OBJECTIVE: Compliance with chronic medications is poor, resulting in lost clinical benefits and wasted resources. Developing evidence-based compliance programs is difficult because the literature is often equivocal, contradictory, or of poor methodological quality. The objectives of this study were 1) to assess the relationship between methodological rigor in program evaluation and probability of intervention success; and 2) to identify the characteristics of compliance interventions that most strongly predict positive outcomes. METHODS: We searched the MEDLINE database for controlled studies of interventions designed to improve patient compliance with prescription medications for asthma, dyslipidemia, hypertension, depression, and diabetes (English-only, published between 1982 and 2002). Each study was abstracted for methodological rigor and intervention design. Compliance programs were considered "successful" if the authors demonstrated a statistically significant improvement in compliance compared to a control group. RESULTS: Sixty-four controlled studies were identified, in which 76 compliance interventions were evaluated. Overall, 68% of the interventions improved compliance. Of 20 interventions evaluated based on self-reported compliance, 9 (45%) were successful; 41 of 54 interventions (76%) that measured compliance via pill count, electronic monitoring, or pharmacy records were successful (p=0.02). There were nine broad categories of intervention types, the most frequent of which were improved dosing convenience (n=20 interventions, 70% successful), intensive clinician management (n=7, 57% successful), counseling (n=7, 57% successful), and reminders (n=6, 66% successful). In addition, 20 of 25 programs (80%) consisting of multiple interventions were successful. CONCLUSION: Reliance on self-reported medication use appears to bias results against the intervention. This may be because patients report exaggerated levels of compliance in both intervention and control groups, creating a "ceiling effect" that reduces the observable difference. Common attributes of successful programs included simplified treatment regimens, facilitation of provider-patient relationships, and multi-faceted patient education methods. Combination interventions had higher rates of success than individual tactics.