Secondary Logo

Journal Logo


Averted HIV infections due to expanded antiretroviral treatment eligibility offsets risk of transmitted drug resistance

a modeling study

Nichols, Brooke E.a; Sigaloff, Kim C.E.b,c; Kityo, Cissyd; Mandaliya, Kishore; Hamers, Raph L.b,c; Bertagnolio, Silviaf; Jordan, Michael R.g; Boucher, Charles A.B.a; Rinke de Wit, Tobias F.b,c; van de Vijver, David A.M.C.a

Author Information
doi: 10.1097/01.aids.0000433239.01611.52



Since 2010, the WHO recommends earlier initiation of combination antiretroviral therapy (ART) for HIV-infected persons in resource-limited countries. A shift in the immunological criteria for treatment initiation from below 200 to below 350 CD4+ cells/μl has resulted in a substantial increase in the number of people eligible to initiate ART [1]. In addition to the individual clinical benefits [2], earlier initiation of ART may reduce HIV incidence rates, and the concept of treatment as prevention has attracted attention as a means to reduce the global HIV epidemic [2].

As access to ART expands in resource-limited countries, concerns surrounding increasing numbers of patients failing treatment and the subsequent emergence of drug-resistant viruses will become increasingly important. Resistant virus selected for during treatment may subsequently be transmitted to newly infected individuals, undermining the effectiveness of currently recommended or available first-line ART [3]. A recent comprehensive assessment of transmitted drug resistance (TDR) reported a significant rise in prevalence of predominantly non-nucleoside reverse transcriptase inhibitor (NNRTI)-associated resistance mutations since ART rollout in east and southern Africa [4]. The highest TDR levels (9–13%) have been detected in East Africa [5,6]. As NNRTIs are the cornerstone of standard first-line ART in sub-Saharan Africa, reduced susceptibility to this drug class is especially worrisome in view of the limited availability of alternative first-line drug options.

Mathematical modeling is an important tool to inform policy makers about the potential consequences in terms of HIV drug resistance as a result of increased ART coverage in sub-Saharan Africa. Models of HIV transmission have been used to predict emergence of TDR [7–10], but these models have not examined the impact of earlier treatment initiation. In this study, we used a compartmental mathematical model to predict whether initiation of ART at different immunological thresholds and the availability of second-line ART have an effect on transmission of drug class-specific resistance in East Africa. Specifically, we examined whether the number of new HIV infections averted by early ART initiation offsets a potential rise in TDR.


Study design and population

To predict time trends of TDR, our model included resistance data from PharmAccess African Studies to Evaluate Resistance (PASER). This comprised two distinct observational studies on transmitted (PASER-Surveillance) and acquired (PASER-Monitoring) drug resistance conducted in Kampala, Uganda, and Mombasa, Kenya (Table 1). PASER-Monitoring used a prospective cohort that assessed HIV drug resistance in individuals about to initiate first-line ART in 2007–2008 and 24 months thereafter [11]. Viral load and drug resistance testing were not used to inform clinical decisions. PASER-Surveillance comprised two cross-sectional surveys among newly HIV-1-diagnosed, antietroviral-naive individuals attending voluntary counseling and testing sites in Kampala and Mombasa in 2009–2010 [5,6]. Specimens collected in the PASER studies were genotyped in two reference laboratories in Uganda, which participated in quality assessment schemes for genotypic drug resistance testing.

Table 1:
Characteristics of patients included in the PASER studies.

Model and calibration

A compartmental deterministic mathematical model was constructed which was described with a total of 77 ordinary differential equations and 123 parameters (Supplemental Digital Content 1,, figure of model structure, and Supplemental Digital Content 2,, full table of parameters). The model stratified disease progression into an acute stage, three chronic stages, and two AIDS stages. Three chronic stages were included to indicate previous (ART initiation at CD4+ cell count <200 cells/μl [12]) [1], current (CD4+ cell count <350 cells/μl) [1], and potential future treatment guidelines (CD4+ cell count <500 cells/μl) [13]. Infectivity varied by stage of infection [14,15]. Other key model parameters are summarized in Table 2. TDR was defined as the number of individuals infected with drug-resistant HIV over total number of HIV-infected, treatment-naive individuals. In the Kampala model, monotherapy with zidovudine was available to a small number of individuals from 1991 to 1996 and dual therapy with zidovudine/lamivudine was available to a small number of individuals from 1996 to 2000. Triple therapy started rollout in the model on a small scale from the year 2000 in both Kampala and Mombasa.

Table 2:
Key additional model parametersa.

The model identified four sexual activity groups ranging in the number of new sexual partners per year [22]. Using Monte Carlo filtering techniques [23], we parameterized the different sexual activity groups and only accepted the simulations that were associated with distinct HIV prevalence from country data [24,25] and TDR prevalence by resistance class in the two PASER-Surveillance sites. For Kampala, this resulted in 1017 out of 50 000 simulations with HIV prevalence between 7.1 and 8.4% between 2005 and 2009 and TDR prevalence between 7.1 and 10% in 2009. For Mombasa, this resulted in 1247 out of 50 000 simulations with an HIV prevalence of 5.8–8.2% between 2006 and 2010, and TDR prevalence between 11.9 and 14.9% in 2009 (see graphs, Supplemental Digital Content 3,, which show the model calibration curves).

Sensitivity analysis

Sensitivity analyses using recursive partitioning [26,27] were conducted to determine the most influential parameters on both TDR prevalence and number of acute infections in Kampala and Mombasa, respectively (see figures, Supplemental Digital Content 4,, which show the recursive partitioning analysis and further description).

Data and additional parameters

The PASER-Monitoring data from antiretroviral-naive patients about to start HIV treatment were used as parameters in the model regarding regimens being prescribed and resistance patterns. In 2008–2009, approximately 40% of individuals from PASER-Monitoring in Kampala were receiving tenofovir-containing regimens, and 60% zidovudine-containing regimens, both combined with emtricitabine or lamivudine, and efavirenz or nevirapine. In Mombasa, just 1% of individuals were on tenofovir-containing regimens, and all others were on zidovudine or stavudine-containing regimens (Table 1). We assumed that stavudine would be phased out and replaced by zidovudine in Mombasa. We also evaluated the impact of instead replacing stavudine by tenofovir in a sensitivity analysis. Specific drug-resistance mutations were assumed to be selected for by tenofovir (i.e. K65R) and zidovudine [i.e. thymidine analog mutations (TAMs)], for lamivudine and emtricitabine (i.e. M184V), as well as efavirenz and nevirapine (i.e. NNRTI-specific mutations). Individuals were classified as receiving either a zidovudine-containing or tenofovir-containing regimens, and within these regimens, there were different probabilities of acquiring the signature mutations of zidovudine/tenofovir, as well as the M184V or NNRTI-specific mutations. The model also included second-line treatment which consisted of a ritonavir-boosted protease inhibitor in combination with two nucleoside reverse transcriptase inhibitors (NRTIs). Each different regimen (zidovudine, tenofovir, or protease inhibitor-based) was assumed to have a different likelihood of transmitting acquired drug-resistant mutations onwards to a susceptible individual [28]. In our model, patients with an acquired drug-resistant virus could either pass on an NRTI mutation (divided further into a TAM, M184V, or K65R), NNRTI mutation, or protease inhibitor mutation. Drug-resistance mutations have a reduced fitness as compared to a wild-type virus that is susceptible to antiretroviral drugs for those with an acquired resistance mutation [29]. The fitness cost can result in a lower viral load. Because viral load is the key parameter explaining transmission [30], drug-resistant viruses are less easily transmitted. A specific fitness cost for K65R, protease inhibitor, and TAMs was estimated by reductions in replication capacity in the literature [28,31]. The fitness cost for the M184V and NNRTI mutations was estimated directly from PASER data as there were a sufficient number of patients who acquired these mutations while on treatment. We calculated the fitness cost using a published formula by taking the difference between baseline viral load and viral load after treatment failure with the respective resistance mutation [32]. We then took the intraquartile range of the fitness costs from the transformed PASER data as the parameter values (see table, Supplemental Digital Content 5,, which shows the proportions of resistance mutations coming from each drug regimen and the mutation fitness costs). The estimates from the PASER data were in line with the literature [33,34]. Reversion to wild type after infection with a drug-resistant virus was also considered (see table, Supplemental Digital Content 6,, which shows the rates of reversion to wild-type HIV-1 after being infected with a drug-resistant HIV virus).

As a baseline scenario, we calculated the TDR prevalence for ART initiation at less than 200 CD4+ cells/μl (including 20% of patients with CD4+ cell count 200–350 cells/μl to represent the current treatment situation). We then predicted the effect on TDR prevalence of ART initiation at CD4+ cell count less than 350 and 500 cells/μl. The total number of people accessing treatment depends on the HIV testing and retention rate, both of which were calibrated in the model. We also investigated the impact of increasing access to second-line therapy. In the model, we assume that second-line therapy is only limitedly available. In accordance with PASER data, the proportions of people who switch to second-line therapy after continued virological failure in Kampala and Mombasa is 33–66 and 15–30%, respectively, during the first 2 years on therapy, as is assumed to not be available thereafter. We expect an increased availability of second-line regimens and viral load testing in the future. Accordingly, we increased the proportions of people who switch to second-line therapy after continued virological failure in Kampala and Mombasa up to 80–100% for the entire duration of ART. We then determined the overall and mutation-specific TDR prevalence over 10 years for all immunological ART initiation cut-offs. Full model description including equations can be found in the Text of Supplemental Digital Content 7 (

In order to investigate whether the prevention of new infections by earlier initiation of ART offsets rising TDR, we calculated the total number of new HIV infections averted for each additional case infected with drug-resistant HIV.


Impact of antiretroviral therapy on the HIV epidemic

Figure 1a shows the effect of earlier first-line ART initiation on estimated HIV prevalence in Kampala and Mombasa. Although the HIV prevalence remains relatively stable when treatment is initiated at less than 200 or 350 CD4+ cells/μl due to reduced mortality of infected individuals, a decline in HIV prevalence is expected when treatment is initiated at less than 500 CD4+ cells/μl. HIV incidence is expected to drop in both cities when treatment is initiated at below 350 and 500 CD4+ cells/μl (Fig. 1b). The decrease in HIV incidence is also reflected in the proportion of infections that can be averted at particular immunological thresholds of ART initiation. Compared to initiating ART at CD4+ cell count below 200 cells/μl, initiating ART at CD4+ cell count below 350 cells/μl averts a median of 12.6% [interquartile range (IQR) 11.3–13.7%] of infections over 10 years in Kampala and averts a median of 11.6% (IQR 10.3–13.0%) of infections in Mombasa. Initiating ART at CD4+ cell count below 500 cells/μl averts a median of 28.8% (IQR 26.0–31.4%) and 26.3% (IQR 23.2–29.5%) in Kampala and Mombasa, respectively.

Fig. 1:
Yearly median of HIV prevalence, incidence, and overall transmitted drug resistance prevalence from 2012 to 2022 when initiating treatment at CD4+ cell count below 200, 350, and 500 cells/μl in Kampala, Uganda, and Mombasa, Kenya.

Prevention of new infections and increase of transmitted drug resistance

When treatment is initiated at CD4+ cell count below 350 cells/μl, a median of 18 (IQR 11–31) infections in Kampala and 46 (IQR 30–83) infections in Mombasa will be averted for every additional case infected with drug-resistant virus. Similarly, when treatment is initiated at CD4+ cell count below 500 cells/μl, the estimated number of infections averted per additional case of TDR is a median of 22 (IQR 17–35) in Kampala and 32 (IQR 21–57) in Mombasa. The larger number of infections averted in Mombasa as compared to Kampala is in line with the smaller TDR increase predicted in Mombasa.

Evolution of overall transmitted drug resistance prevalence

Figure 1c shows that expanding access to ART by initiating treatment at higher CD4+cell counts is expected to increase TDR prevalence in both Kampala and Mombasa. Between 2012 and 2022, the estimated TDR prevalence in Kampala increases from a median of 8.3% (IQR 7.7–9.0%) to a median of 9.4% (IQR 8.4–10.5%), 12.2% (IQR 10.9–13.8%), and 19.0% (IQR 16.5–21.8%) when initiating ART at CD4+ cell count below 200, below 350 or below 500 cells/μl, respectively. During the same period, the estimated TDR prevalence in Mombasa remains a median of 12.3% (IQR 11.7–13.1%) when starting ART at CD4+ cell count below 200 cells/μl, but increases to a median of 13.6% (IQR 12.5–14.9%) when starting at CD4+ cell count below 350 cells/μl and to a median of 19.2% (IQR 17.1–21.5%) when starting at CD4+ cell count below 500 cells/μl.

Drug resistance by drug class and mutation

In both settings, current TDR is predominantly characterized by resistance to NNRTIs. According to our modeling results, NNRTI mutations are predicted to continue to comprise the majority of the future prevalence of TDR (Fig. 2a and b). In Kampala, the prevalence of transmitted NNRTI resistance is estimated to rise over the next 10 years from a median of 4.4% (IQR 3.8–4.9%) to 6.9% (IQR 6.0–7.9%) when initiating at CD4+ cell count below 200 cells/μl. An increase to a median of 9.1% (IQR 7.8–10.5%) when initiating treatment at CD4+ cell count below 350 cells/μl is predicted, and to 14.3% (IQR 11.9–16.6%) when initiating at CD4 below 500 cells/μl. In Mombasa, NNRTI resistance is estimated to remain stable (median 7.1%; IQR 6.3–7.7% in 2012 to median 6.9%; IQR 6.0–7.9% in 2022) when initiating ART at CD4+ cell count below 200 cells/μl, but to increase to a median of 8.7% (IQR 7.6–9.7%) when initiating at a CD4+ cell count of below 350 cells/μl and to 12.2% (IQR 10.9–13.7%) when starting at below 500 cells/μl. With respect to the NRTI class, in Kampala, estimated prevalence of TAMs decreases between 2012 and 2022, regardless of the immunological threshold used. In contrast, the TAM prevalence in Mombasa is estimated to slightly increase over the same period. A sensitivity analysis in which stavudine was replaced by tenofovir instead of zidovudine yielded similar results (see graph, Supplemental Digital Content 8, for the results of this sensitivity analysis). Transmitted M184V increases in both Kampala and Mombasa when treatment is initiated at CD4+ cell count below 500 cells/μl, although the prevalence was predicted to remain less than 2% in both areas. In both cities, protease inhibitor resistance decreases over time, irrespective of immunological threshold used to initiate therapy. Transmitted K65R will increase slightly at all immunological thresholds but will remain less than 1% between 2012 and 2022 in both settings.

Fig. 2:
Yearly median of transmitted drugresistant mutation prevalence by mutation, from 2012 to 2022, when starting treatment at CD4 + cell count less than 200, 350, and 500 cells/μl in Kampala, Uganda, and Mombasa, Kenya.

Increasing access to second-line treatment

If access to virological monitoring and second-line treatment increases in parallel to the further scale-up of first-line ART, TDR is not expected to increase at any of the immunological thresholds used for ART initiation. In the scenario where 80–100% of individuals with prolonged virological failure are appropriately switched to second-line boosted protease inhibitor therapy, the level of TDR is expected to decline in both locations at all immunological thresholds to below the 2012 TDR level. Increasing access to boosted protease inhibitors will either reduce or stabilize TDR for all drug classes (Fig. 3a–d).

Fig. 3:
Yearly median of transmitted drug resistant mutation prevalence by mutation when increasing access to second-line treatment for those with continued virological failure to 80–100% from 2012 to 2022 when starting treatment at CD4 + cell count below 200, 350, and 500 cells/μl in Kampala, Uganda, and Mombasa, Kenya.

Sensitivity analysis

In the models for both Kampala and Mombasa, higher test rate (greater than 22.5 and 21.4% respectively) was the strongest predictor for a reduction in new infections. This is likely due to the fact that more individuals will get into care sooner, and thus spend a greater amount of time with a suppressed virus. In both Kampala and Mombasa, transmitted drug resistance depends most strongly on the rate of reversion of a drug-resistant mutation in a treatment-naive individual to a wild-type virus. This is because if revertancy is slower, a patient is more likely to infect another person with a resistant virus instead of a wild-type virus. (See Figure, Supplemental Digital Content 4,, which shows full recursive partitioning results.)


We have modeled the impact of ART initiation at different immunological thresholds on the prevalence of TDR in two East African settings over the next 10 years. This is the first model to show that averted HIV infections due to the expansion of ART eligibility offset the risk of increased TDR. We predict that the number of infections that will be averted by earlier ART initiation will far exceed the number of infections with a drug-resistant virus. When the current WHO treatment guidelines of ART initiation at CD4+ cell count below 350 cells/μl are fully implemented in these two settings, TDR prevalence is expected to increase slightly. Expanding treatment by initiating ART at CD4+ cell count below 500 cells/μl will lead to an increasing TDR prevalence. TDR mutations associated with the NNRTI drug class, the cornerstone of current first-line regimens [35], are expected to drive the TDR rise. Importantly, if switches from first to second-line treatment occur for all patients necessitating a switch, then the overall TDR prevalence, including NNRTI resistance, will decrease in the next 10 years. This implies that wider access to virological monitoring and boosted protease inhibitors for second-line therapy will preserve the effectiveness of NNRTI-based first-line treatment in all scenarios.

Current and predicted TDR is predominantly due to NNRTI resistance. This finding corroborates a recent meta-analysis of TDR which estimated an increase of predominantly NNRTI-related resistance of 29% per year in East Africa [4]. This can be explained by the low genetic barrier of NNRTIs for HIV drug resistance as only a single amino acid substitution is sufficient for high-level resistance [36]. In addition, transmitted NNRTI-associated mutations persist for a prolonged period of time [37]. Transmission of NNRTI resistance can have important clinical ramifications as their presence is associated with an increased risk of virological failure of standard first-line treatment and for further selection of drug resistance after treatment initiation [3]. Our modeling study suggests that this can be prevented by increasing access to boosted protease inhibitors in second-line ART. Timely switches to second-line regimens are only possible when routine virological monitoring, that is, at 6 or 12-monthly intervals, is implemented. In agreement, a recent model of HIV transmission predicted that routine virological monitoring in patients on ART can reduce TDR [7]. For this purpose, cheap point-of-care viral load assays should be developed [4,7].

The analysis models the changes in treatment initiation guidelines, and thus even when the treatment initiation threshold changes, many individuals still initiate therapy late in infection. When treatment is initiated at CD4+ cell count below 500 cells/μl, not all individuals initiate treatment early due to test rates and retention of the respective settings. In the model, once treatment at CD4+ cell count below 500 cells/μl is fully scaled up in 2013, 43% of individuals initiate treatment between CD4+ 350 and 500 cells/μl, 38% initiate between CD4+ 200 and 350 cells/μl, and 19% initiate when CD4+ cell count is below 200 cells/μl. In 2022, 49% of individuals will initiate between 350 and 500 cells/μl, 34% will initiate between 200 and 350 cells/μl, and 17% will initiate at CD4+ cell count below 200 cells/μl.

Our mathematical model has several strengths. To our knowledge, our model is the first to examine the impact of initiating ART at different immunological thresholds, including at CD4+ cell count less than 500 cells/μl, on the prevalence of TDR in sub-Saharan Africa. This is particularly relevant in light of increased interest for early initiation of ART as a means to prevent new infections. We have demonstrated that the number of new infections prevented by earlier ART initiation far outweighs the expected number of infections with drug-resistant virus. Second, our analysis predicts future levels of HIV drug resistance in sub-Saharan Africa in terms of the presence of specific mutations to particular drugs, accounting for variation in transmissibility between individual mutations. A small number of mathematical HIV transmission models have examined the impact of antiretroviral drugs on transmitted drug resistance in sub-Saharan Africa [7–10]. Almost all previous models used overall TDR rates to describe HIV drug resistance [8–10]. One model used a classification of HIV drug resistance similar to ours, but only reported overall TDR [7]. Lastly, this model combines data on transmitted and acquired HIV drug resistance from the same geographic areas and time period, collected within the same research project. This is important as resistance acquired during treatment constitutes the pool of variants that can be transmitted within a population. The TDR predictions were largely similar for the two distinct geographic settings even though these two settings have a different history of antiretroviral roll-out. In Uganda, antiretroviral drugs became available at least 5 years ahead of neighboring countries, including limited-scale distribution of mono and dual NRTI-based therapies [38]. This may account for the higher initial rate of TAMs observed in Kampala, but our analysis shows that it is unlikely to impact future rates of TDR.

The study has some potential limitations. First, the data on acquired resistance do not exceed 24 months of follow-up. Data on HIV drug resistance beyond 24 months of ART in resource-limited settings are scarce. Nonetheless, the PASER-Monitoring study provides the most accurate empirical data on acquired resistance patterns currently available in Africa. Second, we did not incorporate the type of ART monitoring to guide switching, that is, clinical, immunological, or virological, as a variable in our model. Instead, based on PASER-Monitoring data we noted that in Kampala 33–66% and in Mombasa 15–30% of patients with virological failure were appropriately switched to second-line regimen during the first 2 years of HIV treatment. Third, baseline HIV test rates were assumed to be 10–30% of the populations. Increasing HIV testing uptake is likely to lead to greater numbers of people initiating ART at higher CD4+ cell counts, although this poses logistical challenges for implementation. Fourth, we assumed that drug regimen use would remain constant for the next 10 years, as it is difficult to make predictions about future drug substitutions. We did, however, account for the fact that stavudine is likely to be phased out. We assumed that stavudine would be replaced by zidovudine in Mombasa, or instead by tenofovir in a sensitivity analysis, with comparable results. Finally, costs are not included in this analysis as we do not have comprehensive costing data for these study sites. Second-line treatment is expensive and using second-line treatment to limit drug resistance can increase costs. Conversely, second-line treatment can also result in reduced HIV transmission (as less resistance emerges) which can be cost-saving due to the substantial lifetime costs associated with a new infection.

To avert infections is a commonly used metric in modeling studies to quantify the impact of an intervention on a population. The ratio of infections averted to TDR gained has been recently described [39], and can quantify infections averted versus gains in TDR. These metrics cannot be validated in large prospective cohort studies, as it is not known how many infections would have occurred in absence of an intervention. The ratio of infections averted to transmitted drug resistance gained does not, however, take into account all potential benefits or detriments of increased treatment. One potential detriment could be higher mortality of those individuals with resistance to first-line or to second-line treatment. Data from resource-rich settings show that there is limited impact of drug resistance on mortality [40,41]. Although no data are available, resistance could increase mortality in resource-limited settings as treatment options are limited.

With regard to our modeling strategy, we assumed that individuals could only be infected with a wild-type virus or a virus containing a single class of resistance mutation, as the PASER-Survelliance data from Kampala and Mombasa showed this. Whereas most transmitted drug resistance is transmission of a virus containing a single resistance mutation, it has been described that in approximately 20% of the time, a virus containing more than one resistance mutation is transmitted. But, importantly, this usually involves mutations from the same class. Transmission of resistance involving more than one class is rare in sub-Saharan Africa [42]. Our assumption that only a single mutation can be transmitted, therefore, is in agreement with epidemiological studies on transmission of drug-resistant HIV [42,43]. In our model, we also assumed a closed population for both cities. We have calibrated the model to the closed population, thus our predictions on the closed populations should be accurate. For this analysis, we do not calculate the number of deaths averted or person-years lived, thus our results should not be influenced by this assumption.

In conclusion, expansion of ART eligibility will lead to increased TDR, but this is not expected to offset the preventive benefit of controlling the HIV epidemic. Transmitted NNRTI resistance can potentially have a profound impact on the effectiveness of first-line treatment. Importantly, we have demonstrated that further increase of NNRTI resistance can be avoided by increasing access to second-line boosted protease inhibitor regimens. Therefore, as ART is scaled up, efforts should be made to make virological monitoring and effective second-line therapy available. Transmitted drug resistance should not be a reason to withhold early initiation of ART as averted HIV infections due to expanded treatment eligibility are predicted to offset the risk of increased transmitted resistance.


The authors are grateful to all participants of the PASER studies and support staff at JCRC, ICRH, PharmAccess Foundation, and AIGHD. PASER was part of the LAASER program (Linking African and Asian Societies for an Enhanced Response to HIV/AIDS), a partnership of Stichting Aids Fonds, The Foundation for AIDS Research (amfAR) – TREAT Asia, PharmAccess Foundation and International Civil Society Support.

Contributors: Dv.V., B.E.N., and K.C.E.S. conceived the study. T.R.W., C.K., and K.M. are principle investigators, and R.L.H. and K.C.E.S. are co-investigators of PASER. B.E.N. programmed the computer simulation model. K.C.E.S. performed descriptive analysis of the PASER-S and PASER-M data. K.C.E.S., B.E.N., and Dv.V. wrote the first draft of the manuscript and C.K., K.M., R.L.H., M.R.J., S.B., T.R.W., and C.A.B. contributed to data interpretation and development of the manuscript. All authors contributed to subsequent drafts and reviewed and approved the final manuscript.

Sources of support: This work was supported by the Aids Fonds Netherlands (2010–035); European Union FP7 CHAIN grant (223131); European Union FP7 DynaNets grant (233847); and National Institutes of Health (K23 AI074423–05, M.R.J.).

Conflicts of interest

S.B. is a staff member at WHO, but the views expressed in this paper do not necessarily represent the decisions or stated policies of WHO. The authors have declared that no competing interests exist.


1. WHOAntiretroviral therapy for HIV infection in adults and adolescents: recommendations for a public health approach: 2010 rev. Geneva:World Health Organization; 2010.
2. Cohen MS, Chen YQ, McCauley M, Gamble T, Hosseinipour MC, Kumarasamy N, et al. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med 2011; 365:493–505.
3. Hamers RL, Schuurman R, Sigaloff KC, Wallis CL, Kityo C, Siwale M, et al. Effect of pretreatment HIV-1 drug resistance on immunological, virological, and drug-resistance outcomes of first-line antiretroviral treatment in sub-Saharan Africa: a multicentre cohort study. Lancet Infect Dis 2012; 12:307–317.
4. Gupta RK, Jordan MR, Sultan BJ, Hill A, Davis DH, Gregson J, et al. Global trends in antiretroviral resistance in treatment-naive individuals with HIV after rollout of antiretroviral treatment in resource-limited settings: a global collaborative study and meta-regression analysis. Lancet 2012; 380:1250–1258.
5. Ndembi N, Hamers RL, Sigaloff KC, Lyagoba F, Magambo B, Nanteza B, et al. Transmitted antiretroviral drug resistance among newly HIV-1 diagnosed young individuals in Kampala. AIDS 2011; 25:905–910.
6. Sigaloff KC, Mandaliya K, Hamers RL, Otieno F, Jao IM, Lyagoba F, et al. Short communication: high prevalence of transmitted antiretroviral drug resistance among newly HIV type 1 diagnosed adults in Mombasa, Kenya. AIDS Res Hum Retroviruses 2012; 28:833–837.
7. Phillips AN, Pillay D, Garnett G, Bennett D, Vitoria M, Cambiano V, Lundgren J. Effect on transmission of HIV-1 resistance of timing of implementation of viral load monitoring to determine switches from first to second-line antiretroviral regimens in resource-limited settings. AIDS 2011; 25:843–850.
8. Baggaley RF, Garnett GP, Ferguson NM. Modelling the impact of antiretroviral use in resource-poor settings. PLoS Med 2006; 3:e124.
9. Wilson DP, Kahn J, Blower SM. Predicting the epidemiological impact of antiretroviral allocation strategies in KwaZulu-Natal: the effect of the urban-rural divide. Proc Natl Acad Sci U S A 2006; 103:14228–14233.
10. Vardavas R, Blower S. The emergence of HIV transmitted resistance in Botswana: ‘when will the WHO detection threshold be exceeded?’. PLoS One 2007; 2:e152.
11. Hamers RL, Wallis CL, Kityo C, Siwale M, Mandaliya K, Conradie F, et al. HIV-1 drug resistance in antiretroviral-naive individuals in sub-Saharan Africa after rollout of antiretroviral therapy: a multicentre observational study. Lancet Infect Dis 2011; 11:750–759.
12. WHOAntiretroviral therapy for HIV infection in adults and adolescents: recommendations for a public health approach, 2006 revision. 2006; Geneva, Switzerland:World Health Organization, (
13. DHHS. PanelDHHS on antiretroviral guidelines for adults and adolescents. Guidelines for the use of antiretroviral agents in HIV-1-infected adults and adolescents. Department of Health and Human Services 2012; [Accessed 17 July 2012]
14. Hollingsworth TD, Anderson RM, Fraser C. HIV-1 transmission, by stage of infection. J Infect Dis 2008; 198:687–693.
15. Ruark A, Shelton JD, Halperin DT, Wawer MJ, Gray RH. Universal voluntary HIV testing and immediate antiretroviral therapy. Lancet 2009; 373:1078[author reply 1080–1071].
16. Pilcher CD, Joaki G, Hoffman IF, Martinson FE, Mapanje C, Stewart PW, et al. Amplified transmission of HIV-1: comparison of HIV-1 concentrations in semen and blood during acute and chronic infection. AIDS 2007; 21:1723–1730.
17. Lodi S, Phillips A, Touloumi G, Geskus R, Meyer L, Thiebaut R, et al. Time from human immunodeficiency virus seroconversion to reaching CD4+ cell count thresholds <200, <350, and <500 Cells/mm(3): assessment of need following changes in treatment guidelines. Clin Infect Dis 2011; 53:817–825.
18. Brinkhof MW, Boulle A, Weigel R, Messou E, Mathers C, Orrell C, et al. Mortality of HIV-infected patients starting antiretroviral therapy in sub-Saharan Africa: comparison with HIV-unrelated mortality. PLoS Med 2009; 6:e1000066.
19. Donnell D, Baeten JM, Kiarie J, Thomas KK, Stevens W, Cohen CR, et al. Heterosexual HIV-1 transmission after initiation of antiretroviral therapy: a prospective cohort analysis. Lancet 2010; 375:2092–2098.
20. Reynolds SJ, Makumbi F, Nakigozi G, Kagaayi J, Gray RH, Wawer M, et al. HIV-1 transmission among HIV-1 discordant couples before and after the introduction of antiretroviral therapy. AIDS 2011; 25:473–477.
21. Landier J, Akonde A, Pizzocolo C, Haidara I, Drabo M, Pizarro L, et al. Switch to second-line ART in West African routine care: incidence and reasons for switching. AIDS Care 2011; 23:75–78.
22. Garnett GP, Anderson RM. Factors controlling the spread of HIV in heterosexual communities in developing countries: patterns of mixing between different age and sexual activity classes. Philos Trans R Soc Lond B Biol Sci 1993; 342:137–159.
23. Rose KA, Smith E, Gardner R, Brenkert A, Bartell S. Parameter sensitivities, Monte Carlo Filtering, and model forecasting under uncertainty. J Forecast 1991; 10:117–133.
24. STD/AIDS Control Programme. The HIV/AIDS Epidemiological Surveillance Report 2010. Kampala, Uganda: Ministry of Health; 2010.
25. National AIDS and STI Control Program. Sentinel surveillance for HIV and Syphilis among pregnant women, 2010. Nairobi: NASCOP; 2010.
26. Venables WN, Ripley BD. Venables WN, Ripley BD. Chapter 9: tree-based methods. Modern Applied statistics with S Springer, 4th ed.New York:2002.
27. Mayo Foundation, Therneau TM, Atkinson EJ. An introduction to recursive partitioning using the RPART routines. 1997; 251, 258.
28. De Luca A. Geretti AM. The impact of resistance on viral fitness and its clinical implications. Antiretroviral resistance in clinical practice. London:Mediscript; 2006; Chapter 12.
29. Nijhuis M, Deeks S, Boucher C. Implications of antiretroviral resistance on viral fitness. Curr Opin Infect Dis 2001; 14:23–28.
30. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li C, Wabwire-Mangen F, et al. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. N Engl J Med 2000; 342:921–929.
31. Xu HT, Martinez-Cajas JL, Ntemgwa ML, Coutsinos D, Frankel FA, Brenner BG, Wainberg MA. Effects of the K65R and K65R/M184V reverse transcriptase mutations in subtype C HIV on enzyme function and drug resistance. Retrovirology 2009; 6:14.
32. Smith RJ, Okano JT, Kahn JS, Bodine EN, Blower S. Evolutionary dynamics of complex networks of HIV drug-resistant strains: the case of San Francisco. Science 2010; 327:697–701.
33. Cong ME, Heneine W, Garcia-Lerma JG. The fitness cost of mutations associated with human immunodeficiency virus type 1 drug resistance is modulated by mutational interactions. J Virol 2007; 81:3037–3041.
34. Armstrong KL, Lee TH, Essex M. Replicative fitness costs of nonnucleoside reverse transcriptase inhibitor drug resistance mutations on HIV subtype C. Antimicrob Agents Chemother 2011; 55:2146–2153.
35. WHORapid advice: antiretroviral therapy for HIV infection in adults and adolescents. Geneva, Switzerland:World Health Organization; 2009.
36. van de Vijver DA, Wensing AM, Angarano G, Asjo B, Balotta C, Boeri E, et al. The calculated genetic barrier for antiretroviral drug resistance substitutions is largely similar for different HIV-1 subtypes. J Acquir Immune Defic Syndr 2006; 41:352–360.
37. Jain V, Sucupira MC, Bacchetti P, Hartogensis W, Diaz RS, Kallas EG, et al. Differential persistence of transmitted HIV-1 drug resistance mutation classes. J Infect Dis 2011; 203:1174–1181.
38. Weidle PJ, Malamba S, Mwebaze R, Sozi C, Rukundo G, Downing R, et al. Assessment of a pilot antiretroviral drug therapy programme in Uganda: patients’ response, survival, and drug resistance. Lancet 2002; 360:34–40.
39. Abbas UL, Glaubius R, Mubayi A, Hood G, Mellors JW. Antiretroviral therapy and preexposure prophylaxis: combined impact on HIV-1 transmission and drug resistance in South Africa. J Infect Dis 2013; 208:224–234.
40. Sabin CA, Smith CJ, Youle M, Lampe FC, Bell DR, Puradiredja D, et al. Deaths in the era of HAART: contribution of late presentation, treatment exposure, resistance and abnormal laboratory markers. AIDS 2006; 20:67–71.
41. Hogg RS, Bangsberg DR, Lima VD, Alexander C, Bonner S, Yip B, et al. Emergence of drug resistance is associated with an increased risk of death among patients first starting HAART. PLoS Med 2006; 3:e356.
42. UK National Guidelines for HIV Testing 2008. British HIV Association; 2008.
43. Nichols BE, Boucher CA, van de Vijver DA. HIV testing and antiretroviral treatment strategies for prevention of HIV infection: impact on antiretroviral drug resistance. J Intern Med 2011; 270:532–549.

antiretroviral treatment; mathematical modeling; primary HIV-1 drug resistance; sub-Saharan Africa

Supplemental Digital Content

© 2014 Lippincott Williams & Wilkins, Inc.