HIV recent infection test-based incidence as a counter-factual for new PrEP trials


BACKGROUND: Clinical trials of new PrEP agents are challenging because it is not ethical to include a placebo-only group. Innovative ways to evaluate new PrEP modalities are needed without impractically large sample sizes (SS) required for non-inferiority trials. HIV recent infection testing algorithms (RITAs) such as the limiting antigen avidity assay (LAg) plus viral load (VL) could be used to derive a 'counter-factual' incidence estimate (CFIE) using specimens from untreated, HIV-positive people identified during screening, to which on-PrEP incidence can be compared. The feasibility of this approach is partly dependent on the SS needed to ensure adequate power, which is impacted by RITA performance, the number of recent infections identified, the expected efficacy of the intervention, and other factors.
METHODS: SS (number of persons screened) required to support detection of an 80% reduction in incidence (null hypothesis: 50% reduction) were calculated based on a test statistic of log incidence ratio ( in different populations, and assuming: 4th generation Ab/Ag testing to identify HIV-positives, 90% enrollment, 90% recency testing success, two years of follow-up on PrEP, significance level 0.05 and power 0.8. Subtype-specific mean durations of recent infection and false recent ratios (FRR) for the LAg + VL RITA were derived from pooled calibration data.
RESULTS: Required SS for three key populations were modeled: women aged 14-17 years or >18 years in South Africa (subtype C), and men who have sex with men in the USA (subtype B). SS for these three populations were 2882, 5463, and 2327, respectively. These SS are comparable to the number of participants in recent phase 3 PrEP trials.
CONCLUSIONS: CFIEs based on recent infection testing can facilitate next-generation PrEP trials, at least in high incidence populations for which RITAs have been calibrated, and where the efficacy of the intervention is expected to be very high. SS may not be feasible in populations with lower incidence, where the FRR is higher (e.g. subtype D), or if PrEP efficacy is expected to be lower. Despite these limitations, generation of a CFIE based on recency assays appears to be feasible, offers high statistical power, and is nearly contemporaneous with the on-PrEP population.