Assessing risk for HIV infection among adolescent girls in South Africa: an evaluation of the VOICE risk score (HPTN 068)

INTRODUCTION: To maximize impact and minimize costs, antiretroviral pre‐exposure prophylaxis (PrEP) interventions should be offered to those at highest risk for HIV infection. The risk score derived from the VOICE trial is one tool currently being utilized to determine eligibility in adolescent PrEP trials in sub‐Saharan Africa. This study is aimed at evaluating the utility of the risk score in predicting HIV incidence among a cohort of adolescent girls in rural South Africa. METHODS: We utilized data from HIV Prevention Trials Network (HPTN) 068, a phase III randomized controlled trial conducted in rural Mpumalanga province, South Africa. School‐attending young women aged 13 to 20 years were enrolled into the trial from 2011 to 2012 and followed for up to three years. A risk score based on individual‐level risk factors measured at enrolment was calculated for HPTN 068 participants who completed a one‐year follow‐up visit and were HIV seronegative at enrolment. Possible scores ranged from 0 to 10. A proportional hazards model was then used to determine if risk score at enrolment was predictive of incident HIV infection at follow‐up and an area under the curve analysis was used to examine the predictive ability of the score. RESULTS AND DISCUSSION: The risk score had limited variability in the HPTN 068 sample. Scores ≥5 identified 85% of incident infections from 94% of the sample, compared to the VOICE sample in which scores ≥5 identified 91% of incident infections from only 64% of participants. The risk score did not predict HIV incidence after one year of follow‐up (hazard ratio = 1.029; 95% confidence interval (CI): 0.704, 1.503, p = .884) and showed poor predictive ability (area under the curve = 0.55; 95% CI: 0.44, 0.65). Certain individual risk factors that comprise the risk score may be context specific or not relevant for adolescent populations. Additional factors should be considered when assessing risk for the purposes of determining PrEP eligibility. CONCLUSIONS: The VOICE risk score demonstrated low utility to predict HIV incidence in the HPTN 068 sample. Findings highlight the need for an age and developmentally appropriate tool for assessing risk for HIV infection among adolescents. Use of the VOICE risk score for determining PrEP eligibility in younger populations should be carefully considered.