To study individual exposures with any degree of accuracy it is usually necessary to attempt studies based on workforces rather than communities, with all the problems that this entails. There are some jobs where workers are expected to stay for a lifetime, and which are closely associated with the communities in which they live. Mining, cotton spinning and farming are good examples. In these situations the biases are not quite so big. Nevertheless biases are still considerable. Those starting work of any kind are in general more healthy than the general population (the healthy worker effect). There is evidence that workers starting exposure to particularly dusty environments have even higher lung function than the general working population. This is illustrated in Figure 5.1 from the French PAARC study,3 originally set up to investigate the effects of air pollution. Households whose head was not a manual worker were sampled, the difference between the FEV1 of workers exposed to dust or fume was compared with age-matched workers without such exposure. The FEV1 was higher in the younger exposed workers and lower in the older exposed workers. If this population was studied cross-sectionally, a population under 40 years old would have higher levels of FEV1 if exposed, those about 40 would have no effect, and those over 50 would show an effect of exposure. Longitudinal studies need to include all those starting the study in the follow-up measurements. Studies of specific workforces in particular have
a major problem in following up workers who have left employment. There is a preferential loss from the workforce of those who show the greatest effect of FEV1 decline. Longitudinal studies tend to underestimate any effect, and cross-sectional studies may show a diminishing effect of higher exposures. This is illustrated in a study of silicaexposed gold miners10 where the prevalence of bronchitis is related to cumulative silica exposure. There was increasing bronchitis with increasing exposures up to 30 000 particle years, then a decrease, probably due to loss of the most affected from the cohort (Figure 5.2). Similar results were obtained in longitudinal studies of British and US coal miners, where those with the highest cumulative coal dust exposure had the lowest rates of longitudinal decline in FEV1.