Relative excess risk: an alternative measure of comparative risk.

The proverbial relative risk may not always be the most suitable measure to compare the risk of two exposures, since it inherently includes a background effect. An alternative comparative measure, the relative excess risk, is introduced. It applies to situations in which an "unexposed" reference group is included in addition to the two exposures under evaluation. This comparative measure is based solely on the component of risk due to the exposures, since it removes the background risk. Estimators of the relative excess risk are presented, along with formulas for the confidence intervals under cohort and case-control designs, using both crude and adjusted rate ratios. This new measure is illustrated with data from epidemiologic studies of the risks of oral contraceptives and antidiabetic drugs.