Adherence to antiretroviral therapy is a powerful predictor of survival for individuals living with human immunodeficiency virus (HIV) and AIDS. Concerns about incomplete adherence among patients living in poverty have been an important consideration in expanding the access to antiretroviral therapy in sub-Saharan Africa. To evaluate estimates of antiretroviral therapy adherence in sub-Saharan Africa and North America. Eleven electronic databases were searched along with major conference abstract databases (inclusion dates: inception of database up until April 18, 2006) for all English-language articles and abstracts; and researchers and treatment advocacy groups were contacted. Study Selection and Data Abstraction To best reflect the general population, studies of mixed populations in both North America and Africa were selected. Studies evaluating specific populations such as men only, homeless individuals, or drug users, were excluded. The data were abstracted in duplicate on study adherence outcomes, thresholds used to determine adherence, and characteristics of the populations. A random-effects meta-analysis was performed in which heterogeneity was examined using multivariable random-effects logistic regression. A sensitivity analysis was performed using Bayesian methods. Thirty-one studies from North America (28 full-text articles and 3 abstracts) and 27 studies (9 full-text articles and 18 abstracts) from sub-Saharan Africa were included. African studies represented 12 sub-Saharan countries. Of the North American studies, 71% used patient self-report to assess adherence; this was true of 66% of the African assessments. Studies reported similar thresholds for adherence monitoring (eg, 100%, >95%, >90%, >80%). A pooled analysis of the North American studies (17,573 patients total) indicated a pooled estimate of 55% (95% confidence interval, 49%-62%; I2, 98.6%) of the populations achieving adequate levels of adherence. Our pooled analysis of African studies (12,116 patients total) indicated a pooled estimate of 77% (95% confidence interval, 68%-85%; I2, 98.4%). Study continent, adherence thresholds, and study quality were significant predictors of heterogeneity. Bayesian analysis was used as an alternative statistical method for combining adherence rates and provided similar findings. Our findings indicate that favorable levels of adherence, much of which was assessed via patient self-report, can be achieved in sub-Saharan African settings and that adherence remains a concern in North America.