A review of methodological issues of the standardized mortality ratio (SMR) in occupational cohort studies.

This paper is a review and clarification of methodological issues related to the standardized mortality ratio (SMR), widely used in occupational epidemiology. Although the SMR seems to be a simple statistic, it can be misused and misinterpreted. The paper discusses SMRs in relation to age-specific mortality ratios, relative risk, life expectancy, and statistics derived from direct standardization. Inter-SMR comparisons and the effect of the choice of the comparison population on the SMR are also discussed. Mathematical and empirical review of SMR use has led to the development of a correction procedure which permits direct comparisons of two SMRs by adjusting for the age distributions of the two populations. The paper also proposes a method for testing the trend of age-specific mortality ratios (MRs). If MRs are homogeneous, this method can also be used to compare two SMRs. If MRs are not homogeneous, the relative risk between two sets of age-specific MRs can be tested for significance, but no summary index, including the SMR, can accurately describe the mortality experience over the entire age range. It is suggested that the summary statistics of a cohort mortality experience should include relative risk, attributable risk, and life expectancy. These statistics are complementary and can be derived or approximated from the data that generated the SMR. A valid risk assessment should weigh the evidence from all three summary statistics.