An improved mathematical method is presented for estimating the amount of fitness reduction due to a chronic disease by using demographic data. It is shown that Cavalli-Sforza and Bodmer's equivalent formula gives an underestimate. Application of the new formula indicates that the selective difference between blood groups O and A, resulting from their association with duodenal ulcer, is 6.4 X 10(-5), i.e. ten times higher than Cavalli-Sforza and Bodmer's estimate.