Reconciling counterfactual placebo HIV incidence rate estimates for an active-controlled pre-exposure prophylaxis trial, PrEPVacc

Background Estimating counterfactual placebo HIV incidence rates (the rate in a placebo arm if it had been included) is critically important for the interpretation of active-controlled HIV pre-exposure prophylaxis (PrEP) trials. The PrEPVacc trial (NCT04066881) enabled this estimation using three approaches. In this paper, we contrast the resulting estimates and explore the potential biases of each. Methods The PrEPVacc HIV prevention trial conducted in Uganda, Tanzania and South Africa included a non-inferiority comparison of two daily oral PrEP regimens—Emtricitabine/tenofovir (TDF/FTC) and Tenofovir alafenamide/Emtricitabine (TAF/FTC) administered for a target duration of 26 weeks (PrEP component). The trial also included a concurrent placebo-controlled, three-arm HIV vaccine trial. The analyses described in this paper were aimed at estimating counterfactual placebo HIV incidence for the active-controlled PrEP component of the trial. Estimates generated using multivariable Poisson regression models and data from a pre-trial registration cohort have been published previously. In this paper we generate estimates using data from the post-PrEP component of the trial and the adherence-efficacy relationship of TDF/FTC and comprehensively review the potential biases of each approach. Results Participants in the PrEP component of the PrEPVacc trial were predominantly female (87%), with 53% aged ≥ 25 years, 60% identifying as sex workers, and 26% diagnosed with a sexually transmitted infection at baseline. Point estimates from the different approaches ranged from 0.7 to 1.9/100 person-years. All three counterfactual placebo incidence estimation approaches had considerable biases, including uncertainties around calendar trends in the pre-trial registration cohort and whether vaccine effects observed in the post-PrEP period of the PrEPVacc trial applied similarly during the PrEP component. The adherence-efficacy approach was particularly imprecise and unreliable due to the small number of HIV infections (two) observed in the TDF/FTC arm and the imprecision of the risk reduction estimates attributed to the observed level of adherence (Tenofovir Diphosphate levels). Conclusion Obtaining a precise estimate of counterfactual placebo HIV incidence in the PrEPVacc trial was challenging. Further development and standardisation of counterfactual placebo HIV incidence estimation approaches would simplify analyses of future HIV pre-exposure prophylaxis trials.