Counterfactual estimation of efficacy against placebo for novel PrEP agents using external trial data: example of injectable cabotegravir and oral PrEP in women

Abstract

Introduction:Multiple antiretroviral agents have demonstrated efficacy for human immunodeficiency virus (HIV) pre-exposure prophylaxis (PrEP). As a result, clinical trials of novel agents have transitioned from placebo-to active-controlled designs; however, active-controlled trials do not provide an estimate of efficacy versus no use of PrEP. Counterfactual placebo comparisons using other data sources could be employed to provide this information. Methods: We compared the active-controlled study (HPTN 084) of injectable cabotegravir (CAB-LA) versus daily oral emtricitabine/tenofovir disoproxil fumarate (FTC/TDF) among women from seven countries in Africa to three external, contemporaneous randomized HIV prevention trials from which we constructed counterfactual placebo estimates. We used direct standardization via analysis weights to achieve the same distribution of person-years between the external study and HPTN 084, across strata predictive of HIV risk (country and selected risk covariates). We estimated prevention efficacy against a counterfactual placebo to provide information on the use of CAB-LA and FTC/TDF compared to no intervention. We compared the counterfactual placebo findings for FTC/TDF to previous placebo-controlled trials, adjusted for observed adherence to daily pills. Results: Distribution of age and baseline prevalence of gonorrhoea and chlamydia were similar among matched counterfactual placebo and observed HPTN 084 arms after standardization. Counterfactual estimates of CAB-LA versus placebo in all three settings showed a consistent risk reduction of 93%-94%, with lower bounds of the confidence intervals above 72%. Observed adherence (quantifiable tenofovir in plasma) in HPTN 084 was 54%-56%, and estimated efficacy of daily oral FTC/TDF against a counterfactual placebo was consistent with a predicted risk reduction of 39%-40% for this level of daily pill use. Conclusions: Counterfactual placebo rates of HIV acquisition derived from external trial data in similar locations and time can be used to support estimates of placebo-based efficacy of a novel HIV prevention agent. External trial data must be standardized to be representative of the clinical trial cohort testing the novel HIV prevention agent, accounting for confounders.

Safety and efficacy of long-acting injectable cabotegravir as preexposure prophylaxis to prevent HIV acquisition

“…This estimate should be interpreted keeping in mind the heterogeneity of study populations and that relatively few HIV infections were identified. Although ethical considerations prevented trials from using a placebo comparator, a modeling study estimated that the efficacy of CAB-LA versus placebo was 93-95% among women [38].…”

Fusing trial data for treatment comparisons: Single vs multi‐span bridging

“…For both (3) and ( 4) the sampling weights ŴSi are applied only to data from the trial conducted in the non-focal population (6) and are not applied to data from the target population (5). Note ( 5) and ( 6) are Hajek estimators corresponding to the identifiable forms of the parameters derived in Section 2.2.…”

“…With placebo counterfactual methods, a standardized estimate of the outcome of interest in the placebo arm of a trial is obtained for comparison with an active control in a different trial. [4][5][6][7][8] While this method can account for some differences in trial populations, covariate adjustments are limited to a few discrete covariates. Bayesian methods have also been proposed to estimate placebo counterfactuals but require strong assumptions from expert knowledge or prior data.…”

“…More recently, researchers have estimated placebo counterfactuals, where a counterfactual refers to what an individual's outcome would have been had they been assigned to a placebo rather than active treatment. With placebo counterfactual methods, a standardized estimate of the outcome of interest in the placebo arm of a trial is obtained for comparison with an active control in a different trial 4‐8 . While this method can account for some differences in trial populations, covariate adjustments are limited to a few discrete covariates.…”