The impact of prevention-effective PrEP use on HIV incidence: a mathematical modelling study

Abstract

Introduction: Models that project the impact and cost-effectiveness of HIV pre-exposure prophylaxis (PrEP) must specify how PrEP use aligns with HIV exposure. We hypothesized that varying PrEP use according to individual-level partnership dynamics rather than prioritization to population subgroups based on average risk will result in larger incidence reductions and greater efficiency.

Methods: We used an individual-based network transmission model calibrated to HIV dynamics in Eswatini to simulate PrEP use among individuals ages 15-34 between 2022 and 2031 under two paradigms of PrEP delivery: "Risk Group" and "Partnership." In the "Risk Group" paradigm, we varied PrEP coverage by risk groups (low, medium and high) defined by average partnership frequency and concurrency. In the "Partnership" paradigm, all individuals are potentially eligible for PrEP, but we assumed use occurs only during partnerships and varied prioritization by partner HIV status (no prioritization to high prioritization with HIV-positive partners). We calculated person-time on PrEP and incidence relative to a no PrEP scenario and estimated efficiency as the person-years of PrEP needed to avert one additional infection (NNT).

Results: In the Risk Group paradigm, restricting PrEP to the high-risk group was the most efficient (NNT = 17), but the number of infections averted was limited by the small size of the high-risk group. Expanding PrEP use to all risk groups averted up to three times more infections but with lower efficiency (NNT = 202). PrEP use under the Partnership paradigm was 2-6 times more efficient (NNT = 33-102) than the Risk Group paradigm with all groups eligible for PrEP. A 33% reduction in incidence among 15- to 34-year-olds was achieved at 46% (95% CI: 39-52%) PrEP coverage in the Risk Group paradigm and 6% (95% CI: 5-7%) to 17% (95% CI: 14-20%) in the Partnership paradigm.

Conclusions: Modelling PrEP use based on risk groups resulted in a sharp trade-off between PrEP efficiency and impact, whereas PrEP use predicated on partnerships resulted in much higher efficiency for widespread PrEP availability. Model estimates of PrEP impact and cost-effectiveness in generalized epidemics are strongly influenced by assumptions about how PrEP use aligns with individual-level HIV exposure heterogeneity.

Keywords: Africa; HIV prevention; PrEP; adherence; cost-effectiveness; modelling.

Figure 1

Illustration of PrEP use assumptions according to two model paradigms. Four example individual trajectories over a ten‐year period are shown in each paradigm. In the Risk Group paradigm, PrEP coverage can vary according to behavioural risk group, but PrEP use within risk group is independent of partnership status. In the Partnership paradigm, PrEP use occurs only during partnerships and may be prioritized with HIV‐positive partners. In this example, PrEP coverage is twice as high during partnerships with HIV‐positive individuals compared to during partnerships with HIV‐negative individuals. The high‐risk group is not shown. PrEP coverage levels are systematically varied in the main analysis. Abbreviation: Pr(PrEP), PrEP coverage.

Figure 2

Relationship between additional person‐time on PrEP and percentage of infections averted relative to a no PrEP scenario, by PrEP paradigm. Person‐years on PrEP and percentage of infections averted are cumulative across a 10‐year period spanning 2022–2031. Shaded regions indicate 95% credible intervals. Abbreviation: Mult, multiplier on the probability that an individual with an HIV‐positive partner will use PrEP relative to the probability that an individual with only HIV‐negative partners will use PrEP.

Figure 3

Relationship between PrEP coverage among all HIV‐negative 15‐ to 34‐year‐olds and the relative risk of HIV infection compared to a no PrEP scenario, by PrEP paradigm. Relative risk is averaged across 2022–2031. Shaded regions indicate 95% credible intervals. Abbreviation: Mult, multiplier on the probability that an individual with an HIV‐positive partner will use PrEP relative to the probability than an individual with only HIV‐negative partners will use PrEP.