Abstract
It is well known that comparisons of indirectly adjusted statistics, such as standardized morbidity ratios (SMRs), can be biased when the joint effect of the study exposure and thecovariates is not multiplicative. This paper describes a related bias in SMR comparisons, one that can arise if the total study population is used as the referent. This bias can occur even if the exposure-covariate effects are multiplicative. (Author/publisher)