Accuracy of measures for antiretroviral adherence in people living with HIV

BACKGROUND: Good patient adherence to antiretroviral (ART) medication determines effective HIV viral suppression, and thus reduces the risk of progression and transmission of HIV. With accurate methods to monitor treatment adherence, we could use simple triage to target adherence support interventions that could help in the community or at health centres in resource‐limited settings. OBJECTIVES: To determine the accuracy of simple measures of ART adherence (including patient self‐report, tablet counts, pharmacy records, electronic monitoring, or composite methods) for detecting non‐suppressed viral load in people living with HIV and receiving ART treatment. SEARCH METHODS: The Cochrane Infectious Diseases Group Information Specialists searched CENTRAL, MEDLINE, Embase, LILACS, CINAHL, African‐Wide information, and Web of Science up to 22 April 2021. They also searched the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) and ClinicalTrials.gov for ongoing studies. No restrictions were placed on the language or date of publication when searching the electronic databases. SELECTION CRITERIA: We included studies of all designs that evaluated a simple measure of adherence (index test) such as self‐report, tablet counts, pharmacy records or secondary database analysis, or both, electronic monitoring or composite measures of any of those tests, in people living with HIV and receiving ART treatment. We used a viral load assay with a limit of detection ranging from 10 copies/mL to 400 copies/mL as the reference standard. We created 2 × 2 tables to calculate sensitivity and specificity. DATA COLLECTION AND ANALYSIS: We screened studies, extracted data, and assessed risk of bias using QUADAS‐2 independently and in duplicate. We assessed the certainty of evidence using the GRADE method. The results of estimated sensitivity and specificity were presented using paired forest plots and tabulated summaries. We encountered a high level of variation among studies which precluded a meaningful meta‐analysis or comparison of adherence measures. We explored heterogeneity using pre‐defined subgroup analysis. MAIN RESULTS: We included 51 studies involving children and adults with HIV, mostly living in low‐ and middle‐income settings, conducted between 2003 and 2021. Several studies assessed more than one index test, and the most common measure of adherence to ART was self‐report. ‐ Self‐report questionnaires (25 studies, 9211 participants; very low‐certainty): sensitivity ranged from 10% to 85% and specificity ranged from 10% to 99%. ‐ Self‐report using a visual analogue scale (VAS) (11 studies, 4235 participants; very low‐certainty): sensitivity ranged from 0% to 58% and specificity ranged from 55% to 100%. ‐ Tablet counts (12 studies, 3466 participants; very low‐certainty): sensitivity ranged from 0% to 100% and specificity ranged from 5% to 99%. ‐ Electronic monitoring devices (3 studies, 186 participants; very low‐certainty): sensitivity ranged from 60% to 88% and the specificity ranged from 27% to 67%. ‐ Pharmacy records or secondary databases (6 studies, 2254 participants; very low‐certainty): sensitivity ranged from 17% to 88% and the specificity ranged from 9% to 95%. ‐ Composite measures (9 studies, 1513 participants; very low‐certainty): sensitivity ranged from 10% to 100% and specificity ranged from 49% to 100%. Across all included studies, the ability of adherence measures to detect viral non‐suppression showed a large variation in both sensitivity and specificity that could not be explained by subgroup analysis. We assessed the overall certainty of the evidence as very low due to risk of bias, indirectness, inconsistency, and imprecision. The risk of bias and the applicability concerns for patient selection, index test, and reference standard domains were generally low or unclear due to unclear reporting. The main methodological issues identified were related to flow and timing due to high numbers of missing data. For all index tests, we assessed the certainty of the evidence as very low due to limitations in the design and conduct of the studies, applicability concerns and inconsistency of results. AUTHORS' CONCLUSIONS: We encountered high variability for all index tests, and the overall certainty of evidence in all areas was very low. No measure consistently offered either a sufficiently high sensitivity or specificity to detect viral non‐suppression. These concerns limit their value in triaging patients for viral load monitoring or enhanced adherence support interventions.