From 1978 to 1985, 470 consecutive male patients with complaints of chest pain underwent a maximal exercise test with a thallium scan and coronary angiography (CA). Patients with a history of myocardial infarction (MI) were excluded. During the follow-up (from 12 to 96 months), 32 patients died and 30 had a non-fatal MI. Survival (SR) and event-free rates (EFR) were estimated by actuarial methods; the influence of non-invasive and invasive variables were examined in univariate and multivariate models using Cox analysis. The five-year SR was 89% and EFR was 81%. Among historical data, age (less than 0.001), type of complaints (less than 0.01) and pretest likelihood of CAD (less than 0.01) were univariate predictors of EFR; by multivariate analysis, age was the only significant predictor (less than 0.001). Most of the maximal-exercise (MEX) test data were good univariate predictors; by multivariate analysis, workload (less than 0.001) and the maximal-exercise test score (less than 0.001) were the significant predictors. From history and maximal-exercise test data, multivariate analysis indicated that the prognostic information was given by age (less than 0.05) and maximal-exercise test score (less than 0.001). Among the invasive data, the number of diseased vessels (less than 0.001) and ejection fraction were the predictors. The combination of invasive and non-invasive data indicated that age, MEX score, number of diseased vessels and ejection fraction contributed significantly and independently to the estimation of prognosis. Among 242 patients with two or three diseased vessels, the prognosis was determined by the maximal workload (less than 0.01); ejection fraction (less than 0.07) was no longer significant.(ABSTRACT TRUNCATED AT 250 WORDS)