Although the effort of Valberg and Watson (1996) to incorporate epidemiologic data into an assessment of the plausibility of carcinogenicity estimates is to be commended, their analysis contains a fundamental flaw. The authors equate comparison of observed rates with expected rates, a common technique in epidemiology, to a comparison of the number of observed events with an expected number of events. The two comparisons are not equivalent. The only circumstance in which they are equivalent is when the same denominator of person-time produces both the observed and the expected events.
