Data for on-demand pre-exposure prophylaxis (PrEP) are scarce. We implemented a cohort study to assess its efficacy, safety, and effect on sexual behaviour.
We invited men and transgender women who have sex with men, previously enrolled in the randomised placebo-controlled ANRS IPERGAY trial at seven sites (six in France and one in Canada), to participate in an open-label extension with on-demand tenofovir disoproxil fumarate (300 mg) and emtricitabine (200 mg) to be taken before and after sexual intercourse. We assessed the incidence of HIV and other sexually transmitted infections (STIs), PrEP adherence, safety, and sexual behaviour. Statistical analyses included comparisons of proportions and incidence between the randomised phase of the ANRS IPERGAY trial and the open-label phase, and all participants were included in safety analyses. ANRS IPERGAY is registered with ClinicalTrials.gov, number NCT01473472.
Between Nov 4, 2014, and Jan 27, 2015, we enrolled 361 participants. Median follow-up was 18·4 months (IQR 17·7–19·1). One participant who discontinued PrEP acquired HIV infection. HIV incidence was 0·19 per 100 person-years (95% CI 0·01–1·08), compared with 6·60 per 100 person-years (3·60–11·05) in the placebo group of the randomised study, indicating a relative reduction of 97% (95% CI 81–100) in the incidence of HIV with on-demand PrEP. Participants used a median of 18 pills of study drugs per month (IQR 11–25), and at the 6 month visit 240 (71%) of 336 participants had tenofovir detected in plasma. Drug-related gastrointestinal events were reported in 49 participants (14%) but were self-limited. Only four participants (1%) discontinued PrEP, three because of an increase in plasma creatinine. The proportion of participants reporting condomless sex at their last receptive anal intercourse significantly increased from 77% (136 of 176 participants) at baseline to 86% (66 of 77 participants) at 18 months' follow-up (p for trend=0·0004). The incidence of a first bacterial STI during this open-label phase did not change significantly compared with the randomised phase (59·0 vs 49·1 per 100 person-years, respectively; p=0·11).