OBJECTIVES The objective of this systematic review was to evaluate the impact of interventions to improve medication compliance in patients on antihypertensive medications. METHODS CINAHL, EMBASE, all EBM Reviews and MEDLINE electronic databases were searched for articles published January 1979 through February 2007 using MeSH keywords relevant to patient compliance and intervention type, yielding 405 citations. Two reviewers independently examined each citation and abstract. We conducted three levels of review resulting in 94 papers for full review. Data were analyzed using qualitative and quantitative methods, including a meta-analysis using a random effects model (REM) and Mantel-Haenszel (MH) summary estimate. RESULTS The most common research design involved randomized controlled trials (RCTs) (60/94, 64%). Seventy-two (77%) studies involved interventions that sought to directly improve medication compliance, 12 (13%) had to do with interventions to improve compliance through a multidisciplinary process, 5 (5%) involved interventions to indirectly improve compliance through changing physician practice, and 5 (5%) were other types. Thirteen studies (14%) involved the use of a theoretical framework to guide the research. A meta-analysis was conducted of RCTs of educational interventions, (N=6). The REM showed a trend toward statistical significance with a risk ratio of 0.83 for non-compliance (95% CI: 0.68 -1.00, p=0.05). The MH summary estimate was statistically significant with a risk ratio of 0.83 for non-compliance (95% CI: 0 .75 - 0.92, p=0.00), favoring patients who received educational interventions to improve compliance. However, the X2 test for heterogeneity was significant; p=0.020 and p=0.016 respectively. CONCLUSIONS Our analysis indicated that most interventions focused on directly improving medication compliance. The meta-analysis illustrated educational strategies provide a significant benefit in reducing non-compliance in patients using antihypertensives. The small sample size may have contributed to the observed heterogeneity and require additional investigation. Our findings have implications for designing future research and implementing educational interventions.