HIV Articles  
HIV-1 Prevention With ART and PrEP: Mathematical Modeling Insights Into Resistance, Effectiveness, and Public Health Impact - Editorial
  Download the PDF here
Download the PDF here
The Journal of Infectious Diseases July 15, 2013
Connie Celum,1 Timothy B. Hallett,2 and Jared M. Baeten1 1Department of Global Health, Medicine and Epidemiology, University of Washington, Seattle; and 2Department of Infectious Disease Epidemiology, Imperial College London, United Kingdom
For both PrEP and ART for HIV-1 prevention, adherence is key to effectiveness.
This model clearly demonstrates that both ART and PrEP, particularly when rolled out together, offer the potential for substantial HIV-1 prevention. Recognizing the potential risks of PrEP and ART, including antiretroviral resistance, is critical for developing mitigating strategies, because the potential benefits of these new prevention strategies are substantial and there is real public health risk in not implementing tools that we know work.

(See the major article by Abbas et al on pages 224-34.) below The development of, and globlal access to, effective antiretroviral medications revolutionized human immunodeficiency virus type 1 (HIV-1) care. In addition to their important life-saving treatment benefits, antiretrovirals have recently been demonstrated to be highly efficacious for HIV-1 prevention as well, when used as antiretroviral therapy (ART) to reduce the infectiousness of HIV-1-infected persons and pre-exposure prophylaxis (PrEP) for uninfected persons who have ongoing HIV-1 exposure. Antiretroviral-based prevention, both ART and PrEP, are among the most promising strategies for reducing the number of new HIV-1 infections globally. Consequently, policymakers are weighing the costs, benefits, and risks of public health implementation of ART and PrEP for HIV-1 prevention. One potential risk of both ART and PrEP is the selection and transmission of HIV-1 variants that are resistant to one or more antiretroviral medications, which can result in HIV-1 treatment failure with associated morbidity and mortality and increased costs (of more complex second- and third-line treatment regimens); thus, there has been considerable speculation about the potential risks of resistance from both ART and PrEP.
In this issue of the Journal of Infectious Diseases, Abbas et al present a mathematical model to estimate the number of HIV-1 infections averted and the number of acquired and transmitted HIV-1 cases of resistance in a setting similar to South Africa and under several scenarios about coverage of ART and PrEP [1]. For PrEP, the authors assumed use of combination emtricitabine-tenofovir, for which efficacy has been demonstrated [2-4]. For ART, the authors modeled first-line regimens containing the same antiretrovirals and assumed that second-line drugs were not available, given limited availability of second-line medications in many resource-limited settings. A number of additional scenarios were analyzed, including having ART initiation at CD4 lymphocyte cell counts at either <200 or <350 cells/μL, reflecting evolving international guidelines on clinical benefits of earlier ART initiation, and allowing for "inappropriate" PrEP use by persons that are already infected with HIV-1, either through PrEP initiation occurring during unrecognized seronegative acute HIV-1 infection or PrEP initiation by persons with undocumented, chronic HIV-1 infection, which could occur if HIV-1 testing is not conducted prior to initiation or through "black market" availability of PrEP. The authors used optimistic scenarios for ART retention and PrEP effectiveness; notably, the model assumed general distribution of PrEP rather than risk-targeted delivery. Not surprisingly, the results of this mathematical modeling article underscore that population-level coverage and effectiveness (which is dependent on adherence) are the main determinants of the number of infections averted with both ART and PrEP, and that implementation of a combination of ART and PrEP prevents more infections in a population than a program that delivers exclusively either ART or PrEP. More interestingly, the model analysis also suggests that HIV-1 drug resistance in a population would be largely driven by ART, not PrEP, in all scenarios modeled, as a result of insufficient ART adherence or lack of viral load monitoring in ART programs, leading to selection of resistant variants during incomplete viral suppression. The model also finds that the population prevalence of resistance as a direct result of PrEP may be very low and that the greatest resistance risks related to PrEP would be from inappropriate use by persons already HIV-1 infected, rather than from PrEP being prescribed for HIV-1 prevention, even anticipating inadvertent prescribing for persons with unrecognized acute HIV-1 infection.
What do these findings mean for potential implementation of PrEP for HIV-1 prevention? First, these results directly address the often-voiced concern that PrEP will lead to substantial HIV resistance in populations. Instead, because high PrEP adherence prevents most HIV-1 infections, it is unlikely to select for resistant variants, although low PrEP adherence does not prevent infection; the model presented by Abbas et al suggests that substantial population-level resistance is unlikely. Moreover, even with optimistic assumptions about ART continuation rates, the amount of resistance generated from ART failure greatly exceeds the resistance selected by PrEP. Indeed, emerging evidence from Africa has demonstrated increasing resistance accompanying ART roll-out over the past decade [5-7]. Nevertheless, if persons on PrEP alternated between periods of good and poor adherence in patterns that increased the risk of them becoming HIV-1 infected and then taking PrEP, the risk of resistance could be greater than predicted in this model. Implementation of PrEP will require ongoing complementary strategies to ensure high quality HIV-1 testing, reduce HIV-1 risk, and maximize PrEP-taking.
A second message from the Abbas et al model is the importance of HIV-1 testing before PrEP initiation to avoid inadvertent exposure for an HIV-1-infected person to what is effectively suboptimal mono- or dual-agent ART. The model assumed that 2.5% of persons with undiagnosed chronic HIV-1 infection would initiate PrEP each year, which is arguably very high. Nevertheless, the model alerts us to the importance of strategies to monitor quality HIV-1 testing and PrEP pharmacovigilance during this period where PrEP is moving from efficacy trials to implementation.
Third, a highly intuitive finding of the model is that less drug resistance could result if ART and PrEP regimens were used that did not include the same antiretroviral agents. The completed, first-generation PrEP trials used tenofovir, alone or in combination with emtricitabine, resting on the substantial body of clinical safety and experience with these agents for testing PrEP as a novel HIV-1 prevention strategy. New PrEP agents are in development, but their use would not be routine for several years. While, hypothetically, it is preferable to utilize PrEP regimens that do not overlap with antiretrovirals used for treatment, there is also a cost of inaction-missing the opportunity to prevent new HIV-1 infections with demonstrated effective tenofovir-based PrEP while waiting for new safe and effective PrEP regimens to be identified. While providing some new insights, there are also limitations to the Abbas et al model. For ART, the key benefit that was not included in this model was its health impact in terms of saving lives. As the primary benefit of ART is to prolong life, and the primary problem of resistance is the loss of efficacy of ART, this is an important gap in the model. For PrEP, a critical operational factor for maximizing impact in terms of infections averted will be "prioritization," in which age, gender, and risk behaviors are incorporated into risk assessments for potential PrEP users to maximize the likelihood that PrEP is provided to those most at risk of HIV-1 infection. To optimally use resources for PrEP, programs will need to prioritize those who are at highest risk of HIV-1 acquisition and are motivated to take PrEP. Whereas the authors of the present model used a coverage level of 30% of the general population and included those that did not adhere to PrEP well, other models have suggested that if PrEP delivery programs can target delivery to those at greater HIV-1 risk and achieve higher adherence in a prioritized population, by reducing the total number of new HIV-1 infections, PrEP could even reduce the prevalence of drug resistance [8]. Thus, the complex mathematical model developed by Abbas et al helps identify some of the next steps for mathematical models and needs for empiric data to clarify policy considerations and implementation priorities for antiretroviral-based HIV-1 prevention through ART and PrEP.
For both PrEP and ART for HIV-1 prevention, adherence is key to effectiveness. For ART, the result is adherence over a lifetime, or until a cure is available. Expanding implementation of ART for HIV-1 prevention will include persons initiating at higher CD4 lymphocyte counts and earlier in their disease course before they have experienced symptoms, and they may face heightened adherence challenges. PrEP adherence has different challenges than ART, namely requiring persons without HIV-1 to perceive their own risk sufficiently to initiate and adhere to PrEP. Thus, PrEP needs to be delivered in a different model than ART, as it is not a commitment to life-long medications, but specially directed to individuals during life periods of highest risk.
While much can be learned from the Abbas et al model about the potential for generation and spread of HIV-1 antiretroviral resistance related to ART and PrEP, equally important is what this model can teach us about the public health impact of these prevention strategies. The authors have moved the discussions about PrEP forward from modeling simply the number of drug resistance cases with their public health perspective in which they present the ratio of cumulative HIV-1 infections averted to prevalent HIV-1 drug resistance, which puts the deleterious effect of drug resistance into context with the benefits of HIV-1 infection prevention. This model clearly demonstrates that both ART and PrEP, particularly when rolled out together, offer the potential for substantial HIV-1 prevention. Recognizing the potential risks of PrEP and ART, including antiretroviral resistance, is critical for developing mitigating strategies, because the potential benefits of these new prevention strategies are substantial and there is real public health risk in not implementing tools that we know work.
Antiretroviral Therapy and Pre-exposure Prophylaxis: Combined Impact on HIV Transmission and Drug Resistance in South Africa
Ume L. Abbas,1 Robert Glaubius,1 Anuj Mubayi,1,a Gregory Hood,2 and John W. Mellors3 1Departments of Infectious Diseases and Quantitative Health Sciences, Cleveland Clinic, Ohio; 2Pittsburgh Supercomputing Center and 3Division of Infectious Diseases, School of Medicine, University of Pittsburgh, Pennsylvania
Background. The potential impact of antiretroviral therapy (ART) and pre-exposure prophylaxis (PrEP) with overlapping and nonoverlapping antiretrovirals (ARVs) on human immunodeficiency virus (HIV) transmission and drug resistance is unknown.
Methods. A detailed mathematical model was used to simulate the epidemiological impact of ART alone, PrEP alone, and combined ART + PrEP in South Africa. Results. ART alone initiated at a CD4 lymphocyte cell count <200 cells/μL (80% coverage and 96% effectiveness) prevents 20% of HIV infections over 10 years but increases drug resistance prevalence to 6.6%. PrEP alone (30% coverage and 75% effectiveness) also prevents 21% of infections but with lower resistance prevalence of 0.5%. The ratio of cumulative infections prevented to prevalent drug-resistant cases after 10 years is 7-fold higher for PrEP than for ART.
Combined ART + PrEP with overlapping ARVs prevents 35% of infections but increases resistance prevalence to 8.2%, whereas ART + PrEP with nonoverlapping ARVs prevents slightly more infections (37%) and reduces resistance prevalence to 7.2%.
Conclusions. Combined ART + PrEP is likely to prevent more HIV infections than either strategy alone, but with higher prevalence of drug resistance. ART is predicted to contribute more to resistance than is PrEP. Optimizing both ART and PrEP effectiveness and delivery are the keys to preventing HIV transmission and drug resistance.
Oral antiretroviral (ARV) pre-exposure prophylaxis (PrEP) is a new biomedical intervention against human immunodeficiency virus (HIV) transmission with proven efficacy [1-3]. There is concern, however, about the potential emergence and spread of HIV drug resistance arising from the rollout of PrEP, particularly in resource-constrained settings, where antiretroviral therapy (ART) options are limited [4]. This concern is amplified by the possibility that the same ARVs will be used for both ART and PrEP. The combination of 2 nucleoside reverse-transcriptase inhibitors, tenofovir (TDF) and lamivudine or emtricitabine (3TC or FTC, respectively), with 1 nonnucleoside reverse-transcriptase inhibitor (NNRTI), efavirenz or nevirapine, is the World Health Organization-recommended first-line ART regimen in several countries worldwide, including South Africa [4], and TDF or TDF + FTC have shown efficacy in HIV prevention trials [1-3]. Thus far, only 9 drug-resistant cases have been observed among clinical trial participants on PrEP, most of whom had unrecognized acute infection at enrollment. However, clinical trials of PrEP are not designed to address the population-level and/or long-term epidemiological impact of PrEP, including consequences of drug resistance. We therefore used a mathematical model [5] to examine the potential impact of orally administered overlapping and nonoverlapping PrEP and ART on HIV transmission and drug resistance in South Africa.
Model Structure

We developed and analyzed a detailed mathematical model to assess the impact of PrEP and ART implementation on the adult population (aged 15-49 years) of South Africa, using deterministic and stochastic modeling techniques and the programming language C/C++. The model describes population and epidemiological stratifications based on gender (male; female), sexual activity (high; medium; low; lowest), PrEP and ART use status (on; not on), infection status (susceptible; infected), stage of HIV infection (acute preseroconversion; acute postseroconversion; early chronic; late chronic; AIDS), and HIV drug susceptibility (drug-sensitive; drug-resistant). Model parameter assignments are made using recent results from PrEP trials [1-3, 6, 7] and data mainly from South/sub-Saharan Africa on HIV disease progression [8], infectivity [9], sexual behavior [10], ART rollout [4, 11-18], and HIV drug resistance [19-33]. The model is calibrated to simulate the HIV epidemic in South Africa with adult HIV prevalence (Supplementary Figure S1) reaching 17% at the end of 2003, having a female-to-male prevalence ratio of 1.6 and HIV incidence near 2.4% [34]. A simplified model structure is shown in Supplementary Figure S2 and model input parameters are shown in Tables 1 and 2 and Supplementary S1. Model equations and details are provided in the Supplementary Text S1.
HIV Drug Resistance
We stratify HIV-infected individuals based on their ARV status, HIV drug susceptibility, type of drug resistance, and virus population dynamics of drug-resistant HIV, including persistence and reversion of resistance [35]. The model tracks individuals infected with different viral variants over time, either untreated, on PrEP, or on ART. We do not explicitly represent different drug-resistant mutants but assume the emergence and transmission of 184V with PrEP use [1, 2, 6]; and although several different mutations may arise with ART use (such as 103N, 106M, 181C, 184V, 65R), 184V is the most common. "Transmitted resistance" may occur from a donor either on PrEP, not on PrEP, on ART, or not on ART, having a majority population of drug-resistant virus, to a recipient either on or not on PrEP. "Acquired resistance" may occur due to de novo selection on PrEP or ART in persons with wild-type infection, reemerge from archived drug-resistant variants on PrEP or ART, or persist/accumulate on ART.
Upon removal of drug pressure, either by discontinuation of ART or PrEP or transmission to a recipient not on PrEP or ART, the drug-resistant virus may revert to drug-sensitive virus after a period of persistence. Prior to reversion, drug-resistant variants comprise the majority population, whereas following reversion, they become a minority population [35].
ARV Interventions, Base-Case Scenarios, and Model Analyses
We simulate 3 different rollout strategies for ARV-mediated HIV prevention-ART alone, PrEP alone (a hypothetical illustration), and ART + PrEP-and compare the epidemiological outcomes with an ARV-naive epidemic. For each strategy, we first construct and analyze a reference-case (base-case) scenario using a defined set of input parameters, including estimates of the effectiveness of ART and PrEP for prevention of HIV from the HPTN 052 clinical trial [36] and the Partners PrEP study [1], respectively; followed by uncertainty and sensitivity analyses [37].
Base-Case Analyses
ART Rollout and Effectiveness

In our model, individuals become treatment eligible at CD4 lymphocyte cell counts <200 cells/μL [11]. Treatment scale-up starts at the end of 2003 [17] and the proportion of eligible persons on ART (ie, coverage) reaches 55% by the end of 2009 [18] and 80% by the end of 2011 [11]. Coverage is then maintained at 80% throughout the simulation [11]. To represent the current situation in South Africa, we simulate 2 additional scenarios of expanded ART rollout in which treatment eligibility threshold changes at the end of 2009 to include individuals with CD4 counts between 200 and 350 cells/μL [4], reaching 66% coverage at CD4 count threshold <350 cells/μL by the end of 2011 [15]. Coverage is then: (1) maintained at the 66% level (termed status-quo coverage) or (2) increased to reach 80% at the end of 2016 [12] and maintained thereafter (termed optimized coverage). We model only first-line ART with conservative coverage to focus on the interplay between first-line ART and PrEP, assuming that access to second-line regimens [38] and drug-resistance testing [39] is limited. In base-case analyses, we assume ART reduces HIV transmission by 96% [36]. Our model represents virologic suppression and failure (with/without drug resistance), dropout, survival, and HIV transmission during the first and subsequent years of ART.
PrEP Rollout and Effectiveness
The effectiveness of PrEP against HIV acquisition is a composite of efficacy and adherence [40]. The Partners PrEP study showed the effectiveness of oral TDF + FTC PrEP to be 75% (95% confidence interval [CI], 55-87); with 90% efficacy of PrEP in those with near-perfect adherence, and only 12% of subjects having less than 80% adherence [1].
We therefore stratify individuals into 2 groups based on their level of adherence to PrEP: high or low. For base-case analyses, we assume that close to 90% of individuals have 95% adherence and about 10% have low (near zero) adherence. However, given the conflicting results from different PrEP trials (TDF + FTC was ineffective in the Fem-PrEP trial [6], and oral TDF was ineffective in the VOICE trial [7]), for uncertainty and sensitivity analyses we use a wide range of input estimates for PrEP efficacy and adherence and the proportion of individuals in the 2 (high/low) adherence groups.
PrEP (TDF + FTC) scale-up starts in 2012 and achieves 30% coverage over a 5-year period that is then maintained. We assume that PrEP is about 90% efficacious against wild-type virus [1, 2] and that the average duration of PrEP use is 5 years in susceptible individuals with HIV testing every 6 months (and PrEP discontinuation if HIV infection occurs). For the ART + PrEP strategy, in addition to our base-case scenario with overlapping drugs (ie, cross-resistance) between PrEP (TDF + FTC) and ART (TDF + FTC + NNRTI), we simulate an alternate scenario with identical model input and structural assumptions except for there being no overlap/cross-resistance between ART and PrEP.
Uncertainty Analyses
We perform uncertainty analyses to estimate the extent of variation in our projections across a broad range of input parameter estimates that include the following assumptions (Tables 1 and 2): ART effectiveness is 73%-99%; PrEP efficacy against wild-type virus is 70%-99%; PrEP adherence among individuals highly adherent is 80%-99% and among poorly adherent is 1%-79%; the proportion of individuals highly adherent is 10%-90%; PrEP coverage is 15%-45%; average duration of PrEP use is 2.5-7.5 years; the frequency of HIV testing under the PrEP program is 3-9 months; and the time by which about 100% of wild-type virus recipients acquire PrEP resistance from inappropriate PrEP use with perfect adherence is 3-9 months with the median time to acquired resistance of about 1 month [41]. We perform 50 000 simulations using Latin hypercube sampling (LHS) for each ARV-based strategy, and compute the epidemiological outcomes (median and interquartile range [IQR]) in comparison with an ARV-naive baseline epidemic. We also calculate the outcomes for the overlapping and nonoverlapping ART + PrEP strategies in comparison with ART alone as baseline.
Sensitivity Analyses
We conduct sensitivity analyses to identify those parameters that exert the greatest influence on the predicted model outcomes for each strategy. For these time-dependent multivariate analyses, we use the input and output data from our uncertainty analyses to derive standardized regression coefficients. In addition, we examine the sensitivity of the model's predictions to the modeling technique by comparative analyses of our stochastic and deterministic model simulations.
Inappropriate PrEP Use
We simulate 2 contexts of inappropriate PrEP initiation and use by previously infected individuals by extending our PrEP-alone and ART + PrEP base-case scenarios. In the first, individuals in the preseroconversion phase of acute HIV infection are started on PrEP ("window use"). In the second, individuals with undiagnosed established HIV infection start PrEP inappropriately at a rate of 2.5% per year ("general use"). The duration of inappropriate PrEP use following seroconversion is determined by the HIV testing interval assumed for the PrEP program (6 months for base-case; LHS range: 3-9 months). For general use, the duration is determined by the frequency of population surveillance (1 year for base-case).
Prevention of HIV Transmission
Base-Case Scenarios

Figure 1A shows the impact of different ARV-based strategies on HIV prevention after 10 years compared with an ARV-naive epidemic. ART alone is projected to prevent 20% of HIV infections (0.92 million). Similarly, PrEP alone prevents 21% (0.96 million) of HIV infections. The combined strategy of ART + PrEP is predicted to be most effective, reducing infections by 35% (>1.6 million) with overlapping regimens and 37% (>1.7 million) with nonoverlapping ARV regimens.
Expanded ART Rollout
The scenarios, which expand treatment rollout to include coverage at a CD4 count <350 cells/μL, result in modest increase in infections prevented when measured against the base-case scenarios of ART alone and overlapping ART + PrEP (Figure 1B). Coverage at 66% (status-quo coverage) respectively prevents 23% and 38% of infections, while 80% coverage (optimized coverage) prevents 28% and 41% of infections versus 20% and 35% for the base-case ART-alone and ART + PrEP scenarios.
Prediction Uncertainty of HIV Prevention
Figure 2A shows the results of uncertainty analyses for the 3 ARV-based strategies. The median decrease in HIV infections with ART alone after 10 years is 15% (IQR, 12%-19%), PrEP alone is 14% (IQR, 10%-18%), overlapping ART + PrEP is 27% (IQR, 22%-31%) and non-overlapping PrEP is 28% (IQR: 23%-33%). Overlapping ART + PrEP (Figure 2C) prevents a median of 12.7% (IQR, 9.1%-17.2%) more infections than ART alone. Results are similar for nonoverlapping ART + PrEP (median, 14%; IQR, 10%-18.9%).
HIV Drug Resistance
Base-Case Scenarios

Figure 3A shows the impact of different ARV-based strategies on HIV drug resistance prevalence compared with an ARV-naive epidemic. After 10 years of PrEP alone, the prevalence of overall resistance is low at 0.5% (20 090 cases). Drug resistance prevalence is higher from the ART-alone strategy at 6.6% overall (307 254 cases) with 4.2% acquired (195 758 cases) and 2.4% transmitted resistance (111 497 cases). The prevalence of resistance increases further from overlapping ART + PrEP to 8.2% (339 895 cases) with the prevalence of acquired and transmitted ART resistance increasing to 4.6% and 3.3%, respectively. With nonoverlapping ART + PrEP, drug resistance prevalence falls modestly to 7.2% due to a lower prevalence of transmitted ART resistance (2.2%). In terms of the number of prevalent cases of drug resistance (data not shown), acquired ART resistance falls modestly from both overlapping and nonoverlapping ART + PrEP, when measured against ART alone; however, transmitted ART resistance rises with overlapping but falls with nonoverlapping ART + PrEP. Both acquired and transmitted cases of PrEP resistance fall from ART + PrEP when measured against PrEP alone.
Expanded ART Rollout
The scenarios of expanded ART rollout result in a modest increase in drug resistance prevalence when measured against the base-case scenarios of ART alone and overlapping ART + PrEP strategies (Figure 3B). Drug resistance prevalence increases to 8.3% and 10.1% in status-quo coverage scenarios and 11.4% and 13.4% in optimized coverage scenarios, versus 6.6% and 8.2% in the base-case scenarios of ART alone and ART + PrEP, respectively.
Ratio of Cumulative Infections Prevented to Prevalent and Incident Drug-Resistant Cases
To compare the resistance consequences of different ARV-based strategies, we calculated ratios of cumulative infections prevented to resistance over 10 years, either defined as prevalent cases (prevailing cases with majority drug-resistant variants; Figure 4A) or incident (new cases of transmitted or acquired drug resistance; Figure 4B). PrEP alone prevents about 48 infections for each prevalent drug-resistant case and more than 5 infections for each incident drug-resistant case. Inappropriate window-use in the PrEP-alone strategy decreases these ratios modestly to 46 and 4.8, respectively. By contrast, inappropriate general-use PrEP markedly reduces the ratios to 10 and 1, respectively. ART alone prevents about 7 infections for each prevalent drug-resistant case and about 1 infection for each incident drug-resistant case, which is 6- to 7-fold lower than for PrEP. The prevention-resistance ratios for prevalent and incident cases are 9.8 and 1.4, respectively, for overlapping ART + PrEP, and 14.7 and 1.7, respectively, for nonoverlapping ART + PrEP.
Prediction Uncertainty of HIV Drug Resistance
Figure 2B shows the results of uncertainty analyses for HIV drug resistance outcomes from different ARV-based strategies. After 10 years, the median overall prevalence of drug resistance from ART alone is 5.9% (IQR, 4.6%-7.4%), from PrEP alone is 0.5% (IQR, 0.3%-0.7%), from overlapping ART + PrEP is 7% (IQR, 5.6%-8.8%), and from nonoverlapping ART + PrEP is 6.5% (IQR, 5.2%-8.1%). These finding are consistent with our base-case scenarios.
Overlapping ART + PrEP compared to ART alone (Figure 2C), increases the number of prevalent overall and transmitted ART-resistant cases after 10 years by a median 8.8% (IQR, 5.8%-13.1%) and 15.9% (IQR, 11.4%-21.9%), respectively, while modestly decreasing the number of acquired ART-resistant cases (median, -0.9%; IQR, -1.8% to 0%). Nonoverlapping ART + PrEP decreases the overall drug resistance prevalence at 20 years (median, -4%; IQR, -7.5% to -0.7%).
Inappropriate PrEP Use
Inappropriate PrEP use by persons infected at baseline increases HIV drug resistance from PrEP. When measured against the overlapping ART + PrEP base-case, an overlapping ART + PrEP strategy that includes inappropriate window-use PrEP prevents almost the same number of infections (1.63 million), with a modest increase (8.3% vs 8.2%) in the prevalence of resistance (data not shown). In contrast, overlapping ART + PrEP with inappropriate general-use PrEP leads to an increase in the overall resistance prevalence from 8.2% to over 10%, with acquired PrEP resistance rising to 1.3% from 0.2% and transmitted PrEP resistance to 0.4% from 0.1% (data not shown). Nonoverlapping ART + PrEP with inappropriate general-use PrEP raises the overall resistance prevalence to 8.5% (data not shown).
Sensitivity Analyses
The results of the sensitivity analyses are described in detail in Supplementary Text S1 and summarized in Table 3.
The important insights derived from our study are several. First, an ART strategy of treatment initiation at a CD4 count <200 cells/μL combined with PrEP prevents more infections than either ART alone or PrEP alone; however, the incremental benefit of PrEP critically depends on PrEP efficacy, adherence, and coverage. Second, the prevalence of HIV drug resistance is largely driven by ART in both ART alone and ART + PrEP strategies. Third, PrEP alone results in low prevalence of drug resistance; high PrEP adherence leads to fewer infections and less opportunity for acquired resistance, while low adherence leads to predominantly wild-type breakthrough infections because of low drug pressure for emergence of acquired resistance. Fourth, use of overlapping ARVs for both ART and PrEP could increase drug-resistance prevalence compared to ART alone due to more frequent transmitted resistance. By contrast, resistance prevalence falls with non-overlapping ART + PrEP; however, this decrease is modest because the principal driver of resistance is ART, not PrEP. Fifth, inappropriate PrEP initiation among individuals with undetectable HIV infection produces only a minor increase in the overall resistance prevalence; however, inappropriate PrEP use among persons with established HIV infection could significantly increase drug resistance from PrEP. Lastly, PrEP prevents many more infections per case of resistance than ART does.
The extent of coverage and the degree of effectiveness against HIV transmission are the principal determinants of the infections prevented with ART. Similarly, PrEP coverage and effectiveness against HIV acquisition are the key determinants of the additional preventive benefit of ART + PrEP. The paradigm of test and treat [42] has gained considerable momentum, and the HPTN 052 trial [36] has provided the needed proof of concept for ART-based prevention, though its population-level impact may be limited by potential reluctance of asymptomatic HIV-infected persons for ART initiation. Notwithstanding scale-up efforts, there is considerable unmet need for ART in resource-constrained settings; about 60% of those eligible did not have access to ART at the end of 2010 [43]. Moreover, the population-level effect of treatment as prevention could be limited by the actual proportion of infected individuals optimally and durably suppressed on ART. In 2010, of the 1.2 million infected persons in the United States, 80% were aware of their status, but 41% were retained in care, and only 28% had virologic suppression [44]. The situation is much worse in sub-Saharan Africa, where about two-thirds of HIV-infected persons are unaware of their seropositive status [45]. In a systematic review [46], fewer than one-third of HIV-positive persons were retained in care between HIV testing and ART initiation. Furthermore, studies show high rates of loss to follow-up among patients starting ART [16]. Thus, PrEP could play an important additional role in controlling the HIV pandemic. Prioritized coverage with effective PrEP of individuals at highest risk of HIV acquisition and spread could potentially yield the optimal public health and cost benefits [40]. ART rollout is also limited by infrequent [18] access to second-line regimens and CD4 cell count, rather than virological monitoring [4]. As a result, there are high levels of drug resistance mutations among individuals with prolonged virological failure [22, 32], which may compromise both first-line [29, 47] and the limited second-line [48] ART regimens available. Our model shows that ART drives the prevalence of HIV drug resistance in both ART alone and ART + PrEP strategies. The principal determinants of the prevalence of acquired resistance include ART coverage, survival on ART with acquired resistance, and the rate of treatment failure. For the prevalence of transmitted resistance, determinants include the infectiousness of persons with acquired ART resistance and the persistence time of transmitted resistance. We find that PrEP is about 6- to 7-fold more efficient in HIV prevention than ART in terms of ratios of infections prevented to incident/prevalent drug-resistant cases generated. Thus, improving the effectiveness of first- and second-line ART is critical for preventing HIV infection and controlling drug resistance.
Our model projects a low prevalence of drug resistance from PrEP. Highly effective PrEP results in few breakthrough infections and a chance for emergence of acquired resistance. By contrast, poorly effective PrEP fails to protect from acquisition of wild-type HIV but also fails to exert selective pressure for emergence of acquired resistance. Both of these phenomena have been observed in recent PrEP trials [1, 2]. However, drug resistance from PrEP at the population level could rise with inappropriate PrEP use among those with undiagnosed HIV infection. While this increase is modest from inappropriate PrEP use during the preseroconversion phase of acute infection, it becomes more pronounced with inappropriate use among persons with established HIV. The latter may be of concern in potential situations of unsupervised PrEP use (eg, black-market drugs and drug sharing [49]) or inaccurate HIV testing [50].
There are some important limitations of our model. The accuracy of our predictions will be affected by variations in the model structure and sexual activity details, for which data are very limited. We therefore employed a well-established template of sexual behavior [40] with robust epidemiological and demographic parameterization, broadly applicable to South Africa. Nevertheless, the HIV epidemic in South Africa is heterogeneous and incompletely understood, with significant differences between the demographic and HIV/AIDS epidemiological estimates predicted by different agencies. HIV incidence is also not precisely known, even when measured directly at the population level. Although there is uncertainty regarding ARV-related parameters, we employed ranges (within plausible bounds) and performed extensive sensitivity and uncertainty analyses. We excluded population stratification by age and analysis of prioritized ARV implementation, as this was addressed in previous work [40]. Because of limited access to both second-line regimens [38] and drug-resistance testing [39] in resource-limited settings, we chose not to represent specific drug resistance mutations or second- or third-line ART regimens, nor do we consider HIV subtype polymorphism. We also did not explicitly include other influences on transmission. These and other refinements will be included in future work, although including such parameters greatly increases model complexity.
A key conclusion of this study is that combined ART + PrEP can have a greater public health impact than ART alone; however, overlapping ARVs for both can increase drug resistance in resource-limited settings. Drug resistance prevalence is predominantly driven by ART and not PrEP; consequently, nonoverlapping strategies will produce only modest declines in resistance. Thus, it is critical to consider the impact of ARVs not only on prevention but also drug resistance. Improved efficacy of first-line therapy and timely switching of ART to effective second-line regimens are critical for controlling HIV drug resistance. In addition, frequent and accurate HIV testing could minimize resistance consequences of PrEP. Our study also highlights that poor adherence to PrEP will undermine its potential impact on HIV prevention. Thus, prioritization of PrEP to groups at most risk of HIV acquisition and counseling about PrEP adherence are likely to maximize efficiency of PrEP and minimize drug resistance.

  iconpaperstack view older Articles   Back to Top