The authors are grateful for the suggestions and

arising from an omitted risk factor. Am J Epidemiol 1989; 129:850-6. The authors describe a form of selection bias that may arise when a second disease selectively removes from the population persons susceptible to the primary disease of interest Two examples of this bias are given: 1) a lack of association between an exposure and the primary disease may appear as an inverse association, and 2) a direct association between exposure and primary disease may be greatly attenuated. These examples of bias require the presence of an unknown risk factor in addition to the exposure of interest epidemiologic methods; risk; risk factors Cause-specific hazard rates are ordinar-ily estimated by the life table method. This method directly estimates the conditional probability of a person's dying from the specific cause in a certain time interval, given that the person survives to the begin-ning of the interval. The cause-specific haz-ard generated by this procedure has been called the "crude " hazard function (1). The life table method rests on the assumption that any other causes of loss to follow-up Received for publication July 21, 1987, and in final form June 13, 1988