INTERNATIONAL PROGRAMME ON CHEMICAL SAFETY
ENVIRONMENTAL HEALTH CRITERIA 27
GUIDELINES ON STUDIES IN ENVIRONMENTAL EPIDEMIOLOGY
This report contains the collective views of an international group of
experts and does not necessarily represent the decisions or the stated
policy of the United Nations Environment Programme, the International
Labour Organisation, or the World Health Organization.
Published under the joint sponsorship of
the United Nations Environment Programme,
the International Labour Organisation,
and the World Health Organization
World Health Orgnization
Geneva, 1983
The International Programme on Chemical Safety (IPCS) is a
joint venture of the United Nations Environment Programme, the
International Labour Organisation, and the World Health
Organization. The main objective of the IPCS is to carry out and
disseminate evaluations of the effects of chemicals on human health
and the quality of the environment. Supporting activities include
the development of epidemiological, experimental laboratory, and
risk-assessment methods that could produce internationally
comparable results, and the development of manpower in the field of
toxicology. Other activities carried out by the IPCS include the
development of know-how for coping with chemical accidents,
coordination of laboratory testing and epidemiological studies, and
promotion of research on the mechanisms of the biological action of
chemicals.
ISBN 92 4 154087 7
The World Health Organization welcomes requests for permission
to reproduce or translate its publications, in part or in full.
Applications and enquiries should be addressed to the Office of
Publications, World Health Organization, Geneva, Switzerland, which
will be glad to provide the latest information on any changes made
to the text, plans for new editions, and reprints and translations
already available.
(c) World Health Organization 1983
Publications of the World Health Organization enjoy copyright
protection in accordance with the provisions of Protocol 2 of the
Universal Copyright Convention. All rights reserved.
The designations employed and the presentation of the material
in this publication do not imply the expression of any opinion
whatsoever on the part of the Secretariat of the World Health
Organization concerning the legal status of any country, territory,
city or area or of its authorities, or concerning the delimitation
of its frontiers or boundaries.
The mention of specific companies or of certain manufacturers'
products does not imply that they are endorsed or recommended by the
World Health Organization in preference to others of a similar
nature that are not mentioned. Errors and omissions excepted, the
names of proprietary products are distinguished by initial capital
letters.
CONTENTS
PREFACE
1. INTRODUCTION
1.1. Interrelationships with toxicological studies
1.2. Design
1.3. Environmental agents and assessment of exposures
1.4. Effects on health
1.5. Organization and conduct
1.6. Analysis and interpretation of results
1.7. Uses of epidemiological information
REFERENCES
2. STUDY DESIGNS
2.1. Introduction
2.2. Preliminary review of state of knowledge
2.3. Descriptive studies and use of existing records
2.3.1. Mortality statistics
2.3.2. Morbidity statistics
2.3.3. Populations at risk
2.3.4. Geographical differences in mortality and morbidity
2.3.5. Time trends
2.3.6. Associations with environmental indices
2.3.7. Case registers
2.3.8. General surveys
2.4. Formulation of hypotheses
2.5. Cross-sectional studies
2.6. Prospective and follow-up studies
2.7. Retrospective cohort studies
2.8. Time-series studies
2.9. Case-control studies
2.10. Controlled exposure studies
2.11. Monitoring and surveillance
REFERENCES
3. ASSESSMENT OF EXPOSURE
3.1. Introduction
3.2. Exposure and dose
3.2.1. Systematic agents
3.2.2. Local exposure
3.2.3. Physical factors
3.3. Combined exposure, physical and chemical
interactions
3.3.1. Same agent, various sources
3.3.2. Various agents, same source
3.3.3. Various agents, various sources
3.3.4. Impurities
3.3.5. Interactions
3.4. Qualitative assessment of exposure
3.5. Environmental assessment of exposure
3.5.1. Quality of data
3.5.2. Monitoring strategy for air pollutants
3.5.2.1 What to sample, how long, how frequently?
3.5.2.2 Representativeness
3.5.3. Monitoring of pollutants in food and water
3.5.3.1 Overall assessment of dietary
intake of toxic elements
3.5.3.2 Indirect assessment of intake
3.5.3.3 Direct assessment of intake
3.5.4. Monitoring of physical factors
3.5.4.1 Noise
3.5.4.2 Vibration
3.5.4.3 Ionizing radiation
3.5.4.4 Non-ionizing radiation
3.6. Personal sampling
3.7. Biological assessment of exposure
3.7.1. Advantages, disadvantages, limitations
3.7.2. Collection for future reference
3.7.3. Index specimens for various pollutants
3.7.4. Example of environmental versus biological
assessment of exposure: inorganic lead
3.7.4.1 Lead in blood (Pb-B)
3.7.4.2 Lead in urine (Pb-U)
3.7.4.3 Lead in faeces (Pb-F)
3.7.4.4 Lead in deciduous teeth (Pb-T)
3.8. Assessment of the subjective environment
3.8.1. Assessment of odour
3.8.2. Assessment of taste
3.8.3. Example of sensory assessment of drinking-water
3.9. Interindividual and intergroup variability in
exposure: population at risk
3.10. Outdoor/indoor exposure
3.11. Time-weighted exposure
REFERENCES
4. HEALTH EFFECTS, THEIR MEASUREMENT AND INTERPRETATION
4.1. Introduction
4.1.1. General comments on effects
4.1.2. General comments on measurements of effects
4.1.2.1 Inter- and intrainstrument variation
4.1.2.2 Inter- and intralaboratory differences
4.1.2.3 Inter- and intraobserver variations
4.2. Mortality and morbidity statistics
4.2.1. Mortality statistics
4.2.2. Routine morbidity statistics
4.3. Cancer
4.3.1. Cancer and environmental factors
4.3.2. Measurements of cancer
4.3.2.1 Incidence and mortality rate
4.3.2.2 Variations of incidence with age
4.3.2.3 Geographical differences
4.3.2.4 Cancer and life-style
4.3.2.5 Cancer in migrants
4.3.2.6 Time trends
4.3.2.7 Correlation studies
4.3.2.8 Hospital data
4.3.2.9 Cancer and occupation
4.3.2.10 Case reports
4.3.2.11 Epidemiological uses of pathological findings
4.4. Respiratory and cardiovascular effects
4.4.1. Symptom questionnaires
4.4.2. Tests of system function
4.4.3. Standardization of methods
4.4.4. Radiographic measurements
4.4.5. Hypersensitivity measurements
4.4.6. Example: Effects of manganese on
respiratory and cardiovascular systems
4.5. Effects on nervous system and organs of sense
4.5.1. Central and peripheral nervous systems
4.5.2. Ear: Effects of sound
4.5.3. Eye and vision
4.6. Behavioural effects
4.6.1. Effects of environmental exposure
4.6.2. Indicators and measurements of effects
4.6.3. Interpretation of data
4.7. Haemopoietic effects
4.7.1. Environmental agents inducing direct toxic
effects in the haematological system
4.7.2. Environmental agents inducing indirect
toxic effects in the haematological system
4.7.3. Measurements and their interpretation
4.7.4. Example: Effects of low lead
concentrations on workers' health
4.8. Effects on the musculoskeletal system and growth
4.8.1. Effects of environmental exposure
4.8.2. Identification of effects
4.8.3. Intrinsic liability
4.8.4. Extraneous influences
4.8.5. Development states
4.8.6. Example: Endemic fluorosis
4.9. Effects on skin
4.9.1. Environmentally caused skin diseases
4.9.2. Epidemiological methods of study
4.10. Reproductive effects
4.10.1. Effects on reproductive organs
4.10.2. Genetic effects
4.10.2.1 Assessment of genetic risks
4.10.3. Fetotoxic effects
4.10.3.1 Measurement of fetotoxic effects
4.10.4. Registries of genetic diseases and
malformations
4.10.5. Example: EEC study of congenital
malformations
4.11. Effects on other major internal organs
4.11.1. Renal system
4.11.1.1 Detection of renal diseases
4.11.2. Bladder
4.11.3. Gastrointestinal tract
4.11.3.1 Oesophagus
4.11.3.2 Stomach and duodenum
4.11.3.3 Intestines
4.11.4. Liver
4.11.5. Pancreas
REFERENCES
5. ORGANIZATION AND CONDUCT OF STUDIES
5.1. Introduction
5.2. Study protocol
5.2.1. Description of problems and hypothesis formulation
5.2.2. Description of methods
5.2.3. Evaluation of institutional-based data sources
5.2.4. Analysis and reporting of data
5.2.5. Resources required
5.2.6. Studies in developing countries
5.3. Ethical and legal considerations
5.3.1. Medical confidentiality
5.4. Time schedule of study
5.4.1. Preparatory phase
5.4.2. Pilot study
5.4.3. Main study
5.5. Composition of the study team
5.5.1. Team leadership and epidemiology
5.5.2. Clinical specialist
5.5.3. Statistical expertise
5.5.4. Environmental scientists
5.5.5. Interviewers and technicians
5.5.6. Support staff
5.5.7. Special considerations for developing countries
5.5.8. Example: Study teams of Itai-Itai disease
and chronic cadmium poisoning
5.6. Implementation of study
5.6.1. Arrangements with local authorities and study population
5.6.2. Picking samples
5.6.2.1 Example: Sampling procedures
5.6.3. Designing recording forms and questionnaires
5.6.4. Planning for control of data and computer programming
5.6.5. Training of personnel
5.6.6. Pilot study
5.6.6.1 Example: Testing of spirometers
and assessment of observer error
5.6.6.2 Example: Assessment of X-ray observer error
5.6.7. Main study
5.6.7.1 Advance contact
5.6.7.2 Interview studies
5.6.7.3 Medical and laboratory examinations
5.6.7.4 Environmental measurements
5.6.7.5 Linkage and evaluation of data
5.6.7.6 Reporting of results
5.6.8. Examples of cohort studies
5.6.8.1 Michigan polybrominated biphenyls study
5.6.8.2 Study on air pollution and
adverse health effects in Bombay
5.6.8.3 Tucson chronic obstructive lung disease study
5.6.8.4 The Tecumseh community health study
5.6.8.5 Late effects of atomic bomb radiation
5.7. International collaborative studies
5.7.1. Study protocol and timetable
5.7.2. Organizational and sampling procedures
5.7.3. Questionnaires
5.7.4. Standardization of measurement instruments
and methods and quality assurance
5.7.5. Reporting forms
REFERENCES
6. ANALYSIS, INTERPRETATION AND REPORTING
6.1. Introduction
6.2. Data preparation
6.2.1. Coding
6.2.2. Key punching
6.2.3. Data monitoring and editing
6.3. Data description (or reduction)
6.3.1. Purpose
6.3.2. Frequency distributions and histograms
6.3.3. Bivariate distributions and scattergrams
6.3.4. Discrete variables and contingency tables
6.3.5. Independent and related data
6.3.6. General points on tables and graphs
6.3.7. Summary statistics and indices
6.3.7.1 Averages
6.3.7.2 Scatter (or dispersion)
6.3.7.3 Morbidity and mortality indices
6.3.7.4 Standardization
6.3.7.5 Proportional mortality
6.3.7.6 Relative risk and attributable risk
6.3.7.7 Concluding remarks about summary
statistics and indices
6.4. Analysis and interpretation
6.4.1. Statistical ideas about the interpretation of data
6.4.2. Errors
6.4.3. Analysing results from cross-sectional studies
6.4.3.1 Qualitative data
6.4.3.2 Quantitative data: response and
explanatory variables
6.4.3.3 Statisticians and computers
6.4.3.4 Analysis of variance
6.4.3.5 Correlations
6.4.3.6 Multiple regression
6.4.3.7 Additive linear models
6.4.3.8 More complicated models
6.4.3.9 Dummy variables
6.4.3.10 Selection of variables
6.4.3.11 Evaluating "goodness of fit"
6.4.3.12 Evaluating the stability of models
6.4.3.13 Predicted normal values
6.4.3.14 Other methods for studying multivariate data
6.4.4. Analysis of data from prospective and
follow-up studies
6.4.4.1 Nomenclature
6.4.4.2 Time as a measured variable
6.4.4.3 Person-years method
6.4.4.4 Modified life-table method
6.4.4.5 Overlap of exposure and observation periods
6.4.4.6 Lagged exposures
6.4.4.7 Measures of latency
6.4.4.8 Some analytical techniques
6.4.5. Analysis of data from case-control studies
6.4.5.1 Relative and absolute risks
6.4.5.2 Relation between prospective and
case-control studies
6.4.5.3 Analysis of stratified samples
6.4.5.4 Analysis of matched samples
6.4.5.5 Effect of ignoring the matching
6.4.5.6 Alternative methods of analysis
6.4.6. Drawing conclusions from analyses
6.5. Reporting
6.5.1. The variety of epidemiological reports
6.5.2. Main scientific report
6.5.2.1 Introduction
6.5.2.2 Methods
6.5.2.3 Results
6.5.2.4 Discussion
6.5.2.5 Abstract
6.5.3. Non-technical reports
REFERENCES
7. USES OF EPIDEMIOLOGICAL INFORMATION
7.1. Introduction
7.2. Communication with the public
7.3. Important features and limitations of epidemiological information
7.4. Standard setting
7.4.1. Factors in standard setting
7.4.2. Interim standards
7.5. Assessment of effectiveness of control measures taken
7.6. Policy of openness
REFERENCES
NOTE TO READERS OF THE CRITERIA DOCUMENTS
While every effort has been made to present information in
the criteria documents as accurately as possible without
unduly delaying their publication, mistakes might have
occurred and are likely to occur in the future. In the
interest of all users of the environmental health criteria
documents, readers are kindly requested to communicate any
errors found to the Manager of the International Programme on
Chemical Safety, World Health Organization, Geneva,
Switzerland, in order that they may be included in corrigenda
which will appear in subsequent volumes.
In addition, experts in any particular field dealt with in
the criteria documents are kindly requested to make available
to the WHO Secretariat any important published information
that may have inadvertently been omitted and which may change
the evaluation of health risks from exposure to the
environmental agent under examination, so that the information
may be considered in the event of updating and re-evaluation
of the conclusions contained in the criteria documents.
PREFACE
For more than a century, epidemiological studies have
played an important part in investigations on the ways in
which infectious diseases spread through the community. At
the same time, intensive experimental work has resulted, in
many cases, in the identification of the bacterial, viral, or
other biological agents involved and therapeutic, preventive,
and control procedures have been introduced. Epidemiological
studies of biological agents and the effective control of
infectious diseases have paved the way to the use of
epidemiological methods in studying effects of non-biological
agents present in the environment. However, it is generally
more difficult to reach unequivocal conclusions in studies
involving physical and chemical agents than in those involving
biological ones. This is because the various factors involved
in studies of non-biological agents and their interactions are
usually more complex.
The relationship between environmental hazards and the
health of human communities is of growing concern and
increasing interest to governmental and public health
administrators, politicians, and the public. Despite the
substantial efforts made during the past two decades to expand
epidemiological studies on the effects of environmental
agents, there is a paucity of good studies that are useful in
establishing health criteria and a lack of adequate guidance
for the design and execution of studies and for the evaluation
of results. There are numerous papers and published reports
scattered throughout the literature. However, perhaps because
of the rapid development of the subject or of the pressure for
action once a new problem has been recognized, a large
proportion of the studies published to date have suffered from
deficiencies in design, analysis, or interpretation. There is
an urgent need, therefore, for a publication that sets out
appropriate methodology for such studies, which would help
Member States, relevant scientific institutions and individual
research workers to conduct epidemiological studies in a more
correct manner.
The concept of a monograph on the "Guidelines on Studies
in Environmental Epidemiology" was born at the meeting of the
WHO Study Group on Epidemiological Methods for Assessment of
the Effects of Environmental Agents on Human Health convened
in Geneva from 7-13 October 1975. The Study Group considered
that the underlying principles of environmental epidemiology,
particularly with regard to biological agents, had been
sufficiently well established, and that the main problem was
their proper application to the study of the health effects of
physical and chemical agents present in the environment.
The Study Group therefore recommended, among other
proposals, the preparation of a WHO monograph to provide
guidelines on epidemiological methods for assessing the
effects of environmental agents on human health. This
recommendation was unanimously endorsed by the Executive
Board of WHO at its fifty-eighth session in 1976.
In order to fulfil this recommendation, the first meeting
of the editorial group to prepare the monograph was held at
the Medical Research Council Toxicology Unit, St. Bartholomew's
Hospital Medical College, London, on 30 January - 1 February
1978.a The group agreed on the outline of the monograph,
on the tentative list of over 60 contributors or chapter
coordinators, and on the time schedule for preparation of
publication.
The second meeting of the editorial group was held in
London on 26-29 February l980. The group reviewed the first
draft of chapters prepared by coordinators, based on
contributions received, and made suggestions for further
editorial work. It was decided to make clear that the chapters
were not the work of individual coordinators, as the basic
material was being prepared by numerous contributors and
some material was being transposed from one chapter to another
during the course of editing. The chapter coordinators had a
dual role, compiling the chapters from individual contributions,
and also serving both as editors of individual chapters and
members of the editorial group for the monograph as a whole.
The members of the editorial group and all contributors are
shown on pages 14 - 18, respectively. Individual contributions
do not generally appear in their original form, as they have
been merged into various chapters; however, some of them have
been more extensively quoted and appear almost as originally
written.
The editorial group last met for two days at the Institute
of Occupational Medicine in Edinburgh from 20-2l August 1981.
The members of the editorial group were able to consider and
comment on the structure of the individual chapters and on the
volume as a whole. The arrangements for the joint IEA/WHO
workshop on the monograph, which was to take place the
following week during the IXth Scientific Meeting of the IEA,
were also discussed and the work plan for the final editing of
the monograph was agreed upon.
Thirty-eight environmental health scientists, including
six members of the editorial group, attended the joint IEA/WHO
workshop, coming from 16 countries and from the Commission of
-----------------------------------------------------------------
a Professor J. Kostrzewski, then the President of the
International Epidemiological Association (IEA),
participated in this meeting and was elected chairman of
the editorial group. He emphasized the preparedness of
the IEA to offer full technical support to the project.
European Communities. The workshop was held in Edinburgh on
24 August 1981. The members reviewed the drafts of individual
chapters and the monograph as a whole, making valuable comments
and suggestions for improvement. A list of the participants
appears on pages 19 and 21.
The final discussion on the draft of the monograph was held
in Moscow from 30 November - 3 December 1981, attended by a group
of international experts from all WHO regions, five members of
the editorial group, and two WHO staff members. The participants
are shown on pages 22 and 23. This meeting was convened with
the financial assistance of the United Nations Environment
Programme (UNEP) and was hosted by the Centre for International
Projects of the USSR State Committee on Science and Technology
and the A.N. Sysin Institute of General and Communal Hygiene of
the USSR, Moscow, which is also a WHO Collaborating Centre for
for Environmental Health Effects. Comments and proposals made by
this group greatly helped in making the monograph a more cohesive
work.
Thus, a total of 102 experts from 26 Member States were
involved in the preparation of the monograph.
In 1980, WHO, together with UNEP and ILO, launched an
International Programme on Chemical Safety (IPCS) and the present
project has become one of the priority projects within the frame-
work of this new international programme.
Although efforts have been made to avoid inconsistency in
terminology, uniformity has not been possible; indeed, this is
something beyond the scope of the present monograph. However,
within the IPCS, a project is under way to compile internationally
agreed definitions for the terms most frequently used in toxicological
and epidemiological studies. At present, therefore, it is important
to understand that some terms may have various meanings and
implications in different countries or in different scientific
circles and that it may be highly misleading to employ them outside
the national pattern of use or outside the context of a specialized
field, without precise definition. The reader is therefore warned
to be wary of the uncritical transfer of technical terms from one
set of circumstances to another.
The present monograph should serve as a useful guide to
the conduct of epidemiological studies on the effects of
non-biological agents on the health of human communities. An
epidemiological investigation would involve not only
epidemiologists but also other experts (e.g., clinicians,
statisticians, engineers, chemists, and nurses), and the
monograph is intended to address such a broad audience. The
editorial group believes that this publication will pave the
way for future studies.a
The final editing of the monograph was carried out during
1982 by M. Lebowitz and R. Waller, assisted by J. Kostrzewski,
M. Jacobsen, and Y. Hasegawa.
A special tribute should be paid to the late Dr F. Sawicki,
in recognition of his valuable contribution to this monograph,
and his extensive work on environmental factors in relation to
respiratory diseases.
The Editorial Group
* * *
Partial financial support for the development of this
criteria document was kindly provided by the Department of
Health and Human Services through a contract from the National
Institute of Environmental Health Sciences, Research Triangle
Park, North Carolina, USA - an IPCS Lead Institution.
---------------------------------------------------------------------------
a For instance, collaborative epidemiological studies on the
health effects of several chemicals being initiated under
the coordination of the WHO Regional Office for Europe
will apply the principles set out in this publication.
The editorial group was aware that a manual on epidemiology
for occupational health was being prepared jointly by the
World Health Organization in Geneva and the WHO Regional
Office for Europe; although the present Monograph has cited
several occupational health studies, it is addressed to
problems in the general population rather than among
specific occupational groups.
GUIDELINES ON STUDIES IN ENVIRONMENTAL EPIDEMIOLOGY
Editorial Groupa
Dr K.P. Duncan, Health and Safety Executive, London, England
(Coordinator for Chapter 7)
Dr M. Greenberg, Health and Safety Executive, London, England
(Coordinator for Chapter 4)
Dr Y. Hasegawa, Environmental Hazards and Food Protection,
Division of Environmental Health, World Health
Organization, Geneva, Switzerland (Secretary)
Professor I.T.T. Higgins, School of Public Health, University
of Michigan, Ann Arbor, Michigan, USA (Coordinator for
Chapter 2)
Dr M. Jacobsen, Institute of Occupational Medicine, Edinburgh,
Scotland (Coordinator for Chapter 6)
Professor J. Kostrzewski, Department of Epidemiology, National
Institute of Hygiene, Warsaw, Poland (Chairman)
Professor P.J. Lawther, MRC Toxicology Unit, Clinical Section,
St. Bartholomew's Hospital Medical College, London, England
Professor M.D. Lebowitz, Division of Respiratory Sciences,
Health Sciences Center, University of Arizona, Tucson,
Arizona, USA (Coordinator for Chapter 5)
Dr I. Shigematsu, Radiation Effects Research Foundation,
Hiroshima, Japan
Mr R. E. Waller, Toxicology and Environmental Protection,
Department of Health and Social Security, London, England
(Coordinator for Chapter 1)
Professor R.L. Zielhuis, University of Amsterdam, Coronel
Laboratorium, Netherlands (Coordinator for Chapter 3)
---------------------------------------------------------------------------
a Professor W.W. Holland and Dr C. du V. Florey, St.
Thomas's Hospital Medical School, London, were initially
the Coordinators for Chapter 4. As they were unable to
continue the work it has been taken over by Dr Greenberg.
The Secretariat wishes to thank Professor Holland and Dr
Florey for their efforts.
GUIDELINES ON STUDIES IN ENVIRONMENTAL EPIDEMIOLOGY
Contributors
Professor M. Alderson, Institute of Cancer Research, Surrey,
England (contributor to section 4.2)
Professor K. Biersteker, Agricultural University, Wageningen,
Netherlands (contributor to Chapter 2)
Professor N.P. Bochkov, Institute of Medical Genetics, Moscow,
USSR (contributor to section 4.10)
Professor D.S. Borgaonkar, North Texas State University,
Denton, Texas, USA (contributor to section 4.10)
Dr N. Breslow, International Agency for Research on Cancer,
Lyon, France (contributor to Chapter 6)
Dr D.G. Clegg, Food Directorate, Department of National Health
and Welfare, Ottawa, Canada (contributor to section 3.5)
Dr J.H. Cummings, Dunn Clinical Nutrition Centre, Addenbrookes
Hospital, Cambridge, England (contributor to section 4.11)
Professor W.J. Eylenbosch, University of Antwerp, Wilrijk,
Belgium (contributor to sections 4.10 and 4.11)
Dr C. Favaretti, Institute of Hygiene, University of Padua,
Italy (contributor to Chapter 7)
Dr A.J. Fox, The City University, London, England (contributor
to Chapter 6)
Mrs M. Fugas, Institute of Medical Research and Occupational
Health, Zagreb, Yugoslavia (contributor to Chapter 3)
Professor J.R. Goldsmith, Ben Gurion University, Beer Sheva,
Israel (contributor to Chapter 3)
Professor B.D. Goldstein, New York University Medical Center,
New York, USA (contributor to section 4.7)
Professor I.F. Goldstein, Columbia University, School of
Public Health, New York, USA (contributor to Chapter 6)
Professor M. Hashimoto, Graduate School of Environmental
Sciences, Tsukuba University, Ibaraki-ken, Japan
(contributor to Chapter 7)
Professor M.W. Higgins, School of Public Health, University of
Michigan, Ann Arbor, Michigan, USA (contributor to section
5.6)
Contributors (contd.)
Dr A.W. Hubbard, Ministry of Agriculture, Fisheries and Food,
London, England (contributor to section 3.5)
Dr A. Jablensky, Division of Mental Health, World Health
Organization, Geneva, Switzerland (contributor to section
4.6)
Dr A. Jakubowski, Institute of Occupational Medicine and Rural
Hygiene, Lublin, Poland (contributor to section 3.5)
Dr H. P. Jammet, Centre for Nuclear Studies,
Fontenay-aux-Roses, France (contributor to section 3.5)
Professor F. Kaloyanova, Institute of Hygiene and Occupational
Health, Sofia, Bulgaria (contributor to Chapter 3)
Professor S.R. Kamat, Department of Chest Medicine, K.E.M.
Hospital, Bombay, India (contributor to sections 4.8 and
5.6)
Dr H. Kato, Radiation Effects Research Foundation, Hiroshima,
Japan (contributor to section 5.6)
Professor L.T. Kurland, Mayo Clinic, Rochester, Minnesota, USA
(contributor to section 4.5)
Dr J.F. Kurtzke, Veterans Administration Hospital, Washington,
DC, USA (contributor to section 4.5)
Dr P.J. Landrigan, National Institute for Occupational Safety
and Health Cincinnati, Ohio, USA (contributor to Chapter 5)
Professor M.F. Lechat, Catholic University of Louvain,
Brussels, Belgium (contributor to section 4.10)
Professor S.R. Leeder, University of Newcastle, New South
Wales, Australia (contributor to section 4.4)
Dr D.T. Mage, Health Effects Research Laboratory,
Environmental Protection Agency, Research Triangle Park,
North Carolina, USA (contributor to Chapter 3)
Dr P.B. Meijer, TNO Research Institute for Environmental
Hygiene, Delft, Netherlands (contributor to Chapter 3)
Dr W.E. Miall, MRC Epidemiology and Medical Care Unit, Harrow,
England (contributor to Chapter 5)
Dr C.S. Muir, International Agency for Research on Cancer,
Lyons, France (contributor to Chapter 2 and section 4.3)
Profesor M. Nikonorow, National Institute of Hygiene, Warsaw,
Poland (contributor to section 3.5)
Contributors (contd.)
Dr B.R. Ordo]ez, Autonomous Metropolitan University, Mexico
City, Mexico (contributor to Chapter 7)
Professor B. Paccagnella, Institute of Hygiene, University of
Padua, Italy (contributor to Chapter 7)
Dr R.F. Packham, Water Research Centre, Medmenham, England
(contributor to section 3.5)
Professor B.S. Pasternack, New York University Medical Center,
New York, USA (contributor to Chapter 6)
Professor W.O. Phoon, University of Singapore, Singapore
(contributor to Chapter 5)
Dr I. Purcell, University of Newcastle, New South Wales,
Australia (contributor to section 4.4)
Professor A.V. Roscin, Central Institute for Advanced Medical
Training, Moscow, USSR (contributor to section 4.7)
Dr M. Saric, Institute for Medical Research and Occupational
Health, Zagreb, Yugoslavia (contributor to section 4.4)
Dr F. Sawicki, National Institute of Hygiene, Warsaw, Poland
(contributor to Chapter 5)
Dr M.A. Schneiderman, National Cancer Institute, Bethesda,
Maryland, USA (contributor to Chapter 2)
Dr I. Shigematsu, Radiation Effects Research Foundation,
Hiroshima, Japan (contributor to section 5.5 and Chapter 7)
Dr C. Silverman, Bureau of Radiological Health, US Food and
Drug Administration, Rockville, Maryland, USA (contributor
to section 3.5)
Professor F.H. Sobels, University of Leiden, Leiden,
Netherlands (contributor to section 4.10)
Dr E. Somers, Environmental Health Directorate, Department of
National Health and Welfare, Ottawa, Canada (contributor
to Chapter 7)
Dr J. H. Stebbings, Jr., Los Alamos Scientific Laboratory, Los
Alamos, New Mexico, USA (contributor to Chapter 6)
Dr A. H. Suter, Occupational Safety and Health Administration,
Department of Labor, Washington, DC, USA (contributor to
section 3.5)
Dr H. Tamashiro, Institute of Public Health, Tokyo, Japan
(contributor to section 5.5 and Chapter 7)
Contributors (contd.)
Dr M.P. van Sprundel, University of Antwerp, Wilrijk, Belgium
(contributor to sections 4.10 and 4.11)
Dr M. Violaki-Paraskeva, Ministry of Social Services, Athens,
Greece (contributor to Chapter 7)
Professor D. Wassermann, The Hebrew University Hadassah
Medical School, Jerusalem, Israel (contributor to Chapter
4)
Professor M. Wassermann, The Hebrew University Hadassah
Medical School, Jerusalem, Israel (contributor to Chapter
4)
Professor W.E. Waters, Community Medicine, Southampton General
Hospital, Southampton, England (contributor to section 5.3)
Dr J.A.C. Weatherall, Office of Population Censuses and
Surveys, London, England (contributor to section 4.10)
Professor P.H.N. Wood, The Arthritis and Rheumatism Council
Epidemiology Research Unit, University of Manchester,
Manchester, England (contributor to section 4.8)
Dr M. Zaphiropoulos, Ministry of Social Services, Athens,
Greece (contributor to Chapter 7)
IEA/WHO JOINT WORKSHOP: MONOGRAPH ON GUIDELINES ON STUDIES IN
ENVIRONMENTAL EPIDEMIOLOGY
Members
Dr E. Bennet, Health and Safety Directorate, Commission of the
European Communities, Luxembourg
Dr S. Beresford, Royal Free Hospital School of Medicine,
London, England
Dr G. W. Brebe, Clinical Epidemiology, National Cancer
Institute, National Institutes of Health, Bethesda,
Maryland, USA
Professor L. Breslow, School of Public Health, University of
California Los Angeles, California, USA
Dr D. Brille, Studies and Research Mission, Ministry of the
Environment, Paris, France
Dr P.G.J. Burney, Department of Community Medicine, St Thomas'
Hospital Medical School, London, England
Miss M. Deane, Epidemiological Studies Section, California State
Department of Health Services, Berkeley, California, USA
Dr A.R. Eltom, Faculty of Medicine, Khartoum, Sudan
Professor W.S. Eylenbosch, Department of Epidemiology and
Social Medicine, University of Antwerp, Wilrijk, Belgium
Professor G.M. Fara, Institute of Hygiene, Milan, Italy
Dr J.J. Feldman, Analysis and Epidemiology, National Center
for Health Statistics, Hyattsville, Maryland, USA
Dr I.F. Goldstein, Environmental Epidemiology Research Unit,
Columbia University, School of Public Health, New York, USA
Dr Y. Hasegawa, Environmental Hazards and Food Protection,
Division of Environmental Health, World Health
Organization, Geneva, Switzerland
Dr N. M. Hanis, Epidemiology Unit, Research and Environmental
Division, Medical Department, Exxon Corporation, E.
Millstone, New Jersey, USA and Cornell University Medical
School, New York, USA
Professor I.T.T. Higgins, School of Public Health, University
of Michigan, Ann Arbor, Michigan, USA
Professor M.W. Higgins, Department of Epidemiology, School of
Public Health, University of Michigan, Ann Arbor,
Michigan, USA
Members (contd.)
Dr M. Hitosugi, Department of Public Health, School of
Medicine, Kitasato University, Sagami-hara City,
Kanagawa-Ken, Japan
Professor A.C. Irwin, Department of Preventive Medicine,
Dalhousie University, Halifax, Nova Scotia, Canada
Dr M. Jacobsen, Institute of Occupational Medicine, Edinburgh,
Scotland
Professor H. Kasuga, Department of Public Health, School of
Medicine, Tokai University, Isehara-Shi, Kanagawa-Ken,
Japan
Dr M. Khogali, Faculty of Medicine, Kuwait University, Safat,
Kuwait
Professor M.A. Klinberg, Department of Preventive and Social
Medicine, Tel-Aviv University School of Medicine, Ramat
Aviv, Israel
Professor J. Kostrzewski, Department of Epidemiology, National
Institute of Hygiene, Warsaw, Poland
Professor L.T. Kurland, Department of Medical Statistics and
Epidemiology, Mayo Clinic, Rochester, Minnesota, USA
Professor R.A. Kurtz, Faculty of Medicine, Kuwait University,
Safat, Kuwait
Professor M. Lebowitz, Division of Respiratory Sciences,
Health Sciences Center, University of Arizona, Tucson,
Arizona, USA
Dr S. Mazumdar, Department of Biostatistics, University of
Pittsburgh, Pittsburgh, Pennsylvania, USA
Dr U. G. Oleru, College of Medicine, University of Lagos,
Lagos, Nigeria
Professor B.S. Pasternack, Department of Environmental
Medicine, New York University, Medical Center, New York,
USA
Professor M.R. Pandey, Thapathali, Katmandu, Nepal
Professor W.O. Phoon, Department of Social Medicine and Public
Health, National University of Singapore, Singapore
Dr M.P. Sprundel, Department of Epidemiology and Social
Medicine, University of Antwerp, Wilrijk, Belgium
Members (contd.)
Professor R. Steele, Department of Community Health and
Epidemiology, Queen's University, Kingston, Ontario, Canada
Dr H. Tamashiro, National Institute for Minamata Disease,
Minamata City, Kumamoto-Ken, Japan
Professor K.W. Tietze, Federal Health Office, Berlin (West)
Dr M. Wahdan, Regional Adviser on Epidemiology, WHO Regional
Office for Eastern Mediterranean, Alexandria, Egypt
Mr R.E. Waller, Toxicology and Environmental Protection,
Department of Health and Social Security, London, England
Professor W. Winkelstein, School of Public Health, University
of California, Berkeley, California, USA
FINAL REVIEW MEETING ON GUIDELINES ON STUDIES IN ENVIRONMENTAL
EPIDEMIOLOGY
Members
Dr L.K.A. Derban, Medical Officer, Volta River Authority,
Accra, Ghana
Dr C. Favaretti, Institute of Hygiene, University of Padua,
Italy
Dr M. Jacobsen, Institute of Occupational Medicine, Edinburgh,
Scotland
Professor S.R. Kamat, Department of Chest Medicine, K.E.M.
Hospital, Bombay, India
Professor J. Kostrzewski, Department of Epidemiology, National
Intitute of Hygiene, Warsaw, Poland (Chairman)
Professor M.D. Lebowitz, Division of Respiratory Sciences,
Health Sciences Center, University of Arizona, Tucson,
Arizona, USA (Co-rapporteur)
Professor A. Massoud, Department of Community, Industrial and
Environmental Medicine, Ain Shams University, Cairo, Egypt
Dr B.R. Ordonez, Environmental Health Programme, Autonomous
Metropolitan University, Mexico City, Mexico
Professor A.V. Roscin, Central Institute for Advanced Medical
Training, Moscow, USSR
Dr I. Shigematsu, Radiation Effects Research Foundation,
Hiroshima, Japan
Academician G.J. Sidorenko, A.N. Sysin Institute of General
and Communal Hygiene, Academy of Medical Sciences of the
USSR, Moscow, USSR (Vice Chairman)
Mr R.E. Waller, Toxicology and Environmental Protection,
Department of Health and Social Security, London, England
(Co-rapporteur)
WHO Secretariat
Dr I. Farkas, Promotion of Environmental Health, WHO Regional
Office for Europe, Copenhagen, Denmark
Dr Y. Hasegawa, Medical Officer, Environmental Hazards and
Food Protection, Division of Environmental Health, WHO,
Geneva, Switzerland (Secretary)
Other participants
Dr I.R. Golubev, Department of Public Health, USSR State
Committee for Science and Technology, Moscow, USSR
Dr Z.P. Grigorievskaya, A.N. Sysin Institute of General and
Communal Hygiene, Moscow, USSR
Dr Y.E. Korneyev, Laboratory of Epidemiological Methods of
Study, A.N. Sysin Institute of General and Communal
Hygiene, Moscow, USSR (Co-rapporteur)
Dr N.N. Litvinov, A.N. Sysin Institute of General and Communal
Hygiene, Moscow, USSR
Dr Y.I. Prokopenko, Department of the Influence of
Environmental Factors of Public Health, A.N. Sysin
Institute of General and Communal Hygiene, Moscow, USSR
Dr Ya.I. Zvinjackovskij, Laboratory of the Influence of
Environmental Factors of Public Health, Marseev Institute
of General and Communical Hygiene, Kiev, USSR
* * *
OTHER REVIEWERS
Dr A. David, Office of Occupational Health, Division of
Noncommunicable Diseases, WHO, Geneva, Switzerland
Mr J. Duppenthaler, Division of Epidemiological Surveillance
and Health Situation and Trend Assessment, WHO, Geneva,
Switzerland
Dr K. Hemminki, Institute of Occupational Health, Helsinki,
Finland
Dr J. Stjernswärd, Cancer, Division of Noncommunicable
Diseases, WHO, Geneva, Switzerland
Dr C. Xintaras, Office of Occupational Health, Division of
Noncommunicable Diseases, WHO, Geneva, Switzerland
1. INTRODUCTION
1.1. Interrelationships with Toxicological Studies
In some respects the present volume is intended to complement
an earlier publication in the Environmental Health Criteria series,
"Principles and methods for evaluating the toxicity of chemicals -
Part I", which dealt with experimental work using mainly animals
and other biological assay systems (WHO, 1978). There are some
parallels between such laboratory studies and epidemiological
investigations of the effects of hazardous substances on human
populations. The object, in each case, is to compare the effects
on groups subjected to different levels of the suspect agent,
always ensuring that the groups are matched as far as possible in
respect of other relevant factors (which may include sex, age,
temperature etc). Much experimental work is indeed done on human
subjects, restricted to doses that will evoke only relatively minor
physiological or biochemical responses that are readily reversible.
The borderline between laboratory experimentation and epidemological
work is not clearly defined. For the present purposes, however,
straight-forward toxicological studies on human beings, in which
the effects of specified doses of suspect agents administered
to small groups of subjects in the laboratory are examined, will
not be considered.
There are extensions of this approach which form a bridge
between laboratory work and that in the general environment, and
some mention should be made of these. Environmental chambers have
been constructed by some research groups, where small numbers of
subjects may spend periods of hours, days, or weeks under closely
controlled conditions. These have application in studies on acute
effects, and have been used, for example, to investigate the
effects of exposure to polluted urban air, drawn in from the
general atmosphere and carrying out control experiments with clean
air (Kerr, 1973). In isolated instances, this approach can be taken
a little further, for example, studies on the effects of lead
intake can be done by controlling the air and/or diet of groups
such as prisoners living in confined conditions (Cole & Lynam,
1973). More generally, however, the investigator cannot control
either the exposure of the subjects or their activities. Advantage
must then be taken of existing contrasts in environmental exposures
to obtain evidence on effects on health. In many cases, both
toxicological and epidemiological data are essential in
establishing sound health criteria and they are complementary to
each other.
1.2. Design
Perhaps as an over-reaction to a number of environmental
"disasters" that have occurred around the world, there has been a
tendency in recent years to carry out epidemiological imvestigations
without first posing any specific questions. There is indeed a
place for exploratory studies, often based on existing routinely
collected data on mortality or morbidity together with general
observations on environmental factors, but further studies
need to be carefully designed to test specific hypotheses. Then
one has to ask:
Who should be studied? Are particular subgroups of the
population at risk? How should control groups be selected?
What should be measured? Can specific agents be
identified? Is there a single pathway (for example, via
inhalation) or have several ways of entry to be considered
simultaneously? How are effects on health to be assessed?
Where has the study to take place? Should geographical
position, altitude, meteorology, etc., be taken into account
in selecting a locality? Are there existing monitoring
stations or sets of data relating to the environmental factors
in question?
When should the study be carried out? Are seasonal
effects likely to be important? Is the available time-span
long enough to provide a satisfactory estimate of long-term
exposures? Should exposures be averaged over months or years,
or are short-term peaks relevant in some cases?
In designing a special investigation or survey in the field of
environmental health, the objects of the exercise must first be
considered carefully. Without advocating a strict cost/benefit
approach to such studies, the question of the amount of time and
money spent in relation to the probable yield of information must
obviously be of importance. At one end of the spectrum, one might
consider monitoring the health records of the whole population, and
linking the information with as many data on environmental factors
as possible. Certainly, the monitoring of national death
statistics and of some aspects of morbidity records is possible,
looking particularly for the emergence of new trends or patterns of
distribution in congenital abnormalities and relatively rare
diseases. The need for this was underlined by the thalidomide
episode and, although the use of therapeutic drugs is not being
considered specifically in the present context, there are many
parallels between present-day enquiries into the safety of drugs
and the conduct of epidemiological studies on environmental agents.
However, to go beyond a broad surveillance such as this, with
enquiries into "health and habits" on a national scale, might be
regarded as an intrusion on personal privacy, apart from the
prohibitive cost. Even so, there is evidence that careful scanning
of linked records maintained on a regional, if not national, scale
can reveal new problems, as in the case of the occurrence of nasal
cancer among furniture makers (Acheson et al., 1967).
In this particular example, suspicions had been aroused by
clinical investigations on a few cases; however often, with a
relatively rare disease the "clustering" of a few cases in one area
or within one small subgroup of the population is sufficient to
give a positive lead on a new environmental hazard. In general,
when there is some indication of adverse effects of a particular
agent, the most effective way to conduct further epidemiological
studies is to concentrate attention on groups of people considered
to be particularly at risk. An example of this for a physical
agent - noise - is the investigation of exposure to "pop" music,
conducted among young college students (Hanson & Fearn, 1975). In
this study, dose-response relationships were examined within the
group selected, but, in general, it may be necessary to include
appropriate control groups, not exposed to the suspect agent.
An alternative approach, still directed towards high-risk
subgroups, is to consider a specific disease or effect, and to
compare the available information on exposures to environmental
agents with those of a control group. This is the "case-control"
type of study that was so successful in the early stages of the
investigation of the role of environmental factors in the
development of lung cancer (Doll & Hill, 1950; Wynder & Graham,
1950).
On wider issues, where the interrelationships between agents
and effects are more diffuse or more tenuous, relatively expensive
general community surveys may be needed, based on random samples of
the population concerned, or of particular age or occupational
groups. This technique has been particularly valuable in studies
on the role of environmental factors in the development of
bronchitis.
The types of survey that have proved to be of value in the
study of the effects of environmental agents are described in
Chapter 2. The dividing lines between them are not always clearly
defined, and there may be advantages in combining several
approaches within a single survey. The choice will depend on the
objects of the study and on the resources available.
1.3. Environmental Agents and Assessment of Exposures
As indicated in the preface, epidemiological methods were
developed initially to investigate the distribution and determinants
of communicable diseases, but their scope has now been widened to
include all aspects of health and wellbeing in relation to
biological or non-biological agents. Much of the discussion that
follows in later chapters on the design, conduct, and analysis of
epidemiological studies could apply to any field of interest, but
the prime concern here is with effects of chemical and physical
agents. The interaction of bacteria, viruses, fungi, yeasts,
protozoa, and higher animal agents or vectors with non-biological
agents is, however, recognized as contributing to human disease.
The term "agent" is a neutral one with no intrinsic implication
of "beneficial" or "adverse" characteristics. Most agents have the
potential for one or other or both of these effects, varying with
the precise nature of the agent, the level and duration of
exposure, and the state of nutrition and other acquired or
inherited characteristics of the subject. Thus, for example, the
chemical agents constituting vitamins and their analogues, that may
serve as essential food factors, are claimed to offer protection
against certain diseases (e.g., vitamin A against carcinogenesis),
or to have severe toxic effects, according to dose, to the state of
nutrition and acquired characteristics of the subject, and to other
agents operating coincidentally.
When studying these chemical and physical agents, it is
necessary to characterize them and to determine their absorption,
concentration in air, water, etc. with careful attention to
precision. For example, when describing a mineral, it is not
enough to give the name, chemical formula, and dose. An adequate
description involves specifying its contaminants, its physical form
(amorphous, crystalline, discrete particulate or fibrous), and
particle size distribution, and sometimes its physicochemical
surface properties. Due consideration has to be given to demo-
graphic and sociocultural factors that may affect the degree of
exposure or uptake as well as to special host characteristics,
including immunological status, before extrapolating the experience
in one population to that of another.
Chemical agents involved in environmental considerations have
been characterized as natural and manufactured organic (but not
living) and inorganic substances occurring in food, air, water,
soil, and other media. While living materials are excluded from
this category, their products are widely distributed in the
environment, in the form of metabolites, cell bodies, or bio-
chemical extracts. Thus, many foodstuffs are infested by, or
require for their synthesis, micro-organisms that are also found in
the wild and may contaminate the general environment.
Physical agents that impinge on man may occur naturally or be
man-made or man-intensified. They include ionizing and non-
ionizing radiation, the latter ranging from ultraviolet through
visible light and infrared to microwave, radio frequency and
extremely low frequency electromagnetic fields. Climatic
conditions of temperature and humidity play important direct and
indirect roles in environmental health. Noise and vibration at the
intensities experienced occupationally are associated with
objective evidence of damage; lesser intensities occurring outside
occupational environments, apart from affecting amenity, are a
source of concern in case they present health hazards.
The assessment of exposures is the most difficult aspect of
epidemiological research on environmental agents, and the one that
requires most careful thought, if any attempt is to be made to
establish "exposure/effect" relationships. For mixtures of
pollutants that are found under actual environmental conditions,
some integrated approach would be required for adequate exposure
assessment. However, such an approach has still to be developed.
The commonest practice is to monitor concentrations or
intensities (in air, water, etc.) at fixed points in order to make
estimates of the exposure of the community being investigated.
Measurements may have to be specially made for each investigation,
but advantage can often be taken of existing monitoring networks.
There has been a rapid expansion of monitoring activities
throughout the world in recent years: many reports have been
written about national and international programmes (Munn, 1973;
Department of the Environment, 1974), and a computer-based record
of current work is maintained in the United States of America
(Whitman, 1975). Although it is possible to monitor environmental
variables continuously at a large number of sites, it is impossible
to use that information in an undigested form in epidemiological
studies in which the health indices are generally crude. Provision
must therefore be made for statistical analyses of the data, once
collected, and, in some schemes operating with automatic
instruments, statistical analysers are incorporated, or the
instruments are "on line" to a central computer. A scheme of this
type in the field of air pollution monitoring has been described by
Lauer & Benson (1974). There is a risk however of becoming
overwhelmed with data from such complex networks and in most
epidemiological studies, it is more important to consider what is
the minimum requirement for a reasonable assessment of exposures
than to collect a vast array of data from which to select a few
figures.
"Personal" monitors may sometimes be applicable. This is
particularly true in assessing exposures to ionizing radiation, as
simple integrating devices (such as film badges) are available. It
is more difficult to measure most other environmental agents with
light portable equipment, but personal samplers for air pollutants,
such as suspended particulate matter and sulfur dioxide, are
available. Even so, the initial cost and maintenance problems
associated with these are at present deterrents to their use in
large-scale epidemiological studies.
For multimedia and non- or less-degradable pollutants, such as
metals and many organochlorine compounds, the biological monitoring
method, namely, the measurement of levels of polllutants in
tissues and fluids, has proved to be a useful tool for exposure
assessment.
Procedures for the assessment of exposures and for their
quality control are described in detail in Chapter 3: in general
the principles involved are meant to apply to any type of
environmental agent, though measurement methods are specific to
the one in question.
1.4. Effects on Health
Physical and chemical agents generated by man's activities
may have various effects on human being. Some substances may not
produce any adverse effects, while others, may be liable, if
exposures are sufficient, to affect such basic phenomena as
growth and development. Sometimes, environmental exposures may
affect host susceptibility or resistance, or produce functional or
prepathological changes. Behaviour may be modified by exposure,
especially to physical agents such as noise, light, and heat. A
wide range of pathological states in different organs may be
induced by exposure to environmental agents, and even death may be
caused or hastened by such exposures.
The starting point for many studies on the effects of
environmental agents has been the examination of existing records
of mortality or morbidity. The interpretation of findings from
these may itself be hazardous, but to determine which effects
should be studied, this retrospective approach is often considered
first.
In most countries, there are well-established systems of
registration of deaths, in which the cause of death is reported
(with varying accuracy) along with the age, date and place of
death, place of usual residence, marital status, occupation, and,
in some cases, additional information that may allow links to be
established with birth registration or other particulars of the
same individual.
Examination of long-term trends in death rates, or of
differences between countries, can occasionally give leads on
suspect environmental agents, but the most fruitful analyses in the
past have been those of local and regional differences in death
rates from specific diseases, within single countries. Thus, an
excess of cancer of the oesophagus could be seen in certain areas
of France, or of bronchitis in the industrial towns of the United
Kingdom, and, in these and many other examples, the findings have
been confirmed and investigated further in carefully designed
epidemiological surveys.
Occupational mortality studies can also be very valuable, but
those based on nationally-collected statistics are difficult to
interpret, since "occupation" may be inadequately described by the
relatives who have to give the information entered on death
certificates. Adelstein (1972) has also drawn attention to the
difficulty in distinguishing occupational risks from those of
"social" origin (notably tobacco smoking) in the more recent
records, and occupational studies are now better done as special
surveys within industries.
Short-term changes in death rates can provide information on a
limited range of agents that are subject to large variations in
intensity over periods of months, weeks, or days, and are
potentially lethal to some sections of the community. Many
diseases spread by bacterial and viral agents fall into this
category, but the main examples among physical and chemical agents
are air pollution and climatic conditions. Tabulations of deaths
on a monthly or weekly basis may be of some value in seeking any
evidence of acute effects of these factors, but ideally daily
tabulations are required, for selected areas containing large
populations.
Death is a crude but clearly defined index of response and it
is the one that has been most widely used in studies of the effects
of environmental agents. Where a small number of otherwise healthy
people die suddenly in one incident, perhaps as the result of an
accidental release of toxic materials in industry, cause and effect
relationships are easily established, but, in the general
community, associations are usually far more tenuous. It is
commonly the weakest sections of the community that are most
sensitive to potentially lethal effects of environmental agents:
the very old, the chronic sick, and the very young. In studies of
acute effects, it may be sufficient to study changes in the total
number of deaths in a given area, but specificity can often be
improved by considering deaths within limited age-ranges or for
certain causes only. Surprisingly, even the effects of major
insults to health may not be immediately obvious, if they impinge
mainly on the very old, among whom death rates are normally
relatively high. In the London fog of December 1952, there were
general indications of an exceptional death rate, such as a
shortage of coffins and flowers for funerals, but it was not until
all the returns of deaths from local registrars were collected
together and scrutinized that it was realized that the number of
deaths during and just after the fog was about 4000 more than would
normally have been expected (Ministry of Health, 1954).
Illness, as defined in various ways in routinely collected
morbidity statistics, can be regarded as a further index of
response. The more it is qualified in terms of age-range and
disease category the better it is, but there are many hazards in
accepting information collected largely for administrative
purposes, because of possible biases. Morbidity data are far more
subject to interference from social factors than mortality data;
weekends and holidays, for example, have little effect on death
rates, but they have a profound effect on consultation rates with
general practitioners and on hospital admissions. Provided these
reservations are borne in mind, it is still possible to make use of
some existing information for epidemiological studies.
The range of indices of effects on health available for use in
specially designed epidemiological surveys is very wide and covers
all organ systems. It is described in detail in Chapter 4 and it
can include death, the onset or prevalence of specific illnesses,
measurements of developmental, behavioural, functional, and
prepathological changes, and biochemical indices. There are,
however, limitations of costs, usage, and acceptability of some of
the tests.
1.5. Organization and Conduct
There are many practical problems to consider in the
organization and conduct of epidemiological studies and the
recommended procedures are described in Chapter 5. Both the level
of study to be conducted (simple to complex) and the resources
required to do it must be considered. As in the rest of the
monograph, studies are described which can be conducted in various
settings in the world. There is often preliminary work to do in
contacting organizations that may be able to provide, or help in
collection of, the health and environmental data, or may merely
need to be made aware of the aims and existence of the survey.
Some advance publicity may be desirable to gain the cooperation of
subjects and, where occupational groups are concerned, discussions
with managements and unions are essential.
Having selected the population required for the study, initial
contacts with individuals may need to be made by letter, prior to
any interview or examination. One of the major difficulties is to
obtain an adequate response from the population selected. Particu-
larly in studies of chronic effects, where contrasts are being
sought between people living in different areas, it is essential to
ensure that failure to contact or to follow up some of the subjects
does not bias the result. The possibility of observer bias must
also be considered. Where a number of observers are engaged in
interviewing or examining subjects in different localities, joint
training sessions are required to ensure uniformity of approach and
it may be necessary to interchange the teams to reduce risk of
bias. Even where objective assessments are being made, for example
with peak flow measurements in surveys of respiratory disease, it
is important to ensure uniformity of procedure and to check
regularly the performance or calibration of the instruments being
used. Careful standardization of methods of measuring the related
environmental agents is also required and, if biological indices of
effects are being used, it may be necessary to ensure that all
these measurements are made in a single laboratory.
The conduct of the field work itself will depend on the nature
of the survey. Surveys can be simple or complex. Subjects may be
seen only by field workers, but preferably by those who know the
community and its culture. Subjects may be asked to come to one of
several bases that might be set up in the areas near where the
subjects live or to a central laboratory or clinic. In surveys
requiring instruments for the measurement of lung function, etc.,
mobile laboratories are sometimes used. Where the subjects are
grouped together, for example in selected schools, offices, or
factories, the survey team will normally visit them there, by prior
arrangement with the authorities concerned. The most labour-
intensive survey, but often the most satisfactory, where the
effects of common environmental agents are being studied on
samples of the general population, is where the field workers
visit the subjects in their own homes.
Ethical problems sometimes arise; for example, if some of the
tests involved are regarded as intrusive. In surveys of exposure
to lead, blood samples may be required and, although there is
relatively little difficulty in taking these with the prior
permission of the subject in the case of adults, problems arise
with children, for parents and others cannot properly give
permission for samples to be taken in this way, if it is not for
the benefit of the child. Apart from this, confidentiality of
all information obtained in surveys must be maintained at all
times, hence it is common practice to exclude names and addresses
at all stages of the preparation and analysis of results, beyond
the original survey form.
1.6. Analysis and Interpretation of Results
In some surveys of modest size and quite often in the case of
studies on acute effects of environmental agents, the findings may
be tabulated manually and/or presented graphically in a straight-
forward manner. More generally, however, the data will be
transcribed on to punched cards, paper tape, or magnetic tape for
analysis by computer. Procedures for the preparation of the data,
and for the analysis and interpretation of findings are described
in detail in Chapter 6, in which the need for the close involvement
of statistical staff throughout the study is stressed.
The statistical analysis of epidemiological studies has been
revolutionized by the application of "package" programmes. These
have been written as general purpose statistical routines and
survey analyses and, although they may be handled by people with
relatively little statistical and computing experience, it is
essential to have expert advice and guidance to avoid misapplying
the techniques or misinterpreting the findings. The more complex
the technique, the more necessary it is to pause to consider the
relevance of the data, and, if possible, to try to provide some
visual presentation of the main features, for example, in the form
of a graph that may be displayed on a screen linked with the
computer, or plotted out on microfilm or by line-printer.
A fundamental point in relation to the control of
environmental pollutants is whether there is any kind of level of
exposure, below which effects of an environmental agent are not
detectable (with the techniques used), but beyond which effects
increase gradually in a defined relationship. A feature such as
this may be extremely difficult to establish, since the effects of
very low levels of exposure cannot be assessed with a degree of
precision great enough to allow much discrimination between
alternative hypotheses.
The greatest risks of mistaken interpretation occur in multiple
regression analysis where attempts are made to assess the extent to
which each of several variables affects some index of health; for
example, the prevalence of respiratory symptoms in a number of
communities may be studied in relation to several different
measures of air pollution, to climatic factors, and to the levels
of cigarette smoking. In such cases, there is a need to consider
whether a linear relationship is appropriate for each of the
variables, but beyond that, if some of those variables included are
correlated with one another (as is likely with measures of air
pollution and climate) then the regression coefficients cannot be
determined with any satisfactory degree of precision, and there is
a serious risk of overestimating the effect of one variable at the
expense of another (McDonald & Schwing, 1973).
Even when a significant correlation is found between an index
of health and one or more environmental factors, the relevance of
this must be considered carefully, for example in terms of
biological plausibility. If the number of observations in a study
is large, a correlation coefficient as low as 0.2 may be
statistically significant, but it would account for only 4% of the
variance in the health index, leaving 96% to be explained some
other way, perhaps in part by environmental factors that were not
measured.a In such cases, it may be necessary to consider
whether the assessments of environmental exposures were adequate,
or whether the overall effect of any environmental factors may have
been trivial in relation to that of other determinants.
Above all, the fact that correlation does not necessarily imply
causation must be recognised. Many unrelated factors exhibit
similar time trends or geographical distributions, and much
supporting evidence is required before there can be any presumption
of causation.
1.7. Uses of Epidemiological Information
The problem of considering whether a statistical association,
observed between indices of health and various chemical and
physical agents in the environment, suggests any cause and effect
relationship, is much more difficult than in the case of classical
epidemiological studies concerned with communicable diseases. The
basic difficulty is that few of the non-biological agents have
unique effects on health, and conversely the effects considered may
often be related to a wide range of factors. Thus, when decisions
have to be made about the need for control of suspect agents,
within industry or in the community at large, many aspects of the
situation may have to be taken into account, such as the strength
and consistency of associations seen in epidemiological studies,
related toxicological and clinical findings, and economic or social
implications of control measures.
Clearly many different disciplines become involved at this
stage and a full discussion is beyond the scope of the present
monograph, but this very important facet, which should involve the
scientist as well as the administrator, is introduced in Chapter 7.
--------------------------------------------------------------------------
a In general terms, the proportion of variance explained by
a regression is r2, where r is the correlation
coefficient; hence r = 1 (perfect agreement), all the
variance is explained: for r = 0.2, r2 = 0.04 (i.e. 4%).
The standard error of a correlation coefficient is
1
approximately -- where n is the number of observations.
n´
1 1
Hence for n = 10000, -- = --- = 0.01 and a coefficient r
n´ 100
in excess of 0.02 would be significant at the 5% level.
REFERENCES
ACHESON, E.D., HADFIELD, E.H., & MACBETH, R.G. (1967) Carci-
noma of the nasal cavity and accessory sinuses in wood-
workers. Lancet, 1: 311-312.
ADELSTEIN, A.M. (1972) Occupational mortality: cancer. Ann.
occup. Hyg., 15: 53-57.
COLE, J.F. & LYNAM, D.R. (1973) ILZRO's research to define
lead's impact on man. In: Environmental aspects of lead,
Luxembourg, Commission of the European Communities, pp.
169-187.
DEPARTMENT OF THE ENVIRONMENT (1974) The monitoring of the
environment in the United Kingdom. London, Her Majesty's
Stationery Office.
DOLL, R. & HILL, A.B. (1950) Smoking and carcinoma of the
lung. Br. med. J., 2: 739.
HANSON, D.R. & FEARN, R.W. (1975) Hearing acuity in young
people exposed to pop music and other noise. Lancet, 2:
203-205.
KERR, H.D. (1973) Diurnal variation of respiratory function
independent of air quality. Experience with an environ-
mentally controlled exposure chamber for human subjects.
Arch. environ. Health, 26: 144-152.
LAUER, G. & BENSON, F.B. (1974) The CHAMP air quality moni-
toring program. In: Proceedings of the International
Symposium, Recent Advances in the Assessment of the Health
Effects of Environmental Pollution (Paris). Luxembourg,
Commission of the European Communities.
MCDONALD, G.C. & SCHWING, R.C. (1973) Instabilities of
regression estimates relating air pollution to mortality.
Technometrics, 15: 463-481.
MINISTRY OF HEALTH (1954) Mortality and morbidity during the
London fog of December 1952. London, Her Majesty's Stationery
Office.
MUNN, R.E. (1973) Global environmental monitoring systems.
Toronto (SCOPE Report 3).
WHITMAN, J. (1975) More on monitoring. Environ. Sci.
Technol., 9: 611.
WYNDER, E.L. & GRAHAM, E.A. (1950) Tobacco smoking as a
possible etiologic factor in bronchiogenic carcinoma. J. Am.
med. Assoc., 143: 329.
WHO (1978) Environmental Health Criteria 6: Principles and
methods for evaluating the toxicity of chemicals. Part I.
Geneva, World Health Organization.
2. STUDY DESIGNS
2.1. Introduction
This chapter is concerned with the type of approach to be used
in an epidemiological study, starting with exploratory investigations,
which may be based on existing mortality or morbidity records, on
general health surveys, or sometimes on quite small-scale clinical
observations, and are aimed at seeking indications of the role of
environmental factors in a particular disease or condition. Such
investigations may be of value in formulating hypotheses that can
be followed up by studies designed specially to test them and,
where appropriate, to try to assess relationships between exposure
and effect in a quantitative manner.
Generally, it is an unusual distribution of disease in a
locality or a particular population that prompts the enquiry (which
could be regarded then as "effect-oriented"), though sometimes
concern arises because of some characteristic of the environment
that is thought, either on toxicological or more general grounds,
to have adverse effects on health ("agent-oriented"). In the
former category, an example is the recent epidemic of a severe
respiratory and generally debilitating disease in Spain (Tabuenca,
1981; Aldridge & Connors, 1982). This affected people over a wide
range of ages in several parts of the country, and it was at first
thought to be due to a respiratory infection. Astute clinical
enquiries concentrating attention on infants in the first instance,
because of their more closely confined environments and more
readily specified diets, revealed that each case was related to the
use of a particular supply of cooking oil that proved to be
chemically contaminated. These initial enquiries constituted the
exploratory study that generated a hypothesis capable of being
tested by both toxicological and epidemiological techniques.
The investigation of long-term effects of exposure to ionizing
radiation, following the 1945 atomic bomb explosions in Japan,
could be regarded as falling in the agent-oriented category. While
immediate effects were disastrous and there was every indication
that survivors would be liable to develop further radiation-induced
illnesses over the years, the exact nature of the effects and the
form of exposure/effect relationships were unknown. A longitudinal
study, in which defined populations were to be followed through to
death, was designed to examine these questions, and this is
referred to in detail in section 5.6.8.5.
In the following sections, some of the more commonly used types
of design in epidemiological studies are described, but it is
essential to stress that they are not alternatives that can be
chosen freely for any given situation. The choice of design
depends primarily on the questions being asked (the objectives of
the study) and on constraints imposed by factors such as resources
available, the time limit within which at least provisional answers
are required, accessibility of the population to be studied, and
ethical considerations. It is vital that a sensible hypothesis,
supported wherever possible by toxicological evidence, is
formulated first and the art of good survey design is to reconcile
conflicts between the ideal and what is possible in a way that will
maximize the acquisition of useful data.
2.2. Preliminary Review of State of Knowledge
The available literature on the clinical features and natural
history of the disease or condition being considered, on what is
known of its causes and distribution in the population, and on
trends with time, should be critically reviewed. Often, there are
conflicting findings between different published studies in the
field of environmental epidemiology and it is important to try to
establish which findings can be regarded as reasonably well-
founded.
At the same time, a review is required of information on all
the relevant environmental factors, including physical and chemical
properties, possible interactions with other agents, and anything
known about their spatial and temporal distribution. Any data
available on toxicological properties from animal experiments or
other biological testing procedures also needs to be examined
carefully.
In some instances, where new problems are encountered suddenly
and immediate action is required, as in the Spanish cooking-oil
problem cited above, or the Seveso accident in which dioxin was
dispersed in the vicinity of a chemical works (see section 7.3),
there may be little prior information on the agents concerned or
their effects, and, in any case, little time to study it. Even so,
it remains vitally important to consider carefully the types of
epidemiological studies that could and should be undertaken. A
false move in the beginning could completely undermine the chances
of yielding results that would contribute to the identification of
causal agents, and to the specification of exposure/effect
relationships.
2.3. Descriptive Studies and Use of Existing Records
Investigations of the general distribution of disease and of
possible environmental determinants on the basis of existing
records are referred to as descriptive studies: they describe the
situation as it exists in the community, without special efforts to
investigate symptoms, physiological functions, or exposures to
particular agents in defined groups. They may be included among
the exploratory investigations mentioned above, but they can
nonetheless be major undertakings in their own right, as in the
case of the construction of the detailed atlases of cancer
mortality that have now been prepared in a number of countries
(Mason et al., 1975; Editorial Committee for the Atlas of Cancer
Mortality in the People's Republic of China, 1979; Japan Health
Promotion Foundation, 1981).
Although past records frequently suffer from lack of
reliability, they also have certain advantages and have been used
not only for descriptive studies but for other types of
epidemiological studies including retrospective studies and case-
control studies (sections 2.7 and 2.9). For example, many of the
diseases and conditions of importance in environmental health
studies, as in the case of a number of cancers, occur many years
after significant exposure has taken place. In these circumtances,
it is usually wise to consider using information about the effects
of past exposures as the basis for providing answers to the
questions of interest.
Another advantage of existing records is economics. In most
situations, it will be found that the length of time required to
gather relevant new data would justify some initial investment of
effort in the study of past records.
There are two further reasons why such an approach should
always be considered. First, environmental hygiene changes with
time; recent exposures are generally at lower levels than those in
the more distant past. The effects of exposure are likely to be
more evident in people exposed to higher levels than in those
exposed to lower levels. If, therefore, the aim is to seek an
answer to a preliminary question as to whether or not there is a
real association between the hazard and the suspected environmental
agent, then attention must be focused initially on so-called "high-
risk" groups who are most likely to demonstrate an effect, if it
exists. The second reason is based on ethical considerations.
Knowing that a group of people has been exposed to a certain toxic
substance, it seems incumbent on society to assess the possible
health effects from such exposures in order to take preventive
action.
2.3.1. Mortality statistics
The routine collection of national mortality data commenced in
a number of countries in the mid-nineteenth century; for example,
since 1837, material has been collected for virtually every death
occurring in the United Kingdom. The World Health Organization
(WHO) has been responsible for sponsoring and encouraging the
collection of accurate mortality statistics throughout the world,
and the majority of developing countries now have some system for
the recording, collection, processing, and production of mortality
data.
All sets of routine data have disadvantages however. The
majority of deaths are certified by the practitioner attending the
patients, or sometimes by an official responsible for investigations
in cases of doubt or of violent or unnatural death, which may
include occupationally-associated disease. Though many systems
suffer from delay in data collection, legal requirements to
register the death and the establishment of registrars responsible
for handling this material usually result in a steady flow of data
into the central processing system. Insofar as autopsy contributes
to accurate diagnosis, varying rates will affect the validity of
comparisons between different countries and different periods
(Moriyama et al., 1966). Diagnostic vogues and differing vigilance
may also introduce bias.
Waldron & Vickerstaff (1977) have reviewed the subject of the
accuracy of diagnoses of fatal conditions and the quality of
certification. Although a clinician may be clear in his own mind
about the diagnosis, he does not always record it on the death
certificate in a way that can be appropriately coded. For a number
of years, it has been recognized that death is commonly the result
of a complex of diseases, and the international system for the
derivation of a single underlying cause of death from a full death
certificate can produce unrealistic statistics. This issue has
been discussed by a number of authors, for example, Alderson
(1976). For all its imperfections, the International Statistical
Classification of Diseases, Injuries and Causes of Death (WHO,
1977) is of great value. It contains definitions and recommendations
together with rules for medical certification, for the clerical
coding of primary causes of death and for quality control.
If death certificates themselves are used for epidemiological
purposes rather than the officially published statistics, then the
person undertaking the coding of cause of death should check his
performance against that of national coding staff. It is possible
to undertake analyses of morbid conditions mentioned on death
certificates apart from the primary cause of death. These can be
of value in studying health service requirements as well as their
relationship with environmental hazards. In some countries (e.g.,
Scotland, Sweden, and the USA) it is considered worth coding all
the conditions mentioned on the death certificates.
2.3.2. Morbidity statistics
A wide range of routine morbidity statistics is now available
in many developed countries. These may include data on abortion,
cancer, congenital abnormalities, hospital inpatients, infectious
diseases, school health, and sickness absence, including accidents
at work and occupational diseases.
WHO plays a major role in the standardization of morbidity
statistics. Various contributions to the World Health Statistics
Quarterly have discussed aspects of the methods required to
collect, analyse, and present material on all aspects of health
care. A general review of this topic has been published by WHO
(1965). Wagner (1976) reviewed 91 projects in 25 European
countries, concerning processed data on patients discharged from
hospital in-patient care. This report provides detailed
information about the capture, coding, and processing of the data
but limited indication of how the output from these systems was
used. A conference of the Commission of the European Community
discussed the relationship between health interview surveys, health
examination surveys, and routinely processed data on hospital
inpatient discharge records; Armitage (l977) indicated the
possibilities of international collaboration and the topics for
which this seemed feasible.
Despite the extensive data base on morbidity in a number of
countries, much care is generally required in using this type of
information, even for exploratory studies in environmental
epidemiology. The records may not provide complete coverage of the
population and there may be many in-built biases, particularly in
relation to socioeconomic class. Thus official sickness/absence
records show large variations in the apparent extent of illness
between different occupations, but these are often connected with
social factors or the amount of physical or mental effort required
in the job rather than with specific hazards precipitating illness.
Data assembled at cancer registries can, however, provide a
valuable supplement to those obtained from mortality records. Each
newly diagnosed case of cancer enters the system and near-complete
coverage of the population has been achieved in many countries.
The techniques involved have been reviewed by McLennan and co-
workers (1978). While both cancer registry and mortality data
suffer from differences in diagnostic standards and practice that
make international comparisons difficult, the former avoids some of
the problems introduced by different treatment regimes in the
interpretation of mortality statistics, and they are particularly
valuable for studies on conditions such as skin cancer that have a
low fatality rate.
2.3.3. Populations at risk
Occasionally, the absolute numbers of deaths or cases of a
particular disease can be of value in establishing relationships
with environmental factors without reference to the size or age
structure of the population at risk. This is particularly true of
rare conditions: for example, the identification of just a few
cases of angiosarcoma of the blood vessels of the liver was
sufficient, coupled with experimental animal studies, to
demonstrate a clear link with occupational exposure to vinyl
chloride. Similarly, clusters of cases of mesothelioma of the
pleura demonstrated links with particular types of fibres
(crocidolite asbestos, among occupational groups in South Africa
and elsewhere, and a local volcanic rock with an unusual fibrous
structure in the case of a village community in Turkey). Also, the
proportion of deaths attributed to a certain cause among all deaths
in a defined group can provide useful clues about environmental
factors, providing that basic data on sex and age are taken into
account (section 6.3.7.5).
More generally, however, detailed information on the size, sex,
and age structure of the population at risk is required for the
proper interpretation of mortality and morbidity statistics. The
calculation of appropriate rates is discussed further in section
6.3.7, and it is necessary here only to stress the importance of
obtaining adequate information on the denominators (the populations
at risk) as well as on the numerators (the numbers of deaths, or
cases of disease).
In most countries, complete censuses of the population are done
at intervals of the order of 10 years, and estimates of changes in
the intervening period are made from records of births, deaths, and
migration. Such records are capable of providing a detailed break-
down by sex and age, not only on the national scale but also for
individual towns and smaller communities within them. Even so,
much care is required in studies confined to small local areas and
it may be necessary to check or supplement the official data, even
to the extent of carrying out an unofficial census. This type of
approach may, in any case, be necessary in countries where census
data are incomplete or where internal migration rates are high.
2.3.4. Geographical differences in mortality and morbidity
Contrasts in appropriately standardized mortality and morbidity
rates (section 6.3.7.3) can be made between countries, or within
countries between groups characterized by their area of residence
or any other qualifier (such as ethnic group) that may be included
on the official records. These characteristics may, to a limited
extent, provide a qualitative guide to exposures to environmental
agents, thus allowing some exploratory studies to be done.
International comparisons are, however, fraught with difficulties,
due to differences in diagnostic practice or other factors. For
example, in the 1950s, mortality from bronchitis was about 25 times
higher in Scandinavia than in the United Kingdom. It was suspected
that this was partly an artifact of definition, and it led to
studies on variations between countries on the certification of
bronchitis and emphysema on death certificates (Fletcher et al.,
1965). In this particular case, it appeared that while differences
in terminology and in rules for assigning cause of death explained
quite a large part of the difference in mortality between the
United Kingdom and other countries, environmental factors probably
also contributed. To pursue this question further, however, it was
necessary to set up specially designed studies (Holland et al.,
1965).
In general, geographical contrasts between areas within a
single country are likely to be less than those between countries,
but they can be more revealing in relation to environmental
influences. Possibly, one of the most exciting intracountry
variations hitherto uncovered is the 30-fold difference in
oesophageal cancer risk for women in different areas along the
Caspian Littoral of Iran where, in the high incidence areas, this
form of cancer, generally rare in females (Kmet & Mahboubi, 1972),
is two to three times commoner than the relatively high incidence
of breast cancer in North American and European women.
It is not only in developing countries that such variations are
to be found. In England, stomach cancer is 50% commoner in
Liverpool than in Oxford. While some of the differences
demonstrated in the recently published maps of cancer morality,
referred to at the beginning of section 2.3, will turn out, when
examined closely, to be due to artefacts, others will prove to be
real and suitable for study.
Some of these contrasts can be linked with differences in
social class distribution between areas, implying effects of broad
environmental factors related to lifestyle, to concentrations of
recent immigrants or ethnic groups or to the selective migration of
relatively fit members of the community in or out of the areas
concerned.
Studies based on routinely collected mortality and morbidity
data usually have to be confined to comparisons based on area of
residence at the time of death or of occurrence of the illness in
question, and this is a limiting factor in studies on chronic
diseases, particularly in countries with high internal migration
rates. However, in the case of migrants between countries,
official records of country of origin are often maintained, and it
is possible to compare the experience of migrants with that of
their compatriots in both the country of origin and that of
subsequent residence. This sheds some light on the relative roles
of environmental and genetic factors in the development of disease.
The migrant exchanges one environment and its associated
exposures for another. If the international differences in various
disease risks observed are due to genetic factors, then incidence
should not be influenced by migration. Yet, as the pioneer studies
of Haenszel & Kurihara (1968) have shown, cancer morbidity and
mortality rates in migrant populations gradually come to approximate
those of the host country.
2.3.5. Time trends
Long-term trends with time in the mortality or morbidity rates
for specific diseases can be of value in indicating possible
effects of environmental factors, though interpretation is
complicated by the effects of improvement in therapeutic treatment,
etc. There has, for example, been a massive decline in mortality
from pulmonary tuberculosis in most developed countries during the
present century. It is difficult however to separate out all the
factors responsible: much of the decline occurred before the
really effective treatment by antibiotics became available, and, to
some extent, it can be attributed to environmental factors in the
broadest sense, i.e., to improved housing and social conditions and
to better medical care generally.
In many countries, the incidence of cancer of the breast, lung,
pancreas, and prostate is rising. It has been suggested, particularly
for lung cancer, that these increases are artefactual, being due to
better diagnosis, changes in classification, etc. (Percy et al.,
1974). While such factors probably have had some influence, it is
very difficult to believe that for an organ as accessible as the
breast they explain more than a small proportion of the observed
increase. The increase in malignant melanoma of the skin, a very
accessible cancer, has been carefully investigated by Magnus (1973)
and others who conclude that the rise is real. In the United
States of America, cancer of the oesophagus has doubled in persons
of Negroid origin, since 1935. Nonetheless, it is worth while
remembering that were it not for tobacco-caused lung cancers, the
overall cancer mortality in the USA for Caucasian males would be
falling and that for Caucasian females, the overall cancer
incidence is falling slowly (Devesa & Silverman, 1978).
When examining trends over an extended period, it is always
important to ensure either that sex/age specific rates are used or
that the data are standardized with respect to age (section
6.3.7.4), since there have been considerable changes in the age-
structure of the population in most countries during recent
decades. Sometimes, contrasts in trends between men and women can
provide clues about the factors responsible, as in the case of lung
cancer, for which death rates began to increase sharply sooner in
men than in women (consistent with an effect of cigarette smoking).
2.3.6. Associations with environmental indices
Apart from the general guidance that can be obtained from the
examination of geographical differences and trends in mortality and
morbidity, it is often possible to use observations on dietary
factors, or on air or water pollution, etc. to carry out further
descriptive studies.
For example, a large number of studies concerned with
associations between mortality and routine observations of urban
air pollution have been reviewed by Holland and co-workers (1979).
While most of these indicate positive correlations with measurements
of pollutants such as smoke, total suspended particulates or sulfur
dioxide, there is probably an interaction with other confounding
factorsa not taken into account, notably tobacco smoking. These
initial studies were however valuable as exploratory ones, leading
to the development of studies designed specifically to test the
hypothesis that exposure to urban air pollution contributes to the
development of chronic respiratory disease.
2.3.7. Case registers
As mentioned in section 2.3.3, it is sometimes possible to
identify environmental agents related to the development of
relatively rare conditions, simply from the clustering of a few
cases in local areas or in particular occupations. It is seldom
possible to recognize associations between common exposures and
common conditions in this way, but one effect-oriented approach is
to establish case registers through hospitals and/or general
practitioners for selected conditions for which there is already
some indication (e.g., an irregular geographical distribution)
that environmental factors may play a part. It may then be
possible, through careful enquiry into domestic and occupational
histories, to identify some common factor that can be followed up
further with additional epidemiological and toxicological
studies.
---------------------------------------------------------------------------
a Defined in section 6.4.5.3.
In developing countries, careful appraisal of a wider range of
cases and their associated histories can, however, help to provide
background information in the absence of comprehensive official
statistics. Even so, with the 3000-5000 people for whom a single
primary health care worker may be responsible, the wide random
fluctuations in morbidity or mortality rates that would be likely
to occur, would have little real meaning, and it would probably be
necessary to assemble information at a district or provincial level
in order to seek evidence of unusual local patterns of disease
(WHO, 1982).
2.3.8. General surveys
While survey techniques, considered in greater detail in
subsequent sections of this chapter, form an essential part of most
of the study designs, in many countries, regular surveys of the
population, made for administrative purposes, can be of value as
exploratory studies in relation to environmental factors. Thus, in
the United Kingdom, there is a General Household Survey that
enquires into family expenditure on foods, etc., and within this,
questions are asked on recent illnesses. There are possibilities
of adding additional questions on matters that may affect health
and, in this way, data have been obtained that could be used in
conjunction with mortality records to demonstrate strong
interrelationships between smoking and occupation and, in turn,
with lung cancer mortality (Office of Population Censuses and
Surveys, 1978). The application of information from other types of
surveys is discussed further in section 4.2.2.
2.4. Formulation of Hypotheses
Studies are most likely to be productive if they are based on
clearly stated hypotheses. These can be developed from the results
of various descriptive studies, as discussed above. Basically,
this is to try to demonstrate an association between carefully
specified effects on health and assessments of exposure to
specified environmental agents. Epidemiological studies cannot by
themselves prove that a particular agent causes a particular health
effect; they may, however, demonstrate quantitatively the strength
of an association between the presence of the agent and the
occurrence of the hypothesized effect. Appropriate statistical
analyses may in turn determine the probability that an association
as strong as that observed might have occurred by chance (section
6.4.1). Whether the correct agent has been identified or whether
the apparent association has arisen artefactually, because of
correlations with exposure to other agents or factors that were not
studied, is a question requiring further epidemiological studies
and, where possible, also toxicological work.
Most investigations in the field of environmental epidemiology
are necessarily of an observational nature, that is, they are
observations based on existing situations. Associations can be
demonstrated most clearly if it is possible to compare groups
exposed to several levels of the agent in question, but, in the
last resort, hypotheses about the exact form of exposure/effect
relationships can be tested effectively in experimental situations,
where the research worker has some control over exposures.
While the working hypothesis must be as simple as possible, it
has to be recognized that causes of ill-health are commonly multi-
factorial, and that the environment, though it comprises many
individual components, acts as an entity, having effects liable to
be greater than the total of those of the components. It may be
that, in the subsequent statistical analysis, a complex variable
can be developed to describe the combined effects of exposures to a
range of different agents as measured within the study (Cassell &
Lebowitz, 1976), but such ideas are difficult to incorporate into
the initial hypotheses.
It may be helpful to view the formulation and testing of
hypotheses in environmental epidemiology as an example of the
essentially iterative process of science, which comprises an
initial (or crude) hypothesis, assembling of data from available
sources or from planned investigations, testing of the validity of
the hypothesis, rejection of the hypothesis leading to its revision
or refinement, and the further assembling of data to test a revised
version.
The main types of study designs in environmental epidemiology
and some of the salient features of each are presented in Table
2.1, and described further in the sections below.
2.5. Cross-sectional Studies
Cross-sectional studies, sometimes called prevalence studies,
provide information on disease frequency (prevalence) at a given
time. Estimates of exposures, and measurements of personal
characteristics and biological effects may be made at the same time
or may be derived from existing records. Thus, for example, an
investigator might pose the question: Are small opacities on a
chest radiograph more often found in welders than in other men (of
the same age)? He might attempt to answer this question by
obtaining 1000 chest radiographs of welders and 1000 chest radio-
graphs of other men. After mixing the films to ensure blinding
with respect to which film was of a welder and which of a non-
welder, the 2000 films would be examined and categorized by two
independent readers and then the changes observed would be
compared, between welders and the non-welders, within 5- or 10-year
age groups. This would be a pure cross-sectional study. In
practice, it is seldom that a cross-sectional study is so precisely
limited with respect to time.
Usually, historical information is collected so that a
retrospective component is included in the study. Thus, information
would be collected on past as well as current smoking habits, an
occupational history would be taken, comprising details of all jobs
held since leaving school, and often residential details of each
community in which the subject had lived, and dietary information
and data on any other present and past exposures of potential
significance would be obtained. On the disease side of the
equation, attempts are often made to establish the time of onset,
mode of development and course of disease, and any relevant
antecedent conditions. Thus, although the information may be
collected at one time, it often refers to events that may have
taken place over a period of years. Hospital records, information
from physicians about past episodes of disease, and any measurements
that may have been made on relevant environmental factors may be
used, if they are likely to contribute useful information to the
study.
(a) Choice of population
Cross-sectional studies are often designed to compare the
prevalence of disease in different places and in different groups
of people according to their measured, assessed, or surmised
exposures. The two most common population types that need to be
considered are: the general population, comprising the whole
community or some segment of it, based on age, sex, and race; and
the occupational group. The former will usually be more
appropriate to the investigation of wider community exposures (air
pollution, water quality and contaminants, effects of hot or cold
weather, neighbourhood pollution from some plant or factory).
Sometimes, the families of workers may be exposed to pollutants of
industrial origin not only because of local emissions, but also
through dust being brought home on the workers' clothes. Many
studies concentrate on the health of children (for example, Golubev
et al., 1979; Dantov et al., 1980); apart from the importance of
this topic in its own right, where concern is primarily with
general environmental agents, the confounding effects of
occupational exposures and of smoking can be minimized in this way.
Among adults, a single occupational group may be chosen for the
investigation of community problems, in order to avoid interference
from specific occupational factors.
Table 2.1. Major features of various study designs in environmental epidemiology
---------------------------------------------------------------------------------------------------------
Study Population Exposure Health effect Confounders Problems Advantages
design are:
---------------------------------------------------------------------------------------------------------
Descrip- Various Records Mortality and Difficult Hard to establish Cheap, useful
tive sub- of past morbidity to sort cause-result and to formulate
study populations measure- statistics, out exposure-effect hypothesis
ments case regist- relationships
ries, etc.
Cross- Community Current Current Usually easy Hard to establish Can be done
sectional or special to measure cause-relation- quickly; can
study groups; ship; current use large popu-
exposed vs. exposure may be lations; can
non-exposed irrelevant to estimate extent
groups current disease of problem
(prevalence)
Prospec- Community Defined at To be deter- Usually easy Expensive and Can estimate
tive or special outset of mined during to measure time consuming; incidence and
study groups; study (can course of exposure cate- relative risk;
exposed vs. change dur- study gories can can study many
non-exposed ing course change; high diseases; can
groups of study) dropout rate infer cause-
result rela-
tionship
Retro- Special Occurred in Occurred in Often Changes in Less expensive
spective groups such past - need past - need difficult exposure/effect and quicker
cohort as occupa- records records of to measure over time of than cohort
study tional groups, of past past diagnosis because of study; need to prospective
patients, measure- and measure- retrospective rely on records study giving
and insured ments ments nature (e.g., that may not be similar
persons past smoking accurate enough response, if
habits) sufficient
past records
are available
---------------------------------------------------------------------------------------------------------
Table 2.1. (contd.)
---------------------------------------------------------------------------------------------------------
Study Population Exposure Health effect Confounders Problems Advantages
design are:
---------------------------------------------------------------------------------------------------------
Time- Large com- Current, Current, Often Many confounding Useful for
series munity with e.g., daily e.g., daily difficult factors, often studies on
study several mil- changes in variations in to sort out, difficult to acute effects
lion people; exposure mortality e.g., effects measure
susceptible of influenza
groups such
as asthmatics
Case- Usually small Occurred Known at Possible to Difficult to Relatively
control groups; in past start eliminate generalize cheap and
study diseased and deter- of study by matching due to small quick; useful
(cases) vs. mined by for them study group; for studying
non-diseased records or some incor- rare diseases
(controls) interview porated biases
Experi- Community Controlled/ To be measured Can be Expensive; Well accepted
mental or special known during course measured; ethical results; strong
(inter- groups of study can be consider- evidence for
vention) controlled by ation study causality
study randomization subjects'
of subjects compliance
required;
drop-outs
Monitor- Community Current Current Difficult Difficult to Cheap when
ing and or special to sort out relate exposure using existing
surveil- groups data with monitoring and
lance effects surveillance
data
---------------------------------------------------------------------------------------------------------
(b) Assessment of exposure and effects on health
The index of occurrence of disease in a cross-sectional study
is prevalence, or the prevalence rate, i.e., the number of persons
in the group who are affected, expressed as a proportion of the
total number in the group. For physiological or biochemical
variables, the average and the distribution are the parameters of
interest (section 6.3.7.1). However, as in the case of exposure,
some assessment of the onset, development, and progression of the
effect may be obtained from judicious questioning; available
information may also be sought from records.
(c) Confounding variables
It is not possible to list all the confounding factors that
need to be considered. These will vary from study to study
according to the condition under investigation and, in many cases,
it may not be possible to avoid confounding factors entirely.
However, it is necessary to ensure that potential confounding
variables are identified at the design stage and that all the
available information on them is recorded. Unless a single sex/age
group is being examined, it may be necessary to ensure that the
contrasting groups that are being selected for study have similar
age and sex distributions, by stratified random sampling. For age,
10-year groups are sufficient for most purposes. Smoking has been
found to be such an important factor in so many of the effects
likely to be investigated, that it should always be recorded. Some
index of social circumstances, number of years of education,
occupation, type and quality of housing, degree of overcrowding and
so on should often be included. Other factors will need to be
considered in certain studies though not necessarily in all. In
short, the appropriate attention to confounding factors can only be
given if the epidemiological and other knowledge about the causation
of the effect of interest is carefully reviewed before and during
the design stage.
(d) Analysis
In many parts of the world, only limited and non-specialized
statistical help may be available for research workers. The
absence of elaborate statistical facilities should not deter would-
be researchers from undertaking prevalence surveys. Full exploitation
of results from such studies may require the application of fairly
complex methods, but important new knowledge about relationships
between environmental factors and indices of health can be
established without sophisticated statistics. The essential
requirements are: attention to the principles of study design
mentioned above; conscientious adherence to protocols and survey
methods (chapter 5); and careful description of the results, as
discussed in section 6.3.
(e) Advantages and disadvantages
A cross-sectional study may provide the answers to many
questions. Thus, this method has been extensively used to compare
the prevalence of respiratory symptoms and levels of lung function
in different groups of people, living in different places and
working in different jobs with various potential levels of
exposure. Prevalence studies have been used to study such diverse
chronic conditions as rheumatoid arthritis, asymptomatic bacteriuria,
diabetes mellitus, hypertension, peptic ulcer, stroke, and coronary
disease. In the occupational setting, cross-sectional studies of
exposure to chemicals, dusts, fumes, and gases have often provided
valua