BACKGROUND: Heart rate variability (HRV) is reported as a surrogate index for clinical outcome in trials of secondary prevention strategies for coronary artery disease (CAD), but a standardized guide for interpreting HRV change is not established. DESIGN: We evaluated HRV change in trials with CAD patients who received conventional medications (beta-blockers, calcium channel blockers, angiotensin converting enzyme inhibitors), biobehavioral treatment (psychotropics, biofeedback, relaxation) or exercise training. METHODS: Medline, Pubmed, Psycinfo, the Cochrane database, and Embase were searched until July 2007, without language restriction. We identified 33 randomized controlled trials. Two reviewers independently abstracted all trials using a standardized form. A hierarchy of frequency and time domain HRV indices defined outcome. RESULTS: A random-effects model yielded an overall pooled standardized mean difference (SMD) between treatment and control groups of moderate magnitude across treatment classes, based on a composite of time and frequency domain indices (SMD=0.40, P<0.0001), or only time or frequency indices (SMD=0.37 and 0.43, respectively, both P<0.0001). This change was equivalent to an increase in standard deviation of all normal-to-normal RR intervals of 9.0 ms (95% Confidence Interval, CI, 7.3, 10.7 ms) or a relative increase of 15.9% (95% CI, 13.2, 18.6%). To detect HRV change of this magnitude, a hypothetical trial would require a sample size of 660 patients for conventional medications or 1232 patients for all treatment classes. CONCLUSION: Pharmacologic, biobehavioral and exercise strategies for secondary prevention of CAD significantly increase HRV. This review provides a framework to assist efforts to evaluate the contribution of HRV change to CAD prognosis.