Sex and test verification bias. Impact on the diagnostic value of exercise echocardiography

Abstract

Background: The use of exercise echocardiography for the diagnosis of coronary artery disease (CAD) has been validated in pilot studies but is not documented in clinical practice and in women comparatively with men. The objectives of this study were to determine the effects of sex and of test verification bias on the diagnostic performance of exercise echocardiography.

Methods and results: Three thousand six hundred seventy-nine consecutive patients (1714 women, 1965 men) who underwent an exercise echocardiographic study were studied; the observed sensitivity, specificity, and correct classification rate were calculated among 340 patients (244 men, 96 women) who underwent angiography; to study the effect of test verification bias, sensitivity and specificity were estimated for all patients who underwent exercise echocardiography including those not referred to angiography. In the angiographic group, the prevalence of CAD was 60% in women and 80% in men. The observed sensitivity and specificity of exercise echocardiography was 78% and 44% in men and 79% and 37% in women. After adjustment for test verification bias, the estimated sensitivity was lower in women (32% versus 42% in men), whereas specificity was similar in both sexes. The positive predictive value was lower in women (66%) compared with men (84%).

Conclusions: In clinical practice, test verification bias results in a lower observed specificity and a higher sensitivity of exercise echocardiography. In women, positive predictive value and adjusted sensitivity are lower compared with that in men.