Long-term mortality study of oil refinery workers: V. Comparison of workers hired before, during, and after World War II (1940-1945) with a discussion of the impact of study designs on cohort results.

The mortality experience of a large refinery cohort (1937-1978) was examined by dividing it into three subcohorts according to hire dates: those hired before 1940, those hired during the period 1940-1945, and those hired after 1945. These three periods are approximately equivalent to before, during, and after World War II and span a total hiring period of more than 75 years. The results showed that a substantial portion of the cohort (3,330 or 27%) had been recruited during 1940-1945, and they contributed 980 or 28% of the total deaths. However, their mortality experience was quite different from the rest. A series of significant increases were seen among the external causes for accidents, suicide, and homicide. In terms of overall mortality and in contrast to the rest of the cohort, no "healthy worker effect" was seen (SMR = 1.00). They also showed increases in several types of cancer including cancers of the pancreas and prostate and leukemia. These unusual experiences cannot be explained either on the basis of their war-related deaths or on their period of employment (one-half were terminated within 1 year from date of hire), and data is insufficient to separate the role of hiring practices or their socioeconomic status. However, their life-styles were probably quite different judged from the fact that alcoholism-related deaths were increased as much as fivefold. Almost two-thirds of the total deaths occurred among 4,080 workers in the before 1940 subcohort. Further, the 5,117 workers of the after-1945 subcohort contributed only 5% of the total deaths. Thus, the results of the original refinery cohort (1937-1978) primarily reflect the experience of those employees hired before 1940. Given the same cohort method (historical prospective), cohort results vary widely according to different study designs, and this has implications for "generalizable" risk assessment or risk projections. A prospective study of new hires with 30 years of follow-up is rather inefficient because it would yield only a small number of deaths, with a strong healthy worker effect. The same is true for studies based on active workers with a short period of follow-up. Studies based on time of hire, however, provide a means for controlling time-related occupational exposures.