Meta-analysis of trials of interventions to improve medication adherence.

The effect of tools and methods designed to enhance medication adherence that have been evaluated in randomized controlled trials was studied. A literature search was performed with MEDLINE, International Pharmaceutical Abstracts, PsychLIT, ERIC, and EMBASE for the period from 1966 to December 2000. Only randomized, controlled trials with at least 10 subjects per intervention group were included. Of 484 articles evaluated, only 61 met the criteria for the meta-analysis. Multiple interventions or study samples were identified in 23 of the articles. Each intervention was counted as a separate study, yielding 95 cohorts totaling 18,922 subjects. Of these subjects, 9,604 (51%) received interventions and 9,318 served as controls. Cohorts reported between 1990 and 1999 accounted for 53% of the sample; 56% of all cohorts were based in physician offices and 26% involved hypertensive patients. Behavioral interventions accounted for 41 cohorts (8,885 subjects), educational interventions for 22 cohorts (6,392 subjects), and combined interventions for 32 cohorts (3,645 subjects). Homogeneity of groupings and effect sizes (ESs) were calculated for each type of intervention. Overall, the data were not homogeneous, so conclusions could not be derived from the entire body of data. The educational intervention and combined intervention cohorts were nonhomogeneous (p < 0.001 and p < 0.01, respectively); however, the behavioral intervention cohort was homogeneous (Q = 42.48, d.f. = 40, p = 0.36). The overall ES for behavioral interventions was 0.07 (95% confidence interval [CI] = 0.04-0.09). There were no significant differences among the behavioral interventions. Educational interventions had an overall ES of 0.11 (95% CI = 0.06-0.15); there were no significant differences among the educational interventions. The overall ES of the combined interventions was 0.08 (95% CI = 0.04-0.12). When stratifying the combined intervention group by type of behavioral intervention, mail reminders had the largest impact (ES = 0.38). Meta-analysis of studies of interventions to improve medication adherence revealed an increase in adherence of 4-11%. No single strategy appeared to be best.