Prospective application of the multivariate approach to enhance the accuracy of the treadmill stress test

Thirty normal subjects and 84 patients with coronary artery disease were subjected to stress testing and the results assessed by multivariate analysis. Probabilities for the presence or absence of disease were determined based on the results of a previous retrospective study. The multivariate analysis approach significantly improved the sensitivity, from 73% to 88%, and correct classification rate, from 78% to 88%, when compared to the ST response (P less than 0.05). The specificity remained significantly unchanged at 93% and 87% respectively in comparison to the ST response. The predictive value of a positive and negative test were not significantly different from those values from the ST response. Although the probability of disease could separate normal from diseased subjects, it could not separate single vessel from multi-vessel disease. The correct classification rate remained the same for all probabilities greater than or equal to 50% suggesting that a higher probability was not more accurate in detecting disease. The multivariate approach did not improve the accuracy of the stress test over the ST response in women, but the females in this study were a small and highly select group. Multivariate analysis appears to be a valuable method in detecting disease and appears to improve diagnostic accuracy over the ST response alone. However, further studies are needed to determine the final role of such a system, particularly in female subjects.