In our article on scientific and ethical issues of genetic screening of workers, we reported an incorrect calculation of the proportion of disease attributable to genetic predisposition [l]. The formula attributable risk = [(OR - l)/OR] x P,  should correctly include the proportion (Pe,) of slow acetylators among the cases caused by the relevant exposure ("excess cases"), and not among the whole exposed population. The proportion of slow acetylators among the excess cases may be 100%, if only the slow acetylators are susceptible to the action of aromatic amines. In this extreme case, the attributable risk would exceed 50% when the OR is greater than 2.0, and approach 100% as the OR increases. Therefore, there is not a 50% upper bound for the proportion of (excess) cancers attributable to the slow genotype. Nonetheless, a key difference exists between preemployment selection of susceptible workers and primary prevention. Primary prevention will prevent, by definition, the whole excess of cases occurring above the number expected, based on general population rates, whereas screening would achieve this same result only in the extreme situation in which the whole excess is concentrated the slow acetylators (no excess case occurring among the exposed fast acetylators) and the screening test has 100% sensitivity and specificity. Short of these rather unrealistic circumstances the screening strategy will always prevent a smaller number of (excess) cases than the primary prevention strategy.