Estimated long-acting PrEP effectiveness in the HPTN 084 cohort using a model-based HIV incidence in the absence of PrEP


BACKGROUND: HPTN 084 is a randomized double-blind controlled-superiority study assessing the safety and efficacy of long-acting injectable cabotegravir for pre-exposure prophylaxis (CAB-LA) for preventing HIV in African women aged 18-45 years. Daily oral tenofovir disoproxil fumarate and emtricitabine (TDF/FTC) was an active comparator; there was no placebo control. We estimate the incidence in a hypothetical placebo control arm and project the effectiveness of CAB-LA compared to placebo.
METHODS: Our model-based counterfactual predicts HIV risk in cohorts of sub-Saharan Africa (SSA) women based on individual VOICE risk scores and HIV incidence, prevalence, and viral load suppression among adult males in the communities of each trial site. HIV risk is used to predict cumulative HIV incidence over one year of follow-up. This model was calibrated to data from the VOICE trial and previously validated by comparing predicted HIV incidence to that observed in HPTN 035, FEM-PrEP, ASPIRE, and ECHO.

RESULTS: Overall, we project a counterfactual placebo incidence of 2.2% (95% cred. int. 1.7% - 2.8%) in the HPTN 084 study cohort compared to 0.2% (95% conf. int. 0.06-0.52%) in the CAB-LA arm and 1.86% (95% conf. int. 1.3-2.57) in the TDF/FTC arm, suggesting an effectiveness against HIV acquisition of 91% (95% cred. int. 76%-97%) and 15% (95% cred. int. -26%-44%), respectively, compared to placebo.
CONCLUSIONS: For ethical reasons effectiveness of new HIV prevention products, such as CAB-LA, must be compared to an approved PrEP product such as TDF/FTC, whose effectiveness depends on adherence. Counterfactual estimates of incidence allow for comparison of such products against a hypothetical placebo control. This model-based counterfactual, using contemporary epidemic data and participant risk factors, provides additional assurance that CAB-LA reduced HIV acquisition risk by 90% among women in SSA. This effectiveness estimate can be further refined and validated with additional counterfactuals using other data and methodologies.