University of California, San Francisco Logo

University of California, San Francisco | About UCSF | Search UCSF | UCSF Medical Center

Home > UCSF Institute for Global Health - Literature Digest > Adherence
Adherence to Antiretroviral Therapy in Sub-Saharan Africa and North America: A Meta-Analysis
Global Health Sciences Literature Digest
Published September 16, 2006
Journal Article

Mills EJ, Nachega JB, Buchan I, Orbinski J, Attaran A, Singh S, Rachlis B, Wu P, Cooper C, Thabane L, Wilson K, Guyatt GH, Bangsberg DR. Adherence to Antiretroviral Therapy in Sub-Saharan Africa and North America: A Meta-Analysis. JAMA. 2006 Aug;296(6):679-90.


This is a systematic review comparing levels of adherence to antiretrovirals among patients in sub-Saharan Africa to those who are part of more established treatment programs in North America.


Fifty-eight prospective studies assessing adherence rates as a primary or secondary outcome among general population samples of HIV seropositive persons in North America and Africa were included: 31 from North America (28 full-text articles, three abstracts) and 27 from sub-Saharan Africa (nine full text articles, 18 abstracts). The following were excluded: studies not representative of the "general population" of HIV-positive individuals, meaning those that assessed only homosexuals, the homeless, children or drug users, those containing experimental adherence interventions, those presenting adherence as the mean of all doses taken.

Studies measured "adherence" differently. Patient self-report was used to assess adherence in 71% of the North American and 66% of the sub-Saharan studies. Other measures included pharmacy claims, medical records, MEMS-cap (pill-counting devices), and clinician and staff reports. Time periods for measurement varied greatly, from the previous 24 hours up to 365 days, and including the previous three days, four days, one week, two weeks, one month, two months and three months.

The definition of "being adherent" varied by study, based on an "adherence threshold," from 100% adherence in some studies to >80% in others. For North American studies, 15/31 (48%) considered 100% adherence as "adherent"; 9/31 (29%) studies ≥95% adherence; 4/31 (13%) studies ≥90% adherence; 3/31 (10%) studies ≥80% adherence. The number of studies from Africa using each adherence threshold was the following: 11/27 (41%) studies ≥100% adherence; 11/27 (41%) studies ≥95% adherence; 2/27 (7%) studies ≥90% adherence; 3/27 (11%) studies ≥80% adherence.

Search Strategy

Eleven electronic databases (Medline, EMBASE, Cochrane CENTRAL, AIDSLINE, AMED, CINAHL, TOXNET, Development and Reproductive Toxicology, Hazardous Substance Databank, PsycINFO, and Web of Science), major HIV conference abstract databases, lay publications and websites were searched. Individual clinical researchers, AIDS cohort trial groups, and treatment advocacy groups were contacted to inquire about unpublished studies. Searches were conducted from the inception of the particular database until April 18, 2006.


This review included studies of general (male and female) HIV-positive populations in North America and Africa. The median number of participants per study from North America was 219 (total=17,573 patients), and from Africa was 100 (total=12,116 patients).



Outcome Measures

This review provided pooled and comparative analyses of the proportion of persons who were "adherent" across studies, stratified by data source (abstract or full-text), continent (North America or Africa), and adherence threshold used (100% to >80%). Multivariable random effects logistic regression was used to determine independent predictors of adherence; predictor variables included continent, adherence threshold used, clinic setting, and whether treatment was free or paid.


Meta-analysis revealed significant heterogeneity across studies. The combined analysis of data from all studies revealed an adherence of 64% (95% CI, 59%-70%), meaning that 64% of persons were adherent, based on whatever definition (adherence threshold) was used in each study. Pooled estimates from North American studies showed 55% of persons studied were adherent (95% CI, 49%-62%). A pooled adherence estimate from Africa studies showed 77% of persons were adherent (95% CI, 68%-85%). Independent predictors of adherence included African study, adherence threshold used (100% and 95%), using more than one adherence measure, and using MEMS.


The authors concluded that favorable levels of adherence can be achieved in sub-Saharan Africa. They state that concerns about suboptimal adherence are not supported by the data, and should not be used as a reason to delay access to treatment. However, these higher levels of adherence may be due in part to African individuals being on less complicated regimens, and being early in treatment. The initial experience of dramatic clinical improvement without the long-term side effects could be related to better adherence. Thus, these findings may not persist as access becomes more widespread.

Quality Rating

Based on the QUOROM grading system for systematic reviews, this analysis was of high quality. The main limitation of this analysis was the quality of the studies. There was a very wide range in how adherence was measured, little validation of measures, and many studies from Africa did not provide complete data (they were from abstracts).

In Context - Reviewer Comment

This is the first meta-analysis of adherence data from sub-Saharan Africa, and comparing it to data from the US and Canada. There has been no evidence to indicate that adherence in resource-poor settings is significantly worse than in regions where antiretroviral therapy (ART) is widely available or healthcare infrastructure more established. Data from homeless HIV-infected persons in San Francisco indicate that even though adherence may be only 80%, clinical improvement can occur. This review and meta-analysis excluded the vast majority of those high-risk persons (such as MSM and IDU) who are on ART in North America. The population of heterosexual or mixed gender populations may or may not be more similar to African populations.

Programmatic Implications

The findings of this analysis argue against the concern that poor adherence in Africa would be a rationale for delaying the expansion of ART programs. However, this does not mean that evaluation of adherence, and interventions to improve and maintain high levels, should not be instituted in African settings. Both are still required. Ensuring reliable drug supply and distribution networks are also important in maintaining high ART adherence rates.