Pharmacy refill data can be used to predict virologic failure for patients on antiretroviral therapy in Brazil

Introduction: Pharmacy adherence measures such as pharmacy dispensing ratios (PDRs) have previously been shown to be predictive of virologic outcomes. We aimed to determine the optimal interval of PDR assessment for predicting virologic failure for HIV-infected patients on antiretroviral therapy (ART). Methods: Using national Brazilian ART pharmacy refill data, we examined PDRs for patients ≥18 years of age with at least one HIV RNA level ≥180 days after ART initiation on or after 1 January 2011. Patients with a documented ART change ≤270 days prior to viral load test date were excluded. Logistic regression models were used to describe associations between virologic failure, defined as an HIV RNA level ≥400 copies/mL and PDRs, defined as the number of days index drug dispensed (non-nucleoside reverse-transcriptase inhibitor or protease inhibitor) per 180- and 90-day, interval preceding viral load testing, adjusting for sex, age, race, time since ART initiation and index drug. Backward elimination of insignificant variables was performed after adjusting for PDR. A predictive probability of virologic failure was calculated using the corresponding odds ratios for the PDR and any other significant variables. The diagnostic performance of the PDR interval was assessed by calculating the area under the receiver operating characteristic curve (AUROC) for the predictive probability with respect to virologic failure. Results and Discussion: A total of 1,025 patients were included (68% were male, median age 40 years, median time on ART 3.4 years). The PDR was found to be significantly associated with virologic failure for all of the PDR intervals (p < 0.001). There was an increased risk of virologic failure for all PDRs <0.95. The 90–180 days interval had a AUROC of 0.842, compared to 0.841 and 0.829 for the 0–180 days and 0–90 days intervals, respectively. The PDR performed well as a predictive tool to identify patients in virologic failure with the 90–180-days interval prior to viral load testing being marginally more predictive. Conclusions: The validation and use of the pharmacy dispensing ratio using public pharmacy refill data could aid in early identification of patients with poor adherence and prevent development of treatment failure and drug resistance in Brazil.