Fig. 2: Summary of clinical trials evaluating FTC/TDF-based PrEP in cis women.

a–e, The total observation time in person-years (indicated by clock) in the respective trials (HPTN 084 (a), TDF2 (b), Parters-PrEP (c), VOICE (d) and Fem-PrEP (e)) dichotomized into ‘detected’ and ‘undetected’ based on the fraction of measurements with detectable plasma TFV. Likewise, the total number of infections (red ribbon) were proportionally assigned based on the fraction of infected individuals who had detectable plasma TFV. We then stochastically simulated the ‘no-drug intervention’ (crossed-out pill, details in Methods) to compute the number of infections in the ‘no-drug intervention’ arm. From both the total observed number of infections and the simulated number of infections in the ‘no-drug intervention’ arm, we could compute the number of infected individuals with detectable plasma TFV. f, Incidence in placebo (blue error bars) and ‘drug-undetected’ subcohort of the PrEP intervention arms (gray error bars) with sample sizes indicated in a–e. Error bars show the mean incidence rates and the 95% confidence intervals (CIs) computed using Wilson’s method. Purple error bars show incidences (and their 95% CI) computed from 10,000 stochastic clinical trial simulations with sampled incidence rate parameters (Methods), depicting uncertainty in the simulation parameter, as well as intrinsic randomness in trial outcomes due to rare events. g, Violin plots indicating the probability distributions of average PrEP efficacy from clinical trial simulation only taking data provided in the respective studies (Supplementary Fig. 8). The width indicates the likelihood of a particular efficacy. Square-shaped violin plots indicate uninformative clinical trials (FEM-PrEP, VOICE study), that is, no conclusion can be drawn with regard to the PrEP efficacy, whereas sharply concentrated distributions (HPTN 084 and Partners-PrEP studies) are informative with regard to PrEP efficacy. h, Relative odds of particular efficacy ranges (related to g).