Comparison of the sensitivity and specificity of exercise electrocardiography in biased and unbiased populations of men and women.

To assess for sex-related differences in posttest referral bias, we compared the accuracy of exercise electrocardiography in biased (coronary angiography only) and unbiased (all unselected) populations with possible coronary disease. A retrospective analysis of clinical and exercise test data from 4467 patients (788 who underwent angiography) was performed (2824 men and 1643 women). The accuracy of a positive exercise test result was assessed in the entire unbiased group with a method that used disease probability (derived with a logistic algorithm) rather than angiography results. We found that the sensitivity and specificity were significantly greater in men than in women with use of the biased or unbiased groups. When the results for the unbiased and biased groups were compared, the sensitivities for the unbiased group were significantly lower and the specificities were significantly higher than those of the biased group. These differences reflect the effects of posttest referral bias. The amounts that sensitivity decreased and specificity increased, however, was not different for men and women. Therefore, we conclude that the accuracy of exercise electrocardiography is lower in women than men irrespective of whether a biased or an unbiased group is used. However, these differences cannot be explained on the basis of sex-related differences in posttest referral bias.