Of the 118 patients included in the analysis, 26 (22%) had cardiac events, nine during the training period and 17 during the follow-up. All events were nonfatal. As shown in (Table le6), cardiac events were more frequent in the control group (19 vs. 7, p = 0.008). During the initial six months after CA, two patients in the training group and seven patients in the control group had cardiac events. The two patients in the training group were withdrawn from the exercise program because of unstable angina. Both patients had coronary restenosis. One of them underwent a new angioplastic procedure at the site of previous PTCA, while the other had CABG (three-vessel disease). Of the seven untrained patients, five underwent CS, three at the site of a previous PTCA and two on a new significant stenosis; two patients had an AMI with no evidence of angiographic restenosis. During the follow-up period, of the trained patients, two had a new procedure (i.e., CS); two underwent CABG, and one had an AMI. Among untrained patients, six underwent a new PTCA; five underwent CABG, and one had an AMI. In the survival model (Figure 1), trained patients had a significantly lower event rate than control patients (log-rank 8.15, p < 0.005). Multivariate analysis demonstrated that the improvement in events was not collinear with changes in the coronary risk profile and thallium uptake, and the only independent predictor of events was exercise training (p = 0.008). Trained patients also had a lower rate of hospital readmission than control patients group (18.6% vs. 46%, RR: 0.69, 95% CI: 0.55 to 0.93, p < 0.001) and a reduced need for further revascularization (10% vs. 27%, RR: 0.78, 95% CI: 0.66 to 0.89, p = 0.03). Causes of hospital readmission were angina pectoris (6 patients from group T, 17 patients from group C), atypical chest pain (three patients from group C) and low threshold myocardial ischemia during exercise testing (five patients from group T, one patient from group C). As shown in (Figure 2), there was a significant difference between hospital readmission curves when separated by exercise training (log rank: 12.51, p < 0.001).