Conditional Probability in the Diagnosis of Coronary Artery Disease: A Future Tool for Eliminating Unnecessary Testing?

Approximately 30% of the 300,000 coronary arteriograms done annually in the United States yield normal results. To see if the use of conditional probability in the diagnosis of coronary artery disease (CAD) could aid in predicting these normal results, we retrospectively assessed 96 patients with chest pain but without prior myocardial infarction. The probability of CAD was calculated by a computer program (CADENZA) using data from the history, exercise electrocardiography, and thallium-201 scintigraphy. All patients had coronary arteriography for definitive diagnosis. Based on preangiogram probabilities, different cutoff points were used to separate patients with angiograms likely to be normal from those likely to be abnormal. A preangiogram probability of 20% appeared to be the best separator. If the computer program, with this cutoff point, had been used to influence the decision to perform angiography, 38 of the 42 patients with negative angiograms would have been advised against it, and only two of the 54 patients who later were found to have CAD would have been missed. In view of the costs and risks of the CAD diagnostic process, the ability to identify the likelihood of positive angiograms could be useful in planning testing strategy, both in terms of avoiding “unnecessary” angiograms and avoiding redundant noninvasive procedures.