Overview

Intervention scale-up strategies

We considered strategies where ART, PrEP, or both are scaled up to recruit 25%, 50%, 75% or 100% of remaining eligible individuals annually. For ART scale-up, we considered strategies where ART was offered either to individuals with CD4 cell counts ≤350 cells/μL only (Guidelines), or to all identified HIV-infected individuals (Universal) following a national testing campaign [30]. Thus, the 100% Guidelines strategy models outcomes under full implementation of the current HIV treatment guidelines, whereas the 100% Universal strategy reflects the outcomes under an ideal ‘Universal Test and Treat’ program [12, 31].

To investigate the role of PrEP in controlling South Africa’s HIV epidemic, we considered its use in the general adult population (General). We also investigated the health and economic outcomes of using PrEP only in individuals at high-risk of acquiring HIV, defined here as having four or more sexual partners per year and low condom use (Focused).

Antiretroviral therapy

For the Guidelines strategy, we used a treatment eligibility threshold of 350 cells/μL. In the base case (status quo), 40% of treatment-eligible individuals receive ART, and PrEP is not used [24]. We assumed that 10% of untreated treatment-eligible individuals start ART each year. For the Universal strategy, any infected individual identified and linked to care started ART.

ART is associated with both personal and transmission benefits. Treated individuals have a longer remaining life expectancy than untreated individuals (41.3 remaining years for a 15-year-old if treatment is initiated early, versus 12.2 if untreated) [32, 33]. HIV-related mortality rates for each disease stage and ART status are in Additional file 1 – Model details: Table S2. In addition, treated individuals are 95% less likely to infect their sexual partners [9]. We assumed a 2% annual rate of loss-to-follow-up throughout the time horizon. Although studies report a higher rate of loss-to-follow-up in the short term, our estimate is consistent with long-term experience, and the higher efforts to retain patients in care [34, 35]. Individuals who discontinue ART do not re-start treatment.

Pre-exposure prophylaxis

For the General strategies, we assumed that all uninfected adults in the population are eligible to start PrEP. While achieving high rates of PrEP coverage among the entire population may be unfeasible for practical reasons, we explored these possibilities to gain insight and establish an upper bound for the benefits that could be achieved with a large-scale PrEP program. In the Focused strategies, we assumed PrEP is offered only to a population of 3 million uninfected individuals at highest risk for acquiring HIV. This limited approach may render PrEP more feasible. We adjusted the risk profile of the ineligible uninfected population in the Focused strategy to match the overall population risk in the General strategy.

We assumed uninfected individuals taking PrEP have a 60% lower chance of acquiring HIV from an infected sexual partner [15, 16]. Because of the uncertainty over PrEP’s real-world effectiveness highlighted in recent trials, we broadly varied PrEP’s effectiveness in the sensitivity analysis (10% to 90%) [17, 18, 36]. We assumed a 5% annual drop-out rate beyond the non-adherence that is reflected in our estimate of PrEP effectiveness. We also assumed that individuals who were infected while on PrEP and later placed on ART had similar benefits from treatment as those who never received PrEP [15].

Model structure

We divided South Africa’s adult population (15 to 49 years old, 26.3 million people) into 11 groups defined by HIV disease stage (uninfected, CD4 count >350 cells/μL, 200 to 350 cells/μL, and <200 cells/μL), ART status (untreated, treated, and lost-to-follow-up), and PrEP status (receiving or not receiving PrEP). A schematic diagram of the model is included in Additional file 1 – Model details.

HIV transmission and progression

We modeled heterosexual HIV transmission based on the average number of annual partnerships per individual, assuming random homogeneous mixing in the population. The annual probability of HIV transmission in a partnership depends on the disease stage and ART use of the infected partner, PrEP use by the uninfected partner, and the rate of condom use and effectiveness.

HIV progression rates prior to and while receiving ART were based on published studies from South Africa [32, 33]. ART was assumed to reduce mortality and lower infectivity. We assumed no significant differences in disease progression rates for individuals who previously received PrEP or for those who discontinued ART versus individuals who were never treated.

Model calibration

Our model was calibrated to match information about South Africa’s HIV epidemic, including initial HIV prevalence (17.9%), HIV incidence (1.4%), and population growth [24, 37]. Prevalence in individuals entering the model at age 15 was estimated at 5% [25]. Using this data, we estimated that HIV prevalence will decline to 10.4% over the next 20 years, due to a reduction in incidence from 1.4% to 0.8%, while population growth will average 1.1% annually.

Health outcomes and costs

We measured costs and QALYs for each strategy over 20 years, discounted at 3% annually. We included costs and benefits incurred over the time horizon, plus discounted future lifetime costs and QALYs incurred by the population at the end of the time horizon.

We assumed lower quality of life as HIV progresses, and that ART increases quality of life by 10% of the difference between uninfected and infected untreated individuals, to include both the benefits and the side effects of ART.

We assumed annual ART costs of $150, and annual PrEP costs of $80, reflecting the reduced prices of HIV drugs accessible in South Africa [28, 29]. All individuals incurred an annual non-HIV medical cost of $200. In addition, HIV-infected individuals incurred $1,000 in HIV-related annual medical costs [27].

Health benefits of scaling up ART and PrEP

Figure 1 presents the number of HIV infections averted and incremental costs versus the status quo for 100% scale-up of all strategies considered, individually and in combination. In general, strategies involving Universal ART scale-up avert more infections than strategies involving General PrEP for the same rate of scale-up.

In the extreme scenario of 100% Universal ART without PrEP, 75% of new infections are averted over 20 years, whereas 100% coverage with Guidelines ART averts 38% of new infections. Without PrEP, up to 50% more people start treatment in the Universal ART strategies compared to the corresponding Guidelines strategies, and 2 to 3.4 times as many infections are averted, depending on program scale (Table 2).

Scaling up General PrEP alone to 100% while keeping ART at current recruitment levels would avert 63% of new infections over 20 years. Most of PrEP’s benefits are achieved with General PrEP programs with less aggressive recruitment rates; increasing PrEP recruitment rate from 50% to 100% in the general population for any given ART level averts only an additional 2% to 7% of new infections. This is because a significant proportion of uninfected individuals will start PrEP over the time horizon even with a less aggressive program recruitment, and the incremental benefits of an initially less aggressive program approach the benefits of a more aggressive program over time. Figure 2 shows the infections averted by all combinations of General PrEP and Guidelines ART (Figure 2A) or Universal ART (Figure 2B). Note that as indicated in Figure 2B, a 10% Universal ART program offers incremental benefits versus the status quo due to expanded ART eligibility criteria (1,530,000 infections averted); adding General PrEP averts up to an additional 2,270,000 infections.

Implementing PrEP in 50% of the uninfected population in the presence of Guidelines ART programs averted 61% to 69% of new infections, depending on the scale of the ART program, versus 16% to 38% for Guidelines ART alone. The marginal benefit of PrEP was smaller when added to Universal ART programs, since treating those with early HIV reduces the size of the population that transmits HIV. Thus, 50% PrEP incremental to Universal ART strategies averted 75% to 85% of new infections, versus 53% to 75% for Universal ART alone. This suggests that Universal ART and General PrEP function as partial substitutes and the benefits of PrEP may be more significant if Universal ART cannot be accomplished. If it is achievable, Focused PrEP can be an effective intervention: 100% Focused PrEP required reaching 10% of the adult population, and averted 57% of all new infections even without any further ART scale-up. When added to 50% ART programs, Focused PrEP averted a significant number of additional infections. Scaling up to 50% Guidelines ART and 100% Focused PrEP averted 2.4 times as many infections as the ART program alone. These effects were smaller for larger ART programs and Universal ART, but in all scenarios Focused PrEP averted 20% or more incremental HIV infections compared to the ART program alone (Table 2). These results, however, critically depend on the ability to identify the high-risk population and the effectiveness of PrEP in this group, both of which are uncertain.

Cost-effectiveness

Universal ART had higher total costs, but also added significantly more QALYs than Guidelines ART: Universal ART cost $310 to $320 per QALY gained versus the status quo, whereas Guidelines ART scale-up cost $410 to $420 per QALY gained versus the status quo, with total costs as well as benefits lower than the corresponding Universal ART strategy. Figure 3 shows the benefits (QALYs) and costs for single and combination programs (Guidelines and Universal ART, General and Focused PrEP).

With our base case assumptions of PrEP effectiveness, strategies consisting of Focused PrEP alone were cost-saving compared to the status quo and offered significant benefits. Combinations of Focused PrEP and ART dominated ART alone, offering greater benefits at lower costs: Universal ART scale-up to 100% in addition to Focused PrEP cost $150 per QALY gained. This was also the most effective of all strategies (Table 2). However, if Focused PrEP is infeasible, Universal ART is the alternative with the most attractive cost-effectiveness profile.

We estimated that General PrEP strategies would cost $1,200 per QALY gained compared to the status quo. For similar levels of benefit, General PrEP strategies were more costly than ART programs. Adding 50% General PrEP to Guidelines ART scenarios gained QALYs versus ART alone at costs between $1,600 per QALY gained (for 25% Guidelines ART) and $2,650 per QALY gained ( for 100% Guidelines ART). When added to Universal ART, 50% General PrEP added QALYs at $4,300 per QALY gained (25% Universal), but becomes more expensive at $13,300 per QALY gained for larger ART programs (100% Universal).

While all strategies met traditional thresholds of cost-effectiveness, our analysis suggests that some combinations of strategies are much more attractive, given South Africa’s gross domestic product per capita of $10,700 [38].

Sensitivity analysis

We performed extensive one-way sensitivity analyses on all model parameters. We found that the results varied with the effectiveness of the two interventions, discontinuation rate for both, the cost of PrEP, the number of sexual partners, and the rates of condom usage in the population. However, the same alternatives were preferred in terms of infections averted and cost-effectiveness. In particular, we found that a PrEP effectiveness of 10% limited the cost savings of Focused PrEP by 60% and reduced the cost-effectiveness of General PrEP by 50%.

Higher rates of PrEP attrition played little role in the effects of the programs, assuming enough recruitment to the program occurred to compensate for the attrition. Higher rates of ART attrition (which made individuals ineligible for ART) did not significantly decrease the cost-effectiveness of ART scale-up, even at relatively high levels (10% versus 2% in the base case).

The Focused PrEP strategy was no longer cost-saving for costs of PrEP above $150. This price increase may reflect additional costs to target high-risk individuals, improve adherence, and monitor for adverse effects of PrEP. To be as cost-effective as universal ART ($310 per QALY), Focused PrEP annual cost would be $260 per person, and $580 to cost as much per QALY ($1,200 per QALY) as General PrEP at $80 in the base case.