A multivariate analysis (MVA) system was derived retrospectively from a population of 76 males with coronary artery disease and 18 control subjects. Posterior probabilities were then derived from such a system prospectively in a new male population of 11 subjects with normal coronary arteries and hemodynamics and 63 patients with coronary artery disease. The sensitivity was 84% compared to that for change in ejection fraction (delta EF) greater than or equal to 5 criterion of 71% (p less than 0.01), the specificity was 91% compared to 73% for the delta EF greater than or equal to 5 criterion (p greater than 0.05), and the correct classification rate was 85% compared to 72% for the delta EF greater than or equal to 5 criterion (p less than 0.01). The significant variables were: change in EF with exercise, percent maximal heart rate, change in end-diastolic volume (delta EDV) with exercise, change in R wave, and exercise duration. Application of the multivariate approach to radionuclide imaging with stress, including both exercise and nuclear parameters, significantly improved the diagnostic accuracy of the test and allowed for a probability statement concerning the likelihood of disease.