Cost-effectiveness of tenofovir gel in urban South Africa: model projections of HIV impact and threshold product prices

Abstract

Background: There is urgent need for effective HIV prevention methods that women can initiate. The CAPRISA 004 trial showed that a tenofovir-based vaginal microbicide had significant impact on HIV incidence among women. This study uses the trial findings to estimate the population-level impact of the gel on HIV and HSV-2 transmission, and price thresholds at which widespread product introduction would be as cost-effective as male circumcision in urban South Africa.

Methods: The estimated 'per sex-act' HIV and HSV-2 efficacies were imputed from CAPRISA 004. A dynamic HIV/STI transmission model, parameterised and fitted to Gauteng (HIV prevalence of 16.9% in 2008), South Africa, was used to estimate the impact of gel use over 15 years. Uptake was assumed to increase linearly to 30% over 10 years, with gel use in 72% of sex-acts. Full economic programme and averted HIV treatment costs were modelled. Cost per DALY averted is estimated and a microbicide price that equalises its cost-effectiveness to that of male circumcision is estimated.

Results: Using plausible assumptions about product introduction, we predict that tenofovir gel use could lead to a 12.5% and 4.9% reduction in HIV and HSV-2 incidence respectively, by year 15. Microbicide introduction is predicted to be highly cost-effective (under $300 per DALY averted), though the dose price would need to be just $0.12 to be equally cost-effective as male circumcision. A single dose or highly effective (83% HIV efficacy per sex-act) regimen would allow for more realistic threshold prices ($0.25 and $0.33 per dose, respectively).

Conclusions: These findings show that an effective coitally-dependent microbicide could reduce HIV incidence by 12.5% in this setting, if current condom use is maintained. For microbicides to be in the range of the most cost-effective HIV prevention interventions, product costs will need to decrease substantially.

Figure 1

Model-projected HIV prevalence and incidence trends over time and data used to fit model*. a: HIV prevalence trends among general population males and females in 2002 and 2005. b: HIV prevalence trends among the general population and among female sex workers in 2008. c: HIV incidence trends among general population males and females. *The points with error bars for males and females in 2002 and 2005 represent the mean and 95% confidence intervals derived from data [12,26]. The dotted line represents the model-projected female HIV prevalence while the solid line represents the model-projected male HIV prevalence. The shaded areas represent the range of prevalence estimates spanned by all model fits.

Figure 2

Projected population-level HIV impact in Gauteng for different assumptions about gel efficacy, uptake and use, and its influence on condom use**. **Main shaded bars show best-fit model projections and error bars indicate range spanned by 95% of all model fits (2.5% to 97.5% percentile range or 95% credibility interval), with the dark bars illustrating the main scenario (54% HIV efficacy and 71% HSV-2 efficacy, gel uptake reaching 30% by year 10, gel used in 72% of sex-acts, and no reduction in condom use).