The impact of competing risks on relative risks in occupational cohort studies.

This paper develops a method for adjustment of competing causes of death in the calculation of relative risk. It has identified three factors determining the significance of competing risks: (1) the magnitude of the overall mortality risk of the study population; (2) differential risk (or the adjustment factor) for mortality other than cause of interest between two populations; and (3) age intervals used in the mortality calculation. Thus, the impact of competing risks is increased if the study cohort has a high mortality risk, if the mortality other than the cause of interest has a large differential risk or if the mortality calculation uses wide age intervals. Two examples from a refinery cohort and the US national population show that among certain age groups unadjusted for competing risks the relative risk is overestimated by 9%. The impact of competing risks in these two particular examples is relatively small. Furthermore, if relative risk is expressed in terms of the ratio of mortality rates, competing risks can be ignored.