Epidemiology is recognized both as the science basic to preventive medicine and one that informs the public health policy process. Several operational definitions of epidemiology have been suggested. The simplest is that epidemiology is the study of the occurrence of disease or other health-related characteristics in human and in animal populations. Epidemiologists study not only the frequency of disease, but whether the frequency differs across groups of people; i.e., they study the cause-effect relationship between exposure and illness. Diseases do not occur at random; they have causes—quite often man-made causes—which are avoidable. Thus, many diseases could be prevented if the causes were known. The methods of epidemiology have been crucial to identifying many causative factors which, in turn, have led to health policies designed to prevent disease, injury and premature death.
What is the task of epidemiology and what are its strengths and weaknesses when definitions and concepts of epidemiology are applied to occupational health? This chapter addresses these questions and the ways in which occupational health hazards can be investigated using epidemiological techniques. This article introduces the ideas found in successive articles in this chapter.
Occupational epidemiology has been defined as the study of the effects of workplace exposures on the frequency and distribution of diseases and injuries in the population. Thus it is an exposure-oriented discipline with links to both epidemiology and occupational health (Checkoway et al. 1989). As such, it uses methods similar to those employed by epidemiology in general.
The main objective of occupational epidemiology is prevention through identifying the consequences of workplace exposures on health. This underscores the preventive focus of occupational epidemiology. Indeed, all research in the field of occupational health and safety should serve preventive purposes. Hence, epidemiological knowledge can and should be readily implementable. While the public health interest always should be the primary concern of epidemiological research, vested interests can exercise influence, and care must be taken to minimize such influence in the formulation, conduct and/or interpretation of studies (Soskolne 1985; Soskolne 1989).
A second objective of occupational epidemiology is to use results from specific settings to reduce or to eliminate hazards in the population at large. Thus, apart from providing information on the health effects of exposures in the workplace, the results from occupational epidemiology studies also play a role in the estimation of risk associated with the same exposures but at the lower levels generally experienced by the general population. Environmental contamination from industrial processes and products usually would result in lower levels of exposure than those experienced in the workplace.
The levels of application of occupational epidemiology are:
- surveillance to describe the occurrence of illness in different categories of workers and so provide early warning signals of unrecognized occupational hazards
- generation and testing of an hypothesis that a given exposure may be harmful, and the quantification of an effect
- evaluation of an intervention (for example, a preventive action such as reduction in exposure levels) by measuring changes in the health status of a population over time.
The causal role that occupational exposures can play in the development of disease, injury and premature death had been identified long ago and is part of the history of epidemiology. Reference has to be made to Bernardino Ramazzini, founder of occupational medicine and one of the first to revive and add to the Hippocratic tradition of the dependence of health on identifiable natural external factors. In the year 1700, he wrote in his “De Morbis Artificum Diatriba” (Ramazzini 1705; Saracci 1995):
The physician has to ask many questions of the patients. Hippocrates states in De Affectionibus: “When you face a sick person you should ask him from what he is suffering, for what reason, for how many days, what he eats, and what are his bowel movements. To all these questions one should be added: ‘What work does he do?’.”
This reawakening of clinical observation and of the attention to the circumstances surrounding the occurrence of disease, brought Ramazzini to identify and describe many of the occupational diseases that were later studied by occupational physicians and epidemiologists.
Using this approach, Pott was first to report in 1775 (Pott 1775) the possible connection between cancer and occupation (Clayson 1962). His observations on cancer of the scrotum among chimney-sweeps began with a description of the disease and continued:
The fate of these people seems singularly hard: in their early infancy, they are most frequently treated with great brutality, and almost starved with cold and hunger; they are thrust up narrow, and sometimes hot chimneys, where they are bruised, burned and almost suffocated; and when they get to puberty, become peculiarly liable to a most noisome, painful, and fatal disease.
Of this last circumstance there is not the least doubt, though perhaps it may not have been sufficiently attended to, to make it generally known. Other people have cancer of the same parts; and so have others, besides lead-workers, the Poitou colic, and the consequent paralysis; but it is nevertheless a disease to which they are peculiarly liable; and so are chimney-sweeps to cancer of the scrotum and testicles.
The disease, in these people, seems to derive its origin from a lodgement of soot in the rugae of the scrotum, and at first not to be a disease of the habit … but here the subjects are young, in general good health, at least at first; the disease brought on them by their occupation, and in all probability local; which last circumstance may, I think, be fairly presumed from its always seizing the same parts; all this makes it (at first) a very different case from a cancer which appears in an elderly man.
This first account of an occupational cancer still remains a model of lucidity. The nature of the disease, the occupation concerned and the probable causal agent are all clearly defined. An increased incidence of scrotal cancer among chimney-sweeps is noted although no quantitative data are given to substantiate the claim.
Another fifty years passed before Ayrton-Paris noticed in 1822 (Ayrton-Paris 1822) the frequent development of scrotal cancers among the copper and tin smelters of Cornwall, and surmised that arsenic fumes might be the causal agent. Von Volkmann reported in 1874 skin tumours in paraffin workers in Saxony, and shortly afterwards, Bell suggested in 1876 that shale oil was responsible for cutaneous cancer (Von Volkmann 1874; Bell 1876). Reports of the occupational origin of cancer then became relatively more frequent (Clayson 1962).
Among the early observations of occupational diseases was the increased occurrence of lung cancer among Schneeberg miners (Harting and Hesse 1879). It is noteworthy (and tragic) that a recent case study shows that the epidemic of lung cancer in Schneeberg is still a huge public health problem, more than a century after the first observation in 1879. An approach to identify an “increase” in disease and even to quantify it had been present in the history of occupational medicine. For example, as Axelson (1994) has pointed out, W.A. Guy in 1843 studied “pulmonary consumption” in letter press printers and found a higher risk among compositors than among pressmen; this was done by applying a design similar to the case-control approach (Lilienfeld and Lilienfeld 1979). Nevertheless, it was not until perhaps the early 1950s that modern occupational epidemiology and its methodology began to develop. Major contributions marking this development were the studies on bladder cancer in dye workers (Case and Hosker 1954) and lung cancer among gas workers (Doll 1952).
Issues in Occupational Epidemiology
The articles in this chapter introduce both the philosophy and the tools of epidemiological investigation. They focus on assessing the exposure experience of workers and on the diseases that arise in these populations. Issues in drawing valid conclusions about possible causative links in the pathway from exposures to hazardous substances to the development of diseases are addressed in this chapter.
Ascertainment of an individual’s work life exposure experience constitutes the core of occupational epidemiology. The informativeness of an epidemiological study depends, in the first instance, on the quality and extent of available exposure data. Secondly, the health effects (or, the diseases) of concern to the occupational epidemiologist must be accurately determinable among a well-defined and accessible group of workers. Finally, data about other potential influences on the disease of interest should be available to the epidemiologist so that any occupational exposure effects that are established from the study can be attributed to the occupational exposure per se rather than to other known causes of the disease in question. For example, in a group of workers who may work with a chemical that is suspected of causing lung cancer, some workers may also have a history of tobacco smoking, a further cause of lung cancer. In the latter situation, occupational epidemiologists must determine which exposure (or, which risk factor—the chemical or the tobacco, or, indeed, the two in combination) is responsible for any increase in the risk of lung cancer in the group of workers being studied.
If a study has access only to the fact that a worker was employed in a particular industry, then the results from such a study can link health effects only to that industry. Likewise, if knowledge about exposure exists for the occupations of the workers, conclusions can be directly drawn only in so far as occupations are concerned. Indirect inferences on chemical exposures can be made, but their reliability has to be evaluated situation by situation. If a study has access, however, to information about the department and/or job title of each worker, then conclusions will be able to be made to that finer level of workplace experience. Where information about the actual substances with which a person works is known to the epidemiologist (in collaboration with an industrial hygienist), then this would be the finest level of exposure information available in the absence of rarely available dosimetry. Furthermore, the findings from such studies can provide more useful information to industry for creating safer workplaces.
Epidemiology has been a sort of “black box” discipline until now, because it has studied the relationship between exposure and disease (the two extremes of the causal chain), without considering the intermediate mechanistic steps. This approach, despite its apparent lack of refinement, has been extremely useful: in fact, all the known causes of cancer in humans, for instance, have been discovered with the tools of epidemiology.
The epidemiological method is based on available records —questionnaires, job titles or other “proxies” of exposure; this makes the conduct of epidemiological studies and the interpretation of their findings relatively simple.
Limitations of the more crude approach to exposure assessment, however, have become evident in recent years, with epidemiologists facing more complex problems. Limiting our consideration to occupational cancer epidemiology, most well-known risk factors have been discovered because of high levels of exposure in the past; a limited number of exposures for each job; large populations of exposed workers; and a clear-cut correspondence between “proxy” information and chemical exposures (e.g., shoe workers and benzene, shipyards and asbestos, and so on). Nowadays, the situation is substantially different: levels of exposure are considerably lower in Western countries (this qualification should always be stressed); workers are exposed to many different chemicals and mixtures in the same job title (e.g., agricultural workers); homogeneous populations of exposed workers are more difficult to find and are usually small in number; and, the correspondence between “proxy” information and actual exposure grows progressively weaker. In this context, the tools of epidemiology have reduced sensitivity owing to the misclassification of exposure.
In addition, epidemiology has relied on “hard” end points, such as death in most cohort studies. However, workers might prefer to see something different from “body counts” when the potential health effects of occupational exposures are studied. Therefore, the use of more direct indicators of both exposure and early response would have some advantages. Biological markers may provide just a tool.
The use of biological markers, such as lead levels in blood or liver function tests, is not new in occupational epidemiology. However, the utilization of molecular techniques in epidemiological studies has made possible the use of biomarkers for assessing target organ exposures, for determining susceptibility and for establishing early disease.
Potential uses of biomarkers in the context of occupational epidemiology are:
- exposure assessment in cases in which traditional epidemiological tools are insufficient (particularly for low doses and low risks)
- to disentangle the causative role of single chemical agents or substances in multiple exposures or mixtures
- estimation of the total burden of exposure to chemicals having the same mechanistic target
- investigation of pathogenetic mechanisms
- study of individual susceptibility (e.g., metabolic polymorphisms, DNA repair) (Vineis 1992)
- to classify exposure and/or disease more accurately, thereby increasing statistical power.
Great enthusiasm has arisen in the scientific community about these uses, but, as noted above, methodological complexity of the use of these new “molecular tools” should serve to caution against excessive optimism. Biomarkers of chemical exposures (such as DNA adducts) have several shortcomings:
- They usually reflect recent exposures and, therefore, are of limited use in case-control studies, whereas they require repeated samplings over prolonged periods for utilization in cohort investigations.
- While they can be highly specific and thus improve exposure misclassification, findings often remain difficult to interpret.
- When complex chemical exposures are investigated (e.g., air pollution or environmental tobacco smoke) it is possible that the biomarker would reflect one particular component of the mixture, whereas the biological effect could be due to another.
- In many situations, it is not clear whether a biomarker reflects a relevant exposure, a correlate of the relevant exposure, individual susceptibility, or an early disease stage, thus limiting causal inference.
- The determination of most biomarkers requires an expensive test or an invasive procedure or both, thus creating constraints for adequate study size and statistical power.
- A biomarker of exposure is no more than a proxy for the real objective of an epidemiological investigation, which, as a rule, focuses on an avoidable environmental exposure (Trichopoulos 1995; Pearce et al. 1995).
Even more important than the methodological shortcomings is the consideration that molecular techniques might cause us to redirect our focus from identifying risks in the exogenous environment, to identifying high-risk individuals and then making personalized risk assessments by measuring phenotype, adduct load and acquired mutations. This would direct our focus, as noted by McMichael, to a form of clinical evaluation, rather than one of public health epidemiology. Focusing on individuals could distract us from the important public health goal of creating a less hazardous environment (McMichael 1994).
Two further important issues emerge regarding the use of biomarkers:
- The use of biomarkers in occupational epidemiology must be accompanied by a clear policy as far as informed consent is concerned. The worker may have several reasons to refuse cooperation. One very practical reason is that the identification of, say, an alteration in an early response marker such as sister chromatid exchange implies the possibility of discrimination by health and life insurers and by employers who might shun the worker because he or she may be more prone to disease. A second reason concerns genetic screening: since the distributions of genotypes and phenotypes vary according to ethnic group, occupational opportunities for minorities might be hampered by genetic screening. Third, doubts can be raised about the predictability of genetic tests: since the predictive value depends on the prevalence of the condition which the test aims to identify, if the latter is rare, the predictive value will be low and the practical use of the screening test will be questionable. Until now, none of the genetic screening tests have been judged applicable in the field (Ashford et al. 1990).
- Ethical principles must be applied prior to the use of biomarkers. These principles have been evaluated for biomarkers used for identifying individual susceptibility to disease by an interdisciplinary Working Group of the Technical Office of the European Trade Unions, with the support of the Commission of the European Communities (Van Damme et al. 1995); their report has reinforced the view that tests can be conducted only with the objective of preventing disease in a workforce. Among other considerations, use of tests must never.
- serve as a means for “selection of the fittest”
- be used to avoid implementing effective preventive measures, such as the identification and substitution of risk factors or improvements in conditions in the workplace
- create, confirm or reinforce social inequality
- create a gap between the ethical principles followed in the workplace and the ethical principles that must be upheld in a democratic society
- oblige a person seeking employment to disclose personal details other than those strictly necessary for obtaining the job.
Finally, evidence is accumulating that the metabolic activation or inactivation of hazardous substances (and of carcinogens in particular) varies considerably in human populations, and is partly genetically determined. Furthermore, inter-individual variability in the susceptibility to carcinogens may be particularly important at low levels of occupational and environmental exposure (Vineis et al. 1994). Such findings may strongly affect regulatory decisions that focus the risk assessment process on the most susceptible (Vineis and Martone 1995).
Study design and validity
Hernberg’s article on epidemiological study designs and their applications in occupational medicine concentrates on the concept of “study base”, defined as the morbidity experience (in relation to some exposure) of a population while it is followed over time. Thus, the study base is not only a population (i.e., a group of people), but the experience of disease occurrence of this population during a certain period of time (Miettinen 1985, Hernberg 1992). If this unifying concept of a study base is adopted, then it is important to recognize that the different study designs (e.g., case-control and cohort designs) are simply different ways of “harvesting” information on both exposure and disease from the same study base; they are not diametrically different approaches.
The article on validity in study design by Sasco addresses definitions and the importance of confounding. Study investigators must always consider the possibility of confounding in occupational studies, and it can never be sufficiently stressed that the identification of potentially confounding variables is an integral part of any study design and analysis. Two aspects of confounding must be addressed in occupational epidemiology:
- Negative confounding should be explored: for example, some industrial populations have low exposure to lifestyle-associated risk factors because of a smoke-free workplace; glass blowers tend to smoke less than the general population.
- When confounding is considered, an estimate of its direction and its potential impact ought to be assessed. This is particularly true when data to control confounding are scanty. For example, smoking is an important confounder in occupational epidemiology and it always should be considered. Nevertheless, when data on smoking are not available (as is often the case in cohort studies), it is unlikely that smoking can explain a large excess of risk found in an occupational group. This is nicely described in a paper by Axelson (1978) and further discussed by Greenland (1987). When detailed data on both occupation and smoking have been available in the literature, confounding did not seem to heavily distort the estimates concerning the association between lung cancer and occupation (Vineis and Simonato 1991). Furthermore, suspected confounding does not always introduce non-valid associations. Since investigators also are at risk of being led astray by other undetected observation and selection biases, these should receive as much emphasis as the issue of confounding in designing a study (Stellman 1987).
Time and time-related variables such as age at risk, calendar period, time since hire, time since first exposure, duration of exposure and their treatment at the analysis stage, are among the most complex methodological issues in occupational epidemiology. They are not covered in this chapter, but two relevant and recent methodological references are noted (Pearce 1992; Robins et al. 1992).
The article on statistics by Biggeri and Braga, as well as the title of this chapter, indicate that statistical methods cannot be separated from epidemiological research. This is because: (a) a sound understanding of statistics may provide valuable insights into the proper design of an investigation and (b) statistics and epidemiology share a common heritage, and the entire quantitative basis of epidemiology is grounded in the notion of probability (Clayton 1992; Clayton and Hills 1993). In many of the articles that follow, empirical evidence and proof of hypothesized causal relationships are evaluated using probabilistic arguments and appropriate study designs. For example, emphasis is placed on estimating the risk measure of interest, like rates or relative risks, and on the construction of confidence intervals around these estimates instead of the execution of statistical tests of probability (Poole 1987; Gardner and Altman 1989; Greenland 1990). A brief introduction to statistical reasoning using the binomial distribution is provided. Statistics should be a companion to scientific reasoning. But it is worthless in the absence of properly designed and conducted research. Statisticians and epidemiologists are aware that the choice of methods determines what and the extent to which we make observations. The thoughtful choice of design options is therefore of fundamental importance in order to ensure valid observations.
The last article, by Vineis, addresses ethical issues in epidemiological research. Points to be mentioned in this introduction refer to epidemiology as a discipline that implies preventive action by definition. Specific ethical aspects with regard to the protection of workers and of the population at large require recognition that:
- Epidemiological studies in occupational settings should in no way delay preventive measures in the workplace.
- Occupational epidemiology does not refer to lifestyle factors, but to situations where usually little or no personal role is played in the choice of exposure. This implies a particular commitment to effective prevention and to the immediate transmission of information to workers and the public.
- Research uncovers health hazards and provides the knowledge for preventive action. The ethical problems of not carrying out research, when it is feasible, should be considered.
- Notification to workers of the results of epidemiological studies is both an ethical and methodological issue in risk communication. Research in evaluating the potential impact and effectiveness of notification should be given high priority (Schulte et al. 1993).
Training in occupational epidemiology
People with a diverse range of backgrounds can find their way into the specialization of occupational epidemiology. Medicine, nursing and statistics are some of the more likely backgrounds seen among those specializing in this area. In North America, about half of all trained epidemiologists have science backgrounds, while the other half will have proceeded along the doctor of medicine path. In countries outside North America, most specialists in occupational epidemiology will have advanced through the doctor of medicine ranks. In North America, those with medical training tend to be considered “content experts”, while those who are trained through the science route are deemed “methodological experts”. It is often advantageous for a content expert to team up with a methodological expert in order to design and conduct the best possible study.
Not only is knowledge of epidemiological methods, statistics and computers needed for the occupational epidemiology speciality, but so is knowledge of toxicology, industrial hygiene and disease registries (Merletti and Comba 1992). Because large studies can require linkage to disease registries, knowledge of sources of population data is useful. Knowledge of labour and corporate organization also is important. Theses at the masters level and dissertations at the doctoral level of training equip students with the knowledge needed for conducting large record-based and interview-based studies among workers.
Proportion of disease attributable to occupation
The proportion of disease which is attributable to occupational exposures either in a group of exposed workers or in the general population is covered at least with respect to cancer in another part of this Encyclopaedia. Here we should remember that if an estimate is computed, it should be for a specific disease (and a specific site in the case of cancer), a specific time period and a specific geographic area. Furthermore, it should be based on accurate measures of the proportion of exposed people and the degree of exposure. This implies that the proportion of disease attributable to occupation may vary from very low or zero in certain populations to very high in others located in industrial areas where, for example, as much as 40% of lung cancer can be attributable to occupational exposures (Vineis and Simonato 1991). Estimates which are not based on a detailed review of well-designed epidemiological studies can, at the very best, be considered as informed guesses, and are of limited value.
Transfer of hazardous industries
Most epidemiological research is carried out in the developed world, where regulation and control of known occupational hazards has reduced the risk of disease over the past several decades. At the same time, however, there has been a large transfer of hazardous industries to the developing world (Jeyaratnam 1994). Chemicals previously banned in the United States or Europe now are produced in developing countries. For example, asbestos milling has been transferred from the US to Mexico, and benzidine production from European countries to the former Yugoslavia and Korea (Simonato 1986; LaDou 1991; Pearce et al. 1994).
An indirect sign of the level of occupational risk and of the working conditions in the developing world is the epidemic of acute poisoning taking place in some of these countries. According to one assessment, there are about 20,000 deaths each year in the world from acute pesticide intoxication, but this is likely to be a substantial underestimate (Kogevinas et al. 1994). It has been estimated that 99% of all deaths from acute pesticide poisoning occur in developing countries, where only 20% of the world’s agrochemicals are used (Kogevinas et al. 1994). This is to say that even if the epidemiological research seems to point to a reduction of occupational hazards, this might simply be due to the fact that most of this research is being conducted in the developed world. The occupational hazards may simply have been transferred to the developing world and the total world occupational exposure burden might have increased (Vineis et al. 1995).
For obvious reasons, veterinary epidemiology is not directly pertinent to occupational health and occupational epidemiology. Nevertheless, clues to environmental and occupational causes of diseases may come from epidemiological studies on animals for several reasons:
- The life span of animals is relatively short compared with that of humans, and the latency period for diseases (e.g., most cancers) is shorter in animals than in humans. This implies that a disease that occurs in a wild or pet animal can serve as a sentinel event to alert us to the presence of a potential environmental toxicant or carcinogen for humans before it would have been identified by other means (Glickman 1993).
- Markers of exposures, such as haemoglobin adducts or levels of absorption and excretion of toxins, may be measured in wild and pet animals to assess environmental contamination from industrial sources (Blondin and Viau 1992; Reynolds et al. 1994; Hungerford et al. 1995).
- Animals are not exposed to some factors which may act as confounders in human studies, and investigations in animal populations therefore can be conducted without regard to these potential confounders. For example, a study of lung cancer in pet dogs might detect significant associations between the disease and exposure to asbestos (e.g., via owners’ asbestos-related occupations and proximity to industrial sources of asbestos). Clearly, such a study would remove the effect of active smoking as a confounder.
Veterinarians talk about an epidemiological revolution in veterinary medicine (Schwabe 1993) and textbooks about the discipline have appeared (Thrusfield 1986; Martin et al. 1987). Certainly, clues to environmental and occupational hazards have come from the joint efforts of human and animal epidemiologists. Among others, the effect of phenoxyherbicides in sheep and dogs (Newell et al. 1984; Hayes et al. 1990), of magnetic fields (Reif et al. 1995) and pesticides (notably flea preparations) contaminated with asbestos-like compounds in dogs (Glickman et al. 1983) are notable contributions.
Participatory research, communicating results and prevention
It is important to recognize that many epidemiological studies in the field of occupational health are initiated through the experience and concern of workers themselves (Olsen et al. 1991). Often, the workers—those historically and/or presently exposed—believed that something was wrong long before this was confirmed by research. Occupational epidemiology can be thought of as a way of “making sense” of the workers’ experience, of collecting and grouping the data in a systematic way, and allowing inferences to be made about the occupational causes of their ill health. Furthermore, the workers themselves, their representatives and the people in charge of workers’ health are the most appropriate persons to interpret the data which are collected. They therefore should always be active participants in any investigation conducted in the workplace. Only their direct involvement will guarantee that the workplace will remain safe after the researchers have left. The aim of any study is the use of the results in the prevention of disease and disability, and the success of this depends to a large extent on ensuring that the exposed participate in obtaining and interpreting the results of the study. The role and use of research findings in the litigation process as workers seek compensation for damages caused through workplace exposure is beyond the scope of this chapter. For some insight on this, the reader is referred elsewhere (Soskolne, Lilienfeld and Black 1994).
Participatory approaches to ensuring the conduct of occupational epidemiological research have in some places become standard practice in the form of steering committees established to oversee the research initiative from its inception to its completion. These committees are multipartite in their structure, including labour, science, management and/or government. With representatives of all stakeholder groups in the research process, the communication of results will be made more effective by virtue of their enhanced credibility because “one of their own” would have been overseeing the research and would be communicating the findings to his or her respective constituency. In this way, the greatest level of effective prevention is likely.
These and other participatory approaches in occupational health research are undertaken with the involvement of those who experience or are otherwise affected by the exposure-related problem of concern. This should be seen more commonly in all epidemiological research (Laurell et al. 1992). It is relevant to remember that while in epidemiological work the objective of analysis is estimation of the magnitude and distribution of risk, in participatory research, the preventability of the risk is also an objective (Loewenson and Biocca 1995). This complementarity of epidemiology and effective prevention is part of the message of this Encyclopaedia and of this chapter.
Maintaining public health relevance
Although new developments in epidemiological methodology, in data analysis and in exposure assessment and measurement (such as new molecular biological techniques) are welcome and important, they can also contribute to a reductionist approach focusing on individuals, rather than on populations. It has been said that:
… epidemiology has largely ceased to function as part of a multidisciplinary approach to understanding the causation of disease in populations and has become a set of generic methods for measuring associations of exposure and disease in individuals.… There is current neglect of social, economic, cultural, historical, political and other population factors as major causes of diseases.…Epidemiology must reintegrate itself into public health, and must rediscover the population perspective (Pearce 1996).
Occupational and environmental epidemiologists have an important role to play, not only in developing new epidemiological methods and applications for these methods, but also in ensuring that these methods are always integrated in the proper population perspective.