Of 22 randomized trials of rehabilitation with exercise after myocardial infarction (MI), one trial had results that achieved conventional statistical significance. To determine whether or not these studies, in the aggregate, show a significant benefit of rehabilitation after myocardial infarction, we performed an overview of all randomized trials, involving 4,554 patients; we evaluated total and cardiovascular mortality, sudden death, and fatal and nonfatal reinfarction. For each endpoint, we calculated an odds ratio (OR) and 95% confidence interval (95% CI) for the trials combined. After an average of 3 years of follow-up, the ORs were significantly lower in the rehabilitation than in the comparison group: specifically, total mortality (OR = 0.80 [0.66, 0.96]), cardiovascular mortality (OR = 0.78 [0.63, 0.96]), and fatal reinfarction (OR = 0.75 [0.59, 0.95]). The OR for sudden death was significantly lower in the rehabilitation than in the comparison group at 1 year (OR = 0.63 [0.41, 0.97]). The data were compatible with a benefit at 2 (OR = 0.76 [0.54, 1.06]) and 3 years (OR = 0.92 [0.69, 1.23]), but these findings were not statistically significant. For nonfatal reinfarction, there were no significant differences between the two groups after 1 (OR = 1.09 [0.76, 1.57]), 2 (OR = 1.10 [0.82, 1.47]), or 3 years (OR = 1.09 [0.88, 1.34]) of follow-up. The observed 20% reduction in overall mortality reflects a decreased risk of cardiovascular mortality and fatal reinfarction throughout at least 3 years and a reduction in sudden death during the 1st year after infarction and possibly for 2-3 years. With respect to the independent effects of the physical exercise component of cardiac rehabilitation, the relatively small number of "exercise only" trials, combined with the possibility that they may have had a formal or informal nonexercise component precludes the possibility of reaching any definitive conclusion. To do so would require a randomized trial of sufficient size to distinguish between no effect and the most plausible effect based on the results of this overview.