Routine Antiretroviral Pharmacy Refill Information Can Predict Failure Postpartum in Previously Suppressed South African Women With HIV.

Background: Detection of antiretrovirals (ARVs) in biological specimens is a reliable, objective way to measure adherence. However, routine ARV testing is not feasible in many high-burden settings. This study explored if pharmacy data could accurately predict HIV viremia postpartum in previously virally suppressed women. Methods: South African women with HIV who initiated antiretroviral therapy (ART) during pregnancy and achieved viral suppression (VS; viral load [VL]≤50 copies/mL) were followed postpartum; during follow-up, plasma VL was measured and ARV adherence self-reported. A portion of samples were tested for the presence of ARV using mass spectrometry. Patient-level routine pharmacy data were used to classify if women should have the drug in hand for the past 7 days before the visit date. Logistic regressions were used to calculate associations between adherence and viral nonsuppression (VNS; VL > 50) or failure (VF; VL > 1000) at the first study visit of women who had ARV measured. Data for all women were examined for associations of self-reported adherence and drug in hand with VS and VF at 2, 6, and 12 months postpartum. Results: Women with no ARV detected were significantly more likely to have VNS (odds ratio [OR], 26.4). Having no drug in hand for 7 days was also predictive of VNS in these same women (OR, 7.0) and the full cohort (n = 572) at 3 (OR, 2.9), 6 (OR, 8.7), and 12 months (OR, 14.5). Similar results were seen for VF. Conclusions: These data show that routine pharmacy data can act as a highly predictive mechanism for identifying patients at risk of VNS and VF due to nonadherence.