Variability occurs as a result of differences between the characteristicsof different people or populations. Uncertainty arises as a result of lack of data. Both uncertainty andvariability need to be considered in risk assessments.


xv

Key Principles in Environmental Health Risk AssessmentThere are a number of key principles for environmental health risk characterisation (EPA NSW, 1998;US EPA, 1995):

1. Actions should always adequately protect public health and the environment, putting theseresponsibilities before all other considerations.

2. Risk assessments should be transparent. The nature and use of default values and methods,assumptions and policy judgements in the risk assessment should be clearly identified. Conclusionsdrawn from the evidence should be separated from policy judgements.

3. Risk characterisations should include a summary of the key issues and conclusions of each of theother components of the risk assessment, as well as describing the likelihood of adverse healtheffects. The summary should include a description of the overall strengths and limitations(including uncertainties) of the assessment and conclusions.

4. Risk characterisations (and risk assessments) should be consistent in general format, but recognisethe unique characteristics of each specific situation

5. Health risk assessment must be undertaken with an appreciation that the health risk assessment isoften part of a larger assessment that encompasses ecological risk assessment.

6. To protect public health and the environment an appropriate degree of conservatism must beadopted to guard against uncertainties.

7. Ensure that comparisons have been made against environmental health criteria that have beenendorsed by the relevant Commonwealth, State or Territory environmental health agencies.

8. Where there are no Environmental Health Criteria for a particular agent refer to the administrativeauthority at the relevant Commonwealth, State or Territory level.

9. Ensure that human health risk assessments are undertaken, where necessary, according to methodsin this document, or its revisions as published from time to time

10. When deriving environmental health criteria use toxicological data or exposure criteria fromagencies or organisations relevant to the State or Territory (e.g. local or Commonwealth healthagencies such as NHMRC, or the enHealth Council) or to which Australia is party (e.g. WorldHealth Organization). (See Toxicity Assessment Section 5.4)

11. Ensure that human health risk assessments are undertaken using national toxicological assessments(e.g. NHMRC) or WHO assessments or, where neither has been made, methods agreed to by theadministrative authority for contaminated sites at the relevant Commonwealth, State or Territorylevel.

12. The risk assessor's knowledge of the peer-reviewed scientific literature relevant to risk assessmentand the practical aspects of risk assessment should be up-to-date.

13. Variations in risk assessments as a result of particular statutory requirements, resource limitations,and other specific factors should be explained as part of the risk characterisation. For example, areason will be required to explain why certain elements are incomplete.


xvi

Risk Management, Risk Communication and Community ConsultationRisk assessment may lead into risk management. Risk management is about a broader evaluation of theresults of the risk assessment and takes into account not only scientific data, but also social, economicand political considerations. It is important that the basis of decision making is clearly documented.

Public involvement should be an inherent part of risk assessment and management not only because theyhave a right to know but also because they have local knowledge such as sources of exposure, patterns ofbehaviour and local concerns that may be missed by generic risk assessments and models. The nature andextent of the risk must therefore be communicated in terms understandable by all parties.

Risk communication about the estimated risks is an essential process that should be incorporated beforeand throughout risk assessment and management. It is the deliberate exchange of information about thenature, severity, or acceptability of risks and the decisions taken to combat them.

Risk communication should be seen as a process that enables all stakeholders to make an informedjudgement about a risk and its management. The process must involve a frank and open presentation ofall relevant background information to the stakeholders, in a manner understandable by all. This processalso involves listening to stakeholders. There are three perspectives to ‘risk’: actual, estimated andperceived (McKone and Bogen, 1991). The estimated risk is the outcome of the risk assessment withits uncertainties. The actual level may never be known because there may not be instruments availableto ‘measure’ it or because actions will change the course of events. All stakeholders will have their ownperceptions of the risk. Good risk assessment and risk communication minimise the mismatch betweenthese three perspectives of risk and assist in efficient risk management. They will also address disquietand will highlight significant risks where they may not be apparent to the stakeholders.


xvii

AbbreviationsACDP Advisory Committee on Dangerous Pathogens

ADI Acceptable Daily Intake (WHO)

ANZECC Australia and New Zealand Environment and Conservation Council

ASCEPT Australasian Society of Clinical and Experimental Pharmacologists and Toxicologists

BMD Benchmark Dose

BMDL Lower confidence limit on BMD

BMR Benchmark Risk (Response)

DOH Department of Health (United Kingdom)

DNA Deoxyribonucleic acid

EA Environment Australia

ECETOC European Centre for Ecotoxicology and Toxicology of Chemicals

FDA Food and Drug Administration (USA)

GCP Good Clinical Practice

GLP Good Laboratory Practice

HSE Health and Safety Executive (United Kingdom)

IARC International Agency for Research on Cancer

ICRP International Commission on Radiological Protection

IPCS International Programme on Chemical Safety

IRIS Integrated Risk Information System

JECFA Joint FAO/WHO Expert Committee on Food Additives

JMPR Joint FAO/WHO Meeting on Pesticide Residues

LED Lowest Effective Dose

LOAEL Lowest Observed Adverse Effect Level

MAC Maximum Allowable Concentration

mg/kg bw/d mg/kg bodyweight/day

MTD Maximum Tolerated Dose

MRL Maximum Residue Limit

MRT Maximum Residue Tolerance

NEPC National Environment Protection Council (Australia)

NHMRC National Health and Medical Research Council (Australia)

NICNAS National Industrial Chemicals Notification and Assessment Scheme (Australia)

NOAEL No Observed Adverse Effect Level


xviii

NOEL No Observed Effect Level

NOHSC National Occupational Health and Safety Commission (Australia)

NTP National Toxicology Program (USA)

OECD Organisation for Economic Co-operation and Development

PCB Polychlorinated biphenyl

PM10 Particulate Matter 10µ

PTWI Provisional Tolerable Weekly Intake (WHO)

q1* The 95 per cent upper confidence limit of the slope estimate used for the linearised multi-stage model

QRA Quantitative Risk Assessment

RfD Reference Dose (US EPA)

SF Safety Factors

SAR Structure Activity Relationship

TDI Tolerable Daily Intake (WHO)

TWP Technical Working Party

US EPA United States Environmental Protection Agency

WHO World Health Organization

Glossary(Adapted from NHMRC 1997)

Absorbed dose The amount of chemical that, after contact with the exchange boundary (skin, lungs,gut), actually penetrates the exchange boundary and enters the circulatory system.The amount may be the same or less than the applied dose.

ADI Acceptable Daily Intake. The daily intake of a chemical which, during a lifetime,appears to be without appreciable risk, on the basis of all the facts known at the time.It is expressed in milligrams per kilogram of body weight per day (mg/kg/day).(WHO, 1989) For this purpose, ‘without appreciable risk’ is taken to mean thatadverse effects will not result even after a lifetime of exposure. Furthermore, for apesticide residue, the acceptable daily intake is intended to give a guide to themaximum amount that can be taken daily in the food without appreciable risk to theconsumer. Accordingly, the figure is derived as far as possible from feeding studies inanimals. (See also ‘Tolerable Daily Intake’ and ‘Reference Dose’)

Adverse effect The change in morphology, physiology, growth, development or life span of anorganism which results in impairment of functional capacity or impairment ofcapacity to compensate for additional stress or increase in susceptibility to theharmful effects of other environmental influences. Some adaptive changes are notgenerally considered to be adverse e.g. some changes in enzyme levels.

Adduct A chemical moiety which is covalently bound to a large molecule such as DNA orprotein. (DOH, 1991)

Agent Any chemical, physical, biological or social substance or factor being assessed, unlessotherwise noted.

Applied dose Amount of an agent presented to an absorption barrier and available for absorption.The amount may be the same or more than the absorbed dose.

Bias A process resulting in a tendency to produce results that differ in a systematic valuefrom the true values. Also known as systematic error. (Beaglehole et al, 1993)

BMD Benchmark Dose. The dose associated with a given incidence (e.g. 1 per cent, 5 per centor 10 per cent incidence) of effect, the Benchmark Risk, based on the best-fittingdose–response curve.

BMR Benchmark Risk. A predetermined incidence of adverse response that determines theBenchmark dose.

Background ‘naturally-occurring, ambient concentrations in the local area of a site’concentration (ANZECC/NHMRC, 1997)

Bioavailability The ratio of the systemic dose to the applied dose.

Biological Measurement of a contaminant or metabolite in body tissue, fluid, blood, expired air,monitoring breast milk and sweat. It is usually used as a marker or indicator of exposure to

environmental chemicals.

Biomarker Any measurement reflecting an interaction between a biological system and anenvironmental agent, which may be chemical, physical or biological (WHO, 1993).Often used to describe measurements used in biological monitoring.


xix


Carcinogen Chemical, biological or physical cancer-causing agent.

Carcinogenesis The origin, causation and development of tumours. The term applies to all forms oftumours (e.g. benign and malignant).

Carcinogenicity The ability to produce tumours, which may be benign or malignant. (IEH, 1999b)

Chronic toxicity The ability to produce an adverse effect which persists over a long period of time,whether or not it occurs immediately upon exposure to a chemical or is delayed, or aneffect which is only induced by prolonged exposure to a chemical. (IEH, 1999b)

Confidence Weight assigned by the evaluator to the quality of the information available (high,medium or low confidence) to indicate that a chemical possesses certain toxicologicalproperties.

Confidence A range of values determined by the degree of presumed random variability in a set limits of data, within which the value of a parameter, e.g. the mean, lies, with a specified

level of confidence or probability (e.g. 95 per cent). The confidence limit refers to the upper or lower value of the range. (DOH, 1991)

Confounding A factor that distorts the apparent effect or magnitude of the effect of a study factor factor or risk. Such factors must be controlled for in order to obtain an undistorted

estimate of a given effect. (DOH, 1991)

Critical effect(s) The adverse effect judged to be the most important for setting an acceptable humanintake or exposure. It is usually the most sensitive adverse effect, i.e. that with thelowest effect level, or sometimes a more severe effect, not necessarily having thelowest effect level. (IEH, 1999b)

Default value A pragmatic, fixed or standard value used in the absence of relevant data.

Dermal Of the skin, through or by the skin.

Developmental The ability to produce an adverse effect in embryo, fetus or immature organism, whichtoxicity is induced and/or manifest either prenatally or postnatally before sexual maturity.

(IEH, 1999b)

Dose A stated quantity or concentration of a substance to which an organism is exposed overa continuous or intermittent duration of exposure. It is most commonly expressed as theamount of test substance per unit weight of test animal (e.g. mg/kg body weight).

The applied dose is the amount of chemical in contact with the primary absorptionboundaries (e.g. skin, lungs, and gastrointestinal tract) and available for absorptionThe absorbed dose is the amount crossing a specific absorption barrier (e.g. theexchange boundaries of skin, lung, and digestive tract) through uptake processes. Theamount of the chemical available for interaction by any particular organ or cell istermed the delivered dose of that organ or cell. (EPA, 1992, p. 22933). The systemicdose is the dose to which the whole, or extensive parts, of the body is exposed. Theabsorbed dose may not be the systemic dose as substances absorbed in the digestivetract may be removed by the liver and not enter the systemic circulation.

Dosage A general term comprising the dose, its frequency and the duration of dosing. Dosageis properly applied to any rate or ratio involving a dose. Dosages often involve thedimension of time (e.g. mg/kg/day), but the meaning is not restricted to thisrelationship. (Hayes, 1991)

xx

Dose–response Determination of the relationship between the magnitude of the dose or level of assessment exposure to a chemical and the incidence or severity of the associated adverse effect.

(IEH, 1999b)

Dose–response The correlative association existing between the dose administered and the response relationship (effect) or spectrum of responses that is obtained. The concept expressed by this term

is indispensable to the identification, evaluation, and interpretation of most pharmacological and toxicological responses to chemicals. The basic assumptions which underlie and support the concept are: (a) the observed response is a function of the concentration at a site, (b) the concentration at a site is a function of the dose,and (c) response and dose are causally related (Eaton and Klaassen, 1996). The existence of a dose–response relationship for a particular biological or toxicological response (effect) provides a defensible conclusion that the response is a result of exposure to a known substance.

Endpoint An observable or measurable biological event used as an indicator of the effect of achemical on a biological system (cell, organism, organ etc.).

Environmental Those aspects of human health determined by physical, chemical, biological and health social factors in the environment. Environmental health practice covers the

assessment, correction, control and prevention of environmental factors that can adversely affect health, as well as the enhancement of those aspects of the environment that can improve human health.

Environmental The monitoring of the concentration of substances in the physical environment of air,monitoring water, soil and food.

Epidemiology The study of the distribution and determinants of health-related states or events inspecified populations, and the application of the study to the control of healthproblems (Last, 1988)

Expert An expert has (1) training and experience in the subject area resulting in superiorknowledge in the field (2) access to relevant information, (3) an ability to process andeffectively use the information, and (4) is recognised by his or her peers or thoseconducting the study as qualified to provide judgements about assumptions, models,and model parameters at the level of detail required. (NCRP, 1996).

Exposure Contact of a chemical, physical or biological agent with the outer boundary of anorganism, e.g. inhalation, ingestion or dermal contact.

Exposure The estimation (qualitative or quantitative) of the magnitude, frequency, duration,assessment route and extent (for example, number of organisms) of exposure to one or more

contaminated media for the general population, for different subgroups of the population, or for individuals.

Exposure The course a chemical or physical agent takes from a source to an exposed organism.pathway An exposure pathway describes a unique mechanism by which an individual or

population is exposed to chemicals or physical agents at or originating from a site.Each exposure pathway includes a source or release from a source, an exposure point,and an exposure route. If the exposure point differs from the source, a transport/exposure medium e.g. air or media (in cases of inter-media transfer) also is indicated. (US EPA, 1989, p. 62)


xxi


Exposure route The way a chemical enters an organism after contact e.g. by ingestion, inhalation, ordermal absorption. (EPA, 1992, p. 22933)

Extrapolation For dose–response curves, an estimate of the response at a point outside the range ofthe experimental data. Also refers to the estimation of a response in different speciesor by different routes than that used in the experimental study of interest.

Factor A single factor or product of several single factors used to derive an acceptable intake.These factors account for adequacy of the study, interspecies extrapolation, inter-individual variability in humans, adequacy of the overall data base, nature and extentof toxicity, public health regulatory concern and scientific uncertainty.

False negative A result that is erroneously negative.

False positive A result that is erroneously positive.

Gene The DNA molecule of inheritance of characteristics including susceptibility todisease.

Genotoxic Agents for which a direct activity is the alteration of the information encoded ingenetic material. (Butterworth, 1990)

Genotoxic A chemical which induces tumours via a mechanism involving direct damage to carcinogen DNA. (IEH, 1999b)

Genotoxicity A broad term describing the ability to produce damage to the genetic material(DNA) of cells or organisms.

Guidance values Values such as concentrations in air or water, which are derived after appropriateallocation of Tolerable Intake (TI) among the possible different media of exposure.Combined exposure from all media at the guidance values over a lifetime would beexpected to be without appreciable health risk. The aim of a guidance value is toprovide quantitative information from risk assessment for risk managers to enable themto make decisions concerning the protection of human health. (WHO, 1994, p. 16)

Guideline dose The average daily intake of a chemical, which, during a lifetime, is unlikely to result incancer, based on a comprehensive expert assessment of the best information availableat the time. The guideline dose is derived by regulatory authorities using cancer riskassessment according to guidelines developed by national health advisory bodies.

Hazard The capacity of an agent to produce a particular type of adverse health orenvironmental effect, e.g. one hazard associated with benzene is that it can causeleukemia; or

The disposition of a thing, a condition or a situation to produce an adverse health orenvironmental effect; or an event, sequence of events or combination of circumstancesthat could potentially have adverse consequences (adapted from ACDP, 1996).

Hazard The identification, from animal and human studies, in vitro studies and structure-identification activity relationships, of adverse health effects associated with exposure to an agent.

(IEH, 1999b)

Health Health is a state of complete physical, mental and social well being and not merelythe absence of disease or infirmity (WHO, 1946).

xxii

Health The concentration of a soil contaminant (arrived at using appropriate sampling,investigation analytical and data interpretation techniques) above which further appropriate level investigation and evaluation will be required to determine whether a significant

health risk exists

Health risk The process of estimating the potential impact of a chemical, biological, physical or assessment social agent on a specified human population system under a specific set of conditions

and for a certain timeframe.

Health risk The process of evaluating alternative actions, selecting options and implementing management them in response to health risk assessments. The decision making will incorporate

scientific, technological, social, economic and political information. The process requires value judgements, e.g. on the tolerability and reasonableness of costs.

Hormesis Demonstrated beneficial effects of an agent at low (but not homeopathic) doses butwith toxicity occurring at higher doses.

Immunotoxicity The ability to produce an adverse effect on the functioning of organs and cellsinvolved in immune competence. (IEH, 1999b)

IRIS Integrated Risk Information System. The computerised database of the US EPA,which provides the Agency's adopted hazard and dose–response assessment forchemical and radiological agents. Used as guidance and to provide consistency in theAgency’s regulatory decisions designed to reduce risk related to environmentalexposures (see abbreviations).

LD50 The quantity of a chemical compound that, when applied directly to test organisms,via inhalation, oral or dermal exposure is estimated to be fatal to 50 per cent of thoseorganisms under the stated conditions of the test.

Number of microorganisms of a particular species that are fatal in 50 per cent of thehost organisms.

LED10 Lowest Effective Dose. The lower 95 per cent confidence limit on a dose associatedwith an estimated 10 per cent increased tumour or relevant non-tumour response.(US EPA, 1996)

LOEL Lowest Observed Effect Level. The lowest concentration or amount of a substance,found by experiment or observation, that causes alterations of morphology, functionalcapacity, growth, development or life span of target organisms.

WHO (1990) define it as the lowest dose of a substance which causes changesdistinguishable from those observed in normal (control) animals.

LOAEL Lowest Observed Adverse Effect Level. The lowest concentration or amount of asubstance, found by experiment or observation, that causes adverse alterations ofmorphology, functional capacity, growth, development or life span of targetorganisms.

Level of The minimum concentration or mass of analyte that can be detected at a known detection confidence level

Level of The value calculated from the instrumentation detection limits and with appropriate reporting scale up factors applied. The scale up factors are affected by the procedures, methods

and the size of the sample.


xxiii


Lifetime Covering the average life span of an organism (e.g. 70 years for humans).

Metabolite A substance that is the product of biochemical alteration of the parent compound inan organism.

Model A mathematical representation of a biological system intended to mimic thebehaviour of the real system, allowing description about empirical data andpredictions about untested states of the system.

Mutagenicity The ability to produce a permanent, heritable change in the amount or structure ofgenetic material of cells or organisms. (IEH, 1999b)

Neurotoxicity The ability to produce an adverse effect in the central or peripheral nervous system.(IEH, 1999b)

NOAEL The No Observed Adverse Effect Level is the highest dose of a substance at whichno toxic (i.e. adverse) effects are observed. (WHO, 1990) It may also be worded inmore detail thus: The NOAEL is defined as the highest exposure at which there is nostatistically- or biologically-significant increase in the frequency of an adverse effectwhen compared to a control group. (National Academy of Sciences/NationalResearch Council, 1994) The definition of NOEL is equivalent, but with theremoval of the term, ‘adverse’. Often, the difficult issue in the use of the termsNOEL or NOAEL is in deciding whether a compound-related effect noted in aparticular study is necessarily an ‘adverse’ effect. Alterations of morphology, functionalcapacity, growth, development or life span of the target organism may be detectedwhich are judged not to be adverse.

Nongenotoxic A chemical which induces tumours via a mechanism which does not involve direct carcinogen damage to DNA (IEH, 1999b).

Physiologically- Modelling the dose or degree of exposure to a chemical at a target tissue, cell or based pharmaco- receptor, by integration of pharmacokinetic data with anatomical, physiological andkinetics (PBPK) biochemical data (IEH, 1999b).

NOEL The ‘No Observed Effect Level’ or ‘No Observable Effect Level’ (NOEL) is thehighest dose of a substance administered to a group of experimental animals at whichthere is an absence of observable effects on morphology, functional capacity, growth,development or life span, which are observed or measured at higher dose levels usedin the study. Thus, dosing animals at the NOEL should not produce any biologicallysignificant differences between the group of chemically exposed animals and anunexposed control group of animals maintained under identical conditions. TheNOEL is expressed in milligrams of chemical per kilogram of body weight per day(mg/kg bw/day) or, in a feeding study, in ppm in food (converted to mg/kg bw ofcompound intake by measured or estimated food intake over the period of the study)

The NOEL has been simply defined as the highest dose of a substance which causes nochanges distinguishable from those observed in normal (control) animals (WHO, 1990).

PM10 Particulate Matter 10µm: the fraction of particles passing an inlet with a 50 per centcut-off efficiency at an aerodynamic diameter of 10µm

xxiv

PTWI Provisional Tolerable Weekly Intake. The tolerable intake of a chemical expressed as aweekly amount. The term was established by WHO (1972) for several heavy metalswhich ‘are able to accumulate within the body at a rate and to an extent determinedby the level of intake and by the chemical form of the heavy metal present in food.’(WHO, 1989)

Public health The science and art of preventing disease, prolonging life and promoting healththrough the organised efforts of society.

Reproductive The ability to produce an adverse effect on any aspect of reproductive capacity,toxicity function or outcome. It includes effects on the embryo, fetus, neonate and prepubertal

organism and on adult reproductive and neuroendocrine systems (IEH 1999b).

RfD Reference Dose (RfD). An estimate (with uncertainty factors spanning perhaps anorder of magnitude) of the daily exposure (mg/kg/day) to the general humanpopulation (including sensitive sub-groups) that is likely to be without an appreciablerisk of deleterious effects during a life time of exposure. It is derived from theNOAEL or the LOAEL by application of uncertainty factors that reflect varioustypes of data used to estimate RfD and an additional modifying factor, which is basedon professional judgement of the entire data base of the chemical. (IRIS, 1996).Usually doses less than the RfD are not likely to be associated with adverse healthrisks, and are therefore less likely to be of regulatory concern. As the frequencyand/or magnitude of the exposures exceeding the RfD increase, the probability ofadverse effects in a human population increases. However, all doses below the RfDare not assumed to be ‘acceptable’ (or risk-free) and nor are all doses that exceed theRfD necessarily ‘unacceptable’ (i.e. result in adverse effects) (US EPA)

Risk The probability that, in a certain timeframe, an adverse outcome will occur in aperson, group of people, plants, animals and/or the ecology of a specified area that isexposed to a particular dose or concentration of a hazardous agent, i.e. it depends onboth the level of toxicity of the agent and the level of exposure.

Risk assessment The process of estimating the potential impact of a chemical, physical,microbiological or psychosocial hazard on a specified human population or ecologicalsystem under a specific set of conditions and for a certain timeframe

Risk An interactive process involving the exchange among individuals, groups and communication institutions of information and expert opinion about the nature, severity, and

acceptability of risks and the decisions taken to combat them.

Risk The process of evaluating alternative actions, selecting options and implementing management them in response to risk assessments. The decision making will incorporate scientific,

technological, social, economic and political information. The process requires value judgements, e.g. on the tolerability and reasonableness of costs.

Safety factor See factor. Safety factor usually refers to health-related concerns.

Skin irritancy A local inflammatory reaction affecting the skin

Stochastic A random probabilistic phenomenon

Structure activity The relationship between the biological activity of a chemical or series of chemicalsrelationship and their structure. The relationships can be described qualitatively and quantitatively.


xxv


Systemic dose Amount of a substance that is eventually distributed in the blood after absorption,distribution and storage.

Teratogenicity The ability to produce a structural malformation or defect in an embryo or fetus(IEH, 1999b)

Threshold The lowest dose or exposure level which will produce a toxic effect and below whichno toxicity is observed (IEH, 1999b).

Threshold dose The lowest dose which produces an effect and below which no biological effect occurs.The acceptability and usefulness of the concept of the experimental NOEL/NOAELdepends on the scientific rationale supporting the existence and demonstrability of athreshold for responses produced by biologically active agents. As used here, the term‘threshold’ designates that level of a stimulus which comes just within the limits ofperception, and below which level a recognisable response is not elicited.

TDI Tolerable Daily Intake. An estimate of the intake of a substance which can occur overa lifetime without appreciable health risk. It may have different units depending onthe route of administration. (WHO, 1994). (Imray and Langley, 1996, p. 18). Theterm, ‘acceptable daily intake’ is used for chemicals such as pesticides (herbicides,insecticides, antifungals etc.) which are deliberately used on food crops or food-producing animals and for which some level of residues may be expected to occur infood. The term ‘tolerable daily intake’ is used when the chemical is a potential food orenvironmental contaminant. Whilst exposure should not occur, a TDI is anestablished health limit below which lifetime exposure should not have any adversehealth effects. (See also ‘Acceptable Daily Intake’ and ‘Reference Dose’)

TWI Tolerable Weekly Intake. The TI expressed as a weekly amount

Tolerable intake An estimate of the intake of a substance that over a lifetime is without appreciablehealth risk. (WHO, 1994). Examples are the ADI, TDI and Reference Dose.

Toxicity The quality or degree of being poisonous or harmful to plant, animal or human life.

Transformation The process by which a normal cell acquires the capacity for neoplastic orcarcinogenic growth. It is thought to occur in several stages.

Tumour A mass of abnormal, disorganised cells, arising from pre-existing tissue, which ischaracterised by excessive and uncoordinated cell proliferation or growth and byabnormal differentiation (specialisation). There are two types of tumours, benign andmalignant. Benign tumours morphologically resemble their tissue of origin, growslowly (may also stop growing) and form encapsulated masses; they do not infiltrateother tissues, they do not metastasise and are rarely fatal. Malignant tumoursresemble their parent tissue less closely and are composed of increasingly abnormalcells genetically, morphologically and functionally. Most grow rapidly, spreadprogressively through adjacent tissues and metastasise to distant tissues.

Tumour The first step in carcinogenesis whereby a small number of cells (or one cell) are initiation irreversibly changed due to genetic damage.

Tumour The stage in carcinogenesis when tumours acquire the features of malignant growth.progression

xxvi

Tumour The process by which initiated cells undergo clonal expansion to form overt tumours.promotion

Uncertainty The lack of knowledge about the correct value, e.g. a specific exposure measure orestimate.

Uncertainty A numerical factor applied to the no-effect level to derive an exposure level factor considered to be without appreciable risk to health (the NEL is divided by the

uncertainty factor). The magnitude of the uncertainty factor depends on the nature of the toxicity observed, the quality of the toxicological data available, and whether the effects were observed in humans or animals (IEH, 1999b).

Variability Measurable factors that differ e.g. height is variable across populations. The majortypes of variability are temporal, spatial and interindividual. They may be discrete(e.g. albinism) or continuous (e.g. body weight). It may be readily identifiable (e.g. presence of albinism) or difficult to identify (e.g. ability to detoxify a particularchemical metabolite)

Volume of Is the relationship of plasma chemical concentration and the amount of chemical distribution distributed throughout the body. This is not a real volume in the true sense, but an (Vd) apparent (mathematical) volume which can be estimated by:

DVd =

Cp

where D = dose administered, and Cp = plasma concentration (Gossel and Bricker, 1989).


xxvii


1

Risk assessment is the process of estimating thepotential impact of a chemical, physical,microbiological or psychosocial hazard on aspecified human population or ecological systemunder a specific set of conditions and for a certaintimeframe.

Risk assessment is intended ‘to provide completeinformation to risk managers, specificallypolicymakers and regulators, so that the bestpossible decisions are made’ (Paustenbach, 1989,p. 28). There are uncertainties related to riskassessment and it is important to make the bestpossible use of available information.

Risk assessment gathers and organisesinformation and enables:

• risks at a point in time (including baselinerisks) and changes in risk over time to beestimated and whether action is necessary;

• health Guidance Values to be estimated forenvironmental hazards that can be used andwhich will adequately protect public health;

• assessments of new types of risk;

• assessments of different types of risk;

• a comparison of the potential health impactsof various environmental health interventions(thus enabling cost-effectiveness estimates);

• the identification and comparison of differentfactors that affect the nature and magnitudeof the risk;

• risk-based standards setting for regulatoryexposure limits, and clean-up standards

• prioritising issues according to their levels of risks;

• questionable theories, methods and data to bechallenged and addressed by providing aclearly documented and open process(Covello and Merkhofer, 1993);

• risk-based policy making; and

• consistent, transparent appraisal andrecording of public health risks.

Risk assessment may be done as a relatively rapid‘desk top’ study for simple issues or may be a largeand complex process where there are significanthealth concerns. There are numerous models ofrisk assessment to suit the many contexts inwhich risk assessments are undertaken.

Even limited measures of the level of risk can bevaluable for identifying complex cause and effectprocesses and the most efficient means ofaddressing the risks.

It is important that assessors, users, regulators andmembers of the public recognise risk assessmentmay not always provide a compelling or definitiveoutcome. There are criticisms of risk assessmentsome of which are:

• default values and assumptions are notrealistic. A series of such unrealistic values orassumptions compounds the inaccuracy sothat risks may be seriously overstated orunderstated if the default values are tooconservative or insufficiently conservative,respectively;

• interactions between agents (i.e. mixtures ofagents) and the variability of responsebetween individuals may be insufficientlytaken into account;

• the use of default values and assumption maybecome too rigid so that situation-specificdata are not applied;

• the nature of the population to whom the riskassessment is to be applied is sometimespoorly defined;

• the uncertainties of risk assessment are ofteninadequately described e.g. specific pointestimates are given which do not recogniseuncertainty, or simplistic upper boundestimates of uncertainty are used;

• there is an emphasis on cancer risk to thepossible neglect of other adverse effects e.g.reproductive and developmental outcomes;

• in some situations there may be insufficientscientific knowledge to be able to performcredible risk assessments;

Background to Risk Assessment

1


• risk assessment can be perceived to betailored to provide a desired or predeterminedoutcome (NRC, 1994);

• excessive emphasis is given to the process ofrisk assessment rather than its content. In theUSA, a US$150 000 risk assessment wasundertaken on the results of a singleenvironmental sample;

• the risk assessment process is used as a‘whitewash’;

• Tal (1997) indicates that environmentalgroups identify a number of problems withthe way risk assessments have been practisedincluding disempowerment and potentialregulatory delays. Risk assessments should bedesigned and undertaken in ways thatminimise these pitfalls; and

• risk assessment is used to justify thecontinuation or increase of pollutingactivities.

1.1 When to Undertake RiskAssessment

The Issues Identification phase will determinewhen to undertake a risk assessment. The need toundertake a risk assessment will be influenced bysituation-specific factors. As such, the followinglist is indicative and not exhaustive. In generalrisk assessments will be needed for products,processes, situations and activities where there is aplausible increased risk of significant healthconsequences for the human population.Examples are:

• new additives to food or potable orrecreational waters;

• changes to climate, landform, geography ordemography that may impact on diseasevectors and parasites;

• situations where environmental standards orguidelines are unavailable;

• environmental changes that will increasetraffic flow and may increase the risk of injuryor air pollution e.g. new traffic corridors;

• changes where impacts on environmentalhealth factors may be permanent andirreversible;

• changes which may impact on themicrobiological or chemical safety of foodchains and food supplies;

• situations where there is a high level of publicinterest in and/or concern aboutenvironmental health issues;

• situations where vulnerable populations maybe affected by environmental health issuese.g. the placement of schools;

• planning new towns or communities;

• situations involving planning modification orapproval;

• legislative changes;

• policy changes;

• designating housing set-backs from industryand transport corridors;

• assessment of existing installations to improveexisting risk treatment practices;

• locating new airports;

• locating new power generation plants;

• locating intensive horticulture;

• locating toxic waste disposal plants;

• locating sewage treatment plants;

• designation of watersheds; and

• where Health Impact Assessments areundertaken.

1.2 Types of Risk Assessment1.2.1 Individual and population risk

assessmentsRisk assessments may assess individual orpopulation risks. Individual risks may be for theaverage (i.e. typical) individual or the highlyexposed or particularly susceptible individual andthe risks may be estimated for various durations of

2

exposure (e.g. per year or per lifetime) or fordifferent locations. Individual risk can only beassessed for a hypothetical individual withassumed characteristics. Assessing the risk for anyreal individual will be frustrated by the fact thatrisk predictions for an individual can never bevalidated by experience. Any real individual willeither experience the negative outcome or will not.Neither of these results can validate any riskprediction other than a probability of one or zero.

Population risk may relate to the number ofadverse health effects (e.g. fatalities, cancers, orillnesses) in a population over a specified periodof time or the rate of adverse effects for a givenlocation or sub-population (Covello andMerkhofer, 1993).

1.2.2 Qualitative and quantitativerisk assessments

The level of risk can be described eitherqualitatively (i.e. by putting risks into categoriessuch as ‘high’, ‘medium’ or ‘low’) or quantitatively(with a numerical estimate). Current riskassessment methods do not enable accuratequantitative estimates of risk for low levels ofexposure to environmental hazards. Numericalestimates of risk will rarely be feasible because ofvariability in the agent and population andlimitations in toxicological and exposure datawhich will be reflected in the uncertaintyassessment, but a degree of quantification may bepossible for some components such as datacollection and exposure assessment. Furtherdiscussion of qualitative and quantitative riskassessment is given in Section 9.3 of ‘RiskCharacterisation’.

1.3 Assessing Risk AssessmentMethods

There are various criteria for assessing riskassessment methods (Covello and Merkhofer,1993) including:

• the logical soundness of the method (e.g. itsjustification based on theoretical argumentsor scientific knowledge, and the validity ofthe underlying methodological assumptions);

• completeness (e.g. whether it can address allaspects of the problem and the degree towhich it excludes issues because they are hardto accommodate);

• accuracy (e.g. the precision reflected in theconfidence level associated with the results;biases resulting from undue weight given tospecific interests or considerations; and thesensitivity of results to untested or untestableassumptions);

• acceptability (e.g. compatibility with existingprocesses; whether it is viewed as rational andfair; the level of understanding for all partiesaffected by it; and the confidence andfamiliarity of those who will use it);

• practicality (e.g. the level of expertise, timeand input data required); and

• effectiveness (e.g. usefulness of results; rangeof applicability across different risks andproblem areas; the generalisability of theconclusion to other problem areas; andeffectiveness and efficiency of linkage withother types of methods).

1.4 Models of Risk AssessmentThere are various models available for theenvironmental health risk assessment. The USNational Academy of Sciences (1983) model waspublished relatively early in the development ofrisk assessment processes and the model has beenparticularly influential as a template:

‘risk assessment…mean(s) estimating the potentialadverse health effects of human exposures toenvironmental hazards. Risk assessments includeseveral elements: description of the potentialadverse health effects based on an evaluation ofresults of epidemiological, clinical, toxicological,and environmental research [hazard identification];extrapolation from those results to predict the typeand estimate the extent of health effects in humansunder given conditions of exposure [dose responseassessment]; judgements on the number andcharacteristics of persons exposed at variousintensities and durations [exposure assessment];and summary judgements on the existence and


3


overall magnitude of the public health problem[risk characterisation]. Risk assessment alsoincludes characterisation of the uncertaintiesinherent in the process of inferring risk.’

Several other international models are detailed inAppendix 6 (‘International Models of RiskAssessment’). The models are generally similaralthough the use of differing definitions of termssuch as ‘risk’, ‘hazard’ and ‘assessment’ can beconfusing. (The International Programme onChemical Safety has commenced a process toharmonise definitions and methodologies.)

1.5 Bayesian Tools for RiskAssessment

Bayesian approaches have been proposed for riskassessment. Given that in some risk assessmentsthere will be a considerable element of judgementand different experts will have different priorbeliefs, the Bayesian approaches incorporate thesein a formalised way into the risk assessment byusing simulations with different weightings sothat prior knowledge, assumptions andjudgements can be formalised and explicitly usedin the risk assessment. This approach can be asvalid as conventional statistic techniques forestimating probabilities (IEH, 1999b).

1.6 Australian Models of RiskAssessment

There are a variety of models used for riskassessment in Australia by government agenciesand consultants (Appendix 5).

This document uses a model of risk assessmentthat involves five stages. The model follows areview of various models and is based largely onthe National Academy of Sciences model (1983)with the addition of a preliminary step, ‘IssueIdentification’:

• issue identification;


• dose–response assessment;

• exposure assessment for the relevantpopulation; and

• risk characterisation.

These five stages are closely linked and highlydependent on the preceding stages. The model isillustrated in Figure 1, below. The terminology issimilar to terminologies used by other majormodels (See Figure 1, Appendix 6).

1.6.1 Issue identificationIssue Identification identifies issues amenable torisk assessment and assists in establishing acontext for the risk assessment by a process ofidentifying the problems that the risk assessmentneeds to address. It includes identifying:

• what is the concern;

• what is causing the identified concern;

• why is the concern an issue;

• how the concern was initially identified;

• how the concerns were raised;

• whether the issue is amenable to riskassessment; and

• whether risk assessment is appropriate.

1.6.2 Hazards vs issues‘Hazards’ need to be distinguished from ‘issues’.The determination of the issues is necessary toestablish a context for the risk assessment andassists the process of risk management. Issues havedimensions related to perceptions, science,economics and social factors. Examples of issuesare: community concerns over emissions from asmelter; community outrage over the proposeddevelopment of a communications tower;assessment of a new water treatment chemical; andchanges to a microbiological food standard.‘Hazards’ relate to the capacity of a specific agent toproduce a particular type of adverse health orenvironmental effect. Examples of hazards are: thecapacity of benzene to cause leukemia; the capacityof solar radiation to cause skin cancer; the capacityof Salmonella to cause vomiting and diarrhoea.

4

Figure 1: Risk assessment model


5

Issue Identification

Risk Characterisation

Hazard Assessment Exposure Assessment

Review and

reality check

Review and

reality check

- identification of key issues amenable to risk assessment

- Analysis of hazard locations- Identification of exposed populations- Identification of potential exposure pathways- Estimation of exposure concentration for pathways- Estimation of contaminant intakes for pathways- Uncertainty analysis for exposure assessment step

- Characterise potential for adverse health effects to occur- Evaluate uncertainty- Summarise risk information

Risk Management- Define the options and evaluate the environmental health, economic, social and political aspects of the options- Make informed decisions- Take actions to implement the decisions- Monitor and evaluate the effectiveness of the action taken

Hazard Identification

- Collection and analysis of relivant data- Uncertainty analysis for hazard identification step

Dose–response Assessment

- Collection and analysis of relevant data- Uncertainty analysis for dose-response assessment step

Engage the Stakeholders, Risk Communication and Community Consultation


1.6.3 Hazard assessment‘Hazard Assessment’ is comprised of ‘HazardIdentification’ and ‘Dose–Response Assessment’.

1.6.4 Hazard identificationHazard identification involves determining:

• what types of (adverse) health effects mightbe caused by the agent; and

• how quickly the adverse health effects mightbe experienced and their duration (HealthCanada, 1999).

The data for hazard identification will come froma range of toxicological, epidemiological, in vitroand mechanistic studies. Not only the agent mayneed to be assessed but, in the case of chemicals,the breakdown products e.g. acrolein as well asbutadiene when doing environmental monitoring;the four metabolites of atrazine (desethylatrazine,desisopropylatrazine, diamonochlorotriazine andhydroxyatrazine) when monitoring atrazinecontamination of water catchments.

1.6.5 Dose–response assessmentDose–response assessment considers bothqualitative and quantitative toxicity informationto determine ‘the incidence of adverse effectsoccurring in humans at different exposure levels’(US EPA, 1989, p. 1.6). Where available, humanand animal evidence will be assessed as part ofthis process. Risk assessment cannot be donewithout good dose–response information.

Whereas constant doses can be used in animalstudies, long term human exposures may bevariable. This may be a significant source ofuncertainty and there is a need to develop anintegrated estimate of long term exposure.

1.6.6 Exposure assessmentExposure assessment involves the determinationof the frequency, extent duration and character ofexposures in the past, currently, and in the future.There is also the identification of exposedpopulations and particularly sensitive subpopulations, and potential exposure pathways.Environmental monitoring and predictive models

can be used to determine the levels of exposure atparticular points on the exposure pathways. Thecontaminant intakes from the various pathwaysunder a range of scenarios can then be estimated(US EPA, 1989).

Where the risk assessment is being done as partof a protective and pro-active risk assessment,exposure assessment data may not be availableand may have to be estimated. Modelled data mayalso be used where the data package is limited.

1.6.7 Risk characterisationRisk characterisation provides a qualitative and/orquantitative estimate, including attendantuncertainties, of the nature, severity and potentialincidence of effects in a given population basedon the hazard identification, dose–response andexposure assessments.

1.6.8 Follow upRisk assessment is an iterative process that will bereviewed as the risk assessment progresses. Afterrisk assessment is completed there may be a needto review the situation from time-to-time as newinformation becomes available or circumstanceschange to ensure that the risk assessment is stillrelevant and protective.

1.6.9 Aims of the health riskassessment method

The method is intended to assist risk assessmentpractitioners and those evaluating riskassessments.

The aims of the method are:

• to identify information needed to makedecisions;

• to make the decision-making process moreexplicit by identifying the specific elementsaffecting risk so that more objective andscientific decisions can be made;

• to make the decision process moretransparent to promote confidence by thecommunity, industry and scientists aboutdecisions and actions taken;

6

• to increase consistency in risk assessment sothat different people assessing similar problemswill come to comparable conclusions;

• to have the ability to account for the range ofrisks that are present or could arise as theresult of actions;

• to refine the assessment and management ofrisk so that better decisions are made andmore rigorous risk assessment andmanagement occurs;

• to enable the adoption of futureimprovements to risk assessment; and

• to enable risk-benefit analysis and theevaluation of the outcomes of riskmanagement decisions about current andpossible future risks (ACDP, 1996).

1.7 Risk Assessment andHealth Impact Assessment

Health impact assessment is described in theHealth Impact Assessment Guidelines (enHealth,2001) as a systematic process to assess the actualor potential, and direct or indirect, effects on thehealth of individuals, groups or communitiesarising from environmental conditions or hazardsarising from policies, objectives, programs, plans,or activities. It is usually a process undertaken aspart of an Environmental Impact Assessment fora significant project and looks at both positiveand negative impacts on health. The definition ofhealth is taken to be broader than the mereabsence of disease or infirmity but ‘a completestate of physical, mental and social wellbeing’(WHO Constitution).

Environmental Health Risk Assessment providesa tool for appraising health risks in the broaderprocess of Health Impact Assessment.

1.8 Principles of RiskAssessment

In conducting risk assessments there are severalguiding principles:

• Risk assessment is to inform the riskmanagement process.

• Risk assessors and risk managers should besensitive to distinctions between riskassessment and risk management.The assessors should (a) generate a credible,objective, realistic, and scientifically balancedanalysis; (b) present information on theseparate components of the risk assessment;and (c) explain the confidence in eachassessment by clearly delineating strengths,uncertainties and assumptions, along with theimpacts of these factors (e.g. confidencelimits, use of conservative/non-conservativeassumptions) on the overall assessment. Therisk assessors do this without consideringissues such as cost, feasibility, or how thescientific analysis might influence theregulatory or site-specific decision (US EPA, 1995, p. 2).

• Risk assessment processes should be coherentand transparent. It is important that the basisof the decision-making is clearlydocumented. This formal record should beclear comprehensive and concise and includea summary of the key data which influencedthe risk assessment and an appraisal of itsquality (ACDP, 1996, p. 5).

• Risk assessment information is only one ofseveral kinds of information used fordecision-making. The risk managementdecision will not be determined only by therisk assessment but a range of other factorsincluding ‘technical feasibility (e.g.treatability, detection limits), economic, social,political,’ and legislation when determiningwhether to regulate and, if so, to what extent (US EPA, 1995, p. 2).

• Consultation with the community to identifytheir concerns.

• Scientific judgements and policies must beclearly identified. Inevitable gaps inknowledge will be filled by scientificjudgements and policies. These must beclearly identified so that others mayunderstand the role of judgement ininterpreting the evidence.


7


1.9 Dealing with Uncertaintyand Variability

The use of conservatism should be carefullyconsidered. An appropriate level of conservatismis important for policy makers because, in general,underestimating a particular risk is likely to havegreater health, environmental, economic andsocial losses than overestimating the same risk.However, applying conservatism to riskassessment may distort the results particularlywhere there is layer upon layer of conservativeassumptions, the compounding effect of whichmay be an overly cautious risk assessment.Addressing an excessively cautious risk assessmentmay have significant opportunity costs on acommunity i.e. the extra money could be moreeffectively spent on other health interventions(Covello and Merkhofer, 1993). If undulyconservative risk estimates are misinterpreted asthe expected risks, considerable anxiety may becreated in the community.

A degree of conservatism is warranted wherethere are significant uncertainties about exposureor toxicological data or if the variability inpopulations has not been taken into account. Thevariability can arise from heterogeneity in factorssuch as:

• uptake e.g. due to variations in diet andinhalation rates;

• pharmacokinetic heterogeneity resulting indifferences in concentration over time in theblood or at the site of action e.g. due todifferences in metabolism or excretion;

• response, where there are differences at thesite of action (Hattis and Silver, 1994 p. 422);and

• the degree of conservativeness should bemade quite clear in the risk assessment sothat risk managers are fully aware of theprecautions inherent in the risk assessmentand do not add unnecessarily further levels ofconservatism for the risk management step.

1.10 Risk Assessment and thePrecautionary Principle

The Precautionary Principle was first introducedin 1984 at the First International Conference onProtection of the North Sea and has since beenintegrated into numerous internationalconventions and agreements. In Australia, theInter Governmental Agreement on theEnvironment (May, 1992) describes the principlein the following way:

Where there are threats of serious or irreversibleenvironmental damage, lack of full scientificcertainty should not be used as a reason forpostponing measures to prevent environmentaldegradation. In the application of thePrecautionary Principle, public and privatedecisions should be guided by:

i) careful evaluation to avoid, whereverpracticable, serious or irreversible damage tothe environment; and

ii) an assessment of the risk-weightedconsequences of various options.

The Precautionary Principle is particularlyrelevant during the risk management phase. Riskassessment provides a process for applying thePrecautionary Principle by providing informationabout the nature and magnitude of the ‘threats ofserious or irreversible environmental damage’.

The European Commission (1998) describes theprecautionary approach as a risk management toolto be used in the face of scientific uncertainty andwhere there is a need for action in the case of apotentially serious risk without awaiting theresults of scientific research. The European Unionhas detailed six guidelines (ibid) as the basis forthe approach:

1. Implementation of an approach based on theprecautionary principle should start with anobjective risk assessment, identifying at eachstage the degree of scientific uncertainty.

2. All the stakeholders should be involved in thedecision to study the various managementoptions that may be envisaged once theresults of the risk assessment are available andthe procedure be as transparent as possible.

8

3. Measures based on the precautionaryprinciple must be proportionate to the riskwhich is to be limited or eliminated.

4. Measures based on the precautionaryprinciple must include a cost/benefitassessment (advantages/disadvantages) withan eye to reducing the risk to a level that isacceptable to all the stakeholders.

5. Measures based on the precautionaryprinciple must be able to establishresponsibility as to who must furnish thescientific proof needed for a full riskassessment.

6. Measures based on the precautionaryprinciple must always be of a provisionalnature, pending the results of scientificresearch performed to furnish the missingdata and perform a more objective riskassessment.

1.10.1 Key factors in riskassessments

Key factors that are to be considered in riskassessments (ACDP, 1996) include:

1. Hazard assessment

• interactions with other agents in theenvironment

• immediate or delayed onset of healtheffects

• severity of health effects

• reversibility of health effects

• presence of a clear threshold for effects

• potency of agent

2. Exposure

• duration of exposure

• frequency and consistency of exposure

• patterns of exposure

• past, current and future exposure

• timing of exposure

• exposure route (ingestion vs inhalation vsdermal contact) may influence outcome

• intergenerational exposures

• cumulative vs non-cumulative exposures

• failure of exposure controls

• quality of exposure data

• quality of exposure models

3. Population

• genetic variability

• individual host characteristics (e.g. age,gender, body weight, pre-existing poorhealth, immune status, nutritional status,previous exposures, reproductive status)

• population characteristics (e.g. herdimmunity and social behaviours forcommunicable diseases, social mobility forexposure to air and soil contaminants,recreational patterns for exposure tocontaminated recreational waters)

4. Environment

• intervention strategies (e.g. containment ofcontaminated soil, chlorination of water,pasteurisation of food)

• transport mechanisms (e.g. meteorologicalfactors affecting air pollution, vectors forcommunicable diseases)

• factors affecting persistence (e.g. photolysisand volatilisation of chemicals, desiccationof microorganisms)

• breakdown of public health measures (e.g. flooding affecting waste control andpotable water treatment

1.11 Risk Assessment and Particular Population Groups

Sensitivity of individuals is likely to be affected byage, sex, nutritional and pregnancy status, andcombinations of these (IEH, 1999c).


9


1.11.1 Risk assessment and childrenChildren may differ from adults in a range ofbehavioural and physiological parameters thatmay need to be taken into account in the riskcharacterisation phase of risk assessments.

The principal factors causing these potentialdifferences are (Roberts, 1992):

• growth, development and maturational rates;

• children have greater potential futuredurations of life, which is relevant to thepotential for accumulation or exceedinglatency periods;

• dietary differences—children can eat muchgreater quantities of particular foods(particularly dairy products, soft drinks andsome fruit and vegetables) than adults on abody weight basis (Rees, 1999);

• exposure factors—the surface area to bodymass ratio will change markedly with ageing.In the newborn the ratio is typically 0.067(m2/kg) decreasing to 0.025 in an adult.While the respiratory volume remains fairlyconstant at 10 ml/kg/breath, the surface areaof the alveoli increases from 3m2 in an infantto approximately 75m2 in an adult and therespiration rate drops from 40 breaths perminute to 15 breaths per minute (Snodgrass,1992). Children have unique exposurepossibilities e.g. placental transfer and breastmilk (Kimmel et al, 1992);

• behavioural factors, e.g. children are morelikely to indulge in soil eating behaviours;

• available parameters for toxicity assessment,e.g. techniques for assessing dizziness,intelligence and hearing impairment aredifferent between children and adults;

• biochemical and physiological responses—children have a higher metabolic rate, morelimited ability to control body temperature,more rapid growth rate, a higher percentageof water in the lean body tissue;

• disposition of the agent within the body, e.g.transit time, pH and enzyme activity in the

gut are different for children as are tissue-chemical bindings;

• liver function related to detoxificationmatures after birth, as does the renalexcretion of foreign compounds;

• differences in gut microflora;

• the immaturity of children’s immune systems;and

• differences in the clearance of chemicals—thehigher clearance of certain chemicals fromthe body in children compensates in part forthe greater sensitivity for their developingorgan systems (Renwick, 1999) but for someother chemicals, clearance may be lower.

The potential impact of these differenceshighlights the need for agent-by-agent appraisal.

1.11.2 Risk assessment and older persons

For the ageing, there is a lack of functionalreserve in the physiological and psychologicalsystems. Distribution of chemical agents isaffected by changes in body composition withage: body fat increases and body water decreaseswith age. The clearance of renally eliminatedcompounds is reduced because of changes in renalfunction. Liver function can be reduced in theelderly affecting biotransformation of chemicalagents. Increased sensitivity to the central nervoussystem in the aging population from many drugshas been reported (Crome 1999). Changes willoccur to the immunological system often resultingin reduced immunocompetance.

Ageing populations are very heterogeneous interms of their general health. For those withimpaired health, there may be a variety ofconditions present.

Cognitive impairment is common in the very oldand affects their abilities to recognise, interpretand react to acute and chronic environmentalhazards. They are higher consumers ofpharmaceuticals and there is a potentialinteraction with these pharmaceuticals and other agents.

10

1.11.3 Risk assessment and genderGender differences may need to be taken intoconsideration when identifying potential exposurepathways in the exposure assessment phase andcharacterising potential adverse health effects inthe risk characterisation phase of the riskassessment process.

There are anthropometric (e.g. height, weight,body surface area) and body compositiondifferences (e.g. fat content, muscle mass)between males and females that may affectexposure concentrations of agents from differentpathways. These differences may also influencethe absorption, distribution, metabolism andelimination of xenobiotics and have a significantinfluence on toxicity (Silvaggio and Mattison,1994). Some of the factors which influence theseprocesses are summarised in Tables 1–4.

Men and women also differ in many lifestylefactors (e.g. alcohol drinking and cigarette

smoking) dietary patterns and how they spendtheir time (American Industrial HealthFoundation, 1994; US EPA, 1996) and theiroccupational exposures. These factors mayinfluence the exposure and effect of an agent onthe individual.

For many chemical toxicants there are importantdifferences between males and females inexperimental studies. Calabrese (1985) identified200 toxicants where toxicological data analysis ofanimal studies suggest there are importantdifferences between males and females in theexpression of toxicity.

There have been reports of differences whencomparing men and non-pregnant women intheir response to toxic levels of lead, berylliumand benzene. Gender differences have also beenreported to occur from exposure to ionisingradiation, noise and vibration and extremetemperature changes (i.e. heat and cold stress)(Hunt, 1982).


11

Table 1: Factors influencing the absorption of chemicals

Parameter Physiological difference Toxicokinetic impact

Gastric juice pH M < F < pregnant F Absorption of acids/bases modified by change in pH

Gastric juice flow M > F > pregnant F Absorption modified by decreasing flow

Intestinal motility M > F > pregnant F Absorption increases with decreasing motility

Gastric emptying M > F > pregnant F Absorption and gastric metabolism increase with decreasing gastric emptying

Dermal hydration Pregnant F > M, F Altered absorption in pregnant F

Dermal thickness M > F Absorption decreases with increasing dermal thickness

Body surface area M > pregnant F > F Absorption increases with increasing body surface area

Skin blood flow Pregnant F > M, F Absorption increases withincreasing skin blood flow

Pulmonary function M > pregnant F > F Pulmonary exposure increases with increasing minute volume

Cardiac output M > pregnant F > F Absorption increases withincreasing cardiac output

F, female; M, male


Table 2: Factors influencing the distribution of chemicals in the body


Plasma volume Pregnant F > M > F Concentration decreases with increasing volume

Total body water M > pregnant F > F Concentration decreases with increasing body water

Plasma proteins M, F > pregnant F Concentration fluctuates with changes in plasma proteins and protein binding

Body fat Pregnant F > F > M Body burden of lipid-soluble chemicals increases with increasing body fat

Cardiac output M > pregnant F > F Distribution rate increases with increasing cardiac output

F, female; M, male

Table 3: Factors influencing the rate of metabolism of chemicals


Hepatic metabolism Higher BMR in M, Metabolism generally increases with BMRfluctuating hepatic metabolism in pregnant F

Extra-hepatic metabolism Metabolism by fetus/placenta Metabolism fluctuates

Plasma proteins Decreased in pregnant F Elimination fluctuates with changesin plasma proteins and protein binding

BMR, basal metabolic rate; F, female; M, male

Table 4: Factors influencing the elimination of chemicals from the body


Renal blood flow, GFR Pregnant F > M > F Renal elimination increases with increasing GFR

Pulmonary function M > pregnant F > F Pulmonary elimination increases with increasing minute volume

Plasma proteins Decreased in pregnant F Elimination fluctuates with changes in plasma protein and protein binding

GFR, glomerular filtration rate; F, female; M, male(tables adapted from Government/Research Councils Initiative on Risk Assessment and Toxicology, 1999)

12

1.11.4 Risk assessment and reproductive status

The human reproductive system is susceptible toenvironmental factors that can produce a varietyof adverse effects during the production of ova(oocytogenesis) and viable sperm(spermatogenesis); on fertilisation; onimplantation within the uterus; and growth anddevelopment of the embryo and fetus.

Reproductive status is also influenced by theextent of exposure and adverse effects fromoccupational and environmental agents.Teratogenesis (abnormal development of theembryo and fetus) is a risk for the fetus that maybe exposed to environmental agents. The principalfactors that determine an agent’s risk ofteratogenicity and which need to be considered inrisk assessment include (Goldfrank et al, 1990):

• the nature of the agent;

• access of the agent to the fetus;

• the onset and duration of exposure;

• the level and duration of dosage; and

• the genetic constitution of the fetus.

Substances that inhibit mitosis,(e.g. antineoplastic agents such as vincristine and vinblastine) are also of a particular risk topregnant women and exposure to such agents maylead to teratogenicity and embryotoxicity. Thefemale fetus is sensitive to toxins affectinggametogenesis which, in humans, finishes by theseventh month.

Access of an agent to the fetus is determined byits molecular weight. Generally the larger themolecular weight of a substance, the less likely itwill cross the placental barrier. Most teratogeniceffects are also dose related, that is, the larger thedose, the more likely and severe the effect. Highdose exposures to polychlorinated biphenyls(PCBs) have been associated with fetalabnormality.

Timing of exposure is particularly important. Thecritical period for organogenesis is in the firsttrimester (between days 18 and 55 of gestation).

This is the time of greatest cell differentiation andenvironmental agents may have a profound effecton development at this stage. For example, withthalidomide the period of greatest sensitivityappears to be between days 21 and 33 ofgestation. The effects of thalidomide on the fetusdo not appear to be dose related and teratogeniceffects appeared in over 80 per cent of the fetusesexposed during the critical period (AmericanAcademy of Pediatrics, 1999)

The extent of the toxicity effect will also dependon the genetically determined detoxificationmechanisms (i.e. enzyme systems) of individuals.

Exposure of environmental or occupational agentscan also occur at the postnatal stage. Theproduction of milk during nursing and breast-feeding is one pathway for the excretion ofcontaminants such as lead, mercury, PCBs andorganochlorine pesticides (e.g. DDT) stored inother body tissues. Kinetic processes such asabsorption, distribution and elimination willinfluence the passage of agents into breast milk.Milk has a high fat and protein concentration andlipid-soluble or protein-bound contaminants passreadily to milk and are dissolved in or bound tothe milk fat and protein. (Hunt, 1982).

1.11.5 Risk assessment and lifestyle factors

Lifestyle factors may have an impact onindividual risk assessments and population riskassessments if the activity is widespread. For thisreason, the potential influence of lifestyle factorsneeds to be clearly identified in risk assessments.Specific lifestyle factors that may have an effecton risk assessment include:

• tobacco smoking;

• diet; and

• hobbies.

Tobacco smoking will affect the exposureassessment component of the risk assessmentprocess because there will be an increase inbackground exposure to substances found insmoke e.g. cadmium, cyanide and polycyclicaromatic hydrocarbons (PAHs).


13


Tobacco smoking also affects the toxicityassessment component. Maternal cigarettesmoking and passive smoking have beenassociated with respiratory illness, acute toxicityand cardiotoxicity among newborns. Furthermoreepidemiological studies have shown evidence ofsynergistic interaction between humancarcinogens and long term cigarette smoking. Thebest studied interactions have included jointexposure to tobacco and radon and tobacco andasbestos, respectively. Results fromepidemiological studies of joint exposure to radonand cigarette smoke have shown an additive orpossibly a multiplicative increase in the number ofcancers induced and a synergistic decrease in thelatency period for tumour induction.Epidemiological studies have shown that asbestosand tobacco administered together can producean increased incidence in lung cancer that isgreater than from the administration of eitheragent alone and the interaction is considered to bemultiplicative by most investigators (NRC, 1994).

Diet will also influence the stages of the riskassessment process particularly the toxicity andexposure assessment stages. Interactions betweentoxic metals and essential metals from the diethave being known to affect the risk of toxicity.Absorption of toxic metals from the lung andgastrointestinal tract may be influenced by thepresence of an essential metal or trace element ifthe toxic metal shares the same homeostaticmechanism. Examples are lead and calcium, andcadmium and iron. Other dietary interactionsinclude an inverse relationship between proteincontent of the diet and cadmium and leadtoxicity. Vitamin C in the diet also reduces leadand cadmium absorption.

Different types of food will have differentamounts of agents and hence cause a range oftoxic effects depending on dietary habits. Forexample the major pathway of exposure to manytoxic metals in children is food and childrenconsume more joules per kg of body weight than adults do. Furthermore, children have ahigher gastrointestinal absorption of metals,particularly lead.

Alcohol ingestion may influence toxicityindirectly by altering diet and reducing essentialmineral intake. The ingestion of alcoholicbeverages (ethanol), fats, protein, calories andaflatoxins has been implicated in carcinogenesis.(Klaassen, 1996).

Homegrown produce such as vegetables has beenassociated with contamination of heavy metalssuch as lead, arsenic and cadmium. Free-rangepoultry tissue (e.g. meat, fat, skin) and eggs (eggyolk) have been associated with contamination byorganochlorine pesticides such as aldrin, dieldrinand DDT. Hence the consumption of these foodtypes may result in an increased exposure to theseagents (Cross and Taylor, 1996).

The type of diet can also influence the risk toexposure to hazardous agents. Individuals who arevegetarians will have a reduced exposure to zinc.Individuals who consume barbecued foods can beexposed to relatively large amounts of PAHs fromthe charcoal used to cook the food. Populations(e.g. general population and fishermen) whoconsume seafood may be exposed to heavy metalssuch as mercury in fish and zinc in shellfish (e.g. oysters).

The exposure to a hazard may also be influencedby lifestyle and hobbies. For example the amountof time spent indoors (e.g. in the home, workenvironment/office, factory), outdoors ortravelling in the car, bus, aeroplane, train will alsoinfluence the amount of exposure of agents andthe risk to health (e.g. lead, benzene levels in thecar, cosmic radiation in aeroplanes etc). Hobbiessuch as pistol shooting in indoor shooting ranges;antique furniture restoration, lead soldering, boatbuilding and lead lighting can result in anincreased exposure to lead (Lead Safe, 1997).House renovating can also result to an increaseexposure to hazardous agents such as lead andasbestos. Other hobbies involving paint strippingusing methylene chloride can cause exposure to itsmetabolic breakdown product, carbon monoxide,and car maintenance can also result in an increasein exposure to hydrocarbons and heavy metals.

14

Community Consultation



Risk Management




Review and

reality check

Review and

reality check


16

Figure 2: Relationship of risk assessment and risk management

(adapted from P/CCRARM, 1997; Patton, 1998; NRC, 1983)

An Australian Framework for Risk Assessment

2

2.1 Context of Risk Assessment

The relationships of environmental health riskassessment, risk management, stakeholderengagement, risk communication and communityconsultation are detailed in Figure 2 whichdemonstrates the links between the steps and theoverlap with risk management.

2.2 Community Consultationand Involvement

There is a growing awareness of the need forappropriate community consultation andinvolvement. The process may not lead toconsensus, but it is likely to ultimately smooth thepassage of a proposal to increase the validity ofthe risk management process, and to provideinformation that is useful throughout the riskassessment and management steps.


17

Engage the Stakeholders, Risk Communication and Community Consultation

Risk Assessment Risk Management

Identifying the Issues



and Dose–response

Assessment

Exposure Assessment

Informed decision making

Implement decisions

Scientific, Technological,

Social, Economic and Political Information is a major

influence on Risk Management

but also affects Risk Assessment

Review and reality

check

Review process

Review and reality

checkMonitor and evaluate the effectiveness of the actions

taken


18

A partnership approach is most appropriate andthe scope and nature of community involvementshould be commensurate with the potentialeffects on the community. Triggers for communityinvolvement should be identified.

There are usually time, money and resourcepressures on a development but urgency toachieve short term goals can ultimately lead todelays (See Box 1).

Risk cannot be managed without addressinghuman behaviour so that community involvementis essential in the process. Communityconsultation can provide assistance at each step ofthe risk assessment process. Examples follow ofhow it can be used at each stage but it does notneed to be tightly compartmentalised.

Issue identificationAt first contact community involvement canprovide a range of information about the site,

health concerns and potential value conflicts.A communication plan can be prepared at thistime. Further matters where information can beexchanged with the community include:

• why the risk assessment is being undertaken;

• what the risk assessment will consider (i.e. what risks deserve attention);

• what information may be available from the community;

• how the risk assessment will be performed(i.e. what process will be used); and

• what will happen to the risk assessment (i.e. how does it fit into the risk managementprocess).

Hazard identification• To provide information about data gaps, local

perceptions of hazards and the applicability ofassumptions to the community.

Box 1: Management of scheduled wastes

By the late 1980s Australia had accumulated a substantial store of waste organic pollutants such as theorganochlorine pesticides (e.g. DDT, aldrin and heptachlor), hexachlorobenzene and PCBs. These wastes werecalled ‘intractable wastes’ as they were persistent and difficult to destroy. In an attempt to dispose of them adecision was made by the Australian Government to develop high temperature incinerators and to locate themin New South Wales, Victoria, Western Australia or the Northern Territory. There were at least twelve attemptsto establish a high temperature incinerator (HTI) for scheduled wastes. These attempts to site a HTI failed asthe communities were unwilling to accept the risks associated with toxic waste incineration.)

In July 1992, after the final proposal to establish a HTI had failed, the Australian and New ZealandEnvironment and Conservation Council (ANZECC), acting on the advice of an independent panel, decided toabandon the proposal to establish a centralised high temperature waste disposal facility.

The rejection of the HTI created significant problems for the disposal of the scheduled wastes in Australia butalso created significant opportunities as alternative methods of destruction had to be investigated andestablished. While in some cases these technologies were more expensive to establish, as pollution controlrequirements for HTIs expanded, the price differentiation decreased.

In response to this situation two committees were established; the Scheduled Wastes Management Group(SWMG) and the National Advisory Body (NAB). The SWMG was made up of senior officers of the State,Territory and Commonwealth environment departments. The NAB comprised stakeholders from industry,government, unions and environmental groups. Ultimately, through a process of negotiation and broadcommunity consultation, a series of national management plans were developed that were agreed to by themembers of the advisory body. These plans have been endorsed by ANZECC and are currently being adoptedinto State legislation and implemented.

Dose–response relationships• The community’s attitudes towards the range

and type of technical data and the assumptionsmade in the interpretation of the data.

Exposure assessment• Information about: the community’s attitudes

to biological monitoring, and healthmonitoring; local knowledge of the range andnature of exposures; relevant exposuresettings; the community’s attitudes tosampling design and environmentalmonitoring and to the uncertainties and assumptions in the exposure assessment phase.

Risk characterisation• Information of the community’s concepts of

risk and safety

Evaluating the actions taken• Community involvement will affect how

environmental monitoring may be undertakento ensure that the best decisions are made.

Risk management• Information of the community’s concepts of

acceptable risk and safety. Communityconsultation is an integral part of riskmanagement.

The objectives for each stage of the riskassessment process should be examined todetermine the nature of the communityconsultation.

At a legislative and administrative level therequirements for, and practise of, communityconsultation vary. For example, there is limitedcommunity consultation required for certainaspects of the approval of therapeutics.

2.3 Risk Perception and Risk Communication

Ideally ‘actual’, ‘estimated’ and ‘perceived’ risksshould be closely aligned. This presents animmediate problem as actual risks areunquantifiable and unknowable. The aim of risk

assessment should be to achieve the alignment ofactual and estimated risk and the aim of good riskcommunication should be the alignment ofperceived and actual risk.

All parties, both expert and non-expert will haveperceptions of risks. Experts and non-expertsalike are influenced by emotion, beliefs and theirviews of the world (Thomas and Hrudey, 1997).

A simple numerical estimate of risk portrayed asthe ‘real risk’ ignores the subjectivity and multipledimensions of risks (Thomas and Hrudey, 1997).People see risk as multi-dimensional and notrepresented by a numerical value and will judge itaccording to its characteristics and context. Forexample trauma or death as the result of aninvoluntary catastrophic reaction is likely to bedreaded more than the situation where theadverse consequences are the result of a situationwhere the risk is assumed voluntarily and theperson feels some degree of control (e.g. motorvehicle crashes).

Concerns about risk will be heightened by risksthat are:

• involuntary or imposed on the community;

• man made rather than natural;

• inescapable;

• controlled by parties outside the community;

• have little or no benefit to the community;

• unfairly distributed;

• related to an untrusted source;

• exotic or unfamiliar;

• affect children or pregnant women;

• affect identifiable rather than anonymouspeople;

• the cause of insidious and irreversibledamage;

• the cause of dreaded health effects such ascancer;

• poorly understood by science; and


19


• subject to contradictory statements fromresponsible sources (or, even worse, from thesame source) (DOH, 1998).

Concerns about risk will be lessened when:

• the risks are voluntarily assumed;

• the risks have a natural origin;

• individuals or the community feel able toexert some control over the risks;

• there are clear benefits from the risks;

• the risks are fairly distributed;

• the risks are associated with a trusted source;

• the risks are familiar;

• the risks only affect adults;

• the risks are understood; and

• the process of how the risks are determined isunderstood.

Risk communication is often seen as a one-wayprocess aimed at rectifying incongruities betweenthe community’s perceptions and the opinions ofregulators. However it should be recognised that allparties will have perceptions about a situation andthe ultimate aim is to draw these perceptions aboutrisk, the estimated levels of risks, and the actuallevels of risks as closely together as possible.

Risk communication should not be seen as aretrospective form of community involvement andconsultation. It is an interactive process involvingthe exchange among individuals groups andinstitutions of information and expert opinionsabout the nature, severity and acceptability ofrisks and the decisions taken to combat them.

Good risk communication and consultationresults in an outcome where there is a high levelof agreement between the affected parties. It alsoentails knowing how to respond to public concernand is a genuine process conducted with thecommunity’s interest in mind. Good riskcommunication and community involvement willenable government and industry to betterunderstand public perceptions and to more readilyanticipate community responses. It will increase

the effectiveness of risk management decisionsand reduce unwarranted tension. It will explainrisks more effectively and constructively informcommunities.

Information about risks needs to take intoaccount their complexities and uncertainties andbe constructed so that they can result inmeaningful interpretation by all parties. People’sresponses to risk will be strongly influenced bytheir wider values so that isolated facts about risksmay have limited impact on their acceptability(DOH, 1991) especially when they are perceivedto have little benefit.

The communication process need not always beto reduce concern about risks. Many public healthinterventions are intended to increase publicconcerns about risks such as smoking or excessivealcohol consumption. In communities withregional lead contamination (e.g. Port Pirie orBroken Hill), public health activities have beendesigned to increase concerns about what areoften subtle effects and to provide informationabout specific activities that can be undertaken toprotect children.

Some of the key principles of effective riskcommunication are:

• accepting and involving the public as apartner and stakeholder;

• carefully planning and evaluating the natureand content of the risk communicationundertaken so that it is relevant andunderstandable;

• listening to the public’s specific concerns.Trust, credibility, competence, fairness andempathy are often as important to thecommunity as statistics and scientific details.Trust and credibility are very difficult toregain if lost. Experts do not commandautomatic trust;

• being honest, realistic and open;

• appreciating that intentional communication isoften only a minor part of the message actuallyconveyed. The manner of delivery and its tonemay be more important than its content;

20

• ensuring that information is accurate,consistent between agencies, and notspeculative;

• effectively communicating with the media;

• acknowledging the public concerns and theeffects on the community; and

• focusing on issues and processes rather thanpeople and behaviours.

(adapted from US EPA 1988, DOH 1998)

Even with good community consultation and riskcommunication there may be disagreementbetween parties.

In designing community consultation and riskcommunication programs the following issuesshould be addressed:

• What is the purpose of the consultation? Is itto gain information, ideas and options? Is itto build credibility? Is it to meet regulatoryrequirements? Is it to provide maximumopportunity for public involvement?

• Who is the audience? Anybody who perceivesthemselves to be affected should be able toparticipate in the process. ‘The community’ isdiverse, with different groups regarding risk indifferent ways. They may need a range ofmessages and styles of delivery;

• How will industry be involved?

• What does the community want to know?Local community leaders, environmentalgroups, and environmental health officersmay often be able to provide broaderinformation about particular concerns (See Box 2);

• How will communication occur? Smaller,informal meetings are often more effectivethan large impersonal meetings. At largemeetings some members of the communitymay feel apprehensive about asking questionsor expressing opinions. There is a need toavoid partisan Chairs for meetings andwritten materials may have more credibility

than the spoken word. Materials need to bepre-tested before they are printed anddistributed and evaluated afterwards. There isa need to determine how industry andgovernment will listen to concerns and howinformation about concerns will be sought.If the community is not listened to, it willcease to listen;

• Do not seek more feedback than you are ableto use as this will lead to communitydisillusion and loss of trust (adapted fromChess and Hance, 1994); and

• Seeking grudging approval from thecommunity will be far less productive thangenuinely seeking feedback that will be used,asking for comments in a situation whereplans can be changed.

Avoid problems by anticipating issues such as:

• lack of communication skills (by any of theparties);

• limited resources and time and staffing (byany of the parties);

• confusion between the ‘risk assessment’ and‘risk management’ phases;

• cultural differences;

• legal considerations;

• external politics, hidden agendas and politicalpressures;

• conflicting interests within the varying partiesconcerned;

• impacts from the media; and

• evaluation of the consultation. Evaluation is acontinuous process designed to avoid midcourse corrections and repeating failures.Evaluation may cover: whether thecommunication was timely; whether thecommunication was sufficient; whether thepublic was empowered; and whether thecredibility and trust of the organisation was enhanced (adapted from Chess andHance, 1994).


21


Box 2: Aluminium smelting and the community

Public consultation and commitment to an independent study resulted in a successful resolution of public healthconcerns in the Portland community.

In 1994, Portland Aluminium sought approval to increase sulfur dioxide emissions by nearly 30 per cent so that itcould increase production at its aluminium smelter in Portland.

Members of the community were opposed to any increase in emissions with the central issue being the effect of sulfurdioxide on health. There was a widespread belief that asthma levels were high in the Portland area. There were alsosimilar concerns about the levels of sore and itchy eyes and skin irritations, as well as odours and acid smells.

Portland Aluminium stated that, with increased emissions, the use of taller stacks would improve air quality atground level by allowing sulfur dioxide to disperse higher into the atmosphere.

Many residents had concerns about the reliability of air monitoring within the Portland area and believed theywere not given complete information about the potential health effects associated with aluminium production.

In response to these concerns, the Victorian Department of Human Services established a Health ProfessionalsAdvisory Committee which included local health professionals, a respiratory physician and Departmentrepresentatives.

The role of the committee was to organise and oversee an independent health study to assess the potential for anyadverse health effects from the proposed increase in sulfur dioxide emissions from the smelter.

A proactive program of community consultation was established and local residents were interviewed and giventhe opportunity to raise key areas of concern. The committee then ensured that these concerns were addressed inthe study’s terms of reference.

The Victorian EPA then granted Portland Aluminium approval to replace the low stacks at the smelter with sixtall stacks and to monitor their emissions for 12 weeks. The findings of the health study and the results ofmonitoring of emissions from the old stacks and new, tall stacks were to be evaluated before the application toincrease sulfur dioxide emissions was granted.

The health study involved a literature review and a health survey. To determine whether there was an increase inasthma and itchy eyes in Portland, the consultants surveyed residents of Portland and Warrnambool (a similarpopulation) using a questionnaire which covered a range of health symptoms.

The study also reviewed the measurements of ground level concentrations of sulfur dioxide that resulted fromemissions from the older low stacks and the new tall stacks, after they were built.

The literature review found that there was no evidence that sulfur dioxide caused people to become asthmatic butit did cause symptoms such as wheeze to occur more often. The survey showed that other health symptoms suchas itchy eyes, cough, stuffy noise, sore throat and skin rash were more common in Portland but there was nosignificant difference in the proportion of people with asthma and wheeze, although both cities had high rates.

Monitoring data for 1995, 1996 and 1997 showed that the one-hour ‘acceptable level’ for sulfur dioxide at groundlevel was exceeded four times over this period. However, monitoring of emissions from the new tall stacks showedmuch lower levels.

The monitoring results were used to predict the ground level concentrations of sulfur dioxide that would occurwith the proposed 30 per cent increase in smelter emissions. The levels in Portland and surrounding areas werepredicted to be well below the standard.

The results of the study were discussed with the community at a public meeting and a report of the study wascirculated. The study concluded that there was no evidence that the proposed increase in sulfur dioxide emissionsfrom the taller stacks would be detrimental to health.

The report was well received by the community. Portland Aluminium was given EPA approval to increase sulfurdioxide emissions from the smelter and ongoing monitoring of air pollutants would be a condition of the licence.

22



Risk Management




Review and

reality check

Review and

reality check


24


3

3.1 IntroductionIssue identification identifies issues for which riskassessment is useful and establishes a context forthe risk assessment by a process of identifying theconcerns that the risk assessment needs toaddress. Issue Identification draws on all relevantlines of information.

Issue Identification comprises several phases:

1. identification of environmental health issues(or an individual issue) and determiningwhether there are hazards amenable to riskassessment. This will involve demarcating‘hazards’ from ‘issues’ and may requireenvironmental sampling (See Section 8.5);

2. putting the hazards into their environmentalhealth context (clarification and prioritisingof problems and hazards);

3. identification of potential interactionsbetween agents; and

4. stating clearly why risk assessment is neededand the scope and objectives of the riskassessment. This will involve identifyingproblems for which information is, or can be,available to undertake adequate riskassessments and problems which riskassessment cannot assist (ACDP, 1996;P/CCRARM, 1997).

At this stage it often becomes apparent that thesetting for the risk assessment is a situation where:

• there are multiple, interacting hazards ratherthan an isolated hazard;

• there are concerns about a range of potentialhealth effects from various hazards;

• there is variable and often superficialinformation on exposure and the level ofhealth problems; and

• there is an environment of public anxiety,anger and impatience.

A consideration of conflicts will assist in providinga context for effective risk assessment, riskmanagement, risk communication and communityconsultation. Examples of these value conflicts are:

• economic activity (e.g. jobs, property values)vs conservation and health protection;

• personal experiences and perceptions vs so-called ‘objective’ evidence;

• quality of life and aesthetics vs defineddisease problems;

• local control and involvement vs externalcontrol structures;

• local concerns vs national/statewide/regionalconcerns;

• monitoring and health data vs personalexperience;

• personal experience vs scientific literature inmaking causal inferences;

• broad community concerns vs narrow interestgroups;

• urgency vs priority determination;

• political activism vs incremental, scientificanalysis; and

• voluntary exposure hazards vs involuntaryexposure hazards.

Communication and consultation is important sothat these conflicts are resolved.

When issues have been identified, a preliminaryqualitative risk assessment may be carried out toprioritise issues for more detailed study. This willconsider the likelihood of exposure and thepossible consequences taking into account thingssuch as biological plausibility, evidence ofexposure and community concerns.

3.2 Identification ofEnvironmental Health Hazards

Environmental health hazards may be caused byphysical, chemical, biological or social factors inthe environment.

Physical factors include heat, cold, noise,mechanical hazards, solar radiation, ionisingradiation (e.g. X-rays) and non-ionising radiation(e.g. microwaves), noise and vibration.


25


Chemical factors include synthetic and naturallyoccurring substances.

Biological factors include viruses, prions, bacteria,parasites and vermin.

Social factors include poverty and unemployment.

Hazardous agents may be identified from range ofdata sources including:

• environmental monitoring (e.g. of food, air,water and soil);

• emissions inventories (e.g. the NationalPollutant Inventory);

• biological monitoring (e.g. of children’s bloodlead levels or Ross River Virus antibodylevels);

• disease surveillance (e.g. of Salmonella typesfor food poisoning, skin cancer rates,pregnancy outcomes);

• health monitoring (e.g. of lung functiontesting to detect the onset ofenvironmentally-caused asthma);

• epidemiological studies (e.g. of particulardisease rates in certain populations such asworkers) to identify previously unknownhazards; and

• information about analogous hazards.

3.3 Environmental Samplingand Analysis

Environmental sampling and analysis is a keyfactor in identifying the agents that may bepresent, their concentrations and distributions.The results of initial environmental sampling andanalysis will assist in identifying issues and willinfluence the direction of the risk assessment.It will be particularly important in the Exposure

Assessment phase and detailed information isprovided in Section 8.

3.4 Putting the Hazards into their EnvironmentalHealth Context

This entails a consideration of:

• Whether the hazard has a single or multiplesources (e.g. atrazine contamination of adrinking water supply from a chemical spill vsparticulates in ambient air arising from dieselengines, wood stoves and environmentaltobacco smoke);

• Whether the contaminant affects multipleenvironmental media (e.g. lead smelteremissions contaminating soil, air, water and food);

• How do stakeholders perceive the problem? Do different groups have different perceptions? A stakeholder group comprised of workers ata smelter who are also nearby residents mayhave complex perceptions; and

• How do the hazards compare to otherenvironmental hazards affecting thecommunity? This component of the appraisalwill be affected by objective data (e.g. ofdifferent disease rates) and subjectiveperceptions by the stakeholders(P/CCRARM, 1997). It enables the priorityorder of risk assessment to be determined.

There may be multiple iterations of hazardappraisal as the risk assessment proceeds and newinformation and perspectives emerge.

3.5 Identification of Potential Interactionsbetween Agents

There may be interactions between the physical,chemical, biological and social hazards that needto be identified and considered as part of the riskassessment. For example malnutrition mayincrease the absorption of cadmium and hencethe risk of renal dysfunction. A high zinc intakemay reduce the gastrointestinal absorption ofcadmium reducing the risk from highenvironmental levels. People who carry the sicklecell anaemia gene have a reduced risk of malaria,while people with the genetic condition ofWilson Disease will have a greatly increased riskfrom environmental copper.

26

There are several potential types of interactionbetween hazardous agents:

• Additive where the combined effect of two ormore agents is equal to the sum of theindividual effects e.g. 2+3=5. An example ischolinesterase inhibition from simultaneousexposure to two organophosphorusinsecticides;

• Synergistic where the combined effect of twoor more agents is much greater than the sumof the individual effects e.g. 2+2=20.Examples are risk of lung cancer fromasbestos and smoking and the hepatotoxicityof carbon tetrachloride and ethanol;

• Potentiation where one agent alone does nothave a toxic effect but, when given withanother agent, results in a much greater toxiceffect from the other agent e.g. 3+0=8.An example is risk of cancer from an initiatorand a promoter (tobacco smoke containsboth); and

• Antagonistic where the combined effect oftwo or more agents is less than the sum ofthe individual effects (Hodgson et al, 1998).An example is risk of cyanide toxicity fromcyanide after receiving an antidote such asKelocyanor (Klaassen, 1996).

The potential hazards from interactions betweenchemicals are widely discussed but there are nogenerally accepted methods for predictiveappraisal of interactions as part of the riskassessment process (See Section 6.9).

3.6 Stating Why RiskAssessment is Needed

In some instances the hazard and need for actionwill be so obvious to all stakeholders that riskassessment will be undertaken only to determinethe effect and cost-effectiveness of the variousmanagement options. In this situation, theopportunity costs of undertaking a riskassessment to determine whether action isnecessary are considerable. In other instances riskassessment will be inappropriate as the solutionsto the problem will not be based on addressing

risk but on addressing other factors such as socialand political concerns.

In deciding to undertake a risk assessment thefollowing matters must be clear:

• what is the concern?

• why is it a concern?

• how urgent is the concern?

• how do stakeholders perceive the concern?(P/CCRARM, 1997)

Risk assessment is inappropriate when it is aritual rather than a meaningful process andshould not be undertaken when:

• there is no data or an insufficient amount ofdata;

• there is an inability to take action or it is toolate to take action;

• there are insufficient resources; and

• it is politically or socially unacceptable.

Of relevance to risk assessment is Bardwell’sreference (1991) to a study that indicates that ‘about90 per cent of real world problem solving is spent:

• solving the wrong problem;

• stating the question so that it cannot beanswered;

• solving a solution;

• stating questions too generically; or

• trying to get agreement on the answer beforethere is agreement on the question’(Bardwell, 1991 cited in Thornton andPaulsen, 1998, p. 799).


27


3.7 Limitations andUncertainties

At this stage it may become apparent that thereare limitations to the proposed risk assessmentsuch as:

• information gaps e.g. effects of mixtures, lowlevel and variable exposures over time;relative contributions of ‘lifestyle’ factors vsother environmental hazards; variations insensitivity;

• poor exposure information e.g. complexmixtures of hazards with complex behavioursin the environment; limited knowledge aboutthe actual or potential population andsensitive sub populations; geographicvariations in exposure;

• limitations of toxicological andepidemiological research e.g. smallpopulations, limited exposure information;multifactorial causes of many diseases;‘background noise’ affecting research intocommon diseases or symptoms; populationheterogeneity; expensive and time consuming;

• complexity e.g. large number of combinationsof hazards, exposures and health states;

• complex causality for many health conditions;

• confidentiality of health and commercialinformation;

• charged atmosphere of fear, antagonism anddistrust; and

• value conflicts.

28



Risk Management




Review and

reality check

Review and

reality check


30

Hazard Assessment— Part 1:Hazard Identification—Toxicology

4

4.1 IntroductionThere are two elements to the toxicologicalassessment: hazard identification anddose–response assessment.

Hazard identification examines the capacity of anagent to cause adverse health effects in humansand other animals (US EPA, 1995). It is aqualitative description based on the type andquality of the data, complementary information(e.g. structure–activity analysis, genetic toxicity,pharmacokinetic), and the weight of evidencefrom these various sources (ibid). Key issuesinclude (ibid):

• nature, reliability and consistency of humanand animal studies;

• the availability of information about themechanistic basis for activity; and

• the relevance of the animal studies tohumans.

The dose–response assessment examines thequantitative relationships between exposure andthe effects of concern. ‘The determination ofwhether there is a hazard is often dependent onwhether a dose–response relationship is present’(ibid). Important issues include:

• the relationship between the extrapolationmodels selected and available information onbiological mechanisms;

• how appropriate data sets were selected fromthose that show the range of possiblepotencies both in laboratory animals andhumans;

• the basis for selecting interspecies scalingfactors to account for scaling doses fromexperimental animals to humans’;

• relevance of the exposure routes used in thestudies to a particular assessment and theinterrelationships of potential effects fromdifferent exposure routes;

• environmental conditions (pH, organicmatter, clay content, temperature);

• the relevance to the assessment of ‘theexpected duration of exposure and theexposure durations in the studies forming thebasis of the dose–response assessment’; and

• ‘the potential for differing susceptibilities inpopulation subgroups’ (ibid)

Both qualitative and quantitative toxicityinformation is evaluated in assessing ‘the incidenceof adverse effects occurring in humans at differentexposure levels’ (US EPA, 1989, p. 1.6).

Hazard identification uses:

• Animal data. This is usually assessed bytoxicological methods.

• Human data. This is usually assessed byepidemiological methods when groups ofpeople are involved, or by toxicologicalmethods when using case studies and acutechamber studies.

• Other data. This includes data such asstructure-activity data or in vitro dataassessed by toxicologists.

The data may come from a range of sources suchas: ad hoc data, anecdotal data, case-report dataand data collected from epidemiological registries(such as cancer or pregnancy outcome data). Ineach instance the quality of the study design andmethodology and the resulting data will need tobe rigorously assessed.

This chapter on toxicological evaluation is basedin part on the draft OECD Monograph, GuidanceNotes for the Analysis and Evaluation of Repeat-Dose Toxicity Studies, prepared for the OECD bythe Chemicals Unit, Department of Health andAgeing, Canberra, Australia, in cooperation withthe US EPA and the Canadian Pest ManagementRegulatory Agency (PMRA).

This chapter focuses on chemical hazards and inparticular on some of the problems and pitfallswhich may arise during an assessment of possiblecompound-related changes in parametersmeasured in toxicology studies conducted on achemical or substance. It is intended to provideguidance on the process of hazard identificationand assessment.


31


Toxicology studies have been designed to permitdetermination of toxic effects associated withexposure to chemical hazards. Such studies canprovide information relating to toxic effects andpotential health hazards likely to arise from singleor repeated exposures, in terms of predictingpotentially important toxicity end points andidentifying potential target organs or systems.It is important to note that, over time, the scientificcommunity will gain a better understanding of themechanisms of toxicity and this may lead tochanges in both methodology and interpretation ofhazard data; analysis and evaluation of toxicitystudies should reflect scientific consensus at thetime the data are reviewed.

4.2 Toxicity Testing—Major in vivo Study Types

Hazard identification mostly relies on the resultsof in vivo toxicity studies conducted according tostandard protocols e.g. OECD Test Guidelines(OECD, 1998). The following types of studiesare defined:

• Acute toxicity studies are studies whichinvestigate the effects of single doses of asubstance. The LD50 test, or medium lethaldose test (OECD test Guideline 401) whichrecords gross toxicity and mortality data overa 14 day post-dosing period, has beencommonly employed, but newer tests (‘limit’tests and ‘up-and-down’ dosing methods) arenow favoured as they reduce the numbers ofanimals required and reduce the sufferingseen in the classical LD50 test. OECD TestGuideline 420 covers acute oral toxicitydetermination by the ‘Fixed Dose Method’,TG 423 by the ‘Acute Toxic Class Method’,and TG 425 by the ‘Up-and-DownProcedure’.

The standard acute toxicity studies includetests for: acute oral, dermal and inhalationaltoxicity, eye irritation, skin irritation and skinsensitisation. Such studies may serve as thebasis for classification and labelling of aparticular chemical or mixture, and serve asan initial guide to possible toxic modes of

action and in establishing a dosing regimen insub-chronic toxicity studies.

• Sub-chronic toxicity studies are short termrepeat-dose studies. A short-term study hasbeen defined (WHO, 1990) as ‘having aduration lasting up to 10 per cent of theanimal’s lifespan, 90 days in rats and mice, or 1 year in dogs’, although the US EPAconsiders a 1 year dog study to be a chronicstudy. The main purpose of sub-chronic testingis to identify any target organs and to establishdose levels for chronic exposure studies.

• Chronic toxicity studies, or long-termstudies, are defined as studies lasting for thegreater part of the lifespan of the testanimals, usually 18 months in mice, 2 years in rats (WHO, 1987; 1990). The protocol forthese studies may cover the investigation ofchronic toxicity or carcinogenicity, or both.

• Reproductive toxicity studies are studiesdesigned to provide general informationabout the effects of a test substance onreproductive performance in both male andfemale animals, such as effects on matingbehaviour, gonadal function, oestrous cycling,conception, implantation, parturition,lactation, weaning and neonatal mortality.These studies may also provide someinformation about developmental orteratogenic effects of the test substance. Theconduct of and the results from these studiesare very important to assess with care, sincethe reproductive process is critical forperpetuation of the species and factors oragents that alter or disrupt this process canhave devastating consequences, both in factand in public perception (Korach, 1998). Forinformation on study design, refer to OECDTest Guideline 415, One-GenerationReproduction Toxicity Study; and 416, Two-Generation Reproduction Toxicity Study:(OECD, 1998)

• Developmental toxicity studies are studieswhich examine the spectrum of possible inutero outcomes for the conceptus, includingdeath, malformations, functional deficits and

32

developmental delays (Tyl and Marr, 1997).Exposure during sensitive periods may alternormal development resulting in immediateeffects, or may subsequently compromisenormal physiological or behaviouralfunctioning later in life. Since somedevelopmental processes can occur peri- orpostnatally, protocols for developmentalstudies are being reviewed with the possibilityof extending the dosing period indevelopmental toxicity studies from theperiod covering major organogenesis to coverthe perinatal and early postnatal period.

• Genotoxicity studies are designed todetermine whether test chemicals can perturbgenetic material to cause gene or chromosomalmutations. A large number of assay systems,especially in vitro systems, have been devisedto detect the genotoxic or mutagenic potentialof agents. Most authorities consider that aminimum set of data is required to define amutagen/non-mutagen. These data usuallyconsist of gene mutations in bacteria andmammalian cells and in vitro and in vivocytogenetics. Newer assays which couldprovide additional information include theComet assay, mutations in transgenic animals,fluorescent in situ hybridisation and celltransformation (IARC, 1999).

4.3 Important Issues in ToxicityTesting and Assessment

4.3.1 Study protocol and design

Dosing regimen

The purpose of toxicity studies is the detection ofvalid biological evidence for any toxic and/oroncogenic potential of the substance beinginvestigated. Therefore, protocols shouldmaximise the sensitivity of the test withoutsignificantly altering the accuracy andinterpretability of the biological data obtained.The dose regimen has an extremely importantbearing on these two critical elements.

Since the determination of dose responses for anyobserved effects is one of the objectives of repeat-

dose studies, at least 3 dose levels are normallyrequired, as well as controls. US EPA guidelinesallow a limit dose of 1000 mg/kg in chronic andsub-chronic studies; if this dose produces noobservable toxic effects and if toxicity is notexpected, based upon data on structurally-relatedcompounds, then a full study using three doselevels might not be considered necessary. Ideally,the dose selection should maximise the detectionof potential dose–response relationships andfacilitate the extrapolation of these to potentialhazards for other species including humans. Thelargest administered dose should not compromisebiological interpretability of the observedresponses. For example, it is generally consideredthat the upper dose should not:

a) cause a body weight decrement fromconcurrent control values of greater than10–12 per cent;

b) in a dietary study, exceed 5 per cent of thetotal diet because of potential nutritionalimbalances caused at higher levels or;

c) produce severe toxic, pharmacological orphysiological effects that might shortenduration of the study or otherwisecompromise the study results;

d) in a carcinogenicity study, alter survival in asignificant manner due to effects other thantumour production.

The International Life Sciences Institute (ILSI)‘Risk Sciences Working Group on Dose Selection’has published its deliberations on the selection ofdoses in chronic rodent bioassays (Foran JA andthe ILSI Risk Sciences Working Group on DoseSelection, 1997).

Although it has been argued that responsesobserved at doses far in excess of levelsexperienced under real or potential exposureconditions legitimately fall within the classicaldose–response concept, there are valid scientificconcerns that such doses introduce a further layerof uncertainty into the already difficult task ofevaluating animal dose responses and theassessment of their relevance to human hazardidentification and risk (Paynter, 1984). High


33


doses which overwhelm normal mechanisms formetabolism, detoxification and/or excretion, orproduce severe tissue damage (i.e. necrosis,demyelination) can make interpretation difficultor lead to inappropriate conclusions about theextent of the hazard.

With respect to the selection of the low dose, it iscommonly accepted that the lowest dose shouldnot produce any evidence of toxicity (i.e. allowsthe establishment of an NOAEL).

Dosing route

For repeat-dose studies, the most convenientroute of administration is by dietary admixture.However, depending on the possible route ofexposure of the public or occupationally exposedworkers to a chemical or an environmentalcontaminant, it may need to be investigated bythe dermal and/or inhalational route.

For dermal exposure the material, in a suitablevehicle, is applied to the clipped skin of rats,rabbits or guinea-pigs; OECD test guidelines (no. 410) recommend even application to an arearepresenting about 10 per cent of the total bodysurface area. The site is generally occluded withpolyethylene sheeting and gauze patches, or semi-occluded, in order to prevent dislodgment ofmaterial and oral ingestion, which could affect thevalidity or usefulness of the study. For volatile orsemi-volatile materials, application and coveringprocedures should minimise the possibility ofevaporation. Useful chapters or sections ondermal toxicity testing may be found in textbookson toxicology e.g. Derelanko and Hollinger(1995) and Hayes (1994).

The surface area of the respiratory membrane islarge, estimated at approximately 50–100 squaremetres in the normal adult compared with theestimated area of the small intestine at 250 squaremetres (Guyton, 1991) and much more air (about5000 times, by volume) is inhaled each day thanfood or water is ingested (McClellan andHenderson, 1989). Thus, exposure to airbornematerial is potentially greater than via dermal ororal exposure. Airborne material can be gases orvapours, liquid droplets or solutions, aerosols

(solid and vapour components), or dry fibres orpowders. As a consequence, to conductinhalational toxicity studies, mechanisms neededto deliver chemicals to a test chamber in a formthat can be inhaled are quite complex, particularlywhen coupled with the need to include measuringdevices which can establish particle size,concentration and form of the material in theexposure chamber. Furthermore, many factors caninfluence the inhalation, deposition and retentionof inhaled materials in the respiratory tract. Thus,the conduct of inhalational studies is considerablymore complex than equivalent studies by thedietary or dermal routes.Of critical importance, in both the conduct andassessment of such studies, in the need toestablish what portion of the material delivered tothe exposure chamber was in a respirable form. Inaddition to standard toxicology texts, some usefulspecific references on inhalation toxicologyinclude McClellan and Henderson (1989),Mohr et al (1988) and Salem (1987).

Study findings—Physiological,

pharmacological, or toxic?

In conducting an hazard assessment, the evaluatorneeds to determine whether effects seen in studiesare physiological, pharmacological or toxic.

Responses produced by chemicals in humans andexperimental animals may differ according to thequantity of the substance received and theduration and frequency of exposure e.g. responsesto acute exposures (a single exposure or multipleexposures occurring within twenty four hours orless) may be different from those produced bysub-chronic and chronic exposures. Not allobserved responses within a study, irrespective ofexposure duration or frequency, will representtoxicity per se. They may encompass a range ofeffects from physiological throughpharmacological and toxic manifestations.Although it sometimes may be difficult to make aclear distinction between these responses, anattempt to do so should be made. If an evaluatoris uncertain of the type or the biologicalsignificance of a response, he/she should nothesitate to obtain competent advice for resolving

34

the uncertainty. It is essential that all relevanttoxicity end points (statistically and/orbiologically significant) be identified forconsideration when evaluating data for thepresence or absence of non-toxic levels.

Physiological responses vary within limits whichare in accord with the normal functioning of aliving organism; examples of such response are theusual respiratory and pulse rate increasesassociated with increased physical activity,systemic changes associated with normalpregnancy, and those associated with homeostaticmechanisms. These variable factors are notimportant toxicity end points in sub-chronic andchronic exposure studies unless their fluctuationsare abnormally altered by a dose regimen. If suchalterations occur at a particular dose or are part ofa dose–response relationship, they should becorrelated with other toxicity end points whichmay be present.

Pharmacological responses are alteredphysiological functions arising from interaction ofa substance with a cellular receptor site, arereversible, and are usually of limited durationfollowing removal of the stimulus. Whilst some ofthese responses may be undesirable under certaincircumstances, they are distinguished from toxic(adverse) responses by generally not causinginjury. An example of this type of response is theincreased activity of the hepatic cytochromeP-450 containing mono-oxygenase systems(enzyme induction) caused by exposure to manypesticides, industrial chemicals, and drugs (noting,however, that while not a direct adverse effect, acytochrome P-450 inducer can, for example, alterhormonal homeostasis and effect tumourpromotion, or increase an organism’s susceptibilityto other chemical exposures).

Toxic responses may be reversible or irreversiblebut are distinguished from other types ofresponses by being injurious and therefore adverseand harmful to living organisms or tissues. Achemical which causes a physiological orpharmacological effect may produce a toxicresponse if the exposure is prolonged and/or if thedose is increased beyond a certain level.

The reversibility or otherwise of such responsesmay also depend on these two factors. Thereversibility or irreversibility of a histopathologicalchange will depend on the ability of the injuredorgan or tissue to regenerate. For example, liverhas a relatively great ability to regenerate andmany types of injury to this organ are reversible.By contrast, differentiated cells of the centralnervous system are not replaced and many injuriesto the CNS are irreversible.

4.4 Assessment of the Qualityof the Data Characterisingthe Hazard

The following considerations address theacceptability of experimental studies and thedocumentation provided.

1. The adequacy of the experimental design andother experimental parameters, including: theappropriateness of the observational andexperimental methods; frequency andduration of exposure; appropriateness of thespecies, strain, sex and age of the animalsused; the numbers of animals used per dosagegroup; justification of dose, route andfrequency of dosing; and the conditions underwhich the substance was tested.

2. There are many guidelines to the generationof scientifically valid data which concerngood experimental design, laboratory practiceand reporting e.g. OECD and US EPAguidelines, and accepted codes of GoodLaboratory Practice, or GLP (OECD, 1982;US EPA, 1983). They can be useful as aids indetermining report and data acceptability.However, the evaluator needs to make ajudgement about how well the study, in toto,facilitates the identification of potentialadverse effects, or lack thereof, of thesubstance being evaluated, rather than howprecisely it fits a prescribed test guideline or‘recipe’. The experience of senior evaluatorscan be helpful in resolving concerns aboutacceptability of study conduct and/orreporting.


35


3. The competency and completeness of theconduct and reporting of the study.

4. The effects of modifying factors which mayresult in major inequalities between controland test animals.

This qualitative consideration has more to dowith the evaluation and interpretation of datathan with acceptability of documentation. It isplaced here because determination of the factorswhich may have a major influence ontoxicological data needs to be made prior to theanalysis of the data. There are many factorsinfluencing the responses of experimental animalsto experimental treatment; some of these arediscussed by Doull (l980). Some influences maybe quite subtle, as exemplified by studiesperformed by Thompson et al (1982), in which itwas noted that the onset of acute pulmonaryoedema in rats being used in immunehypersensitivity studies was sudden and seasonal.Circadian rhythms and seasonal physiologicalvariations can subtly influence experimentalresults. Such factors influencing animal responsescan be troublesome when their effects areconfused with or misinterpreted as toxic responsesto treatment. For further discussion ofenvironmental effects on experimental parameterssee Herrington and Nelbach (1942).

The acceptability of reports and other technicalinformation is primarily a scientific judgement.Therefore, the rationale for rejecting a hazardassessment study should be succinctly stated inthe evaluation document.

4.5 Analysis and Evaluation ofToxicity Studies

Useful guidance documents for evaluating dataand conducting assessments include the IPCSEnvironmental Health Criteria (EHC)monographs viz. EHC 6, 70, 104 and 141(WHO, 1978; 1987; 1990; 1992).

4.6 Analysis and Evaluation ofMajor Study Parameters

Not all observed effects of test substances aretoxic effects. Rather, they may be adaptive (e.g.liver enzyme induction leading to some hepaticenlargement) or may be a manifestation of apharmacological effect (e.g. in an animal colonysuffering from various low-grade infections, anantibiotic will lower leucocyte counts in treatedanimals relative to controls; obviously it is notappropriate to describe this as a leukopaenic effectof the chemical).

Concurrent control groups should always be used;notwithstanding the value of historical controlranges in tumorigenicity studies. It is generallynot appropriate to rely on statistical comparisonswith historical controls since the incidence ofspontaneous lesions can vary significantly overtime (and even between concurrent randomisedcontrol groups). Controls must be age-matchedbecause some forms of toxicity represent no morethan acceleration and/or enhancement ofage-related changes. Examples of pathologicalchanges in aged rats which may be affected bycompound administration include chronicprogressive glomerulonephropathy, peripheralnerve degeneration, amyloidosis and variousneoplasms.

The use of non-treated and vehicle-controlgroups aids assessment of effects due to vehicle orexcipients. When a vehicle is used to deliver thedoses of the agent understudy (e.g. a lipophilicagent delivered in corn oil), the need for vehicle-treated controls is paramount. Since someparameters can be affected by animal handling(e.g. serum ALT was raised in mice which weregrasped around the body compared withunhandled or tail-handled mice; Swaim et al,1985), control animals should be treated in thesame way as test animals.

Control animals must receive as much attentionduring the analysis and evaluation process as dothe treated ones. Any untreated animal or groupmay exhibit some signs of abnormality or driftfrom the norm for that species or strain. Becauseof the possibility that statistically significant

36

differences between treated and control groups arethe result of abnormal values among the controls,such differences should usually be dose-relatedand should delineate a trend away from the normfor that stock of animals, if they are to beindicative of a true compound-related effect.

Historical control data may be useful whenevaluating the acceptability of the ‘normal’ dataobtained from control groups (Haseman et al,1984; Paynter, 1984; Sumi et al, 1976; Tarone,1982). Any departure from the norm by thecontrol groups should be taken into consideration,especially during the conduct of any statisticalanalysis. The finding of consistent departuresfrom the norm in control groups may necessitateinvestigation of the source of the animals.

Ideally, historical control data should be takenfrom the same laboratory, utilising the samestrain, age and sex of animals obtained from thesame supplier, and only include those studiesconducted within a 2 to 3-year span on either sideof the study under review, with identification ofstudy methodology (e.g. pre-sampling conditionssuch as fasting or non-fasting, assay methodologyfor study parameters, histopathological criteria forlesion identification, time of terminal sacrificeetc.) which could have affected the results.

Weil and McCollister (l963) analysed toxicity endpoints, other than oncogenicity, from short- andlong-term tests and concluded that only arelatively small number of end points wereeffective in delineating the lowest dosageproducing an effect in such tests. Body weight,liver weight, kidney weight, and liver pathologydelineated this dosage level in 92 per cent of testchemicals in short-term (sub-chronic) studies and100 per cent in long-term (chronic) studies. Toreach 100 per cent efficiency in short-termstudies, renal and testicular histopathology had tobe included. Heywood (l981) surveyed thetoxicological profiles of fifty compounds in rodentand non-rodent species and confirmed theseobservations.

4.6.1 Mortality/survivalReasonable efforts should be made to determinethe cause or likely cause of individual deaths. Theevaluation of pathological lesions ormorphological changes in belatedly-observeddeaths are frequently complicated by post-mortem autolysis. The separation of deathscaused by factors unrelated to exposure to the testagent (e.g. acute or chronic infections, age ordisease-related degenerative processes, anatomicalabnormalities, negligent handling or accident)from toxicity-induced deaths is important. Alldata relating to moribund or dead animals duringtheir study life, as well as the results of post-mortem examinations, should be scrutinised in anattempt to make this distinction. Note that USEPA guidelines state that the highest dose used insub-chronic studies with non-rodents should notproduce an incidence of fatalities which wouldprevent meaningful evaluation.

Analysis of mortality requires more than astatistical treatment of incidence at termination ofa study. Survival/mortality data can be influencedby factors other than the test substance. Changesin the protocol during the course of a study cancomplicate the analysis e.g. alterations in dosagelevels can produce a confusing mortality pattern.

Any unusual mortality pattern should beexplained by the test laboratory on biological ortoxicological grounds. If overall mortality is high(i.e. significantly greater than expected for theparticular colony and strain) for any repeat-dosestudy, or for a particular group within a study, acredible explanation should be provided.

An evaluation of mortality patterns within eachgroup is important. Such patterns may indicatethat mortality is clustered early or late in thecourse of the study, is intermittent and scatteredthroughout the duration of the study, or has ahigher incidence in one sex than in the other. Theanalysis of the cause of individual deaths will aidin determining the toxicological significance ofthese various patterns. Early deaths within treatedgroups may just reflect deaths of the moresusceptible animals in the test population.Alternatively, it may indicate changes in


37


compound intake per unit body weight, in thoseexperiments in which the quantity of testsubstance in the diet is kept constant. Relative tobody weight, young rats ingest more food thanolder rats and hence, young rats ingest relativelymore of the test substance than do older rats.Early deaths may therefore be the result of thehigher exposure, on a mg/kg/d basis, of younganimals compared to older animals.

Deaths which are clustered at a specific time periodmay reflect a spontaneous epidemic diseasesituation of limited duration. High mortalityassociated with infectious agents in treated groups,in the absence of such evidence in the concurrentcontrol group, could indicate an immunosuppressiveaction of the chemical being tested.

The effect of dietary intake on mortality needs tobe considered. A compound administered in thediet may make the laboratory chow more or lesspalatable, may have a pharmacological stimulantor depressant effect on appetite, or may affect thepartitioning of the nutrients in the food.Likewise, decreased water consumption (e.g. inthe case of an unpalatable compoundadministered in the water) will lead to reducedfood consumption. These effects may significantlyinfluence longevity since it has been clearly shownin animal species that long-term dietaryrestriction very significantly increases lifespan(e.g. Tucker, 1979). Conversely, excessive adlibitum intake of highly nutritious diets canreduce lifespan compared with the expectedaverage lifespan for an animal species/strain. Todate, regulatory authorities have not come to anydecision on recommending restricted diets vs. adlibitum feeding in toxicity study guidelines; someuseful references on this topic include Keenan(1998; see also other related articles by thisauthor), Klinger et al (1996), Masoro (1992), andThurman et al (1995).

4.6.2 Clinical observationsAdverse clinical signs (gross observations) notedduring the exposure period may correlate withtoxicity end points or disease processes. These canbe used as supportive evidence for dose–responserelationships and may play a role in the

determination of the NOEL/NOAEL. However,not all adverse clinical signs will correlate withpathological or morphological changes in organsor tissues. Some will be caused by biochemical orphysiological effects i.e. incoordination, muscletwitching, tremor, or diarrhoea may indicateacetylcholinesterase inhibition without anymorphological changes being evident in nervous tissue.

Many of these qualitative signs can be counted,scored for intensity, and tabulated as incidences.However, statistical analysis is of limited value.The evaluator must rely on the number ofindividuals per group exhibiting signs of aparticular type, as well as the intensity of theresponses, to gain an impression of a dose–response relationship.

Clinical observations such as palpable tumours orthose which might be associated with neoplasia(e.g. haematuria, abdominal distension, orimpaired respiration) may be useful in definingthe time a tumour was first suspected as beingpresent. Such signs might aid in the evaluation ofdecreased tumour latency in long-term rodentstudies. They may also aid in determining causeof death. A statement of the correlations, or thelack thereof, between clinical signs and specifictoxicity end points should be made in theevaluation.

Useful information on gross behaviouralobservations in laboratory animals and abnormalbehaviour patterns can be found in Bayne (1996).

The revised OECD test guidelines for 90-dayoral toxicity studies in rodents and non-rodents(Test Guidelines 408 and 409; adopted 21 September 1998) have placed additionalemphasis on neurological end-points i.e. studiesshould allow for the identification of chemicalswith the potential to cause neurotoxic effects,which may warrant further in-depth investigation.The reader is referred to the references cited inTest Guideline 408 relating to neurotoxicityassessment, including sensory reactivity to stimuliof different types (auditory, visula,proprioceptive), grip strength, and motor activity.

38

4.6.3 Body weight changes,food and water consumption

Body weight changes (gains or losses) forindividual animals and groups of animals whencompared to concurrent control changes duringthe course of a study are a criterion of someimportance (Heywood, 1981; Roubicek et al,1964; Weil and McCollister, 1963). Such changesare usually related to food intake, and analysis ofone without an analysis of the other is of limitedvalue. Weight decrement may not always berelated to toxicity per se (Seefeld and Petersen,1984). Occasionally the incorporation of the testsubstance into the diet will reduce the palatabilityof the diet to many individuals in all treatmentgroups or to the majority of individuals in thehigher dietary level groups. Food spillage needs tobe considered in the evaluation of foodpalatability and compound intake. The sameconsiderations apply if the compound isadministered in drinking water.

This effect is often evidenced during the first twoor three weeks of the study. Sometimes animals inthe affected groups(s) are able to accommodate tothe diet and a gradual increase in group weightgain will occur (Nolen, 1972). In sub-chronicstudies, the lag in group weight gain may persist,even though the individual animal gains per gramof food consumed (food utilisation efficiency) arefavourable after the accommodation, and producea statistically significant difference between theaffected group and the concurrent controls whichis not related to toxicity of the test substance(McLean and McLean, 1969). Sometimes theaddition of the test substance will interact withone or more essential nutritional elements in thediet thereby producing weight gain decrements oralterations of toxic responses (Casterline andWilliams, 1969; Conner and Newbern, 1984;Rogers et al, 1974). This phenomenon may beencountered in sub-chronic studies and whenidentified, can usually be overcome by acceptablemeans before a chronic study is initiated.Infrequently, control values for weight gain (at one or more time points) can be low, causingthe other value to appear unusually high.

Diet composition, food and water consumption,and body weight gains per se can also have animportant influence on many aspects of animalresponses including shifts in metabolic, hormonal,and homeostatic mechanisms (Kennedy, 1969) aswell as disease processes (Berg and Simms, 1960;Paynter, 1984; Ross and Bras, 1965; Tannenbaum,1940) and maturation (Innami et al, 1973), andshould be considered when unusual effects areobserved in the absence of any indication ofinjury to organs and other vital systems.

The evaluation of body weight changes andattendant effects is significantly aided by thegraphical presentation of group mean bodyweights and food consumption vs compoundconsumption (on a mg/kg body weight basis).This allows quick identification of any unusual orsudden changes in gain or loss by any group.

4.6.4 Haematological,clinical chemistry,and urinary measurements

Regulatory guidelines generally suggest thathaematological, clinical chemistry, and urinaryparameters be routinely measured in sub-chronicand chronic toxicity studies.

Normal biological variation in inter-animal valuesand their alteration in response to a variety ofinputs means that evaluators will have to contendwith much ‘noise’ in this area, and will frequentlybe presented with scattered, statisticallysignificant effects, in the absence of any evidenceof clinically significant relationships to specifictoxicity end points. To deal with ‘noise’ there is aneed to examine whether the effect noted iswithin the normal range of variation (concurrentand historical controls). Note that some of theseparameters can vary significantly with no clinicalmanifestations but others (e.g. serum potassium)have a very narrow normal clinical range andsmall differences can be important.

Frequently these data show apparently ‘random’changes in individual group(s) or, less commonly,non dose-related trends in changes across severalgroups. If using historical control data as an aid toevaluation, only values produced by the identical


39


methods from the same laboratory are valid insuch comparisons. Literature values for normalranges which do not specify the method by whichthey were obtained should be used with caution.

A good review of factors (physiological,environmental etc.) which can complicate theinterpretation of findings in a toxicity study maybe found in the Handbook of ToxicologicPathology (Bucci, 1991).

To gain maximum information from enzymedeterminations it is important to consider themost appropriate enzymes. It is important thatorgan distribution and location of the enzyme inthe cell is known. ALT (Alanine aminotransferase)is found in greatest concentration in the liver inrats, even more so in dogs. AP (Alkalinephosphatase, ALP) is virtually absent from theliver in these two species, being mainly confinedto the kidney, intestine and bone. CPK (Creatinephosphokinase, CK) is mainly located in skeletaland heart muscle, whilst AST (Aspartateaminotransferase) is found in variousconcentrations in most organs. It is clear thatCPK is the most appropriate enzyme to detectmuscle damage, while changes in ALT wouldprobably reflect some liver necrosis. AlthoughAST is not organ-specific, it serves to confirmorgan damage, especially for muscle and liver, ifits activity changes in parallel with otherenzymes. In dogs, AP is a sensitive test for biliaryfunction but in the rat it is of little diagnosticvalue since it is absent from the liver andprincipally derived from the intestines. Forhepatocellular evaluation, ALT, AST, SDH(Sorbitol dehydrogenase) and GLDH (Glutamatedehydrogenase) are the most appropriate, whilefor hepatobiliary evaluation, AP, 5’-nucleotidase,GGT (Gamma glutamyl transferase) and totalbilirubin are the most appropriate measurements.It is important to understand that many of thesetypes of serum enzyme tests and urinalysis fail todetect minor injury or may reflect only transientor reversible changes. Therefore, evaluation andinterpretation of the test results must beperformed carefully and correlated with morespecific, sensitive, and reliable histopathologicalfindings.

Sensitivity and specificity of the enzyme changesas diagnostic of organ pathology are greatlyinfluenced by the species selected for testing (seee.g. Clampitt, 1978; Tyson and Sawhney, 1985).For example, in mammalian species, aspartatetransaminase is not specific to any tissue andthereby elevated plasma AST activity may suggestdamage to any one of many tissues. In contrast,alanine transaminase is relatively specific to theliver in the cat, dog, ferret, mouse, and rat,whereas in primates, ALT is present in heart,skeletal muscle, and liver. Plasma alkalinephosphatase measurement has been less useful indetecting liver cell necrosis in the dog, sheep, cow,and rat but may be indicative of other types ofliver damage, particularly those of a cholestaticnature in a number of species. It is evident thatspecies differences are of great importance whenspecific clinical chemistries are selected forinclusion in toxicity studies.

When analysis and evaluation of clinical dataindicate a dose response relationship or abiologically important drift from concurrentcontrol values, the effects observed should becorrelated with other manifestations of toxicity.The evaluator should indicate that a correlationcould not be made, if that is the situation.

Standard veterinary (e.g. Bush, 1991; Duncan et al, 1994; Evans, 1996; Fox et al, 1984; Jain,1993) and human clinical manuals (e.g.Fischbach, 1996; Henry, 1984; Tyson andSawhney, 1985; Walach, 1996) should beconsulted for information about laboratorydiagnostic tests and to assist in the evaluation ofpotential correlations between clinical chemistry,haematological, urinary data, and adverse effects.

The deliberations of a joint internationalcommittee, established to provide advice forclinical pathology testing of laboratory animalspecies used in regulated toxicity and safetystudies, has published its recommendations,including those parameters which should bemeasured (Weingand et al, 1996). Whilst theserecommendations have not been formallyincorporated into national or internationalguidelines at this stage, they are noted, as follows:

40

For repeated-dose studies in rodent species,clinical pathology testing is necessary at studytermination. Interim study testing may not benecessary in long-duration studies provided that ithas been done in short-duration studies usingdose levels not substantially lower than those usedin the long-duration studies. For repeated-dosestudies in non-rodent species, clinical pathologytesting is recommended at study termination andat least once at an earlier interval. For studies of 2to 6 weeks in duration in non-rodent species,testing is also recommended within 7 days ofinitiation of dosing, unless it compromises thehealth of the animals. If a study contains recoverygroups, clinical pathology testing at studytermination is recommended.

The core haematology tests recommended aretotal leukocyte (white blood cell) count, absolutedifferential leukocyte count, erythrocyte (redblood cell) count, evaluation of red blood cellmorphology, platelet (thrombocyte) count,haemoglobin concentration, haematocrit (orpacked cell volume), mean corpuscular volume,mean corpuscular haemoglobin, and meancorpuscular haemoglobin concentration. In theabsence of automated reticulocyte countingcapabilities, blood smears from each animalshould be prepared for reticulocyte counts. Bonemarrow cytology slides should be prepared fromeach animal at termination. Prothrombin timeand activated partial thromboplastin time (orappropriate alternatives) and platelet count are theminimum recommended laboratory tests ofhaemostasis. The core clinical chemistry testsrecommended are glucose, urea nitrogen,creatinine, total protein, albumin, calculatedglobulin, calcium, sodium, potassium, totalcholesterol, and appropriate hepatocellular andhepatobiliary tests. For hepatocellular evaluation,measurement of a minimum of two scientificallyappropriate blood tests is recommended,e.g. alanine aminotransferase, aspartateaminotransferase, sorbitol dehydrogenase,glutamate dehydrogenase, or total bile acids.For hepatobiliary evaluation, measurement of aminimum of two appropriate blood tests isrecommended, e.g. alkaline phosphatase, gamma-glutamyltransferase, 5’-nucleotidase, total

bilirubin, or total bile acids. Urinalysis should beconducted at least once during a study. Forroutine urinalysis, an overnight collection(approximately 16 h) is recommended. It isrecommended that the core tests should includean assessment of urine appearance (colour andturbidity), volume, specific gravity or osmolality,pH, and either the quantitative or semi-quantitative determination of total protein andglucose. For carcinogenicity studies, only bloodsmears should be made from unscheduledsacrifices (decedents) and at study termination, toaid in the identification and differentiation ofhaematopoietic neoplasia.

4.6.5 Absolute and relative organ weights

It is generally considered that histopathology ismore sensitive for establishing the lowest doseproducing an effect than are organ or body weightchanges. Organ weights are usually reported asabsolute organ weights and as relative organweights (relative to body weight and/or brainweight). Relative organ weight comparisons areused since body weights are often affected bycompound administration.

Experimentally controllable and uncontrollablefactors (i.e. circadian rhythms, food intake,dehydration, nature of the diet, age of animals,organ workload, stress, and method of killing)have an influence on organ and body weights andthe variability of such data. A review of thissubject by Weil (1970) should be consulted. Themost important influencing factor appears to bethe method of killing and the timing of necropsy.The killing method used not only affects theappearance of the tissue, important in describinggross necropsy observations, but also, inconjunction with the timing of necropsies, maycause postmortem shifts in organ weights (Boydand Knight, 1963; Pfeiffer and Muller, 1967).

A not uncommon problem in interpretation ofstudy findings is the misinterpretation of relativeorgan weight changes e.g. there is no sense inreporting an increase in relative brain weight in atoxicity study in which the chemical is having asignificant effect in causing bodyweight loss or


41


reducing body weight gain because the brain willbe spared under conditions leading to reducedbodyweight, the relative brain weight willobviously increase. Similarly, other organs maychange in relative weight in a manner dependentupon body weight rather than as a result of aspecific compound effect: useful Tables of therelationship of relative organs weights to variouslevels of reduced bodyweights (produced bydietary restriction) may be found for rats inSharer (1977). When growth is markedly affectedin a toxicity experiment, alterations of organweight:body weight ratios have to be expected asa physiological response of the organism todecreased nutrient intake; such changes must bedifferentiated from organ weight changesresulting from primary toxic effects of thecompound being tested.

The interpretation of organ weight changes mustnot be made solely on the determination of astatistically significant difference between theconcurrent control value and a treatment groupvalue. A proper evaluation will also includeconsideration of any correlation between organweights (absolute and relative), histopathologicaland metabolic/pharmacodynamic data.

4.6.6 Post mortem observationAlthough much progress has been made in thestandardisation of nomenclature, to minimise anydifficulties in this area, an experienced pathologistwill describe each significant lesion type, at leastonce, in such detail that another competentpathologist can perceive a mental picture of thelesion and form a judgement as to its relevance tothe histopathology induced by the chemical being tested.

To assist in the uniform description ofpathologies, a series of articles on pathologynomenclature have been published, under the titleStandardized System of Nomenclature andDiagnostic Criteria Guides for ToxicologicPathology by the US Society of ToxicologicPathologists (STP), in cooperation with theArmed Forces Institute of Pathology (AFIP) andthe American Registry of Pathology (ARP).

Age-associated, especially geriatric, changes canhave an extremely important effect onhistopathology, as well as clinical chemistry,metabolic and pharmacokinetic parameters (Grice and Burek, 1983; Mohr et al, 1992; 1994;1996) and therefore, important overt, andfrequently subtle, influences on observedphysiological, pharmacological, and toxicresponses during the latter part of any long-termstudy. It is essential in all cases where spontaneousand/or age associated lesions are present, todifferentiate between such lesions and treatmentinduced lesions. References such as Grice andBurek (1983) and Benirschke et al (1978)(containing detailed descriptions of potentialhistopathological changes caused by toxicsubstances, spontaneous or degenerative and otherdiseases, and their incidences in experimentalanimals) are very helpful in this respect, as isadvice from a competent and experiencedpathologist.

An overview of factors (physiological,environmental etc.) which can complicate theinterpretation of morphological findings in atoxicity study may be found in the Handbook ofToxicologic Pathology (Bucci, 1991).

4.6.7 Analysis and evaluation ofstudy parameters in acute,developmental, reproductiveand special toxicity studies

Acute toxicity studies

Important end-points in acute toxicity studies areclinical signs, gross necropsy signs, and mortality;each of these end-points are discussed in detail atSections 4.6.1, 4.6.6 and 4.6.2 respectively. Sincethe purpose of acute toxicity studies has movedaway from the establishment of a strict,quantitative number for the median lethal dose toan estimate of the likely toxicity range, theemphasis is more on clinical signs and gross organpathology than on mortality.

Reproductive toxicity studies

Sections 4.6.1–4.6.6 describe the major studyparameters to be considered in repeat-dosetoxicity studies and these end-points may also

42

apply to the sires and dams in developmentaltoxicity studies. However, the critical end-pointsrelate to potential toxic effects on reproductiveparameters, including effects on mating behaviour(both sexes), on fertility (both sexes), theimplantation of blastocysts, embryonic and fetaldevelopment and survival, parturition, lactation,and postnatal survival and development. Thus, aplethora of reproductive parameters need to beassessed in one or more generations, dependingon whether the study is a one-generation (OECDTest Guideline 415), two-generation (TG 416) orthree-generation test. Important end-points toassess within each generation include: time afterpairing to mating; mating behaviour; percentageof females pregnant; number of pregnancies goingto full term; litter size; number of live births;number of stillborns; pup viability and weight atparturition, and postnatal days 4, 7, 14 and 21days of age; the fertility index (percentage ofmatings resulting in pregnancy); gestation index(percentage of pregnancies resulting in livelitters); viability index (percentage of pups thatsurvive 4 or more days); and lactation index(percentage of pups alive at 4 days that survivedto day 21 i.e. weaning); gross necropsy andhistopathology on some parents (sires and dams),with attention paid to the reproductive organs;and gross necropsy on weanlings. It is beyond thescope of this guidance to go into detail about theprocedures for the assessment of these end-points,but guidelines and procedures are welldocumented e.g. Korach (1998).

Developmental toxicity studies

The critical end-points in developmental toxicitystudies relate to potential developmental effects inutero, including death, malformations, functionaldeficits and developmental delays in fetuses. Thusthe following parameters need to be assessed; no.of live litters; no. of live fetuses/litter (total and bysex); sex ratio of fetuses; fetal body weights; litterweights; no. and percentage of fetuses withmalformations; no. and percentage of litters withmalformations; no. and percentage of fetuses withvariations; no. and percentage of litters withvariations; no and percentage of fetuses/litter withmalformations; no and percentage of fetuses/litterwith variations; and types of malformations and

variations. It is beyond the scope of this guidanceto go into detail about the procedures for theassessment of malformations and deviations butguidelines and procedures for soft tissue andskeletal examination are well documented e.g. Tyl and Marr (1997). In addition to the above developmental parameters, it is appropriateto investigate other reproductive parameters,including the following; number of. femalespregnant; number of corpora lutea/dam; numberof implants/dam; and number and percentage ofpre-implantation loss/litter. Whilst the dosingperiod in a standard teratology study commencesafter mating, conception and commencement ofimplantation, it is appropriate to check theseparameters to see that the study has not beencompromised by factors other than the compoundunder test. OECD Test Guideline 414 outlinesthe protocol for a standard developmental or‘teratology’ study.

Special studies

Different classes of chemicals may require specialtoxicology studies which are not part of the‘standard’ package of studies. For example, it iscommon to test organophosphate (OP) pesticidesfor their ability to cause delayed neuropathy byconducting tests in hens (OECD Test Guideline419), since this species is especially sensitive toinhibition of neuropathy target esterase (NTE) byOPs. Furthermore, sponsors of particularchemicals may conduct further in vitro and in vivo studies to attempt to resolve possiblemechanisms for toxic effects seen in the standardtoxicology test battery. Because of the wide rangeof types of studies which may be classified in thiscategory, it is not possible to comment on theassessment of particular end-points, but theevaluator should apply sound scientific judgementin reviewing these studies.

Toxicokinetic and metabolism data

Toxicokinetic (absorption, distribution andelimination) and metabolic data on the handling ofthe substance in the test species, can be very usefulin the evaluation and interpretation of sub-chronicand chronic exposure study data, as discussed byPaynter (1984) and references cited therein.


43


References in this paper also discussdose-dependent effects in the absorption processand in biotransformation interactions (Levy, 1968),the potential difficulties presented by impurities,the overloading of detoxification mechanisms(Munro, 1977) and other important experimentalconsiderations (Dayton and Sanders, 1983).

A number of toxicology textbooks includechapters on pharmacokinetics and toxicologyassessment e.g. Sharma and Coulombe (1996).The publication, Science and Judgement in RiskAssessment (National Academy of Sciences(NAS)/National Research Council (NRC), 1994),has useful sections on the impact ofpharmacokinetic information in risk assessment.

4.6.8 Interspecies scaling of doses(from NHMRC, 1999)

Where animal bioassays are the source of data, anestimate or measure of the human equivalent doseis required for assessing the health risks posed byenvironmental agents. To derive a humanequivalent dose from animal data, the preferredoption is to use toxicokinetic data which providesbiologically equivalent doses.

In the absence of such data, the recommendedprocedure is to scale the daily applied dose inproportion to body weight. That is, milligramsper kilogram of body weight of the experimentalanimal in the bioassay would be equivalent tomilligrams per kilogram body weight in humans.

Where oral doses are expressed in parts permillion (ppm) in the diet or drinking water, thedosage needs to be converted to mg/kg bodyweight using appropriate estimates of food orwater consumption and body weights (see WHO,1987; Faustman and Omenn, 1996).

4.6.9 Route-to-route scaling(from NHMRC, 1999)

Often the toxicological data are not available forthe most appropriate route of exposure forhumans. For example, only oral carcinogenicitydata may be available, whereas exposure to soil

contaminants by oral, dermal and inhalationalroutes may be important. Thus, extrapolationfrom one route of exposure to another may benecessary; this needs to be assessed on a case-by-case basis depending on the available data.

One important consideration in route-to-routeextrapolation is determining whether the adversehealth effects are localised to the exposure site orwhether they are a consequence of systemicdistribution. If the effects are localised at theexposure site and not a consequence of thesystemic distribution of the agent, then it may beinappropriate to extrapolate the dose to a differentroute of exposure. If the effects are consequent toabsorption and systemic distribution of the agent,then dose scaling between routes of exposureneeds to account for the bioavailability of theagent by the different routes.

Therefore, bioavailability is an importantconsideration when extrapolating the applied doseto different routes of exposure. However,additional factors may need to be considered, suchas physiological differences between species whenextrapolating, for example, from inhalationalexposure in animals to oral exposure in humans orvice versa. The assessor should includeinformation about the bioavailability of thechemical agent in the experimental studies in thefinal report.

In cases where bioavailability data are notavailable, important clues may be gained from thephysical and chemical properties and physicalstate of the agent (e.g. liquid, solid or gas).

4.6.10 Other factors in scaling of doses(from NHMRC, 1999)

For inhalational exposure, doses expressed asmg/m3 or ppm must be converted to mg/kg bodyweight in the test species by calculations based onthe physical properties of the agent and minutevolumes and respiration rates of the animal(Kennedy and Valentine, 1994). The procedurefor deriving a human equivalent dose for inhaledparticles and gases is as described by Di Marcoand Buckett (1996).

44

4.6.11 Extrapolating occupationaldata to the general public

Occupational data is often derived from arelatively homogeneous group: usually male, agedbetween 20 and 65 years and relatively healthy.When applying this data to the generalpopulation the differences between the exposedpopulations should be taken into account as thegeneral population will contain females, andpeople who are not in the workforce because oftheir age (young or old) or poor health.

4.6.12 Statistical testsThe objective of a toxicology study is todemonstrate responses of biological importance.Where statistical analyses are used in thejudgement process, an awareness of the validity ofthe test and the degree of certainty (confidence)pertaining within the context of the study shouldbe demonstrated.

There are limitations associated with the use ofstatistics in toxicology (Gad and Weil, 1986):

1. statistics cannot make poor data better;

2. statistical significance may not implybiological significance;

3. an effect that may have biologicalsignificance may not be statisticallysignificant;

4. the lack of statistical significance does notprove safety.

The importance and relevance of any effectobserved in a study must be assessed within thelimitations imposed by the study design and thespecies being studied (See also Section 5; ‘HazardAssessment Part 2: Hazard Identification—Epidemiology’).

If statistical tests have not been used, ifinappropriate tests appear to have been used, or iftests not commonly employed have been used,then this should be noted and action taken e.g.data re-analysis.

A number of textbooks and papers on theapplication of statistics in experimental toxicologyand the life sciences are available; these include

Dickens and Robinson (1996), Gad and Weil(1986), Gad and Weil (1989), Lee (1993),Salsburg (1986), Tallarida and Murray (1987) and Waner (1992).

4.6.13 General commentsDetailed comments about the analysis andevaluation of toxicology studies have been madeabove. The following further general commentsmay be made.

If possible, compound-related changes inbiochemical, haematological or urinalysisparameters should be linked with organ weight,gross pathology and/or histopathological changes.

The following points also should be noted inevaluating repeat-dose toxicity data:

Findings should be considered on the basis ofboth statistical significance and likely biologicalsignificance. The variability of biological datamust be remembered in assessing a statistically-significant result. Conversely, a finding that is notstatistically significant may have biologicalsignificance when considered in the light of thelikely toxicological or pharmacological action ofthe compound, or when combined with resultsfrom other studies. Thus, evaluators should notetrends or transient changes in parameters if thereis an indication that these may be related todosing with the compound in some way. Thisinformation may be useful when comparingresults across studies and in the consideration ofthe overall significance or relevance of anobserved effect i.e. in one study an effect may beonly a trend whilst in another study it may bevery clearly treatment-related.

A difficult problem for evaluators is the fact thatsome studies producing either clearly positive ornegative results may have to be considered asflawed. In any long-term study there may bequestionable components of the study and theexperienced toxicologist must learn to recognisewhat is useful and discard what is not. The use ofa seriously flawed negative study may provide onlya false sense of security. On the other hand, aflawed positive study may be entitled to someweight; how much is a matter of judgement (Task Force of Past Presidents, 1982).


45


Data obtained from studies carried out many yearsago should not be dismissed out-of-hand simplybecause they do not meet today’s standards; theymay provide some useful information. Again, thisis a matter for scientific interpretation andjudgement on a case-by-case basis.

4.6.14 Completion of hazard analysisAt this point the assessor should have formulatedjudgements and supporting rationale concerning:

a) the acceptability of the study and its database;

b) the existence of biologically importantadverse effects;

c) the relevance of any factors noted during theevaluation which might have had somebearing on the outcome of the study andmodified the findings in some way; and

d) the likelihood that any of the observed effectswere induced by the administered substance.

The evaluator should succinctly summarise thecritical toxicokinetic and toxicological data,together with any modifying factors for the studyunder review. The lowest, or most appropriateNOEL/NOAEL, or the absence thereof, shouldbe stated, with a clear indication of the effect(s)on which it was based (i.e. the lowest-observedeffect level or LOEL should be apparent). It isimportant to correlate findings seen in differentstudies; whilst this is done within the finalsummary of all toxicity studies, it will often beappropriate to make some mention of cross-studycorrelations (or the unexpected/unexplainedabsence of them) within individual studysummaries. Possible or proven mechanisms oftoxicity should also be discussed and included inthe summary.

4.7 Evaluation of theWeight-of-Evidence and Consideration of theToxicology Database in toto

The essential purpose of toxicity studies is thedetection of valid biological evidence of thehazard potential of the substance beinginvestigated. In this document, the evaluation ofthe weight of evidence1 produced by toxicitystudies is that process which considers thecumulative data pertinent to arriving at a level ofconcern about the potential adverse effects of asubstance. It is composed of a series ofjudgements concerning the adequacy, validity, andappropriateness of the methods used to producethe data base, and those judgements which bringinto causal, complementary, parallel, or reciprocalrelationships, all the data considered. Because ourknowledge about mechanisms of toxicity is stilldeveloping, because good epidemiologicalevidence is seldom available, and because animalstudies are not always conclusive, the informationavailable at a given time may provide only‘persuasive’ rather than ‘hard’ evidence of adefensible presumption, one way or the other,about the potential health effects of a substanceunder given conditions of exposure. Therefore, itis necessary to succinctly discuss the rationale forjudgements and conclusions contained in riskassessments together with any associateduncertainties. This becomes important when newdata or new scientific knowledge requires re-evaluation of the database or a change in aprevious risk assessment or regulatory action.

At present, there is no acceptable substitute forinformed judgement, based on sound scientificprinciples, in the analysis, evaluation,interpretation, and weighting of biological andtoxicological data derived from animal toxicitystudies conducted according to currently availableprotocols.

46

1 ‘Strength of evidence’ is commonly taken to mean the degree of conviction regarding the outcome of anexperiment eg NTP’s ‘clear evidence’, ‘some evidence’, ‘equivocal evidence’ and ‘no evidence’ of carcinogenicity.‘Weight of evidence’ involves integration of all available data, not just one study.

It is also accepted practice to apply safety oruncertainty factors to the NOEL/NOAELderived from animal studies when estimating anADI (or TDI) as an aid in evaluating theacceptability of actual or potential humanexposures. For a further discussion on this, seeSections 11.2 and 11.3 (also Dourson and Stara,1983; Paynter and Schmitt, 1979; Weil, 1972).

In addition to identifying toxic effects and thedoses at which these effects do or do not occur,toxicity studies may yield insight into the mode-or mechanism of action of a chemical toxicant.The evaluator may be able to combineinformation from a number of studies within thedatabase (e.g. metabolic/toxicokinetic, acute,short-term repeat-dose, subchronic,chronic/carcinogenicity, developmental,reproductive, and genotoxicity studies), to adduceinformation about the mode or mechanism oftoxic action of the substance.

It is at the point of overviewing the entiretoxicology database the WHO/IPCS ConceptualFramework for Cancer Risk Assessment (seeAppendix 7) is intended to be applied. This‘Framework’ is an analytical tool providing alogical, structured approach to the assessment ofthe overall weight of evidence for a postulatedmode of carcinogenic action. Use of theFramework should increase the transparency ofthe analysis by ensuring that the facts andreasoning have been documented clearly,including any inconsistencies and uncertainties inthe available data.

Note that although the Conceptual Frameworkhas been developed to assist in the assessment ofcarcinogenic end-points, the principles uponwhich it is based are broad, and should enable itsuse in analysing modes of action of non-neoplastic effects of chemicals. Irrespective of thenature of the disease process, characterising themode of action will facilitate subsequentjudgements about the human relevance of thetoxicological findings, the possible need forfurther data, risk quantification, and settingappropriate regulatory standards for the chemical.

4.8 Methods for the HazardIdentification of Carcinogens

4.8.1 Evaluation of carcinogensA variety of risk assessment methods has beenused elsewhere, for example by the United StatesEnvironmental Protection Agency (US EPA,1986), and the World Health Organization(WHO, 1993a).

Advances in biological knowledge are enablingmechanistic data, pharmacokinetic data and otherrelevant data to be increasingly taken into accountin classifying and assessing the risks ofcarcinogens.

Existing methodologies have difficulties inconveying the broad range of health implicationsof exposure to environmental pollutants. This,combined with a high ‘dread factor’ for cancer, hasresulted in many cases in a disproportionateregulatory, political and public focus on cancer ascompared to other-than-cancer health effects.

Australia uses a variety of methods for classifyingcarcinogens including the International Agencyfor Research on Cancer’s method for theclassification of carcinogens (IARC, 1978).

The International Agency for Research on Cancer(IARC) developed the first system for qualitativelycategorising chemical carcinogens (IARC, 1978).Initially, the approach was to adopt a strength-of-evidence scheme to decide whether, for humansand experimental animals separately, there wassufficient or limited evidence of carcinogenicity fora substance, mixture, or exposure circumstance, orwhether data were inadequate for classification(prior IARC monographs essentially onlysummarised existing tumourigenicity studies).Since then, the scheme has evolved whereby nowall data, including human, animal and in vitrostudies are assessed for an overall weight-of-evidence evaluation of human carcinogenicity(Vainio and Wilbourn, 1992).

A major contributor to this evolution was thedecision that ‘in the absence of adequate data onhumans, it is reasonable, for practical purposes [it is biologically plausible and prudent (IARC,


47


1987)], to regard chemicals for which there issufficient evidence of carcinogenicity in animalsas if they presented a carcinogenic risk to humans’(IARC, 1983; IARC, 1987). Thus considerableweight is given to the animal cancer bioassays,though some researchers are not convinced of thevalidity of this philosophy.

Another recent decision by IARC was toincorporate information on the mechanism ofaction of chemicals in the evaluation process(Vainio et al, 1992). For example, in practicalterms, this means that category Group 1(sufficient evidence for carcinogenicity inhumans) ‘could be extended to include agents forwhich the evidence of carcinogenicity in humansis less than sufficient but for which there issufficient evidence of carcinogenicity inexperimental animals and strong evidence inexposed humans that the agent acts through arelevant mechanism of carcinogenesis’ (Vainio et al, 1992). This aspect of the evaluation processwill become increasingly important as theunderstanding of mechanistic pathways improves;great advances are being made, especially with theadvent of sophisticated laboratory moleculartechniques. Essentially four descriptivedimensions of mechanistic data are proposed:

1. evidence of genotoxicity (i.e. structuralchange at the level of the gene);

2. evidence of effects on the expression ofrelevant genes (i.e. functional changes at theintracellular level);

3. evidence of relevant effects on cell behaviour;and

4. evidence of time and dose relationships ofcarcinogenic effects and interactions betweenagents. (Fitzgerald 1993, p. 51)

4.9 The Hazard Identification Report:Structure and Format

The hazard assessment component is likely to bebased on a number of studies, conducted indifferent species within each toxicology study typee.g. acute, chronic, developmental, or reproductivetoxicity. The report must be transparent,accountable and defensible. The quality of theHazard Identification report often determineswhether the Hazard Identification stands or falls.

4.9.1 Study identificationThe toxicity studies [or review(s)/monograph(s)]on which the hazard identification andassessment are based should be clearly identifiedin the risk assessment report. This information isimportant for the identification of the basic data[or review(s)/monograph(s)] on which the riskassessment is based.

4.9.2 Layout and formattingThe report should be structured to allow for readyaccess to all significant and relevant points arisingfrom the assessment.

Reports should be as concise and precise aspossible, consistent with adequate and transparentreporting.

48



Risk Management




Review and

reality check

Review and

reality check


50

Hazard Assessment—Part 2:Hazard Identification—Epidemiology

5

5.1 IntroductionEpidemiology and toxicology are complementaryin risk assessment. Epidemiology is the directhuman evidence component and, if based onsound epidemiological methods, can provide themost important evidence in characterising risk.Epidemiology is the principal driver inmicrobiological risk assessment.

Epidemiology is ‘the study of the distribution anddeterminants of health related states or events inspecified populations, and the application of thestudy to the control of health problems’(Last, 1988).

An excellent introductory text is:

• Beaglehole R, Bonita R, Kjellstrom T (1993).Basic epidemiology. World HealthOrganization: Geneva.

Epidemiological methods are used to investigatethe cause of adverse health effects; the naturalhistory of health conditions; the description of thehealth status of populations; and to evaluatehealth related interventions (Beaglehole et al,1993). In the context of environmental health,epidemiological methods may also be used tocharacterise population exposures, investigateperceived clusters of disease, to develop healthsurveillance programs to establish a baseline, andto monitor the consequences of risk managementactivities.

Epidemiology can assist risk assessment in severalof its stages, including:


• dose response assessment; and

• exposure assessment.

At the same time, there are often unrealisticexpectations of what an epidemiological studymay be able to achieve.

The purpose of this chapter is to provide a basisfor understanding the strengths and weaknesses ofEpidemiology in supporting risk assessment. AsMundt et al (1998) noted, if the limitations ofepidemiological studies are not understood by the

risk assessment team, the validity of anassessment might be compromised by includinginappropriate, possibly misleading,epidemiological data. The systematic appraisal ofepidemiological studies is intended to answer thequestion ‘Is there any other way of explaining theset of facts before us [i.e. the study results], isthere any other answer equally, or more, likelythan cause and effect?’ (Hill 1965 in WHO2000). Alternative explanations may result fromchance, bias and confounding (WHO 2000).

5.2 Bias and Confounding:Key Concepts in EnvironmentalEpidemiology

There are many ways in which error can beintroduced into epidemiological studies. Errormay be random (due to chance alone, andpotentially reduced by improving sample size), orsystematic (and not reduced by increasing samplesize). Whilst this section does not attempt to dealwith the subject of systematic error in any depth,the two key concepts of bias and confoundingmust be highlighted. The size of the statisticalconfidence intervals will provide an indication ofthe potential for random sampling error, butstatistical confidence intervals do not representuncertainty arising from bias or confounding.

Bias occurs if there is a systematic tendency by astudy to produce results that diverge from thetruth. There are many sources and varieties ofbias, but the most important include selectionbias and measurement (or classification) bias. Thereader is referred to Beaglehole et al (1993) for asuccinct account of bias. It may be difficult toprecisely estimate the effect bias has in a study,but it is vital for risk assessors to look for andattempt to identify the potential size anddirection of bias in interpreting a study’s findings.

Confounding is the distortion of the effect of theagent of interest by an extraneous factor(Moolgavkar et al, 1999). This may occur ifanother exposure exists in the study populationthat is associated with both the disease


51


(or outcome) and the exposure being studied e.g. a third factor (‘confounding variable’) thatindependently affects the risk of developing the disease.

There are specific approaches for the control ofconfounding that can be used in both the designand analysis of analytic studies providing that theconfounding variables have been identified andmeasured.

5.3 Types of EpidemiologicalStudy—An Overview

Broadly speaking, epidemiological activity can beeither ‘descriptive’ (reporting and describing thedistribution of exposure and effect) or ‘analytical’(designed to analyse and understand the degree ofassociation between exposure and effect).Descriptive studies include case reports, caseseries and cross-sectional surveys. Cross-sectionalsurveys measure exposure and effect in anindividual at the same point in time and thus areunable to support causal inference.

In practical terms in environmental epidemiologythere are four main categories of analytical study:

• cohort (longitudinal) studies;

• case-control studies;

• cross-sectional studies; and

• ecological studies (including a specialsubgroup known as time-series studies).(from Moolgavkar et al, 1999)

Cohort, cross-sectional and case control studiesdiffer from ecological studies in that informationon exposure and disease is available on anindividual basis. With ecological studies thisinformation is only available on a group basis, sothe community or region is the unit of analysis.

In case-control studies, exposure and otherattributes of cases of the disease underinvestigation are compared with those from asuitable control or comparison group of personsunaffected by the disease, and analysed to yieldeffect estimates. The selection of appropriatecontrols to avoid bias is a significant challenge

with case-control studies. They are relativelyinexpensive, ideal for studying rare diseases anduseful for investigating multiple, differentexposures (Gregg, 1996).

Cross-sectional studies measure the prevalence ofdisease and measure exposure and effect at thesame time. They are relatively easy and economicalto conduct and are particularly useful formeasuring fixed characteristics of individuals suchas socioeconomic status (Beaglehole et al, 1993).

Cohort studies follow cohorts or groups ofindividuals, defined in terms of their exposures,over time to see if there are differences in thedevelopment of new cases of the disease ofinterest (or other health outcome) between thegroups with and without exposure. Such studiescan be carried out by either reviewing past records(retrospective) or by tracking people into thefuture (prospective cohort). The essential featureof these longitudinal studies is that for eachindividual prior exposure information can berelated to subsequent disease experience (Breslowand Day 1987).

Ecological studies involve the investigation of agroup of people such as those living within ageographical area such as a region or state. Forexample, place and time of residence may be usedto create surrogate measures of the real exposureof interest (Elliott et al, 1992). Rates of diseaseand average exposure levels to a particular agentare determined independently, and on a groupbasis. This may give rise to spurious apparentcorrelation, called the ecological fallacy. Because itis not ascertained whether individuals who havebeen exposed to the agent are the sameindividuals who developed the disease, statementsabout causal associations are inappropriate.However ecological studies are relativelyinexpensive for linking available health data setsand environmental information and are useful forhypothesis-generation (Yassi et al, in press).Examples of ecological studies are the assessmentsof the relationship between tobacco sales indifferent countries and lung cancer rates, andfluoride in water supplies and dental caries.

52

A subset of ecological studies, known as timeseries studies, is regarded as very helpful inunderstanding the influence of short-termfluctuations in air pollutants on day-to-daychanges in population morbidity and mortalityafter controlling for factors such as season and airtemperature. However disentangling the effects ofindividual pollutants as measured in a mixturesuch as urban air pollution can be quite difficult.

To strengthen the design of ecological studies,Nurminen (1995) recommended the selection ofareas with populations that:

• are homogeneously exposed (to minimisewithin-area exposure variation);

• represent different extremes of exposuredistribution (to maximise between-areaexposure variations);

• are comparable with respect to co-variatedistributions (e.g. socio-economic status,demography); and

• use the smallest possible sampling units forecological analysis.

The largest number of environmentalepidemiology studies found in the literature are ofthe ecological or cross-sectional type, because they

are easier to carry out and cost less (Thomas andHrudey 1997). However, as noted above anddiscussed further below in relation to assessment ofcausality, such studies may be useful for identifyingpotential hazards or hypothesis generation, butthey cannot determine cause and effect.

Characteristics of the various study types aresummarised in Table 5: Epidemiological studiesare rarely definitive and a single epidemiologicalstudy cannot establish causality. A ‘weight ofevidence’ approach is generally required, involvingthe interpretation of integrated information.

Unfortunately experimental interventions such asrandomised controlled trials are rarely available toassist environmental health risk assessment. Anexample of an experimental intervention is arandomised trial of lead abatement proceduresundertaken in Broken Hill (S. Corbett, personalcommunication)

Epidemiological studies, depending on theirdesign, may serve two purposes; hypothesis-generation or assessment of a causal relationship.Their ability to evaluate a causal relationship maybe limited by a lack of control of potentialconfounders or a lack of power (which is usuallythe result of limited sample sizes) (Samet et al, 1998).


53


Tabl

e 5:

Stu

dy d

esig

ns in

env

iro

nmen

tal e

pide

mio

logy

tha

t us

e th

e in

divi

dual

as

the

unit

of

anal

ysis

Stud

y D

esig

nPo

pula

tion

Exp

osur

e H

ealth

Eff

ect

Con

foun

ders

Prob

lem

s A

dvan

tage

s

Cas

e re

port

s,C

omm

unit

y or

R

ecor

ds o

f pas

tM

orta

lity

and

Diff

icul

t to

sort

out

Har

d to

est

ablis

hC

heap

,use

ful t

oca

se s

erie

s an

d va

riou

s m

easu

rem

ents

mor

bidi

ty s

tatis

tics;

expo

sure

-eff

ect

form

ulat

e ot

her

desc

ript

ive

sub-

popu

latio

nsca

se r

egis

ters

;re

latio

nshi

pshy

poth

eses

stud

ies

othe

r re

port

s

Cro

ss-s

ectio

nal

Com

mun

ities

or

Cur

rent

Cur

rent

Usu

ally

Cur

rent

exp

osur

eC

an b

e do

nest

udy

sp

ecia

l gro

ups;

may

be

irre

leva

ntqu

ickl

y;ca

n us

eex

pose

d vs

to

cur

rent

dis

ease

larg

e po

pula

tions

;no

n-ex

pose

d

can

estim

ate

prev

alen

ce

Cas

e-co

ntro

lD

isea

sed

(cas

es)

Rec

ords

or

inte

rvie

wK

now

n at

sta

rtIf

con

foun

ders

can

Diff

icul

t to

gene

ralis

e;R

elat

ivel

y ch

eap

stud

yvs

non

-dis

ease

d of

stu

dybe

iden

tifie

d an

dm

ay in

corp

orat

ean

d qu

ick;

(con

trol

s)

mea

sure

d th

ey m

ay

bias

es;c

anno

t der

ive

part

icul

arly

use

ful

be a

ddre

ssed

ra

tes

for

stud

ying

rar

e di

seas

es

Tim

e-se

ries

stu

dy

Lar

ge c

omm

unit

y C

urre

nt e

.g.d

aily

Cur

rent

e.g

.dai

lyO

ften

diff

icul

t to

Man

y co

nfou

ndin

gU

sefu

l for

stu

dies

(sev

eral

mill

ion

chan

ges

in e

xpos

ure

vari

atio

ns in

sort

out

,e.g

.eff

ects

fact

ors,

ofte

n di

ffic

ult

on a

cute

eff

ects

peop

le);

susc

eptib

le

mor

talit

yof

influ

enza

to m

easu

regr

oups

suc

h as

as

thm

atic

s

His

tori

cal

Spec

ial g

roup

s e.

g.R

ecor

ds o

f pas

tR

ecor

ds o

f pas

t or

Oft

en d

iffic

ult

Nee

d to

rel

y on

Les

s ex

pens

ive

(ret

rosp

ectiv

e)

wor

kers

,pat

ient

s,m

easu

rem

ent

curr

ent d

iagn

osis

beca

use

ofre

cord

s th

at m

ayan

d qu

icke

r th

anco

hort

stu

dy

insu

red

pers

ons

retr

ospe

ctiv

e na

ture

;no

t be

accu

rate

pros

pect

ive

stud

y;de

pend

s on

ava

ilabi

lity

can

be u

sed

toof

pre

viou

sly

stud

y ex

posu

res

obta

ined

dat

a th

at n

o lo

nger

exi

st

Pros

pect

ive

Com

mun

ity o

r spe

cial

Def

ined

at o

utse

tT

o be

det

erm

ined

Usu

ally

eas

y to

Exp

ensi

ve,

Can

est

imat

eco

hort

stu

dy

grou

ps;e

xpos

ed v

s of

stu

dy(c

an c

hang

edu

ring

stu

dym

easu

retim

e co

nsum

ing;

inci

denc

e an

dno

n-ex

pose

d

duri

ng s

tudy

) ex

posu

re c

ateg

orie

sre

lativ

e ri

sk;c

anca

n ch

ange

;hig

h st

udy

man

y di

seas

esdr

opou

t rat

e po

ssib

lein

one

stu

dy;c

an

desc

ribe

ass

ocia

tions

th

at s

ugge

st c

ause

effe

ct r

elat

ions

hips

Exp

erim

enta

lC

omm

unity

or

Con

trol

led/

To

be m

easu

red

Can

be

cont

rolle

d E

xpen

sive

eth

ical

Wel

l acc

epte

d(in

terv

entio

n st

udy)

spec

ial g

roup

s kn

own

alre

ady

duri

ng s

tudy

by

ran

dom

isat

ion

cons

ider

atio

ns;

resu

lts;s

tron

gof

sub

ject

st

udy

subj

ects

ev

iden

ce fo

rco

mpl

ianc

e re

quire

d ca

usal

ity

or e

ffic

acy

of in

terv

entio

n

(ada

pted

from

WH

O,1

991)

54

Table 6: Applications of different observational study designs

Ecological Cross-sectional Case-control Cohort

Investigation of rare disease + + + + - + + + + + -

Investigation of rare cause + + - - + + + + +

Testing multiple effects of cause + + + - + + + + +

Study of multiple exposures + + + + + + + + + + +and determinants

Measurement of time relationship + + - +a + + + + +

Direct measurement of incidence - - +b + + + + +

Investigation of long latent periods - - + + + +c/-

Key: +…+ + + + + indicates the degree of suitability- not suitablea If prospective b If population-basedc If retrospective or historical cohort study

(adapted from Beaglehole et al, 1993)

Table 7: Advantages and disadvantages of different observational study designs

Ecological Cross-sectional Case-control Cohort

Probability of:selection bias N/A Medium High Low recall bias N/A High High Low loss to follow-up N/A N/A Low High

Confounding High Medium Medium Low

Time required Low Medium Medium High

Cost Low Medium Medium High

(from Beaglehole et al, 1993)


55

5.3.1 Observational studiesDifferent observational study designs havedifferent applications, advantages anddisadvantages (see Table 6 and 7). Thesecomparisons assume the different types of studies

are equally well designed. Even so, designvariations may affect their performance andprovide exceptions. See Beaglehole et al (1993)for a more detailed description.


5.4 Assessing the Relationshipbetween a Possible Causeand an Outcome

A cause is ‘an event, condition, characteristic or acombination of these factors which plays animportant role in producing the disease’(Beaglehole et al, 1993, p. 76).

Causation of adverse health effects is affected byfour types of factor:

• predisposing factors such as immunedeficiencies, gender and previous illness;

• enabling factors such as poor nutrition and badhousing may favour the development of disease;

• precipitating factors such as the exposure to aspecific disease agent; and

• reinforcing factors such as repeated exposuremay aggravate an established disease or state(Beaglehole et al, 1993).

The term ‘risk factor’ is commonly used to describefactors that are positively associated with the risk ofdevelopment of a disease but that are not sufficientin themselves to cause the disease. A ‘sufficient’cause is one which inevitably produces or initiates adisease and a ‘necessary’ cause is one for which adisease cannot develop in its absence (Beaglehole etal 1993). In the biological sciences there is often aconstellation of components acting in concert for acause to create an effect, and many of thecomponents of a ‘sufficient cause’ may be unknown(Rothman and Greenland 1997). At the low levelsof exposure commonly encountered in theenvironment and where there may be a range ofcontributory factors present, it may be difficult orinappropriate to assign this nomenclature to anagent even though the agent is accepted as causinga specific effect with high exposures.

As with other scientific disciplines, epidemiologyhas attempted to define a set of causal criteria tohelp distinguish causal from non-causalassociations. In the first place other explanationsfor a potentially causal association must beexcluded (such as chance, selection or measurementbias, or confounding, as mentioned in Section 5.2).Particularly rigorous scrutiny should be given tostudies giving a positive but not statisticallysignificant result. Figure 3 illustrates this process.

Figure 3: Assessing the relationshipbetween a possible causeand an outcome when anassociation is observed

(from Beaglehole et al, 1993)

56

No

Could it be due to selection or

measurement bias?

No

Could it be due to confounding?

Probably not

Could it be the result of chance?

APPLY GUIDELINES AND MAKE JUDGEMENT

Could it be causal?

If alternative explanations such as bias andconfounding can be excluded, it is then useful tosystematically apply Beaglehole et al’’s (1993)guidelines for assessing causation as shown inTable 8 below. The concepts in these guidelinesderive from work by Hill (1965) and others.However, as Rothman and Greenland (1997)note, apart from temporality (whereby a putativecause must precede the effect) there are nonecessary and sufficient criterion for determiningwhether an observed association is causal. Thusthe term ‘guidelines’ is more appropriate than theslightly more absolute ‘criteria’; and there is notnecessarily an easy epidemiological road-map tofinally determine causation.

The guidelines are similar in some respects toKoch’s postulates for determining whether anorganism is causal of a particular disease. Herequired that the putative organism was to befound in every case of the disease and that itcould be isolated and grown from cases of thedisease. Further the isolate should produce a likedisease in a new host and that in turn theorganism could be recovered from that case.Modern microbiology is now finding instanceswhere these postulates cannot be fully compliedwith, e.g. the organism can be detected using

molecular methods, but cannot be isolated, orgrown in the laboratory. Notwithstanding this,causality may be imputed by the strength andconsistency of evidence.

With environmental health in particular, muchdecision-making rests on a ‘weight of evidence’approach rather than definitive proof of cause,which is commonly not available—hence the finalconcept, ‘Judging the evidence’ in Table 8, isparticularly relevant.

These guidelines are ordered in a logical sequencefor making judgements on causality. They are notweighted equally, and their relative contributionto a final judgement will vary from one situationto another (Thomas and Hrudey, 1997)

Consistency can be demonstrated if severalstudies give the same result, especially if a varietyof designs is used in different settings since thisreduces the likelihood that all studies are makingthe same mistake. However other factors such asdifferent exposure levels or study conditions mayneed to be taken into account, and the best-designed studies should be given the greatestweight. It is important to note that inenvironmental epidemiology, reliance on a singlepivotal study is the exception rather than the rule.


57

Table 8: Guidelines for the assessment of causation

Temporal relation: Does the cause precede the effect (essential)

Plausibility: Is the association consistent with other knowledge? (mechanism of action; evidence from experimental animals)

Consistency: Have similar results been shown in other studies?

Strength: What is the strength of the association between the cause and the effect? (In general, relative risks greater than 2 can be considered strong)

Dose–response Relationship: Is increased exposure to the possible cause associated with increased effect?

Reversibility: Does the removal of a possible cause lead to reduction of disease risk?

Study design: Is the evidence based on a strong study design?

Judging the Evidence: How many lines of evidence lead to the conclusion?

(adapted from Beaglehole et al, 1993, p. 76)


The technique of meta-analysis grew out of theneed to reduce random error in clinical trials.Meta-analysis in the context of systematic reviewscan be used to pool the data from well-designedstudies, each of which may deal with a relativelysmall sample size, in order to obtain a betteroverall estimate of effect. Meta-analysis haspitfalls if poor quality studies are included, andneeds to be applied with caution to observationalstudies - which are less able to control forconfounding than randomised trials. Standardmethods for conducting and reporting systematicreviews have been published (Greenhalgh, 1997).The reader is also referred to an excellent resourcepublished by NHMRC (2000), ‘How to reviewthe evidence: systematic identification and reviewof the scientific literature’.

The strongest evidence comes from well-designedand competently conducted randomisedcontrolled trials. The National Health andMedical Research Council (1999) places strongestemphasis on evidence obtained from systematicreviews of all relevant (and well-conducted)randomised controlled trials (‘level I’).2

However there are relatively few such trialsavailable for environmental health hazards, thatcould form the basis for a systematic review. Mostapply to the effects of treatment or preventioncampaigns. A rare example is the Melbourne

Water Quality Study which was a blinded studyinvolving real and sham domestic reverse osmosiswater filters and an assessment of acutegastrointestinal disease (Hellard, 1999).

In practice, most evidence comes fromobservational studies (e.g. nearly all the evidenceon the health effects of smoking). In well-conducted cohort studies bias is minimised. Casecontrol studies are subject to several forms of biasand weaknesses related to time-sequence but, ifwell designed, may still provide useful evidencefor the causal nature of an association. Cross-sectional studies are weaker as they provide nodirect evidence on the time sequence of events.

Ecological studies are the least satisfactorybecause of the dangers of incorrect extrapolationto individuals from data derived from regions orcountries. However where certain exposurescannot normally be measured individually (e.g. airpollution, pesticides residues in food, fluoride indrinking water) evidence from ecological studiesmay be important in environmental healthdecision making (Beaglehole et al, 1993). Time-series studies demonstrating health outcomesassociated with fluctuating air pollutant levelsmay be one particularly useful example.

The above principles about strength of evidenceobtained from various study types are summarisedin Table 9.

58

2 The NHMRC document referred to is oriented towards clinical interventions and clinical practice guidelinedevelopment. At present there are no comparable endorsed ‘levels of evidence’ to guide assessment ofepidemiological evidence for environmental health practice, although the National Public Health Partnershipwith input from NHMRC is constructing a similar approach to public health interventions.

Table 9: Relative ability of different types of study to ‘prove’ causation

Type of study Ability to ‘prove’ causation

Randomised controlled trials Strong

Cohort studies Moderate

Case-control studies Weak/Moderate

Cross-sectional studies Weak

Ecological studies Weak

(adapted from Beaglehole et al, 1993)

The ranking in this Table assumes that studies arewell designed and well conducted in each case.Even the presence of a strong ability to ‘prove’causation should be supplemented by mechanisticknowledge to be confident of causation.

5.5 The Strengths andLimitations of ObservationalEpidemiology versusExperimental Toxicology

Epidemiological studies are crucial for assessingeffects directly in humans and estimatingpopulation attributable risks. However theirpower of resolution is limited, mainly because ofthe difficulties in estimating exposure preciselyand in controlling bias. Toxicological studies arenecessary for elucidating causal mechanisms,which may be important for determiningdose–response relations and extrapolating to lowdoses in risk assessment. On the other hand,direct generalisations to humans based on animaldata are often uncertain (Pershagen, 1999).

Epidemiological studies are often given increasedweighting because they come from humans but,compared to toxicological studies of animals, maybe more costly and time consuming and morelikely to result in ambiguous findings (Samet et al,1998). However substantive findings have beenobtained at times through opportunistic study ofhighly exposed groups - such as occupationalcohorts or communities that have beeninadvertently exposed to contaminants e.g. viafood or water. These can be either observationalepidemiological studies, or what Lilienfield(1980) called ‘natural experiments’.

5.5.1 Hazard identificationEpidemiology has a number of potentialadvantages over animal toxicology in the area ofhazard identification:

• it directly assesses human health risk;

• absorption, metabolism, detoxification andexcretion may vary between humans and theanimal species studied does not need to betaken into account in epidemiological studies;

• sample sizes for human studies may be muchlarger than those available for animal studies;

• genetic diversity may be broad in humanscompared to selected animal strains used intoxicological studies;

• epidemiological studies may include differentgroups (e.g. the young, old and susceptible)that may not be included in the usuallyrelatively homogeneous groups used intoxicological studies; and

• effects on some aspects of mental function orbehaviour, and more subjective effects such asnausea or headache, can be better assessed inhuman studies.

Differences in hazard identification based onepidemiological and toxicological data may beseen in the matter of ‘site concordance’. Theepidemiological data may suggest lung cancer isof concern whereas the toxicological data maysuggest liver cancer. Similar conflicts can arisewhere there are suggestions of a problem fromepidemiological data unsupported by toxicologicalevidence (Samet et al, 1998).

5.5.2 Dose–response assessmentEpidemiological data may assist in assessingdose–response relationships. Advantages ofepidemiology over animal studies may include:

• reduced uncertainty about interspeciesvariability in metabolism, lifespan, andgenetic diversity;

• complex temporal patterns of exposure anddoses in situations requiring risk assessmentmay be impossible to replicate in animalstudies; whereas some epidemiological studiesmay be more useful for understanding thesecomplex dose–response relationships; and

• the ability to assess large numbers of peopleexposed to low levels of an agent. The dosesfrom exposure to a hazardous agent inepidemiological studies are often considerablyless than in toxicological studies. This mayhave the advantage of providing informationabout the exposure range of interest although,


59


if they are the result of (prolonged) adultoccupational exposures, the exposures arelikely to be considerably more than thoseexperienced by people in the generalpopulation. With appropriate tools smalldifferences in relative risk in large populationsmay be able to be assessed.(Roseman, 1998;Samet et al, 1998)

However, epidemiological studies are oftenlimited by the amount of data available on doseand tend to address exposure–responserelationships (i.e. they are based on whether ornot exposure occurred) rather than dose–responserelationships. Doses are usually discontinuous andvariable in epidemiological studies compared tocontrolled toxicological experiments. Anintegrated measure of exposure may need to bedeveloped to represent the non-uniform doses.

Quantitative description of dose–responserelationships may be hampered by incompleteinformation on exposure (especially forbiologically relevant time windows), by exposureor dose misclassification, or by the use ofsurrogate markers of exposure. Incorrectinformation about the exposure may bias thedescription of the exposure-response relationship.If there are wide confidence intervals around theresults there can be substantially different policyendpoints depending on whether the upperbound, the lower bound or the midpoint has beenchosen for policy making (Samet et al, 1998).

Commonly too there are insufficientepidemiological data to discriminate betweenalternative models that could describe thedose–response relationship. This is particularlyimportant at very low exposure levels and this iswhere both epidemiological and toxicological dataare often limited. Surrogate measures of outcome(e.g. nerve conduction or tremor) and arelationship between the surrogate measures andhealth outcomes may need to be established inorder to interpret the significance of a study,although care needs to be taken that the surrogateoutcomes do relate to clinically meaningfuloutcomes.

The reviewer or risk assessor should answer thebasic question of whether the epidemiologic data,in an individual study or cumulatively, areadequate for use in dose–response evaluation.There is no formula or quantitative weightingscheme prescribed for making this judgement.

If epidemiologic data adequate for dose–responseevaluation are not available, and a risk assessmentis being developed for use in making animportant regulatory decision, and if it is feasibleto develop new epidemiologic data, or to extractnew data from existing studies, an effort shouldbe made to develop and provide goodepidemiologic dose–response data that can beused together with, or in preference to, high-doseanimal data.

The following ‘London Principles’ (Federal Focus,1996) may be used to guide the choice of studiesin this critical area:

• Principle 1. Dose–response assessment shouldinclude a range of reasonable dose measures,an explanation why any were rejected, andprovide a rationale if any particular dosemetric is preferred. In evaluations of bothhuman and animal data, several differentmeasures of dose should be evaluated (if possible).

• Principle 2. In the selection of adose–response model, the greatest weightshould be given to models that fit theobserved animal and human data and areconsistent with the biologically relevantmode(s) of action (genotoxic, non-genotoxic,unclassified). When mechanistic knowledge isuncertain or limited, several plausibledose–response models should be consideredand the most plausible ones, based onavailable data and professional judgement,should generally be used in dose–responseevaluation.

• Principle 3. When extrapolating cancer riskto exposure levels below the observable range,mechanistic data should be used tocharacterise the shape of the dose–responsefunction.

60

• Principle 4. When the availableepidemiologic data are not adequate toperform dose–response analyses, causing low-dose estimates of risk to be derivedexclusively from animal data, every effortshould still be made to use the availablehuman data in assessing the validity of low-dose risk estimates. To the extent feasible,heterogeneity in the human populationshould be accounted for. Whenever feasible,human data on metabolic biomarkers andother biological measures should be employedto adjust the risk estimates for knowndifferences between species and between highand low doses. If possible, data onsusceptibility should be included.

• Principle 5. When epidemiologic studies areselected for dose response assessment, higherquality studies should be given preference,especially those with precise and accurateexposure information. The availability ofinformation with respect to timing ofexposure and response (time/age of firstexposure, intensity of exposure, time totumour), adjustment for confoundingvariables, and potential interaction with othereffect modifiers is particularly important.

• Principle 6. A properly conducted meta-analysis, or preferably an analysis based onthe raw data in the original studies, may beused in hazard identification anddose–response evaluation when suchcombination includes an evaluation ofindividual studies and an assessment ofheterogeneity. The combined results ought toprovide, more than any single study, preciserisk estimates over a wider range of doses.Before using these tools, the gains should bejudged sufficient to justify potential errors ininference resulting from combining studies ofdissimilar design and quality.

• Principle 7. When epidemiological data areused in dose–response assessment, aquantitative sensitivity analysis should beconducted to determine the potential effectson risk estimates of confounders,measurement error, and other sources ofuncontrolled bias in study design.

• Principle 8. Scientific understanding ofdifferentials in human susceptibility to disease(racial/ethnic/gender/genetic differences,genetic polymorphisms, etc.) should be usedto refine the low-dose extrapolationprocedures when such phenomena areadequately understood.

5.5.3 Exposure assessmentA lack of good exposure data is a common pitfallof environmental epidemiological studies, to theextent that such studies tend only to be as good astheir exposure data. The association of particularhealth effects and specific patterns of exposure, ifin keeping with knowledge of pathophysiology,can provide strong support for causalinterpretations (WHO, 2000).

Illustrating this problem, Saunders et al (1997)reviewed 14 key relevant studies selected from ashort list of 43 analytical studies assessing humanhealth effects in relation to hazardous waste sites,and found that poor exposure measurement was amajor factor in the overall lack of convincingevidence of causation from these studies. It isoften the case that only a broad indication of thelevel or nature of exposure may be deduced fromepidemiological studies.

Experimental toxicological studies on the otherhand generally have the advantage of control andaccurate measurement of exposures. Nevertheless,at times environmental epidemiological studiesmay be the only way of determining thedistribution of ‘real-life’ exposures in terms of:

• magnitude;

• duration;

• temporal patterns;

• routes;

• size of exposed population; and

• nature of exposed population.

Future studies should be designed in such a wayas to better capture such information.


61


5.5.4 Risk characterisationThe term ‘risk’ tends to be used in a subtlydifferent and more specific way in epidemiologythan in risk assessment. In epidemiology, riskdescribes the ‘frequency of occurrence of a diseasein one population compared with another, eitheras a difference in rates (attributable risk) or as aratio of rates (relative risk)’ (ACDP, 1996, p. 20).The feature distinguishing the two populations byits presence or distribution is referred to as a ‘riskfactor’. The reliance on comparisons of diseaserates between populations creates substantiallimitations for the sensitivity of relative riskdetermination for common diseases (Thomas andHrudey 1997, p. 206).

In risk assessment, the characterisation of ‘risk’may be arrived at by a wider variety of meansthan in epidemiology.

5.6 Critical Evaluation ofPublished Research

The following section is reprinted, with minoradaptation, from ‘Introduction to research in thehealth sciences’ by Polgar S and Thomas SA(1991), p. 302–306 by permission of the publisherChurchill Livingstone. Italicised questions arefrom Riegelman 1981, p. 73 and British MedicalJournal 1988, p. 50 (with minor amendments).

5.6.1 Critical evaluation of theintroduction

The Introduction of a paper essentially reflectsthe planning of the research. Inadequacies in thissection might signal that the research project waserroneously conceived, or poorly planned. Thefollowing issues are essential for evaluating thissection:

• Adequacy of the literature review. The literaturereview must be sufficiently complete so as toreflect the current state of knowledge in thearea. Key papers should not be omitted,particularly when their results could havedirect consequences for the researchhypotheses or aims. Researchers must beunbiased in presenting evidence which isunfavourable to their points of view;

• Clearly defined aims or hypotheses. The aims orhypotheses of an investigation should beclearly and operationally stated. If this islacking, then how the evidence obtained inthe investigation is to be used for conceptualadvances in the area, will be ambiguous;

• Selection of an appropriate research strategy. Informulating the aims of the investigation, theresearcher must have taken into account theappropriate research strategy. For instance, ifthe demonstration of causal effects isrequired, a survey may be inappropriate forsatisfying the aims of the research; and

• Selection of appropriate variables. Theoperational definition of the variables beinginvestigated calls for selecting appropriatemeasurement strategies. If the selection of thevariables is inappropriate to the constructsbeing investigated, then the investigation willnot produce useful results.

5.6.2 Critical evaluation of themethods section

A well-documented Methods section is anecessary condition for understanding, evaluatingand perhaps replicating a research project. Ingeneral, the critical evaluation of this section willreveal the overall internal and external validity ofthe investigation.

Subjects

The section shows if the sample wasrepresentative of the target population and theadequacy of the sampling model used.

• Sampling model used. A number of samplingmodels can be employed to optimise therepresentativeness of a sample. If thesampling model is inappropriate, then thesample might be biased, raising questionsconcerning the external validity of theresearch findings.

• Sample size. Use of a small sample is notnecessarily a refutation of an investigation, ifthe sample is representative. However, given ahighly variable, heterogeneous population,a small sample will not be adequate to ensure

62

representativeness. Also, a small sample sizecould decrease the power of the statisticalanalysis.

• Description of the sample. Was there a power-based assessment of adequacy of sample size?A clear description of key sample variables(for example, age, sex, type and severity ofcondition) should be provided. Whennecessary and possible, demographicinformation concerning the populationshould be provided. Was the population ofadequate composition to answer the studyquestions? If not, the reader cannot judge therepresentativeness of the sample. Also, thereaders might not be able to decide if thefindings are applicable to the specific groupsof patients being treated.

Instruments/apparatus

The validity and reliability of observations andmeasurements are fundamental characteristics ofgood research. In this section, the investigatormust demonstrate the adequacy of the equipmentused for the data collection.

• Validity and reliability. The investigatorshould use standardised apparatus, orestablish the validity and reliability of newapparatus used. The lack of proven validityand reliability will raise questions about theadequacy of the empirical findings.

• Description of instrumentation. Full descriptionof the structure and use of novelinstrumentation should be presented so thatthe instrument can be replicated byindependent parties.

Procedures

Full description of how the investigation wascarried out is necessary for both replication and forthe evaluation of its internal and external validity.

• Adequacy of the design. A good design shouldcontrol for alternative interpretations of thedata. That is, a poor design will result inuncontrolled influences by extraneousvariables, negating the unequivocal evaluationof causal effects. A variety of threats tointernal validity must be considered whencritically evaluating an investigation.

• Control groups. A specific way of controllingfor extraneous effects is the use of controlgroups. If no control groups are employed,then the internal validity of the investigationmight be questioned. Also, if placebo oruntreated groups are not present, the size ofthe effects due to the treatments might bedifficult to estimate.

• Subject assignment. When using anexperimental design, care must be taken inthe assignment of subjects so as to avoidsignificant initial differences betweenexposure groups. Even when quasi-experimental or natural comparison strategiesare used, care must be taken to establish theequivalence of the groups.

• Was there a satisfactory statement given of thesource of subjects?

• Was a satisfactory response rate achieved?

• Was the assignment of people to study and controlgroups appropriate?

• Could selection bias have occurred?

• If the study was experimental, were random andblind assignment maintained?

• Regardless of the study type, were the study andcontrol groups comparable with respect tocharacteristics other than the study factors(s)?

• Exposure parameters. Was exposure adequatelydefined? It is important to describe all theexposures experienced by the different groups.If the exposures differ in intensity or in thequality of the administering personnel, thenthe internal validity of the project isthreatened.


63


• Rosenthal and Hawthorne effects. Wheneverpossible, studies should be double or singleblind. If the subjects, experimenters orobservers are aware of the aims and predictedoutcomes of the investigation, then thevalidity of the investigation will be threatenedthrough bias and expectancy effects.

• Settings. The setting in which a study iscarried out has implications for external(ecological) validity. An adequate descriptionof the setting is necessary for evaluating thegeneralisability of the findings.

• Times of exposures and observations. Thesequence of exposures and observations mustbe clearly indicated, such the issues such asseries effects and confounding can bedetected. Identification of variability intreatment and observation times caninfluence the internal validity ofexperimental, quasi-experimental or n=1designs, resulting in, for instance, internalvalidity problems.

5.6.3 Assessment of outcome• Was the assessment of outcome properly performed

in the study and the control groups?

• Was the measure of outcome appropriate to thestudy aims?

• Was the measure of outcome precise?

• Was the measure of outcome complete?

• Did the process of observation affect the outcome?

5.6.4 Critical evaluation of statistical analysis

• Was there a statement adequately describing orreferencing all statistical procedures used?

• Were the statistical analyses used appropriate?

• Was the presentation of statistical materialsatisfactory?

• Did the analysis properly compare the outcomesin the study and the control groups?

• Were the results adjusted to take into account theeffect of possible confounding variables? Commonconfounders are age and sex, regional differences,socio-economic differentials, smoking, occupation,ethnic differences.

• Was a significance test properly performed toassess the probability that the difference was dueto chance?

• Was a proper measure of the size of the differencepresented?

• Was a proper measure of the degree of overlap ofthe differences presented?

• Were the confidence intervals given for the mainresults?

Motulsky (1995) provides a useful checklist ofcommon pitfalls to bear in mind when readingresearch papers that include statistical analysis,which has been adapted as follows:

• Look at the data—summary statistics mayresult in the loss of useful information;

• Beware of very large and very small samples—large samples may generate statisticallysignificant but unimportant findings; smallsamples have little power to detect importantdifferences;

• Beware of multiple comparisons—whenanalysing random data, on average 1 out of20 comparisons will be statistically significant(p < 0.05) by chance;

• Don’t focus on averages alone: variability mayreflect real biological diversity, and outliersmay be more important;

• ‘Garbage in, Garbage Out’—statistical tests donot tell whether the study was conductedproperly;

• Confidence limits are as informative as p values(and may be more so, particularly whendealing with hazards);

• Statistical significance does not necessarilyindicate biological importance;

64

• p<0.05 is not sacred—it is an arbitrary cutoffvalue; and

• Correlation or association does not implycausation.

Section 3.7.12 also provides information on thestatistical analysis of data.

5.6.5 Critical evaluation of the results

The ways in which epidemiological data areproperly presented and analysed goes beyond thescope of this document in terms of complexityand depth, and reference should be made tostandard texts. However, the following generalpoints can be made.

The results should represent a statistically correctsummary and analysis of the data. Inadequacies inthis section could indicate that inferences drawnby the investigator were erroneous.

• Tables and graphs. Data should be correctlytabulated or drawn and adequately labelledfor interpretation. Complete summaries of allthe relevant findings should be presented.

• Selection of statistics. Both descriptive andinferential statistics must be selectedappropriately. The selection of inappropriatestatistics could distort the findings and leadto inappropriate inferences.

• Calculation of statistics. Both descriptive andinferential statistics must be correctlycalculated. The use of computers generallyensures this, although some attention must bepaid to gross errors when evaluating the data.

5.6.6 Critical evaluation of thediscussion

In the discussion, the investigator drawsinferences from the data in relation to the initialaims or hypotheses of the investigation. Unlessthe inferences are correctly made, the conclusiondrawn might lead to useless and dangeroustreatments being offered to clients.

• Drawing correct inferences from the data. Theinferences from the data must take account ofthe limitations of descriptive and inferentialstatistics. Correlations do not necessarilyimply causation, or that a lack of significancein the analysis could imply a Type II error(see below).

• Logically correct interpretations of the findings.Interpretation of the findings must followfrom the statistical inferences, withoutextraneous evidence being introduced. Forinstance, if the investigation used a n=1design, the conclusions should not claim thata procedure is generally useful.

• Protocol deviations. In interpreting the data,the investigator must indicate and take intoaccount unexpected deviations from theintended design. For instance, aplacebo/active treatment code might bebroken, or ‘contamination’ between controland experimental groups might be discovered.If such deviations are discovered byinvestigators, they are obliged to report these,so that the implications on the results mightbe taken into account.

• Generalisation from the findings. Strictlyspeaking, the data obtained from a givensample are generalisable only to thepopulation from which the sample wasdrawn. This point is sometimes ignored byinvestigators, and the findings are generalisedto subjects or situations which were notconsidered in the original sampling.

• Statistical and practical significance. Statisticalsignificance does not necessarily imply thatthe results of an investigation are applicablein practical terms. In deciding on practicalsignificance factors such as the size of effect,side effects, cost effectiveness, as well as valuejudgements concerning outcome must beconsidered.

• Theoretical significance. It is necessary to relatethe results of an investigation to previousrelated findings, as identified in the literature


65


review. Unless the results are logically relatedto the literature, the theoretical significanceof the investigation remains unclear.The processes involved in comparing thefindings of a set or related papers areintroduced in the next sub-section.

• Was a valid interpretation drawn from thecomparisons made between the study and controlgroups during analysis?

• Did the investigators properly reject or fail toreject the null hypothesis?

• Did the investigators consider the possibility ofType I and Type II errors in interpreting themeaning of the significance test? (Type I errorsare the result of chance and are the rejection of thenull hypothesis when no true difference exists inthe larger population. Type II errors result fromchance or too small a sample size and are thefailure to reject the null hypothesis when a truedifference exists in the larger population.)

• Were the size of the differences and the degree ofoverlap taken into consideration in theconclusions reached about the meaning ofobserved differences?

• In interpreting the meaning of any relationship,was the concept of cause and effect (causation)properly applied?

• Were the extrapolations to individuals notincluded in the study properly performed?

• Did the investigators stay within the limits ofthe data when extrapolating the results?

• If the investigators extrapolated from populationdata to individual data, was this appropriateand correct?

• Did the researchers take into considerationdifferences between the study population and thepopulation to which they extrapolated their data?

5.7 Evaluation of Meta-AnalysesMeta-analysis is the process of undertaking aquantitative review of the literature, seekingconsistent patterns among, and sources ofdiscrepancies between, studies (WHO, 2000). Anassessment should consider the homogeneity ofthe studies examined and whether summaryeffects estimates will be calculated and by whatmethods (ibid).

WHO (2000) has recommended that thefollowing features be considered whenconducting, or assessing the findings of, a meta-analysis:

• establishing or noting a protocol specifyingthe objectives of the review and the methodsto be employed;

• having inclusion criteria that are moreinclusive than exclusive, so enablingsensitivity analysis using different levels oninclusion to be undertaken;

• avoiding a single quality score of studies andpresenting, instead, an assessment of a rangeof characteristics;

• weighting according to the precision of thestudy;

• assessing and addressing the impact ofpublication bias;

• systematically quantifying the heterogeneityof the studies which can enable theidentification of sources of variability in theresults of studies from factors such as thechoice of methodology, and the inclusion ofsusceptible subgroups or unusual exposureconditions;

• using sensitivity analyses of factors such asdifferent analytic approaches, differentmethods of extracting results from the studiesor the inclusion or exclusion of particularstudies or types of studies; and

• appraising methods used to obtain qualitativeand quantitative summary estimates from acollection of studies.

66

5.8 Common Omissions andErrors in PublishedResearch

Rushton (2000, p. 2) provides a report on some ofthe most serious omissions and errors in paperspresented in recent years to the journal,‘Occupational and Environmental Medicine’.These are:

Design

• Authors unclear about type ofepidemiological study;

• Adequacy of sample size not considered;

• Bias in selection of subjects;

Execution

• Data collection problems and missing datanot adequately reported;

• Non-respondents not investigated;

• Sample selection and exclusions inadequatelyjustified;

Analysis

• Parametric tests carried out on obviouslyskewed data;

• Use of multiple paired tests;

• Inappropriate analysis of repeated measuresor longitudinal data;

• Incorrect analysis of matched case-controlstudies;

• Modelling incorrect—e.g. inadequateadjustment for confounders, interaction termsnot included, only significant variables frompreliminary analyses included;

Presentation

• Inadequate description of the methodologyand statistical procedures;

• Inappropriate summary statistics for non-normal data;

• No presentation of risk estimates—e.g. oddsratios—and confidence intervals;

Interpretation

• Potential bias due to sample selection, no orpoor response, missing values, exclusions;

• Lack of statistical power not considered;

• No allowance made for multiple testing; and

• Misunderstanding and misinterpretation ofresults from models.

Table 10 summarises some of the potentialproblems and their implications which mightemerge in the context evaluation of aninvestigation. A point which must be kept inmind is that even where an investigation isflawed, some useful knowledge might be drawnfrom it. The aim of critical analysis is not todiscredit or tear down published work, but toensure that the reader understands itsimplications and limitations.


67


Table 10: Checklist for evaluating published research

Problems which might be identified in a research article Possible implications

1. Inadequate literature review Misrepresentation of the conceptual basis for the research

2. Vague aims or hypothesis Research might lack direction;interpretation of evidence might be ambiguous

3. Inappropriate research strategy Findings might not be relevant to the problem being investigated

4. Inappropriate sampling method Measurements might not be related to concepts being investigated

5. Inadequate sampling method Sample might be biased, investigation could lack external validity

6. Inadequate sample size Sample might be biased; statistical analysis might lack power

7. Inadequate description of sample Application of findings to specific groups or individuals might be difficult

8. Instruments lack validity or reliability Findings might represent measurement errors

9. Inadequate design Investigation might lack internal validity; i.e. outcomes might be due to uncontrolled extraneous variables

10. Lack of adequate control groups Investigation might lack internal validity;size of the effect difficult to estimate

11. Biased subject assignment Investigation might lack internal validity

12. Variations or lack of control Investigation might lack internal validityof treatment parameters

13. Observer bias not controlled (Rosenthal effects) Investigation might lack internal and external validity

14. Subject expectations not controlled Investigations might lack internal and external validity (Hawthorne effects)

15. Research carried out in inappropriate setting Investigation might lack ecological validity

16. Confounding of times at which observations Possible series effects; investigation might lack and interventions are carried out internal validity

17. Inadequate presentation of descriptive statistics The nature of the empirical findings might not be comprehensible

18. Inappropriate statistics used to describe Distortion of data; false inferences might be drawn and/or analyse data

19. Erroneous calculation of statistics False inferences might be drawn

20. Drawing incorrect inferences from the data False conclusions might be made concerning the outcome of an investigation

21. Protocol deviations Investigation might lack external or internal validity

22. Over-generalisation of finding External validity might be threatened

23. Confusing statistical and clinical significance Treatments lacking clinical usefulness might be encouraged

24. Findings not logically related to Theoretical significance of the investigation remainsprevious research findings doubtful

(adapted from Polgar and Thomas, 1991)

68

5.9 Undertaking Health Studies

The material in the following sections is adaptedfrom ATSDR (1996).

In some situations there will be a need toundertake health studies as part of a riskassessment. A risk assessment may have beenprompted by health studies undertaken by thecommunity. The design of health studies shouldbe underpinned by epidemiological principles. Arange of factors need to be considered beforeembarking on a health study.

• Public health significancePublic health significance is a key factor inconsidering the merits of a proposed healthstudy. Issues for consideration include: thetoxicity of the agent; the pathways of humanexposure; severity and biological plausibilityof the health outcome; need for newinformation (beyond what is already knownor what has already been done); size andsusceptibility of the population affected;ability to prevent or mitigate exposure orhealth outcomes; and relevance to othersituations with similar agents and exposurepathways.

• Community perspective and involvementCommunity involvement is critical to thesuccess of any proposed health study. Variouscommunity involvement methods can be usedfor health studies. Issues for considerationinclude: an ability to involve key communitystakeholders; an understanding of communityhealth concerns; an understanding of theapproach and limitations of proposedactivities; and community support for thestudy being conducted.

• Scientific importanceScientific importance is closely related topublic health significance. Issues forconsideration include: the ability to providenew knowledge or information about anexposure-outcome relationship; to addressspecific exposures or outcomes that have not

been adequately studied; to allow newlaboratory tests or study methods to be usedor evaluated; to generalise to other situationsor populations; and provide confirmation oradditional support to a preliminaryhypothesis or theory.

• Ability to provide definitive resultsSince health studies may end up withinconclusive findings, it is important toconsider how definitive the study might be inproviding scientifically useful results relatedto specific exposure-outcome relationships.Issues for consideration include the ability to:obtain appropriate measurements of exposureand to document health outcomes andexposures; use adequate control orcomparison populations; obtain communitysupport to ensure an adequate participationrate; state clearly the study objectives andspecific hypothesis to be tested; havesufficient statistical power to detect predictedeffects, obtain data on important potentialconfounders, and evaluate a dose–responserelationship or gradients of exposure.

• ResourcesResources are critical to the support, conduct,and completion of any proposed health study.Issues for consideration include: theavailability of qualified personnel andtechnical support; an ability to obtainnecessary data and health information; andthe availability of appropriate projecttimelines and resources;

• Authority and supportIt is critically important that local, state, andfederal health agencies be involved early indiscussions about potential health studies.Issues for consideration include: the ability tosupport or provide technical assistancerequested by the local or state health agency;the ability of local and state health agenciesto address the community problem andhealth concerns; and the involvement ofappropriate agencies with legislative andregulatory backing.


69


5.10 Nature of the Health StudyWhen the decision to conduct a health study isbeing considered, several criteria are used todetermine the type of health study. These relate towhether the relevant research hypothesis requires:

• the characterisation of environmentalcontaminants by type, media, andconcentration levels;

• documented evidence of human exposure at alevel of concern;

• level of current knowledge about therelationship between exposure and specificadverse health outcomes; and/or

• documented excess of an adverse healthoutcome, when known.

The health studies can be grouped into Type 1and Type 2 studies.

5.10.1 Type-1 health studiesType-1 health studies explore or generatehypotheses about exposure-outcome associationsand address specific exposures, community healthconcerns, or specific information needs. Examplesof Type-1 health studies follow.

• Cross-sectional studies. These are surveys of asample of residents to obtain informationabout current and past health orenvironmental exposures, or both. Thesestudies can include comparison populationswith demographics similar to those of theexposed (target) population.

• Pilot investigations collect additionalinformation to assess the feasibility and valueof conducting a full-scale health study. Theinvestigation might include; assessments ofdata completeness and quality; the level ofdocumentation of exposures or healthoutcomes; methods to identify and trackindividuals, study size and statistical powerissues; and the availability of a controlpopulation or comparison.

• Cluster investigations evaluate the reportedoccurrence of a specific disease or condition isabove the expected number for a given

geographic location and time period. Theseinvestigations can be conducted to confirmcase reports, determine an unusual diseaseoccurrence, and explore potential risk factors.

• Comprehensive case reviews are medical orepidemiological evaluations of the medicalstatus of one or more individuals throughmedical record reviews, interviews orbiomedical testing to determine additionalinformation about their health status orpotential for exposure.

• Situation-specific surveillance is designed toassess the specific occurrence of one or moredefined health conditions among a specificpopulation potentially exposed to hazardousagents in the environment. Data collectionmight include using existing records of healthevents or records from relevant health careproviders.

• Health statistics reviews use available healthand demographic information to assess theoccurrence of specific health effects in definedgeographic areas and determine if the ratesare elevated compared to similar populationselsewhere. Available information mightinclude: death certificate, birth certificate, andcensus data; tumour or disease registry data;and health surveillance or disease notificationdata. A health statistics review may beperformed in response to a reported cluster ofspecific diseases or conditions.

• Exposure investigations use environmental orbiological testing, or both, for the hazardousagent(s) of interest. The biological test mightmeasure the level of the hazardous agent, ametabolite of a hazardous substance oranother marker of exposure in human bodyfluids or tissues. The purpose of thisinvestigation is to assess individual exposurelevels to a specific agent associated with thesituation. The levels identified should becompared with that of a relevant referencegroup or with a known standard referencelevel. Depending on the hazardous agent, theinvestigation can be used to explore forevidence of past or current exposure.

70

• Disease and symptom prevalence surveys areused to measure and compare the occurrenceof self-reported diseases, in some instancesusing medical records or physicalexaminations to validate adverse healthconditions. Addressing potential healthconcerns raised by the community, the surveycompares an exposed population (target area)with an unexposed population (control area)with similar demographic characteristics. Thepurpose is to determine the need for furtherhealth studies in the target area, providedthere are statistically significant excesses thatare clinically important. Depending on thecontaminants and circumstances, biologicaltesting of exposure or effect, or both, mightalso be collected as part of the survey.

When a Type-1 health study is consideredappropriate, there are several attributes that areconsidered necessary in order to ensure thequality of the study effort:

• a reasonable ability to document andcharacterise exposure in the target area;

• an adequate study size for the type of studyrecommended;

• an ability to identify and locate subjects andrecords;

• appropriate comparisons for rates ofoccurrence or levels of exposure; and

• an ability to control confounding factors andbiases (when possible).

5.10.2 Type-2 health studiesType-2 health studies are specifically designed totest scientific hypotheses about the associationsbetween adverse health outcomes and exposure tohazardous substances in the environment.Examples of Type-2 health studies follow:

• Case-control studies are designed to collectinformation and compare differences inexposures and other risk factors in two groupsof people: persons with specific illnesses orconditions (cases) and persons without theillnesses or conditions (controls). The controls

are selected to represent the population fromwhich the cases were identified. Usually thecases and controls are identified first, andthen information is collected about pastexposures and other risk factors.

• Cohort studies are designed to collectinformation from a group of people followedover a period of time, and information on theoccurrence of specific illnesses or conditionsis collected. Cohort studies can beprospective, meaning that individualsinvolved in the study are followed into thefuture, or cohorts can be retrospective,meaning that the cohort is reconstructedfrom historical records and then followedover a specified time period. They areexpensive, require long periods of time, andlarge numbers of people must be followed forrarer outcomes to provide enough cases foranalysis.

• Nested case-control studies are anotherapproach that uses both of the study designspreviously mentioned. The nested case-control study uses cohort individuals whohave developed a specific illness or condition(case) and persons sampled from the cohortwho have not developed the illness orcondition (control). The case-control methodis then used to collect additional informationand analyse the differences between these twogroups.

There are several attributes of Type-2 healthstudies that are considered necessary in order toensure valid scientific findings including:

• an ability to reasonably estimate or documentindividual exposures;

• an ability to document or validate humanhealth outcomes;

• an adequate study size and statistical power;

• an ability to identify and locate subjects andrecords;

• availability of an appropriate control orcomparison population;


71


• an ability to control confounding factors andminimise biases; and

• an ability to determine influence ofenvironmental, behavioural, or other factors.

5.11 Ensuring the Quality of aHealth Study

To ensure a useful and appropriate outcome thefollowing factors should be met:

• The group conducting the health study mustbe capable and fully responsible forconducting the health study;

• Personnel conducting the health study mustbe identified and have appropriate trainingand experience;

• The facilities and resources must beappropriate for the successful completion ofthe health study;

• Contractors for a health study must followwritten and approved work plans and theirwork must be carefully reviewed by thesponsoring group;

• For complex studies, a detailed study protocolshould be written and undergo scientific peerreview;

• Ethical issues relating to the protection ofhuman subjects, consent, and dataconfidentiality procedures must be addressed;

• Reports of complex health studies may needto undergo scientific peer review prior to anypublic release of information;

• Community involvement and knowledge ofthe health study are necessary: theinvolvement process will assist in ensuringthat the community understands and supportsthe study focus and design, and itslimitations.

• Depending on the community involvementapproach, public meetings might be held topresent and discuss the study methods andfindings. However, final study methods mustbe scientifically valid before proceeding;

• All study reports, data files and relateddocumentation should be kept in the officialrecords; and

• Any environmental sampling or biologicaltesting must follow existing standards forcollection, handling, chain of custody, storage,analysis, and reporting by an approvedlaboratory(ies): all standard quality controland quality assurance procedures must befollowed and documented.

5.12 Contents of a Health StudyProtocol

The following components should be consideredin drafting a report. Protocols for health studiesmight not need to contain all of the items withinthis outline. The listing is more comprehensive inorder to cover the wide variety of studyapproaches.

1. Title and identification page

2. Introduction and overview

3. Background

• Situation description

• Demographics

• Contaminants and pathways

• Community health concerns

• Literature review

4. Purpose

5. Study objectives

6. Methods

• Rationale for study design

• Study description

• Eligibility criteria

• Selection of target area and population

• Selection of comparison area andpopulation

• Sample size and statistical power estimates

72

• Participant selection and definitions

• Enrollment procedures

• Location(s) of data and specimencollection

• Informed consent procedure

• Questionnaire procedures

• Interviewer training and methods

• Methods for measurement of exposure

• Collection of biological specimens

• Additional data collection or sources

• Chain of custody and shipping

• Laboratory methods and quality control

• Privacy protection

• Findings of immediate significance

• Follow-up of abnormal lab results

• Data analysis

• Data entry, editing, and management

• Data transformation

7. Data analysis plan and methods

8. Study timelines

• Key activities or milestones

9. Community involvement and notification

10. Interpretation of results

11. Limitations of the study

12. References

13. Tables and figures

14. Attachments

• Data collection forms and questionnaire

• Study letters of notifications and consentform

15. Specimen collection and shipping protocols


73



Risk Management




Review and

reality check

Review and

reality check


74

Figure 4: Hypothetical curve for an animal carcinogenicity study

(adapted from Levy and Wegman, 1988)

Hazard Assessment—Part 3:Dose–Response Assessment

6

6.1 IntroductionThe following section uses material from theNHMRC’s Toxicity Assessment Guidelines forCarcinogenic Soil Contaminants (1999) andKlaassen (1996).

There are different ways of characterising doseresponse relationships including:

• effect levels (e.g. LD50, LC50, ED10) and noobserved adverse effect levels (NOAELs);

• margins of safety;

• therapeutic indices; and

• models to interpolate high dose experimentaldata to the low doses likely to be experiencedin the environment (Klaassen, 1996).

There are often limited human exposure data andanimal bio-assay data are most often used for doseresponse assessment. The use of these data requiresextrapolations from animals to humans andinterpolations from high doses to low doses (ibid).

The shape of the dose response curve below theexperimental range can have multiple shapesdepending on the model used (Figure 4). Thechoice of the model should, where possible, bebased on mechanistic information.


75

0

Range of exposures in the general environment

?

?

?

Range of animal experimental doses

Increasing dose or exposure

% o

f pop

ulat

ion

unde

r st

udy

resp

ondi

ng

0.01

Axes not to scale.

0.1

10

15

20

25

95

100

0.001x

0.01x

0.1x

x 10x

100x


The dose response curves for different effects willhave different shapes and will occur at differentdoses (See Figure 5). The shape of the doseresponse curve will be different again whendealing with, for example, an essential traceelement such as copper where there will be at lowdoses a dose response curve for the effects ofdeficiency and at higher doses another doseresponse curve describing the effects of excess.

6.2 MethodologiesAll methodologies make the distinction betweenneoplastic and non-neoplastic end-points in riskassessment. The impetus for this distinction wasthe concept of a lack of threshold in thedose–response for carcinogens based on the initialpremise that all carcinogens are mutagens (Ameset al, 1973). One mutation or one DNA damageevent was considered sufficient to initiate theprocess that leads to the development of cancer.

In contrast, the dose–response was assumed tohave a threshold for non-cancer effects,the assumption being that, for non-cancer effects,there is a dose below which the risk of adverseeffects will be nil; in effect, that there is a‘safe’ dose.

In recent times, as it has become evident that notall carcinogens have genotoxicity as their primemode of action (Ashby and Tennant, 1991), thedose–response curve has been assumed to be non-threshold for genotoxic carcinogens and thresholdfor non-genotoxic carcinogens (Vermeire and van der Heijden, 1990; Health Council of The Netherlands, 1994; Moolenaar, 1994a).Accordingly, non-threshold and threshold modelshave been applied to genotoxic and non-genotoxiccarcinogens, respectively, in some countries such asCanada and some European countries (Whysnerand Williams, 1992; Health Council of TheNetherlands, 1994; Moolenaar, 1994a, 1994b,

76

Figure 5: Different dose–response curves for different effects from a hypothetical substance

0

Enzyme (1) level

Enzyme (2) level

Liver pathology

Death

Increasing dose or exposure

50

100

Perc

enta

ge o

f pop

ulat

ion

unde

r st

udy

resp

ondi

ng

New Zealand Ministry of Health, 1995) and inthe WHO Drinking Water Guidelines (WHO,1993). More recently, however, the WHOconsidered that these approaches are not suitableto the development of generic guidance values inEnvironmental Health Criteria documentsbecause they ‘…require socio-political judgementsof acceptable health risks.’ (WHO, 1994).

In these examples, the distinction between agenotoxic carcinogen and a non-genotoxiccarcinogen is a science policy decision forregulatory purposes and does not necessarilyreflect the mechanism of carcinogenesis. It doesnot mean that a non-genotoxic carcinogen doesnot affect the genetic material of the cell undersome circumstances, nor that a genotoxic effect isthe only event required for the development ofcancer by a genotoxic carcinogen.

With advances in biological knowledge,mechanistic data, pharmacokinetic data and otherrelevant data are increasingly being taken intoaccount in classifying and assessing the risks ofcarcinogens. The US EPA (1996) is in the processof revising its guidelines for cancer riskassessment and, whilst relying almost exclusivelyon the non-threshold, low dose extrapolation forcancer risk assessment as in the past, also seemsto be accepting an approach which considersmode of action and multiple dose–responserelationships.

6.3 Threshold ApproachesA threshold is considered to occur because ofbiological mechanisms such as the ability tometabolise or excrete a toxin or to repair damageup to a certain dose.

The approach with these models is to deriveexposure limits such as an ADI, a ProvisionalTolerable Weekly Intake (PTWI), Tolerable DailyIntake (TDI) or RfD (Barnes and Dourson, 1988;WHO, 1993; WHO, 1994; Dourson et al, 1996).This approach makes no attempt to calculate alevel of risk at low exposures. Rather, it derives adose which is apparently without effect in ahuman population or suitable animal model, andthen applies a factor to derive an exposure which

has a high likelihood that no effect will occur inthe general human population.

These exposure limits are derived by firstdetermining the No Observed Adverse EffectLevel (NOAEL) or, if the NOAEL cannot bedetermined, the Lowest Observed Adverse EffectLevel (LOAEL) and dividing the value by factorsto account for:

• interspecies differences (extrapolating fromanimals to humans);

• intraspecies differences (differing sensitivitiesbetween individuals);

• the severity of the adverse effect; and

• the quantity and quality of the scientific data.

Traditionally, safety factors for intraspecies andinterspecies differences have each been assignedvalues of ten, and the other two have beenassigned values between 1 and 10. An additionalfactor of ten is sometimes used if the NOAELwas not established in the study. The individualfactors are then multiplied to determine an overallsafety factor by which the NOAEL is divided togive the ADI, PTWI, TDI or RfD.

Historically, the most common overall factor usedby a number of regulatory bodies is 100, if a largetoxicological database has been assessed, althoughthe overall factor can range from 10 to 10 000.From the data available on humans andexperimental animals, it appears that interspeciesand intraspecies differences are in general lessthan 10, hence the often-used overall safety factorof 100 for these two factors is conservative andadequately protective of public health ( Johannsen,1990; Renwick and Walker, 1993).

The decision on the magnitude of factors to use ispredominantly based on expert or informedjudgement. Whilst this approach to selecting thenumber and magnitude of the safety factorsappears to be arbitrary, current knowledge of thebiological processes which cause inter- andintraspecies variation (e.g. metabolic and otherpharmacokinetic rate differences) support thechoice of safety factors.


77


6.4 Non-Threshold ApproachesThese approaches do not recognise the possibilityof a threshold effect and are appropriate forradiation and for some genotoxic carcinogens. Itis, as a science policy decision, applied to allcarcinogens by the US EPA.

Non-threshold models assume linearity betweenthe lowest experimentally derived dose and thezero dose (the origin). This implies that there is acalculable probability of an adverse effect (risk) nomatter how small the dose. This does not meanthat there is no dose that could be considered safeunless safety must be equated with zero risk(Hrudey and Krewski, 1995).

Numerical estimates of risk probabilities aregenerated by fitting one or more mathematicalmodels to the data in the experimental dose rangeand extrapolating to the low environmentalexposure doses. For example, low-doseextrapolation using a linear model is a defaultapproach for cancer risk assessment in the USA(US EPA, 1986; 1996) and is one approach whichhas been used by the WHO for genotoxiccarcinogens in deriving drinking water guidelines(WHO, 1993).

The outcomes are estimates of either:

i) the dose at a predetermined acceptable risklevel (note that this requires some judgementon what constitutes an acceptable level ofrisk); or

ii) the estimated risk level at any particular dose.

6.5 Threshold vs Non-Threshold Approaches

This area of scientific debate largely centres onthe management of carcinogens.

The important conceptual distinction betweennon-threshold methods and those which derivean acceptable exposure from the NOAEL using asafety factor is that the safety factor approachmakes no attempt to determine a finite level ofrisk at low exposures whereas the linear methodsmake an estimate of the risk at low exposures.

The NOAEL is assumed to be the threshold dosefor the effect. Both approaches have advantagesand disadvantages.

The advantages of the threshold approach are thatthe NOAEL is relatively easy to determine, andthe process is simple to use, easy to understandand allows the use of expert judgement. In thefew cases where epidemiological data havebecome available, the ADIs derived by thismethod have been validated (Lu and Sielken Jr,1991). Additionally, the approach has beenapplied seemingly in a consistent fashion by theWHO in the last three decades in deriving ADIsfor pesticides (Lu, 1995). The safety factorapproach remained essentially unchanged until1994 (WHO, 1994), although a number ofarticles were published suggesting modificationsor improvements (e.g. Zbinden, 1979; Crump,1984; Johannsen, 1990; Lewis et al, 1990; Lu andSielken Jr, 1991; Calabrese and Gilbert, 1993;Calabrese and Baldwin, 1994).

Because it provides numerical estimates of risk atall doses, the non-threshold approach, inprinciple, has the potential advantages (if theestimates are correct) of allowing: computation ofcomparative risks in the sub-experimental range,which may be a useful tool in risk managementand communication; potency comparisonsbetween chemical agents at a particular risk level;and estimates of the increased risks if a particulardose is exceeded. It has been argued (McMichael,1991) that risk estimates by this approachapproximate those seen in humans in some casesand where there are disparities they areoverestimates of the risks.

Both the threshold and non-threshold methods,however, are likely to be unduly influenced by theselection of doses. The choice of the NOAEL islimited to one of the doses included in theexperimental design. The biological no effect dosemay occur at this dose or at a dose not includedin the study. The closeness with which theselected NOAEL truly reflects the actual no effectdose has an obvious impact on the degree ofprotectiveness in the derived ADI, PTWI orRfD. Furthermore, the NOAEL is influenced bythe biological effects monitored, the number of

78

animals in the test groups, the spontaneousincidence of the adverse effect, and the criteriaused to determine when the incidence in a testgroup exceeds that in the controls (Renwick andWalker, 1993).

Additional limitations of the threshold approachinclude: the NOAEL is often perceived as abiological threshold, whereas it is a thresholdlimited by the experimental protocol; risk isexpressed as a fraction of the guidance dose (e.g. ADI); it makes limited use of thedose–response slope; the choice of safety factorshas been arbitrary to some extent and the processdoes not generate a range of estimates of risk, butrather a single estimate of a dose below which noadverse effects are likely to be produced.

Dose selection in non-threshold models has beendiscussed by Lovell and Thomas (1996) whosuggest that the estimate of q1* (the 95 per centupper confidence limit of the slope estimate usedfor the linear multi-stage model) is so dependenton the doses selected that it is almostindependent of, or at least insensitive to, theactual tumour incidences in the dose groups.Specifically, the highest dose in an animalbioassay has overwhelming influence on theestimate of q1*, thus leading to the overestimationof risk at very low doses, with the extent ofoverestimation increasing as the environmentalexposure becomes lower. Typically, the highestdose in a carcinogenicity bioassay is the maximumtolerated dose (MTD), a dose that causes nomore than a ten percent decrease in body weightand no other overt toxicity. The MTD is verymuch greater than doses expected from non-occupational environmental exposures. Therefore,the dose which is the least relevant toenvironmental risk assessment has the greatestinfluence on low dose risk estimates.

Non-threshold models currently in use areinflexible and generally do not take account of thecomplexities of the events between exposure to anagent and the induction of a neoplasm. Risksestimated at doses below the range ofexperimental data can vary considerablydepending on the model used, even though thevarious mathematical models used generally fit

the experimental data equally well (Crump, 1985;Paustenbach, 1995). The numerical expression ofthe estimated level of risk falsely gives theimpression that it represents an exact measure ofactual risk. This numerical expression provideslittle or no information on the uncertaintiesrelated to the estimated level of risk, nor does itallow comparison with values for non-cancerhealth effects.

Low-dose linearity assumes a positive slope of thedose–response curve at zero dose and implies thata single, irreversible genetic event at the initiationstage of carcinogenesis leading to transformationof a cell, is sufficient by itself to lead to thedevelopment of cancer. The major difficulty inthis debate is the impossibility of testingexperimentally the shape of the dose–responsecurve at extremely low doses (Purchase andAuton, 1995).

A transformed cell which has acquired thepotential to develop into a tumour, will probablyrealise that potential only rarely (US EPA, 1996),most likely because of the natural large scalerepair of DNA damage and other defencemechanisms of the body (DOH, 1991; Abelson,1994). Furthermore, whilst it is generally acceptedthat mutagens and mutations play a role in thedevelopment of cancer, carcinogenesis is morethan mutagenesis, with a number of non-mutagenic as well as mutagenic events takingplace during the process (Bishop, 1991). Theshape of the dose–response curve at any one ofthese steps, not just the mutagenic events, caninfluence the shape of the dose–response curve forthe carcinogenic response. Factors, such as geneticmake-up, lifestyle and other environmentalfactors, may also have a modifying influence onthe processes of carcinogenesis.

6.6 Mechanistically DerivedModels

These use models, which describe biologicalmechanisms by mathematical equations. Theyassume that the toxic effect results from therandom occurrence of one or more biologicalevents. These are known as stochastic events(Klaassen, 1996).


79


Mechanistically-derived models have beenparticularly used for cancer modelling andespecially based on radiation exposures. Thesimplest form is a ‘one hit’ linear model in whichonly one ‘hit’ or critical cellular interaction resultsin the alteration of a cell. This model wouldpropose that a single molecule of a genotoxiccarcinogen would have a ‘minute but finite chanceof causing a mutational event’ (Klaassen, 1996).From these models more complex models basedon multihits or multistage events have beenderived. Although conceptually based onbiological mechanisms, most of these models donot rely on independently validated parametersdescribing the mechanisms but rely on fittingcurves to empirically observed data.

More recently these models have been adapted totake into account information based onknowledge of the relevant physiology andtoxicokinetics (Physiologically-basedtoxicokinetics (or pharmacokinetics) modelling—PBTK ). These models take into account the

effective dose at the target organ. A furtherdevelopment has been to make generalisedmechanistic models take into account specificbiological processes. An example of thesebiologically based dose response models is theMoolgavkar-Venson-Knudson model that uses a two-stage model for carcinogenesis(Klaassen, 1996).

6.7 Benchmark Dose ApproachThe benchmark dose (BMD) approach has beenused in dealing with both cancer and non-cancerend points. It is described in EHC170 and amodified version for use with carcinogenic soilcontaminants is described in NHMRC (1999). Thebenchmark dose corresponds to a predeterminedincrease (usually 5 per cent) of a defined effect in atest population. Mathematically it is the statisticallower confidence limit on the dose that correspondsto that predetermined increase although someagencies are using a best estimate rather than alower confidence limit (IEH, 1999b).

80

Table 11: Models used in risk extrapolation

Statistical or distribution models

• Log-probit

• Logit

• Weibull

Mechanistic models

• One-hit

• Multihit

• Multistage

• Linearised multistage

• Stochastic two-stage model (Moolgavkar-Venson-Knudson)

Model enhancement

• Time to tumour response

• Physiologically based toxicokinetic models

• Biologically based dose–response models

(from Klaasen, 1996, p. 82)

Figure 6: Graphical illustration of the benchmark dose

(adapted from WHO, 1994)


81

In this example, LED5 = BD, and LED5 is lowerconfidence limit of the effective dose causing a 5 per cent increase in a defined effect.

For developmental toxicity the BMD5 values havebeen similar to statistically-derived NOAELs fora wide variety of developmental toxicity endpoints (Klaassen, 1996). BMD approaches arealso being developed and tested in regard to acuteinhalation toxicity (Fowle et al, 1999), to therelationship between the BMD and the MTD(Gaylor and Gold, 1998), and to addressingstatistical procedures available for calculatingBMDs and their confidence limits for non-cancerendpoints (Gaylor et al, 1998).

Particular advantages of the BMD approachinclude:

• taking into account information from theentire dose response curve rather thanfocussing on a single test dose such as is donewith the NOAEL approach;

• the use of responses within or near theexperimental range versus relying onextrapolations to doses considerably belowthe experimental range;

• the use of a consistent benchmark responselevel that cross a range of studies andendpoints;

• it is less influenced than NOAEL approachesby the arbitrary selection of doses (Crump, 1984);

• it is able to be rigorously described; and

• it uses all available relevant information.

Its disadvantages are that it may not be possibleto define the shape of the dose response curvebecause of limited dose groups or the number ofanimals per group and it also requires greaterstatistical expertise than the NOAEL typeapproach (IEH, 1999b)

Use of a benchmark dose with 5 per cent extrarisk provides a more data sensitive and less modelsensitive endpoint than using 1 per cent extrarisk. (Klaassen, 1996; NHMRC, 1999)

When the benchmark response is within or nearthe experimental range of data the correspondingvalues of the benchmark doses are not greatlysensitive to the choice of the model used but thebest scientific choice of a model would be abiologically based mechanistic model.

Exce

ss p

ropo

rtio

n of

ab

norm

al r

espo

nses

Dose

0.05

0LED5 ED5

Upper confidence limit on estimated risk

Does-response fitted to experimental data


6.8 Inter- and Intra-SpeciesConsiderations

The following material is from WHO (1999b, p. 20) with slight adaptation.

The strains and species of laboratory animalsexposed in toxicity studies have been selected toshow minimum inter-individual variability. Incontrast to laboratory animals, humans representa very heterogeneous population with bothgenetic and acquired diversity.

Therefore, two principal areas are consideredwhen interpreting dose–response data acquired inanimal species in relation to human risk:

a) Inter-species consideration: comparison ofthe data for animals with a representativehealthy human. Species differences resultfrom metabolic, functional and structuralvariations; and

b) Intra-species or inter individualconsideration: comparison of therepresentative healthy human with the rangeof variability present within the humanpopulation in relation to the relevantparameter(s).

For each of these areas, there are two aspects tobe considered in assessing risk, i.e. toxicokinetics(the delivery of the compound to the site ofaction) and toxicodynamics (the inherentsensitivity of the site of action to the chemical).Any approach that allows for the incorporation ofadequate data on toxicokinetic or toxicodynamicdifferences between test animal and humans, orbetween different humans, will increase thescientific validity of risk assessment.

Sources of inter-species and inter-individualvariations in toxicokinetics include: differences inanatomy (e.g. gastrointestinal structure andfunction); physiological function (e.g. cardiacoutput, renal and hepatic blood, glomerularfiltration rate and gastric pH); and biochemicaldifferences in, for example, enzymes involved inxenobiotic metabolism. Sources of inter-speciesand inter-individual differences in toxicodynamics(or inherent sensitivity) also include anatomicalconsiderations. For example, the effect may occur

in an organ of questionable relevance to humans,such as the rodent forestomach. Physiologicaldifferences, such as the hormonal control of thetarget organ, and biochemical differences,e.g. species differences in key biochemicalcomponents such as α2µ-microglobulin and itsrole in rat kidney cancer, may also be relevant(Flamm and Lehman McKeeman, 1991).

In some cases, it may be possible to conclude thateffects detected in animals are unlikely to berelevant to humans. In other cases, there may bedata to indicate that humans are likely to be moreor less sensitive than animal species; thisinformation is important for consideration in theselection of critical effects.

If compound-specific toxicokinetic data areintroduced into risk assessment, then it isessential that these are related to the species,protocol and active chemical entity (e.g. parentcompound or metabolite) involved in the toxicitythat is the basis for the hazard identification(Monro, 1990, 1993; Renwick, 1993).

6.8.1 Species differencesMetabolism and structural/functional variationsare both important determinants of speciesdifferences. Common areas of metabolic variationbetween species are digestive tract enzymes, levelsof circulating enzymes, liver enzymes anddetoxification processes.

In extrapolating between species, three aspectsneed to be considered:

1. differences in body size, which requires dosenormalisation or scaling (often done byexpressing the dose per kg body weight);

2. differences in toxicokinetics, particularlybioactivation and/or detoxification processes;and

3. the nature and severity of the target for toxicity.

Inter-species normalisation (or scaling) isgenerally based on physical characteristics (e.g. body weight, body surface area), althoughoccasionally it is based on caloric demand or,where there are data in four species, multiplespecies regression.

82

When clearance of the parent substance is limitedby enzyme activity rather than blood flow orwhen metabolites are the toxic agents, moresophisticated physiologically-basedpharmacokinetic models are more appropriate,provided that adequate data are available.Currently, such data are available for only a smallnumber of substances.

6.8.2 Human variabilityAlthough data from animal studies using fairlyhomogeneous populations may provide limitedinformation on inter-individual variability withinthe test species, the greater potential variability inheterogeneous human populations must beaddressed in risk assessment. Sources of inter-individual variability in human populationsinclude, for example, variations in geneticcomposition, nutrition, disease state and lifestyle.

6.9 MixturesCurrently there is no agreed Australian approachto assessing mixtures of agents. Where data(including mechanistic data) are available on theinteraction of agents this should be taken intoaccount in the risk assessment.

Environmental exposures can involve more thanone type of agent and may require a differentmode of assessment than for single agentexposures (See Section 3.5). For such complexexposure scenarios, risk assessment considerationsare most advanced for chemical mixtures.

Humans continue to be consistently exposed to acomplex, ever-changing mixture of environmentalchemicals in the air they breathe, the water theydrink, the foods and beverages they consume, thesurfaces they touch and the products they use(Sexton et al, 1995). Moreover, most currentestablished exposure standards are for singlecompounds only (Lang, 1995). An increasedunderstanding of chemical mixtures is importantespecially as exposure from multiple sourcesincreases.

The Agency of Toxic Substances and DiseaseRegistry (ATSDR) in the US uses one approachthat includes performing a critical synthesis ofrelevant data and then identifying generalisable

rules that can be used in site-specific assessmentsof health risk following exposure to mixtures.This approach allows research to: identify whatchemical mixtures may affect public health;evaluate the potential for exposure of humanpopulations to chemical mixtures; study thepharmacokinetic behaviour of chemical mixtures;identify various end points that would be effected;study the mechanism of action, progression andrepair; and identify (both generic and specific)that would allow the determination of the healthof the organism and develop qualitative andquantitative health assessment methods so as toassess multiple health effects (Hansen et al, 1998).

Where chemicals share structural similarities suchas Dioxins, PCBs and Polycyclic AromaticHydrocarbons the use of Toxic Equivalency Factorshas been proposed. Different substances are giventoxicity ‘scores’ that are fractions of the toxicity ofanother in the chemical group for which there isadequate toxicity data. Given a mixture of thesubstances, a cumulative toxicity score can bedetermined. The IPCS has published TEFs forseveral dioxins and these are shown in Table 12.

World Health Organization TEFs derived fromexpert meeting held in Stockholm on 15–18 June1997. Cited in Van den Berg, et al (1998).


83

Table 12: Toxic equivalencyfactors (TEFs) for human andmammalian risk assessment

Congener (Dioxins) WHO TEF

2,3,7,8 – TetraCDD 1

1,2,3,7,8 – PentaCDD 1

1,2,3,4,7,8 – HexaCDD 0.1

1,2,3,6,7,8 – HexaCDD 0.1

1,2,3,7,8,9 – HexaCDD 0.1

1,2,3,4,6,7,8 HeptaCDD 0.01

OctaCDD 0.0001

CDD: Chlorinated dibenzodioxin


Biological methods such as bioassays are beingappraised for their application to the assessmentof the toxicity of mixtures. Rodents or othermammals may be administered extracts as toxicityvalues such as LD50s can be determined. Thesemethods are expensive and time consuming whichusually precludes their use in risk assessment.Similar techniques using aquatic species such asDaphnia are less expensive and time consumingbut are disadvantaged by the greater toxicokineticand toxicodynamic differences between thespecies used and humans (Pollak, 1996). There arein vitro tests such as the Microtox test and theSubmitochondrial Particle Test but these requirevalidation for use in risk assessment.

Useful information for exposures to mixtures ofhazards may be available from epidemiologicalstudies of closely similar mixtures.

Sexton and colleagues (1995) propose nine ‘needs’crucial to mixture-related research and which aresummarised as follows:

1. A need for a balanced research approachbetween whole mixture analysis (‘top-down’approach) in which research begins with theentire mixture and attempts to characterisethe mixture’s toxicity by studying it infractions according to toxicity or by itsentirety and component analysis or (‘bottom-up’ approach) in which research begins withindividual components and attempts topredict the toxicity of the whole mixture bydeveloping an understanding of theinteractions among the individualcomponents;

2. A need for mechanistically- andphysiologically-based predictive models toimprove the ability to extrapolate realistically(e.g. from animals to humans) and ensurethat studies simulate actual conditions;

3. A need for well-designed toxicological studiesthat help to explain, for example, how oftenchemical interactions are dependent onexposure conditions;

4. A need to determine boundaries for mixture-related outcomes such as quantifying theresults of mixture-related research todetermine when, why and how often mixturesdeviate from additivity;

5. A need to identify and focus on high prioritymixtures so as to protect public healththrough criteria such as expected exposureconditions, the number and types of peoplelikely to be affected or seriousness of healthoutcomes;

6. A need to develop better understanding ofexposure-related variables (e.g. routes,duration, frequency). The more closelytoxicological studies mimic real-life, the morerelevant the results would be for screening,protecting and predicting due to thecomplicated nature of human exposures;

7. A need for new and better methods that takeadvantage of advances in informationtechnology to ensure greater precision andcertainty;

8. A need to identify, quantify and expressuncertainty in mixture-related risk assessmentso as to develop approaches and methods thatwill allow for quantitative estimation ofscientific uncertainty associated with scientificpolicy decisions related to specific mixtures;and

9. A need for collective and comprehensiveresearch strategies by researchers andregulators.

Potential health effects caused by concurrentexposures to multiple chemicals requireconsideration in a risk assessment. However thisprocess is not straightforward and further work isbeing conducted to address the issues involved.

6.10 HormesisIt is important to be mindful of the phenomenonof hormesis, i.e. demonstrated beneficial effects ofan agent at low (but not homeopathic) doses butwith toxicity occurring at higher doses. Hormesis

84

is known for several chemicals and generallyrequires very detailed dose–responsecharacterisation for its detection. Presently thereis no clear guidance on how to incorporate suchinformation into health risk assessment, butdealing with exposures to an agent elicitingpossible health benefits and health risks willrequire especial cooperation between riskmanagers and assessors (see Calabrese, 1996;Calabrese and Baldwin, 1998; Belle, 1998).

A separate but related issue is the concept of theAcceptable Ranges of Oral Intakes for EssentialTrace Elements where there is a need to ensurethat TIs do not fall below the minimumrequirements. (WHO (1996) regards iron, zinc,copper, chromium, iodine, cobalt, manganese,molybdenum and selenium as unequivocallyessential for human health.)


85



Risk Management




Review and

reality check

Review and

reality check


86

Hazard Assessment—Part 4:Hazard Assessment Reports

7

7.1 IntroductionThe overall toxicity assessment report shouldconsider:

• the nature of adverse effects related to theexposure;

• the dose–response relationship for variouseffects;

• the weight of evidence for effects such ascarcinogenicity.

This Section covers:

• A checklist for the hazard assessment report(Section 7.1)

• Sources of toxicological data and their likelyacceptability ranking (Section 7.2)

7.1.1 Checklist for toxicologicalappraisals

The following checklist is adapted with slightmodification from US EPA (1995) and should bethe basis of toxicological appraisals. Asummarised version can be used if TolerableIntake data from WHO or NHMRC are used.

7.1.2 Hazard identification1. What is the key toxicological study (or

studies) that provides the basis for healthconcerns?

• How good is the key study?

• Are the data from laboratory or fieldstudies? Are the data for single species ormultiple species?

• If the hazard is carcinogenic, comment onissues such as: observation of single ormultiple tumour sites; occurrence ofbenign or malignant tumours; certaintumour types not linked to carcinogenicity;use of the maximum tolerated dose.

• If the hazard is other than carcinogenic,what endpoints were observed, and what isthe basis for the critical effect?

• Describe other studies that support this finding.

• Discuss any valid studies which conflictwith this finding. See also Section 4.4 forfurther information on assessing data.

• As many relevant studies as possibleshould be collated and rigorously assessedas to their strengths and weaknesses todetermine the key studies. This isparticularly important where quantitativerisk estimates will be undertaken or wherethere are apparently contradictory studies;in the latter case, the studies that areconsidered to be adequate in their designand interpretation will need to beappraised to determine the overall weight-of-evidence. See Section 4.7 for furtherinformation on weight of evidence.

2. Besides the health effect observed in the keystudy, are there other health endpoints ofconcern?

• What are the significant data gaps?

3. Discuss available epidemiological or clinicaldata. For epidemiological studies:

• What types of studies were used,i.e. ecologic, case-control, cohort?

• Describe the degree to which exposureswere adequately described.

• Describe the degree to which confoundingfactors were adequately accounted for.

• Describe the degree to which other causalfactors were excluded.’

For further information refer to Section 5.4‘Assessing the relationship between a possiblecause and an outcome’.

4. How much is known about the biologicalmechanism by which the agent producesadverse effects?

• Discuss relevant studies on mechanisms of action which may include metabolismstudies.


87


• Does this information aid in theinterpretation of the toxicity data?

• What are the implications for potentialhealth effects?

5. Comment on any negative or equivocalfindings in animals or humans, and whetherthese data were considered in the hazardidentification.

6. Summarise the hazard identification anddiscuss the significance of each of thefollowing.

• confidence in conclusions;

• alternative conclusions that are alsosupported by the data;

• significant data gaps; and

• major assumptions.

7.1.3 Characterisation of dose–response

1. What data were used to develop thedose–response curve? Would the result havebeen significantly different if based on adifferent data set?

If animal data were used:

• What species were used? The mostsensitive, average of all species, or other?

• Were any studies excluded? Why?

If epidemiological data were used:

• Which studies were used? Only positivestudies, all-studies, or some othercombination?

• Were any studies excluded? Why?

• Was a meta-analysis performed tocombine the epidemiological, studies?What approach was used? Were studiesexcluded? Why?

2. What model was used to develop thedose–response curve? What rationalesupports this choice? Is chemical-specificinformation available to support thisapproach?

For non-carcinogenic hazards:

• How was the Tolerable Intake (or theacceptable range) estimated?

• What assumptions or uncertainty factorswere used?

• What is the confidence in the estimates?

For carcinogenic hazards:

• What dose–response model was used?What is the basis for the selection of theparticular dose–response model used? Arethere other models that could have beenused with equal plausibility and scientificvalidity?

• What is the basis for selection of themodel used in this instance?

3. Discuss the route and level of exposureobserved in the toxicology or epidemiologystudies, as compared to the expected humanexposures in the situation under appraisal.

• Are the available data from the same route ofexposure as the expected human exposures?If not, are pharmacokinetic data available toextrapolate across route of exposure?

• What is the degree of extrapolation fromthe observed data in the toxicological orepidemiological studies to the expectedhuman exposures in the situation underappraisal. (one to two orders of magnitude?multiple orders of magnitude)? What isthe impact of such an extrapolation?

7.2 Sources of Toxicologicaland Tolerable Intake Data

The following sources are grouped into ‘levels’which are given in order of preference. In general,published Australian ADIs should be used butother data may be used with appropriatejustification. Different agencies are likely to haveused differing risk assessment and standards-setting methodologies and these differencesshould be appraised by the risk assessor. Alldocuments, particularly those in the second andthird categories require rigorous appraisal forrelevance, validity and accuracy.

88

7.2.1 Level 1 sources1. National Health and Medical Research

Council documents and documents fromother joint Commonwealth, State andTerritory health organisations. These may bea source of Australian guidance values.

2. ADI List from the Therapeutic GoodsAdministration (regularly updated atwww.health.gov.au/tga/docs/html/adi.htm).

3. World Health Organization (WHO)documents. Australia is a party to the WHOprocess and has incorporated their material ina variety of environmental health criteria. Arange of documents include those from theWHO/ILO/UNEP International Programmeon Chemical Safety (IPCS) which producesEnvironmental Health Criteria monographs,and Concise International ChemicalAssessment documents (CICADs).Documents detailing international AcceptableDaily Intakes (ADIs), Tolerable Daily Intakes(TDI) or Tolerable Weekly Intakes (TWI)may be found in evaluations by theWHO/FAO Joint Meeting on PesticideResidues ( JMPR) and by the JointFAO/WHO Expert Committee on FoodAdditives ( JECFA).

4. enHealth Council documents.

5. National Environmental Health Forumdocuments distributed by the CommonwealthDepartment of Health and Ageing.

6. International Agency for Research on Cancer(IARC) monographs.

7. WHO/FAO Joint Meeting on PesticideResidues ( JMPR) Monographs.

8. NICNAS Priority Existing Chemical (PEC)reports.

9. US Agency for Toxic Substances and DiseaseRegistry (ATSDR) documents for generaltoxicological reviews and Reference Doses.

10. National Toxicology Program (NTP)carcinogenicity appraisals which report indetail the results of carcinogenicity tests on awide range of chemicals.

11. OECD Standard Information Data Sets (SIDS)and SIDS Initial Assessment Reports (SIAR).

12. EPA Reference Doses.

7.2.2 Level 2 sources

Peer-reviewed journals

These may provide opinions that do not meetgeneral scientific agreement. With justification,and acceptance by the local jurisdiction, they maybe suitable for use if no Guidance Values areavailable.

Industry publications

With justification, and acceptance by the localjurisdiction, they may be suitable for use:

1. European Centre for Ecotoxicology andToxicology of Chemicals (ECETOC):Monographs, JACC Reports and TechnicalReports

2. Chemical Industry Institute of Toxicology(CIIT) reports

3. Unpublished industry reports submitted forregulatory purposes. These may have restrictedavailability but information may be availablein evaluation reports from regulatory agenciesthat have reviewed individual reports.

Occupational health and safety sources

These may be a useful source for toxicologicaldata and reviews but occupational exposurecriteria must not be used in a general publichealth context without appropriate adjustment forthe different durations of exposure, the inclusionof susceptible sub-population in the generalcommunity (e.g. children) and the methodologicaldifferences in the setting of criteria.

7.2.3 Level 3 sourcesThese are sources not covered in Levels 1 and 2.The use of this information requires justificationthat no other sources are available and an appraisalof the methodology detailing the level ofconservatism and range of uncertainties inherentin the approach. With justification, and acceptanceby the local jurisdiction, they may be suitable foruse if no Guidance Values are available.


89



Risk Management




Review and

reality check

Review and

reality check


90

Exposure Assessment

8

8.1 IntroductionExposure assessment involves the determinationof the magnitude, frequency, extent, character andduration of exposures in the past, currently, and inthe future. There is also the identification ofexposed populations and potential exposurepathways. Environmental monitoring andpredictive models can be used to determine thelevels of exposure at particular points on theexposure pathways. The contaminant intakes fromthe various pathways under a range of scenarios,including worst case situations, can then beestimated (US EPA, 1989).

An initial requirement for exposure assessment isan understanding of the presence (or absence) ofan agent and its concentrations and distribution.

Accurate and useful exposure assessment requiresa detailed understanding both of the strengthsand weaknesses of the exposure assessmenttechniques, and the specific exposure factors usedin the assessment. Considerable effort needs to bemade to accurately characterise the population towhich the exposure assessment is relevant.

‘Direct measurement of the exposures of the(potentially) affected population provides the bestexposure data but this is not always available orpracticable and default exposure factor data areoften required.’ (Langley, 1993, p. 90)

An understanding of transport and fate modelsfor the agent is also important. Transport and fatewill be affected by:

• environmental medium (e.g. air, surface,water, soil, ground water or biota);

• geographic scale (e.g. global, national,regional or local);

• pollutant source characteristics (e.g.continuous or instantaneous releases fromindustrial, residential and commercial pointor area sources);

• the nature of the risk agent (e.g. whether it isa specific agent or group of agents);

• the receptor population (e.g. humans,animals, plants, microorganisms, and habitats,as well as specific sub-populations exposed to

high levels of the agent or who areparticularly sensitive to exposure);

• exposure routes (e.g. ingestion, dermalcontact or inhalation);

• environmental conditions (e.g. pH, presence oforganic matter, clay content, temperature); and

• the timeframe (e.g. retrospective, current orprospective).

(Fiksel and Scow, 1983, cited in Covello andMerkhofer, 1993)

There are various components to estimatingexposure and these are shown in Figure 7.

8.2 Issues in ExposureAssessments

1. Children usually receive a higher exposure toenvironmental agents per unit body weightthan adults because of behavioural andphysiological factors (e.g. hand-to mouthactivities for soils, higher respiration rates perunit body weight, increased gastrointestinalabsorption of some substances).

2. For soil, ingestion is usually by far the mostimportant exposure route for small children(ibid).

3. One exposure route will normallypredominate (ibid).

4. In large-scale contamination (i.e. regional)more exposure pathways will be involved thanin small-scale (very localised) contamination.

5. All exposure pathways must be considered forhealth risk assessment. As the total amount ofa chemical absorbed by a persons bodyinfluences the risk to health exposureassessment must take into account all sourcesof exposure irrespective of whether these arefrom food, water, the work place, outdoor airor a combination of these and other sources(Langley, 1991; IEH, 1999a).

6. Bioaccumulation may be a significant concernfor some substances with long biological half-lives e.g. cadmium, organochlorine pesticidesand this factor should be considered.


91


Figure 7: Components of exposure assessment

(adapted from National Academy of Sciences (NAS), 1991)

92

Exposure Assessment Approaches

Direct Methods

Personal Monitoring

Biological Markers

Migration Measures

Environmental Monitoring

Models Questionnaires Diaries

Pharmacokinetic and

Pharmacodynamic Models

Indirect Methods

Exposure Models

Exposure Estimate

8.3 Environmental DistributionFor the development of sampling plans forchemical agents and the process of exposureassessment an understanding and the movementof chemical agents between environmentalcompartments and the effects of environmentalpartitioning will be necessary.

Partitioning will reflect the fact that substanceswill move to the environmental compartment forwhich they have the most affinity (Calamari,1993; Calamari, 1999). Transformation may occurin any environmental compartment.

Fugacity modelling (Mackay, 1991) enables anestimation of which compartment will containmost of the agent and where the highestconcentrations in the ‘unit of world’ are. Mackay’s

‘unit of world’ is a box 1km square and 6km deepthat includes air, terrestrial and aquatic biomass,soil, water and sediment.

Especially where monitoring data are inadequate,fate models are useful for estimating chemicalconcentrations. These models can span a widerange of complexity in terms of spatialdimensions and temporal assumptions (i.e. steady-state versus non-steady-state).Types of fate models include:

• simple dilution models where a measuredconcentration in an effluent is divided by adilution factor or the chemical release rate isdivided by a dilution factor or the chemicalrelease rate is divided by the bulk flow rate ofthe medium;

• equilibrium models which predict thedistribution of a chemical in the environmentbased on partitioning ratios or fugacity (theescaping tendency of a chemical from oneenvironmental phase to another);

• dispersion models which predict reductions inconcentrations from point sources based onassumed mathematical functions or dispersionproperties of the chemical; and

• transport models which predict concentrationchanges over distance and can representdispersion, biochemical degradation andabsorption (from WHO, 1999b, p. 42).

8.4 Environmental PersistenceThe terrestrial, aquatic and atmospheric fate ofagents needs to part of the exposure assessment.The agents may be relatively inert (e.g. asbestos)or subject to biodegradation and abioticdegradation. Persistent substances, or those withlong half-lives in the environment or biota, willincrease the opportunities for exposure over time.

The source of the data and the relevantenvironmental comparisons with Australianconditions needs to be taken into account. Forexample, Australian soils and climatic factors mayresult in different environmental persistence forsome pesticides in Australia compared to NorthAmerican or Northern European conditions.

8.5 Environmental Samplingand Analysis

Data collection entails the acquisition and analysisof information about hazards on a site that mayaffect human health and which will be the focusfor the particular risk assessment (US EPA, 1989).

Adequate data collection is the foundation to anacceptable health risk assessment.

Sampling is often carried out to more clearlydefine detected or suspected contamination and, ifremediation occurs, to verify that contaminatedmaterial has been removed and that anycontamination remaining does not constitute ahealth or environmental risk.

The greatest concern, in collecting samples, is toensure that the samples taken adequatelyrepresent potential exposures for the situation.Consequently it is essential to be fully apprised ofthe context of the risk assessment, the objectivesof the task, the environmental conditions at thesite locations and what analytes will be tested ineach sample, before sampling commences (Lock, 1996).

Inappropriate sample collection procedures ‘yieldsamples that are not representative of thepopulation of interest; are of little use; seriouslycompromise the purpose of sampling; andcontribute to the uncertainty of the analyticalresults’ (Keith, 1990, p. 610).

Laboratory errors can occur and if an aberrant oran unexpected result is provided the potential forlaboratory error should always be considered.

An important aspect of Environmental Samplingand Analysis is that the environment is not staticand sampling results can vary over time. Theinterpretation of Environmental Sampling datashould take this into account.

8.5.1 Data quality objectivesData quality objectives ‘provide critical definitionsof the confidence that must be inherent in theconclusions drawn from the data produced by thewhole project’ and determine the degree ofuncertainty or error that can be tolerated in thedata (Keith, 1990, p. 611).

Data Quality Objectives which clearly specify theamount, nature and quality of the data to becollected should be detailed. Data QualityObjectives will be situation-specific. More detailis given in EPA QA/G 4 Guidance for the dataquality objectives process. Washington: USEnvironment Protection Agency.EPA 600R96055.

The criteria for both accepting and rejecting datashould be rigorous.

Consideration will need to made as to whetherroutinely collected historical data will be asappropriate to use as data collected de novo for therisk assessment.


93


8.5.2 Sampling strategiesIn sampling, the statistical considerations need tobe matched to expertise in situation assessmentand a knowledge of the particular situation (Lord, 1987). The sampling plan and decisionsregarding the number, type and location ofsamples need to be developed with anunderstanding of the potential exposure pathways and routes (US EPA, 1989).

Sampling will be influenced by, and willinfluence, possible risk management decisions(Heyworth, 1991). The proposed human activitiesfor the particular setting will critically affect thenature of the sampling program.

The reasons for sampling include (Heyworth, 1991):

1. determining the nature of contamination;

2. determining the concentration anddistribution of the agent;

3. monitoring site conditions to determine ifremedial actions are required;

4. designing and implementing remedialactions; and

5. determining if remedial actions have beeneffective.

There are often three phases of sampling:

• an initial assessment to determine if detailedinvestigation is necessary;

• a detailed sampling and analysis plan; and

• post-remedial validation.

For any of these phases, a sampling program withmultiple stages may be required, especially forlarge and complex situations.

8.5.3 Sampling methodologiesNumerous techniques are available forenvironmental sampling.

General references are:

• Keith LH (1990). Environmental sampling: asummary. Environmental Science andTechnology. 24(5), 610-617;

• Keith LH (ed) (1988). Principles ofEnvironmental Sampling. Washington:American Chemical Society; and

• Perkins JL (1997). Modern IndustrialHygiene. Volume 1. Recognition andEvaluation of Chemical Agents. New York: VanNostran Reinhold.

Sampling is often most effectively done as astaged and iterative procedure where earlierresults can focus later sampling stages.

Fugacity modelling (Section 8.3) may provideassistance in determining where elevatedconcentrations of an agent are likely to be found.

Some key issues are (Keith, 1990):

• When sampling water, allowance should bemade for the fact that stratification can occurin bodies of water particularly in lakes deeperthat 5 metres, deep rivers, and in situationswhere two streams merge, such as where aneffluent enters a river;

• Groundwater contamination is affected by‘depth to water, recharge rate, soilcomposition, topography (slope), as well asother parameters such as the volatility andpersistence’ of the substance (Keith, 1990,p. 614). There is always a significant risk ofcross contamination of aquifers when sinkingbores and special precautions should be madeto protect against this;

• The contamination of water samples is alwaysa problem and this is most pronounced whenvery low concentrations are being sought; and

• Considerable variation in an environmentalmedium over time may occur andenvironmental sampling may need to be spreadover a period of time to give an accuraterepresentation of potential human exposures.

8.5.4 Sampling patternsSampling plans will depend on the medium beingsampled. If there is sufficient information about asituation, random sampling may be inappropriateor inefficient and judgemental sampling may bemore appropriate. Air and water over a small areaare likely to be more homogeneous than soil.

94

Further general information on sampling plans isavailable from:

• Gilbert RO (1987) Statistical Methods forEnvironmental Pollution Monitoring. NewYork: Van Nostrand Reinhold;

• Heyworth J (1991) Sampling and StatisticalAnalysis for Assessing Contaminated Sites.In: El Saadi O, Langley AJ. The Health RiskAssessment and Management of ContaminatedSites. Adelaide South Australian HealthCommission. p. 15–30;

• Keith L.H (1990). Environmental sampling:a summary. Environmental Science andTechnology. 24(5), 610–617; and

• Keith LH (ed) (1988) Principles ofEnvironmental Sampling. Washington:American Chemical Society.

8.5.5 Sampling density‘Statistical equations are tools to be used as aids tocommon sense and not as a substitute for it’ (Keith,1990, p. 612). Statistical formulae for determiningsampling density are usually based on therequirements that the results will be normallydistributed (i.e. in a bell-shaped curve) and that aparticular concentration is equally likely to occur atany point. Some analytical techniques require anestimate of the mean of the results and the standarddeviation of the results before sampling density canbe calculated. These requirements can rarely be metduring the stages of initial and detailedinvestigations as sites are often heterogeneous witha highly skewed distribution of results.

Sampling is a screening process and false positiveand false negative results will occur. From a healthperspective the aims of sampling are to reduce thelikelihood of a false negative that could resultultimately in significant adverse health effects,and to enable the identification and adequateremediation of contaminated sites sufficient toprotect human health.

A considerable amount of expert judgement isrequired to determine the amount of sampling.The final amount will depend on an integratedappraisal of factors including:

1. proposed or current human activities;

2. the number of stages of sampling consideredfeasible;

3. the scale and distribution of potential humanexposures; and

4. potential remediation and managementstrategies.

The sampling density requirements will vary frommedium to medium.

8.5.6 Sample handling, storage and transport

Sample handling and transport should be doneaccording to relevant regulatory documents orAustralian Standards.

Some key issues are (Keith, 1990):

• contamination may arise from substances inthe sampling devices and storage containers.PVC and plastics other than teflon tend tosorb organics and leach plasticisers and otherchemicals used in their manufacture. Somepesticides and halogenated compoundsstrongly adsorb to glass (Keith, 1990);

• the loss of volatile analytes or reducedconcentrations from irreversible absorption onthe walls of sampling containers can be asignificant problem; and

• sample preservation can be of considerableimportance. If incorrectly stored, materialscan have accelerated breakdown, chemicalsmay be lost by volatilisation, and proliferationor diminution of microbiological organismscan occur. The nature of the storagecontainer, its seal, and the degree ofrefrigeration needed should always beconsidered and addressed.

Specific references for air, food, water, and soil aredetailed in Appendices 1–4.

8.5.7 Chain of custodyThe consultant’s report must provide thefollowing chain of custody information (EPANSW, 1997, p. 12):


95


1. the sampler;

2. nature of the sample;

3. collection date;

4. analyses to be performed;

5. sample preservation method;

6. departure time from site; and

7. dispatch courier(s).

AS 4482.1-1997 Appendix H provides a Chainof Custody form.

8.5.8 Analytical methodologiesGood (1993) considers an appropriate testmethod must be (p. 45):

• ‘accurate: it must be shown to give resultswhich differ little from the concentration wewould accept as the ‘true’ value. This isgenerally demonstrated by comparison withother, well respected techniques;

• precise: it gives results which show acceptablysmall variation from batch to batch andanalyst to analyst when applied as prescribed;and

• robust: results are not unduly affected byminor variations in test conditions.’

If these three criteria have been measured, themethod can be relied upon to provide an answerwithin a predictably narrow range around theaccepted ‘true’ value for a given sample. For amethod to be widely useful however, it must alsobe (ibid, p. 45):

• ‘not too complex: A procedure so complex as to be only useable by a few highly trainedpersons will probably be of limited practicalvalue;

• not expensive: The costs of site assessmentsare already high;

• reasonably comprehensive: Methods shoulddetermine a reasonably wide range ofcompounds potentially present; and

• available: Even the best method is of littleuse if its use is restricted by copyright orother instrument, or it resides in an obscurejournal unknown to potential users.’

The original analytical records (e.g. traces,chromatographs) should be retained and shouldbe reviewed when the data are about to drive asignificant action.

General reference

• Manahan SE (1993). Fundamental ofEnvironmental Chemistry. Boca Raton:Lewis Publishers.

• Perkins JL (1997). Modern IndustrialHygiene. Volume 1. Recognition andEvaluation of Chemical Agents. New York:Van Nostran Reinhold.

Specific references

Specific references for air, food, water, and soil aredetailed in Appendices 1–4.

8.5.9 Choice of analytesThe choice of analyte will be principally governedby the ‘Issues Identification’ stage for theparticular situation.

8.5.10 Field instrumentsField instruments should be regarded only as ascreening tool and their results require laboratoryvalidation.

Field instruments require regular maintenanceand calibration, and skilled and diligent use.

The accurate use of such instruments relies onfactors including:

• the method of sampling;

• the nature of the contaminant;

• the presence of interfering gases or vapoursresulting in overestimates or underestimatesof environmental concentrations;

• the type and make of the instrument;

96

• the type of calibrant used;

• the length of time since the last calibration;

• the cleanliness of the instrument; and

• the skill and knowledge of the operator.

They may be useful for assisting in theidentification of areas where sampling should beconcentrated. They do not replace analysis in alaboratory.

Examples of field instruments are PhotoIonisation Detectors (PIDs) and X-rayFluorescence (XRF). Information on time, dateand method of calibration should be providedwith reports.

8.5.11 Quality assurance of dataused in site-specific healthrisk assessment

The following information is adapted from Good(1993, p. 44).

• Quality assurance (QA)All of the actions, procedures, checks anddecisions undertaken to ensure therepresentativeness and integrity of samplesand accuracy and reliability of analysis results.

In the field this includes selection ofappropriate sampling methods, documentationand sample storage, cleaning of tools beforesampling and between samples, cleaning ofcontainers, and maintenance of sampleenvironment to minimise samplecontamination and analyte losses.

In the laboratory, QA involves proper samplecontrol, data transfer, instrument calibration,selection of properly trained staff and suitableequipment, reagents and analytical methods.

• Quality control (QC)Those parts of QA which serve to monitorand measure the effectiveness of other QAprocedures by comparison with previouslydecided objectives. In the field, this mayinclude checking of sampling equipmentcleanliness by keeping rinses for analysis,cross-checking of sample identities, duplicatesampling of sites and performance of ‘field

blanks’ and ‘field spikes’. In the laboratory,QC procedures involve measurement of thequality of reagents, cleanliness of apparatus,accuracy and precision of methods andinstrumentation by regular analysis of‘blanks’, sample replicates, ‘spiked recoveries’and standard reference materials (SRMs),with proper recording of results for thesechecks and immediate investigation ofobserved problems.

According to these definitions, ‘adequate QAis achieved when the results of QCdemonstrate that agreed objectives such asfreedom from contamination, methodaccuracy and precision can be reliablyachieved. An important decision then is thecorrect level of QC’ (ibid, p. 47).

‘As a general rule, the level of required QC isthat which adequately measures the effects ofall possible influences upon sample integrity,accuracy and precision, and is capable ofpredicting their variation with a high degreeof confidence. QC is more often performedinadequately than very well’ (ibid, p. 47).

• BlanksA reagent blank (or preferably two for verylow level analysis), prepared by processingreagents only in exactly the manner used foreach sample. The aim of the blankdetermination is to establish the magnitudeof that component of the analytical signalwhich can be ascribed to contributions fromreagents, glassware, etc. The contributionestablished should be subtracted from thegross analytical signal for each analysis beforecalculation of sample analyte concentration.

• Replicate analysis (matrix duplicate)Repeat analysis of at least one sample. Thevariation between replicate analyses should berecorded for each batch to provide anestimate of the precision of the method.

• Recovery check or reference material analysisrecovery check (matrix spike)Analysis of one or more replicate portions ofsamples from the batch, after fortifying theadditional portion(s) with known quantities of


97


the analyte(s) of interest. Recovery checkportions should be fortified at concentrationswhich are easily quantified but within the rangeof concentrations expected for real samples.

• Reference material analysisAnalysis of a sample similar in matrix type tothe samples, with accurately knownconcentration of the analyte(s) of interest.Results of recovery checks and referencematerial analyses for each batch should berecorded so that the bias of a method may beestimated, and day-to-day method efficiencymay be monitored.

• Surrogate spikes and internal standardsWherever appropriate, especially forchromatographic analysis of organics, the useof surrogate spikes and internal standards ishighly recommended. Inclusion into methodsrequires little additional effort and greatlyenhances confidence in qualitative andquantitative results obtained.

• Surrogate spikesSurrogate spikes are known additions, to eachsample, blank and recovery/reference sampleanalysis, of known amounts of compoundswhich are similar to the analytes of interest interms of:

• extractability;

• recovery through clean-up procedures; and

• response to chromatographic or othermeasurement.

but which:

• are not expected to be found in realsamples;

• will not interfere with quantification ofany analyte of interest; and

• may be separately and independentlyquantified by virtue of (e.g.)chromatographic separation or productionof different mass ions in a GC/MS system.

Surrogate compounds may be alkylated orhalogenated analogues or structural isomersof analytes of interest.

The purpose of surrogate spikes, which areadded immediately before the sampleextraction step, is to provide a check for everyanalysis that no gross processing errors haveoccurred which could have led to significantanalyte losses or faulty calculation.

• Internal standardsImmediately prior to instrumental analysis,each sample, blank and recovery or referencematerial extract is fortified with a set amountof one or more compounds which:

• are not found in real samples;

• will not interfere with quantification ofanalytes of interest; and

• may be separately and independentlyquantified.

The purpose of internal standards inchromatograms is to provide extra peakswhich serve to check the consistency of theanalytical step (e.g. injection volumes,instrument sensitivity and retention times forchromatographic systems). Analyteconcentrations may be determined bymeasuring the RATIO of the analyteresponse to that of an internal standard, withmarked improvements in quantitativeprecision.

• Control chartsNadkarni (1991) claims that the heart of aQA/QC program is a control chart. ‘This is anumerical picture (a plot) of the variation ofmeasured QC parameter (e.g. blank andrecovery values). Data are plotted in thesequence in which they were obtained, andreviewed frequently in order to detect anyproblem with minimal delay. The use of thesecharts is highly recommended.’(Good, 1993, p. 47)

8.5.12 Safety plansThe safety of people assessing a situation andnearby residents must always be considered inenvironmental sampling. Site safety plans shouldbe developed where there may be risks.

98

A general reference, framed around contaminatedsites but applicable to a wider range of situations is:

• National Environment Protection Council(1999). National Environment ProtectionMeasure for the Assessment of SiteContamination. Guideline 9. Protection ofHealth and the Environment during theAssessment of Site Contamination. Adelaide:National Environment Protection Council.This is accessible at www.nepc.gov.au.

8.5.13 Assessment of summarystatistic data andpresentation of data

(adapted from Langley, 1993a, p. 23–28)

Vast amounts of data can be generated about asingle environmental health investigation. To enablean efficient and accurate appraisal of a situationrequires that the data be collated in a form thatallows an understanding of the location, extent,trends, and likely ‘behaviour’ of any environmentalhazards. Mapping of data is essential.

An adequate understanding of what is (and willbe) occurring is almost impossible to achieve frompages of raw data especially where there areabnormal results or more than a handful ofresults. At its worst sample identificationnumbers, sampling points, technical logs, andresults for each analyte will be on separate pages.

There is a constant tension between consultantswho wish to maintain individuality to theirreports and government agencies which seekuniform reports. A uniform approach to thelocation and presentation of data makes for morerapid and accurate assessments of reports.

The major problems that can occur with data setsand assessments are:

• a failure to collate data and to condense intocomprehensible tables;

• providing cluttered data sets, tables and graphs;

• treating the sum of the data as somewhatgreater than the sum of its parts. This isexemplified by elaborate contour maps basedon a very limited number of data points;

• providing fairly definitive conclusionsinsufficiently underpinned by supportingdata;

• considering the numbers in isolation fromother data important to interpretation e.g.situation history and characteristics of thesampled medium; and

• inappropriate ‘compositing’ of data.

Summary statistics

No single summary statistic (e.g. an arithmeticmean or the median) fully characterises asituation. Instead a range of summary statistics isneeded to build up a picture of potential agentsand exposures.

Each summary statistic will have a contribution,but will also have certain limitations. For example,the mean is affected by each individual score andis particularly sensitive to extreme scores.However it is less sensitive to sampling variationthan the median or mode i.e. it is less affected byrepeated series of random samples from the onepopulation. The median is less sensitive than themean to extreme scores and usually more sensitiveto sampling variation (but less so than the mode)(Pagano, 1986).

Given the complex nature of most data sets, arange of summary statistics needs to be presentedas the mix of summary statistics will be moreuseful than a single summary statistic. Examplesof ways of presenting summary statistics,particularly where there are multiple agents areshown in the Summary Statistics section forContaminated Sites (Appendix 1).

As much of our sampling is judgemental ratherthan random, caution needs to be taken with theuse of conventional statistical methods whichusually assume the random collection of data andthe use of normally distributed data. Much riskassessment data is log normally distributed or hasskewed distributions and this will requiredifferent statistical methods for analysis.

Outlying data should be appropriately consideredand not neglected.


99


Contouring

While graphical representations of contours canprovide useful information about situations suchas the distribution and ‘trends’ of environmentalhazards, contouring often is based on extremeextrapolations from inadequate amounts of data.If the distribution of the environmental hazard isheterogeneous it is unlikely that there will besufficient data points or sufficient associationsbetween adjoining points for contouring to beused with any confidence in its meaning (e.g. most contaminated sites are likely to have aheterogeneous distribution of contamination).Where there is widespread or relativelyhomogeneous distribution of an environmentalhazard contours may provide fairly usefulinformation on a macro scale. Examples areplumes of regional contamination such as arounda lead smelter or sewer outfall.

When contouring is used, there is a need todemonstrate that the model used for contouring is valid.

Mapping of data

Mapping the results is essential but poor designcan cause clutter that obscures important data.

If there is ‘too much’ data available, this may beaddressed by putting only significant results ontothe map. However, this should be done cautiouslyas ‘censoring’ some of the data can obscure trends.‘Normal’ results are important if elevated resultswere anticipated and may need to be included toprovide a useful comparison to the abnormalresults. Other superficially unimportant data canprovide surrogate information about theenvironmental hazards.

A series of transparent overlays, each with adifferent data, can be very useful to reducecluttering.

Geographic information systems

Geographic information systems (GIS) allowspatial relationships between populations andhazards to be examined and it can be useful forthe Hazard Identification and Exposure

Assessment phases of risk assessment. ModernGIS tools allow visualisation of relationshipsbetween data in two or three dimensions, forinstance, it can allow visualisation of certainsymptoms or diseases in regard to theirgeographic location. The relationships may bebetween health, environment and socio-economicdata at many geographic scales, starting with theindividual person e.g. a person’s place of residenceor work. Data can be aggregated for ageographical area and patterns betweengeographical areas visualised.

GIS may also allow certain complex analyses tobe done such as shortest path or best pathanalysis. Path analysis allows predictions ofpopulation behaviour in relation to geographicalvariations. Path analysis will allow traffic flowpatterns and densities to be predicted to assess thevariations in exposures to benzene across a city.

In the case of a specific hazard, path analysis mayallow the estimation of exposure to givenpollutants, allowing opportunities for strategicpublic health interventions to be undertaken. Forexample, estimating shopping location patterns toidentify the populations most likely to have beenexposed to a Legionella-contaminated coolingtower or enabling preliminary rankings of risk fora number of towers when a case of Legionella isreported. The Agency for Toxic Substances andDisease Registry (ATSDR) in the USA has usedGIS to identify populations residing nearhazardous waste sites.

3D representations of data

Data presented as 3D illustrations can beparticularly useful in uncluttering information andproviding a ‘picture’ of what is occurring. Incontrast 3D graphs of data are often misleadingcompared to 2D graphs.

Some principles of graphical representation

Tufte (1983) points out that ‘graphical excellenceis that which gives to the viewer the greatestnumber of ideas in the shortest time with theleast ink in the smallest space’ (p. 51).He goes on to say that ‘graphical excellence is the well-designed presentation of interesting

100

data—a matter of substance, of statistics, and ofdesign…and consists of complex ideascommunicated with clarity, precision, andefficiency’.

Tufte (1983) censures cluttered tables and otherfailings of graphic design such as:

• excessive zeal in the use of computer softwaregraphics packages so that bold cross-hatchingand the use of wavy lines lead to ‘optical art’effects; and

• overdoing the use of horizontal and verticallines in tables. Tufte quotes Tschichold(1935), ‘tables should not be set to look likenets with every number enclosed’.

Some basic principles of graphic representationare given in Table 13.

For effective graphic presentation, Cleveland(1994) recommends:

• avoid excessively complicated graphs;

• avoid pie charts, perspective charts (3D bar and pie charts, ribbon charts),pseudo-perspective charts (2D bar or linecharts);

• use colour and shading only when necessaryand then, only very carefully;

• when possible, accompany graphs with tablesof data;

• if probability density or cumulativeprobability plots are presented, present themwith identical horizontal scales (preferably onthe same page), with the mean clearlyindicated on the curves; and

• do not depend on the audience to correctlyinterpret any visual display of data: provide anarrative in the report interpreting theimportant aspects of the graph.(US EPA, 1997)


101

Table 13: Useful vs not useful graphics

Useful Not useful

No cryptic abbreviations Numerous abbreviations requiring searching the text for explanation

No elaborate encoding

Words run in natural left to right direction Words run vertically or in several directions.Letters running vertically may be even worse

No elaborate shadings, cross hatchings and overpowering colouring.

Simple labelling or graphic means Elaborate or obscurely coded patterns requiring continualno legend or key is required return to legend or key.

Simple, upper and lower case font with serifs, Multiple overbearing fonts, in upper case sans serifmodestly and consistently used.

Clearly printed Murky and clotted printing

Enlightens and arouses curiosity Graphic repels interest and obscures meaning.

(Langley, 1993a; adapted from Tufte, 1983)


It needs to be absolutely clear when log ratherthan logarithmic scales are being used on the axesof graphs.

Besides the work by Tufte (1983, 1990), otheruseful general references are:

• Kosslyn SM (1994). Elements of graph design.New York: WH Freeman and Co.

• Cleveland WS (1994). The elements ofgraphing data. Summit, New Jersey:Hobart Press.

Cost of graphics

Graphic work is usually time-consuming and thecost of this may be significant. However,particularly for large and complex situations, someform of graphic representation is imperative forthe assessor and other stakeholders to visualiseaccurately a model of what is occurring on a site.Without such representations inaccurate (andprobably costly) decisions will be made and riskcommunication and community consultation willbe much more difficult.

Photography

A photographic record that is well-labelled fordate, location and orientation is a valuablereference during the inspection (e.g. topography,soil staining, stack emissions, algal blooms,industrial processes, plant toxicity, proximity ofhousing), and assessment (e.g. the soil stratademonstrated in test pits and soil cores). Goodphotography will provide considerable assistancefor those unable to undertake an inspection of thesituation.

Supplying data on disc

Consultants, assessors and government agenciesshould have access to data on disc or otherelectronic formats as it:

• avoids a further source of transcription error;and

• facilitates the further analysis of data usingother software packages.

8.5.14 Censored dataCensored data are those which are below the levelof detection. Summary statistics can be biasedaccording to the values substituted intomathematical formulae to allow calculations of,for example, means. Often the value of the levelof detection is substituted, upwardly biasing thesample statistics. The approach to censored datamust be clearly stated.

Levels of reporting

The first step in dealing with censored data is toensure that the levels of detection or levels ofreporting are appropriate. The levels of reportingmust be less than the relevant criteria againstwhich the results will be assessed. A level ofreporting of no more than 10 per cent of therelevant criterion should be adopted. Where thismay entail substantial costs, a higher level may betolerable.

Diminishing levels of reporting

Improved analytical techniques have led to levelsof reporting decreasing enormously. For example,the detectability of benzene in water has increased by over 10 000 fold since the 1960s (Hrudey,1998). For some substances picogramconcentrations (10–12g) can be detected incommercial laboratories and femtogram (10–15g)in research institutions.

Dealing with censored data

Heyworth (1991, p. 24) summarises Helsel (1990)and provides a summary of the three essentialmethods for dealing with censored data:

1. Simple substitution methodsSimple substitution methods refer to thosemethods which substitute a single value, suchas one-half the detection limit for eachcensored value. While these methods arecommonly used they have no theoreticalbasis. The choice of the substitution value isessentially arbitrary and the estimates ofsummary statistics will be biased by thesefabricated results.

102

2. Distribution methodsThe distribution method uses thecharacteristics of the assumed distribution ofthe data. For environmental monitoring thelog normal distribution is usually assumedand values of data above and below thereporting limit are assumed to follow thisdistribution.

Estimates of the mean and standard deviationare computed using the best match from theobserved data and percentage that fall belowthe limit. Estimation methods includemaximum likelihood estimation andprobability plotting procedures. Thesemethods will produce unbiased estimates onlywhen the observed data fits the distributionexactly and the sample size is large. This, ofcourse, is a rare case. However, they providebetter estimates than those obtained bysimple substitution.

3. Robust methodsThe robust method combines the observeddata above the detection limit withextrapolated below-limit values to computesummary statistics. In contrast to thedistribution methods the actual data abovethe reporting limit are used to fit adistribution rather than assuming adistribution.

This method has the advantage that estimatesof extrapolated values can be directlyretransformed and summary statisticscomputed in the original units, therebyavoiding transformation bias. Also this methodis not as sensitive to the fit of the distributionfor the largest observations because actualobserved data are used to fit the distribution.

The probability plotting method used to fitthe distribution in robust methods can becomputed quite readily by most commerciallyavailable statistical packages.

Helsel (1990) recommends the use of robustmethods, particularly when the data cannotbe assumed to follow a defined distribution.He concludes that the use of these methods,rather than simple substitution methods for

environmental data, should reduce estimationerrors for summary statistics substantially.

‘Simple substitution is an inappropriatemethod of dealing with less than detectablevalues as it has no theoretical basis’(Heyworth, 1991, p. 25). Simple substitutionmethods of dealing with censored data mayresult in significant over-estimates of risk ifthe level of reporting is used as a value forcensored data and the concentration of theagent does not approach the level ofreporting, or is not present at all. Under-estimates of risk can occur if, for example, avalue of half the level of reporting is used butactual concentrations of the agent are actuallygreater than this.

The use of either the distributional or robustmethods is recommended, but the latter ispreferred. Commonly available statisticalpackages readily enable the use of robustmethods for dealing with censored data.

The values for the median and interquartilerange generally are not affected by censoreddata (ibid).

8.6 Meteorological DataMeteorological data will be particularly importantin the evaluation of both point source andgeneralised air pollution and potential exposuresof populations. When environmental monitoringis being undertaken, there is a need to haveconcurrent meteorological data.

8.7 Content of EnvironmentalSampling and AnalysisReports

8.7.1 Integration of reportsWhere there is a series of reports, each succeedingreport should summarise the important andrelevant points from the preceding reports. Thiswill assist in the rapid comprehension of newmaterial by all parties involved.

Non-integrated reports result in far less efficientappraisals of data.


103


8.7.2 Sampling issuesThe basis of the sampling program should beclearly justified.

8.7.3 Analytical issues

Chain of custody

An uninterrupted chain of custody should bepresent to ensure the quality of the results.

Accreditation of laboratories

Laboratories should be accredited by anappropriate body for the particular analyses beingundertaken. A broad form of accreditation maynot be applicable for the particular test.

Liaison with laboratories

There should be one person responsible tocoordinate samplers, risk assessors and laboratoriesto ensure that appropriate data ensues. This personshould be identified in the report. This coordinationshould also ensure that analytical methods andsampling protocols deliver data useful to the riskassessment and that the communication occurringbetween the various parties is meaningful andextends beyond the transfer of samples and data.

Choice of analytes

The analytes chosen must be applicable to therisk assessment, either directly (e.g. lead near alead smelter) or as surrogate measures of anenvironmental hazard (e.g. turbidity as a measureof water quality, coliforms as a measure of foodsafety).

Analytical techniques

The analytical techniques should comply withtechniques described in relevant protocols.

8.7.4 Situation descriptionsA situation description should, where relevant,contain the following (adapted from Edwards et al, 1994, p. 5; EPA NSW, 1997, p. 8):

• Situation definition and description.Where this applies to specific geographicalareas, the boundaries of these should be clearlyand accurately identified with respect to roads,adjacent properties and geographical features.

A current plan of the site, with scale bar,indicating the site orientation (including north)and general contours of the property, localwater drainage and other environmentallysignificant features is essential as well as alocality map. If historical factors are relevant tothe site are important, a series of aerialphotographs with dates may be warranted;

• Zoning. This will include previous presentand proposed zoning and relevantdevelopment and building approvals records;

• Present and past industrial and non-industrial activities/uses with as muchprecision as possible. For industrial activitiesthis may require details of: raw materials;products; intermediate products andbyproducts; and wastes;

• Present (and previous) buildings andstructures where relevant;

• Waste Disposal Practices and Locations.Locations of solid waste disposal areas andliquid waste lagoons, settling tanks and sumpsshould be identified in the maps and figures;

• Discharges to land, air and water;

• Product Spills, Losses, Incidents andAccidents (including fire). These should belisted chronologically together with anindication of the material spilled, estimates ofquantity, extent of fire damage, andcommunities and structures affected;

• Sewer and underground service plans;

• Chemical storage and transfer areas;

• Adjacent Land Uses. The emissions andcontaminant plumes form adjacent land usesshould be considered;

• Relevant history of complaints;

• Local knowledge of residents and staff;

• Details of building and related permits,licences, approvals and trade wasteagreements; and

• Validity and integrity assessment of the aboveinformation.

104

8.7.5 Site inspectionsThe remaining sections are adapted from EPANSW (1997)

Where relevant a site inspection performed aspart of a risk assessment should contain thefollowing:

• topography;

• conditions at the site boundary;

• visible signs of environmental hazards e.g.discolouration and staining of soil, plant andanimal toxicity;

• presence of waste materials;

• odours;

• condition of structures e.g. presence ofabnormal deterioration, moulds and fungi,adequacy of ventilation;

• quality of surface waters;

• flood potential; and

• details of relevant local sensitive environmentse.g. rivers, lakes, creeks, wetland, local habitatareas, endangered flora and fauna.

8.7.6 Sampling and analysis plan andsampling methodology

Where relevant, a risk assessment report shouldcontain the following:

• Sampling, analysis and data quality objectives(DQOs);

• Rationale for the selection of:

- sampling pattern;

- sampling density including an estimatedsize of the residual contamination;

- spots that may remain undetected;

- sampling locations including locationsshown on a site map;

- sampling depths for soil and water, heightfor air;

- samples for analysis and samples notanalysed;

- analytical methods; and

- analytes for samples.

• Detailed description of the sampling methodsincluding:

- sample containers and type of seal used;

- container pretreatment;

- sampling devices and equipment e.g. augertype;

- equipment decontamination procedures;

- sample handling procedures;

- sample preservation methods and referenceto recognised protocols, e.g. APHA (1992)or US EPA SW 846;

- sample pretreatment; and

- detailed description of field screeningprotocols.

8.7.7 Field quality assurance andquality control (QA/QC)

Where relevant a risk assessment should containthe following:

• details of sampling team;

• decontamination procedures carried outbetween sampling events;

• logs for each sample collected—includingtime, location, initials of sampler, duplicatelocations, duplicate type, chemical analyses tobe performed, site observations and weatherconditions;

• chain of custody fully identifying—for eachsample—the sampler, nature of the sample,collection date, analyses to be performed,sample preservation method, departure timefrom the site and dispatch courier(s);

• sample splitting techniques;

• statement of duplicate frequency;


105


• field blank results;

• background sample results;

• rinsate sample results;

• laboratory- prepared trip spike results forvolatile analytes;

• trip blank results; and

• field instrument calibrations (when used).

8.7.8 Laboratory QA/QCWhere relevant a risk assessment should containthe following:

• a copy of signed chain-of-custody formsacknowledging receipt date and time, andidentity of samples included in shipments;

• record of holding times and a comparisonwith method specifications;

• analytical methods used;

• laboratory accreditation for analyticalmethods used;

• laboratory performance in inter-laboratorytrials for the analytical methods used, whereavailable;

• description of surrogates and spikes used;

• per cent recoveries of spikes and surrogates;

• instrument detection limit;

• matrix or practical quantification limits;

• standard solution results;

• reference sample results;

• reference check sample results;

• daily check sample results;

• laboratory duplicate results;

• laboratory blank results; and

• laboratory standard charts.

8.7.9 QA/QC data evaluationWhere relevant a risk assessment should containthe following:

• Evaluation of all QA/QC information listedabove against the stated Data QualityObjectives, including a discussion of:

- documentation completeness;

- data completeness;

- data comparability (see next point);

- data representativeness; and

- precision and accuracy for both samplingand analysis for each analyte in eachenvironmental matrix informing data usersof the reliability, unreliability, or qualitativevalue of the data.

• Data comparability checks, which shouldinclude bias assessment—which may arisefrom various sources, including:

- collection and analysis of samples bydifferent personnel;

- use of different methodologies;

- collection and analysis by the samepersonnel using the same methods but atdifferent times;

- spatial and temporal changes (because ofthe environmental dynamics); and

- relative per cent differences for intra- andinter-laboratory duplicates.

8.7.10 Basis for assessment criteriaWhere relevant a risk assessment should containthe following:

• table listing all selected assessment criteriaand references;

• rationale for and appropriateness of theselection of criteria; and

• assumptions and limitations of criteria.

106

8.7.11 ResultsWhere relevant, a risk assessment should containthe following:

• summary of previous results, if appropriate;

• summary of all results, in a table that:

- shows all essential details such as samplenumbers and sampling depth;

- shows assessment criteria; and

- highlight all results exceeding theassessment criteria;

• site plan showing all sample locations, sampleidentification numbers and sampling depths;and

• site plan showing the extent of soil, air andwater contamination exceeding selectedassessment criteria for each sampling depth.

8.8 Modelling ExposuresModelling is used in exposure assessment ‘as ameans of forecasting human or other exposures inthe absences of complete monitoring or other data’(WHO, 1999). Modelling provides ‘a mathematical expression representing asimplification of the essential elements of exposureprocesses’ (ibid). Point estimates and probabilitydistributions are used in exposure modelling.

8.9 Use of Point Estimates andProbability Distributions

8.9.1 IntroductionPoint estimates are most commonly used inAustralia for exposure assessments. A pointestimate is a single value chosen to represent apopulation e.g. 70kg as the weight of an adult.Point estimates are usually typical values for apopulation or an estimate of an upper end of thepopulation’s value e.g. 70 years as the duration ofresidence on a property. An upper end value maybe chosen for reasons of conservatism and/or toprovide a ‘worse case’ scenario.

Where a risk assessment uses a series of upperend estimates, the result can be a worse than‘worse case’ scenario due to the compounding

effects of the estimates e.g. the person with theupper end value for weight is unlikely to alsohave: the upper end value for water consumption;the upper end value for contamination; the upperend value for duration of residence; the upper endvalue for soil ingestion, etc.

The estimate of the point estimate of a mean isusually more certain than a point estimate of thelevel intended to represent the 95th or 99thpercentile. This will present problems if there arelimited data for the use of point estimates if thepoint estimate is intended to be, for example, the95th percentile. For the same reason, similarproblems will arise if the tails of a probabilitydistribution are to be estimated.

Increasing attention has been paid to the use ofMonte Carlo-type exposure assessments and suchmethods have been acknowledged by the US EPAand the UK Department of the Environment (USEPA, 1992a; Ferguson, 1994).

‘While methods using probability distributionsare ‘more informative and inherently morerepresentative’ (Ruffle et al, 1994, p. 403) thanpoint estimates, if applied appropriately pointestimates still have a major role in exposureassessment as they are readily understood andapplied, and may incorporate safety factors thatcould be lost with Monte Carlo-type exposureassessments.’ (Langley and Sabordo, 1996)

The Monte Carlo-type exposure assessments relyon the use of probability distribution functions.A ‘distribution of possible values for each of theparameters (is) described along with theprobability of occurrence of each value’(Alsop et al, 1993, p. 407). Using standardmathematical formulae several thousand iterationsof a mock mathematical model are performed.

For each iteration, values for each parameter areselected randomly from each distribution basedupon the probability of occurrence. The estimatedrisk values are combined to provide a frequencydistribution of possible risk (ibid).(from Langley et al, 1998).

Figure 8 demonstrates the process of using theMonte Carlo method to estimate the probabilitydistribution of exposures in a population.


107

Figure 8: Principles of the Monte Carlo method

(adapted from Ferguson, 1994)


108

Establish probability

distributions for exposure factors in

a population

Repeated random sampling to build

output distribution of exposure

Sample randomly from probability distributions to create a single

estimate of exposure

Derive probability distribution for

combined exposure factors for population

COMPUTE

COMPUTE

COMPUTE

1 2 3 4

Monte Carlo analysis may add value to a riskassessment (US EPA, 1997, p. 5) when:

• exposures and risks are likely to approach orbe above levels of concern;

• screening assessments using conservativepoint estimates fall above levels of concern;

• it is necessary to disclose the degree of biasassociated with point estimates of exposure;

• when exposures and exposure pathways needto be ranked;

• where there is a need to appraise the relativevalues of collecting different types of furtherinformation (Cullen and Frey, 1999);

• when the costs of action are likely to be highand the gains are likely to be marginal;

• where the outcomes of action affect differentexposure pathways and the benefits need tobe ranked (Cullen and Frey 1999);

• sensitivity analysis is needed to apprise theimpact of default values and key pathways;and

• when the consequences of simplistic exposureassessments are likely to be unacceptable.

Monte Carlo analysis may not add value to a riskassessment (Cullen and Frey, 1999, p. 8) when:

• exposures and risks are likely to be negligible;

• when the costs of reducing the exposure andrisk are smaller than the cost of probabilisticanalysis;

• when safety is an urgent concern and actionmust be taken rapidly;

• when probability distributions are souncertain and/or incomplete that detailedprobabilistic judgements are unreasonable;and

• when there is little variability or uncertaintyin the data.

If a Monte Carlo assessment is performed themethodology must be ‘transparent’ or problemswill arise in community consultation. As with anyform of risk assessment, the basic principles of themethod must be able to be understood by theaffected community.

For small scale situations, the use of Monte Carlomethods is likely to be too complex and/or costlyand it may be more appropriate to do directmeasurements of exposure. The exposures of highend exposure ‘outliers’ must always beacknowledged in risk assessments and ways ofidentifying and accommodating them must beconsidered. This is particularly important in theassessment of an existing situation (e.g. acontaminated site where housing has already beendeveloped), rather than a forecast exposurescenario, where the presence of an ‘outlier’ willseverely test the credibility of a risk assessmentthat does not accommodate a range of exposurescenarios. (Langley and Sabordo, 1996)

8.9.2 Monte Carlo vs Latin Hypercube

Monte Carlo-type methods use ‘random (orpseudo-random) numbers to sample from theinput distribution … [so that] … samples are

more likely to be drawn from values that havehigher probabilities (e.g. near the mode)’ (AIHC,1994, p. 3.3). This could be important if there isconcern about exposures represented by the tailsof the distributions (e.g. 99 percentile exposures).Large numbers of iterations are required in anattempt to overcome this. Even so, Monte Carlo-type methods are more likely to result in undulyfrequent combinations of modal exposurescenarios (ibid).

Latin Hypercube techniques use randomsampling within equiprobable intervals of thedistribution so that there will not be clusteredsampling near the mode. It also ‘maintainscomplete independence of the variables’ but thisalso means if correlations are intended betweenvariables appropriate mathematical actions mustbe taken (ibid).

8.9.3 Use of Monte Carlo typetechniques in Australia

To date there has been very limited use of MonteCarlo techniques in Australia for EnvironmentalHealth Risk Assessment.

8.9.4 Weaknesses with the MonteCarlo technique

Some of the key limitations of the Monte Carlotechnique are:

1. ComplexityWhile the Monte Carlo method has a verygeneral applicability, changing one variablemay mean large amounts of recalculationbecause of the extent of the iterative processwhen using this model. The complexityreduces the ‘transparency’ of the method. Thismay create difficulties in communityconsultation and risk communication; itobscures errors, and creates difficulties forchecking by both the modellers andadministering authorities.

2. Loss of factor distinctionsThe method does not indicate ‘whichvariables are the most important contributorsto output uncertainty’ (US EPA, 1992,p. 22928).


109


3. Unrealistic probability assessmentsUS EPA (1992) notes that simulations suchas that found with the Monte Carlo modeloften ‘include low probability estimates at theupper end that are higher than those actuallyexperienced in a given population, due toimprobability of finding these exposures ordoses in a specific population of limited size,or due to non-obvious correlations amongparameters at the high ends of their ranges’.This results in overestimation of exposuredose or risk. The Science Advisory Board ofthe US EPA has noted that ‘for largepopulations, simulated exposures, doses andrisks above the 99.9 percentile may not bemeaningful when unbounded log normaldistributions are used as a default’(ibid, p. 22922).

4. Assessment endpointsWith Monte Carlo-type assessments there isstill a need to determine what is an acceptablelevel of exposure. Smith (1994) considers that‘the level of exposure exceeded by 1 in 20exposed persons would seem to be anappropriate reasonable maximum’ (p. 438).This would allow 5 per cent of the populationnot to be included in the exposure assessment.

5. Variability-uncertainty confusionSmith (1994) highlights the need todistinguish between ‘variability’ (measurablefactors that differ across populations such asheight) and ‘uncertainty’ (unknown, difficultto measure factors such as frequency oftrespassing on a site). Currently availablesoftware packages do not distinguish betweenvariability and uncertainty. An administratorreviewing a Monte Carlo risk assessment will,however, need to appreciate the differencesbetween variability and uncertainty and thenature and extent of both (ibid).

6. Limited exposure dataLimited information is available about manyvariables for the exposure assessments. As aconsequence of this, many input variables aredescribed as triangular distributions. Smith(1994) stresses the need to collect and verifydistributions from many currently undescribed

input assumptions (p. 438) to improve accuracy.The use of Monte Carlo methods may beinappropriate where the predictions of exposureare so dominated by uncertainties. McKone(1994) gives the example of benzo(a)pyrene,where information on benzo(a)pyrene exposureis ‘not readily available’ (p. 461) so that the useof Monte Carlo methods to assess variability inpopulation exposures is somewhat redundant.

7. Simplification of complex situationsExposure assessments are comprised ofcombinations of modelling, sampling, andmodelling/sampling combinations (McKone,1994). Even the use of complex models stillprovides a static picture of a dynamic worldalbeit a more elaborate representation ofreality (McKone, 1994) and such a picturemust be placed within a sound theoreticalframework.

8. Misleading precisionThe use of more complex models does notnecessarily increase precision (McKone, 1994,p 461). The costs of collecting and analysingdata, and constructing new models must bebalanced by the value of the informationobtained (ibid, p. 461). There is a need toappraise the value of information along withits uncertainties in defining the capabilitiesand limits of exposure models (p. 461).(Langley and Sabordo, 1996, p. 141).

9. Characterisation of extreme valuesThe 50th percentile can always be estimatedwith less uncertainty than the 99th percentile(Finley et al, 1994). Problems in estimatingthe extreme percentiles can come fromlimitations in the measurement techniques(e.g. incorrect and implausible estimates ofdietary consumption may be accepted intothe survey); the duration over which exposuredata was collected (see short term and longterm variation, below); and whether there aresub-populations who may have unusualexposures (e.g. vegetarians, subsistencefishermen) (Finley et al, 1994). Estimatingextreme percentiles can be a very time-consuming process.

110

8.9.5 Estimating distributions forexposure factors

Several factors affect the choice of distributions(Finley et al, 1994):

• Variability and uncertaintyVariability, as an inherent characteristic of apopulation, will not be reduced with additionaldata but will be more accurately characterised.Uncertainty however, will be reduced withadditional data.

Uncertainty may arise from factors intrinsic tothe available data (e.g. limitations of studydesign and analytical techniques) or from theapplication of data to non-sampled populations(e.g. extrapolating Scandinavian data to anAustralian population) (Finley et al, 1994).

The characterisation of uncertainty related toexposure factors has been developed furtherthan two other areas of uncertainty that may infact be more significant: the relationshipbetween the absorbed dose and the ultimatedelivered dose to a target organ; and theuncertainty about the response to the dose(Finley et al, 1994).

• Factor inter-dependenceSome factors such as body weight and skinsurface area are interdependent and this needsto be considered. Age specific data should beused as the factor may be strongly related toage (e.g. inhalation rates).

• Short-term and long-term variationInterpersonal variability will be decreased if thelength of time over which a factor is measuredis increased. Short term data tend tooverestimate inter-individual variation (ibid).For example, the 95th percentile of dietaryintakes from studies taken over one to threeday periods will be significantly higher than forstudies taken over longer periods such as onemonth to one year. This has been seen in thestudies of tap water consumption and fishconsumption (Finley et al, 1994). It can beparticularly marked for rare exposures (e.g.rarely eaten food such as shellfish).

Studies of shellfish consumption taken overshort periods of time may suggest only a verysmall proportion of the population consumesthe foodstuff and, if the common practice ofexcluding all non consumers is undertaken,there will be a poor characterisation of thevariability in the general population (Finley et al, 1994).

• Parametric versus non-parametric distributioncharacterisationFor data to meet parametric distributions (e.g.normal or log normal), appropriate statisticaltests must be met. Theoretically normal or lognormal density distributions do not have anupperbound limit yet for many factors (e.g.height, weight, fluid consumption) there areobviously physiological limitations to thefactors. Some of the currently available softwareenables such factors to be set within the model.

• Shapes of distributionsTriangular shape distributions are often used inMonte Carlo-type assessments but may beviewed as conservative characterisations oftruncated normal or log normal distributions(Finley et al, 1994, p. 535).

When establishing probability distributions,the distributions should be determined, wherepossible, from relevant data sets. If there is aneed to estimate a probability distribution, itshould be appreciated that that manyenvironmental health factors are likely to belognormally distributed rather thansymmetrically distributed. Examples of riskvariables that have been characterised bylognormal distributions are (Murphy, 1998):

• Body weight (each sex)

• Bioaccumulation

• Breathing rate

• Cancer potency factors

• Concentrations in

- Air

- Soil

- Tissue


111


- Water

• Drinking water rate

• Exposed skin

• Fish consumption

• Lifetime

• Residence time

• Shower duration

• Shower water use

• Soil ingestion rate

• Surface area/ body weight

• Total water use

• Toxic susceptibility

Much environmental data is lognormally ratherthan normally distributed. Table 14 gives someexamples of output variables that can berepresented by probability distributions.

8.9.6 Selecting appropriate data setsFor describing a probability distribution, therelevant studies and the quality of the dataproduced may vary considerably. Unless data setsare rigorously scrutinised the resulting uncertaintyin the range of risk estimates could be greaterthan obtained using point estimates (Finley et al,1994, p. 536).

Finley et al (1994) recommend the followingcriteria for assessing data:

• consistency with other studies;

• relevance of the survey population to thegeneral population or the population beingappraised as part of a risk assessment;

• minimisation of confounding variables; and

• whether there are sufficient data toadequately characterise variability and theextremes of the distribution.

Haimes et al (1994) propose several approaches tothe development of distributions when objectivedata is missing or scarce or not quite relevant:

• when data are sparse but relevant expertjudgement can be used to propose percentilesusing available data as ‘collaborators’ of theexpert judgement;

• where data are not quite relevant to propose adistribution for a parameter, expert judgementagain can be used collaborating thejudgement with analogous data; and

• where there is an absence of data the formalelicitation of expert judgement to construct adistribution (p. 693) can be used.

If there are a variety of studies then the purposes,designs and methodologies that are similar maybe able to be combined (Finley et al, 1994).

Haimes et al (1994) highlight the need toexamine the tails of probability distributionfunctions and submit them to a ‘reality check’ andexamine the combination of factors that resultedin the extreme values. They highlight the extremesensitivity of these tail values to assumptions andreinforce the need to assess the sensitivity of thetails to the assumptions. The assumptions need tobe examined as to whether they are mutuallyconsistent (ibid).

8.9.7 Principles for the use of Monte Carlo-type techniques

The purpose and scope of the risk assessmentshould be clearly articulated in the IssuesIdentification section. Burmaster and Anderson(1994) stress that any method of exposureassessment must have a clearly defined assessmentend point and provide all relevant information sothat the assessment can be reproduced andevaluated (p. 477). Burmaster and Anderson(1994) detail fourteen principles for good practicein Monte Carlo assessments.These are:

1. Detail all formulae.

2. Detail point estimates of exposure wherethese are demanded by regulatory agencies.

3. Detail sensitivity analyses to enable theidentification of relevant and important inputvariables. Those variables which will drive the

112


113

Table 14: Some key variables for which probability distributions might be needed

Model component Output variable Independent parameter variable

Transport Air concentration Chemical emission rateStack exit temperatureStack exit velocityMixing heights

Meteorological factors Wind speedWind direction

Deposition Deposition rate Dry-deposition velocityWet-deposition velocityFraction of time with rain

Overland Surface-water load Fraction of chemical in overland runoff

Water Surface-water concentration River dischargeChemical decay coefficient in riverMixing depth

Groundwater Groundwater concentration Predictions of plumes

Soil Surface-soil concentration Surface-soil depthExposure durationExposure periodCation-exchange capacityDecay coefficient in soil

Food chain Concentration in animal products Soil ingestion ratesPlant to animal bioconcentration factors

Plant concentration Plant interception fractionWeathering elimination rateCrop densitySoil-to-plant bioconcentration factor

Fish concentration Water-to-fish bioconcentration factor

Dose Inhalation dose Inhalation rateBody weight

Ingestion dose Plant ingestion rateSoil ingestion rateBody weight

Dermal-absorption dose Exposed skin surface areaSoil absorption factorExposure frequencyBody weight

(adapted from NRC, 1994, p. 169; adapted from Seigneur et al, 1992.)


risk assessment must obviously be included inthe Monte Carlo analysis but reasons forexcluding insignificant variables must also bedetailed.

4. Use probabilistic techniques (which may bedemanding in terms of time, money andother resources) only where exposurepathways are likely to be significant.

5. Provide detailed information about inputdistributions. The minimum stated byBurmaster and Anderson is:

• a graph showing the full distribution andthe location of the point value used in the[point estimate] risk assessment; and

• a table showing the mean, standarddeviation, the minimum (if one exists), the5th percentile, the median, the 95thpercentile, and the maximum (if one exits)(p. 478). There needs to be a sufficientjustification of the selected distributionwhich should be based on adequatelyreferenced sources and the statistical,physical, chemical, and biologicalmechanisms relevant to the distribution.

6. Detail how the input distributions captureand represent both the variability and theuncertainty in the input variables (p. 478) soas to enable both variability and uncertaintyto be described and analysed separately.

7. Use measured data to test the relevance of theinput distribution to the population, placeand time of the exposure assessment. Furtherdata may need to be gathered to supplymissing information or supplementincomplete information.

8. Describe the methods by which measureddata were used to derive a probabilitydistribution.

9. Detail any correlations between data wherethere are relatively high correlations.Sensitivity analysis may be necessary todetermine the effects of correlations betweenvariables on the exposure analysis.

10. Provide detailed information and graphs foreach output distribution. Burmaster andAnderson suggest the following as aminimum:

• a graph of the variable withadministratively set allowable risk criteriaas annotations and point estimates of riskusing the administratively set pointestimates of exposure; and

• A table of the mean, the standarddeviation, the minimum (if one exists),the 5th percentile, the median, the 95thpercentile, and the maximum (if one exists)(p. 479).

11. Provide records of sensitivity analyses andtheir impact that will enable thedetermination of the most important inputvariables (or groups of variables).

12. Assess the numerical stability of the centralmoments (mean, standard deviation,skewness, and kurtosis) and the tails of theoutput distributions. The latter areparticularly sensitive to the nature of the tailsof the input distributions and, as theystabilise very slowly, sufficient iterations arerequired to demonstrate the numericalstability. Burmaster and Anderson suggestthat commonly more than 10 000 iterationsare required. Software that enables Latinhypercube sampling results in more rapidstability of these output tails. Burmaster andAnderson state that the changes in the tails ofonly a few input distributions contributestrongly to changes in the upper tail of theoutput distribution (p. 480).

13. Detail the name and statistical quality of therandom number generator used. Somegenerators are inadequate because of shortrecurrence periods.

14. Interpret the results and detail the limitationsof the methodology such as the effects ofbiases not elsewhere interpreted.

Burmaster and Anderson state that the principlesare not mutually exclusive nor collectivelyexhaustive (Langley and Sabordo ,1996, p. 140–1).

114

8.9.8 Administrative requirementsfor the use of Monte Carlomethods

Regulatory authorities in Australia will requireassessments using Monte Carlo methods to meetthe following criteria:

1. Meeting the 14 principles of good practicedetailed above.

2. The provision of adequate information to theauthority to enable review of the assessment.This may require the provision of thesoftware (and underlying formulae) and data.

3. A demonstration of the relevance of theexposure data to the site: data from othercountries or cultural backgrounds may not be relevant.

4. An explanation of the data and methodwhich will be able to be understood by therelevant community.

5. The use of data that accounts for age andgender differences and takes into accountsusceptible populations.

On a large site divided into housing lots, theresults for specific housing lots that may beaffected by atypically elevated concentrationsshould not be obscured by averaging or MonteCarlo techniques applied to the entire site. Inmany instances, Monte Carlo methods will onlybe relevant to large sites or sites where directmeasurements of exposure are not practicable(Langley and Sabordo, 1996, p. 141).

The range of total acceptable exposures and riskwill need to be defined on a situation-specificbasis after consultation with stakeholders.Depending on how it is applied, the Monte Carlomethod may lose much of the conservatismusually inherent in point estimates.

8.10 Environmental Monitoring8.10.1 Personal monitoringWhere practicable, personal monitoring may playa central role in the exposure assessmentcomponent in the risk assessment process.

Monitoring methods used in exposure assessmentcan be categorised into direct and indirectapproaches. In the indirect approach to exposuremonitoring, factors that affect exposure aremeasured rather than exposure itself. Fixed-sitemonitors are used to measure pollutants in media,especially air and water (Covello and Merkhofer,1993).

Personal monitoring is a direct approach wherebyindividual human exposures at the point ofcontact are measured directly by instruments(personal exposure monitors or PEMs) thataccompany the individual (Wallace and Ott,1982). PEMs are designed to measure theconcentrations of agents in the air, water, or infood. In the case of food and water, individualsactively test the water or food before theyconsume it. PEMs are available for a limitedrange of agents.

Personal monitoring has been commonly used tomeasure exposures to carbon monoxide, volatileorganic compounds, to electromagnetic fields, andby radiation workers, who routinely carrydosimeters or film badges that measure exposuresto radiation.

Personal monitoring can be used to address someof the problems of exposure monitoring. It maybe able to collect data that integrates the greatdiversity of exposure pathways. Similarly it cantake into account the natural variability of theenvironment over time and space that makes itdifficult to translate measurements obtained fromfixed monitoring stations into actual exposuresexperienced by people that move from place toplace.

If enough subjects are selected for monitoring, apopulation exposure can be constructed. Becauseof time and cost constraints of portable samplingdevices, such techniques are not often used forassessing exposures in the general environment.They are used more frequently in theoccupational setting.

Passive and active personal samplers are available.Passive monitors capture the ambient air samplewithout mechanical assistance. Active samplers


115


direct the sample to the monitor via a pump thatis calibrated to pump air at a certain rate. Activesamplers provide more information than passivesamplers, because pollutant concentrations or thedose can be estimated more directly using activesampling. Passive samplers do provide time-weighted average concentrations rather thanspecific concentrations, but they are less costlyand bulky than active samplers and they are usefulin screening to determine if exposure has occurred(NRC, 1994).

When personal monitoring has been used in arisk assessment the following factors should beconsidered:

• the duration and frequency of monitoring(transient periods of high exposure may notbe detected if monitoring is not conducted atthe relevant times);

• environmental changes that may have affectedthe monitoring such as wind shifts;

• the accuracy and timing of calibration ofequipment;

• the accuracy and sensitivity of the monitoringtechnique (many techniques are designed forthe occupational setting and may beinsufficiently sensitive for assessing generalenvironment exposures where criteria areusually much lower);

• confounders e.g. a formaldehyde PEM mayrespond to a range of aldehydes (such asacrolein from wood smoke) and will alsodetect formaldehyde from a range of sourcesincluding cigarette smoke. This lack ofsource-specificity may present a limitationwhen particular sources need to be addressedby risk assessment;

• the relationship between short-term samplingand long-term exposures;

• bulky equipment may affect behaviour andhence exposure;

• tampering with equipment;

• time and cost constraints (i.e. it may be timeconsuming and costly to obtain enough direct

measurements to establish an accuratefrequency distribution of exposures within apopulation);

• bias in sample selection and poor responserates (this can lead to results which cannot begeneralised to the relevant population);

• accurate, valid and practical measuringmethods must be available (the number ofsubstances which can be reliably measuredwith personal monitoring is still small); and

• some analyses require specialised laboratories(and there may also be laboratory inaccuracy).

Personal monitoring is usually only feasible whererelatively small numbers of people are exposed toa limited range of substances. It is usedparticularly in the occupational environment. Inother situation such as when assessing exposure toair pollutants, environmental monitoring will bethe usual method.

8.10.2 Biological monitoring Biological monitoring is a measuring procedurewhereby validated indicators of the uptake ofcontaminants, or their metabolites, and people’sindividual responses are determined andinterpreted. By comparison, environmentalmonitoring measures the composition of theexternal environment around a person, biologicalmonitoring measures the amount of contaminantabsorbed into the body.

Biological monitoring may be direct e.g. themeasurement of lead in blood, or indirect e.g. themeasurement of the breakdown product ofnicotine, cotinine, in urine. Biological monitoringmay measure a biological effect, such as enzymedepression, or a physiological effect such astremor. The monitoring may be used to identifywhether exposure has occurred at all, or theamount of exposure.

If biological monitoring is practicable it will bemore valuable than environmental monitoring indetermining the level of risk from an environmentas it will measure whether exposure is occurringand the level of exposure (Langley 1991a). It canbe useful in identifying highly exposed individualsor sub-populations.

116

The biological samples used for monitoringinclude: blood, urine, fat, breast milk, hair, andexpired air.

Biological monitoring should not be commencedbefore:

1. The objective of the biological monitoring isdefined clearly.

2. A reference range of results is established thatis applicable for the population under study.This is often not available (or a control groupis not available to establish a reference range).The relationship of body burden levels andexposure (or risk) are unavailable for manysubstances.

3. Consideration has been given as to howresults are to be managed. Significant anxietymay be caused by factors such as: delays inproviding information; and an inability totake action if the person is distressed byelevated levels, perceives that any measure ofexposure is unsatisfactory or equates exposureto a health effect may cause.

4. The correct timing of sampling has beenestablished. Correct timing is critical forsubstances with short biological half-lives or aparticular exposure is of concern.

5. A process has been established to enableconsistent analysis and epidemiologicalappraisal of results.

6. The ethical and confidentiality aspects ofcollecting, maintaining and distributinginformation and results are fully considered.

Results should always be available to participantsin biological monitoring with an explanation ofthe results.

Several aspects must be considered:

• A good biological monitoring test result maynot correlate well with environmental levels(mainly because of human factors);

• The number of substances which can be usedreliably for biological monitoring is still small;

• Irritative, locally or rapidly acting substancesare usually unsuitable as the systemicabsorption may be minimal and/or irrelevantto the level of local reaction (e.g. SO2,ammonia, direct skin exposure to PAHscausing skin cancer);

• The substance must be in some tissue or fluidsuitable for sampling;

• Accurate, valid and practical measuringmethods must be available;

• The result should be interpretable in terms ofhealth risk; and

• The results may have more value for a groupthan an individual. (ibid)

8.11 Choice of Tissue8.11.1 Blood

• Depending on the biological half-life of asubstance, blood analysis may provide anindication of exposure from recent hours toseveral years. Levels are often transient if thehalf-life is not prolonged.

• The process of blood taking may beunacceptable for some people, includingchildren.

• When the volume of distribution is high,concentrations in blood are often too low tobe measured.

• Samples may require careful procedures suchas plasma separation and freezing.

• Substances measurable in the plasma may notbe responsible for the toxic effect which,instead, arises from a metabolite.

8.11.2 Urine• Only a limited number of substances can be

measured in urine because of degradation ofthe parent substance to breakdown products.

• Urine samples, in general, provide a moreintegrated assessment of exposure than bloodfor periods of recent hours or days.


117


• 24 hour sample collections may be moreappropriate than spot samples but manypeople find 24 hour collections are onerous.

• Urine samples require rapid processing andcooling.

8.11.3 Hair and toenails• Hair and toenails can provide an integrated

measure of exposure over a more prolongedperiod than blood or urine.

• Hair and toenails are only useful for chemicalsknown to accumulate in those tissues.

• Hair and toenails are inappropriate tissues forbiological monitoring on or nearcontaminated environments. Externalcontamination of the hair cannot beadequately removed during samplepreparation and an accurate measure ofexcretion via hair cannot be performed.

• Hair analysis may be useful for assessingintake from purely dietary sources when thereis no general environmental contamination.

8.11.4 Breast milk• The collection of breast milk is usually easy

and acceptable to nursing mothers.

• Breast milk provides an integrated exposurefor very lipid soluble compounds for timeperiods related to the biological half-life ofthe substance. Breast milk measurements ofPCBs, organochlorine pesticides and dioxinshave been used for exposure assessments.

• The concentrations must be standardised forfat content and may vary according to theperiod since breast feeding first commenced.

8.11.5 Expired air• Expired air is used to determine exposures to

ethanol and some solvents.

8.12 Choice of a TestOptimally, a biological monitoring test would givea result which reflected the exposure, theconcentration of the substance in the target organand the risks of adverse effects (Friberg, 1985).Few tests are available which approach this ideal(Langley, 1991a).

Where exposures from the environment are lowthis creates problems concerning accuratemeasurement at low levels and the possibility ofresults being overwhelmingly influenced by othersources of exposure (e.g. the influence ofcadmium in food, tobacco smoke and theoccupational environment will generally be fargreater than the influence of cadmiumcontamination of soils).

For many substances, biological monitoring isimpracticable because:

1. Analytical techniques are not available or areinaccurate at low levels or in the tissues orfluids being tested.

2. Insufficient information is available on inter-and intra-individual toxicokinetics andthresholds of health effects to enable riskassessment of results.

3. Insufficient epidemiological studies have beendone to determine normal ranges.

Substances for which biological monitoring ofgeneral environmental exposures is practicable aredetailed in Table 15.

There is a range of other substances for whichbiological monitoring may be available: the testsshould be assessed and used on their individualmerits for a particular situation. Biologicalmonitoring has been applied to a range ofsituations: tobacco use (polycyclic aromatichydrocarbons, aromatic amines and specificnitrosamines), dietary exposures (e.g. aflatoxins,N-nitrosamines, heterocyclic amines), medicinalexposures (e.g. cisplatin, alkylating agents,8-methoxypsoralen, ultraviolet photoproducts),occupational exposures (e.g. benzene, ethyleneoxide, styrene oxide, vinyl chloride, aromaticamines, polycyclic aromatic hydrocarbons).

118


119

Table 15: Substances likely to be suitable for biological monitoring

Substance Fluid/ tissue Comments

Lead Blood Urinary lead does not accurately reflect either recent exposures or body burden. Substantial data available on level of risk for particular blood lead ranges. Numerous Australian studies provide comparison data.Levels of concern available for both general population and occupational groups (WHO, 1986; NHMRC, 1987).

Cadmium Urine/ Blood Urinary levels tend to reflect body burden, blood levels reflect recent exposures. Urinary levels need to be adjusted for changes in urinary flow rates (results often given as µgCd/g Creatinine or µgCd/24hr). Laboratory inaccuracy has always been a major problem, particularly prior to 1980.Limited Australian studies to provide comparison data. Most international studies have concentrated on occupational exposures. Very limited data on children, especially for those less than 5 years. World Health Organisation (cited in Mueller et al, 1989) has set levels of concern. General diet and smoking will tend to have a major influence on levels.

Arsenic Urine Short biological half life—study must be done before study must be done during exposure (or at most within 1–2 days afterwards). Considerable interference from organic sources of arsenic (e.g. seafood)—dietary sources from the environment not under study need to be excluded and testing for inorganic arsenic undertaken. Limited comparison data and no set levels of concern.

Mercury Blood, Urine At equilibrium, the concentration of mercury in the blood reflects daily intake and is probably the single best indicator of exposure. This measure will also include methylmercury from fish and a fractionated analysis of mercury salts and alkylmercuric compounds may be required (Aitio et al, 1988). Methylmercury exposure will not affect urinary mercury levels although urinary levels show significant diurnal variation. Some international comparison data is available (ibid).

Polychlorinated Blood, adipose Long biological half-life so that historical exposures (i.e. body burden) may biphenyls tissue (fat) be able to be monitored. Different PCBs will have different behaviours in (PCBs) the body and different biological half-lives. Some comparison data

available. It is difficult to obtain adipose tissue samples and blood sampling is usually preferred.

Organochlorine Blood, Adipose Long biological half-life so that body burden can be assessed.pesticides e.g. tissue (fat) Some comparison data available, especially for blood. It is difficult to aldrin, dieldrin, obtain adipose tissue samples and blood sampling is usually preferred.chlordane,heptachlor

Organophosphorus Blood Cholinesterase levels will enable physiological response to be monitored.pesticides e.g. Wide range of normal values require individual baseline values to enable malathion, an assessment of ‘normality’.chlorpyrifos

(Langley, 1991a)


120

Besides the pesticides mentioned in Table 15specialised tests may be available from somelaboratories for pesticides such as glyphosate.

Most organic contaminants are not amenable tobiological monitoring in general environmentalsituations because of the low levels of exposureand the lack of comparison data compared tooccupational situations. Specialised studies maymake biological monitoring for some inorganicsubstances practicable (e.g. manganese, radioactiveisotopes).

A good knowledge of the toxicokinetics of asubstance is required for the correct choice ofmethod and interpretation of results. The durationof persistence of the agent will be important as isthe volume of distribution e.g. many very lipidsoluble substances with a very high volume ofdistribution have such low blood levels that theycan’t be measured in blood but can be identified inbreast milk. Individual results may be distorted ifthere is not constant exposure or equilibriumwithin the body (Langley et al, 1998).

Cytogenetic testing may occasionally be of valuebut is often difficult to interpret as only smallnumbers of cells are usually examined so thatthere is the potential for considerable confidencelimits around the results and because there canrarely be a link made to specific agent (oneexception is aflatoxin). Tests such as SisterChromatid Exchange and Micronuclei arenon-specific tests. There are problems with

confounding, distinguishing recent from historicalexposures, quantifying exposures and dealing witha finite background incidence of chromosomalabnormalities.

Under the NOHSC National Model Regulationsfor the Control of Workplace HazardousSubstances (adopted by the States andTerritories), health surveillance is required forspecified substances. Biological monitoringmethods developed for some of these methods aredetailed in the NOHSC series ‘Guidelines forHealth Surveillance’.

8.13 BiomarkersThe term ‘biomarker’ has been used in recenttimes to describe the measurements used inbiological monitoring. The term refers broadly toalmost any measurement reflecting an interactionbetween a biological system and an environmentalagent, which may be chemical, physical orbiological (WHO, 1993). Three classes ofbiomarker are identified by WHO (1993, p. 12):

• biomarker of exposure: an exogenoussubstance or its metabolite or the product ofan interaction between a xenobiotic agent andsome target molecule or cell that is measuredin a compartment within an organism;

• biomarker of effect: a measurablebiochemical, physiological, behavioural orother alteration within an organism that,depending upon the magnitude, can berecognised as associated with an establishedor possible health impairment or disease; and

• biomarker of susceptibility: an indicator of aninherent or acquired ability of an organism torespond to the challenge of exposure to aspecific xenobiotic substance.

For many environmental pollutants, the flow ofevents between exposure and health effects is notwell understood. Biomarkers help address thisproblem by improving the sensitivity, specificityand predictive value of detection andquantification of adverse effects at low dose andearly exposure (ECETOC, 1989; Fowle, 1989;Fowle and Sexton, 1992; NRC, 1992). Sensitivesubpopulations can be better pinpointed bybiomarkers that measure increased absorption rateor a more severe biological response to a givenenvironmental exposure (Lauwerys, 1984;ECETOC, 1989; Fowle and Sexton, 1992;Hemminki, 1992; NRC, 1992) (from WHO,1999b, p48).

8.14 Health Monitoring Health monitoring is the organised medicalassessment of individuals and groups of people.The medical assessment will consist of historytaking and clinical examination, and whereindicated, particular tests (e.g. lung functiontesting where there is a concern about the effectof air pollutant). The epidemiological aspects ofhealth surveys are covered in Section 4.8–4.11.

In Australia, health effects are likely to be foundin only a limited number of situations ofenvironmental contamination. Subtle effects mayonly be able to be determined on a group basisrather than on an individual basis (e.g. subtleneurodevelopmental effects determined bysophisticated testing in groups of children withdifferent lead exposures). Similar problems ofcausation relating to individual findings ratherthan group findings arise if the putative effects arecommon in the general population e.g. headache,fatigue. Health effects are rarely as specific to anexposure as chloracne with PCB or dioxinexposure.

Health monitoring for specific health effects iswarranted where environmental or biologicalmonitoring has indicated a significant risk ofeffects e.g. specific tests of renal function ifurinary cadmium levels above the levels ofconcern are detected in biological monitoring.

When health monitoring is done it should rarelybe done in isolation from environmental and/orbiological monitoring. Clearly defined healtheffects should be sought with specific case-definition criteria. Records of other symptomsand clinical findings should also be kept to enableepidemiological assessment of other potentialhealth effects (Langley, 1991a).

Before health monitoring is undertaken, thefollowing issues should be considered:

• How to ensure that all parties involved do nothave unreasonable expectations about theability of health monitoring to resolve issuesof causation or to detect any subtle effect.The studies rarely provide because of theirsize and biases;

• Confidentiality of information;

• How and when information will be madeavailable to participants. The informationmust be released to participants;

• Access to information (by whom and throughwhat mechanisms);

• Interpretation of information (at anindividual and group level);

• Release of findings (which should be at agroup rather than individual level for reasonsof confidentiality if the results are madepublic); and

• How the information will be used to addressthe relevant environmental health issues.

8.15 Exposure Assessment of Volatile Agents

Volatile agents require specialised samplingtechniques to ensure that the contaminants arenot lost during and after sampling so thatanalytical results accurately represent theconcentrations present. The inhalation route willbe more important than for non-volatilecontaminants. It is often impractical to undertakeenvironmental (i.e. air) sampling because of theconstant variations over time of theconcentrations as a result of fluctuation intemperature, wind speed and direction. Otherfactors that will have a significant effect are: soildisturbance; the physico-chemical properties ofthe soil and contaminants; and whether there is arenewable source or whether the contaminationwill dissipate over time. Exposure assessment willoften depend on modelling.

Currently, field monitoring data are the mostappropriate data to use in assessing exposures tovolatile substances. Environmental fate andmodelling characteristics present problems for theuse of short term field monitoring data. This isparticularly marked for the decay of exposures tofinite sources of volatile substances.


121


8.16 Default Values for ExposureAssessments

Defaults used in air, water, and food riskassessments and standard setting are detailed,where available in Appendices 2–4. It ispopulation and site-specific data. Stakeholderconsultation may be useful in establishing site-specific data. Defaults may be useful for ‘backof the envelope’ appraisals to establish whetherthere is a need to move to site-specific appraisals.

The use of a default of 100 per cent bioavailability hasto be employed in the absence of data onbioavailability but such defaults may misrepresent thetrue situation for toxins such as metals. Bioavailabilitymay be significantly different for differing exposureroutes (inhalation, ingestion and dermal exposures)and differing exposure circumstances (e.g. ingestionwhen fasting or with food).

The following are reprinted from ICRP No 23.Report of the Task Group on Reference Man.Copyright (1975), with permission from ElsevierScience.

8.16.1 Body weight, kgAdult male = 70

Adult female = 58

Average = 643

8.16.2 Daily fluid intake (milk, tap water, otherbeverages), ml/day

Normal conditions:

Adults = 1000–2400,representativefigure = 19004

Adult male = 1950

Adult female = 1400

Child (10 years) = 1400

High average temperature (32oC):

Adults = 2840–3410

Moderate activity:

Adults = 3700

8.16.3 Respiratory volumes

8-h respiratory volume, litres

Resting:

Adult male = 3600

Adult female = 2900


Light/non-occupational activity:

Adult male = 9600

Adult female = 9100


Daily inhalation volume, m3

(8-h resting, 16-h light/non-occupational activity)

Adult male = 23

Adult female = 21

Average Adult = 22


Proportion of time spent indoors = 20 h/day

8.16.4 Dietary intakeThere are six principal sources of dietary intakedata for Australia that can be used for exposureassessments:

• National Dietary Survey of Adults: 1983 1.Foods consumed;

• National Dietary Survey of School Children(aged 10–15 years): 1985 1. FoodsConsumed;

• Apparent Consumption of Foodstuffs andNutrients - Australia: 1996–1997;

• Victorian Dietary Survey 1985

• CSIRO State Nutrition Surveys; and

• National Nutrition Survey—Australia 1995.

122

3 WHO uses 60kg for calculation of acceptable daily intakes and water quality guidelines (WHO, 1987, 1993)

4 WHO and NHMRC use a daily per capita drinking-water consumption of 2 litres in calculating water qualityguidelines (WHO, 1993; NHMRC/ARMCANZ, 1996)

8.16.5 Soil exposuresThe following default values have been used inexposure models since 1991 to derive Health-based Soil Investigation Levels. They are adaptedfrom Langley and Sabordo (1996, p184). Thesevalues should be used unless values morepertinent to the relevant population can beprovided and justified. Factors should be relevantto the population about whom the exposureassessment is being done.

1. Dermal absorption factors

• Where available, substance specific data forbioavailability and dermal adherenceshould be used.

• A child’s soil contact area will beequivalent to the area of both hands, bothlegs and both feet (Hawley, 1985). Thisarea of skin will be taken as 0.21m2 (ibid).

• The child will wash once each day.

• The soil adherence factor will be 11 mgper 21.5cm2 (ibid) i.e. a total of 1 074milligrams of soil on the exposed skin.

• Australian washing/bathing values are tobe used where available (See Langley et al,1998).

2. Inhalation factors

• Inspirable soil particulates inside a housewill be 75 per cent of the level ofinspirable particulates outdoors (Hawley,1985). US EPA (1989) found indoorairborne lead levels were 0.3 to 0.8 outdoorlevels for houses without air-conditioning.

• 75 per cent of the inhaled dust will beretained in the respiratory tract and 25 percent will be exhaled (Hawley, 1985).

• Half the inspirable dust will be sufficientlysmall to reach the pulmonary alveoli. Thiswill be the respirable dust fraction and willbe considered to have a diameter of lessthan 10 microns).

• Australian dust values are to be used whereavailable. The data could be used for any

State but local data should be available foreach capital city. It is proposed that areasonable point estimate to use, based onthis data, is 50 micrograms per m3 forrespirable dust. This is conservative butplausible as it is the highest range forwhich data were recorded for the urbanarea and the second highest range for thesuburban area in Adelaide. (Langley andSabordo, 1996).

3. Ingestion factors

• Where bioavailability data for ingested soilcontaminants is unknown, the value of 100 per cent absorption will be used.If bioavailability data are available it can beused providing the values are able tojustified.

• Soil ingestion rates are taken to be:

Age (years) Soil intake (mg/day)

0-1 Negligible

1-5 100*

5-15 50*

Adult 25*

*conservative estimates.(ANZECC/NHMRC, 1992)

• Consistent soil eating behaviour (geophagia)is considered rare although intermittenteating of unusual substances (pica) includingsoil is more common. There should be anawareness of these behaviours and specificbehavioural and environmental managementmeasures may be indicated to reduce theexposures if a particular individual isidentified with these behaviours (Imray andLangley, 1999).

4. Duration of residency

• While the median duration of occupancymay be around 10 years in Australia, aperiod of 70 years should be used forduration of residency to reflect that asignificant number of people will spendprolonged periods in a residence.


123


5. Duration in a workplace

• While the median duration in a workplace isdecreasing in Australia, a period of 30 yearsshould be used for duration in a workplaceto reflect that a significant number of peoplewill spend prolonged periods in a workplace.National Occupational Exposure Standardshave been developed with an undefinedcareer duration.

8.17 Sources of ExposureAssessment Data

Data must be pertinent to the relevantpopulation. Where available, data from Australianpopulations are preferred.

Sources of information and data include:

• Taylor R and Langley A (1998). ExposureScenarios and Exposure Setting.

• Langley AJ and Sabordo L (1996). ExposureFactors in Risk Assessment.

• Langley AJ, Taylor A and Dal Grande E(1998). 1996 Australian Exposure Factors.

• enHealth (in press) The Australian ExposureAssessment Handbook

• The Australian Bureau of Statistics canprovide a range of Australian data.

The American Industrial Health Council’s‘Exposure Factors Sourcebook’ (1994) providesexamples of probability distributions for a rangeof exposure factors. These largely relate to the USpopulation. These, and similar US-based data,should only be used if they can be demonstratedto be relevant to the Australian population.

8.18 Appraising ExposureAssessments

These are modified from US EPA (1992).

Factors that tend to result in underestimates ofexposure:

• Overlooking a significant pathway;

• Failure to evaluate all contaminants ofconcern in the mixture;

• Comparison of exposure-related data againstcontaminated media or exposed populationsrather than against appropriate backgroundlevels;

• Using insufficiently sensitive detection limitsso that meaningful values are reported as notdetected;

• Composite sampling;

• Failure to consider the additive effects ofmultiple pathways;

• Relevant individual pathways within the sameexposure route may not have been summed; and

• Use of multiple lower range point estimates.

Factors which can cause overestimates of exposureinclude:

• The use of unrealistically conservativeexposure parameters;

• Portraying hypothetical potential exposures asexisting exposures;

• Failure to consider route specificbioavailability;

• The use of 100 per cent default bioavailabilityvalues;

• The cumulative effect of using multiple upperrange point estimates (e.g. at the 90 per centor 95 per cent level); and

• Attributing a significant value to results thatfall below an appropriate detection limit.Substituting such values may create theimpression of values where none exist.

Factors that may cause underestimates oroverestimates include:

• Computational errors;

• Inaccurate analytical data;

• Use of inappropriate factors e.g. for intakeroutes;

• Insufficient uncertainty assessment to put theexposure assessment in perspective;

124

• Use of an inappropriate number of significantfigures for the numeric estimates in asituation where using more than onesignificant figure may imply more confidencein the results than is warranted;

• The unthinking and uncritical use of models.While the concept of ‘garbage in, garbageout’ is well accepted, some risk assessmentmodels result in ‘quality in, garbage out’ (seeCalabrese and Kostecki, 1992);

• The failure to take into account correlationsamong input distributions when usingsimulations such as Monte Carlo. It will beunnecessary to use Monte Carlo simulation ifthe relationship between variables is known;and

• Inappropriate monitoring or sampling (e.g.grab sampling vs static or period sampling).

8.19 Exposure AssessmentReports

The following checklist details matters that shouldbe appropriately addressed in an exposureassessment. With justification, particular materialmay be omitted. It is adapted from US EPA (1995):

1. What are the most significant sources ofenvironmental exposure?

• Are there data on sources of exposure fromdifferent media? What is the relativecontribution of different sources ofexposure?

• What are the most-significantenvironmental pathways for exposure?

2. Describe the populations that were assessed,including the general population, highlyexposed groups, and highly susceptiblegroups.

3. Describe the basis for the exposureassessment, including any monitoring,modelling, or other analyses of exposuredistributions such as Monte-Carlo orkrieging.

4. What are the key descriptors of exposure?

• Describe the range of exposures to groupssuch as: ‘average’ individuals, ‘high end’individuals, general population, highexposure group(s), children, susceptiblepopulations.

• How was the central tendency estimatedeveloped? What factors and/or methodswere used in developing this estimate?

• How was the high-end estimatedeveloped?

• Is there information on highly-exposedsub-groups? Who are they? What are theirlevels of exposure? How are theyaccounted for in the assessment?

5. Is there reason to be concerned aboutcumulative or multiple exposures because ofethnic, racial, or socioeconomic reasons?

6. Summarise exposure conclusions and discussthe following:

• results of different approaches, i.e.modelling, monitoring, probabilitydistributions;

• limitations of each, and the range-of mostreasonable values; and

• confidence in the results obtained, and thelimitations to the results.


125



Risk Management




Review and

reality check

Review and

reality check


126


9

9.1 IntroductionRisk characterisation is the final step in the riskassessment process that:

• integrates the information from hazardassessment and exposure assessment;

• provides an evaluation of the overall qualityof the assessment and the degree ofconfidence the risk assessors have in theestimates of risk and conclusions drawn;

• describes the risks to individuals andpopulations in terms of nature, extent andseverity of potential adverse health effects;

• communicates results of the risk assessment tothe risk manager (US EPA, 1995, p. 4); and

• provides key information for riskcommunication.

The final risk characterisation is rarely accuratelyquantitative because of the limitations of the dataand this will be reflected in the uncertaintyassessment. The process requires considerableexpertise. If data are collected and analysedaccording to the principles and guidelines in thisdocument the process will become moretransparent and consistent. Some parts of the riskassessment process such as ‘data collection’ and‘exposure assessment’ will be, at least in part,quantitative. These guidelines are intended toassist the qualitative process of determiningwhether environmental health intervention isrequired or not required. Due to the complexitiesof the matter, the risk characterisation processcannot be reduced to a ‘cookbook’.

Risk characterisation may involve comparingenvironmental data, exposure data, intakes, andbiological monitoring results with establishedcriteria.

9.2 Key Principles inEnvironmental Health RiskCharacterisation

There are a number of key principles for healthrisk characterisation (adapted from EPA NSW,1998; US EPA, 1995):

1. Actions should always adequately protectpublic health and the environment, puttingthese responsibilities before all otherconsiderations.

2. Risk assessments should be transparent. Thenature and use of default values and methods,assumptions and policy judgements in therisk assessment should be clearly identified.Conclusions drawn from the evidence shouldbe separated from policy judgements.

3. Risk characterisations should include asummary of the key issues and conclusions ofeach of the other components of the riskassessment, as well as describing the natureand likelihood of adverse health effects. Thesummary should include a description of theoverall strengths and limitations (includinguncertainties) of the assessment andconclusions.

4. Risk characterisations (and risk assessments)should be consistent in general format, butrecognise the unique characteristics of eachspecific situation.

5. Health risk assessment must be undertakenwith an appreciation that the health riskassessment is part of a larger assessment thatencompasses ecological risk assessment.

6. To protect public health and the environmentan appropriate degree of conservatism mustbe adopted to guard against uncertainties.

7. Ensure that comparisons have been madeagainst environmental health criteria thathave been endorsed by the relevantCommonwealth, State or Territoryenvironmental health agencies.

8. Where there are no Environmental HealthCriteria for a particular agent refer to theadministrative authority at the relevantCommonwealth, State or Territory level.

9. Ensure that human health risk assessmentsare undertaken, where necessary, according tomethods in this document, or its revisions aspublished from time to time.


127


10. When deriving environmental health criteriause toxicological data or exposure criteriafrom agencies or organisations relevant to theState or Territory (e.g. local orCommonwealth health agencies such asNHMRC, or the enHealth Council) or towhich Australia is party (e.g. World HealthOrganization). (See Section 4 HazardAssessment—Hazard Identification—Toxicology).

11. Ensure that human health risk assessments areundertaken using national toxicologicalassessments (e.g. NHMRC) or WHOassessments or, where neither has been made,methods agreed to by the administrativeauthority for contaminated sites at the relevantCommonwealth, State or Territory level.

12. The risk assessor’s knowledge of the peer-reviewed scientific literature relevant to riskassessment should be up-to-date.

13. Variations in risk assessments as a result ofparticular statutory requirements, resourcelimitations, and other specific factors shouldbe explained as part of the riskcharacterisation. For example, a reason will berequired to explain why certain elements areincomplete.

9.3 Quantitative andQualitative RiskCharacterisation

The level of risk can be described eitherqualitatively (i.e. by putting risks into categoriessuch as ‘high’, ‘medium’ or ‘low’) or quantitatively(with a numerical estimate). Current riskassessment methods do not enable accuratequantitative estimates of risk for low levels ofexposure to environmental hazards. Numericalestimates of risk will rarely be feasible because ofvariability in the agent and population andlimitations in toxicological and exposure datawhich will be reflected in the uncertaintyassessment, but a degree of quantification may bepossible for some components such as datacollection and exposure assessment.

Estimates do not have to depend on the use ofnumbers to be useful; ordinary language may beused to indicate the level of risk. A finely dividedranking system can give a relatively accurateindication of quantity without using numbers(ACDP, 1996). Clearly defined qualitativecategories can enable reliable and effective riskmanagement decisions.

Tolerable Intakes are a form of quantitative riskcharacterisation as they are an estimate of theintake of a substance that over a lifetime iswithout appreciable health risk. (WHO, 1994).

Numbers may give a misleading implication ofaccuracy, especially when based on poor oruncertain information. The generation of a precisevalue in QRA should not be mistaken foraccuracy (IEH, 1999b). The problems arecompounded where results are interpolated overseveral orders of magnitude and whereinformation on the mechanisms of tumourinduction is limited.

The most conservative mathematical models usedin QRA are virtually insensitive to the actualexperimental data and should be viewed only as arisk management solution, not a risk assessmenttechnique (IEH, 1999b, p. 34).

A detailed critique of quantitative cancer riskassessment is provided in Hrudey (1998).

While qualitative risk conclusions can avoid thefalse sense that the extent of the risk is knownprecisely, the use of terms such as ‘high’, ‘medium’or ‘low’ may have different interpretations todifferent groups and they should be clearlydefined. This is often best achieved by being putin context or compared to other risks relevant tothe community. If comparisons do not directlyrelate to alternative options, they should be usedcautiously, especially if like is not compared tolike or if comparisons are being used to implyacceptability. Flippant comparisons arecounterproductive (DOH, 1998). Many riskcomparisons are inappropriate because of a weakevidentiary base supporting the estimate orbecause they are perceived by the community asirrelevant (e.g. a recent risk assessment used acomparison with ‘death in the Balkan War’).

128

Comparisons should be used only where theevidentiary base and the method for riskestimation are similar and where the uncertaintiesin all the comparative estimates are shown(Thomas and Hrudey, 1997, p. 218–219).

It is important to consider contingent risks. Thisrequires not looking at risks in isolation so that,for example, the risks of immunisation (orchlorination) are considered in the context of therisks of not having immunisation (orchlorination).

9.4 Risk ConclusionsThe following is adapted from US EPA 1995.

Risk conclusions

What is the overall picture of risk, based on datacollection, exposure assessment, toxicityassessment and risk characterisation?

What are the major conclusions and strengths ofthe assessment in each of the three main areas(i.e. hazard identification, dose–response, andexposure assessment)?

What are the major limitations and uncertaintiesin the three main areas?

What are the science policy issues in each of thethree major areas?

What alternative risk assessment approaches wereevaluated?

What is the basis for the selection of options?

Risk context

What are the qualitative characteristics of thehazard (e.g. voluntary vs. involuntary,technological vs. natural, etc.)? Comment on thefindings, if any, from studies of risk perceptionthat relate to this hazard or similar hazards.

What are the alternatives to this hazard (e.g. arethere other water treatment processes or foodadditives available)? How do the risks compare?

How does this risk compare to other risks?

How does this risk compare to other similar risksthat the regulatory agency has made decisionsabout?

Where appropriate, can this risk be comparedwith past regulatory agency decisions or commonrisks with which people may be familiar?

Describe the limitations of making thesecomparisons.

Comment on significant relevant communityconcerns that will influence public perception ofrisk for the hazards addressed in the riskassessment.

Existing risk assessments

Comment on other risk assessments that havebeen done on this agent by Commonwealth, Stateor Territory agencies, or other organisations. Arethere significantly different conclusions that meritdiscussion?

Other risk assessments

Comment on risk estimates generated by differentstakeholders (P/CCRARM, 1997).

Other information

Is there other information that would be useful tothe risk manager, or the public in this situationthat has not been described above?

9.5 UncertaintyUncertainty is always present and this reinforcesthe need for a systematic and rigorous approachthat most accurately portrays the level of actualrisk. Uncertainty is a key reason why riskassessment is being performed. Uncertaintyanalysis must be addressed for each step of therisk assessment and for its cumulative effect fromall of the steps.

The assessment of uncertainty is a critical part ofthe risk assessment process. Uncertaintycharacterisation is an essentially qualitativeprocess relating to the selection and rejection ofspecific data, estimates, scenarios, etc (US EPA,1992). Uncertainty assessment can be morequantitative and it may be represented by moresimple measures such as ranges, simple analytical


129


methods such as sensitivity analysis and mayprogress to complex measures and techniques(Langley, 1993).

Uncertainty (i.e. the lack of knowledge about thecorrect value, for example a specific exposuremeasure or estimate) must be distinguished fromvariability (i.e. different levels of exposureexperienced by different individuals).

There are three broad types of uncertainty whenestimating risks (US EPA,1992):

1. Uncertainty arising from missing orincomplete information (scenario uncertainty)e.g. descriptive errors, aggregation errors,errors in professional judgement, andincomplete analysis.

2. Uncertainty affecting a particular parameter(parameter uncertainty) e.g. measurementerrors, sampling errors, variability, and use ofgeneric or surrogate data. If expertjudgements are used they should beincorporated in a ‘consistent, welldocumented manner’.

3. Uncertainties in the scientific theory affectingthe ability of a model to make predictions(model uncertainty).

Uncertainty may need to be addressed by thecollection of further data and uncertainty analysiscan be particularly useful for identifying researchthat will be of value.

130

Table 16: Example of an uncertainty table for exposure assessment

Effect on exposurea

Assumption Potential magnitude Potential magnitude Potential magnitudefor over-estimation for under-estimation for over- or under-of exposure of exposure estimation of

exposure

Environmental sampling and analysis

Sufficient samples may not have been taken Moderateto characterise the media being evaluated,especially with respect to currently available soil data.

Systematic or random errors in the chemical Low–Highanalyses may yield erroneous data.

Exposure parameter estimation

The standard assumptions regarding body Moderateweight, period exposed, life expectancy,population characteristics, and lifestyle may not be representative of any actual exposure situation.

The amount of media intake is assumed to Moderate be constant and representative of the exposed population.

Assumption of daily lifetime exposure Moderate to highfor residents.

a As a general guideline, assumptions marked as ‘low’, may affect estimates of exposure by less than one order ofmagnitude; assumptions marked ‘moderate’ may affect estimates of exposure by between one and two orders ofmagnitude; and assumptions marked ‘high’ may affect estimates of exposure by more than two orders of magnitude.

(adapted from US EPA, 1989a, p. 6–51)

Reasons for addressing uncertainties inassessments include:

• the combination of uncertain informationfrom various sources;

• Having to make decisions about whetherfurther resources should be used on seekingfurther information and data to reduceuncertainty;

• as a means of highlighting biases that mayhave crept into the process;

• as assessment is an iterative process,uncertainty analysis may enhance theoutcome of the process by highlighting areaswarranting further work or consideration;

• risk assessment may be one of severalprocesses involved in a particular situation.Being able to characterise the uncertainty willassist the decision-makers and ultimatelyimprove the decision making (amended fromUS EPA, 1992); and

• uncertainty assessment assists risk assessors tomeet their responsibility to present not justnumbers but also a clear and explicitexplanation of the implications andlimitations of their analyses.

In summarising the output from the uncertaintyassessment, the important implications for riskmanagement need to be highlighted. The riskassessor and the risk manager need to worktogether (or, at least, understand each other’sneeds and limitations). The risk assessor shouldemphasise the following aspects of the uncertaintyassessment results:

• the implications for relying on any pointestimate that might have been producedwithout consideration of uncertainty;

• the shape and breadth of the uncertaintydistribution which will provide informationabout how prudent various risk estimatesmight be;

• their insights regarding the balance betweenthe health costs of overestimating andunderestimating risk;

• the sensitivity of the uncertainty estimates tofundamentally unresolved scientificcontroversies; and

• the implications for research and further datagathering, identifying which uncertainties aremost important; which uncertainties areamenable to reduction by directed researchefforts; and an estimate of the effort thatwould be required to significantly reduceuncertainty.

(adapted from NRC, 1994, p. 168; Finkel, 1990.)

Uncertainty issues to be addressed in eachrisk assessment step

1. Hazard identification: What are theuncertainties about the capacity of theenvironmental agent(s) for causing adverseeffects in laboratory animals and in humansconcerning:

• the nature, reliability, and consistency ofthe particular studies in humans and inlaboratory animals;

• the available information on themechanistic basis for activity; and

• experimental animal responses and theirrelevance to human outcomes.

2. Dose–response assessment: What are theuncertainties about the biological mechanismsand dose–response relationships underlyingany effects observed in the laboratory orepidemiology studies providing data from theassessment relating to:

• the relationship between extrapolationmodels selected and available informationon biological mechanisms;

• how appropriate data sets were selectedfrom those that show the range of possiblepotencies both in laboratory animals andhumans;

• the basis for selecting interspecies dosescaling factors to account for scaling dosefrom experimental animals and humans;and


131


• the correspondence between the expectedroute(s) of exposure and the exposureroute(s) utilised in the hazard studies, aswell as the interrelationships of potentialeffects from different exposure routes.

3. Exposure assessment: What are theuncertainties related to the paths, patterns,and magnitudes of human exposure andnumber of persons likely to be exposed?

• The basis for the values and inputparameters used in each exposure scenario.If based on data, information on thequality, purpose, and representativeness ofthe database is needed. If based onassumptions, the source and general logicused to develop the assumption (e.g.monitoring, modelling, analogy,professional judgement) should bedescribed.

• The major factor or factors (e.g.concentration, body uptake,duration/frequency of exposure) thought toaccount for the greatest uncertainty in theexposure estimate, due either to sensitivityor lack of data.

• The link of the exposure information tothe at-risk population including importantsubgroups of the population such as highlyexposed or highly susceptible groups orindividuals (and the reasons they arehighly exposed or highly susceptible, ifknown). This component includes theconservatism or non-conservatism of theexposure scenarios. In addition,information that addresses the impact ofpossible low probability by possibly highconsequence events may need to beaddressed.

4. Risk characterisation: Detail what otherassessors, decision-makers, and the publicneed to know about the primary conclusionsand assumptions, and about the balancebetween confidence and uncertainty in theassessment? What are the strengths andlimitations of the assessment?

• Numerical estimates, where practicable,should be included with the descriptiveinformation that is integral to the riskassessment. For decision-makers, a completecharacterisation (key descriptive elementsalong with numerical estimates) should beretained in all material relating to anassessment used in decision-making.Differences in assumptions and uncertainties,coupled with non-scientific considerationscalled for in various environmental statutes,can clearly lead to different risk managementdecisions in cases with ostensibly similarrisks, i.e. the level of risk alone does notdetermined the decisions.

• Consideration of alternative approachesinvolves examining selected plausibleoptions for addressing a given uncertainty.The description of the option chosenshould include the rationale for the choice,the effect of option selected on theassessment, a comparison with otherplausible options, and the potentialimpacts of new research.(adapted from USEPA, 1992a)

9.6 Exposure Durations andExceedances of AcceptableDaily Intakes (ADIs)

Appropriate durations of exposure need to beassessed so that transient (short term) andimportant exposures are not obscured by the use,for example, of average lifetime exposures. This isimportant in the Australian context whereAcceptable Daily Intake values from WHO areoften used. The duration and magnitude ofexceedances of the ADIs must be obvious inexposure assessments.

WHO publications provide comment onexceedances of ADIs:

‘Because in most cases, data are extrapolatedfrom life-time animal studies, the ADI relatesto life-time use and provides a margin ofsafety large enough for toxicologists not to beparticularly concerned about short-term useat exposure levels exceeding the ADI,

132

providing the average intake over longerperiods does not exceed it’ (WHO, 1987).

and:

‘It is impossible to make generalisationsconcerning the length of time during whichintakes in excess of the PTWI (provisionaltolerable weekly intake) would betoxicologically detrimental.

Any detrimental effect would depend upon thenature of the toxicity and the biological half-life of the chemical concerned’ (WHO, 1989).

The following discussion on the significance ofexceeding the ADI applies equally to otherrecommended limits of intake or exposure, suchas TDI or PTWI.

Three questions should be considered if there arepotential exceedances of the ADI:

• What proportion of the population should beallowed to exceed the ADI?

• To what extent can the ADI be exceededwithout any real concern?

• How long does the person need to exceed theADI before there is a cause for real concern?’

The significance of any minor excursions ofintake above the ADI can only be put intocontext by reference back to the animal data andto the NOEL which gave rise to the ADI(Renwick and Walker, 1993, p. 464).

Renwick and Walker describe three parametersgoverning the precision of the NOEL:

• the sensitivity of the toxicological end pointwhich depends on the incidence of the lesionin control animals and/or its inter-animalvariability;

• the group size studied which tends to be lessimportant than;

• the increment between doses. There may beconsiderable increments between doses and thiscan result in a NOAL that can be significantlylower than the actual or absolute NOEL.

Commonly, a NOEL will be chosen from adataset which contains a number of repeat-dosetoxicity studies, and this is likely to increase theprecision of the NOEL. For a discussion on issuesin the selection of the most appropriate overallNOEL/NOAL see Section 11.2.

Renwick and Walker (1993) conclude that thesignificance of the exceedance must be assessedon a substance-specific basis and by reference tothe toxicological (and especially the NOEL) data.

9.6.1 Appraisal of short termexposures

Consideration will need to be given as to whetherexcursions above the Tolerable Intake provide anacute risk.

Tolerable intakes refer to long-term, usuallylifetime, exposures. There may be a need to assessrisks from short-term exposures. To assist in suchappraisals, acute Reference Doses are beingdeveloped (WHO, 1997). These are an estimateof the amount of a substance in food or drinkingwater, expressed on a body-weight basis, that canbe ingested over a short period of time, usuallyduring one meal or one day, without appreciablehealth risks to the consumer on the basis of allknown facts at the time of the evaluation (WHO,1997). In the first instance acute Reference Dosesare being set for pesticides.

The most relevant endpoint to establish an acutereference dose depends on identifying the bestrelevant study on a case-by-case basis.

The following types of studies will need to beassessed when establishing an acute referencedose:

• acute oral toxicity (LD50) studies;

• short term studies of toxicity;

• developmental toxicity studies;

• reproductive toxicity studies;

• human data; and

• mechanistic studies (WHO, 1997).


133

Appraisal of Assessments

10

10.1 IntroductionAccurate and timely site-specific health riskassessments depend upon coherent and logicallydeveloped reports. The Commonwealth, Statesand Territories and Local Government shouldrequire standardised formats modelled on materialin Section 8.5 - ‘Environmental Sampling andAnalysis’ part of this document. Section 8.7 canbe used as a checklist to be completed by thehealth risk assessor.

Reports should be:

• timely;

• comprehensive in their appraisal of allrelevant data; and

• clear in their content and conclusions.

Reports do not create confidence in their contentand are likely to experience rejection or delays inappraisal by regulators if they:

• obfuscate;

• do not meet appropriate levels of coherenceor logic; or

• do not meet the requirements of thestandardised formats or the checklist ofhealth risk assessment contents.

The general attributes of a good report are:

• the scope and objectives of the report areexplicitly stated;

• the report’s content is laid out impartially,with a balanced treatment of the evidencebearing on the conclusions;

• the risk assessment presentation includes adescription of any review process that wasemployed, acknowledging specific reviewcommentary;

• the key findings of the report are highlightedin a concise executive summary;

• the report explains clearly how and why itsfindings differ from other risk assessmentreports on the same topic; and

• the report explicitly and fairly conveysscientific uncertainty, including a discussionof research that might clarify the degree ofuncertainty.

As the risk assessment process is intended toassist risk managers in decision-making, a key testwill be whether the risk assessment reportachieves that aim.

(AIHC, 1989)

10.2 General AppraisalA person reviewing or authoring an assessmentwill consider questions such as:

10.2.1 Key aspectsHave the objectives of the report been definedclearly?

• Is there a clear understanding of the relevantcurrent or future human activities andwhether any constraints (e.g. encumbrances,exclusions) will be acceptable?

• Was the environmental sampling reasonablysufficient to identify, to locate, to demarcateand to characterise any potential hazardousagents?

• Is it clear how results of any environmentalsampling plan were analysed and interpreted?

• Have data been analysed en masse or for theappropriate environmental strata?

• Were environmental fate and transportmechanisms understood?

• Have the data been ‘modelled’ to demonstratea three-dimensional understanding of what isoccurring?

• How were abnormal results or findingsmanaged?

• Were the uncertainties of the assessmentidentified and understood?

(adapted from Langley, 1993a, p. 28)


135


10.2.2 Interpretation of sampling data

An appraisal of data must show an understanding of:

• the context of the risk assessment;

• the topography of the situation;

• the demography of the population;

• environmental factors such as stratification ofwater bodies, movement of plumes in air orgroundwater, soil structure (e.g. presence ofclay or fill, and the depths of individualstrata), meteorological factors, groundwaterflows; and

• the relevant current or future humanactivities.

Too often numerical data are considered inisolation from other key parameters such as:

• the levels of detection (and reporting);

• quality assurance for the data;

• the uncertainty about the data;

• the geographical relationship of one sampleto another; and

• the current or potential human activities.

Other key failings in the analysis of numericaldata include:

• Ignoring negative or unexceptional results byfocussing on unusual or elevated results: thedata set needs to be considered in its entirety;

• Inadequately managing censored data e.g. byassigning a zero value to results below thelevel of detection or reporting; and

• Accepting relatively high levels of detection orreporting so that the value of much data isobscured. This may have the consequence offailing to reveal gradients that will help tohighlight the presence and location ofenvironmental ‘hot spots’. Examples have beenseen where environmental health criteria levelshave been treated as the level of reporting.The very existence of levels of detection andreporting results in the censoring of data.

Censoring of data can be particularlyimportant when the maximum permittedcriterion is close to the level of detection (e.g.with potable drinking water standards). Thecensoring of data must be addressed in anappropriate way (See Section 8.5.14).

Given two similar results, the result that can beexplained (e.g. by history, or similarities withresults from similar strata) will tend to be of lessconcern than the result that cannot be explained(Langley, 1993a).

10.2.3 Use of subjective termsThe use of subjective terms in reports (e.g.‘heavy/medium/light contamination’) or termsthat are used in common parlance but may havelegalistic definitions (e.g. ‘contamination’) areconfusing and should be avoided in reports. Theuse of the term ‘hot spot’ can result in misleadingperceptions of concentrations and the term shouldbe used prudently.

10.3 Specific Appraisal The following is a checklist adapted from EPANSW (1998) and Vic EPA (1997a) and should beaddressed in human health risk assessments.

10.3.1 Data collection• Have the objectives of the risk assessment

been stated?

• Has the background to the events leading tothe risk assessment been provided?

• Have all agent of potential concern beenidentified and appraised?

• Have all appropriate sources of informationregarding chemicals of potential concern beenidentified and appraised?

• Has justification been given for the selectionof the agents of potential concern? Hasjustification been given for the omission ofagents from the analysis?

• Have the sources of the agents beenidentified?

• Have the environmental fate and transport ofthe agents been identified?

136

10.3.2 Hazard identification anddose–response relationship

The general components in an acceptable riskassessment are:

Hazard identification

• All relevant information is presented andreviewed.

• The report highlights critical aspects of dataquality.

• A weight-of-the-evidence approach ispresented for judgement as to the likelihoodof human carcinogenic hazard and includes aclear articulation of the rationale for theposition taken.

• The report identifies research that wouldpermit a more confident statement abouthuman hazard.

Dose–response relationship

• Valid data sets and plausible models for high-to-low dose and interspecies extrapolation arepresented in dose–response modelling.

• The report offers an explicit rationale for anypreferred data set(s) and model(s) used indose–response evaluation; strengths andweaknesses of the preferred data sets arediscussed, and scientific consensus or lackthereof is indicated for critical issues orassumptions.

• The report reveals how dose–responserelationships change with alternate data sets,assumptions, and models.

(AIHC, 1989)

Specific considerations are:

• Have all relevant toxicological facts beenchecked for accuracy and currency?

• Has the adequacy of the availabletoxicological database been appraised?

• Have the effects on each significant bodysystem (for example, renal, hepatic,cardiovascular,) and the types of effects

(for example, allergy, genotoxicity andcarcinogenicity, reproductive anddevelopmental) been appraised andsummarised for the relevant exposure routes?

• Has the critical toxic effect(s) and organ/bodysystem been identified?

• Have known toxicity modifying factors (suchas synergistic and antagonistic effectsresulting from exposure to multiplecontaminants) been considered?

• Have toxicologically sensitive sub-populationsbeen identified?

• Has the toxicological basis of the guidancevalue or potency factor, where applicable,been discussed and the uncertainties noted?

• Have NHMRC (where applicable) or WHOtoxicological assessments been considered asthe primary toxicological resource?

• Where relevant, have differences between, forexample, WHO and US EPA toxicologicalassessments been appraised and discussed?

• Has the dose–response relationship for agentsof potential concern been appraised anddiscussed?

• Have the data been presented in a formamenable to efficient interpretation andreview?

10.3.3 Exposure assessmentThe general components in an acceptable riskassessment are:

• the purpose and scope of the exposureassessment and the underlying methodologiesare clearly described;

• the specific populations and sub-populationsthat are the subjects of the assessment areclearly identified, and the reasons for theirselections and any exclusions are given;

• available data are considered and criticallyevaluated, and the degree of confidence in thedata expressed. (Reasons for any dataexclusion are presented.);


137


• if models are used, their bases are described,along with their validation status;

• potential sources, pathways, and routes ofhuman exposure are identified and quantified;the reasons why any are not included in theassessment are presented;

• central estimates and, if possible, upper andlower bounds on exposures for the fullpopulation, and the distribution of exposuresare described; any preferred estimates arenoted, together with supportingdocumentation;

• uncertainties in the estimates are described,and the relative importance of keyassumptions and data is highlighted;

• research or data necessary to improve theexposure assessment are described;

• the purpose and scope of the exposureassessment and the underlying methodologiesare clearly described;

• the specific populations and sub-populationsthat are the subjects of the assessment areclearly identified, and the reasons for theirselections and any exclusions are given;

• available data are considered and criticallyevaluated, and the degree of confidence in thedata expressed. (Reasons for any dataexclusion are presented.);

• if models are used, their bases are described,along with their validation status;

• potential sources, pathways, and routes ofhuman exposure are identified and quantified;the reasons why any are not included in theassessment are presented;

• central estimates and upper and lower boundson exposures or, if possible, the fullpopulation, distribution of exposures aredescribed; any preferred estimates are noted,together with supporting documentation;

• uncertainties in the estimates are described,and the relative importance of keyassumptions and data is highlighted; and

• research or data necessary to improve theexposure assessment are described.(AIHC, 1989)

Specific considerations are:

• Has the potentially exposed population beenidentified?

• Have potentially exposed, unusuallysusceptible sub-populations been identified?

• Have the estimates of chemical exposure foreach significant exposure route and for eachchemical of potential concern beenadequately quantified and tabulated?

• In cases of presumed insignificant exposure, hasthe exposure been demonstrated to be small?

• Has the relative significance of each exposurepathway, based on the risk analysis, beendiscussed?

10.3.4 Risk characterisationThe general components in an acceptable riskassessment are:

• the major components of risk (hazardidentification, dose–response, and exposureassessment) are presented in summarystatements, along with quantitative estimatesof risk, to give a combined and integratedview of the evidence;

• the report clearly identifies key assumptions,their rationale, and the extent of scientificconsensus; the uncertainties thus accepted;and the effect of reasonable alternativeassumptions on conclusions and estimates;

• the report outlines specific ongoing orpotential research projects that wouldprobably clarify significantly the extent ofuncertainty in the risk of estimation;

• the report provides a sense of perspectiveabout the risk through the use of appropriateanalogy;

• the major components of risk (hazardidentification, dose–response, and exposureassessment) are presented in summarystatements, along with quantitative estimates

138

of risk, to give a combined and integratedview of the evidence;

• the report clearly identifies key assumptions,their rationale, and the extent of scientificconsensus; the uncertainties thus accepted;and the effect of reasonable alternativeassumptions on conclusions and estimates;

• the report outlines specific ongoing orpotential research projects that wouldprobably clarify significantly the extent ofuncertainty in the risk of estimation; and

• the report provides a sense of perspective aboutthe risk through the use of appropriate analogy.

(AIHC, 1989)

10.3.5 Equations• Have all equations used in the risk assessment

been presented in the report?

• Are all equations consistent?

• Have all parameters in each equation beenclearly defined?

• Have the correct units been allocated to eachparameter?

• Are all equations dimensionally correct?

• Have all unit conversion factors, whereapplicable, been included in the equations?

• Has all pertinent information been providedto enable calculations to be checked throughin a stepwise process?

10.3.6 Data evaluation• What were the data collection objectives and

are they consistent with the requirements ofthe risk assessment?

• Have the laboratories that did the chemicalanalyses been noted, and do they haveNATA, or equivalent, accreditation toperform the chemical analyses?

• Has laboratory QA/QC been reported andanalysed?

• Has field QA/QC been reported andanalysed?

• Where appropriate, has the size of a ‘hot spot’detectable by the sampling pattern been stated?

• Have statements of the accuracy of thelaboratory data for each agent been made?

10.3.7 Assessment and reportpresentation

• Have all tables and figures been referred tocorrectly in the text of the report?

• Has information from previous reports on thesituation been appropriately selected andincorporated into this report?

• Has irrelevant information from othersituations been excluded from the report?

• Have all assumptions and default data beenidentified and justified?

• Has the analysis been based on an up-to-dateliterature appraisal?

• Have all conclusions been justified?

• If toxicological data and the exposure scenariolead to the conclusion that a highconcentration of agent is permissible, doesthe result violate ecological, aesthetic, land-use or physical principles?

• Has a risk management decision(s) beenmade during the course of the risk assessmentand, if so, how might it (they) haveinfluenced the calculation of risk?

• Has a detailed uncertainty discussion beenincluded in the report?

• Has information been presented coherentlyand in an appropriate sequence, to enableefficient appraisal of the report?

• Does the report include or enable ecologicalrisk assessment as required by regulatoryauthorities?

• What has been the involvement of the public?

• How has information been communicated tothe public?

• What processes of community consultationhave taken place?


139

Setting Environmental Health Criteria

11

11.1 Principles for SettingCriteria

Much of what is talked about in the context ofrisk assessment, i.e. setting of criteria and ADIsand such, amounts to decision-making whichshould more accurately be considered riskmanagement if we are to subscribe to our goal ofkeeping the risk assessors focused on the scientificevidence rather than on making the tough choicesinherent in risk management. There is not a riskassessment model that can set environmentalhealth standards; one should only seek to haveone that can inform risk management decisionsabout criteria.

The principles in this chapter detail the processesby which criteria can be established. Genericenvironmental health criteria developed using thismethodology require endorsement by appropriatenational health bodies. Situation-specificenvironmental health criteria developed using thismethodology require endorsement by theappropriate health agency before being applied toa particular situation.

Elements of the risk assessment methodologyprovide a framework for setting risk-basedenvironmental health criteria. A series of riskassessments using a range of assumptions willprovide the sensitivity analysis for the criterion-setting process i.e. by varying the assumptionsabout dose–response and exposure the effects onpopulation risk from different possible criteria canbe assessed. ‘Hazard Identification’ establishes thekey hazards of concern. ‘Dose–Response’information (from the scientific literature) and‘Exposure Assessment’ (using a range of possiblevalues that will reflect the possible values of thecriterion) provide the basis for the ‘RiskCharacterisation’ which explains the nature andmagnitude of the impact from the key hazards ofconcern at specified exposure levels on thepopulation to whom the criterion will be relevant.‘Risk Characterisation’ will also detail theuncertainties and assumptions underlying thecriterion.

When establishing criteria key issues to beconsidered are:

• Why is a criterion being proposed?

• Is a criterion necessary? Are there alternativemeans of achieving the desired outcome? Thelarge improvements achieved by the CleanAir Act (1956) in the UK occurred withoutany air quality standards;

• How will the criterion be used? Is thecriterion to be used as a guideline or astandard? Standards often have greater legalor regulatory standing than guidelines;

• Is the criterion to be generic (applying tomany situations) or situation-specific?

• Who will be involved in setting the criterion?

• What population(s) will be affected?

• Are there any sensitive or particularlysusceptible sub-populations who are exposed?

• Over what period of time will the populationbe exposed to the agent for which thecriterion is being set?

• What patterns of exposure are likely to occur?Are there likely to be short or long termfluctuations?

• Are background exposures higher than theTolerable Intake? (Given the size of the safetyactors used in the development of TolerableIntakes, has this had any health consequencesfor the population? Should actual experiencebe used in preference to Tolerable Intakes?)

• Are there difficulties in getting relevant andaccurate background exposure data?

• How do you deal with Tolerable Intakes solow that they can’t be measured (e.g. for somealkaloids)?

• What are the consequences of setting criteriaat the level of detection?

• How can Tolerable Intakes, which are usuallybased on ingestion, be applied to otherexposure routes?

• Has the Tolerable Intake been set usinggavage or bolus administrations rather than


141


administration as part of the diet? Has thesubstance been administered with an oilcarrier or food?

• How is exposure occurring?

• How do you deal with multiple pathways ofexposure such as the relatively high exposuresin tobacco smoke compared to dietarysources?

• How do you apportion exposures? In Canada,for criteria setting, 20 per cent of totalexposures is allowed of each of food, air,water, soil and consumer products;

• Can exposures be altered? How?

• What is the critical health effect? What is itsnature, severity and reversibility?

• Are interactions with other agents relevant?

• What are the background levels of exposureto the agent?

• Are there sufficient data to establish acriterion?

A decision tree detailing the use of Health RiskAssessment to develop risk-based environmentalhealth criteria is provided in Figure 9. This modeluses a Guidance Value (e.g. ADI) which isapportioned between background exposures andexposures relevant to a particular exposure pathway(e.g. food, water, air or soil). This approach is basedon chronic or subacute exposures. It may not beapplicable for acute exposures e.g. when dealingwith a respiratory irritant.

A slightly different approach will be requiredwhere acute exposures may cause the hazard ofconcern to become manifest. Examples are:setting microbiological standards for water andfood where acute exposures precipitate disease;the setting of criteria for sulfur dioxide in airwhere acute exposures may cause exacerbation ofasthma; and, the setting soil values for nickel orchromium (VI) both of which may cause allergicreactions from acute exposures in sensitised

individuals. It is particularly relevant where thesusceptible sub-population comprises a significantproportion of the total population; for example,approximately 20 per cent of Australian childrenhave asthma and 10 per cent of women areallergic to nickel. In these situations GuidanceValues (which, by definition, are based aroundlong term exposures) will be irrelevant andunsuitable for use. However the remainder of therisk assessment process will still be relevant forestablishing criteria.

For most criteria there is a significant margin ofsafety between the criterion and typical exposures.Safety is enhanced by a further margin of safetyarising from the process by which GuidanceValues are set. However for some agents, healtheffects may arise from background exposures (e.g.exacerbation of asthma from urban ozoneexposures) and in these instances a riskmanagement decision will need to be taken of theincidence of adverse health effects in thecommunity.

For many agents, there may be several exposurepathways. Copper will be found in water and food(and, of lesser importance, in food and consumerproducts) so that setting a criterion for copper inanother medium (e.g. soil) or a particularfoodstuff (e.g. shellfish) will need to take intoaccount the range of other potential exposurepathways. Intakes may need to be apportionedbetween the different exposure pathways. Theapportioning of intakes raises other questions:

• What percentage of the total Tolerable Intakeshould be used for establishing a set criterion?Inter-agency cooperation will be required toenable appropriate apportionment.

• What is the nature of the backgroundexposures? Are they fixed or changing overtime? Are they able to be altered? Are theyvoluntary (e.g. smoking) or involuntary (e.g.ambient air pollution)? To what degreeshould voluntary background exposures betaken into account?

142

11.2 Determination ofNO(A)ELs,ADIs (RfD) andTDIs for Humans

The determination of an acceptable daily intake(ADI)5 involves the establishment of an overallNOEL/NOAEL for a chemical which isgenerally the lowest NOEL/NOAEL in the mostsensitive species. This approach of using thelowest NOE(A)L is justified unless there isevidence:

1. from pharmacokinetic/metabolic studies thatthe most sensitive species shows a differenttoxicokinetic behaviour than humans and istherefore less relevant as a predictor of humantoxicity than another toxicity test species; or

2. that the toxic effect which has the lowestNOEL/NOEAL is not relevant for humans;or

3. that the lowest NOEL/NOAEL is derivedfrom an inadequate or invalid study.

Thus it is emphasised that the full database must beused and all relevant findings correlated, whendetermining the most appropriate health end-point.

It is important to note also that in public oroccupational health risk assessments, theestablishment of a NOEL/NOAEL is likely to beinfluenced by a consideration of the relevantroute(s) of exposure.

An ADI or TDI is then derived from theNOEL/NOAEL; the qualitative approach takenfollows the principles outlined in the IPCSEnvironmental Health Criteria Monograph No.104 (WHO, 1990). The uncertainty inherent inextrapolation between and within species hasgenerally been dealt with by the use of safety(uncertainty) factors. These factors generally rangefrom 10 to 2000, depending on the source andquality of data, the biological relevance of the end-point, and the hazard assessment (carried out on acase-by-case basis). Safety factors are notnecessarily rigidly applied; the usual safety factor is100, derived by having a factor of 10 for species

extrapolation and a factor of 10 for individualvariation in human populations. In general termsonly, a safety factor of 10 would apply whenappropriate human data were available and,utilising further safety factor of 10–20, an overallsafety factor of 1000–2000 may apply if, forexample, the toxicological database is incomplete orthe nature of the potential hazards indicate theneed for additional caution. The ADI (RfD) iscalculated by dividing the NOEL/NOAEL by thesafety factor. This approach assumes that exposureat less than the ADI is without appreciable risk butthere is no attempt to quantify the level of risk.

For agricultural and veterinary chemicals, once aNOEL/NOAEL has been established and theADI estimated, a maximum residue level (MRL)for food and, in some cases, water can beestablished. The MRL is the maximumconcentration for a residue resulting from the useof a chemical according to good agriculturalpractice that is legally permitted or recognised asacceptable in or on a food, agriculturalcommodity or animal feed; the object ofestablishing an MRL is to keep human intake toa minimum; thus the MRL for a particularchemical may be set well below the level whichwould result in intake equivalent to the ADI.

Where data sets allow appropriate analysis,alternative procedures such as the Benchmark Dose or Effective Dose (EDx) may be used byregulatory agencies in calculating health end-points.

11.3 Determination of Risk-Based EnvironmentalHealth Criteria

Risk-based Environmental Health Criteria shouldbe determined taking into account:

1. The bioavailability of a substance. Thebioavailability should be assumed to be 100 per cent if specific information is notavailable;

2. The Provisional Tolerable Weekly Intake(PTWI) or Acceptable Daily Intake (ADI) asdetermined by the World Health


143

5 US terminology is ‘Reference Dose’, or RfD.


Organisation/Food and AgricultureOrganisation (1987, 1994), or GuidelineDose (GD) for cancer toxic effects asdetermined by national health advisorybodies; and

3. Other potential sources of the substances thatcomprise a proportion of the PTWI or ADI,or GD (e.g. background levels of thesubstance in food, water, air; and the amountof exposure through these routes)(ANZECC/NHMRC, 1992; Imray andLangley, 1998).

The total exposure to a substance ‘X’ can berepresented by the equation:

‘Exposure to substance X = BE+Exposures from contaminated medium by ingestion, inhalation and skin absorption

= BE+amount of substance absorbed from medium.

= BE+(Ming x Cing x Bing + Minh x Cinh x Binh+ Mskin x Cskin x Bskin)

= BE + MEmedium

BE = Background Exposures (e.g. from food and water).

Ming = Amount of medium ingested.

Minh = Amount of medium inhaled and retained.

Mskin = Amount of medium on skin.

Cing = Concentration of substance in medium ingested.

Cinh = Concentration of substance in medium inhaled and retained.

Cskin = Concentration of substance in medium on skin.

Bing = Bioavailability, i.e. percentage absorbed,of substance when ingested.

Binh = Bioavailability of substance when inhaled.

Bskin = Bioavailability of substance when on skin.

MEmedium = Substance exposure from medium.’

(ANZECC/NHMRC, 1992, p. 37)

Different levels of bioavailability will occurbetween the medium ingested, inhaled or incontact with skin.

National environmental health criteria will be setby national health advisory bodies. A variablepercentage of the TI will be allowed for exposureto the contaminated medium. This is consistentwith the IPCS approach and that used in the fourAustralian workshops on the health risk assessmentand management of contaminated sites.

When the PTWI/ADI is used for establishinginvestigation levels for individual contaminants,the basis for the PTWI/ADI should be soughtfrom appropriate documents (e.g. WHO, 1987;WHO, 1989). This information should includetarget organ(s) and effect(s) (e.g. nature,reversibility, and severity, LOAEL for mostsignificant toxic effect); bioavailability; and safetyfactors accounting for variations in humansensitivity and extrapolations from animal studies.

When a Guideline Dose is derived using theNHMRC ‘Toxicity Assessment Guidelines forCarcinogenic Soil Contaminants’, the basis forthe derivation should be fully documented.Guideline Doses for soil contaminants withcancer effects will be determined by nationalhealth advisory bodies or their appointees.

If no PTWI, ADI, or GD is available a specificapproach acceptable to the relevant healthagencies will need to be determined using WHO(1994) for non-carcinogens, or NHMRC‘Toxicity Assessment Guidelines for CarcinogenicSoil Contaminants’ for substances with cancereffects and used for calculations.

It is considered that these methods fordetermining Risk-based Environmental HealthCriteria should protect the entire population withfew exceptions. Where a significant proportion ofthe population demonstrates allergic sensitisation

144

to a substance (e.g. nickel) this will need to beconsidered in criteria setting. People who mayhave unusual sensitivity to agents may need to beconsidered in a risk assessment (Imray andLangley, 1996).

Similar principles can be used for determiningRisk-based Environmental Health Criteria forcontaminants with and without cancer toxiceffects if a Tolerable Intake is available. Themethodology, when applied to soil, was initiallyendorsed in ANZECC/NHMRC (1992).

Qualifications to setting the Risk-basedEnvironmental Health Criteria are:

• ‘In setting a Risk-based EnvironmentalHealth Criteria, total exposure to substanceX, (i.e. the sum of the background exposureand the substance exposure from themedium) should not exceed the ADI orPTWI,[ or GD] i.e. BE+MEmedium < ADIor PTWI, [or GD].’;

• The degree to which exposures at a proposedRisk-based Environmental Health Criteriaare below the ADI or PTWI, or GD will beset by national health advisory bodies and willdepend on factors such as: the nature of theadverse effects, the completeness oftoxicological data, exposure variability withina population and the relative sizes of BE andMEmedium; and

• It should be recognised that ‘…short-termexposure to levels exceeding the PTWI is not acause for concern provided the individual’sintake averaged over longer periods of time doesnot exceed the level set’ (WHO, 1989, p. 9).(adapted from NEPC, 1999)

11.3.1 An Australian model for setting criteria forcarcinogens

The following material is drawn, withamendment, from the NHMRC ‘ToxicityAssessment Guidelines for Carcinogenic SoilContaminants’ (1999, p. 1–16) which wasprepared by a Technical Working Party (TWP).

This approach is consistent with otherinternational risk assessment methodologies. Thedevelopment and use of an agent-specificGuideline Dose is consistent with current riskassessment practice in Australia as well as withinternational practice. For example, the GuidelineDose and its use in risk assessment is analogousto the Acceptable Daily Intake (ADI) and the USReference Dose (RfD). The Benchmark Doseapproach is proposed in the Draft Revision to theguidelines for carcinogenic risk assessment (USEPA, 1996).

11.3.2 BackgroundThe traditional benchmark dose methodology(traditional BMD) has been developed over thelast two decades and is now being given seriousconsideration as a useful tool in risk assessment(Dourson, 1984; Barnes et al, 1995; US EPA,1995a; 1996). More recently, the approach hasalso been proposed for cancer risk assessment(e.g. US EPA, 1996)

The methodology for use in Australia avoidssome of the limitations inherent in existing cancerrisk assessment methods. In particular, themethodology optimises the advantage of using allrelevant scientific data in the decision-makingprocess and provides for a clear separation andjustification of the major components of theprocess: public health policy; professionaljudgement; and scientific principles and data.

The methodology is a two step process. Firstly,the modified-BMD is derived from theexperimental data. Secondly, the modified-BMDis divided by cumulative factors to derive aGuideline Dose for human exposure.

The modified-BMD is set using 5 per cent extrarisk determined from animal or epidemiologicalstudies. This extra risk is then divided by a seriesof modifying factors (potentially up to 50,000)after consideration of all the availabletoxicological data according to a specified decisiontree to derive an agent-specific Guideline Doseprotective of public health. These factors relate tointer- and intraspecies variation, quality of thedatabase and other factors for the seriousness of


145


146

Figure 9: Decision Tree for the development of risk-based environmental health criteria

Identify agent

No

Yes

No Yes

Yes

Is there an ADI, TDI or PTWI for chemicals or

endorsed exposure standards for other agents?

Does the agent pose a health hazard?

Yes

Is it relevant to humans?

Assemble and assess data

Derive generic risk-based environmental health criterion

Derive situation-specific environmental health criterion

using situation-specific Health Risk Assessment

Derive Tolerable Intake or Guideline Dose (chemicals) or Exposure Value (other Agents)

Are there new data/should data

be reassessed

Interpret situation investigation

results

No

No

Information Inputs - Type of users - Single or multiple agents - Agent concentration and distribution - Bioavailability of agents - Physico-chemical properties of agent - Potential exposure pathways - Uncertainty assessment, etc

the carcinogenic response. The factors are derivedaccording to a decision tree which takes intoaccount all of the available data; and usesscientific judgement to address a number of theuncertainties in the risk assessment process and todevelop safety factors.

All available, relevant information should be usedin the risk assessment process. In cases wherethere are few or inadequate data, conservatismmay be justified and the use of conservative(default) assumptions is supported.Recommendations on default assumptions areprovided for cases where the data are incompleteto bridge data gaps and allow the risk assessmentto proceed. All choices, both those based onscientific data and those based on defaultassumptions, must be supported by reasoned andcritical analytical arguments.

The Guideline Dose is established by regulatoryauthorities and is defined as the daily intake of achemical agent which, during a lifetime, is unlikelyto result in cancer, based on a comprehensiveexpert assessment of the best information availableat the time. It is considered that the GuidelineDose is protective of public health.

The Guideline Dose may be used in thedevelopment of health investigation levels,response levels and risk characterisation of humanexposures to contaminants in soil.

The Guideline Dose does not attempt to modelor predict a response incidence at lowenvironmental exposure. It is an estimate of thedose which is considered protective of publichealth (the use of compounding factors assures ahigh level of safety). This places the focus ofregulation on the control of exposure toenvironmental contaminants rather thancalculation or discussions of risk. This approachhas the added benefit of allowing comparisonswith guidance values based on non-cancer healtheffects for chemical agents.

A numerical value that would constitute anacceptable level of risk for low-levelenvironmental exposure to carcinogens is notrecommended. Whilst there has been

considerable debate over the last twenty yearsabout what constitutes an acceptable risk, there isno agreed position internationally on this issue(see Department of the Environment, 1993).The problems of nominating an acceptable levelof risk are compounded by the inability of currentmethods to accurately quantitate risk at low levelsof exposure and hence to provide an accuratevalue that can be compared to ‘an acceptable levelof risk’.

Key points about the methodology are:

• Maximum use is made of scientific information,while not requiring the assessor to make ajudgement regarding the existence of abiological threshold, nor perform mathematicaldose–response modelling well below the rangeof experimental data because the dose associatedwith 5 per cent extra risk is set near the lowerlimit of responses that can be measuredexperimentally. With the proposedmethodology, it is not necessary to resolve theuncertainties, difficulties and controversiesassociated with mathematical extrapolation tolow doses outside the range of experimentaldata.

• The approach is relatively model-independentwhen compared with methods whichextrapolate to extremely low doses in thesense that the values of the modified-BMDwhich are determined are not greatlyinfluenced by the mathematical modelchosen. Therefore, different models can givethe data a similar goodness of fit. In contrast,extrapolation well below the experimentalrange by other quantitative risk assessmentmethods is very much model dependent andresults are highly variable with differentmodels (Maynard et al, 1995).

• The modified-BMD is standardised to onelevel of extra risk (i.e. 5 per cent), allowingcomparisons of potency between carcinogensin the observed dose range in the animalbioassay or other modelled data. In addition,extra risk in the observed range can becompared between carcinogens for a given dose.


147


• The modified-BMD method is applied toboth genotoxic or non-genotoxic carcinogens.In addition, it readily allows for the direct useof mechanistic data when an appropriatemechanistic model relating to dose–responsecan be developed. The TWP considered thatthe distinction between genotoxic andnon-genotoxic features of carcinogens isrelevant to public health protection andshould be considered in the cancer riskassessment, but not in determining the shapeof the dose–response curve at doses wellbelow the experimental range. The genotoxicproperties of an agent are an important partof the assessment and are accounted for inthe consideration of the seriousness of thecarcinogenic response.

• The modified-BMD is a numerical estimateof the dose associated with a particularresponse and by itself does not reflect theuncertainties inherent in biological data. Duecare should be taken to describe theuncertainties (Lu and Sielken Jr, 1991).

The methodology can be compared to non-threshold models currently in use which assumelow dose linearity (e.g. the US EPAmethodology). (See comments on non-thresholdmodel, Section 5.5, page 95).

11.3.3 General principles of themethodology

1. Identify the relevant soil contaminants.

2. For each of the contaminants, check whetheran ADI, PTWI or TDI has been set by theWHO. In cases where an ADI, PTWI orTDI is available, then:

• Ascertain whether there are new datawhich should be assessed or whether thederivation of the tolerable intake should bereviewed. If yes, proceed to step 3.

• If cancer or genotoxicity was not aconsideration in deriving the value andcurrent scientific information does notchange the judgement that cancer shouldnot be considered, then use the ADI,PTWI or TDI as described elsewhere foradverse effects other than cancer(ANZECC/NHMRC, 1992):

• If the carcinogenic or genotoxic propertiesof the chemical agent were assessed andconsidered in deriving the guidance value,the derivation of the guidance value (andany compelling new scientific evidence)should be reviewed and a decision madewhether or not the toxicological propertiesof the substance should be reassessed. Ifyes, proceed to step 3. If no, then use theADI, PTWI or TDI as describedelsewhere for adverse effects other thancancer (ANZECC/NHMRC, 1992).

If no ADI, PTWI or TDI is available, proceed asfollows:

3. Search the peer-reviewed scientific literatureor any other, scientifically sound, availablesource to find all relevant data. Assess theadequacy of data collected to determinewhich will be selected for use in undertakingthe following steps.

4. Based on studies judged to be adequate,determine whether the contaminant poses acarcinogenic hazard.

5. If the agent does not pose a carcinogenichazard or if there is insufficient informationcurrently available to make an assessment, nofurther evaluation of the carcinogenic hazardis needed. Proceed as for adverse effects otherthan cancer for development of a health-based regulatory value (see Addendum 1).Write findings in the report.

148

6. If the agent is considered to pose acarcinogenic hazard, determine whether theobserved carcinogenic hazards are relevant tohumans. If found to be not relevant, nofurther evaluation of the carcinogenic hazardis needed. Proceed as for adverse effects otherthan cancer for development of a health-basedregulatory value. Write findings in the report.

7. If carcinogenic hazards are consideredrelevant to humans, apply the modified-BMDmethod and determine a modified-BMD forall relevant carcinogenic end-pointscorresponding to 5 per cent and 1 per centextra risk.

8. Use route to route extrapolation whereappropriate.

9. Derive and apply appropriate factors tocalculate Guideline Doses for each modified-BMD0.05.

10. Choose the lowest Guideline Dose supportedby the highest possible strength and weight ofevidence.

11. Compare the Guideline Dose for the cancerend-point with the ADI, PTWI or todetermine whether the carcinogenic end-pointis the most sensitive one. Use the lowest ofthese doses for setting health investigationlevels or for site specific risk assessment asoutlined in the ‘Guidelines for Assessmentand Management of Contaminated Sites’(ANZECC/NHMRC, 1992).

12. Write the report.

11.3.4 Further actionsProcesses are being established to develop andprovide Guideline Doses for specific chemicals.As an interim measure, advice on specificchemicals should be sought from the relevantregulatory body. When probabilistic estimates ofrisk are the only guidance available, there needs tobe a full appreciation of the differences between‘real’, ‘estimated’ and ‘actual’ risk and thelimitations of quantitative risk assessment ofcarcinogens (See Hrudey, 1998).


149


Figure 10: Decision Tree for cancer risk assessment

Note: appendices refer to NHMRC (1999) (from NHMRC, 1999)

150

Identify chemical

No

No

NoYes NoIs there an ADI, TDI or PTWI?

Yes

Yes

Yes

Does the agent pose carcinogenic hazard?

Is it relevant to humans?

Is cancer the most sensitive endpoint?

Assemble and assess data

For all relevant endpoints, list BMD0.05

Yes

Yes

Derive modified-BMD

Apply factors to BMD0.05

Choose lowest dose

Write Report

Use Guideline Dose for risk characterisation and further

guideline development

Write Report

Use othe guidance values for risk characterisation and further

guideline development

GUIDELINE DOSE

Are there new data/should data be reassessed?

Were cancer data part of the assessment?

Should data be reassessed?

No

No

No

Yes

Appendices

In the following Appendices, key issues related tothe Environmental Health Risk Assessment aredetailed.

• Appendix 1—Environmental Health RiskAssessment for Contaminated Sites

• Appendix 2—Environmental Health RiskAssessment for Air Pollutants

• Appendix 3—Environmental Health RiskAssessment for Food

• Appendix 4—Environmental Health RiskAssessment for Water

• Appendix 5—Australian Models of RiskAssessment

• Appendix 6—International Models of RiskAssessment

• Appendix 7—WHO/IPCS ConceptualFramework for Cancer Risk Assessment

• Appendix 8—Microbiological RiskAssessment


153


1.1 IntroductionThe risk assessment framework is detailed inGuideline B4 ‘Health Risk Assessment ofContaminated Sites’ in the National EnvironmentProtection Measure for the Assessment ofContaminated Sites (National EnvironmentProtection Council, 1999). Guideline B4 wasdeveloped by health agency personnel assisting inthe development of the National EnvironmentProtection Measure. Guideline B4 builds on thematerial in the ANZECC/NHMRC ‘Australianand New Zealand Guidelines for the Assessmentand Management of Contaminated Sites’ (1992).

The framework uses four stages that arecompatible with the five stages in this document.The four stages are Data Collection; ToxicologicalAssessment; Exposure Assessment; and RiskCharacterisation. ‘Data Collection’ is anequivalent component of ‘Identifying the Issues’and ‘Toxicological Assessment’ is equivalent tothe combination of ‘Hazard identification’ and‘Dose–Response Relationship’.

1.2 Identifying the Issues1.2.1 Data collectionContaminated sites usually have a heterogeneousdistribution of contamination and the samplingmethods and the presentation of the data arecritical to a good appreciation of the nature anddistribution of the contamination.

1.2.2 Sampling

Random sampling

This may lead to large areas of the site beingmissed for sampling because of a chancedistribution of results. It also neglects priorknowledge of the site (Heyworth, 1991).

Stratified random sampling

By dividing the site into areas and randomlysampling within each area, missing large tracts ofland can be avoided (ibid).

Stratified sampling

The site is divided and different samplingpatterns and densities are used in different sub-

areas. It is useful for large and complex sites(Standards Australia, 1997).

Grid (systematic) sampling

This permits the whole of the site to be coveredand for sampling points to be more readilyidentified for further sampling (Heyworth, 1991).

Judgemental sampling

Sampling is localised to areas identified fromknowledge of the site. There needs to be a highlevel of confidence in the reliability of informationabout the site and that the information reflects thecurrent state of the site (ibid).

The statistical advantages of a stratified randomgrid have been discussed by Heyworth (1991).Against the statistical advantages, are thedifficulties in surveying a site (and hence alsorelocating sampling points if this provesnecessary) and the inability to draw vertical cross-sections of the site. Being able to drawcross-sections is useful in predicting the‘dimensions’ of the contamination. (van Alphen,1993). There are some practical difficulties inmapping and conducting the sampling for otherthan square grids.

Ferguson (1992) critically reviewed a range ofsampling designs to determine their performancein detecting a single circular target occupying 5 per cent of the total site area. A herringbonepattern was found more successful than simplerandom, stratified random, stratified systematicunaligned and regular square grid patterns. Searchperformance varied with target shape, size and thenumber of targets. For an elliptical (4:1) target,the performance of a grid pattern was ‘greatlydegraded when the target lies parallel to the griddirections and, to a much lesser extent, the 45°direction’ (p. 36).

Expert judgement will often be the keydeterminant of the sampling pattern. A randomor grid approach should not be used when there isinformation indicating particular patterns orlocalisation of an agent. A grid or randomsampling will be appropriate when samplingcannot be guided by prior knowledge.

154

Environmental Health Risk Assessment for Contaminated Sites

Appendix 1

A combination of judgemental sampling and gridsampling may be the most feasible approach for asite. In these instances sampling will beconcentrated in areas identified by informationwith grid sampling providing coverage for theremainder of the setting.

Random or grid sampling will usually be theapproach of choice to validate cleanup orintervention although judgemental sampling maybe justifiable for assessing particular features inpost-intervention sampling.

Information on sampling is also available in:Standards Australia (1997). Australian Standard.Guide to the sampling and investigation ofpotentially contaminated soil. Part 1: Non-volatileand semi-volatile compounds. AS 4482.1-1997

1.2.3 Sampling density‘Statistical equations are tools to be used as aidsto common sense and not as a substitute for it’(Keith 1990 page 612). Statistical formulae fordetermining sampling density are usually basedon the requirements that the results will benormally distributed (i.e. in a bell-shaped curve)and that a particular concentration is equallylikely to occur at any point. Some analyticaltechniques require an estimate of the mean of theresults and the standard deviation of results beforesampling density can be calculated. Theserequirements can rarely be met during the stagesof initial and detailed investigation, as sites areoften heterogeneous with a skewed distribution ofresults.

A considerable amount of expert judgement isrequired to determine the density of sampling.The final amount will depend on an integratedappraisal of factors including:

• proposed or current activities and users;

• the number of stages of sampling consideredfeasible;

• the size of the site and final subsites if thesite is to be subdivided;

• the distribution of uses on the site and thedisposition of structures;

• the site history; and

• potential remediation and managementstrategies.

If a large site is to be subdivided to smallerresidential sites the sampling density on the finalsites rather than the initial larger site should alsobe considered. While sound predictions may beable to be made from the many results availablefor the larger site which can suggest patterns andtrends, more detailed sampling may be needed toaddress what is occurring on the smaller sites.

1.2.4 Composite samplingComposite sampling usually interferes with healthrisk assessment and is generally unsuitable fordefinitive health risk assessment due to theinherent uncertainties in the resultant data (Lock, 1996).

Localised elevated concentrations can beobscured, the data is harder to interpret withoutundertaking further single samples, and thecompositing process provides an extra opportunityfor bias and error. Proposed cost savings canquickly evaporate when further sampling isrequired to clarify composite results.

Composite sampling should not be used forsituation-specific health risk assessments.

Compositing may create a false negative result.For example, if 4 soil samples have individuallevels of 5, 5, 5, 165mg/kg, a compositeconcentration of 45 mg/kg appears to complywith the unadjusted Health-based InvestigationLevel (HBIL) of 100mg/kg. Comparison with anunadjusted HBIL obscures the presence of alocalised elevation and hence the nature ofcontamination on the site.

Adjusting the soil criterion (e.g. the Health-basedInvestigation Level) may be done in an attempt toaddress this problem. If there are 4 constituentsamples per composite, the HBIL for arseniccould be divided by 4 so that concentrations arethen be compared to [100/4]mg/kg, i.e. 25 mg/kg.However, this may result in false positiveexceedances if, for example, natural backgroundlevels are elevated but acceptable. If the


155


background level of arsenic is 35 mg/kg, theconcentration determined in an uncontaminatedcomposite manufactured from four samples willexceed the adjusted HBIL of 25 mg/kg. In such acase, the use of composite samples withoutdefinitive individual sample analysis may lead toan unnecessary and costly further investigation.

Further information on the problems of, andconstraints on, composite sampling is detailed in:

• Lock WH. (1996). Composite Sampling.National Environmental Health ForumMonographs Soil Series No. 3. Adelaide:South Australian Health Commission; and

• Standards Australia (1997). AustralianStandard. Guide to the sampling andinvestigation of potentially contaminated soil.Part 1: Non-volatile and semi-volatilecompounds. AS 4482.1-1997.

1.2.5 Analytical methodologies

General reference

• Manahan SE (1993). Fundamental ofEnvironmental Chemistry. Boca Raton: LewisPublishers

• Perkins JL (1997). Modern Industrial Hygiene.Volume 1. Recognition and Evaluation ofChemical Agents. New York: Van NostranReinhold.

Specific reference

Analytical methods should be those described inGuideline 3 Laboratory Analysis of PotentiallyContaminated Soils (NEPC 1999). Guidelines forthe laboratory analysis of contaminated sites.Canberra: Australian and New ZealandEnvironment and Conservation Council. Anexception is the method for Total PetroleumHydrocarbons which will need amending toprovide appropriate data for TPH health riskassessment.

1.2.6 Safety plansThe safety of people working on the site andnearby residents must always be considered inenvironmental sampling. Site safety plans shouldbe developed where there may be risks.

Risks may arise from factors such as:

• dealing with unknown substances;

• deep excavations such as backhoe pitspresenting a physical hazard;

• the release of volatile contaminants duringexcavations or their pooling in excavations;

• generation of dust;

• groundwater surveys can createcontamination or cross contamination ofaquifers if bores are not appropriatelyconstructed; and

• the presence of underground storage tankscan cause subsidence if corroded or fire andexplosion hazards.

The specific reference is:

• National Environment Protection Council(1999). National Environment ProtectionMeasure for the Assessment of SiteContamination. Guideline 9. Protection ofHealth and the Environment during theAssessment of Site Contamination. Adelaide:National Environment Protection Council.This is accessible on www.nepc.gov.au.

1.2.7 Assessment of summarystatistic data and presentationof data

Summary statistics

As sites are usually heterogeneous, it is importantto have ways of appreciating the nature anddistribution of complex patterns ofcontamination. A combination of summarystatistics, graphical display and clear explanatorynarrative will be necessary. Table 1 A1 presentsseveral summary statistics for a soil contaminantand highlights the fact that no single summarystatistic (e.g. an arithmetic mean or a median)fully characterises a site. Instead a range ofsummary statistics is needed to build up a pictureof a site.

156

Table 1 A1: Summary statistics for a single analyte and stratum

Stratum: 0–150mm

Chemical name Pb (Lead) mg/kg

Number of samples 53

Range 3 to 5000

Relevant soil criterion 300

Median 60

Arithmetic mean 446

Arithmetic s.d. 1041

Geometric mean 85

Geometric s.d. 6.0

Frequency distributiona n %

less than soil criterion 37 70

≥1 and <2 times soil criterion 6 11



≥10 times soil criterion 4 8

a An arbitrary method is used to categorise data.


157

For multiple analytes, an impression of the typicallevels, location of contaminants within strata,total ‘burden’, and statistical distribution of resultscan be presented as in Table 2 A1.

Censored data

Summary statistics can be biased according to thevalues substituted into mathematical formulae toallow calculations of, for example, means.Sometimes the value of the level of detection issubstituted, upwardly biasing the sample statistics.

Levels of reporting

The first step in dealing with censored data is toensure that the levels of detection or levels ofreporting are appropriate. The levels of reportingmust be less than the relevant criteria against

which the results will be assessed. A level ofreporting of one tenth of the criterion is preferredbut may not be practicable where the criterion isset at a level of detection. They need to be setsufficiently low so as to be able to distinguishtrends from background levels. Lower levels ofdetection may be required for environmentally-based assessments.

The recommended maximum levels of reportingfor health risk assessments are detailed in theNational Environment Protection Measure forthe Assessment of Site Contamination (1999).They are based around the assessment of a‘Standard’ Residential Setting but apply to allinvestigations.


Table 2 A1: Combining sample number, location, depth,multiple analyte results, and extent of variation above the investigation levels in a single table

Locations 7–12: Chemical Screening Results

Depth (mm) Pb As Cd Cr Co Ni Zn Cu Hg pH

Loc 7

0–150 200 3 ~ 100 4 14 210 28 0.25 8.6

150–300 170 3 ~ 80 6 15 220 100* 0.25 8.7

300–450 10 ~ ~ 60 8 20 34 20 0.05 8.6

Loc 8

0–150 36 2 ~ 90 18 75 24 8.0 0.50 8.0

150-300 ~ 2 ~ 110 12 28 46 28 0.05 7.6

Loc 9

0–150 250 3 ~ 90 4 15 310 50 0.55 8.8

150-300 160 2 ~ 85 5 13 240 60 0.40 8.4

750-900 4 ~ 95 11 22 44 26 7.6 -0.05 7.6

Loc 10

0–150 10 ~ ~ 70 ~ 8 16 1.0 -0.05 8.3

150–300 24 5 1 85 5 13 34 1.8 0.05 8.1

300–450 12 3 1 90 7 15 30 1.8 -0.05 8.1

750–900 4 ~ 1 50 6 14 22 1.5 -0.05 8.4

Loc 11

0–150 290 5 ~ 80 4 11 540* 24 0.10 8.3

150–300 450* 10 ~ 85 5 15 760* 1750*** 0.70 8.1

300–450 90 5 ~ 110 9 17 30 1.9 0.05 7.8

12 2 ~ 110 9 17 30 19 0.05 7.8

Loc 12

0–150 100 3 2 85 6 15 80 28 0.25 8.4

150–300 940** 5 ~ 130 7 18 190 60 2.70* 8.4

300–450 46 1 ~ 110 12 24 46 26 0.20 7.8

HIL. 300 100 20 12% 100 600 7000 1000 15 <5 or >9

(1)HIL. = Health-based Investigation Levels (2) All units are in mg/kg except where shown otherwise(3) ~ indicates < Detection Level * denotes ≥ and <2 x HILb, ** denotes ≥2 and <5 x HIL, *** denotes ≥ 5 and < 10 x HIL.

****denotes ≥ 10 x HIL. This is an arbitrary method of categorising data.(adapted from South Australian Housing Trust/South Australian Health Commission format)

158

1.2.8 Content of data collection reports

Integration of reports

Where there is a series of reports, each succeedingreport should summarise the important andrelevant points from the preceding reports. Thiswill assist in the rapid comprehension of newmaterial by all parties involved.

Non-integrated reports result in far less efficientappraisals of data.

Accreditation of laboratories

Laboratories should be accredited by anappropriate body for the particular analyses beingundertaken. A broad form of accreditation maynot be applicable for the particular test.

Analytical techniques

The analytical techniques should comply withtechniques described in NEPC (1999).


159


2.1 IntroductionPoor air quality can have a significant bearing onthe causes and exacerbation of respiratory disease.For example, asthma may be exacerbated by airpollution and over two million or 11 per cent ofAustralians have asthma, including one in fourprimary school children, one in seven teenagersand one in ten adults (AIHW, 2000).

There are several issues that differentiate the riskassessment of air pollutants from pollutants foundin other environmental media.

As individuals we have little control over thequality of the ambient air we breathe. Exposure toair pollutants occurs in all activities, whilstindoors, in motor vehicles, whilst at work andduring recreation. It is important that all sourcesof air pollutants are considered, noting that forsome pollutants, the indoor and occupationalenvironments may contribute the most toexposure. In addition the surface area of theinternal lining of the lungs is 50–70 square metres(about the size of a tennis court) compared to 1–2square metres for the surface area of the skin).There are 300 million alveoli in adult humanlungs and the air-blood barrier (consisting of theaqueous surface, epithelial lining and thininterstitial space) is 0.36 to 2.25mm thickindicating a much larger area for biologicalinteraction to occur (Hrudey et al, 1996).

While the fundamental principles of riskassessment remain the same, different exposureassessment are available when assessing ambientair pollution from diffuse and point source regionsor large areas, localised air pollution from pointsources, and indoor air pollution such as mayoccur in the home or workplace. Where largepopulations are involved, differentepidemiological methodologies such as time-series analysis may be able to be used. The riskassessment of a site-specific situation will differfrom that for the development of a guideline asthe former will usually relate to a specific, definedpopulation while the latter will need to take intoaccount a broader, more diverse population.

Ambient air is usually divided into two groups,the ‘criteria’ pollutants and ‘other’ (refer toSections 2.5.1 and 2.5.2). Criteria pollutants arethose that are common air pollutants, found inrelatively high concentrations. The ‘other’ group ismade up of hazardous air pollutants and otherspecific substances that are found in traceconcentrations, are specific to a particular settingor activity and are monitored on a needs basis.

Irritant effects are often the critical health effectwith criteria pollutants and the irritant effectsmay occur from short exposures with negligiblesystemic absorption. Other non-irritant healtheffects such as carcinogenicity (eg for benzene),mutagenicity and neurotoxicity are receivingincreasing attention.

The criteria pollutants, carbon monoxide (CO),sulfur dioxide (SO2), nitrogen dioxide (NO2),Particulate Matter 10µm (PM10), photochemicalsmog (measured as ozone) and lead (Pb) aretypically monitored via a network of monitoringstations. These networks are usually located tomeet environmental management objectives.Monitoring stations are selected for a range ofreasons including monitoring of emissions fromindustrial facilities, major roadways or where highconcentrations of secondary pollutants may befound. Ambient air exposures to pollutants arehighly dependent on meteorological factors.

An extensive review of risk assessment related tothe development of ambient air criteria forAustralia has been published: Report of the RiskAssessment Taskforce (National EnvironmentProtection Council, 2000). The Report reviews arange of risk assessment methodologies that couldbe applied to the development of ambient aircriteria as well as detailed appendices coveringepidemiological study design and datarequirements, health effects of criteria airpollutants, ambient air monitoring programs inAustralia and provides an example of the possibleuse of risk assessment related to the review ofparticle standards.

160

Environmental Health Risk Assessment for Air Pollutants

Appendix 2

2.2 Identifying the IssuesThe degree of air pollution from criteriapollutants and the identity of the pollutantspresent are obtained usually from the fixed sitemonitoring stations that are intended to provideinformation about the quality of air in an airshedand the population exposure to the pollutants.This takes into account diffuse source emissionssuch as vehicle emissions from motorways as wellas point source emissions from industry. Specificinformation on air quality within a certain regionthat has major point source emissions, such asfrom a factory or a power station, that can havemajor impacts on the surrounding area, needsintensive local monitoring for pollutants that areknown to be emitted from the local source. Suchmonitoring is often carried out by thejurisdictional environment agency or by theindustry itself. New developments usually requireEnvironment Impact Statements (EIS) todetermine the contribution of the new source tothe air quality in that area and the size of thepopulation potentially exposed. EIS usually relyon air modelling data that are derived fromknown emissions and processes.

The levels of pollutants are compared with airquality standards that are usually health-basedcriteria, set either at levels below which adversehealth effects are not expected to occur or atlevels where the incidence of adverse healtheffects is considered acceptable. If the levels arefound to exceed those in the standards,appropriate strategies and actions will need to beinstigated to reduce the pollutants in the airshed.

2.3 Hazard IdentificationFor the criteria pollutants, most of the healtheffect data can be found in the epidemiologicalliterature. For the hazardous air pollutants, bothtoxicological data and usually indoor oroccupational studies are used during this stage.

2.3.1 Analytical methodologies

General references

• Manahan SE (1993). Fundamental ofEnvironmental Chemistry. Lewis Publishers.Boca Raton.

• Perkins JL (1997). Modern Industrial Hygiene.Volume 1. Recognition and Evaluation of ChemicalAgents. Van Nostran Reinhold. New York.

Air

• Clean Air Society of Australia and NewZealand (latest edition). Air pollutionmeasurement manual. A practical guide tosampling and analysis. Volumes 1 and 2.Latest edition. Eastwood, South Australia.

• American Conference of GovernmentIndustrial Hygienists (1988). Advances in AirSampling. Lewis Publishers. Chelsea.

• Lodge JP (1989) Methods of Air Sampling andAnalysis—3rd Edition. Lewis Publishers.Chelsea.

• United States Environmental ProtectionAgency (US EPA) (1999). Compendium ofMethods for the Determination of Toxic OrganicCompounds in Ambient Air—2nd Edition,EPA-625/R-96/010b, Center forEnvironmental Research Information, USEPA. January 1999. Cincinnati, Ohio.

• US EPA (1999). Compendium of Methods forthe Determination of Inorganic Compounds inAmbient Air. EPA 625/R-96/010a, Office ofResearch and Development, US EPA. June1999. Washington DC.

• US EPA (1994). Quality Assurance Handbookfor Air Pollution Measurement Systems.US Environmental Protection Agency.EPA 600/R-94-038b, May 1994.


161


• US EPA (1990). Compendium of Methods forthe Determination of Air Pollutants in IndoorAir. EPA 600/4-90-010, April 1990.

• US EPA (1983a). Technical AssistanceDocument for Sampling and Analysis of ToxicOrganic Compounds in Ambient Air. US EPA,EPA 600/4-83-027, June 1983.

2.3.2 Dose–response assessmentDose–response data for most classical airpollutants have been determined from humanexposure data, (chamber studies), using a singlecompound at a time and various exposuredurations and concentrations. Much of thedose–response data for PM10 (and for ozone to alesser extent) has been obtained fromepidemiological studies. More recent chamberstudies also use mixtures of compounds orsequential dosing with different compounds to tryto model exposures in the ambient environment.Each pollutant is assessed against specific healthendpoints characteristic to that pollutant.

Sensitive populations, such as children, have beenidentified and studied using chamber exposures.

Animal chamber exposure data are used tosupport human exposure data, often to identifyphysiological mechanisms of action.

Dose–response data will often be available fromthe documentation used in the development ofcriteria by organisation such as WHO.

2.3.3 Use of epidemiological dataTime-series epidemiological analyses are beingused increasingly to examine public health effectsof criteria pollutants. These may show short-termvariations in air pollution levels within the usualrange in the levels of hospital admissions, acuterespiratory and cardiovascular disease or mortality.There are considerable problems of interpretationwith these studies and the weak associations foundare generally in conflict with the expected risk fromthe low concentrations of air pollutants present(IEH, 1999b). There is however, some coherencebetween different study types and designs tosupport the findings of time-series studies.

Times series studies require a population ofsufficient number. NEPC (2000) also notes thattime series studies require:

• daily estimates of population exposure for atleast five years at constant locations;

• sufficient fixed sites in the monitoringnetwork to characterise the spatialdistribution of air pollutants in the studyregion, i.e. sub regions within the airshedcontain at least one monitoring site;

• sub-regional monitoring sites provide ameasure of the distribution of populationexposure not peak data;

• daily data from each sub-regional monitoringsite available for at least 75 per cent of days;and

• pollutants are not measured independently sothat potential confounding can be assessed.

2.3.4 Sensitive sub-populationsFor air pollutants, there are particular sub-populations that show increased sensitivity tosome compounds. For example, 10 per cent of thepopulation is particularly sensitive to ozone,severe asthmatics are sensitive to sulfur dioxide,people with angina and coronary heart disease aresensitive to the effects of carbon monoxide, andasthmatics and those with chronic lung diseaseare sensitive to nitrogen dioxide, and carbonmonoxide (IEH, 1999c).

2.3.5 Use of occupational criteriaWorksafe Australia has developed a range ofExposure Standards for the occupationalenvironment. The Exposure Standards should notbe applied to ambient air quality as they areintended to protect healthy adult workers and donot take into account the very old, the very youngand the infirm. In setting general environmentalstandards safety factors are applied to take thesegroups into account. Occupational Standards areoften based around exposures for a 40 hour week(compared to the 168 hours that a person will beexposed to the general environment) and for aworking lifetime, rather than a complete lifetime.

162

Worksafe exposure standards should not beapplied without reference to the Guidance Noteon the Interpretation of Exposure Standards forAtmospheric Contaminants in the OccupationalEnvironment [NOHSC:3008(1995)], and to therelated documentation which explains therationale for the individual criteria.

2.3.6 Exposure assessmentExposure assessment is the most problematic areain the risk assessment process for air pollutantsand is the source of most of the uncertainty in theestimation of risk.

Some of the major difficulties are factors such astime and spatial variation of air pollutants and thefact that personal exposures are likely to rapidlyvary and be affected by mobility and a range ofother meteorological, physical and chemicalprocesses. Exercise, for example, may substantiallyincrease exposure at a given air concentration andexposure may also be affected by pre-existing lungdisease (Samet, 1999).

2.3.7 Criteria pollutantsIt is very difficult to determine from ambient airmonitoring data from representative monitoringstations the number of people exposed, the duration ofexposure and the actual level of exposure to airpollutants. Generally, worst case scenarios are used andexposure estimations are made using the highest dailyor hourly concentrations. An assumption that most ofthe population in that region of the monitoring stationairshed has been exposed is made. The data fromseveral monitoring stations are usually averaged andconsidered representative of the population. In the caseof some criteria pollutants, there are wide variationsbetween concentration ranges due to meteorologicalfactors and geography. Some methods are available tointerpolate between monitoring stations and weightexposures according to population factors. Modellingtechniques to improve exposure estimates are alsoavailable, however the models have not been validatedfor the Australian context.

If available, personal monitoring data wouldprovide a much better indication on actualexposure, however this data is rarely available.Daily diaries of location and activities may also beused to identify where exposures have occurred.This method is costly and labour intensive so onlysmall numbers of persons have been monitored bysuch ways in Australia.

2.3.8 Other pollutantsExposures from point sources can be estimatedusing dispersion modelling techniques, but theydepend on a considerable knowledge of factorssuch as processes and dispersion characteristics.

2.4 Exposure Defaults2.4.1 The volume of exposureThe amount of air moved into and out of thehuman lungs is about 6 litres per minute at rest(8.6m3/day) but can increase up to about 60 litresper minute (86m3/day if this level of exertioncould be sustained). For risk assessments a valueof 22m3/day is used for adults and 15m3/day for achild (10 years). For further information refer toSection 7.16.3.

2.4.2 Dispersion modelling guidanceThe use and interpretation of air modelling datais a specialised field.

Many models are available. Different models willbe required to assess point source versus diffusesources of contamination; for different types ofcontaminants (e.g. gases versus particles); and fordifferent scenarios (e.g. a whole airshed versusproperties neighbouring a factory).

For modelling data, the following minimuminformation is required:

• wind speed;

• wind direction;

• air temperature;

• mixing height (estimated or measured); and

• atmospheric stability.


163


Table 1 A2: Physicochemical properties of chemicals and the atmosphericenvironment important in transport-fate calculations

Properties of chemical Properties of environment

Physical properties: Particulate load

• Molecular weight • For dusts, other solids particulate mater

• Density • For liquids, aerosols

• Vapour pressure (or boiling point) Oxidant level

• Water solubility Temperature

• Henry’s constant (air-water distribution coefficient) Relative humidity

• Lipid solubility (or octanol-water distribution coefficient) Amount and frequency of precipitation

Chemical properties Meteorologic characteristics

• Oxidation • Ventilation

• Hydrolysis • Inversion

• Photolysis Surface cover

• Microbial decomposition • Water

• Other modes of decomposition • Vegetation

Particle properties • Soil type

• Size

• Surface area

• Chemical composition

• Solubility

(adapted with permission from NAS, 1991)

164

Other information to be considered is:

• trapping of plumes in mixed layers of limitedheight;

• vertical plume dispersion in convectiveconditions;

• strength of capping inversions above mixedlayers

• the standard deviation of the direction;

• relative humidity or a related parameter;

• surface layer heat flux, moisture flux andfriction velocity; and

• fumigation of plumes.

(WA Department of EnvironmentalProtection, 1999)

2.5 Risk Characterisation2.5.1 Comparisons with criteriaThe initial comparison should be with theNational Environment Protection Measure(NEPM) for ambient air quality standards andgoals, and the NHMRC indoor air quality goals.

The NEPM criteria are available for ambient airpollution for the substances listed in Table 3 A2.

Table 2 A2: Exposure modelling parameters for point sources.

Input requirements/data needs Output requirements

Release point locations Concentrations at each receptor point

Mass emission rate of compounds to be studied

Concentration of individual or aggregated compounds

Stack height

Stack velocity

Stack temperature

Stack diameter

Geography

Rural/urban site classification

Local meteorological data

Receptor locations for concentration predictions

Frequency and duration of short-term (intermittent) releases

(adapted with permission from NAS, 1991)

Table 3 A2: NEPM standards and goals

Pollutant Averaging Maximum Goal within 10 years period concentration maximum allowable

exceedances

Carbon monoxide 8 hours 9.0 ppm 1 day a year

Nitrogen dioxide 1 hour 0.12 ppm 1 day a year 1 year 0.03 ppm none

Photochemical oxidants 1 hour 0.10 ppm 1 day a year(as ozone) 4 hours 0.08 ppm 1 day a year

Sulfur dioxide 1 hour 0.20 ppm 1 day a year1 day 0.08 ppm 1 day a year1 year 0.02 ppm none

Lead 1 year 0.50 µg/m3 none

Particles as PM10 1 day 50 µg/m3 5 days a year


165

Further information is available atwww.nepc.gov.au

Table 4 A2 details NHMRC criteria for a rangeof substances relevant to indoor air quality.


Table 4 A2: Indoor interim national air quality goals recommended by theNational Health and Medical Research Council

Pollutant Measurement Units Measurement Notes NHMRC(µg/m3)* (ppm) criteria session at

whichrecommended

Carbon 10,000 9 Eight hour This period of 98th (Oct 1984)monoxide average not measurement is(CO) to be exceeded not to be

more than once confused witha year that for Threshold

Limit Values.

Formaldehyde # 120 0.1 Not to be exceeded Within domestic 93rd ( June 1982)premises and schools.

Lead 1.5 - Three months average - 88th (Oct 1979)

Ozone, 210 0.10 Maximum hourly A public warning to 119th ( June 1995)photochemical average not to be be given if ozoneoxidants exceeded more than levels are expected

once a year to rise above 500 g/m3

(0.25 ppm).

170 0.08 Four hour average - 119th ( June 1995)

Radon # 200 becquerels - Annual mean Action level for 109th (May 1990)per cubic metre simple remedial action(5.4 nCi/m3) in Australian homes.

Where the concentration exceeds this level, householders should consult the appropriate State authority for advice.

Sulfates 15 - Annual mean - 104th (Nov 1987)

Sulfur dioxide 700 0.25 Ten minute average 120th (Nov 1995)(SO2)

570 0.20 One hour average 120th (Nov 1995)

60 0.02 Annual mean 106th (Nov 1988)

Total Suspended 90 - Annual mean TSP goal to be read 92nd (Oct 1981)Particulates (TSP) in conjunction with

annual SO2 goal.

Total Volatile 500 - Hourly average A single compound 115th ( June 1993)Organic shall not contributeCompounds more than 50% of

the total

* Expressed at 0oC and 101.3 kPa, # Formaldehyde and Radon are final goals(NHMRC, May 1996)

166

Caution: at theserecommended levels,there still may besome people (forexample, asthmaticsand those sufferingchronic lung disease)who will experiencerespiratory symptomsand may need furthermedical advice ormedication.

For the non criteria pollutants for which there areno NEPM standards or NHMRC goals, theWHO Air Quality Guidelines for Europe(WHO, 1999a) may be used. The WHOguidelines are listed in Table 5 A2. The WHOguidelines are applicable to both ambient andindoor air.

The following text has been taken directly fromthe WHO (1999a) documentation.

‘The WHO ‘Guidelines for Air Quality’values are levels of air pollution below whichlifetime exposure, or exposure for a givenaveraging time, does not constitute asignificant health risk. If these limits areexceeded in the short-term it does not meanthat adverse health effects automatically occur;

however the risk of such effects increases.Although the ‘Guidelines for Air Quality’values are health- or environment-basedlevels they are not standards per se. Airquality standards are air quality guidelinespromulgated by governments, for whichadditional factors may be considered. Forexample, the prevailing exposure levels, thenatural background contamination,environmental conditions such as temperature,humidity and altitude, and socio-economicfactors.’

The tables for health endpoint relationships forPM10 and PM2.5 can be obtained from the WHOwebsite: www.who.int/peh/air/Airqualitygd.htm


167

Table 5 A2: WHO air quality guidelines for Europe organic air pollutants

Compound Guideline value (µg/m3) Averaging time

Acetaldehyde 2000 24 hours

Acrolein 50 30 minutes

Acrylic acid 54 1 year

2-Butoxyethanol 13,000 1 week

Carbon disulfide 100 24 hours

Chloroform 1.3 24 hours

1,2-Dichloroethane 700 24 hours

Dichloromethane 3,000 24 hours

Di-n-butyl phthalate 0.05 24 hours

Styrene 70 (odour) 30 minutes

Tetrachloroethylene 250 24 hours

Toluene 260 1 week 1,000 30 minutes

Xylenes 4,800 24 hours 4,400 30 minutes

(adapted from WHO, 1999a)


Table 6 A2: WHO air quality guidelines for Europe inorganic air pollutants

Compound Guideline value (µg/m3) Averaging time

Cadmium 0.005 1 year

Carbon monoxide 100,000 15 minutes 60,000 30 minutes 30,000 1 hour 10,000 8 hours

Fluorides 1 1 year

Hydrogen sulfide 7 (odour) 30 minutes

Lead 0.5 1 year

Manganese 0.15 1 year

Mercury (inorganic) 1 1 year

Nitrogen dioxide 200 1 hour 40 1 year

Ozone 120 8 hours

Sulfur dioxide 500 10 minutes 125 24 hours 50 1 year

(adapted from WHO, 1999a)

168

Environmental Health Risk Assessment for Food

Appendix 3

3.1 IntroductionThe key reference is the ‘Framework for theassessment and management of food-relatedhealth risks’ (ANZFA, 1996) which is the sourceof most of the material in this Appendix.ANZFA is revising its risk management processesin 2001.

Food-related risks can occur because of a range offactors and in many cases, the interdependence ofthese factors needs to be considered whenassessing risk.

Some of the risk factors associated with food (inalphabetical order)

• agricultural chemical residues;

• biological agents;

• cooking and process-related artefacts;

• environmental contaminants;

• food additives;

• food processing aids;

• marine toxins;

• microbiological agents;

• mycotoxins;

• novel foods;

• novel ingredients;

• nutrient imbalance;

• packaging migrants;

• physical agents;

• plant toxins;

• radionuclides; and

• veterinary chemical residues.

(ANZFA, 1996)

3.2 Identifying the Issues

3.3 Hazard IdentificationThe hazards associated with an agent will beaffected by:

• the structure and associated physicochemicalproperties;

• the metabolism and toxicokinetics of thesubstance; and

• the results of a series of toxicity testsconducted both in animal models and/or in in vitro systems.

For microbiological agents, hazard identificationconsists of identification of the microorganismsand/or microbiological toxins of concern. TheInternational Commission of MicrobiologicalSpecifications for Food (ICMSF) has categorisedthe most serious and common of themicrobiological hazards according to the severityof the hazard they present. Data for classifyingmicrobiological hazards may come from animalstudies, but more commonly from controlledhuman studies, epidemiological studies, or studieson outbreaks of food-borne diseases.

3.4 Chemical Risk AssessmentChemicals in food can be categorised aseither food ingredients, food additives orfood contaminants.

3.4.1 Particular issuesMuch of the chemical risk assessment is basedaround the development and use of AcceptableDaily Intakes (ADIs).

While the use of ADI approach is applicable formany chemicals specifically added to food, otherclasses of chemicals may sometimes require adifferent approach. These include traditionallyused food additives and processing aids, nutrients,genotoxic carcinogens, some naturally occurringchemicals, and environmental contaminants.


169


For many traditionally used food additives andprocessing aids, there is little toxicity data uponwhich to base an assessment of risk. Many suchchemicals have a long history of use in foods orare members of chemical classes known to be oflow toxicity. For some chemical groups, standardtoxicity tests may be inappropriate to assesshazard. Generally, acceptable levels of intake aredetermined using a combination of toxicity data,information on traditional food use,structure/activity relationships, metabolic data andtoxicokinetic data.

Nutrients in food must also be consideredseparately since, regardless of the potentialtoxicity, there is a nutritional requirement thatmust be met. The appropriate total intake of aparticular nutrient, therefore, must fall within arange, the upper bound of which should not bewithin the toxicity range. The margin of safetytherefore varies for each nutrient, and may beinfluenced by age and sex of individuals or geneticdifferences between population groups.Epidemiological data are the most appropriatedata for assessing risk in this instance but oftenlack the degree of sensitivity required to obtainmeaningful conclusions. For essential traceelements there will be deficiency symptoms at lowconcentrations and toxic effects at highconcentrations. For this reason, AcceptableRanges of Oral Intake (AROIs) are beingdetermined which are the ranges between thoseconcentrations causing deficiency and thosecausing toxicity.

Carcinogenic chemicals which are also genotoxic(i.e. capable of causing genetic damage) present aparticular problem with regard to safetyassessment. Because of their ability to produceDNA damage at very low dose levels, a NOELcannot easily be established for such chemicals. Ingeneral, the approach has been to disallow the useof such chemicals in foods. When theiroccurrence in food is unavoidable, either becausethey are naturally occurring or are producedduring processing or cooking, levels should bekept to a minimum. Carcinogenic chemicals forwhich the mechanism is likely to be other thanthrough a genetic change (so-called non-genotoxic carcinogens) can generally be regulated

in a similar manner to other chemicals using aNOEL to establish an ADI.

Food intolerance, including allergenic reactions,can occur to many foods and food components.Satisfactory animal models to predict foodintolerance in humans have not yet beendeveloped, and currently double-blind challengefeeding studies in humans following anelimination diet appear to be the only reasonablyreliable way of identifying factors causing foodintolerance. Food intolerance is restricted to smallsub-populations or individuals and the usualremedy is to provide component labelling of foodto enable sensitive individuals to avoid consumingfoods containing a particular component.

3.5 Microbiological RiskAssessment

There are three models of action wherebypathogenic microorganisms cause disease: the firstis a result of ingesting toxin which is present inthe food as a result of microbiological growth(e.g. Staphylococcus aureus [enterotoxin], Bacilluscereus [emetic toxin]); the second is a result oftoxin formation within the intestinal tract afteringestion of the organism (e.g. Clostridiumperfringens [enterotoxin]); the third is aninfection-type which includes more widespreadsystemic effects (e.g. Listeria monocytogenes).

The International Commission onMicrobiological Specifications for Foods(ICMSF) has grouped the most common andserious of these microbiological hazards into threecategories according to the severity of hazard orseriousness of disease they may cause. Thesecategories include ‘severe hazards’; ‘moderatehazards, potentially extensive spread’; and‘moderate hazards, limited spread’. Foods mayalso be assigned categories of risk based upon thelikelihood that the foodstuff will or will not beinfected from source; whether or not it is able tosupport the growth of the pathogen concerned;whether there is substantial potential for abusivehandling of the food; or whether the food will besubject to a terminal heat process after packagingor before consumption.

170

It is generally recognised that a minimum numberof organisms is necessary to cause illness. This isreferred to as the minimum infective dose orminimum intoxication dose (MID). The MID isinfluenced, however, by a wide variety of host andfood vehicle factors and may vary for differentpopulation groups (such as the young, the elderlyand the immunocompromised), different foodtypes, and different microbiological strains. Onlya limited amount of data is available on therelationship between the number ofmicroorganisms, or amount of toxin, ingested andthe human response so that attempts to quantifymicrobiological hazards are severely limited.Animal models have limited use formicrobiological risk assessment due to the varyingresponses of different species. Human volunteerstudies have some value but, for safety reasons, arelimited to healthy young adults, who are notgenerally the most susceptible to food-bornepathogens. Volunteer studies are also limited bythe variation among microbiological strains andthe effect of food vehicles. Epidemiologicalstudies considered in conjunction with volunteerstudies can provide a more completeunderstanding of the quantitative relationshipbetween exposure to microorganisms and humanresponses. Epidemiological studies generallyinvolve the sub-populations, including high riskgroups, and they involve real food vehicles. Thelimitation is that the number of microorganismsingested by individuals can only be estimated.Outbreaks of food-borne illness can provideinformation on the number of microorganismswhich caused disease in a particular situation, butdo not provide information on the minimum dosenecessary to cause illness.

3.6 Analytical Methodologies3.6.1 FoodFood Standards Code. Latest edition. AustraliaNew Zealand Food Authority. The FSCprescribes AOAC methods for many of itschemical tests and Australian Standard methods(mostly AS1766 Food Microbiology but alsosome AS dairy-specific standards) for most microtests. Where neither is applicable the FSC mostly

describes its own methods. There are howeversome other external references quoted. Forexample there is an FDA method quoted forListeria in poultry.

3.7 Assessment of SummaryStatistic Data

A key source of information is the series ofAustralian Market Basket Survey publicationswhich are published biannually by ANZFA.These provide information about certainsubstances in a range of foods across Australia.

3.8 Censored DataFor contaminants found to contain aconcentration below the Level of Reporting(LOR), the laboratory assigns ‘Not Detected’ or‘Trace Results’. These are given a numerical valuebelow the LOR which results in the ‘calculationof contaminant exposure being overestimated, butit can be considered to be a worst-case scenario.Dietary exposures are also determined using azero value so that a range can be reported forthese contaminants.’ (ANZFA, 1998, p. 15). Inthe 1996 Market Basket Survey, for example, forcadmium the values were (in mg/kg): limit ofreporting, 0.005; trace result, 0.00375; and notdetected, 0.0025.

3.9 Exposure AssessmentNational food consumption data are used indietary exposure assessments and are availablefrom dietary surveys and market basket surveys.Dietary surveys estimate actual food intakes invarious subgroups of the population, taking intoaccount such factors as age and ethnicbackground. Market basket surveys measure theamount of pesticide residues and other selectedcontaminants in freshly prepared and ready-to-eatfoodstuffs. Detailed theoretical daily intakecalculations are carried out when there is apotential for high exposure levels in humans. Theaverage daily food consumption values are used inpredicting pesticide residue intake in comparisonwith the health guideline value (ADI, TDI orTWI). Theoretical daily intake calculations areperformed based on the procedures as outlined in


171


e.g. the Guidelines for Predicting Dietary Intakeof Pesticide Residues (revised) (1997) prepared bythe Global Environment Monitoring - FoodContamination Monitoring and AssessmentProgramme in collaboration with the CodexCommittee on Pesticide Residues, and publishedby the WHO (WHO, 1997). There are formalprograms in Australia to assess human exposureto pesticides and certain contaminants - these arethe National Residue Survey and the AustralianMarket Basket Survey.

3.10 Risk Characterisation3.10.1 Sources of uncertaintyThe principal sources of uncertainty are:

• The hostSome sections of the population are at greaterrisk from exposure to pathogens. These ‘atrisk’ groups include the elderly, the young andthe immunocompromised. Other host factorsthat may affect exposure to a microbiologicalagent include pregnancy, nutritional status,concurrent or recent infections, physiologicalfactors, medication and stress. In most cases,host factors are more important indetermining the severity or outcome of aninfection than in determining the likelihoodof infection.

• The food vehicleThe nature of the food vehicle may influencethe amount of a microorganism needed tocause infection or disease. This includes its fatcontent, iron content, buffering capacity,presence of preservatives, physical state,storage temperature and storage history.

• The level and distribution of microbiologicalcontaminationThe number of microorganisms present in aproduct at the time of sampling formicrobiological analysis may have littlerelation to the number of organisms presentat the time of consumption. The number ofmicroorganisms varies as a result of storage,handling, or preparation. In addition, thedistribution of microbiological agents withina food product may vary due to factors suchas surface contamination and colony

formation.

• The potential for mishandlingMany cases of microbiological food-borneillness result from mishandling in the homeor food service establishment, associated withactions such as under-cooking of foods,allowing cross-contamination of cooked anduncooked food, and storage of foods atincorrect temperature or for excessive periodsof time.

• Terminal heat processesA food subjected to a heat process prior toconsumption is generally regarded as having alower risk although in certain circumstancessuch a food may harbour heat stable toxins.The risk associated with these foods mayincrease if subsequent cooking or handlingprocedures are inadequate.

• Person-to-person transmissionSecondary spread of microorganisms is animportant factor in some types ofmicrobiological food-borne diseases. Infectedpeople may also unsuspectingly contribute tothe spread of an outbreak by contaminationthrough handling or serving food.

• Variety of disease syndromesSome microbiological agents can cause a widerange of disease symptoms. In addition, whilemost microbiological agents involved short-term risks, some food-borne diseases can havelong-term sequelae, such as reactive arthritis.

172

Environmental Health Risk Assessment for Water

Appendix 4

4.1 IntroductionThere is a wide range of water types, water usesand possible routes of transmission of waterbornehazards to humans. In undertaking health riskassessments the characteristics and potential usesof water bodies need to be determined. Watersources include fresh, estuarine, marine and wastewaters. Water uses can include supply of potablewater for drinking and bathing, recreation,aquaculture or irrigation of crops. Humanexposure to waterborne contaminants can include:

• direct exposure through ingestion, dermalcontact, inhalation of aerosols or sprays; or

• indirect exposure through foods contaminatedby irrigation water or water used foraquaculture and seafoods contaminated bywaste water discharges. Health risk associatedwith food contamination via a waterborneroute is within the scope of Appendix 3addressing the risk assessment of food.

4.2 Identifying the IssuesHealth risks associated with various types ofwater are considered in a range of guidelinesdetailed in the following text.

Drinking water is generally the highest and mostdirect source of human exposure to waterbornecontaminants and accordingly it usually receivesthe most attention in water-related health riskassessment. Drinking water quality issues areaddressed in detail within the AustralianDrinking Water Guidelines (NHMRC andARMCANZ, 1996). These guidelines (ADWG)provide an excellent description of water qualitymanagement needs from source water to tap.They also provide detailed fact sheets describingthe rationale and health risk evidence for settingof guideline numbers for all individual qualityparameters currently covered.

The ADWG are undergoing rolling revisions andtheir current status and draft guidelines can befound at www.health.gov.au/nhmrc/advice/water.htm.

Health risks associated with recreational water areconsidered in published guidelines for recreationalwater quality (US EPA, 1999) and in guidelinesbeing prepared by WHO and NHMRC. Healthrisks associated with use or reclaimed water areconsidered in draft guidelines for use of reclaimedwater from sewerage systems (NHMRC,ARMCANZ and ANZECC) which should bepublished in early 2000. In addition annual reviewsof health effects associated with waste waterdisposal and reuse are published in WaterEnvironment Research (e.g. Froese and Bodo 1999)

4.2.1 Drinking waterDrinking water quality is usually categorised interms of physical, chemical or biologicalparameters. For drinking water, the health-relatedphysical parameters are primarily radiological(NHMRC and ARMCANZ, 1996). Turbidity, asa physical parameter in its own right is anaesthetic concern. However, in drinking waterproduced by filtration plants, turbidity is also usedas an indicator of treatment efficiency for theremoval of pathogenic microorganisms (e.g. seeUS EPA, 1998). The chemical parameters ofconcern in drinking water are categorised intoinorganic and organic chemicals. The latter aresub-categorised into disinfection by-products,pesticides and other organic compounds(NHMRC and ARMCANZ, 1996). Thebiological parameters are focused on pathogenicmicroorganisms which are categorised intobacteria, protozoa, toxic algae (cyanobacteria) andviruses. The water quality parameters currentlycovered by the ADWG are listed in Table 1 A4.

The overall relationship among common waterquality parameters can be summarised in a varietyof ways, with one approach presented in Table 2A4 (Hrudey, 1999). This perspective shows howthe physical characteristics of water contaminants(suspended vs. dissolved, volatile vs. non-volatile)relate to their chemical character as well as theirclassification as biological or chemicalcontaminants.


173


Table 1 A4: Parameters covered by the Australian Drinking Water Guidelines (NHMRC and ARMCANZ, 1996)

Micro-organisms Physical Inorganic Organic characteristics chemicals chemicals

174

Bacteria• Aeromonas • Campylobacter • Coliforms • Escherichia coli • Klebsiella • Legionella • Mycobacteria • Pseudomonas aeruginosa • Salmonella• Shigella • Vibrio• Yersinia

Protozoa• Acanthamoeba • Cryptosporidium species • Giardia• Naegleria fowleri

Toxic algae• CyanobacteriaViruses • Adenovirus• Enteroviruses • Hepatitis viruses• Norwalk virus• Rotavirus, para

rotaviruses andreovirus

Radionuclides • Radium-226,

Radium-228 • Radon-222• Uranium • Other beta-gamma-

emitting radioisotopesPhysical parameters • Dissolved oxygen• Hardness• pH • Taste and odour• Temperature • Total dissolved solids• True colour • Turbidity

• Aluminium • Ammonia • Antimony • Arsenic • Barium • Beryllium • Boron • Bromate • Cadmium • Chloride • Chlorine • Chlorine dioxide,

Chlorite• Chlorate • Chromium• Copper • Cyanide • Fluoride • Hydrogen sulfide,

Sulfide• Iodine, iodide• Iron • Lead• Manganese • Mercury • Molybdenum• Monochloramine • Nickel • Nitrate and nitrite • Selenium • Silver • Sodium • Sulfate • Tin • Zinc

Disinfection by-products • MX • Chloroacetic acids • Chloroketones • Chlorophenols • Chloropicrin • Cyanogen chloride • Formaldehyde • Haloacetonitriles • Trichloroacetaldehyde • Trihalomethanes (THMs)

Other organics• Acrylamide • Benzene • Carbon Tetrachloride • Chlorobenzene • Dichlorobenzenes • Dichloroethanes • Dichloroethenes • Dichloromethane • Epichlorohydrin • Ethylbenezene • EDTA • Hexachlorobutadiene • Nitrilotriacetic acid • Organotins • Plasticisers • PAHs • Styrene • Tetrachloroethylene • Toluene • Trichlorobenzenes • 1,1,1-trichloroethane • Trichloroethylene • Vinyl chloride • Xylenes

Pesticides• Aldrin and dieldrin• Atrazine• Chlordane• 2,4-D• DDT and derivatives • Heptachlor, heptachlor

epoxide • Lindane• Plus 113 others

Table 2 A4: Classification of water quality parameters


175

non-infectious - crustaceans - other non-pathogenic invertebrates

infectious - helminths

non-infectious - non-pathogenic - Fe-bacteria - S-bacteria

infectious - pathogenic - protozoa - amoebae - bacteria - viruses

organic inorganic - grit - particulate metals

organic inorganic - clay turbidity - particulate metals

organic - THMs

inorganic - Hydrogen sulfide

organic - geosmin - MIB

inorganic organic - humic acids - amino acids - haloacetic acids

inorganic - soluble - metals - anions

suspended matter

colloidal matter

volatile

viable non-viable - coarse detritus

viable non-viable - fine detritus

Water Contaminants

semi-volatile non-volatile

dissolved matter

Comments:

• Exchange of contaminants between phases isgoverned by a dynamic equilibrium that isdependant on temperature and the relativeconcentration of contaminant in each phase

Operational definitions:

• dissolved—passes microfiltration but not reverseosmosis

• colloidal—not removed by sedimentation or directgranular filtration

• volatile—air strippable• semi-volatile—steam strippable• viable—can replicate under favourable conditions• infectious—can infect a susceptible mammal

representative of humans

Technology specification examples:

• chlorination/ozonation—converts organic matterinto new organic compounds with some oxidised toinorganic products and disinfects by making viableorganisms non-viable

• coagulation—converts colloidal and suspendedmatter so that sedimentation and granular filtrationcan remove suspended and colloidal matter

• granular filtration—removes some fraction ofsuspended and coagulated colloidal matter

(Hrudey, 1999)

(reprinted by permission of S Hrudey)


The management of water quality requiresattention ranging from sound management ofland use and human activities in catchmentsthrough to the application of treatmenttechnology to achieve safe drinking water. Avariety of measures may be considered for theprotection and enhancement of source waterquality including (Reinert and Hroncich, 1990):

• assessment of safe yield in terms of waterquantity;

• catchment land ownership;

• land use controls or management agreements;

• in situ treatment (mixing, aeration, algaecontrol);

• wildlife control;

• forest or agricultural management practices;

• emergency response measures;

• routine sanitary surveys and catchmentinspection;

• access control (e.g. fencing); and

• public education.

The application of risk management principlestowards a total water quality management systemfor Australian drinking water systems is beingevaluated under the rolling revisions to theADWG program. The results of a series of pilotstudies will be developed to include expandedadvice on hazard identification and risk assessmentfor drinking water quality management systems,including principles derived from the HazardAnalysis Critical Control Point (HACCP)approach which has been widely adopted for riskmanagement in the food industry.

Water treatment has evolved from classicaltechnology developed in the early 1900s that wasoriginally based on coagulation, filtration anddisinfection. These basic approaches have beenrefined and improved through the use of anumber of technological alternatives so that amodern water treatment process scheme can be

developed that will satisfy specified finished waterquality requirements given identified source waterquality challenges. The menu of drinking watertreatment process types that are available forconsideration now generally includes (AWWA,1990; Dezuane, 1997):

• air stripping / aeration;

• coagulation processes (destabilisation, mixingand flocculation);

• sedimentation and flotation;

• filtration;

• ion exchange and inorganic adsorption;

• chemical precipitation;

• membrane processes;

• chemical oxidation;

• adsorption of organic compounds; and

• disinfection.

Within each of these process categories there arean expanding range of technologies and operatingstrategies being developed. Research has beendedicated to improving the capability andreliability of these technologies to meet drinkingwater quality criteria that may be specified byhealth risk assessments.

4.2.2 Recreational waterRecreational water quality is usually categorised interms of microbiological parameters althoughphysical features that could represent a hazard tobathers are also considered in some guidelines(NHMRC, 1990) and will be included andprobably expanded in those being prepared byWHO and NHMRC.

As for drinking water the management of waterquality requires attention to sound managementof potential sources of pathogens from adjoiningdrainage areas and catchments. Pathogens can betransported through point sources such as wastewater outfalls or diffuse sources wherecontamination is related to rainfall events.

176

A practical guide to the design andimplementation of assessments and monitoringprogrammes associated with recreational water is:

• Bartram J. and Rees G. (eds) (2000).Monitoring Bathing Waters. E&FN Spon,London.

4.2.3 Reclaimed waterLike drinking water, reclaimed water quality isusually categorised in terms of physical, chemicaland biological parameters. The management ofrisk is usually based on a combination oftreatment, controlled use and controlled exposure.

Treatment processes have not changed a greatdeal and are based on a combination of primary,secondary and tertiary treatment and disinfection.Detention in lagoons is a low technology butrobust method of treating sewage and can be usedto provide primary and/or secondary treatment.Tertiary treatment generally involves filtrationand is required for uses with potential for higherexposures such as residential non-potable use.

4.3 Hazard IdentificationHealth risk assessment for chemical parameterswas first documented in some detail in a series ofpublications of the National Research Council ofthe U.S. National Academy of Sciences (NAS,1977; 1986; 1987; 1989). These expert panelreports, published as separate volumes in theseries Drinking Water and Health addressed theemerging key issues underlying risk assessment ofchemical contaminants in drinking water andprovided a framework which has guided theevolution of health risk assessment of drinkingwater contaminants.

The most common and likely source of humanhealth effects associated with water is throughexposure to microbiological pathogens. The abilityof a pathogen to cause illness is usually wellestablished but in assessing water quality there is atendency to ignore species variability. For example,while Cryptosporidium as a generic group has beenidentified as a cause of waterborne illness only onespecies C. parvum is regarded as causing humaninfections. In addition it is likely that only subtypes of C. parvum are infectious for humans.

Water has been documented as a source of largedisease outbreaks such as the 1993 Milwaukeeoutbreak of Cryptosporidiosis that was estimatedto have infected over 400,000 people (Mackenzieet al, 1994). However, quantitative risk assessmentfor microorganisms has only recently developed incomparison with chemical risk assessment (Haaset al, 1999). Models for microbiological pathogenshave been developed for a few organismsincluding Cryptosporidium, Giardia and sometypes of viruses but the models are limited. Arange of factors that are generally not yetadequately considered include: human variabilityin the form of immune status and partial or totalimmunity through prior exposure, variations invirulence and variations in seriousness of illnessoutcomes. With a few exceptions (e.g. Legionella,Naegleria fowleri) water borne pathogens tend tobe transient contaminants and not free-livingorganisms.

Limited quantitative microbiological riskassessment has also been undertaken for use ofreclaimed water. As for drinking water, modelshave been developed for Cryptosporidium, Giardiaand some types of viruses.

Drinking water health risk assessment provides acompelling example of risk tradeoffs becauseacute microbiological disease is almost certain toarise with surface water supplies that are notsubjected to disinfection (Singer, 1999). However,since the discovery in 1974 that chloroform andother trihalomethanes are produced as by-products of chlorine disinfection, there has been aphenomenal growth in the identification ofdisinfection by-products (DBPs), as summarisedin the organic DBP section of Table 1 A4. As aresult there have been numerous epidemiologicalstudies, some of which suggest possible linksbetween DBPs and adverse health effects rangingfrom bladder cancer to adverse reproductiveoutcomes. Maintaining a sensible balancebetween the known infectious disease risks thatcan be controlled by disinfection and thehypothesised health effects associated with DBPshas presented the drinking water industry with asubstantial challenge in the assessment and trade-off of competing risks (Craun, 1993). Issuesassociated with potential conflicts between


177


compliance with requirements for microbiologicaland DBP control are discussed in a US EPAGuidance Manual (1999).(www.epa.gov/safewater/mdbp/mdbptg.html)

Drinking water sources or recreational water aremost likely to be polluted with human, animal,agricultural and industrial wastes. The hazardsassociated with human and animal wastes arepredominantly microbiological while thoseassociated with industrial wastes are generallychemical. Agricultural wastes can bemicrobiological (animal) or chemical (fertilisers,

pesticides). In addition pollution can exertsecondary effects through the support of increasedgrowth of naturally occurring organisms such ascyanobacteria which may pose health risks fromboth drinking water and recreational water useperspectives (Chorus and Bartram, 1999). Thisrecent World Health Organization monograph,prepared with substantial input from Australianexperts, provides an excellent illustration of apractical approach to hazard identification andpreliminary risk assessment for cyanobacterialhazards to drinking water supplies. This issummarised as a generic approach in Figure 1 A4.

178

Figure 1 A4: A generic rationale for hazard identification and preliminary risk assessment for drinking water health risks

(adapted from Bartram et al, 1999)

Hazard (Potential to cause harm)

Controls on Exposure to

Hazard

Low

No Barriers

Single Barrier

Multiple Barriers

Advanced Multiple Barriers

Medium High

Risk

(Like

lihoo

d of c

ausin

g har

m)

low

high

A similar approach is likely to be included inguidelines being prepared by NHMRC andWHO for recreational water.

4.4 Multiple Barrier Approachto Reduce Contaminationand Health Risks

The provision of barriers to the transmission ofpathogens and contaminants is important inreducing health risks associated with water. Themultiple barrier approach relies on the concept ofusing more than one type of protection ortreatment in a series in a water treatment processto control contamination and provide overallprocess reliability, redundancy and performance.

An example of the multiple barrier approach toprotect drinking water occurs in normalcatchment-to-tap management. The barriersinclude the following:

• Protection of source water fromcontamination with an active catchmentprotection program;

• Long detention times within reservoirs(weeks to months);

• Water treatment e.g. coagulation, settling andfiltration;

• Finished water to be disinfected before itenters the distribution system;

• Maintenance of an adequate disinfectionresidual throughout the distribution system;and

• Maintenance of the integrity of thedistribution system i.e. no breaks in the pipes,roofs on water tanks etc.

• Monitoring for microbiological quality shouldbe regarded as a check that the barriers aremaintained.

4.5 Monitoring MethodologiesThe most widely accessible comprehensivereference for techniques of water analysis andsampling is the publication, ‘Standard Methods forthe Examination of Water and Waste water’

(Clesceri et al, 1998). Explicit guidance on thefrequency of monitoring for parameters that arecovered by the ADWG has been provided in theADWG document (NHMRC and ARMCANZ,1996). Water quality monitoring has long beenbased on indicator or surrogate parameters torepresent agents of health concern. For example, thepresence of microbiological pathogens, which havebeen impractical to monitor directly, has beeninferred by indicator bacteria such as the total orthermotolerant coliform bacteria. The presence ofindicator organisms has been taken as a sign thatwater quality may have been compromised with thepossible presence of enteric pathogens. Confidencein the reliability of these indicator organisms hasbeen undermined by the finding of protozoanpathogens like Cryptosporidium and Giardia speciesthat are substantially more resistant to chemicaldisinfection than the indicator organisms. Thisreality means that inactivating the indicatororganisms by disinfection does not assureinactivation of the pathogens. These circumstancesare further complicated by the lack of reliablemethods to monitor for viable and infective strainsof the resistant pathogens so that the relevance ofnon-specific monitoring data remains a challenge.

In the case of chemical contaminants, someparameters like the trihalomethanes (THMs) maybe only indicator or surrogate measures for otherchemical agents that may or may not pose healthrisks. There are also issues about the chemicalspecies that are measured in water qualitymonitoring. For example, arsenic toxicity variesover a thousand fold depending on the chemicalform of the arsenic. The more common inorganicforms of arsenate or arsenite that are most likelyto be present in drinking water are believed topose the greatest health concerns.

Guidance on sampling water has also beenprovided by Keith (1988; 1991; 1992). Waterbodies are often not homogeneous mixtures and anumber of issues need to be addressed indesigning sampling programs includingdifferences between stream flows andembayments, water depth, stratification and theimpacts of silts and sediments as sources ofmicrobiological and chemical contaminants.


179


4.6 Assessment of SummaryStatistics and Presentationof Data

The appropriate format for presenting waterquality data depends on the nature of the hazard.Microbiological hazards generally pose an acuterisk so that short term monitoring is needed andtransient excursions above guideline levels canpose the danger of waterborne diseasetransmission. Because real time (instantaneous)monitoring of microbiological parameters iscurrently not possible, factors which can bemonitored frequently or continuously, likedisinfectant residual and turbidity, are often usedto document treatment performance and therebyinfer acceptable microbiological control. Chemicalparameters that pose a chronic risk, such assuspected carcinogens, are usually judged inrelation to standards based on lifetime exposure.For these parameters, long-term average exposureis generally considered for dose, although it isusually expressed on a daily basis (i.e. as anaverage daily dose for a lifetime). The more recentinterest in the possibility of adverse reproductiveoutcomes associated with various water qualityparameters has changed this perspective makingshort-term (possibly even peak) exposures morerelevant.

4.7 Censored Data and Levels of Reporting

As a general guide, reporting to a sensitivity ofone tenth of the guideline level is preferred butmay not be practicable for some substances, suchas pesticides, where the guidelines have been setat a level of detection. Reporting levels need to beset sufficiently low so as to be able to distinguishparameter trends from background levels. Likemost environmental data, water quality data areoften highly skewed because sub-detection valuescannot exist so that data sets are truncated at thedetection limit. Often, log normal distributionsmay fit the data, unless a few extreme values skewthe data more than a log normal distribution willreadily accommodate.

4.8 Dose–Response AssessmentThere are difficulties associated withdose–response assessments for both chemical andmicrobiological contaminants. Microbiologicaldose–response assessments are faced with thedifficulty of considering human andmicrobiological variation. The likelihood ofcontracting an infection is influenced by factorssuch as immune status, immunity imparted byprevious exposure and virulence of the specifictype or strain of microorganism.

Measurement of chemical parameters in water isgenerally well developed. However, evaluatingcausal linkages and dose–response relationshipsbetween estimated doses and disease generallyinvolves a great deal of uncertainty. Predictionsare often based on extrapolation of high doseanimal toxicology data to chronic low levelexposures in humans. Thus, uncertainty arisesfrom both interspecies extrapolation and high tolow dose extrapolation. For chronic diseases likecancer, the causal linkages must be inferred fromobservational epidemiology studies for whichdrinking water contaminant exposures mustusually be reconstructed from limited historicaldata. These realities raise considerable uncertaintyabout actual exposure levels, as well as theuncertainty arising from bias and confounding onthe estimation of relative risk.

4.9 Exposure AssessmentThere is a tendency to simplify exposureassessment by focusing on ingestion ofstandardised volumes of water. However, exposureto drinking water and recreational watercontaminants can occur through ingestion,inhalation and dermal contact. For example, withvolatile and lipophilic organic contaminants (e.g.THMs) doses from showering and bathing maybe higher than via ingestion (Weisel and Jo,1996). The scope and complexity of drinkingwater contaminant exposure assessment hasrecently been addressed in some detail (Olin,1999) and is summarised in Figure 2 A4.

180

Figure 2A4 removed from electronic version due to copyright restrictions.If required, it is available in the hard copy of the document – page 181.


In determining guideline values for drinkingwater a standard daily consumption is used whichin Australia has been judged to be 2 litres per dayfor an adult. However, there is variation inconsumption and individuals may derive drinkingwater from a number of sources including tapwater at home and work, bottled water andrainwater. More detailed exposure evaluationsmust take into account the full range ofcontaminant exposure routes.

In recreational guidelines an estimated maximumingestion of 100 mL per recreational session is usedfor both marine and fresh waters (NHMRC, 1990).

In regard to reclaimed water maximum exposureshave been calculated for a number of usesincluding irrigation of edible crops (10mL),irrigation of public areas, golf courses etc (1mL)(Asano et al, 1992).

Determination of potential doses in terms of theconcentration of contaminants in water is betterdeveloped for chemicals than for pathogens. Forsome chemicals (e.g. some pesticides), there maybe health concerns at concentrations near routinedetection limits. It is possible that these limits arehigher than for other chemicals which can bedetected in parts per billion or even lowerconcentrations. However, for manymicroorganisms there are no reliable or sensitivequantification methods and in some casesmethods have not been developed to identifyspecies or strains that can cause human infections.

4.10 Risk CharacterisationThe characterisation of risk that is inherent insetting drinking water guideline levels can begenerally summarised according to:

Where:

Guideline Level = the guideline concentration ofcontaminant in water

RL = the reference toxicity level (often a no effect level)

BM = the body mass,often 70 kg for an adult

AF = what proportion of total exposure can be attributed to drinking water

ED = exposure duration (if exposure is less than continuous)

IR = an ingestion rate,often taken as 2L per day

UF = uncertainty factors applied to reduce the RL

AT = an averaging time of exposure,will equal ED if exposure is continuous

Health-related criteria have been established for awide range of chemical contaminants in waterand initial comparison of estimated exposurelevels should be with the ADWG (NHMRC andARMCANZ, 1996). An additional source ofinformation is provided by the US EPA DrinkingWater Health Advisories within the IntegratedRisk Information System (IRIS) which can befound at: www.epa.gov/ngispgm3/iris/dwater.html.These Health Advisories provide data fordrinking water exposures for up to one day, 14days, 7 years and lifetime exposures.

182

Guideline Level =RL x BM x AF x ED

IR x UF x AT

Australian Models of Risk Assessment

Appendix 5

A range of models are used by State, Territoryand Federal agencies in Australia for regulatory,administrative and investigation purposes.Approaches dealing with contaminated soil, air,food and water have been identified inAppendices 1–4.

5.1 Chemical risk assessmentSeveral bodies are involved in the process forassessing chemical risk assessment. Nationalchemicals legislation and responsible authoritiesare outlined in Table 1 A5.


183

Table 1 A5: Chemical risk assessment in Australia

Industrial Agricultural Medicine Food additives,chemicals and veterinary and medical contaminants,

chemicals products processingand toxins

Agency National Industrial National Registration Therapeutic Goods Australia New ZealandChemicals Notification Authority (NRA) Administration Food Authorityand Assessment for Agricultural and (TGA)Scheme (NICNAS) Veterinary Chemicalswithin TGA

Ministry Health and Ageing Agriculture, Fisheries Health and Ageing Health and Ageingand Forestry

Scope Assessment only, not Assessment and Product Assessment and Assessment and Productregistration based Registration Product Registration Registration

Relevant Industrial Chemicals Agricultural and Veterinary Therapeutic Goods Australia New Zealandlegislation (Notification and Chemicals (Code) Act Act 1989 Food Authority Act 1994

Assessment) Act 1989 1994, Agricultural and Food Standards CodeVeterinary Chemicals Administration Act 1994

About the Industrial chemicals Agricultural products Therapeutic goods Chemicals are added tochemicals are varied and cover, include chemicals which included prescription food for a number of

for example, dyes, generally destroy/repel and non-prescription reasons, for instance assolvents, adhesives, pests or plants.Veterinary (OTC) medicines. a processing agent,plastics, laboratory products are used to OTCs included preservation or as achemicals, paints, prevent, diagnose or complementary flavouring or colouring.as well as chemicals treat disease in animals medicines (herbal, These are known as foodused in cleaning (Toxicology and public vitamins, minerals additives.products and cosmetics health OH&S and and homeopathicand toiletries environmental assessments preparations), and

conducted for the NRA by some sterilants the Chemicals Unit of the and disinfectants.TGA, NOHSC, and Environment Australia respectively).


5.2 Occupational6 RiskAssessment

The following processes are adopted by theChemical Assessment Division of NOHSC.

5.2.1 NICNAS (existing chemicals)

Data requirements

For existing industrial chemicals, the datarequirements depend on the assessment type.A standard dataset for a full Priority ExistingChemical (PEC) comprises informationconfirming the identity of the chemical, thephysicochemical properties and use of thechemical (including import/manufacturevolumes), all availabletoxicological/epidemiological data, detailedexposure information for workers, the public andthe environment and risk management initiatives.The toxicological package includes availablehuman, animal and in vitro data and ecotoxicityand biodegradability/fate data for theenvironmental assessment.

The dataset for a preliminary PEC may or maynot include a detailed toxicological package ordetailed exposure data. Risk assessment, in termsof a formal risk characterisation for specific uses isnot carried out for preliminary assessments.

The dataset for a secondary notificationassessment is determined in accordance with a setof circumstances (criteria) as set out in the ICNAAct. Should these circumstances require a re-evaluation of the risk(s) assessed in the originalPEC report, then a formal risk characterisation isusually carried out.

Exposure data

Occupational exposure data is provided asstatutory obligation (under the ICNA Act) fromapplicants (for assessment). This information issupplemented from literature review, site visits,international reports (e.g. OECD SIARs) andwhere data is lacking from modelling. The model

that has been used to date is the UK HSE EASEmodel, which provides estimates of airborne anddermal exposure for different occupationalscenarios.

Where exposure by inhalation is the major routeof exposure, and the toxicological databaseincludes good quality inhalation data (human oranimal), the common practice is to use ‘external’exposure data (i.e. not to attempt to extrapolate to‘internal’ dose) in the risk characterisation process(see below). When ‘external’ exposure data areused/determined, no adjustment is made toaccount for reduced personal exposures resultingfrom the use of personal protective equipment(e.g. respiratory protection, gloves etc.). However,where mechanical ventilation is installed, this canbe factored into the EASE model, should suitablemonitoring data (i.e. measured when ventilationhas been installed and is operational) not beavailable. The quality of the monitoring datashould also be a factor considered in the riskcharacterisation and exposure standard settingprocesses (see below).

Where dermal exposure is an important route ofexposure and/or where the toxicological databasedoes not provide an inhalation study, internal(dose) exposure may be estimated, utilising theavailable pharmacokinetic data, and used in therisk characterisation process.

5.2.2 Toxicological dataToxicological and epidemiological/casestudy/clinical data is also provided as statutoryobligation (under the ICNA Act) from applicants(for assessment). This data is supplemented fromliterature review and international reports (e.g.OECD SIARs, IPCS, IARC, ECETOC).

Currently, available toxicological andepidemiological data are evaluated in conjunctionwith available pharmacokinetic data, to estimatethe critical NOAEL or, if not determined, theLOAEL for both acute and chronic exposures foreach relevant route of exposure (i.e. oral, dermal

184

6 Under NICNAS, environmental and public health risk assessment (RA) is carried out by EA and TGA,respectively. Although there are differences in exposure calculation methods, the methodologies currentlyadopted for public health and environmental risk characterisations (i.e. NOAEL/Estimated human exposureratio and PNEC/PEC ratio) are consistent with the margin of exposure (MOE)—also referred to as margin ofsafety (MOS)—approach adopted for OHS risk characterisation.

and inhalation). The health hazards for each end-point are classified in accordance with theNOHSC Approved Criteria for ClassifyingHazardous Substances.

The quality of the toxicological database shouldbe a factor considered in the risk characterisationand exposure standard setting processes (seebelow).

5.2.3 Risk characterisationThe current methodology utilised by NICNAS isthe ‘margin of exposure’ (MOE) approach.

In deriving the MOE, direct comparison is madeof the critical NOAEL with themeasured/estimated exposures for eachoccupational scenario of relevance to manufactureand use in Australia. This is carried out separatelyfor inhalation and dermal exposure (whererelevant) i.e. by using NOAELs derivedspecifically from each route of exposure.

Where exposures may be significant by bothroutes, the combined estimated internal dose maybe used. In this case, the oral NOAEL (for thecritical effect) is usually considered moreappropriate NOAEL for deriving the MOE.

The resulting MOE, is then evaluated (for eachroute), taking into account the quality of theavailable database (e.g. whether derived fromhuman data, uncertainties in the database etc.)and nature/severity of effect (e.g. carcinogen,sensitiser etc.). No specific values are assigned tocomponent uncertainty factors (this is usually partof the exposure standard setting process carriedout by NOHSC—see below). However, the riskcharacterisation process takes these uncertainties(NB these are identified in the report) intoaccount in evaluating the adequacy of the MOE.

Based on the magnitude of the MOE, currentrisk management initiatives are assessed andwhere found inadequate, recommendations foradditional exposure reduction measures (controls)or other risk management initiatives arepromulgated. Recommendations may includeregulatory action by NOHSC or other Agencies(e.g. TGA and EA, where public health and

environmental risks have been identified).Recommendations to NOHSC may include: thesetting of an occupational exposure standard (seebelow), review of an existing exposure standard(see below), scheduling of a substance inaccordance with the model regulations for controlof workplace hazardous substances and, as a lastresort, phase out of use and manufacture.

5.2.4 NICNAS (new chemicals)For new industrial chemicals, the datarequirements depend on the notification categoryand are stipulated under the ICNA Act. Astandard dataset comprises informationconfirming the identity of the chemical, thephysicochemical properties and use of thechemical, detailed exposure information abouthow workers, the public and the environment areexposed to the chemical, and a standardtoxicological package. The toxicological packageincludes animal and in vitro data for the humanhealth assessment and ecotoxicity andbiodegradability data for the environmentalassessment.

Exposure assessment

The occupational exposure assessment isconducted by establishing the use pattern of thechemical and identifying the sources ofoccupational exposure. Exposure is then estimatedby taking into account the routes of exposure, thefrequency and duration of exposure, and measuredworker data, for example, atmospheric and/orbiological monitoring results. Information isneeded for each of the scenarios where workersare potentially exposed to the chemical.

For new industrial chemicals, the occupationalexposure assessment is usually qualitative, asmeasured data is unlikely to be available and thereis insufficient information available to predictreliable quantitative estimates. Modelling, forexample, using EASE, is occasionally used.

Toxicological assessment

Both human and experimental animal data areassessed in accordance with internationalguidelines to identify the critical health effects of


185


the chemical and to determine the dose–responserelationship, with no observed adverse effect levels(NOAELs) established wherever possible. Fornew industrial chemicals, human data is usuallynot available. The health hazards of the chemicalare classified in accordance with the NOHSCApproved Criteria for Classifying HazardousSubstances.

For new chemicals, the toxicological database mayconsist of studies which have been performedwith a structural analogue of the notifiedchemical, or with a formulation. Adequacy andapplicability of the data will be taken into accountwhen performing the assessment. Where datagaps exist, or where toxicological data have notbeen provided, as with some classes of polymer,the toxicological hazard may be predicted fromthe chemical’s physical properties or thecharacteristics of structurally related chemicals,given that factors such as volatility, solubility andmolecular weight can indicate the likely extent ofabsorption across biological membranes.

Risk characterisation

The health risk to workers is characterised byintegrating the occupational exposure andtoxicological assessments. For brief or short-termexposures, human data and information fromacute toxicity studies in animals are taken intoaccount to determine the risk of adverse healtheffects such as acute respiratory effects and skinirritation. For longer term and repeated exposures,the health risk to workers is characterised byfirstly comparing exposure estimates withNOAELs to give a margin of exposure (MOE),and then deciding whether there is cause forconcern.

Matters taken into account when characterisingthe risk, include the uncertainty arising from thevariability in the experimental data and inter- andintra-species variation, the nature and severity ofthe health effect and its relevance to humans andthe reliability of the exposure information.

Where it is not possible to determine a NOAEL orLOAEL, for example, from lack of suitable data, therisk is evaluated on the basis of qualitative or

quantitative exposure relevant to the group ofworkers being considered. For new chemicals, amore qualitative characterisation takes place asexposures are often unknown or more difficult topredict.

5.2.5 NOHSC (agricultural andveterinary chemicals)

The Agvet Section conducts OHS riskassessments on behalf of the NRA, under twoassessment programs, namely ‘ProductRegistration, and ‘Chemicals Review’.

Data requirements

The Agricultural and Veterinary Chemicals CodeAct (1994) makes provision for the evaluation,registration and control of Agvet chemicalproducts. Data required for the OHS assessmentof Agvet chemical products include:

• use pattern of the product;

• formulation composition of product;

• physicochemical properties of the activeconstituent and product;

• toxicology of the active constituent andproduct; and

• exposure data.

Exposure data

It is a requirement under the Act that all availableexposure data and adverse incident reports mustbe provided to the NRA by applicants (forassessment). Exposure data may covermanufacture/formulation of Agvet products andend use situations. Exposure data provided byapplicants is supplemented from literature review,international reports (e.g. US EPA, UK MAFF),field/site visits and modelling. The model used todate is the UK Predictive Operator ExposureModel (POEM). Occasionally, exposure datafrom the US Pesticide Handlers ExposureDatabase (PHED) has been used whereapplicants provide subset exposure data.

The exposure assessment constitutesconsideration of the use pattern of the product,

186

identification of potential exposure scenarios andpredominant route(s) of exposure in each case.For Agvet chemicals, it is generally accepted thatskin contamination is the predominant route ofexposure. In general, inhalation exposurecomprises only a small proportion of totalexposure, except when the product is applied inan enclosed space (e.g. fumigants). Therefore,where the toxicological database includes dermaldosing studies of appropriate quality andduration, ‘external’ exposure data are used in therisk characterisation process.

In the absence of appropriate dermal studies,internal dose is estimated using external dermalexposure data corrected for dermal absorption.Absorption is estimated using in vitro and/or in vivo percutaneous absorption data. In theabsence of chemical specific data, analogue datamay be used, where available. Total body burdenis determined by integrating exposure frominhalation and dermal routes and comparing theresult with systemic effects data to ascertainpotential health risk.

Where biomonitoring data are available, abiological monitoring approach may be used, asabsorbed dose data interpreted with the aid ofpharmacokinetic data are likely to be moreaccurate than the estimation of internal dosegiven by exposure data corrected for dermal andrespiratory penetration.

Pesticide exposure assessments also take intoconsideration the protection afforded by labelspecified protective equipment. Default protectionfactors are utilised in the absence of specific data.

Toxicological data

Toxicological and epidemiological/case study dataprovided by applicants are evaluated by theTherapeutic Goods Administration (TGA). TheTGA evaluation is considered in order todetermine relevant endpoint(s) andNOEL/LOEL(s) for use in the OHS riskassessment. The selection is based on factorsincluding: quality of the database, frequency ofuse of the product, health significance of the

endpoint(s) and predominant route of exposure.

For new Agvet chemicals, the health hazards ofthe chemical are classified in accordance with theNOHSC Approved Criteria for ClassifyingHazardous Substances.

Risk characterisation

The risk assessment takes into consideration thehazard of the chemical and the potential foroccupational exposure. In general, an end use riskassessment is conducted for Agvet products.Potential exposure is determined by the usepattern of the product and currentagricultural/animal husbandry practices (includingexisting exposure mitigation methods such asprotective equipment and engineering controls).

As for industrial chemicals (NICNAS—ExistingChemicals), Agvet assessments currently utilisethe ‘margin of exposure’ (MOE) approach. Thebenchmark MOE is determined on a case by casebasis, following consideration of the quality of thedatabase, nature and severity of the health effectand known variability in human metabolism ofthe chemical. In general, a 10 fold factor isconsidered appropriate to account for interspeciesextrapolation and a similar factor (10x) forintraspecies variability.

Current exposure mitigation methods areevaluated quantitatively, where possible. In theabsence of data or models, qualitative assessmentsare conducted based on generalised informationabout the use pattern and ‘scientific judgement’.Where current exposure assessment methods arefound to result in unacceptable risk, additionalexposure and risk reduction methods may berecommended.

OHS recommendations on regulatory action mayinclude: restrictions on use of the chemical,exposure mitigation methods in accordance withthe hierarchy of controls under HazardousSubstances legislation or review of an existingexposure standard.


187


5.3 Exposure Standards(NOHSC)

5.3.1 Statutory process outlineNational exposure standards for atmosphericcontaminants in the occupational environment aredeclared by the National Occupational Healthand Safety Commission (NOHSC) as guidelinesto be used in the control of occupational healthhazards [NOHSC:1003 (1995)].

The national exposure standards are not dividinglines between safe and dangerous concentrationsof chemicals, neither are they a measure ofrelative toxicity. Appropriately qualified andexperienced persons should undertakeinterpretation of the national exposure standards.

The national exposure standards are referenced inthe NOHSC National Model Regulations for theControl of Workplace Substances [NOHSC:1005(1994)], under regulation 12(4) related toemployers’ duties and the control of exposureprovisions.

Enactment by the Commonwealth, State andTerritory governments of uniform hazardoussubstance legislation, based on the national modelregulations, places the national exposure standardsin a regulatory context across all jurisdictions ofAustralia.

5.3.2 Data requirementsIn 1999, the National Commission approved anew methodology for reviewing and updating thenational exposure standards which maximises theuse and acceptance of overseas standards fromgovernment and non-government sources, andminimises the need to develop standards inAustralia.

Primary sources include the United KingdomHealth and Safety Executive HSE ‘OccupationalExposure Limits’, the American Conference ofGovernmental Industrial Hygienists (ACGIH)Threshold Limit Values, the German Government‘MAK’ values and European Union standards.

These sources were selected after an assessmentagainst several factors including quality and

availability of supporting documentation, integrityof the development process and consistency withthe NOHSC philosophy of transparency,development through consultation, and scientificrobustness.

5.4 Use of Toxicological/Exposure Database

Process is designed to minimise reassessment oftoxicological data, rather relying on integrity ofdevelopment process and inherent sound scienceof overseas systems.

Systems chosen represent an appropriateevaluation of available toxicological andepidemiological data sources by acknowledgedagencies.

Where de novo standards are to be developed, theNOHSC Hazardous Substances Subcommitteewill determine the broad parameters of the datarequirements for any review, without limiting therange of toxicological data to be used by, forexample, a consultant reviewing a nationalexposure standard.

NICNAS PEC Reports, supportingdocumentation from HSE, ACGIH and MAK,overseas scientific publications (ECETOCReports, IPCS Environmental Health Criteria,CICAD reports, IARC Monographs, CriteriaDocuments from the Nordic Expert Group), areregarded as appropriate sources for riskassessment consideration in any review, and willcontinue to be employed in this regard.

5.5 Factors Considered inStandard Setting

Health-based standards are set on the basis of oneor more critical endpoints (e.g. carcinogenicity,irritation). Where available, NOAEL levels forthe critical effect can be used as the basis of thestandard. Arbitrary safety factors (margins ofsafety) may be applied to the standard, andexamples of 33–50 per cent of the value of anobservable effect level (animal or human data)have been recommended on occasion.

188

All national exposure standards are subject to animpact analysis, the depth of which varies withthe estimated impact of the proposed change tothe standard, following a review. From aqualitative impact analysis of identifying whereadditional costs and benefits would be borne as aresult of a change in the standard, through to afull economic impact analysis, NOHSC requiresall changes to include this consideration.

5.6 Role of the Chemicals Unitof the TGA in PublicHealth Risk Assessment

The ‘Chemicals Unit’ which is part of theChemicals and Non-Prescription MedicinesBranch (CNPMB) of the TGA, exists as aprofessional scientific group to provide advice tothe Minister, to the Scientific Director of theUnit, to appropriate Committees of State andCommonwealth Government agencies, and to thepublic, on possible risks to health associated withexposure to chemicals in the environment. Theseinclude agricultural chemicals, veterinary drugs,industrial chemicals and other chemicals whichmay have an impact on public health. The Unit ismade up of two sections, the Chemical ProductsAssessment Section (CPAS) and the ChemicalReview and International Harmonisation Section(CRIHS), together with the Scientific Director ofthe Branch and the secretariat of the AdvisoryCommittee on Pesticides and Health (ACPH).

5.6.1 Chemical products assessmentThe main function of CPAS is to assess thetoxicology and public health aspects ofapplications for registration of new agriculturaland veterinary chemicals. It also has an importantrole in assessing the public health aspects ofnotifications for new industrial chemicals and ofre-assessment of existing industrial chemicalsidentified for priority reviews.

CPAS traces its origin back to 1984 when thethen Toxicology Evaluation Section (TES) wascreated following a Senate enquiry into hazardouschemicals, which noted the increasing use ofchemicals in the environment. Increasing publicconcern and media attention to chemical exposure

demanded greater accountability from thechemical industry and from governmentregulators. In addition, chemical residues inexport produce (e.g. beef, wheat, dairy goods)have important implications for trade; Australiamust be able to ensure standards of chemicalregulation acceptable to international markets, aswell as to domestic consumers. Inacknowledgment of public health and tradeconcerns, chemicals regulation as an area of publicpolicy has developed, recognising that thenumbers of chemicals requiring assessment andthe complexity of the assessment process haveincreased. The TES formed an integral part ofthis process, providing scientific advice to enableappropriate regulation of chemicals in order tosafeguard public health. In 1996, the role of thesection was broadened and the TES was renamedthe Chemical Products Assessment Section(CPAS), with a particular focus on assessing.agricultural and veterinary chemicals, as part ofthe National Registration Scheme (NRS) whichis managed by the National RegistrationAuthority for Agricultural and VeterinaryChemicals (the NRA).

Through the Scientific Director,recommendations on individual chemicals areprovided to the NRA and form an importantcomponent of the decisions by the NRA in theregistration of chemicals. Reports includescheduling recommendations made by theNational Drugs and Poisons Schedule Committee(NDPSC) and may include additional toxicologyadvice provided by the Advisory Committee onPesticides and Health (ACPH). (Therecommendations of NDPSC are formallyincorporated into the ‘Standard for the UniformScheduling of Drugs and Poisons’ (SUSDP)which forms the basis for national uniformity indrugs and poisons scheduling.)

The public health implications associated withthe use of industrial chemicals are also assessed byCPAS staff, in accordance with the provisions ofthe National Industrial Chemicals (Notificationand Assessment) Act. Advice is provided toNICNAS for eventual incorporation into aconsolidated assessment report on occupationalhealth, human health, and environmental effects.


189


Additionally, CPAS may be requested to assessthe toxicological hazards associated with a rangeof natural and synthetic chemicals. Examples ofthese include vitamins, herbs, cosmeticingredients, certain consumer products, and someenvironmental contaminants.

Evaluation Reports: Reports are structured toallow for ready access to the main points arisingfrom the assessment. They are written to providesufficiently detailed information for the reader toform an independent conclusion and aim toobviate the need, during a subsequent review ofthe chemical (or product), to refer back to theoriginal study data.

Reports include:

• Submission Summary—briefly outlines theresults from all studies accompanying theapplication/submission and includes adiscussion of the important findings andappropriate recommendations.

• Main Body of the Report—contains detailedoutlines of the studies conducted with thechemical, including methodology, the extentof monitoring for biological changes, alltreatment-related effects and any otherobservations or information which may bepertinent to the assessment of thesignificance of the findings.

• Consolidated Summary—contains theintegrated summaries of study results fromprevious submissions relating to the particularactive ingredient, plus the newly evaluated data.

• Confidential Business Information—impurity profiles, product ingredients andinformation on additives in formulations etcare included in removable appendices at theend of the report.

In the Main Body of assessment reports onagricultural and veterinary chemicals, thefollowing study types are assessed:

• Toxicokinetics and Metabolism—studies onthe fate of the chemical in laboratory animals.

• Acute Studies—single dose toxicity studies,irritation and sensitisation studies on theactive ingredient and on formulated productscontaining the active ingredient.

• Short-Term Studies—administration ofmultiple doses to test species for less than 90 days.

• Subchronic Studies—duration of dosing atleast 90 days and less than 12 months.

• Chronic/Carcinogenicity Studies—administration for 12 months or longer.Carcinogenicity studies involveadministration for the major portion of theanimal’s lifespan.

• Reproduction Studies—administration priorto, during, and following mating andpregnancy for one or more generations.

• Developmental Studies—administration topregnant animals during the period of majororganogenesis.

• Genotoxicity Studies—studies of the effectson genetic material.

• Other Studies—includes neurotoxicity,immunotoxicity, studies on humans, andtoxicity studies on degradation products andimpurities.

5.6.2 Chemical review andinternational harmonisation(CRIHS)

Until the mid-1990s, there was no formalprogram to review ‘old’ pesticides and veterinarydrugs in Australia. However, two programsprovided an informal mechanism for reviewing anumber of aspects of chemicals safety. Firstly,a review process for individual chemicals occurredon the suspicion of human health and/orenvironmental concerns, or followinginternational regulatory action(s); in practice thisusually involved reviewing limited extra datarelated to the particular issue of concern ratherthan conducting a comprehensive review.Secondly, the Technical Grade Active Constituent

190

(TGAC) Scheme introduced in 1985, whiledesigned primarily to identify the source andascertain the quality of technical grade materialsused for formulating end-use-products (EUPs)used in Australia, had significant elements of areview program in that mammalian toxicologydata and environmental data were collected andreviewed. As a result of the TGAC scheme,toxicology data on over 400 pesticides werereviewed with respect to public healthconsiderations over a 6–7 year period. As aconsequence of the above two programs, Australiawas well placed to develop further more formalreview arrangements.

Australia introduced a formal program to reviewexisting agricultural and veterinary chemicals in1994 under the title of the Existing ChemicalsReview Program (ECRP), managed by the NRA.The program carries out systematic reviews ofexisting agricultural and veterinary chemicals on apriority basis. This program is one of a number ofinitiatives arising from a 1990 senate enquiry intoaspects of the legislative, administrative andregulatory procedures for agricultural andveterinary chemicals. The ECRP stemmed largelyfrom the fact that many registered chemicalsentered the market place based on criteria nowrecognised as outdated by today’s regulatorystandards. The ECRP involves cooperativearrangements between the Chemicals Unit(public health), Environment Australia (EA—environment), the National Occupational Healthand Safety Commission (NOHSC—occupationalhealth and safety) and the NRA (chemistry,efficacy and agricultural issues, residues, andregistration).

The goal of the ECRP is to ensure thatagricultural and veterinary chemicals in use inAustralia can be used safely and effectively. Theprogram operates according to the principles ofopenness, fairness and consistency with regard topublic consultation, selection of chemicals forreview, and standards of assessment. All aspects ofa chemical (public health, OH&S, environmental,efficacy, and animal and crop safety) areconsidered in a review. Thus, the ECRP has beenimplemented to:

• ensure that the chemicals remain safe andeffective when used according to labelinstructions by specifically consideringtoxicity and exposure patterns in relation topublic health, occupational health and safety(OH&S); and environmental controlmechanisms; known and potentialenvironmental impacts; efficacy; safety issuesin relation to target species (animal and crop);management options to reduce identifiedrisks;

• maintain the protection of Australian tradeand commerce in agricultural produce andlivestock;

• address community concerns and generalinterest in agricultural and veterinarychemicals by providing information to thepublic on the use of chemicals and theirenvironmental, public health and OH&Saspects; and

• consider public nomination of chemicals forreview.

Agricultural and veterinary chemicals are selectedfor review on the basis of agreed criteria includingtheir potential health and environment hazard(s),exposure potential, age and adequacy of thedatabase, efficacy, international regulatory actions,and trade and other agricultural implications. TheECRP chemical selection process alsoincorporates a mechanism for public nominationsof chemicals.

The toxicology and public health aspects ofECRP reviews are assessed by staff of theCRIHS. The Section provides toxicological andchemicals policy advice, as required, to theScientific Director of the CNPMB, appropriateCommittees of State and CommonwealthGovernment agencies, and to the public. It alsoprovides scientific secretariat support to theAdvisory Committee on Pesticides and Health(ACPH), an independent expert advisorycommittee to the Department and to theNational Registration Authority for Agriculturaland Veterinary Chemicals (the NRA).


191


The Section also undertakes technical policydevelopment and provides health advice oninternational chemicals treaty negotiations. Itinteracts with the Population Health Division ofthe Department of Health and Ageing on a rangeof environmental health issues.

An important role of CRIHS is to encourage, andwhere practical, to extend, internationalharmonisation of chemicals regulation includingtoxicological reviews and re-registrationprograms. Another important function of theSection is to ensure public access to informationrelevant to the use of chemicals and the hazardsthey pose, especially pesticides.

5.6.3 Assessment processes withinthe chemicals unit

Toxicologists within the Chemicals Unit assessmammalian toxicology and toxicokinetic data andprepare written assessment reports which carrysufficient detail of the studies and findings toallow an independent assessment of the data. Asthe primary emphasis is on independentassessment, limited regulatory status is given tocompany summaries and company sponsored‘expert reports’; it is important that all toxicitydata and the methods by which they are obtainedbe subjected to critical and independent scientificassessment.

Hazard/Risk Assessment: Given the complexityof biological data interpretation and the need forprofessional judgement and a flexible approachwhen assessing the public health risk ofchemicals, it has not been the policy to establishprescriptive methodologies for hazard and riskassessment, although several guidance documentsfor evaluators have been drafted. In general, aqualitative approach is used to assess chemicalrisk. The approach taken to derive an AcceptableDaily Intake (ADI) follows the principlesoutlined in Environmental Health CriteriaMonographs no’s 104 and 210 prepared by theWHO/UNEP/ILO International Programme onChemical Safety (IPCS). (A notable exception tothe use of principles proposed in the EHC 104 isthe endpoint for cholinesterase inhibition byorganophosphorus compounds and carbamates, in

which case Australia has continued to use moreconservative estimates based on inhibition ofplasma cholinesterase rather than inhibition ofred cell or brain acetylcholinesterase.)

Whilst the main focus of agricultural andveterinary chemical assessments is a considerationof human exposure to pesticides throughingestion of residues in food and/or drinkingwater, the direct dermal or inhalational exposureof the public, as users of chemicals (in the homegarden/domestic setting) or as bystanders toagricultural or licensed Pest Control Operator(PCO) use, is also taken into account. (Riskassessments for workers exposed in theoccupational setting are performed by theNOHSC—see Section 5.2)

In general, a classification system for publichealth aspects is not used when regulatingchemicals with potential carcinogenicity. Thepotential human carcinogenicity of chemicals isassessed using a weight-of-evidence approachwhich takes into account epidemiological data,carcinogenic potency in animals, biologicalrelevance and potential human exposure.Australia, in this regard, supports the generalapproach outlined by IPCS. Until there is a betterunderstanding of the factors which influencecarcinogenicity, the basis for a classificationscheme remain unclear and thus the benefits ofusing such a scheme to regulate chemicals in thearea of public health are limited. Indeed, whilstexisting classification schemes for carcinogens arebased on assessment of carcinogenic hazard, thereis a danger that these may be misinterpreted as aclassification of carcinogenic risk.

Exposure Assessment: Calculations of likely dailyintakes of pesticide residues [either NationalTheoretical Maximum Daily Intakes (NTMDIs)or National Estimated Dietary Intakes (NEDIs)are based on the procedures as outlined in the‘Guidelines for Predicting Dietary Intake ofPesticide Residues’ (1989) published by theUNEP/FAO/WHO Food ContaminationMonitoring Programme in collaboration with theCodex Committee on Pesticide Residues.

192

Food consumption data is used in dietary riskassessment and is available from the AustralianDietary Survey (ADS) and the Market BasketSurvey (MBS). The ADS estimates actual foodintakes in various subgroups of the population,taking into account such factors as age and ethnicbackground. The MBS measures the amount ofpesticide residue in ready-to-eat food based on atypical diet for different age groups. Estimateddaily food residue intake can be compared withthe ADI.

Whilst there is no formal program to assesshuman exposure to pesticides in the domesticsetting, pesticides for domestic use are restrictedto those of low toxicity and they have appropriatecontrols on availability, packaging and labelling.Additional exposure assessment may be carriedout where a particular concern arises.

Risk management: Toxicological issues may raiseconcerns with respect to supply, availability, anduse of agricultural or veterinary chemicals. Thesupply and availability of chemicals can bemanaged through NRA’s registration process; thatis, approval for pesticide Technical Grade ActiveConstituents (TGACs) may not be granted (or bewithdrawn), approval for particular uses of apesticide or veterinary chemical may not begranted (or be withdrawn), or registrations forparticular products may not be granted (or bewithdrawn), in order to eliminate or reducepotential public exposure.

The use of registered agvet products on themarket can be regulated through poisonsscheduling and appropriate labelling. TheCommonwealth Government, acting on theadvice of its National Drugs and PoisonsSchedule Committee (a committee nowestablished under the TGA Act), recommendsclassification of drugs and poisons which arepublished in the Standard for the UniformScheduling of Drugs and Poisons (SUSDP); thesefederal recommendations are adopted by Statelegislation. The more restrictive schedulesprescribe restrictions on supply and use, as wellthe use of appropriate signal headings on labels.

The poisons schedule classification of a chemicaland its formulated products, together withproduct labelling instructions (first aid and safetydirections), help control the availability ofproducts and help minimise the exposure of users.

5.7 Standards Australia Modelof Risk Management

A risk management Standard has been publishedby Standards Australia (1999). This is directedtowards ‘as wide a range of risk and riskmanagement disciplines as possible’ forapplication to ‘a very wide range of activities oroperations of any public, private or communityenterprise, or group’ so as to establish ‘asystematic risk management program’. TheStandard provides ‘a generic guide for theestablishment and implementation of the riskmanagement process involving establishing thecontext and the identification, analysis,evaluation, treatment, communication andongoing monitoring of risks’

The risk management process outlined in theStandard contains a model of risk assessment anduses the term ‘risk analysis’ to describe theprocess. ‘Risk analysis’ is defined as ‘a systematicuse of available information to determine howoften specified events may occur and themagnitude of their consequences’.

The risk management Standard has been followedby a further Standard ‘Environmental riskmanagement—Principles and process. HB 203:2000 (Standards Australia, 2000). This gives moreextensive detail on environmental riskmanagement using the framework established inAS/NZS 4360: 1999.

Both Standards provide qualitative measures ofconsequence and likelihood. Appendices in HB203: 2000 detail sustainability principles, linksbetween environmental risk and environmentalmanagement systems, discussion of how theacceptability of risk may be considered, sources ofinformation for risk identification and cost-benefit analysis.


193


6.1 Risk Assessment in CanadaIn 1993, Canada published a formal RiskDetermination framework that defined anddescribed the risk assessment and riskmanagement process in a structured way. Theframework reflected practices that had beenoccurring for a number of years. A process ofreview began in 1997 and the term ‘riskdetermination’ was replaced with ‘riskmanagement’.

The revised risk management process:

• has risk assessment as an inherent part of riskmanagement rather than as a separateprocess;

• has a focus on adverse health effects butexamines how information from sources suchas the biophysical sciences, social sciences andeconomics can contribute to anunderstanding of risk;

• clarifies the risk management process,decisions and related information, and theroles of all parties involved in the riskmanagement process; and

• provides broader participation in the riskmanagement process.

After a step that identifies the problem and itscontext, a step for assessing potential risks andbenefits occurs. The risk assessment component isa four stage process comprising:

• Identify hazards. The methods used varydepending on the type of agent involved andwhether it is being reviewed prior to or afterentering the market or environment;

• Characterise hazards. This involves thequalitative and/or quantitative evaluation ofthe nature of the adverse health effect(s) thathumans may experience under expected levelsof exposure. The preferred source of data iswell-designed and conducted epidemiologystudies, combined with documented exposureassessments;

• Assess exposures. Deterministic exposureassessments using generated single-pointestimates of exposure are most common.Probabilistic assessments are used for moreextensive assessments aimed at long termmanagement of a risk e.g. fish and wildlifecontaminants, food-borne microorganismsand consumer products. High exposurescenarios are used occasionally, particularly forexposure assessments involving chemicalhazards such as environmental and food bornechemicals and consumer products for whichextensive laboratory-exposure, epidemiologicalmonitoring and surveillance data are available.This is becoming less common because adecline in available data e.g. from reducedanimal experimentation; and

• Characterise risks using scientific data.

6.2 Risk Assessment in theUnited States of America

The development of regulatory risk assessmentapproaches became prominent in the USA in the1980s but quantitative risk assessment dates to1976 when the brief and generic EPA guidelinesfor cancer risk assessment were published(Hrudey, 1998).

One important landmark was a Supreme Courtdecision in 1980 when an Occupational Safetyand Health Administration (OSHA) standard forexposure to benzene in the workplace was struckdown. The policy had been aimed at reducingexposure as far as technologically possible but didnot consider whether the existing concentrationposed a significant risk to health. The majority ofthe court concluded that under their legislation,OSHA could only regulate if benzene posed asignificant risk of harm. While ‘Whose significantrisk of harm’ was not defined, the decisionhighlighted that some form of quantitative riskassessment was required as the basis for decidingwhether the risk was large enough to warrantregulation (NRC, 1994).

194

International Models of Risk Assessment

Appendix 6

Following from this judgement, Congressinstructed FDA to have the National ResearchCouncil (NRC) appraise federal efforts to use riskassessment in 1981.

Drawing on work done previously by theEnvironmental Protection Agency, the Food AndDrug Administration, the Occupation Safety AndHealth Administration, International Agency forResearch on Cancer and the National CancerInstitute, the NRC report (1983) recommended arisk assessment on specific definitions of riskwithout recommending specific methods for theconduct of risk assessment.

Two key recommendations of the 1983 report were:

• A clear conceptual distinction between riskassessment and risk management should bemaintained. However it was recognised it wasnot necessary nor advisable for a physicalseparation of the two activities.

• The scientific basis for risk assessment shouldbe detailed along with default options. It wasintended that the guidelines should beflexible and allow departures from thedefaults if there was appropriate data toindicate that the default option was notappropriate.

The NRC committee did not recommend aspecific methodology for risk assessment butnoted that there should be opportunities forcontinuing review of the science underlying theguidelines and of the associated default options(NRC, 1994). The report acknowledged thecritical role of science policy judgements and thatthese must be distinguished from scientific facts.

The Office of Science and Technology Policybrought together scientists from regulatoryagencies, the National Institutes of Health andother federal agencies. This body reviewed thescientific basis of risk assessment of chemicalcarcinogens and adopted the framework for riskassessment proposed by the NRC. Only theEnvironment Protection Agency adopted aspecific set of guidelines for carcinogen riskassessment (in 1986). The Environment

Protection Agency has gone on to issue guidelinesfor other adverse health effects (mutagenicity,developmental toxicity, effects of chemicalmixtures, reproductive risk)

An important step in the application of thesemethodologies is to regulatory decision makingwas the EPA’s adoption of risk assessment toguide decisions at major contaminated sites. Itwent on to apply risk assessment methodologiesto decisions regarding pesticide residues in food,carcinogenic contaminants of drinking watersupplies, industrial emissions of carcinogens tosurface waters, and specified industrial chemicals(NRC, 1994).

The linearised multistage model using upperbound estimates has underpinned US regulatoryrisk assessment of carcinogens. It has beenlabelled as ‘one of the most conservative modelsused in QRA’ (IEH, 1999b). The US EPA hasproposed changing from the linearised multi stageapproach to a benchmark dose approach as theirdefault model (US EPA, 1996) but the outcomeof the proposal is not yet determined. Theproposal remains in draft form and a recentattempt by the US EPA to recognise a thresholdfor carcinogenic effects from chloroform waswithdrawn.

6.3 Risk Assessment in theUnited Kingdom

In 1996 the Government/Research CouncilsInitiative on Risk Assessment and Toxicology wasestablished to review current practices formanaging risks to health from chemical and topromote improved risk assessment decision-making. The agencies involved covered a widevariety of risks including those from foodcontaminants and additives, agriculturalpesticides, biocides, veterinary products,occupational exposures, consumer products, airquality, water quality, land quality, and humanmedicines. As a result of the deliberations of theInitiative, a four stage process of risk assessmenthas been proposed consisting of:

1. identifying the properties of chemicals thatcan lead to adverse (toxic) health effects(hazard identification);


195


2. obtaining quantitative information about thehazard including, where possible, informationon dose–response relationships (hazardcharacterisation);

3. assessing exposure to the chemical (exposureassessment); and

4. comparing exposure and hazard information(risk characterisation).

The Initiative has described the variety of riskassessment practices used in different governmentdepartments as a step towards establishing acommon framework for the procedures used,identifying the major areas of uncertainty andweakness in current risk assessment processes, andestablishing where these risk assessment processesmight benefit from harmonisation acrossdepartments.

There are substantial differences betweendepartments and agencies in the degree of cautionincorporated into risk assessment for factors such as:

• the size of uncertainty factors applied whenthere are thresholds for toxic effects;

• the use of mathematical approaches foreffects with or without a threshold.Mathematical modelling (using Probitanalysis of the best available data set) may beused as one component of the riskassessment;

• the approaches to the assessment of genotoxiccarcinogens. The UK has tended to use aqualitative weight of evidence approach to theevaluation of carcinogenic risk and has tendedto avoid the use of mathematical approachesfor quantitatively assessing risks fromgenotoxic carcinogens and they are rarely, ifever, carried out by UK regulatory agencies.The UK Department of Health’s Committeeon Carcinogenicity does not support theroutine use of QRA for chemical carcinogens(IEH, 1999b);

• the treatment of data gaps and efficiency;

• the degree of protection for generalpopulation exposures compared tooccupational exposures; and

• the degree of conservatism built into worsecase exposure estimates (IEH, 1999b).

Particular interest has been taken in strategies fordealing with variability within the humanpopulation as a result of factors such as age, sex,pregnancy status, health status, lifestyle, andgenetic factors. Physiologically-basedpharmacokinetic modelling is considered to helpto highlight and reduce the uncertainties ofestimating the dose of an agent the body or partsof the body may receive after exposure.

A Procedure for Microbiological Risk Assessment(ACDP, 1996) has detailed several stages for theprocess of microbiological risk assessment whichis undertaken after the cause for concern isdetermined:

• Identification of the source(s) of the hazard(s)and the conditions under which adverseconsequences could occur; and

• Reviewing and quantifying the riskconsequent upon each hazard.

6.4 Terminologies used in RiskAssessment

While the fundamental processes are usuallysimilar, slightly different terminologies are usedinternationally to describe components of the riskassessment process. ILSI (2000) has produced acomparison table of proposed models and how theyfit in with the NAS paradigm. (See Figure 1 A6).

196

Figure 1 A6: Terminologies used in risk assessment

(adapted from ILSI, 1996 with permission)


197

OIE Import RA

enHealth EHRA

ILSI RSI UP EPA Ecological RA

NAS NRC RA

Codex RA


Problem Formulation

Problem Formulation



Release Assessment


Exposure Assessment

Characterisation of Exposure

Characterisation of Exposure

Exposure Assessment

Exposure Assessment

Exposure Assessment

Hazard Assessment

Characterisation of Human

Health Effects

Characterisation of Ecological

Effects

Dose-response Assessment

Hazard Characterisation

Consequence Assessment






Risk Estimation

Analysis Phase Analysis Phase


7.1 IntroductionThis section describes the cancer endpoint orendpoints that have been observed and identifieswhich of these is addressed in the analysis.(The nature of the framework is such that onlyone mode of action is analysed at a time; hence,for example, tumour types associated with adifferent mode of action, even if recorded in thesame animals, will require separate frameworkanalyses). However, where different tumours areinduced by related mode of action, they are bestaddressed in a single analysis. It should also benoted that some modes of action will involvemultiple contributing components.

7.2 Postulated Mode of Action(theory of the case)

This section comprises a brief description of thesequence of events on the path to cancer for thepostulated mode of action of the test substance.This explanation of the sequence of events leadsinto the next section which identifies the eventsconsidered ‘key’ (i.e. measurable) given the database available for the analysis.

7.3 Key EventsThis section briefly describes the ‘key events’—i.e. measurable events that are critical to theinduction of tumours as hypothesised in thepostulated mode of action. To support anassociation, a body of experiments needs to define and measure an event consistently.

Pertinent observations: e.g. tumour response andkey events in same cell type, sites of actionlogically relate to event(s), increased cell growth,specific biochemical events, organ weight,histology, proliferation assays, hormone or otherprotein perturbations, receptor-ligand changes,DNA or chromosome effects, cell cycle effects.For example, key events for tumours hypothesisedto be associated with prolonged regenerativeproliferation might be cytotoxicity in as measuredhistopathologically and an increase in labellingindex. Key events for induction of urinary bladdertumours hypothesised to be due to formation ofbladder stones composed primarily of calcium

phosphate might include elevated urinarycalcium, phosphate and pH, and formation ofbladder stones followed by irritation andregenerative hyperplasia of the urothelium

7.4 Dose–Response RelationshipThis section should detail the observeddose–response relationships and discuss whetherthe dose–response for the key events parallels thedose–response relationship for tumours. Ideally,one should be able to correlate increases inincidence of a key event with increases inincidence or severity (e.g. lesion progression) ofother key events occurring later in the process,and with the ultimate tumour incidence.Comparative tabular presentation of incidence ofkey events and tumours is often helpful inexamining dose–response.

7.5 Temporal AssociationThis section should detail the observed temporalrelationships or sequence of events and discusswhether the key events precede the tumourresponse. One should see the key events beforetumour appearance; this is essential in decidingwhether the data support the postulated mode ofaction. Observations of key events at the sametime as the tumours (e.g. at the end of a bioassay)do not contribute to temporal association, but cancontribute to analysis in the next section. Mostoften, complete data sets to address the criterionof temporality are not available.

7.6 Strength, Consistency andSpecificity of Association ofTumour Response with KeyEvents

This section should discuss the weight of evidencelinking the key events, precursor lesions and thetumour response. Stop/recovery studies showingabsence or reduction of subsequent events ortumour when a key event is blocked or diminishedare particularly important tests of the association.Consistent observations in a number of suchstudies, with differing experimental designsincreases that support since different designs may

198

WHO/IPCS Conceptual Framework for Cancer Risk Assessment

Appendix 7

reduce unknown biases or confounding.Consistency, which addresses repeatability of keyevents in the postulated mode of action for cancerin different studies is distinguished fromcoherence, however, which addresses relation ofthe postulated mode of action with observations inthe broader database (see point 7.7). Pertinentobservations: e.g. tumour response and key eventsin same cell type, sites of action logically relate toevent(s), initiation- promotion studies,stop/recovery studies.

7.7 Biological Plausibility andCoherence

The postulated mode of action and the eventsthat are part of it need to be based on currentunderstanding of the biology of cancer to beaccepted, though the extent to which biologicalplausibility as a criterion against which weight ofevidence is assessed is necessarily limited, due toconsiderable gaps in our knowledge in this regard.One should consider whether the mode of actionis consistent with what is known aboutcarcinogenesis in general (biological plausibility)and in relation to what is also known for thesubstance specifically (coherence). For the former,likeness of the case to others for structuralanalogues may be informative (i.e. structureactivity analysis). Additionally, this section shouldconsider whether the database on the agent isinternally consistent in supporting the purportedmode of action, including that for relevant non-cancer toxicities. Some modes of action can beanticipated to evoke effects other than cancer,e.g. reproductive effects of certain hormonaldisturbances that are carcinogenic. Moreover,some modes of action are consistent withobserved lack of genotoxicity. Coherence, whichaddresses relation of the postulated mode ofaction with observations in the broaderdatabase—for example, association of mode ofaction for tumours with that for other endpoints—needs to be distinguished from consistency(addressed in Section 6 above) which addressesrepeatability of key events in the postulated modeof action for cancer in different studies.

7.8 Other Modes of ActionThis section discusses alternative modes of actionthat logically present themselves in the case. Ifalternative modes of action are supported, theyneed their own framework analysis. These shouldbe distinguished from additional components of asingle mode of action which likely contribute tothe observed effect, since these would be addressedin the analysis of the principal mode of action.

7.9 Assessment of PostulatedMode of Action

This section should include a clear statement ofthe outcome with an indication of the level ofconfidence in the postulated mode of action e.g.high, moderate or low.

7.10 Uncertainties,Inconsistencies and Data Gaps

Uncertainties should include those related to boththe biology of tumour development and those forthe database on the compound of interest.Inconsistencies should be flagged and data gapsidentified. For the identified data gaps, thereshould be some indication of whether they arecritical as support for the postulated mode ofaction or simply serve to increase confidencetherein.

This version of the Framework is current at April2001 but may be subject to further development.


199


8.1 IntroductionThe aim of microbiological risk assessment is toestimate the level of disease associated with aparticular pathogen in a given population under aspecific set of conditions and for a certain timeframe.

There is much support for the application anddevelopment of microbiological risk assessment(MRA) (ACDP, 1996). To date, MRA haspredominantly been applied to two exposuresources, food and water, and much of theconceptual development of MRA has resultedfrom the application of MRA to these media.

When compared with the chemical riskassessment, microbiological risk assessment(MRA), and particularly quantitativemicrobiological risk assessment (QMRA), canbest be described as being in their infancy. Forexample, with QMRA, vast data sets need to bedeveloped, modelling needs to be improved (e.g.secondary transmission) and analytical techniquesneed to be refined etc. The framework for MRAis still being developed with different approachesbeing proposed. Terminology specific to MRAand many other issues such as how to extrapolatefrom animal models to human models are yet tobe resolved.

Nevertheless, Haas et al, (1999) have successfullydeveloped a MRA process which follows on fromthe influential US National Academy of Sciences(NAS) 1983 framework.

MRA is a developing field undergoing muchtransformation. With this in mind, the aim of thisAppendix is to provide a brief description of theprinciples of MRA and to outline the MRAprocess. A more detailed description of MRA canbe obtained by referring to the documents in theBibliography.

8.2 DefinitionsMicrobiological risk assessment (MRA) has beendefined by various scientificorganisations/committees as follows.

The Codex Alimentarius commission (formicrobiological hazards in foods):

• A scientifically-based process consisting of thefollowing steps; hazard identification, exposureassessment, hazard characterisation and riskcharacterisation (Codex, CAC/GL-30 (1999))

The International Life Sciences Institute—RiskScience Institute (in conjunction with the USEPA):

• A process that evaluates the likelihood of humanhealth effects occurring after exposure to apathogenic microorganism or to a medium inwhich pathogens exist (ILSI, 2000).

The Advisory Committee on DangerousPathogens (UK)

• A formal structured procedure for identifying andcharacterising microbiological hazard anddetermining the risk associated with it (ACDP, 1996).

Quantitative microbiological risk assessment(QMRA) has been defined as follows:

• The application of principles of risk assessment tothe estimate of consequences from a planned oractual exposure to infectious microorganisms(Hass et al, 1999).

8.3 General PrinciplesThe estimation of risk is sometimes expressed innumerical notation. However, risk can also beexpressed qualitatively by using terms such aslow/medium/high. Risk may also be characterisedby a narrative description of the risk, or whetherit breaches standards or guidelines. In practice,however, a continuum exists from a fullyquantitative through to a wholly narrativeexpression of risk.

At present, MRA cannot always practicallyachieve numerical expression of microbiologicalrisk (ACDP, 1996). This can be due to, forexample, lack of dose–response data or a lack ofunderstanding of the route of entry of a pathogen.Semi-quantitative or qualitative MRA can beapplied in these situations.

200

Microbiological Risk Assessment

Appendix 8

The Codex principles of MRA, as applied tofood, are listed below. These principles can also begeneralised to the other media—water, air, soiland the surfaces of inanimate objects. Most of theprinciples listed are similar to established riskassessment principles, except for item 9 which isunique to MRA.

1. Microbiological Risk Assessment should besoundly based upon science

2. There should be a functional separationbetween Risk Assessment and RiskManagement

3. Microbiological Risk Assessment should beconducted according to a structured approachthat includes Hazard Identification, HazardCharacterisation, Exposure Assessment, andRisk Characterisation.

4. A Microbiological Risk Assessment shouldclearly state the purpose of the exercise,including the form of Risk Estimate that willbe the output.

5. The conduct of a Microbiological RiskAssessment should be transparent.

6. Any constraints that impact on the RiskAssessment such as cost, resources or time,should be identified and their possibleconsequences described.

7. The Risk Estimate should contain adescription of uncertainty and where theuncertainty arose during the Risk Assessmentprocess.

8. Data should be such that uncertainty in theRisk Estimate can be determined; data anddata collection systems should, as far aspossible, be of sufficient quality and precisionthat uncertainty in the Risk Estimate isminimised (sic).

9. A Microbiological Risk Assessment shouldexplicitly consider the dynamics ofmicrobiological growth, survival and death infoods and the complexity of the interaction(including sequelae) between human andagent following consumption as well as thepotential for further spread.

10. Wherever possible, Risk Estimates should bereassessed over time by comparison withindependent human illness data.

11. A Microbiological Risk Assessment may needreevaluation, as new relevant informationbecomes available.

(Codex, CAV/GL-30; 1999, p.2–3).

8.4 Microbiological RiskAssessment—Paradigmsand Frameworks

Microorganisms are living entities and are verydifferent to chemicals and physical hazards by theirnature. Some believe that MRA requires additionalmethods and terminology which are particular tomicrobiological risks (e.g. methods of estimatingsecondary transmission, and infective doses need tobe developed. However, the enHealth model can,in general, be applied to MRA.

Haas et al (1999) have produced the mostcomprehensive attempt at describing the methodsused in QMRA and the particular needs ofMRA. However, they have not developed amodified framework that attempts to encompassthese different needs. Instead, their approach to(Q) MRA loosely follows the National Academyof Sciences framework proposed for chemical riskassessment (NAS, 1983) which broadly includesthe following steps: hazard assessment(comprising of hazard identification anddose–response analysis), exposure assessment, andrisk characterisation.

By contrast, the International Life SciencesInstitute and the US EPA (ILSI, 1996, 2000)have explicitly adapted the NAS framework tosuit the unique challenges presented by MRA.Like the QRMA process described by Haas et al,it essentially follows the standard descriptionprovided by the National Academy of Sciencesparadigm but uses synonymous terms for eachpart of the process.


201


8.5 Microbiological RiskAssessment Bibliography

• Haas, CN et al (1999). Quantitative MicrobialRisk Assessment. New York: John Wiley andSons.

FAO/WHO• Joint FAO/WHO (2000) Expert Consultation

on Risk Assessment of Microbiological Hazardsin Foods, Rome 17–21, July.

• Joint FAO/WHO (1999) Expert Consultationon Risk Assessment of Microbiological Hazardsin Foods. Geneva, Switzerland, 15 to 19March.

• Joint FAO/WHO (1995) ExpertConsultation. Application of risk analysis tofood standards issues. Geneva, Switzerland.World Health Organization.

• FAO/WHO (1995) Food StandardsProgramme. Joint FAO/WHO expertconsultation on the application of risk analysis tofood standards issues. Recommendations to theCodex Alimentarius Commission. ALINORM95/9, p. 15.

• WHO, Geneva (Switzerland). (1995) FAO,Rome (Italy). Application of risk analysis to foodstandards issue. Report of the Joint FAO/WHOExpert Consultation, Geneva, Switzerland,13–17 March. 43 p. No: XF96:355857.

FAO/WHO/Codex• Joint FAO/WHO Codex Alimentarius

Commission, Rome (Italy). Harmonization ofrisk assessment procedures by countries within theRegion. No: XF95:350171.

Codex• Codex. (1999). Principles and Guidelines for

the Conduct of Microbiological Risk Assessment.Cac/Gl-30.

• Proposed Draft Principles and Guidelines for theConduct of Microbiological Risk Management(2000) (At Step 3 Of The Procedure).Cx/Fh 00/6.

ILSI• International Life Sciences Institute—Risk

Science Institute. (2000). Revised Frameworkfor Microbial Risk Assessment Summary reportof an ILSI Risk Science Institute workshop.

• ILSI Risk Science Institute Pathogen RiskAssessment Working Group (1996). Aconceptual framework to assess the risks ofhuman disease following exposure topathogens. Risk Analysis; 16(6): 841-848.

ACDP• Advisory Committee on Dangerous

Pathogens (1996). Microbiological RiskAssessment: an interim report. London:HMSO.

Scientific articles• Gale P J. (1996) Developments in

microbiological risk assessment models fordrinking water. A short review. AppliedBacteriology, 81, 4:403–410.

• Jaykus LA. (1996) The application ofquantitative risk assessment to microbial foodsafety risks. Critical Reviews in Microbiology,22, 4:279-293.

• Lammerding AM. (1996) Microbial food safetyrisk assessment: principles and practice.Proceedings of the Fourth ASEPTInternational Conference, in Laval, France.

• Lammerding AM. (1997) An Overview ofMicrobial Food Safety Risk Assessment.Journal of Food Protection; 60(11):1420–1425.

• Macler BA and Regli S. (1993) Use ofmicrobial risk assessment in setting USdrinking water standards. Int. J. FoodMicrobiology, 18, 4:245–256.

• McNab BW. (1998) A General FrameworkIllustrating an Approach to QuantitativeMicrobial Food Safety Risk Assessment.Journal of Food Protection; 61(9):1216–1228.

202

• Schlundt J. (2000) Comparison ofmicrobiological risk assessment studiespublished. Int. J. Food Microbiology15;58(3):197-202.

• Schothorst M van. (1997). Practicalapproaches to risk assessment. Journal of FoodProtection. 60(11):1439–1443.

• Voysey PA, Brown M, (2000) Microbiologicalrisk assessment: a new approach to foodsafety control. Int. J. Food Microbiology15;58(3):173–9.


203


Advisory Committee on Dangerous Pathogens(ACDP) (1996). Microbiological RiskAssessment: an interim report. HMSO,London.

Aitio A, Jarvisalo J, Riihimaki V and HernbergS (1988). Biologic monitoring. In: Zenz C,ed. Occupational Medicine. Principles andpractical applications. Year Book MedicalPublishers, Chicago.

Alsop WR, Ruffle B, Frohmberg EJ andAnderson PD (1993). Development of risk-based clean-up standards using Monte Carloanalysis. In: Developing clean-up standards forcontaminated soil, sediment, and groundwater.How clean is clean? ATSDR/US EPA.Washington.

American Academy of Pediatrics (1999).Testimony from the Committee on Genetics.Presentation to the Public Advisory Committeeof the Food and Drug Administration.Dermatologic and Ophthalmic Drugs AdvisoryCommittee. Presented September 4, 1997 byJames W. Hanson, MD.www.aap.org/visit/thal01.htm

American Industrial Health Council (AIHC)(1989). Presentation of Risk Assessments ofCarcinogens: Report of the ad hoc StudyGroup on Risk Assessment. AmericanIndustrial Health Council. Washington.

AIHC Task Force (1994). Exposure FactorsSourcebook. Environmental Health RiskAssessment Sub-committee, AmericanIndustrial Health Council. Washington, DC.

Ames BN, Durston WE, Yamasaki E and LeeFD (1973). Carcinogens are mutagens. A simpletest system combining liver homogenates foractivation and bacteria for detection.Proceedings of the National Academy ofSciences of the USA 70: 2281–2285.

American Public Health Association (APHA),American Water Works Association(AWWA), Water Environment Federation(WEF) (Latest edition). Standard methods forthe examination of water and waste water.APHA, AWWA, WEF. Washington.

Australia and New Zealand Environment andConservation Council (ANZECC) andNational Health and Medical ResearchCouncil (NHMRC) (1992). Australian andNew Zealand Guidelines for the Assessment andManagement of Contaminated Sites. ANZECCand NHMRC. Canberra.

ANZECC and (NHMRC) (1997). TheAssessment and Management of ContaminatedSites—Draft Policy Framework.ANZECC/NHMRC. Canberra.

Australia New Zealand Food Authority(ANZFA) (1996). Framework for theassessment and management of food-relatedhealth risks. ANZFA. Canberra.

Australia New Zealand Food Authority(ANZFA) (1997). Food Standards Code andlater amendments. ANZFA.Canberra.

Australia New Zealand Food Authority(ANZFA) (1998). The Australian MarketBasket Survey 1996. ANZFA. Canberra.

Australian Institute of Health and Welfare(AIHW) (2000). The seventh biennial healthreport of the Australian Institute of Health andWelfare. AIHW. Canberra.

Asano T, Leong LYC, Rigby MG and SakajiRH (1992). Evaluation of the California Wastewater Reclamation Criteria using enteric virusmonitoring data. Water Science andTechnology 26: 1513.

Agency for Toxic Substances and DiseaseRegistry (ATSDR) (1996). Guidance forATSDR Health Studies. Division of HealthStudies, ATSDR. Washington.www.atsdr.cdc.gov/HS/gd1.html

American Water Works Association (AWWA)(1990). Water Quality and Treatment—AHandbook of Community Water Supplies. 4thEdition. American Water Works Association.McGraw Hill, New York. p. 1194.

Bardwell LV (1991). Problem-framing: aperspective on environmental problem-solving. Environmental Management (15):603–612.

204

References

Barnes DG, Daston GP, Evans JS, Jarabek AM,Kavlock RJ, Kimmel CA, Park C andSpitzer HL (1995). Benchmark doseworkshop: Criteria for use of a benchmarkdose to estimate a reference dose. Reg ToxPharmacol 21: 296–306.

Barnes DG and Dourson ML (1988) ReferenceDose (RfD): Description and use in healthrisk assessments. Reg Tox Pharmacol 8:471–486.

Bartram J, Burch M, Falconer I, Jones G andKuiper-Goodman T (1999). SituationAssessment, Planning and Management—Chapter 6. (Table 6.3, p. 202), In: Chorus Iand Bartram J (eds). Toxic Cyanoabacteria inWater—A Guide to Their Public HealthConsequences, Monitoring and Management.E&FN Spon on behalf of the World HealthOrganization. London and New York.

Bayne K (1996). Revised guide for the care and useof laboratory animals. American PhysiologicalSociety. Physiologist 39: 199, 208–211.

Beaglehole R, Bonita R, Kjellstrom T (1993).Basic epidemiology. World HealthOrganization: Geneva.

BELLE Advisory Committee (1998).Newsletter of Biological Effects of Low LevelExposures. Human and ExperimentalToxicology, 17: 681–718.

Benirschke K, Garner FM. and Jones TC (Eds)(1978). Pathology of Laboratory Animals. Vol Iand II. Springer Verlag, New York.

Berg BN and Simms HS (1960). Nutrition andlongevity in the rat. II Longevity and onset ofdisease with different levels of food intake.J Nutr 71: 225–263.

Boyd EM and Knight LM (1963). Postmortemshifts in the weight and water levels of bodyorgans. Toxicol Appl. Pharmacol 5: 119–128.

Breslow NE and Day NE (1987). StatisticalMethods in Cancer Research. Volume II—TheDesign and Analysis of Cohort Studies.International Agency for Research on Cancer,Scientific Publication 82, Lyon.

British Medical Journal (1988). British MedicalJournal. 296: 50.

Burmaster DE and Anderson PD (1994).Principles of Good Practice for the Use ofMonte Carlo Techniques in Human Healthand Ecological Risk Assessments. RiskAnalysis, 14(4): 477–481.

Bush BM (1991). Interpretation of LaboratoryResults for Small Animal Clinicians. BlackwellScientific Publications, London.

Butterworth BE (1990). Consideration of bothgenotoxic and non-genotoxic mechanisms inpredicting carcinogenic potential. MutationResearch 239: 117–132.

Calabrese EJ (1985). Toxic Susceptibility.Male/Female Differences. John Wiley andSons, New York.

Calabrese EJ (1996). Expanding the ReferenceDose concept to incorporate and optimizebeneficial effects while preventing toxicresponses from nonessential toxicants.Regulatory Toxicology and Pharmacology, 24:S68–S75.

Calabrese EJ and Baldwin LA (1994). Improvedmethod of selection of the NOAEL. Reg ToxPharmacol 19: 48–50.

Calabrese EJ and Baldwin LA (1998). Can theconcept of hormesis be generalized tocarcinogenesis? Regulatory Toxicology andPharmacology, 28: 230–241.

Calabrese EJ and Gilbert (1993) Lack of totalindependence of uncertainty factors (UFs):Implication for size of the total uncertaintyfactor. Reg Tox Pharmacol 17: 44–51.

Calabrese EJ and Kostecki PT (1992). Riskassessment and environmental fatemethodologies. Lewis Publishers, Boca Raton.

Calamari D (ed) (1993). Chemical exposureprediction. Lewis Publishers, Chelsea.


205


Calamari D (1999). Assessment ofenvironmental distribution—the use offugacity models for this purpose. In:UNEP/IPCS (1999) Chemical RiskAssessment. Training Module No.3.WHO/PCS/99.2. WHO, Geneva.

Casterline JL Jr and Williams CH (1969).Effect of pesticide administration uponesterase activities in serum and tissue of ratsfed variable casein diets. Tox Appl Pharmacol14: 266–275.

CCME (1999). Recreational Water QualityGuidelines and Aesthetics. CanadianEnvironmental Quality Guidelines. CanadianCouncil of Ministers of the Environment.Canada.

Chess C and Hance, BJ (1994) Communicatingwith the public: 10 Questions EnvironmentalManagers should ask. Center forEnvironmental Communication, The StateUniversity of New Jersey. Rutgers, NewJersey.

Chorus I. and Bartram J (eds) (1999). ToxicCyanoabacteria in Water—A Guide to TheirPublic Health Consequences, Monitoring andManagement. E&FN Spon on behalf of theWorld Health Organization, London andNew York.

Clean Air Society of Australia and NewZealand (latest edition). Air pollutionmeasurement manual. A practical guide tosampling and analysis. Vols 1 and 2. Clean AirSociety of Australia and New Zealand.Eastwood, NSW.

Clesceri LS, Greenberg AE and Eaton AD(1998). Standard Methods for the Examinationof Water and Waste water. 20th Edition.American Public Health Association—American Water Works Association -WaterEnvironment Federation.

Codex (1999). Principles and Guidelines for the Conduct Of Microbiological RiskAssessment, Cac/Gl-30.

Conner MW and Newbern PM (1984).Drug-nutrient interactions and theirimplications for safety evaluations. Fund ApplToxicol 4: 5341–S356.

Covello VT and Merkhofer MW (1993). RiskAssessment Methods—approaches for assessinghealth and environmental risks. Plenham Press,New York.

Craun GF (Ed.) (1993). Safety of WaterDisinfection—Balancing Chemical andMicrobial Risks. International Life SciencesInstitute. ILSI Press, Washington, D.C.

Crome P (1999). The Elderly. In: Institute forEnvironment and Health (IEH) (1999). Riskassessment strategies in relation to population subgroups. Institute for Environment and Health.Leicester.

Cross SJ and Taylor ER (1996). HumanExposure to Soil Contaminants through theConsumption of Home-Grown Produce.Contaminated Sites Monograph Series. No 6.South Australian Health Commission,Adelaide.

Crump K (1984). A new method for determiningallowable daily intakes. Fundamental andApplied Toxicology 4: 854–891.

Crump K (1985). Mechanisms leading todose–response models. In: Ricci P (ed).Principles of Health Risk Assessment, PrenticeHall, Englewood Cliffs, New Jersey.

Cullen AC and Frey HC (1999). Probabilistictechniques in exposure assessment. A handbookfor dealing with uncertainty and variability inmodels and inputs. Plenum Press, New York

Dayton PG and Sanders JE (1983).Dose-dependent pharmacokinetics: Emphasison phase I metabolism. Drug Metab Rev 14:347–405.

Department of the Environment (DOE) (1993).Contaminants in soil: collation of toxicologicaldata and intake values for humans. ProjectNumber 7/10/333.

206

Department of Health (DOH) (1991).Guidelines for the evaluation of chemicals forcarcinogenicity. Committee oncarcinogenicity of chemicals in food,consumer products and the environment.Department of Health, Report RHSS 42,HMSO, London.

Department of Health (DOH) (1998).Communication about risks to public health:pointers to good practice. The Stationery Office.London.

Derelanko MJ and Hollinger MA (eds) (1995).CRC Handbook of Toxicology. CRC Press,Boca Raton, Florida.

Dezuane J (1997). Handbook of Drinking WaterQuality. 2nd Edition. J. Wiley and Sons

Di Marco PN and Buckett KJ (1996). Derivationof health investigation levels for beryllium andberyllium compounds. In: Langley A, MarkeyB and Hill H (eds), The Health Risk Assessmentand Management of Contaminated Sites,Proceedings of the Third National Workshopon the Health Risk Assessment andManagement of Contaminated Sites.Contaminated Sites Monograph SeriesNumber 5, South Australian HealthCommission, Adelaide. p. 1–20.

Dickens A and Robinson J (1996). Statisticalapproaches. In: Evans GO (ed). AnimalClinical Chemistry: A Primer for Toxicologists.Taylor and Francis, London, pp 45–48.

Doull J (1980). Factors influencing toxicology.In: Casarett and Doull’s Toxicology, 2nd Ed,pp. 70-83, Macmillan Pub. Co., New York.[Note: This reference is to an old edition ofthis well-known textbook but the latestedition (at page 22) still refers to thisparticular chapter in the 2nd edition fordetailed discussion on factors influencinganimal responses to toxic substances]

Dourson ML and Stara JF (1983). Regulatoryhistory and experimental support ofuncertainty (safety) factors. Reg ToxicolPharmacol 3: 224–237.

Dourson ML (1984). Acceptable Daily Intake(and discussion). In: Approaches to riskassessment for multiple chemical exposures.Proceedings of the Workshop. September 29-30, 1982. EPA-600/9-84-008. USEnvironmental Protection Agency.Washington. p. 1–42.

Dourson et al (1996). Evolution of science-baseduncertainty factors in noncancer riskassessment. Reg Toxicol Pharmacol 24 (2):121–125

Duncan JR, Prasse KW and Mahaffey EA(1994). Veterinary Laboratory Medicine—Clinical Pathology. 3rd Edition. Iowa StatePress, Ames, Iowa.

Eaton DL and Klaassen D (1996). Principles oftoxicology. In: Casarett and Doull’s Toxicology;The Basic Science of Poisons, 5th Ed. p. 13–33.McGraw-Hill, New York.

ECETOC (1989). Alternative approaches for theassessment of reproductive toxicity. EuropeanChemical Industry Ecology and ToxicologyCentre, Monograph No. 12. Brussels.

Edwards J W, van Alphen M and Langley A(1994). Identification and assessment ofcontaminated land. Improving site historyappraisal. Contaminated Sites MonographSeries. No. 3. South Australian HealthCommission. Adelaide.

Elliott P, Cuzick, English D, Stern R (1992).Geographical and Environmental Epidemiology:Methods for Small-Area Studies. World HealthOrganisation, Oxford.

enHealth (2001). Health Impact AssessmentGuidelines. enHealth, Commonwealth ofAustralia, Canberra.


207


Environment Protection Authority (EPA NSW)(1997). Contaminated Sites. Guidelines forconsultants reporting on contaminated sites.NSW Environment Protection Authority,Sydney.

EPA NSW (1998). Draft Guidelines for the NSWSite Auditor Scheme. Environment ProtectionAuthority. Sydney.

European Commission (1998). Guidelines onthe Application of the PrecautionaryPrinciple. Brussels: Directorate GeneralXXIV, Consumer Policy and ConsumerHealth Protection.

Evans GO (1996). Animal Clinical Chemistry: APrimer for Toxicologists. Taylor and Francis,London.

Farrer D. Lyster R. and Pearson L. TheEnvironmental Law Handbook. 3rd edition.NSW: Redfern Legal Centre Publishing, 1999.

Faustman EM and Omenn GS (1996). Riskassessment. In: Casarett and Doull’s Toxicology.The Basic Science of Poisons. Fifth Editionedited by Klaassen CD, McGraw Hill, NewYork. p. 5–88.

Federal Focus (1996). Principles for EvaluatingEpidemiologic Data in Regulatory RiskAssessment. Recommendations for implementingthe ‘London Principles’ and for risk assessmentguidance. —Developed by an Expert Panel atLondon Conference October, 1995. FederalFocus. Washington.

Ferguson CC (1992). The statistical basis forspatial sampling of contaminated land.Ground Engineering. June: 34-38.

Ferguson CC (1994). The Contaminated LandExposure Assessment Model (CLEA): TechnicalBasis and Algorithms. Draft report submittedto Department of the Environment. TheNottingham Trent University Centre forResearch into the Built Environment.Nottingham.

Fiksel JR and Scow KM (1983). HumanExposure and Health Risks assessments usingoutputs of environmental fate models. In;Swann RL and Eschenroedera (EDS). Fate ofchemicals in the environment; Compartmentaland Multi-media models. ACS Symposiumseries 225. American Chemical Society.Washington.

Finkel AM (1990). Confronting uncertainty inrisk management. Resources for the future.Washington.

Finley B et al, (1994). RecommendedDistributions for Exposure Factors FrequentlyUsed in Health Risk Assessment. RiskAnalysis 14(4): 533–53.

Fitzgerald DJ (1993). Carcinogenic soilcontaminants: an Australian approach. In:Langley AJ, van Alphen M (eds), The HealthRisk Assessment and Management ofContaminated Sites—Proceedings of theSecond National Workshop on the HealthRisk Assessment and Management ofContaminated Sites, p. 51–64. SouthAustralian Health Commission, Adelaide.

Flamm WG and Lehman-McKeeman LD(1991). The human relevance of the renaltumor-inducing potential of d- limonene inmale rats: implications for risk assessment.Regul Toxicol Pharmacol 13: 70–86.

Foran JA and The International Life SciencesInstitute (ILSI) (1997) Risk SciencesWorking Group on Dose Selection—Principles for the selection of doses in chronicrodent bioassays. Env Hlth Persp 105: 18–20.

Fowle JR (1989). Summary and perspectives —Panel discussion on toxicological exposureassessment: State of the art.J Am Coll Toxicol 8: 865–870.

Fowle JR, Alexeeff GV and Dodge D (1999).The use of benchmark dose methodologywith acute inhalation lethality data. RegToxicol Pharmacol 29: 262–278.

208

Fowle JR and Sexton K (1992). EPA priotitiesfor biological research in environmentalhealth. Environ Health Perspect. 98: 235–241.

Fox JG, Cohen BJ and Loew FM (1984).Laboratory Animal Medicine. Academic Press,Orlando.

Freestone D. Het E (eds) The PrecautionaryPrinciple and International Law: the Challengeof Implementation. Cambridge: Kluwer LawInternational, 1996.

Friberg LT (1985). The rationale of biologicalmonitoring of chemicals—with specialreference to metals. American IndustrialHygiene Association Journal 46: 633–641.

Froese KL and KAM Bodo (1999). Healtheffects associated with waste water treatment,disposal and reuse. Water EnvironmentResearch. 71(5): 1119–1126.

Gad SC and Weil CS (1986). Statistics andExperimental Design for Toxicologists. TelfordPress, NJ.

Gad SC and Weil CS (1989). Statistics fortoxicologists. In: Hayes AW. (Ed). Principlesand Methods of Toxicology. 3rd ed. RavenPress, New York. Chapter 7, p. 221–274.

Gaylor DW and Gold LS (1998). Regulatorycancer risk assessment based on a quickestimate of a benchmark dose derived fromthe Maximum Tolerated Dose. Erg ToxicPharmacy 28: 222–225.

Gaylord DW, Ryan L, Crews D and Zhu Y(1998). Procedures for calculating benchmarkdoses for health risk assessment. Erg ToxicPharmacy 28: 150–164.

Gilbert RO (1987). Statistical Methods forEnvironmental Pollution Monitoring. VanNostrand Reinhold. New York.

Goldfrank LR, Flomenbaum NE, Lewin NA,Weisman RS and Howland MA (eds).(1990). Goldfrank’s Toxicologic Emergencies.Prentice-Hall International Inc. EastNorwalk, Connecticut.

Good J (1993). Towards standardised samplingand analysis for contaminated sites. In:Langley AJ, van Alphen M. (eds) The HealthRisk Assessment and Management ofContaminated Sites. Proceedings of theSecond National Workshop. South AustralianHealth Commission. Adelaide. p. 43–49.

Gossel TA and Bricker JD (1989). Principle of Clinical Toxicology 2nd ed. Raven PressNew York.

Government/Research Councils Initiative onRisk Assessment and Toxicology (1999). Riskassessment strategies in relation to populationsubgroups. Report of a workshop organised bythe Risk assessment and Toxicology SteeringCommittee. Institute for Environment andHealth. Leicester, UK.

Greenhalgh T (1997). Papers that summariseother papers—systematic reviews and meta-analysis. British Medical Journal. 315:672–675.

Gregg MB (ed) (1996). Field Epidemiology.Oxford University Press, New York.

Grice HC and Burek JD (1983). Age associated(geriatric) pathology: Its impact on long termtoxicity studies. In: Current Issues inToxicology, Springer Verlag, New York.

Guyton AC (1991) Textbook of MedicalPhysiology. 8th Edition. WB SaundersCompany, Philadelphia.

Haas CN, Rose JB and Gerba CP (1999).Quantitative Microbial Risk Assessment.Wiley. New York.

Haimes YY, Barry T, Lambert JH (Eds) (1994).When and how you can specify a probabilitydistribution when you don’t know much?Proceedings of a Workshop organised by theUS EPA and the University of Virginia,1993, Charlottesville, Virginia. In: RiskAnalysis, 14(5): 661–706.

Hansen H, De Rosa CT, Pohl H, Fay M andMumtaz MM (1998). Public HealthChallenges Posed by Chemical Mixtures.Environmental Health Perspectives 106,Supplement 6: 1271–1280.


209


Haseman JK, Huff J and Boorman GA (1984).Use of historical control data incarcinogenicity studies in rodents. ToxicolPathol 12: 126–135.

Hattis D and Silver K (1994). Human Inter-Individual Variability. A Major Sourceof Uncertainty in Assessing Risks for NonCancer Health Effects. Risk Analysis,14(4): 421–432.

Hawley JK (1985). Assessment of health risksfrom exposure to contaminated soil. RiskAnalysis, 5(4): 289–302.

Hayes WJ (1991). Dosage and other factorsinfluencing toxicology. In: Hayes WJ andLaws ER (eds). Handbook of Toxicology, Vol. 1,General Principles. Academic Press, SanDiego, p. 40–41.

Hayes AW (ed) (1994). Principles and Methods ofToxicology. 3rd Edition. Raven Press,New York.

Health Canada (1999). Revised Health CanadaRisk Management Framework. Draft forDiscussion. April 1999. Canada.

Health Council of Netherlands (1994). RiskAssessment of carcinogenic chemicals in theNetherlands. Regulatory Toxicology andPharmacology 19: 14–30.

Hellard M (1999). The water quality studyapproaches a conclusion. Water Quality News,Issue 10:2.

Helsel DR (1990). Less than obvious, statisticaltreatment of data below the detection limit.Environmental Science and Technology, 24 (12),1767–1774.

Hemminki K (1992). Significance of DNA andprotein adducts. In: Vainio H, McGregor DBand McMichael AJ. (eds) Mechanisms ofcarcinogenesis in risk identification.International Agency for Research of Cancer,Lyon. p. 525–534.

Henry JB (Ed) (1984). Todd, Sanford,Davidsohn. Clinical Diagnosis andManagement by Laboratory Methods.17th Edition. Saunders Co, Philadelphia.

Herrington LP and Nelbach JH (1942). Relationof gland weights to growth and aging processesin rats exposed to certain environmentalconditions. Endocrinol 30: 375–386.

Heyworth JS (1991). Sampling and statisticalanalysis for assessing contaminated land sites.In: El Saadi O, Langley AJ. (eds) The HealthRisk Assessment and Management ofContaminated Sites. South Australian HealthCommission, Adelaide.

Heywood R (1981). Target organ toxicity. ToxicolLett 8: 349–358.

Hill AB (1965). The environment and disease:association or causation? Proceedings of theRoyal Society of Medicine. 58, 295–300.

Hickey J, and Walker V. Refining theprecautionary principle in internationalenvironmental law. Virginia Env Law J 1995;14 (3): 424–54.

Hrudey SE, Chen W and Rousseaux CG (1996).Bioavailability in Environmental Risk Assessment.CRC Lewis Publishers. Boca Raton.

Hrudey SE and Krewski D (1995). Is there asafe level of exposure to a carcinogen?Environmental Science and Technology,29(8): 370A–375A.

Hrudey SE (1998). Quantitative cancer riskassessment—pitfalls and progress. In: HesterRE and Harrison RM. Issues inEnvironmental Science and Technology, No. 9.Royal Society of Chemistry Risk Assessmentand Risk Management. Cambridge, UK.

Hrudey SE (1999). Water quality parametersummary. Prepared for the CooperativeResearch Centre for Water Quality andTreatment. Adelaide.

Hunt VR (1982). Work and the health of women.CRC Press, Boca Raton, Florida.

International Agency for Research on Cancer(IARC) (1978). Some N-Nitroso CompoundsVol. 17. IARC Monographs on theEvaluation of the Carcinogenic Risk ofChemicals to Humans. IARC, Lyon.

210

IARC (1983). Miscellaneous Pesticides Vol 30.IARC Monographs on the Evaluation of theCarcinogenic Risk of Chemicals to Humans.IARC, Lyon.

IARC (1987). Overall Evaluations ofCarcinogenicity: an updating of IARCMonographs Volumes 1 to 42—Suppl. 7.IARC Monographs on the Evaluation ofCarcinogenic Risks to Humans. IARC, Lyon.

IARC (1999) The use of short—and medium-termtests for carcinogens and data on genetic effects incarcinogenic hazard evaluation No. 146.International Agency for Research on CancerScientific Publications. IARC, Lyon.

Imray P and Langley A (1996, 1998).Health-based Soil Investigation Levels.National Environmental Health ForumMonographs. Soil Series No.1. Adelaide:South Australian Health Commission, 1996.

Imray P and Langley A (1999). Health-basedInvestigation Levels. 3rd Edition. NationalEnvironmental Health Forum MonographsSoil Series No. 1. South Australian HealthCommission. Adelaide.

Innami S, Yang MG, Mickelsen O and HafsHD (1973). The influence of high fat diets onestrous cycles, sperm production and fertilityof rats. Proc Soc Exp Biol Med 143: 63–68.

Institute for Environment and Health (IEH)(1999a). Developing New Approaches toAssessing Risk to Human Health fromChemicals. Risk and Toxicology SteeringCommittee. Institute for Environment andHealth. Leicester.

IEH (1999b). Risk Assessment Approaches used byUK Government for Evaluating Human HealthEffects of Chemicals. Institute for Environmentand Health. Leicester.

IEH (1999c). Risk assessment strategies in relationto population sub groups. Institute forEnvironment and Health. Leicester.

International Life Sciences Institute—RiskScience Institute. 2000, Revised Frameworkfor Microbial Risk Assessment Summaryreport of an ILSI Risk Science Instituteworkshop.

ILSI Risk Science Institute Pathogen RiskAssessment Working Group. A conceptualframework to assess the risks of humandisease following exposure to pathogens. RiskAnalysis. 1996; 16(6): 841–848. BlackwellScientific.

IRIS (1996). Integrated Risk Information System.US Environmental Protection Agency,Washington DC.

Jain NC (1993). Essentials of veterinaryhematology. Lea and Fibiger, Philadelphia, PA.

Johannsen (1990). Risk Assessment ofcarcinogenic and noncarcinogenic chemicals.Critical Reviews in Toxicology 20: 341–367.

Keenan KP, Laroque P and Dixit R (1998).Need for dietary control by caloric restrictionin rodent toxicology and carcinogenicitystudies. J Toxicol Environ Health B Crit Rev1(2): 135–148.

Keith LH (ed) (1988). Principles ofEnvironmental Sampling. American ChemicalSociety. Washington.

Keith LH (1990). Environmental sampling: asummary. Environmental Science andTechnology, 24 (5), 610-617.

Keith LH (ed) 1991. Compilation of E.P.A.’sSampling and Analysis Methods. LewisPublishers. CRC Press, Boca Raton, Florida.

Keith LH (1992). Environmental Sampling andAnalysis—A Practical Guide. Lewis Publishers.Chelsea, Michigan.

Kennedy GC (1969). Interactions betweenfeeding behavior and hormones duringgrowth. Ann NY Acad Sci 157: 1049-1061.

Kennedy Jr Gl and Valentine R (1994). Inhalationtoxicology. In: Principles and Methods ofToxicology, 3rd Edition. edited by Hayes AW.Raven Press, New York. p. 805–838.


211


Kimmel CA, Kavelock RJ and Francis EZ(1992). Normal models for assessingdevelopment toxicity. In: Guzelian PS, HenryCJ, Olin SS (eds) Similarities and Differencesbetween children and adults: implications for riskassessment. Washington: International LifeSciences Institute

Klaassen CD (Ed) (1996). Casarett and Doull’sToxicology; The Basic Science of Poisons, 5th Ed.p. 4. McGraw-Hill, New York.

Klinger MM, MacCarter GD and Boozer CN(1996). Body weight and composition in theSprague-Dawley rat: Comparison of threeoutbred sources. Lab Animal Sci 46: 67–69.

Korach KS (Ed) (1998). Reproductive andDevelopmental Toxicology, Marcel Dekker,New York.

Kosslyn SM (1994). Elements of graph design.WH Freeman and Co. New York.

Lang L (1995). Strange brew: assessing risk ofchemical mixtures. Environmental HealthPerspectives 103(2): 142–145.

Langley A and El Saadi O (1991). Protocol of thehealth risk assessment and management ofcontaminated sites. In: El Saadi O, Langley AJ.The Health Risk Assessment and Management ofContaminated Sites. South Australian HealthCommission. Adelaide. p. 205–240.

Langley AJ (1991). Exposure Pathways. In: ElSaadi O, Langley AJ. The Health RiskAssessment and Management of ContaminatedSites. South Australian Health Commission,Adelaide.

Langley AJ (1991a). Protecting the Public:Biological Monitoring. In: El Saadi O,Langley AJ. The Health Risk Assessment andManagement of Contaminated Sites. SouthAustralian Health Commission, Adelaide.

Langley AJ (1993). Refining exposureassessment. In: Langley AJ, Van Alphen M,eds. The health risk assessment and managementof contaminated sites. Proceedings of theSecond National Workshop on the HealthRisk Assessment and Management ofContaminated Sites. Canberra [ContaminatedSites Monograph Series No. 2]. SouthAustralian Health Commission, Adelaide.

Langley AJ (1993a). The representation of dataand the appraisal of contaminated land. In:Langley A J, van Alphen M. The Health RiskAssessment and Management of ContaminatedSites. Proceedings of the Second NationalWorkshop. Adelaide: South AustralianHealth Commission.

Langley AJ, Sabordo L (1996). ExposureFactors in Risk Assessment. In: Langley A,Markey B and Hill H (eds) The Health RiskAssessment and Management of ContaminatedSites, Proceedings of the Third NationalWorkshop on the Health Risk Assessmentand Management of Contaminated Sites.Contaminated Sites Monograph SeriesNumber 5. South Australian HealthCommission, p. 137–190.

Langley AJ, Taylor A and Dal Grande E (1998).1996 Australian Exposure Factors.In: Langley A, Imray P, Lock W and Hill H(eds). The Health Risk Assessment andManagement of Contaminated Sites,Proceedings of the Fourth NationalWorkshop on the Health Risk Assessmentand Management of Contaminated Sites.Contaminated Sites Monograph SeriesNumber 7. South Australian HealthCommission, Adelaide. p. 137–190.

Langley A et al (in press) Australian ExposureFactors Handbook. National EnvironmentalHealth Forum Monograph. Exposure SeriesNo 1. National Environmental HealthForum. Adelaide.

Last JM (1988). A dictionary of epidemiology.2nd Ed. Oxford University Press, Oxford.

212

Lauwerys R (1984). Basic concepts of humanexposures. In: Berlin A, Draper M,Hemminki K and Vainio H eds. Monitoringexposure to carcinogenic and mutagenic agents.IARC Scientific Publications No. 59. IARC.Lyon. p. 31–36.

Lead Safe (1997). A guide for health careprofessionals. Lead Education Program. NSWLead Reference Centre. NSW.

Lee PN (1993). Statistics. In: Anderson D andConning DM (Eds). Experimental Toxicology:The Basic Issues. Royal Society of Chemistry,Cambridge, UK.

Levy BS and Wegman DH (1988). OccupationalHealth. Recognising and preventing work-related disease. 2nd Edition. Little Brown,Boston.

Levy G (1968). Dose dependent effects inpharmacokinetics. In: Important FundamentalPrinciples in Drug Evaluation p. 141–172.

Lewis SC, Lynch JR and Nikiforov AI (1990).A new approach to deriving communityexposure guidelines for ‘No ObservedAdverse Effect Levels’. Reg Tox and Pharm 3: 17–34.

Lilienfield AM and Lilienfield DE (1980).Foundations of Epidemiology. 2nd edition.Oxford University Press.

Lock WH (1996). Composite sampling. NationalEnvironmental Health Monographs SoilSeries No. 3. South Australian HealthCommission. Adelaide.

Lord DW (1987). Appropriate siteinvestigations. In: Cairney T (ed). ReclaimingContaminated Land. Blackie. Glasgow.

Lovell DP and Thomas G (1996). Quantitativerisk assessment and the limitations of thelinearised multistage model. Human andExperimental Toxicology 15: 87–104.

Lu FC and Sielken Jr RL (1991). Assessment ofsafety/risk of chemicals: inception and evolutionof the ADI and dose–response modellingprocedures. Toxicology Letters 59: 5–40.

Mackay D (1991). Multimedia environmentalmodels: the fugacity approach. Lewis Publishers,Michigan.

Mackenzie WR et al (1994). A massiveoutbreak in Milwaukee of Cryptosporidiuminfection transmitted through the publicwater supply. New England Journal ofMedicine. 331(3): 161–167.

Manahan SE (1993). Fundamental ofEnvironmental Chemistry. Lewis Publishers,Boca Raton.

Masoro EJ (1992). Aging and proliferativehomeostasis: Modulation by food restrictionin rodents. Lab Animal Sci 42: 132–137.

Maynard RL, Cameron KM, Fielder R,McDonald A and Wadge A (1995). Settingair quality standards for carcinogens: analternative to mathematical quantitative riskassessment—discussion paper. Human andExperimental Toxicology 14: 175–186.

McClellan RO and Henderson RF (Eds)(1989). Concepts in Inhalation Toxicology.Hemisphere Publishing Corp, New York.

McKone TE (1994). Uncertainty in Variabilityin Human Exposures to Soil ContaminantsThrough Home Grown Food: A MonteCarlo Assessment. Risk Analysis, 14(4):449–463.

McKone TE and Bogen KT (1991). Predictingthe uncertainties in risk assessment.Environmental Science and Technology.October 1991. 25 (10): 1674 (8).

McLean AEM and McLean EK (1969) Diet and toxicity. Br Med Bull 25: 278–28.

McMichael AJ (1991). Setting environmentalexposure standards: current concepts andcontroversies. International Journal ofEnvironmental Health Research 1: 2–13

Mohr U, Dungworth D, Kimmerle G,Lewkowski J, McClellan R and Stöber W(Eds) (1988). Inhalation Toxicology: TheDesign and Interpretation of Inhalation Studiesand their Use in Risk Assessment. ILSIMonograph. Springer-Verlag, New York.


213


Mohr U, Dungworth DL and Capen CC (Eds)(1992). Pathobiology of the Aging Rat. Vol 1.ILSI Press, Washington, DC.

Mohr U, Dungworth DL and Capen CC (Eds)(1994) Pathobiology of the Aging Rat. Vol 2.ILSI Press, Washington, DC.

Mohr U, Dungworth DL, Ward J, Capen CC,Carlton W and Sundberg J (Eds) (1996).Pathobiology of the Aging Mouse. Vols 1 and 2.ILSI Press, Washington, DC.

Monro AM (1990). Interspecies comparisons intoxicology: the utility and futility of plasmaconcentrations of the test substance. RegulToxicol Pharmacol 12: 137–160.

Monro A (1993). How useful are chronic (life-span) toxicology studies in rodents inidentifying pharmaceuticals that pose acarcinogenic risk to humans? Adverse DrugReact Toxicol Rev 12: 5–34.

Moolgavkar SH, Moller H and Woodward A(1999). Principles of the epidemiologicalapproach to QEP. In: Quantitative Estimationand Prediction of Human Cancer Risks(ed. Moolgavkar SH, Krewski L, Zeise L etal). International Agency for Research onCancer, Scientific Publication 131, Lyon.

Motulsky H (1995). Intuitive Biostatistics.Oxford University Press, Oxford.

Mueller PW, Smith SJ, Steinberg KK and ThunMJ (1989). Chronic renal tubular effects inrelation to urine cadmium levels.Nephron 52: 45–54.

Mundt KA, Tritschler JP, Dell LD (1998).Validity of epidemiological data in riskassessment applications. Human andEcological Risk Assessment. 4:675–683.

Munro IC (1977). Considerations in chronictoxicity testing: The chemical, the dose, thedesign. J Environ Path Toxicol 1: 183–197.

Murphy BL (1998). Dealing with theuncertainty in risk assessment. Human andEcological Risk Assessment. 4 (3): 685–699.

Nadkarni RA (1991). The Quest for Quality inthe Laboratory, Analytical Chemistry, Vol. 63,No. 13, July, p. 675A.

National Academy of Sciences (NAS) (US)(1977). Drinking Water and Health (Vol 1).Safe Drinking Water Committee. NationalResearch Council. National Academy ofSciences. Washington, D.C.

NAS (US) (1986). Drinking Water and Health(Vol. 6). Safe Drinking Water Committee.National Research Council. NationalAcademy of Sciences. National AcademyPress. Washington, D.C.

NAS (US) (1987). Pharmacokinetics in RiskAssessment—Drinking Water and Health (Vol. 8). Safe Drinking Water Committee.National Research Council. NationalAcademy of Sciences. National AcademyPress. Washington, D.C.

NAS (US) (1989). Drinking Water and Health—Selected Issues in Risk Assessment (Vol. 9). Safe Drinking Water Committee.National Research Council. NationalAcademy of Sciences. National AcademyPress. Washington, D.C.

NAS (1991). Human Exposure Assessment forAirborne Pollutants. National Academy Press.Washington.

NAS and National Research Council (NRC)(1994). Science and Judgement in RiskAssessment. National Academy Press,Washington, D.C.

National Environment Protection Council(NEPC) (1999). Draft Health RiskAssessment methodology. In: Draft NationalEnvironment Protection Measure for theAssessment of Site Contamination. NationalEnvironment Protection Council. Adelaide.

National Health and Medical Research Council(NHMRC). (1987) Report of the one hundredand third session. June 1987. National Healthand Medical Research Council, p. 30.

214

NHMRC (1990). Australian Guidelines forRecreational Use of Water. National Healthand Medical Research Council. AustralianGovernment Publishing Office, Canberra.

NHMRC (1996). Ambient Air Quality GoalsRecommended by NHMRC and InterimNational Indoor Air Quality GoalsRecommended by NHMRC, May 1996.NHMRC website:www.nhmrc.health.gov.au/publicat/eh-home.htm

NHMRC (1997). Draft Cancer Risk Assessmentfor Environmental Contaminants NationalHealth and Medical Research Council,Canberra. This document was released forpublic comment. The revised draft has beenretitled Toxicity Assessment Guidelines forCarcinogenic Soil Contaminants.

NHMRC (1999). A Guide to the Development,Implementation and Evaluation of ClinicalPractice Guidelines. Australian GovernmentPublishing Service, Canberra.

NHMRC (1999). Toxicity Assessment forCarcinogenic Soil Contaminants Canberra:National Health and Medical ResearchCouncil. (previously titled Draft Cancer RiskAssessment for EnvironmentalContaminants).

NHMRC (2000). How to review the evidence:systematic identification and review of thescientific literature. Commonwealth ofAustralia, Canberra.

NHMRC and Agricultural and ResourceManagement Council of Australia and NewZealand (ARMCANZ) (1996). AustralianDrinking Water Guidelines—National WaterQuality Management Strategy. NationalHealth and Medical Research Council,Agricultural and Resource ManagementCouncil of Australia and New Zealand.

NHMRC, ARMCANZ and ANZECC (1996).Draft Guidelines for Sewerage Systems: Use ofReclaimed Water. National Health andMedical Research Council, Agricultural andResource Management Council of Australiaand New Zealand. CommonwealthGovernment, Canberra.

National Research Council (NRC) (1983). RiskAssessment in the Federal Government:Managing the Process. National AcademyPress, Washington DC.

NRC (1992). Biological Markers inimmunotoxicology. National Academy Press.Washington D.C.

NRC (1994). Science and judgement in riskassessment. Committee on Risk Assessment ofHazardous Air Pollutants. National AcademyPress, Washington.

New Zealand Ministry of Health (1995).Drinking Water Standards of New Zealand,Wellington.

National Committee on Radiation Programs(NCRP) (1996). NRCP Commentary No.14. A guide for uncertainty analysis in doseand risk assessments related to scientificcontamination. National Committee onRadiation Programs, Washington.

National Environment Protection Council(2000). Report of the Risk Assessment Taskforce.National Environment Protection CouncilService Corporation, Adelaide.

Nolen G (1972). Effects of various restricteddietary regimens on the growth, health andlongevity of albino rats. J Nutr 102: 1477–1494.

Nurminen M (1995). Linkage failures inecological studies. World Health StatisticalQuarterly. 48:78–84.

Organisation for Economic Cooperation andDevelopment (OECD) (1982). GoodLaboratory Practice in the Testing of Chemicals.Organisation for Economic Cooperation andDevelopment, Paris, p. 62


215


OECD (1998). OECD Guidelines for the Testingof Chemicals. Section 4: Health Effects. Vol 2,10th Addendum, October 1998. Organisationfor Economic Cooperation andDevelopment, Paris.

Olin SS (ed) (1999). Exposure to Contaminantsin Drinking Water—Estimating UptakeThrough the Skin and by Inhalation. ILSIPress/CRC Press, Boca Raton, Florida.

Pagano RR (1986). Understanding Statistics inthe Behavioural Sciences. West PublishingHouse, St Paul, Minnesota.

Patton DE (1998). Environmental riskassessment: tasks and obligations. Human andEcological Risk Assessment. 4 (3): 657–670.

Paustenbach DJ (1989). A survey of health riskassessment. In: Paustenbach DJ (ed). The riskassessment of environmental and human healthhazards: a textbook of case studies.John Wiley and Son. New York.

Paustenbach DJ (1995). The practice of healthrisk assessment in the United States(1975–1995): How the US and othercountries can benefit from that experience.Human and Ecological Risk Assessment,1: 29–79.

Paynter OE and Schmitt R (1979). Theacceptable daily intake as a quantifiedexpression of the acceptability of pesticideresidues. Advances in Pesticide Science, Part 3,p. 674–679, Pergamon Press, Oxford.

Paynter OE (1984). Oncogenic potential: Guidancefor Analysis and Evaluation of Long TermRodent Studies. Evaluation Procedure #1000.1Office of Pesticide and Toxic Substances,EPA, Washington, D.C.

Perkins JL (1997). Modern Industrial Hygiene.Volume 1. Recognition and Evaluation ofChemical Agents. Van Nostran Reinhold.New York

Pershagen G (1999). Research priorities inenvironmental health. BMJ. 318:1636–1637.

Pfeiffer CJ and Muller PJ (1967). Physiologiccorrelates dependent upon mode of death.Toxicol Appl Pharm 10: 253-260.

Polgar S and Thomas SA (1991). Introduction toresearch in the health sciences. 2nd Edition.Churchill Livingstone, Melbourne.

Pollak JK (1996). Free radical biochemistry ofcontaminants. In: Langley A, Imray P, LockW, and Hill H (eds). The Health RiskAssessment and Management of ContaminatedSites, Proceedings of the Fourth NationalWorkshop on the Health Risk Assessmentand Management of Contaminated Sites.Contaminated Sites Monograph SeriesNumber 7. South Australian HealthCommission, Adelaide, p. 323–338.

Presidential/Congressional Commission on RiskAssessment and Risk Management(P/CCRARM) (1997). Risk Assessment andRisk Management in Regulatory Decision-making. Presidential/CongressionalCommission on Risk Assessment and RiskManagement.

Purchase IFH and Auton TR (1995).Thresholds in chemical carcinogenesis.Regulatory Toxicology and Pharmacology22: 199–205.

Rees N (1999). Current dietary exposureassessment methodology. In: Institute forEnvironment and Health (IEH) Riskassessment strategies in relation to population subgroups. Institute for Environment and Health.Leicester.

Reinert RH and Hroncich JA (1990). SourceWater Quality Management. Chapter 4. p.189–228. Water Quality and Treatment—A Handbook of Community Water Supplies.4th Edition. American Water WorksAssociation. McGraw Hill, New York.

Renwick A (1993). Toxicokinetics. In:Ballantyne B, Marrs T and Turner P (eds).General and applied toxicology. Stockton PressNew York. Vol 1, p. 121–151.

216

Renwick AG (1999) Human variability and riskassessment/safety assurances. In: Institute forEnvironment and Health (IEH) Riskassessment strategies in relation to population sub groups. Leicester: Institute forEnvironment and Health.

Renwick AG and Walker R (1993). An analysisof the risk of exceeding acceptable ortolerable daily intake. RegulatoryToxicologically and Pharmacology, 18: 463–480.

Riegelman RK (1981). Studying a study andtesting a test. Little, Brown and CompanyBoston.

Roberts RG (1992). Overview of similaritiesand differences between children and adults:implications for risk assessment. In: GuzelianPS, Henry CJ, Olin SS (eds) (1992).Similarities and Differences between childrenand adult: implications for risk assessment.Washington: International Life SciencesInstitute.

Rogers AE, Sanchez O, Feinsod FM andNewberne PM (1974). Dietaryenhancement of nitrosamine carcinogenesis.Cancer Res 34: 96–99.

Roseman JM (1998). What can be done toensure the usefulness of epidemiological datafor risk assessment purposes? Human andEcological Risk Assessment. 4:737–746.

Ross MH and Bras G (1965). Tumour incidencepatterns and nutrition in the rat.J Nutr 87: 245–260.

Rothman KJ and Greenland S (1997).Causation and causal inference. In: OxfordTextbook of Pubic Health 3rd edition (ed.Detels et al). Oxford Medical Publications,New York.

Roubicek CB, Pahnish OF and Taylor RL(1964). Growth of rats at two temperatures.Growth 28: 157164.

Ruffle B, Burmaster DE, Anderson PD andGordon HD (1994). Lognormal distributionsfor fish consumption by the general USpopulation. Risk Analysis, 14(4): 395–404.

Rushton L (2000). Reporting of occupationaland environmental research: use and misuseof statistical and epidemiological methods.Occupational and Environmental Medicine,57: 1–9.

Salem H (Ed) (1987). Inhalation Toxicology:Research Methods, Applications and Evaluation.Marcel Dekker Inc, New York and Basel.

Salsburg DS (1986). Statistics for Toxicologists.Marcel Dekker Inc, New York and Basel.

Samet JM (1999). Risk assessment and airpollution. In: Holgate ST, Samet JM, HillelSK, Maynard RL (eds). Air Pollution andHealth. Academic Press, London.

Samet JM, Schnatter R and Gibb H (1998).Invited commentary: Epidemiology and riskassessment. American Journal of Epidemiology*.

Sandman PM (1993). Responding to CommunityOutrage: Strategies for Effective RiskCommunication. American Industrial HygieneAssociation, Fairfax, Virginia.113pp.

Saunders LD, Chen W and Hrudey SE (1997).Studies of hazardous waste sites and humanhealth: evaluation of internal validity.Environmental Review. 5:167–180.

Seefeld MD and Peterson RE (1984). Digestibleenergy and efficiency of feed utilisation inrats treated with2,3,7,8,-tetrachlorodibenzo-p-dioxin. ToxAppl Pharmacol 74: 214–222.

Seigneur C, Constantinou E and Permutt T(1992). Uncertainty analysis of health riskestimates. Doc.Number. 2460-009-510. Paperprepared for the Electric Power ResearchInstitute, Palo Alto.

Sexton K, Beck BD, Bingham E, Brain JD,DeMarini DM, Hertzberg, RC, O’FlahertyEJ and Pounds JG (1995). Chemical mixturesfrom a public health perspective: theimportance of research for informed decisionmaking. Toxicology 105: 429–441.


217


Sharer K (1977). The effect of chronicunderfeeding on organ weights in rats. Howto interpret organ weight changes in cases ofmarked growth retardation in toxicity tests?Toxicology 7: 45–56.

Sharma RP and Coulombe RA Jr (1996).Pharmacokinetics and risk assessment—Chapter 7. In: Fan AM and Chang LW, eds.Toxicology and Risk Assessment. Marcel DekkerInc, New York, p. 81–99.

Silvaggio T and Mattison DR (1994). Settingoccupational health standards. Toxicokineticdifferences among and between men andwomen. J. Occup Med, 36, 849–854.

Singer PC (Ed) (1999). Formation and Control ofDisinfection By-Products in Drinking Water.American Water Works Association. Denver.

Smith RL (1994). Monte Carlo simulation forhuman exposure assessment at a Superfundsite. Risk Analysis, 14(4): 433-439.

Snodgrass WR (1992). Physiological andbiochemical differences between children andadults as determinance of toxic response to anenvironmental pollutant. In: Guzelian PS,Henry CJ, Olin SS (eds) (1992). Similaritiesand differences between children and adults:implications for risk assessment. Washington:International Life Sciences Institute.

Standards Australia (1997). Australian Standard.Guide to the sampling and investigation ofpotentially contaminated soil. Part 1: Non-volatile and semi-volatile compounds.AS 4482.1-1997. Homebush: StandardsAustralia.

Standards Australia (1999). Australia/NewZealand Standard. Risk management.AS/NZS 4360: 1999.

Standards Australia, Standards New Zealand(2000). Environmental risk management—Principles and process. AS/NZS HB 203:2000.

Sumi N, Stavrou D, Frohberg H and JochmannG (1976). The incidence of spontaneoustumours of the central nervous system ofWistar rats. Arch Toxicol 35: 1–13.

Swaim LD, Taylor HW and Jersey GC (1985).The effect of handling techniques on serumALT activity in mice. J Appl Toxicol 5:160–162.

Tal A (1997). Assessing the environmentalmovement’s attitudes towards risk assessment.Environmental Science and Technology, 31 (10):470A–476A.

Tallarida RJ and Murray RB (1987). Manual ofPharmacologic Calculations with ComputerPrograms. 2nd Ed, Springer-Verlag,New York.

Tannenbaum A (1940). Relationship of bodyweight to cancer incidence. Arch Pathol 30:509–517.

Tarone RE (1982). The use of historical controlinformation in testing for a trend inproportions. Biometrics 38: 215–220.

Task Force of Past Presidents (1982). Animaldata in hazard evaluation: Paths and pitfalls.Fund Appl Toxicol 2: 101–107.

Taylor R, Langley A. (1998). Exposure scenariosand exposure settings. NationalEnvironmental Health Forum MonographsSoil Series No. 2. 2nd ed. South AustralianHealth Commission, Adelaide.

Therapeutic Goods Administration (TGA)(1999). ADI List. Acceptable Daily Intakes forAgricultural and Veterinary Chemicals. TGA,Health and Ageing. Australian GovernmentPublishing Services. Canberra.

Thomas SP and Hrudey SE (1997). Risk ofDeath in Canada: What we know and how weknow it. University of Alberta Press.Edmonton.

218

Thompson GE, Scheel LD and Lowe D (1982).Seasonal alteration in response to stress orphysiological change. Drug and Chem. Toxicol5: 189–199.

Thornton KW and Paulsen SG. (1998). Cananything significant come out of monitoring?Human and Ecological Risk Assessment, 4(4):797–805.

Thurman JD, Moeller RB and Turturro A(1995). Proliferative lesions of the testis inad-libitum-fed and food-restricted Fischer F-344 and FBNF1 rats. Lab Animal Sci 45:635–640.

Tschichold J (1935). Asymmetric Typography.Basel.

Tufte ER (1983). The Visual Display ofQuantitative Information. Cheshire,Connecticut: Graphics Press.

Tufte ER (1990). Envisioning Information.Cheshire, Connecticut: Graphics Press.

Tucker MJ (1979). The effect of long-term foodrestriction on tumours in rodents. Int J Cancer23: 803–807.

Tyl RW and Marr MC (1997) In: Hood RD(Ed) Handbook of Developmental Toxicology.CRC Press, Boca Raton.

Tyson CA and Sawhney DS (1985). Organfunction tests in toxicology evaluation. NoyesPublications, Park Ridge, NJ, USA.

US Environmental Protection Agency (USEPA) (1983) Good Laboratory PracticeStandards: Toxicology testing. Fed Reg 48: page53922, and: Pesticide Programs: GoodLaboratory Practice Standards. Final Rule.Fed Reg 46 (No. 230): p. 53946–53969,Tuesday 11/29/83.

US EPA (1983a). Technical Assistance Documentfor Sampling and Analysis of Toxic OrganicCompounds in Ambient Air. US EPA, EPA600/4-83-027, June 1983.

US EPA (1988). 7 Cardinal rules of riskcommunication. OPA-87-020.Washington: USEnvironmental Protection Agency.

US EPA (1989). Risk assessment guidance forSuperfund, Volume 1. Human health evaluationmanual, (Part A). Interim final. EPA/540/1-89/002. Washington: US EnvironmentalProtection Agency, Office of Emergency andRemedial Response.

US EPA (1989a). Exposure Factors Handbook.EPA 600/8-89/043. Washington: USEnvironmental Protection Agency.

US EPA (1990) Compendium of Methods for theDetermination of Air Pollutants in Indoor Air.EPA 600/4-90-010 April 1990.

US EPA (1992). Guidelines for ExposureAssessment; notice. Part VI. Federal Register 29May 1992 57(104): 22888–22938.

US EPA (1992a). Guidance on riskcharacterisation for risk managers. InternalMemorandum by F.Henry Habicht. USEnvironmental Protection agency, Office ofthe Administrator. Washington.

US EPA (1994) Quality Assurance Handbook forAir Pollution Measurement Systems. USEnvironmental Protection Agency. EPA-600/R-94-038b, May 1994.

US EPA (1995) Guidance for risk characterisation.US Environmental Protection AgencyScience Policy Council. Washington.

US EPA (1995a). The use of the benchmark doseapproach in health risk assessment. RiskAssessment Forum, EPA/630/R-94-007,United States Environmental ProtectionAgency, Washington D.C.

US EPA (1996). Draft Revision to the guidelinesfor carcinogenic risk assessment. Office ofHealth and Environmental Assessment,Office of Research and Development, UnitedStates Environmental Protection Agency,Washington D.C.

US EPA (1997). Guiding principles for MonteCarlo Analysis, Risk Assessment Forum,United States Environmental ProtectionAgency, Washington D.C.


219


US EPA (1998). National Primary DrinkingWater Regulations: Interim Enhanced SurfaceWater Treatment: Final Rule. FederalRegister 40 CFR Parts 9, 141 and 142.

US EPA (1999). Compendium of Methods for theDetermination of Inorganic Compounds inAmbient Air. EPA 625/R-96/010a, Office ofResearch and Development, US EPA. June1999. Washington D.C.

Vainio H and Wilbourn J (1992). Identificationof carcinogens within the IARC Monographprogram. Scand. J. Work Environ. Health, 18(Suppl 1): 64-73.

Vainio H, Heseltine E, McGregor D, Tomatis Land Wilbourn J (1992). Working group onmechanisms of carcinogenesis and theevaluation of carcinogenic risks. Cancer Res.,52: 2357-2361.

Van Alphen M (1993). A format for site historyreports. In: Langley A J, van Alphen M. TheHealth Risk Assessment and Management ofContaminated Sites. Proceedings of theSecond National Workshop. South AustralianHealth Commission. Adelaide.

Van den Berg M, Birnbaum L, Bosveld ATC,Brunstrom B, Cook P, Feeley M, et al (1998).Toxic Equivalency Factors (TEFs) for PCBs,PCDDs, PCDFs for humans and wildlife.Environmental Health Perspectives 106: 775–792.

Victorian EPA (1997). National framework forecological risk assessment of contaminated sites.Part A. Environment Australia. (VictorianEnvironment Protection Authority).Canberra.

Van den Berg M, Birnbaum L, Bosveld ATC,Brunstrom B, Cook P, Feeley M et al (1998).Toxic Equivalency Factors (TEFs) for PCBs,PCDDs, PCDFs for humans and wildlife.Environmental Health Perspectives 106:775–792.

Walach JB (1996). Interpretation of diagnostictests. 6th Ed. Little Brown, Boston.

Wallace LA and Ott WR (1982). PersonalMonitors. A State-of -the-Art Survey. J.AirPollut. Control Assoc. 32, 601.

Waner T (1992). Current statistical approachesto clinical pathology data from toxicologicalstudies. Toxicologic Pathology 20(3), Part 2.

Weil CS (1970). Significance of organ-weightchanges in food safety evaluation. In:Metabolic Aspects of Food Safety. p. 419–454,Academic Press, NY.

Weil CS (1972). Statistics vs. safety factors andscientific judgement in the evaluation ofsafety for man. Toxicol Appl Pharmacol 21:454–463.

Weil SC and McCollister DD (1963).Relationship between short-and-long termfeeding studies in designing an effectivetoxicity test. Agr Food Chem 11: 486–491.

Weingand K, Brown G, Hall R, Davies D,Gossett K, Neptun D, Waner T, MatsuzawaT, Salemink P, Froelke W, Provost JP,Dal-Negro G, Batchelor J, Nomura M,Groetsch H, Boink A, Kimball J, WoodmanD, York M, Fabianson-Johnson E, Lupart Mand Melloni E (1996). Harmonization ofanimal clinical pathology testing in toxicity andsafety studies. The Joint Scientific Committeefor International Harmonization of ClinicalPathology Testing.

Weisel CP and W-K Jo (1996). Ingestion,inhalation and dermal exposures tochloroform and trichloroethene from tapwater. Environmental Health Perspectives.104: 48–51.

Western Australian Department ofEnvironmental Protection (WA DoEP)(1999). Air Quality and Air PollutionModelling Guidelines. June 1999. WA DoE,Australia.

220

World Health Organization (WHO) (1972).Evaluation of certain food additives and thecontaminants mercury, lead, and cadmium.Sixteenth report of the Joint FAO/WHOExpert Committee on Food Additives( JECFA). World Health OrganizationTechnical Report Series 505. World HealthOrganization, Geneva.

WHO (1978). Principles and Methods forEvaluating the Toxicity of Chemicals.Environmental Health Criteria 6.International Programme on Chemical Safety(IPCS)/WHO. Geneva.

WHO (1986). Early detection of occupationaldiseases. Geneva: World Health Organization,Geneva.

WHO (1987). Principles for the safety assessmentof food additives and contaminants in food.Environmental Health Criteria 70,IPCS/WHO. Geneva.

WHO (1989). Evaluation of certain foodadditives and contaminants. Thirty third reportof the Joint FAO/WHO Expert Committeeon Food Additives ( JECFA). World HealthOrganization Technical Report Series 776.World Health Organisation, Geneva.

WHO (1990). Principles for the ToxicologicalAssessment of Pesticide Residues in Food,Environmental Health Criteria 104,IPCS/WHO. Geneva.

WHO (1991). Investigating EnvironmentalDisease Outbreaks—A Training Manual.Environmental and OccupationalEpidemiology Series. WHO/PEP/91.35.World Health Organization. Geneva.

WHO (1992). Quality Management for ChemicalSafety Testing, Environmental Health Criteria141, IPCS/WHO. Geneva.

WHO (1993). Biomarkers and risk assessment:concepts and principles. Environmental HealthCriteria 155. IPCS/WHO. Geneva.

WHO (1993a). Guidelines for drinking waterquality. Volume 1 Recommendations. SecondEdition, WHO. Geneva.

WHO (1994). Assessing human health risks ofchemicals: derivation of guidance values forhealth based exposure limits. EnvironmentalHealth Criteria 170. IPCS/WHO. Geneva.

WHO (1997). WHO/FAO/IASEA Report onTrace Elements in Human Nutrition andHuman Health. WHO. Geneva.

WHO (1997). Food Consumption and RiskAssessment of Chemicals. Report of aFAO/WHO Consultation, Geneva, 10-14Feb 1997. (WHO/FSF/FOS/97.5). WorldHealth Organization. Geneva.

WHO (1999). Chemical Risk Assessment.Training Module No. 3 WHO/PCS/99.2.World Health Organization. Geneva.

WHO (1999a). Guidelines for Air Quality. WorldHealth Organization. Geneva.

WHO (1999b). Principles for the Assessment ofrisks to human health of chemicals.Environmental Health Criteria 210.IPCS/WHO. Geneva.

Yassi A, Kjellstrom T, de Kok T, Guidotti T. (inpress). Basic Environmental Health. WorldHealth Organization, United NationsEnvironment Programme, United NationsEducation Science and CulturalOrganization, Council of Rectors ofEuropean Universities.

Zbinden G (1979). The role of pathology intoxicity testing. In: Progress in Toxicology 2,p. 8–18, Springer Verlag, New York.


221


Acceptable daily intakes (ADIs)determination of 143exceedances of 132

Air monitoring 22Air pollutants

Environmental Health Risk Assessment 160Air quality

guidelines 167Aluminium 22Analysis

environmental 26,93Analytical methodologies 96, 156, 161

food 171Assessment

appraisal of 135key aspects 135presentation 139

Assessment criteriabasis 106

Asthma 22Australian model

chemical risk assessment 183occupational risk assessment 184risk characterisation 185toxicological data 184

Bayesian tools 4Benchmark Dose 80Biological monitoring 116

breast milk 118choice of test 118choice of tissue 116hair 118toenails 118

Biological plausibility and coherence 199Biomarkers 120Blood

biological monitoring 117Body weight 122

changes 39Cancer 198

biological plausibility 199coherence 199dose–response relationship 198other modes of action 199postulated mode of action 199temporal association 198

Cancer risk assessmentdata gaps 199decision tree 150inconsistencies 199key events 198uncertainty 199

Carcinogens 47Australian model 145

Causation 56, 57, 58‘Necessary’ cause 56‘Sufficient’ cause 56

Cause and outcome 56Censored data 102, 171, 180

food 171levels of reporting 102water 180

Chain of custody 95Chemical products

assessment 189Chemical review 190Chemical risk assessment

Australian model 183food 169

Chemicalsabsorption 11agricultural 186distribution 12elimination 12existing 184metabolism 12new 185veterinary 186

Chemicals unitassessment process 192role 189

Children 9Choice of analytes 96Clinical

chemistry 39observations 38

Community consultation xvi, 17Conservatism 8Contaminated sites

composite sampling 155data collection 154environmental health risk assessment 154sampling 154sampling density 155

CRIHS 190Criteria

environmental health 141principles for setting 141Risk-Based Environmental

Health, Determination of 143Criteria Pollutants 163Daily fluid intake 122

222

Index

Datacollection 136metabolism 43toxicokinetic 43

Data collection reportintegration 159content 159

Data quality objectives 93Data requirements

statutory process 188Data sets

selecting 112Database

exposure 188toxicological 188

Default values 122Dietary intake 122Dioxins 83Dispersion Modelling 163Dose–response assessment xiii, 6, 74, 180

water 180Dose–response

characterisation 88Dose–response curves 76Dose–response relationship

cancer 198Dosing

regimen 33route 34

Drinking water 173exposure to 181guidelines 174health risks generic rationale 178

Emissions 22Environmental analysis 26Environmental distribution 92Environmental hazards

identification 25Environmental health

criteria 141hazards 25

Environmental Health Risk AssessmentPrinciples xiv

Environmental monitoring 115Environmental persistence 93Environmental sampling 26

accreditation of laboratories 104analytical techniques 104chain of custody 104choice of analytes 104liaison with laboratories 104

Epidemiological data 162

Epidemiological studytypes of 52

Epidemiology 50bias 51confounding 51dose–response assessment 59exposure assessment 61hazard identification 59observational studies 55observational study designs 55risk characterisation 62study designs 54

Equations 139Evaluation

published research 62, 68Experimental Toxicology

strengths and limitations 59Expired air 118Exposure assessment xii, 6, 90, 137, 163, 180

appraising 124components of 92default values 122food 171reports 125short term appraisal of 133sources of data 124volatile agents 121water 180

Exposure durations 132Exposure factors

estimating distributions 111Exposure standards 188Field instruments 96Food

analytical methodologies 171censored data 171chemical risk assessment 169environmental health risk assessment 169exposure assessment 171hazard identification 169microbiological risk assessment 170particular issues 169risk characterisation 172uncertainty sources 172

Gender 11General appraisal of assessments 135Hazard analysis

completion 46Hazard assessment xiii, 6, 50, 74

considerations 82dose–response assessment 6hazard identification 6, 30


223


human variability 83methodologies 76non-threshold approach 78species differences 82threshold approaches 77

Hazard identification xiii, 6, 50, 87, 137, 161, 169carcinogens 47food 169toxicology 30water 177

Hazard identification report 48layout and formatting 48study identification 48

Hazards 26Health impact assessment

risk assessment 7Health monitoring 121Health studies 69

nature 70protocol 72quality 72type 1 70type 2 71

Hormesis 84Interactions 26International harmonisation 190Issue identification 4, 24

limitations 28Latin hypercube 109Levels of reporting 157

water 180Lifestyle factors 13Limitations 28Mechanistically-derived models 79Meteorological data 103Microbiological risk assessment 170, 200

food 170general principles 200paradigms and frameworks 201

Mixtures 83Modelling exposures 107Models

mechanistic 80statistical 80

Modified Benchmark Dose 145Monitoring methodologies

water 179Monte Carlo 107, 108

adminsitrative requirements 115principles 112techniques in Australia 109weaknesses 109

Mortality 37Multiple barrier approach 179

water 179NICNAS 184, 185NO(A)EL

determination of 143NOHSC 186, 188Observational Epidemiology

Strengths and Limitations 59Occupational Criteria 162Occupational data 45

extrapolating 45Occupational risk assessment

Australian model 184Older persons 10Organ weights 41Parameters

Water Quality 175PCBs 83Personal monitoring 115Point estimates 107Polycyclic Aromatic Hydrocarbons 83Post mortem 42Postulated mode of action 198

assessment 199Precautionary principle 7Pregnancy

absorption 11distribution 12elimination 12metabolism 12

Presentationassessment and report 139

Presentation of data 99contouring 100graphical representation 100mapping of data 100

Probability distributions 107Quality assessment 35Quality Assurance 97, 105

blanks 97data evaluation 106laboratory 106quality control 97recovery check 97reference material 97replicate analysis 97surrogate spikes 98

Quality control 105data evaluation 106laboratory 106

224

Reclaimed water 177Recreational water 176Report 139

analytical issues 104data collection 159environmental sampling and analysis 103hazard identification 48integration 103presentation 139sampling issues 104

Reproductive status 13Respiratory volumes 122Results 107Risk

Extrapolation 80Risk assessment 1

aims 6assessing 3Australian models 3, 183Bayesian tools 4Canada 194cancer 198chemical 183children 10context 17dose–response assessment 6exposure assessment 6gender 11hazard assessment 6hazard identification 6health impact statement 7individual 2international models 194issue identification 4key factors 9lifestyle factors 13methods 3microbiological 200models 3, 5occupational 184older persons 10population 2populations 9precautionary principle 7principles 7qualitative 3quantitative 3reproductive status 13risk characterisation 6types 2United Kingdom 195United States 194when to undertake 2

Risk assessment and particular population groups 9Risk characterisation xiv, 6, 126, 138,

164, 172, 182, 185Australian model 185food 172key principles 127qualitative 128quantitative 128water 182

Risk communication xvi, 19Risk conclusions 129Risk Management xviRisk perception 19Safety plans 98, 156Sample handling

storage and transport 95Sampling

and analysis plan 105composite 155data interpretation 136density 95, 155environmental 26, 93grid (systematic) 154judgemental 154methodology 94, 105patterns 94random 154strategies 94stratified 154stratified random 154

Scaling of dosesinterspecies 44other factors 44route-to-route 44

Scheduled Wastes 18Site inspections 105Situation descriptions 104Soil exposures 123Sources of uncertainty

food 172Specific appraisal 136Standard setting

factors 188Statistical Analysis, critical evaluation 64Statistical data

assessment 99,156assessment—water 180presentation 99, 156presentation—water 171summary 171tests 45water 180


225


Statutory processdata requirements 188outline 188

Study identification 48Study parameters

analysis 36evaluation 36

Subjective Terms 136Summary statistics 99Survival 37TDI

determination of 143Temporal association 198Threshold approach 77, 78

non-threshold approach 78Tissue

choice 116Tolerable intake 88Tolerable intake data

level 1 sources 89level 2 sources 89

Toxic equivalency factors 83Toxicity studies

acute 32, 42analysis 36chronic 32developmental 32, 43evaluation 36genotoxicity 33reproductive 32, 42special 42, 43sub-chronic 32

Toxicity testing 32important issues 33modes of action 32study protocol 33

Toxicological appraisalschecklist 87

Toxicological data 88Australian model 184level 1 sources 89level 2 sources 89level 3 sources 89

Tumour response 198Uncertainty 8, 28, 110, 129, 199

cancer 199sources of 172

Urinary measurements 39Urine

biological monitoring 117Variability 7, 110Volatile agents 121Water

censored data 180dose–response assessment 180environmental health risk assessment 173exposure assessment 180hazard identification 177identifying the issues 173levels of reporting 180monitoring methodologies 179multiple barrier approach 179risk characterisation 182

Weight-of-evidence 46

226

enHealth Council Membership and Terms of Reference

The enHealth Council is the premier advisorybody on environmental health in Australia. Itprovides national leadership on environmentalhealth issues, sets priorities, coordinates nationalpolicies and programs and provides a pivotal linkbetween international fora and environmentalhealth stakeholders in Australia. It is alsoresponsible for the implementation of theNational Environmental Health Strategy.

MembershipChair—agreed by the Australian HealthMinisters Conference

State and Territory Health Departmentrepresentatives:

• Australian Capital Territory—ManagerHealth Protection Service

• New South Wales—Director EnvironmentalHealth

• Northern Territory—Program DirectorEnvironmental Health

• Queensland—Manager EnvironmentalHealth

• South Australia—Director EnvironmentalHealth

• Tasmania—Director Environmental andPublic Health

• Victoria—Manager Environmental Health

• Western Australia—Director EnvironmentalHealth Service

• New Zealand—New Zealand HealthMinistry

Commonwealth Department of Health andAgeing representative—Director ofEnvironmental Health

Aboriginal and Torres Strait Islander Commission

Australian Consumers’ Association

Australian Institute of Environmental Health—National President

Environment Australia

National Indigenous Environmental HealthForum—Chair

Public Health Association of Australia

Secretariat services provided by theEnvironmental Health Section of theCommonwealth Department of Health and Ageing.

Terms of Reference1. Provide national leadership on environmental

health issues by:

i) coordinating and facilitating environmentalhealth policies and programs

ii) establishing strategic partnershipsbetween environmental healthstakeholders

iii) setting priorities for national environmentalhealth policies and programs

iv) providing an open consultative system for policy development

v) facilitating cost effective use ofenvironmental health resources

2. Drive the implementation of NationalEnvironmental Health Strategy

3. Advise the Commonwealth, States andTerritories, Local government and otherstakeholders on national environmentalhealth issues

4. Coordinate the development ofenvironmental health action plans at local,state and national levels.

5. Promote and develop model environmentalhealth legislation, standards, codes of practice,guidelines and publications.

6. Strengthen the national capacity to meet currentand emerging environmental health challenges.

7. Provide a pivotal link between internationalfora and environmental health stakeholders inAustralia and strengthening Australia’scollaboration with countries in the Asia-Pacific region


227

envhazards

Documents

increasing cardiac output

mr sam mangas

total bile acids

dr andrew langley

world health organization

tolerable daily intake

ipcs conceptual framework

public health interventions