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

There is an urgent need for effective HIV prevention methods that can be initiated by women. After a decade of little progress in HIV prevention interventions, recent trials have shown that antiretrovirals can be used to prevent HIV incidence and infectivity [1–4]. In 2010, the CAPRISA 004 trial in Kwazulu Natal, South Africa, showed that the incidence of HIV infection among women randomised to using a topical gel containing 1% of the antiretroviral tenofovir before and after sex, was reduced by 39% (95% CI 6-60%). In addition, women’s risk of acquiring herpes simplex virus type-2 (HSV-2) was reduced by 51% (95% CI 22-70%) [5, 6]. This finding was groundbreaking, being the first clinical trial to show a statistically significant impact of a microbicide product on HIV incidence, and providing a proof of concept that a topically applied antiretroviral microbicide can reduce women’s risk of HIV acquisition [7]. The FACTS 001 trial is now seeking to confirm the CAPRISA results, and assess the generalizability of the findings across different settings [8]. Plans are underway to prepare for the possible introduction and manufacture of tenofovir gel. Once this trial is completed, its data will be combined with the CAPRISA 004 results and presented together for regulatory approval. This study aims to inform this process and the field of HIV more broadly.

Evidence from FACTS 001 and this cost-effectiveness analysis will be evaluated in the context of VOICE trial results that showed that daily administration of the same product and oral tenofovir did not achieve a significant reduction in HIV incidence due to adherence issues [9]. It appears that compliance to a daily regimen of gels or pills is not a viable option for African women in particular for younger women. However, given the positive findings from CAPRISA, there is still hope that a coitally-specific product may provide a feasible strategy for women who are unable to protect themselves with the current range of options available.

In parallel to clinical research, it is important to explore the potential population-level impact that tenofovir gel introduction could have on HIV incidence in different epidemic settings, and identify key factors that may influence the impact and cost-effectiveness of future product introduction. Towards this end, this paper uses modelling and the CAPRISA trial results to estimate the gel’s likely level of protection against HIV and HSV-2 during each sex-act (here termed the ‘imputed efficacy’), and the potential population-level impact on HIV and HSV-2 of gradual product introduction in an urban setting with high HIV prevalence and incidence in South Africa. Using South Africa specific unit cost data, the full cost of introducing microbicides into the public health sector is modelled first. Secondly, threshold product prices are identified at which point the costs per HIV infection averted are equal to that of voluntary medical male circumcision (VMMC) (~$1000 per infection averted) [10], an intervention widely accepted as cost-effective and efficient. A range of scenarios are used to explore the effects of various assumptions on this threshold.

We have used detailed mathematical and economic modelling to estimate the impact and break-even price of tenofovir gel as introduced and distributed in urban South Africa. Our findings suggest that, for the main scenario considered, gel introduction could be an important addition to existing HIV prevention efforts, leading to a 12.5% (95%CrI 11.5-12.7%) relative reduction in HIV incidence and a 4.9% (95%CrI 4.7-5.5%) relative reduction in HSV-2 incidence over 15 years. The impact on HSV-2 is less than the impact on HIV, despite the higher efficacy against HSV-2, due to higher risk of HSV-2-infection for women.

Our findings suggest that the introduction of an effective gel could be cost saving at reasonable gel prices, in particular as ART access is expanded, and highly cost-effective even at current gel prices.

Indeed, using plausible assumptions about the full costs of programme delivery, we predict that the introduction of an HIV efficacious gel would be highly cost-effective ($297 per DALY averted) at a price of $0.25, which is within the cost range ($0.17-$0.27) at large scale production of 100 million applicators [36]. In our model this scale of distribution is achieved from year 9 onwards. If South Africa reaches its target of 80% on ART or a single dose regimen is found to be effective, we can expect microbicides to be around the same cost per HIV infection averted as VMMC.

Even at current FACTS 001 trial production prices ($1.60) [36], the incremental cost per DALY averted would be around $2,660, which is well below the commonly applied cost-effectiveness threshold of 1*GDP ($7,610 for South Africa) [42].

Using the trial estimates of 72% adherence, we impute that the HIV and HSV-2 per sex-act efficacies of tenofovir gel are 54% (95% CrI 8-83%) and 71% (95%CrI 30-97%), respectively. Although these estimates are based upon the trial measures of gel adherence reporting, we considered that these are reasonably reliable, as gel applicator returns were found to correlate well with vaginal tenofovir concentration measured in a pharmacokinetics sub-study [43]. We are also reassured about the validity of our imputed efficacy estimates because a secondary analyses of the effectiveness among ‘high adherers’ in the trial (i.e. women who used both doses of the gel in over 80% of sex-acts) found that the risk of HIV acquisition in this population was reduced by 54% (95%CI 4-80%) [5]. However, the wide bounds on the imputed efficacy estimates of tenofovir gel, due to the wide confidence intervals on the trial effectiveness estimates, lead to a wide range of population-level impact estimates predicted by the model.

This analysis builds upon our previous modelling of the potential impact of microbicide introduction in different settings [44–48], as well as the work by Williams et al. [48], Verguet et al. [39] and Walensky et al. [49–51]. Additional strengths here include that the dynamic model is carefully parameterised and rigorously fitted to a specific setting over multiple time points, including fitting to HIV and STI infection status by sex, as well as incorporating the HSV-2-efficacy of tenofovir found in CAPRISA 004 and the related bi-directional interaction between HSV-2 and HIV and the dynamic effects of gel on the HIV and HSV-2 epidemic. Walensky et al. [50] included only the first generation of HIV infections averted. Numerous programme implementation components were also included in our analysis, such as training, mass media and facility distribution costs, to provide a more comprehensive approach to costs than comparable studies to date. However, we did not account for potential toxicity or ART resistance as in Walensky et al. [50], because the concerns in this area are limited, as the authors themselves acknowledge. Moreover, the scenarios that we have modelled are less optimistic than those considered previously [49]. This study makes lower assumptions regarding the likely coverage that can be achieved, more optimistic assumptions about the impact of ART on life expectancy and the discounting of future DALYs averted (an sometime omitted step [51]. Consequently, we produce a far more conservative estimate of impact and cost-effectiveness.

Our modelling stresses the importance of maintaining condom use following microbicide introduction, to avoid large reductions in impact and cost-effectiveness. This issue is particularly of concern if condom use is already high [48], and for a microbicide with lower efficacy, as concomitant reductions in condom use could even lead to increases in the overall number of infections [48]. Similarly, our analyses highlight the degree to which the costs of gel introduction will be affected by programme costs, including the unit cost of gels, and the required frequency of HIV testing. For this reason, further research is needed to identify the optimal spacing between HIV testing, with the aim of achieving an appropriate balance between what may be required to minimise the potential for resistance and be feasible for women and health systems. Similarly, research to explore whether a single dose of tenofovir-containing gel, applied prior to sex, could be effective at reducing a woman’s risk of acquiring HIV is needed [52], to help reduce the product costs, as well as make gel use less burdensome for women.

The fact that 17% of VOICE participants reported engaging in anal intercourse during the 3-month period prior to enrolment highlights an important area for future modelling work [9]. For such modelling, data is needed from different sites and population groups on the frequency of anal sex over a specific timeframe (e.g. past 3 or 6 months), and whether a condom and/or gel is used during the last anal sex-act.

Although it is tempting to draw broader, national or regional conclusions from this analysis, our model projections may not be generalisable to other settings with different epidemic and behavioural profiles and different price levels. However, the impact we predict compares favourably with previous modelling of a hypothetical microbicide introduced into an area within Gauteng [19], and with modelling studies of the impact of male circumcision on HIV incidence in various settings of sub-Saharan Africa [53–55], matched to the same intervention assumptions made here [56].

The results are also highly sensitive to the discount rate used. Our central estimate used a 3% discount rate as commonly applied in international studies. However, when using the upper bound of 6%, just above the South African national discount rate of 5.5%, the cost-effectiveness of microbicides would worsen dramatically by 77%. Although not a realistic price to expect during mass-production, the programme’s incremental cost-effectiveness, at the current product cost of $1.60 per dose would still be highly cost-effective at $2,825 per DALY averted even using a 6% discount rate.

Some have questioned the viability of tenofovir gel given recent disappointing evidence of low user adherence in the VOICE trial. That users have a difficult time with adherence, however, should come as no surprise. Inconsistent adherence is the norm for both medications and any prevention methods that are user-controlled. Nonetheless, every indication is that the South African government and international donors are prepared to seek licensure for tenofovir gel should the results of the CAPRISA trial be sustained in the on-going FACTS 001 trial. Given that policymakers will soon face a key decision point regarding the possible introduction and scale up of tenofovir gel, the findings of this article are highly relevant. Despite intuitions to the contrary, it appears from the results of VOICE, FEMPREP, and other trials of daily oral or vaginal prophylaxis, that women seem better able to adhere to the coitally-dependent BAT-24 regimen than daily use.

Moreover, even if the attention of the wider field has shifted towards long-acting modalities such as the vaginal ring and injection, these products are still many years from licensure and there is undoubtedly benefit of beginning to prepare consumers for using ARVs to prevent infection. There will always be a market for methods like the vaginal gel that will likely not require on-going HIV testing and that can be positioned as a sexual lubricant with additional HIV fighting powers. Though highly cost-effective prevention options are available for men (VMMC and male condoms), there is still a great need to seek coitally-dependent options that women can initiate, such as microbicides, in whichever formulation.

This study has used epidemiological and cost modelling to estimate the impact of microbicide introduction among women of the general population in the Gauteng Province in South Africa. We used highly conservative assumptions to estimate gel impact and found that microbicides have the potential to significantly reduce HIV incidence among women in Gauteng. This is the most comprehensive model of the costs of microbicide introduction available in the literature to date and, even under the very conservative assumptions this analysis, estimated that microbicide introduction is very likely to be highly cost-effective. However this analysis also highlighted the importance of continued strong condom distribution and counselling programmes to maintain condom use and the need for identifying innovative approaches for supporting microbicide adherence.

^aAdherence estimates based on applicator returns for the remaining 884 women indicate that, on average, 72.2% (median = 60.2%) of self-reported sex-acts in the last 30 days were covered by two doses of gel.

^bIt should be noted that condom use outside the trial setting is lower and our approach will therefore underestimate the impact of microbicides on HIV incidence.

^cThe 6% discount rate is consistent with the South African central bank discount rate of 5.5% [57].