How much prognostic information do exercise test data add to clinical data after acute myocardial infarction?

We compared the prognostic significance of variables from a symptom-limited bicycle exercise test performed three weeks after acute myocardial infarction with clinical variables from the hospitalization utilizing discriminant function analysis for one year follow-up. Clinical exclusion criteria define a high-risk group of patients not eligible for exercise testing. Among 430 patients without test 28.4% died versus 6.6% among 456 patients with test (P less than 0.0001) and 21.4% versus 9.6% experienced a new infarction (P less than 0.001). The most important exercise test variables were duration of exercise and ventricular premature beats. The prediction of death was not significantly different when clinical variables, exercise test variables or the combined set were used for positive predictive value (66.7%, 60.0%, 63.3%) and negative predictive value (71.6%, 67.4%, 71.8%). For new infarction the positive predictive value was significantly higher for exercise test variables (56.8%, P less than 0.03) or the combined set (59.3%, P less than 0.02) compared to clinical variables (45.5%). The negative predictive value was significantly higher for clinical variables (79.6%) compared to exercise data (63.6%, P less than 0.0001) or the combined set (67.0%, P less than 0.001). We conclude that exercise data cannot improve prediction of death but provides higher positive predictive value for the prediction of new infarction.