Interval estimation of the attributable risk for multiple exposure levels in case-control studies.

In case-control studies, exposure to a risk factor often occurs at several levels, so the attributable risk at each level is of interest. In this paper, estimation for the 2 X 2 table (case-control status versus dichotomous exposure) and the 2 X k table (case-control status versus exposure at several levels) are reviewed along with an example. A method for finding confidence intervals for attributable risk in the 2 X k table is proposed, and its application to estimates adjusted across strata (the 2 X k X s case) is indicated. The results of a Monte Carlo study of the procedure demonstrate that the nominal and actual coverage probabilities agree satisfactorily for practical applications.