Evaluation of the proportionate mortality index in the presence of multiple comparisons

We investigate the frequency in occupational studies of falsely identifying significantly elevated cause-specific mortality in the presence of multiple comparisons. When one investigates a standard battery of causes of death, use of a standardized mortality ratio as a summary index results in an excessive experimental error rate, but one can control the error rate with use of standard procedures that limit the type I error. When a major disease category for an occupational cohort has a lower rate than expected, then with use of a proportionate mortality ratio as a summary index, the standard procedures for control of the type I error are inadequate. In this situation, use of a proportionate mortality ratio as a summary index is likely to result in the false identification of an excess cause-specific mortality more often than previously acknowledged.