Epidemiology is a multidisciplinary enterprise which assesses the causes, natural history and treatment of disease. It originates from the study of epidemics of communicable diseases, but is now much more concerned with chronic disease and health care delivery. The change of emphasis is determined by the relative burden of diseases and consequent expenditure of effort. Essentially the objectives of epidemiology are to minimize the unwanted consequence of diseases by informing appropriate policy to enable them to be prevented, or appropriately and efficiently treated.
Thus epidemiology is primarily the science of the causes of disease. It is concerned to study the variation in incidence of (or mortality from) disease with respect to persons, time and place so that patterns can be observed from which the component parts of a biologically plausible aetiological process can be constructed. This involves statistical methods, social observation and understanding, as well as biological knowledge. Epidemiology forms a basic science for public health, preventive medicine, health services research and health promotion.
In the development of epidemiology key figures have been John Snow (1813–58) who in the nineteenth century removed the handle from the Broad (now Broadwick) Street water pump in Soho, London, and at a stroke discovered the method of transmission of cholera, because people were no longer able to draw up the water, which was contaminated by raw sewage (Snow 1949). Florence Nightingale (1820–1910) was a seminal figure in epidemiology by emphasizing in her practice the importance of systematic observation of patients in order to learn the distinctive features that a disease course took.
Thomas McKeown (1979) identified and discussed the broad determinants of disease and sought to attribute the role of organized medicine in alleviating this burden, which he found to be importantly less than was supposed. Archie Cochrane (1972) took the argument further by identifying the extent of knowledge about outcomes consequent upon treatment, which he argued needed much more rigorous evaluation in medical practice by randomized trials. Richard Doll and Bradford Hill elucidated the relationship of cigarettes with disease by identifying a cohort of willing doctors who answered periodic questions about their smoking habits and enabled rigorous comparison of their death rates accordingly (Doll and Hill 1964).
Such is the legacy of epidemiology which currently has several branches and factions. Its methods consist of studying and encouraging the improvement of vital statistics on disease incidence and mortality. Sometimes cross-sectional studies are undertaken to look at particular patterns of disease prevalence and population characteristics. More often case control studies are used to investigate the relationships of past exposure to putative risk factors and a disease. Final assessment of causal relationships often has to await prospective cohort studies in which individuals who have been exposed are compared with those who have not. In exceptional circumstances randomized controlled trials are used to compare outcomes between those allocated at random to a particular exposure.
Since epidemiology is almost entirely itself empirical, the development of the subject depends on the amount of variation in disease incidence which is explained by established risk factors. The important theoretical components of epidemiology are largely statistical, social and biological. The statistical provides efficient means of deriving estimates of qualitatively different measures of risk and the biological constrains and informs the epidemiological investigation and their results. The social informs the transfer from discerning risk factors and measuring health to effective health promotion.
In terms of measuring risk, the most useful is the relative risk. This measure simply compares the incidence of a disease among groups of subjects exposed to a risk factor with that among those not exposed. Thus the relative risk of lung cancer among heavy cigarette smokers is around twenty, which says that over both genders and all ages, in any country, the annual incidence of lung cancer among heavy smokers is twenty times as high as among similar people who have never smoked. This then gives an indication of the potency of the aetiological effect. Clearly a relative risk of twenty is a good deal more potent than a relative risk of two. This latter might describe the potency of obesity with respect to coronary heart disease or, indeed, environmental tobacco smoke and lung cancer. This number is an aggregated statistic and in circumstances when the relative risk for a risk factor and a disease varies importantly the phenomenon is called effect modification. There is no intrinsic reason to suppose constant relative risks across heterogenous populations, but often effect modification is much less important than the aggregate effect itself.
An alternative measure of risk is estimated as the attributable risk per cent. This measures the amount of disease which is attributable to the risk factor being considered. Thus it estimates the proportion by which the current overall incidence of lung cancer would decrease if nobody had ever smoked. Thus around 86 per cent of the current incidence of lung cancer is attributable to cigarette smoking. In this respect epidemiology has achieved this degree of understanding, and the policy implication for an individual is importantly informed.
In contrast, another disease, more common among women in most communities than lung cancer, is breast cancer. For this disease the highest determined relative risk is around three, for women with an early age of starting periods, a late age of first pregnancy and a late menopause for instance. The magnitude of the attributable risks depend not only on the relative risks but also on the prevalence of these risk factors. In this case the attributable risk per cents are no higher than 15 per cent for each risk factor, and worse still the risk factors themselves are not readily amenable to incidental change.
Thus as time progresses the role of epidemiology becomes increasingly difficult as the larger effects are found and validated. The smaller effects of relative risks of around 1.5 are exceptionally difficult to establish epidemiologically because of measurement errors and the problem of confounding, where an association is observed which cannot be distinguished between an association with an unknown risk factor or a direct causal relationship.
However, these small effects are of paramount importance for common diseases because they represent the rule and not the exception and offer the greatest hope for the prevention of disease not amenable to treatment. Since most chronic diseases are not particularly amenable to treatment, progress must happen by constructive synergy between epidemiology with statistics, the social sciences and biology. Thus new genes which predispose to cancer, for incidence, will give rise to more precise estimates of other risks in new epidemiological studies.
Klim McPherson
London School of Hygiene and Tropical Medicine
References
Cochrane, A.L. (1972) Effectiveness and Efficiency, Oxford.
Doll, R. and Hill, A.B. (1964) ‘Mortality in relation to smoking: ten years’ observation of British doctors’, British Medical Journal 269.
McKeown, T. (1979) The Role of Medicine: Dream, Mirage or Nemesis?, Oxford.
Snow, J. (1949) Snow on Cholera, Cambridge; MA.
Further reading
Ashton, J. (1994) The Epidemiological Imagination, Buckingham.
Buck, C., Llopis, A., Najera, E. and Rerris, M. (eds) (1989) The Challenge of Epidemiology: Issues and Selected Readings, Buckingham.
