Reconciling systematic reviews of exercise-based cardiac rehabilitation and secondary prevention programmes for coronary heart disease.

Findings of systematic reviews incorporating metaanalysis are by nature hypothesis generating and should be carefully and judiciously interpreted. This need is epitomized by two of the largest reviews of interventions to support risk factor reduction after a diagnosis of coronary heart disease (Table 1). These reviews both included randomized trials published in a similar timeframe (up to 2003 or 2004) that compared the effects of various risk factor reduction programmes for patients with confirmed coronary heart disease to usual care. In 2004, a Cochrane review found that supervised and unsupervised exercise-based cardiac rehabilitation provided to patients with different forms of coronary heart disease lowered all cause total mortality by 20% and cardiac mortality by 26% in 48 trials of 8940 patients. The reviewers concluded that there was ‘no difference in mortality effect between exercise-only cardiac rehabilitation and comprehensive cardiac rehabilitation, or by exercise dose or duration of followup’. Also, the exercise-based programmes did not reduce recurrent myocardial infarction and the need for repeat coronary revascularization to statistically significant levels. Around the same time, a second review of 63 trials involving 21,295 patients who had undergone secondary cardiac prevention programmes found reductions of total mortality of 15% and of acute myocardial infarction of 17%. However, in contrast to the previous review, the secondary prevention programmes significantly reduced the rate of recurrent heart attacks. Also, benefits did not differ between programmes ‘that incorporated education and counselling about coronary risk factors with a supervised exercise programme, those that included risk factor education or counselling but no exercise component, and those that consisted of only a structured exercise programme’. On first reading, these reviews appear to offer different conclusions on effects on cardiovascular morbidity and regarding the benefits of supervised exercise. Either could be cited selectively to support or refute the benefits of exercise programmes for morbidity or to support or refute the incorporation of supervised exercise into services. Why might these different conclusions have come about and how should they be interpreted? Firstly, it is unlikely that the different conclusions arise from differences in quality as both reviews were of high quality as defined by PRISMA criteria and included many of the same trials: 37 of the 48 (77%) exercise-based trials were included in the 63 (59%) secondary prevention trials. Rather, issues of statistical power can explain the differences noted in morbidity. The forest plot analyses showed that, although exercise-based programmes in the Cochrane review did not reach statistically significant headline ‘effect sizes’, there was a tendency towards reduced recurrent myocardial infarction and the need for repeat coronary revascularization. The larger review of secondary prevention programmes considered over twice the number of events in the analysis of recurrent myocardial infarction and found that secondary prevention programmes significantly reduced the rate of recurrent heart attacks. Hence, focusing only on the presence or absence of statistical