ABSTRACT
Typically, the exposures that are beyond individual control affect many people simultaneously. Measurement of individual exposures is thus difficult and costly. As a result, environmental epidemiological investigations often have to rely on the use of existing data, and to analyse these at the aggregate rather than individual level. It is also important to appreciate that epidemiological studies require more than data on exposure and health. Equally important are data on other known or possible risk factors which may confound relationships with the health outcomes of interest. Environmental exposures often have small effects that may be masked or distorted by the effects of confounding. Observed health effects of air pollution, for example, may be confounded by risk factors such as smoking or occupational exposures. Socio-economic factors act as confounders for many environmental health effects. The assessment of effect modification (i.e. the change of the strength of the association between exposure and health outcome according to some other factor) is also important for generalising observed exposure-effect relationships to other populations. In environmental epidemiology, problems connected with ecological analyses (i.e. problems with inference based on grouped data) call for further methodological work. For example, by obtaining individual-level data on the exposure and certain covariates in samples of selected groups, it might be possible to determine the
* This chapter was prepared by T.Nurminen, M.Nurminen, C.Corvalán and D.Briggs
limits of ecological bias in estimating the health effects (Morgenstern and Thomas, 1993; Prentice and Thomas, 1993).
