Methodological issues in the use of vital statistics state registries in occupational mortality studies
Description
Purpose. To determine the extent to which the MOR and PMR correctly identify true underlying associations between certain occupations/industries and mortality from specific cancers and what the important determinants are that contribute towards valid estimation of risk when they are employed Methodology. A literature review of relevant occupational risk factors for lung cancer, pleural cancer, acute myeloid leukemia and soft tissue sarcoma was conducted to determine the extent to which the study findings were supported by other studies. Death certificate data from 26 states in the United States that routinely code occupation and industry on the death certificates of decedents between 1985 and 1998 were used for this study. Both proportionate mortality and case-control studies were used to investigate the extent to which PMR and MOR were influenced by: (1) the choice of three different reference populations; and (2) reliability of the cause of death information as measured by detection and confirmation rates. It also determined whether the direction of bias remained consistent for cancers with well established (asbestos, benzene) and those with less established (solvents, diesel) carcinogens in certain occupations. Finally, similar but limited analyses were made using industry rather than occupation to determine whether estimates of the MOR and PMR differed Results. The results from the MOR and PMR investigation correctly identified risks for lung cancer, mesothelioma and acute myeloid leukemia, although they did not reflect the estimates published in the literature. An improvement in the odds ratio was observed for only lung and pleural/peritoneal cancer and soft tissue neoplasms when cancers caused by the same exposure under investigation were removed from the control group. In the case of the PMR, the proportionate cancer mortality ratio (PMCR) was an improvement over the PMR since it was possible to control for the healthy worker effect Both the MOR and PMR were influenced by the use of different measures of exposure (occupation titles, industry titles and a combination of both) Conclusions. The choice of reference diseases, type of exposure coding and confirmation and detection rates determines whether the MOR and PMR reliably identify associations between certain occupations and industries and mortality from specific cancers. Investigations based on exclusively on death certificate data ought not to be interpreted without substantial support of other rigorous investigations such as cohort studies, but are a valuable tool in epidemiological investigations for hypothesis generating