WHY AND HOW TO CONTROL FOR AGE IN OCCUPATIONAL EPIDEMIOLOGY

In occupational epidemiology, the need to consider the age factor prop erly influences the choice of study design and analytical techniques, In most studies, age is viewed as a potential confounder. Age is stron gly associated with end points of interest in occupational epidemiolog y (diseases, physiological characteristics, doses of xenobiotics, etc) , bur to measure age as a confounder it must be associated with the ex posure under study. When the exposure of interest is time related-for example, duration of employment, time since first exposure, cumulative exposure-a strong intrinsic association with age can be anticipated, and age will behave as a (usually strong) confounder. When occupationa l exposures without a direct relation with age-for example, job, depar tment, type of exposure-are evaluated, the degree and direction of con founding bias cannot be anticipated. Control of the confounding effect of age can be accomplished in the design phase of a study by way of r andomisation, restriction, and matching. Randomisation is seldom viabl e in occupational settings, Restriction is rarely used in the case of age. Matching is often used in a case-control study as a method to inc rease the study efficiency, but it must be followed by proper matched or stratified analysis. Options for age adjustment in the analysis pha se involve stratification and regression methods. In longitudinal stud ies the modified life table analysis is used to take into account the fact that subjects cross categories of age as the study proceeds. Stab ility of relative measures of effect over age strata favoured the grea ter use of relative risks than risk. differences. In the presence of e ffect modification the influence of age should not be eliminated; its interaction with exposure should be explicitly considered.