The antiretroviral rollout and drug-resistant HIV in Africa: insights from empirical data and theoretical models

The U.S. Government has pledged to spend $15 billion in Africa and the Caribbean on AIDS. A central focus of this plan is to provide antiretroviral treatment (ART) to millions. Here, we evaluate whether the plan to rollout ART in Africa is likely to generate an epidemic of drug-resistant strains of HIV. We review what has occurred as a result of high usage of ART in developed countries in terms of changes in risky behavior, and the emergence and transmission of drug-resistant HIV. We also review how mathematical models have been used to predict the evolution of drug-resistant HIV epidemics. We then show how models can be used to predict the likely impact of the ART rollout on the evolution of drug-resistant HIV in Africa. At currently planned levels of treatment coverage, we predict that (over the next decade) in Africa: (i) the impact of ART on reducing HIV transmission (and prevalence) is likely to be undetectable (unless accompanied by substantial changes in behavior), (ii) the transmission rate of drug-resistant HIV will be below the WHO surveillance threshold of 5%, and (ii) the majority of cases of drug-resistant HIV that will occur will be due to acquired (and not transmitted) resistance. For the next decade, large-scale surveillance for detecting transmitted resistance in Africa is unnecessary. Instead, we recommend that patients should be closely monitored for acquired resistance, and sentinel surveillance (in a few urban centers) should be used to monitor transmitted resistance.

Introduction

The treatment of individuals with HIV infection has reduced the rate of AIDS complications, dramatically limited the vertical spread of HIV from mother to infant, and reduced mortality to such an extent that effective treatment is now viewed as a governmental mandate and a human right ^[1-3]. The US government has recently pledged to spend US$15 billion in Africa and the Caribbean on AIDS prevention and care. A central focus of this plan is to provide antiretroviral treatment to millions of HIV-infected individuals. A major concern that has been raised regarding the plan to roll out antiretroviral drugs in Africa is that it may potentially generate an epidemic of drug-resistant strains of HIV. Here we review what has occurred as a result of high levels of usage of these therapies in developed countries in terms of: (i) changes in risky behavior, and (ii) the emergence and transmission of drug-resistant HIV. Mathematical models, when coupled with uncertainty analyses ^[4,5] have been shown to be useful health policy tools for predicting the future impact of antiretroviral therapies ^[6-11] and HIV vaccines ^[12,13]. Here we discuss how models have been used for providing insights into, and predicting the evolution of, epidemics of drug-resistant HIV. We then use the theoretical models (coupled with uncertainty analysis) to predict the likely impact of antiretroviral rollout on drug-resistant HIV in Africa.

Antiretroviral drugs reduce the infectivity of treated individuals by suppression of the viral load ^[14]. It has been estimated by an analysis of longitudinal cohort data that antiretroviral therapy reduces per-partnership infectivity by as much as 60% ^[15]. Mathematical models have been used to quantify, and to predict, the epidemic-level effect of antiretroviral drugs ^{[6-11,13,16-20]}. These models are specified in terms of mathematical equations, and are often numerically analysed using computers. They provide a theoretical framework for tracking simultaneously the transmission of wild-type drug-sensitive and drug-resistant strains of HIV; the models also include the emergence of drug-resistant strains during treatment, as a result of a variety of patient, drug and treatment factors. The exact values of many of the biological parameters in these models are often only imprecisely known; for example, the fitness of drug-resistant strains. Therefore, these models can be used as predictive tools only if they are analysed using uncertainty analysis based upon Monte Carlo sampling techniques ^[4,5,21]. Accurately specified point estimates of parameters are unnecessary to conduct an uncertainty analysis; only parameter ranges need to be defined. Uncertainty analysis has enabled models to be used as predictive tools, with the predictions presented with 'uncertainty bars' ^[7-13]. Multivariate sensitivity analyses (using non-parametric statistics) have been used to analyse the data generated by the models in order to identify the key factors in decreasing mortality, reducing overall transmission rates, and increasing the transmission and prevalence of drug-resistant strains ^[4,5]. To date, much of the modeling analyses of the impact of antiretroviral therapy on generating antiretroviral resistance have been focused on developed countries, where the usage rate of antiretroviral drugs has been very high. The HIV epidemic in the gay community in San Francisco, where 50-90% of the HIV-infected individuals have received antiretroviral therapy, has been extensively modeled ^[6-8,11,18]. However, only a few studies have modeled the potential impact of antiretroviral therapy in developing countries ^[9,16]. Such studies have shown that the level of transmitted resistance will be very dependent upon the value of the relative fitness of the drug-resistant strains that evolve, and will increase as the proportion of the HIV-infected population that receive treatment increases ^[7,9,11,16].

Antiretroviral therapy and changes in risky behavior

Mathematical modeling of the potential epidemic impact of antiretroviral therapy has shown that a high usage of these therapies could be expected to lead to substantial declines in mortality rates, incidence rates, and (over the longer term) prevalence ^[6,16]. The amount by which the mortality rate, incidence rate, and prevalence level would be reduced has been shown to be directly dependent upon the proportion of HIV-infected patients who receive antiretroviral therapies ^{[6-9,13,16,17]}. The theoretical studies have shown that if high usage rates of antiretroviral therapy are accompanied by reductions in risky behavior then the beneficial effects on reducing HIV epidemics would be substantial, and that (under certain specified conditions) it would even be possible to eradicate high prevalence HIV epidemics using current therapies ^[8]. Modeling analyses have also shown that if risky behavior increases then even a high usage of antiretroviral therapy would not lead to a reduction in incidence and prevalence ^[6,8,9]. Increases in risky behavior will mask antiretroviral drug-induced decreases in transmission ^[6,8,9]. Models have been used to quantify the trade-off between increases in risky behavior and the usage of antiretroviral therapy ^[9], see Fig. 1a. Transmission may be significantly reduced, but the incidence rate may not decrease; the percentage of the HIV-infected population that are on treatment and the average increase in risky behavior will determine whether HIV incidence rates either increase, decrease or stabilize (Fig. 1a). It is noteworthy that if 50% of HIV-infected individuals receive antiretroviral therapy then an increase in risky behavior of greater than 20% will lead to increases in the incidence (Fig. 1a). If risky behavior increases then the prevalence will substantially increase over a period of a decade; the degree to which prevalence will increase will depend upon the magnitude of the increase in risky behavior (Fig. 1b).

Fig. 1

It has been suggested that the rollout of antiretroviral agents in Africa may lead to an increase in risky behavior, and that such behavioral increases may therefore increase incidence and prevalence. Theoretical studies of the HIV epidemic in gay communities in the developed world predicted that incidence rates were likely to increase, and would mask the effect of antiretroviral therapy on reducing transmission ^[6]. These predictions have been found to be true for the San Francisco gay community. The wide use of antiretroviral therapy beginning in 1995 brought profound changes in San Francisco ^[22-24]. The number of individuals dying of AIDS precipitously dropped from its peak in 1994 concurrent with the rapid uptake of antiretroviral therapy ^[22]. The use of antiretroviral therapy among persons living with AIDS (PLWA) increased from 5% in 1995 to 75% in 2001 ^[22,25]. The number of PLWA rose from a plateau of 7432 in 1994 to 9096 by 2001 ^[26]. The uptake of antiretroviral therapy was accompanied by increases in sexual risk behavior among PLWA, as evidenced by increases in sexually transmitted infections (STI), from 0.66% of PLWA in 1995 to 2.02% in 2001 ^[27]. Reports of male rectal gonorrhea increased from 72 cases in 1994 to 237 in 2001 ^[28]. Early syphilis among gay men followed later, increasing from a low of six in 1998 to 115 in 2001 ^[29]. Unprotected anal intercourse increased from 18% in 1994 to 38% in 2001 ^[29,30]. Unprotected anal intercourse with two or more partners of unknown HIV serostatus increased from 10% in 1999 to 15% in 2001 for HIV-negative gay men and from 19 to 25% for HIV-positive gay men ^[29]. Also, direct measures of HIV incidence among gay men in several studies showed an increasing trend in the late 1990s ^[28,31-34]. In aggregate, HIV incidence among gay men was estimated to increase from 1.1% per year in 1997 to 2.2% per year in 2001 ^[34].

Meanwhile in other gay communities, similar increases in HIV incidence or risk were observed during roughly the same time period. Increases in STI were reported in several other cities in the USA, including Seattle, New York, and Los Angeles ^[35-37]. US national surveillance data also reported increases in new HIV diagnoses among men who have sex with men ^[37,38]. London and Sydney reported increases in high-risk sexual behavior among gay men ^[39-42]. Increases in HIV incidence rates were directly measured in gay communities in Ontario, Vancouver, Madrid, and Amsterdam between 1995 and 2000 ^[43-46]. The temporal coincidence of antiretroviral therapy uptake with increasing risk do not prove that antiretroviral therapy caused an increase in HIV transmission. Antiretroviral therapy undeniably caused an increase in the prevalence of HIV simply by increasing survival. Unless the incidence of new infection dramatically dropped in preceding years, increases in survival with antiretroviral therapy will more than offset AIDS deaths as treatment uptake expands. A more controversial point is whether antiretroviral therapy caused an increase in risky behavior and therefore HIV incidence. It has been suggested that the availability of antiretroviral therapy produced 'antiretroviral therapy optimism' or 'disinhibition' towards unsafe sex ^[47,48]. HIV-infected individuals may surmise that they are less contagious; HIV-negative individuals may concur; both may feel that HIV is less serious a disease. Studies have measured such attitudes among gay men, and some have linked them to a greater likelihood of unsafe sex ^[47,48]. However, their low levels overall and inconsistent associations with risky behavior may not convincingly argue for causing an increase in HIV incidence ^[48,49]. Another hypothesis is that PLWA returned to sexual activity in the era of antiretroviral therapy simply because their physical well-being returned ^[28]. A third mechanism is through the co-factor effect of STI. Co-infection of HIV and an STI can increase viral load and viral shedding despite antiretroviral therapy and therefore increase the transmission of HIV ^[50]. HIV-uninfected partners with an STI also become more susceptible to HIV acquisition, which can lead to increases in HIV incidence ^[51]. A variety of data thus show that increases in risky behavior occurred; as antiretroviral therapy decreased per partnership infectivity, these results imply that (as predicted by mathematical models) ^[6,9] the trade-off was in balance and was not likely to be observed (Fig. 1a). Future changes in incidence are to be expected as the level of treatment and risky behavior fluctuate. Recent statistical analyses of data suggest that considerable reductions in infectivity have occurred as a result of antiretroviral therapy ^[15,52].

The vast majority of observed increases in risky behavior after the introduction of antiretroviral therapy have been in gay communities. Increases in risky behavior, STI, and HIV incidence have not been observed in other populations in San Francisco despite high levels of antiretroviral therapy uptake. HIV incidence among heterosexual injection drug users (IDU) has remained near 0.5% per year throughout the 1990s ^[53,54]. HIV transmission among heterosexual individuals in San Francisco has also been persistently low despite the rapid uptake of antiretroviral therapy in this population ^[22]. Although at least one study among IDU found the perception that HIV treatments reduce transmission was associated with unprotected sex (although not risky injection practices) ^[55], large studies of HIV incidence among IDU in New York City and Baltimore found decreasing incidence rates during the 1990s ^[56,57]. Studies among women and heterosexual men found mixed associations between antiretroviral therapy use and risky sexual behavior ^[58-60], with one study finding that HIV-positive individuals taking protease inhibitors were significantly less likely to have unprotected sex within serodiscordant partnerships ^[61]. So far, the available data suggest that 'disinhibition' as a result of antiretroviral therapy is a phenomenon that has occurred primarily in gay communities, and is unlikely to occur in Africa as a result of the rollout. In Africa it may well be possible to decrease levels of risky behavior as increased contact between treated patients and their healthcare providers increases the opportunities for the prevention of secondary transmission. Moreover, increasing the availability of treatment for AIDS may reduce despair over seeking HIV testing, providing further opportunities for risk reduction counseling of HIV-positive and negative individuals. Therefore, with foresight, there is hope that HIV prevention may be strengthened as antiretroviral therapy use increases in Africa. However, if increases in risky behavior occur in Africa then it is quite possible that the epidemic level benefits of antiretroviral therapy in reducing transmission will be masked and incidence rates will increase.

Epidemiological data on resistance from the developed world

The development of acquired resistance is complex and is caused by many factors, including the potency of the agents, achievable drug concentrations, susceptibility of the virus to the drugs, and perhaps viral replication characteristics ^[62]. The ability to achieve durably suppressive drug concentrations is partly related to the pharmacokinetic and pharmacodynamic properties as well as the adherence of the patient to the medications. The use of specific combinations of agents that incompletely suppress replicating HIV has been associated with high rates of resistance in treated individuals ^[63]. Antiretroviral therapy was first made widely available in the developed world in 1996. As antiretroviral therapy has been introduced in most communities in the developed world layered over spans of dual (or even mono) antiretroviral agents the patterns of acquired resistance that have been observed reflect outmoded therapeutic strategies ^[64]. Nevertheless, it is instructive to review the published studies of transmitted and acquired resistance (Table 1 and Table 2). Communities that had access to treatments earlier in the course of the epidemic have developed a high prevalence of resistance in comparison with communities that had a later access to treatments. Communities in which the standard was to use less potent antiretroviral therapy combinations (because the regimens were the only ones available or with the expectation that the regimen was better tolerated or less expensive) have also developed a high prevalence of resistance.

Table 1

Table 2

Table 1 summarizes published studies on transmitted resistance from key primary infection cohorts. There are fairly consistent data from these studies. Most studies show an increasing incidence of drug resistance among newly infected individuals. The three largest studies in north America demonstrate an increase in drug resistance within each class of antiretroviral medication, and an increase in the prevalence of multiresistant strains. The picture is slightly less clear in Europe, however a French study and a study in the United Kingdom demonstrated a high level of resistance or an increasing rate of resistance among recent seroconverters. The rates of resistance among newborns appear to show a similar trend, and an increasing number of studies have demonstrated an increase in the transmission of viral resistance from mother to newborn infant. It is important to note that not all groups demonstrated an increasing prevalence of resistance. No resistance was noted among a cohort of IDU in British Columbia, and investigators in Sydney demonstrated a decrease in the transmission of genotypic resistance in the reverse transcriptase domain. The spread in a localized outbreak may thus not contain resistant isolates, and the controlled use of therapies that are maximally suppressive may reduce the development of acquired resistance; both of these phenomena may influence the level of transmitted resistance.

Table 2 contains data from studies on the prevalence of genotypic resistance (and phenotypic resistance if available) among treated patients. Studies were included if they represented an attempt to evaluate the prevalence of resistance that developed among individuals not selected because of recent HIV infection and outside of clinical trials that focused on the development of resistance. The studies are grouped by geographical area and then sub-grouped on the basis of the treatment history (treatment naive or treatment experienced). Trends can be discerned from Table 2. First, among individuals who have not received antiretroviral therapy, the prevalence of HIV resistance was low and ranged from 0 to 17%. This low rate was similar among the various geographical locations analysed. In addition, this low prevalence was similar to the prevalence observed among individuals with primary HIV infection (Table 1). These prevalences were influenced by the time from infection as the time to present for care may have allowed the virus to revert back from resistant to susceptible ^[98]. Infection by resistant isolates, compared with susceptible isolates, may influence the initial viral load, CD4 cell level and response to treatment, and these factors ultimately influence the rate of disease progression ^[67,99]. Another trend observed in Table 2 is that the prevalence of resistance is high and increasing. The studies demonstrated that high rates of resistance are associated with treatment failure among patients receiving effective care and treatment with antiretroviral medications. The prevalence of resistance for an individual class of antiretroviral medications approached 75-90% and the rates of two and three-class resistance varied from 'greater than 25%' to 62%. Clearly, as the prevalence of resistance increases the rate of transmitted resistance can be expected to increase.

The data in Table 1 and Table 2 show the temporal changing and complex epidemiology of drug resistant HIV in the developed world. The epidemic of drug-resistant HIV is continuing to evolve. High rates of poorly suppressive therapy used in certain communities have induced a high prevalence of resistance in treated patients. An increasing prevalence of resistance has led to increases in transmitted resistance. After resistance has developed within a community, changing therapies will not eliminate the resistant isolates among individuals who now harbor the archived resistant isolates. However, the widespread use of maximally suppressive therapies will probably delay resistance, reduce the emergence of transmitted resistance, and may even reduce the overall incidence of HIV. The use of agents such as ritonavir-boosted protease inhibitor-based therapy may be a substantial improvement towards long-term maximally suppressive therapy because of its favorable pharmacokinetics. A reduction in HIV resistance may thus occur. However, the toxicities of the medications are formidable and therefore resistance, or at least treatment exhaustion, may ultimately again lead to increases in HIV replication and treatment failure.

Drug resistance in primary HIV infection has implications for antiretroviral therapy beyond simply limiting the choice of active agents. For individuals with drug-resistant virus before the initiation of antiretroviral therapy, the time to full viral suppression is longer and the time to virological failure is shorter ^[67]. The persistence of drug resistance in primary HIV infection is somewhat different than in drug-experienced patients. In patients who develop drug resistance while on therapy and subsequently discontinue antiretroviral therapy, wild-type (non-resistant) HIV begins to re-emerge at a mean of 6 weeks after antiretroviral therapy discontinuation ^[100]. This is believed to represent the re-activation of archived wild-type virus from latently infected reservoirs, which has a fitness advantage over drug-resistant virus in the absence of drug pressure. In contrast, drug resistance in primary HIV infection, in which there is no latent reservoir of wild-type virus, is likely to persist for many years ^[101,102], which will influence antiretroviral therapy treatment response in these individuals and may allow for secondary transmission of this drug-resistant variant. Are resistant isolates (and by definition less evolutionary favored than a susceptible virus) associated with a poorer prognosis? This is unclear because resistant isolates may have a lower replication capacity and may be associated with higher initial CD4 cell levels. Only longer term studies that evaluate the relative advantage or disadvantage conferred by infection with resistant or susceptible isolates will answer questions regarding viral fitness, patient survival, and the implications for optimal treatment management. However, it is clear that resistance isolates are archived, and once the medications are begun, the resistant isolates will rapidly re-emerge, resulting in treatment failure.

Insights from mathematical models: evolution of drug-resistant HIV epidemics

Mathematical modeling studies (coupled with uncertainty analysis) of the public health impact of antiretroviral therapy in developed countries has shown that the widespread usage of these therapies has: (i) substantially reduced mortality rates; (ii) prevented a significant number of new infections (by reducing viral load in treated individuals); and (iii) increased the transmission and prevalence of drug-resistant strains ^{[6-9,11,13,16,17,19,20]}. On the basis of the modeling analyses it has been argued that a high usage of antiretroviral therapy should be viewed as an effective public health prevention strategy, as well as an effective therapeutic strategy ^[6,9]. Models have been used to predict both the incidence of resistance (i.e. the number of cases of transmitted resistance per year), and the prevalence of resistance in San Francisco ^[6,7,11]. Surprisingly, it was predicted that even with a very high usage of antiretroviral therapy (assuming that a median of 70% of HIV-infected men received therapy) transmitted resistance would initially increase but would then fairly quickly stabilize at a relatively low level (between 6 and 21%). Once empirical data on transmitted resistance had been collected, the modeling predictions were compared with the empirical data and were found to be in close agreement ^[11]. Although the modeling studies predicted that transmitted resistance would remain relatively low, they predicted that the prevalence of resistance would rise to high levels; it was predicted that 42% of San Franciscans with HIV infection could be expected to have drug-resistant virus by 2005 ^[7]. A relatively low rate of transmitted resistance occurs, but a high prevalence of drug resistance emerges because the majority of cases arise as a result of acquired resistance ^[7]. Careful multivariate analyses of the theoretically generated data have been performed in order to suggest how best to minimize the problem of drug-resistant HIV ^[7,16]. Landscape policy analysis (LPA) has recently been used to design robust health policies for implementing antiretroviral usage in developing countries ^[16]. LPA involves generating three-dimensional graphs by simultaneously varying two program variables (for example, the usage rate of antiretroviral therapy and the rate of development of acquired resistance). These analyses suggest that it is necessary to define country-specific 'acceptable' rates for acquired resistance in treated patients ^[16]. LPA also provides a robust theoretical framework for designing health policies that have conflicting health policy goals ^[16].

As the public health impact that antiretroviral therapy has is dependent upon the fraction of the HIV-infected population that receives treatment ^[6-9,16], the epidemic-level impact that antiretroviral therapy will have in Africa is likely to be small. Currently, there are estimated to be 25-28 million individuals infected with HIV in Sub-Saharan Africa (Table 3) ^[103], and 15-20% of these individuals are likely to be eligible for treatment. The rollout of antiretroviral therapy in Africa plans to provide treatment to 2-3 million individuals; thus, approximately only half of the treatment eligible may receive antiretroviral therapy (i.e. only 5-10% of infected individuals will receive antiretroviral therapy) ^[104]. Fig. 2 shows the predicted impact of antiretroviral therapy in Africa as a function of the percentage of HIV-infected individuals who receive treatment. Predictions are only made for 10 years into the future, as it is unlikely that the same treatment regimens will be used beyond this; thus it would be misleading to make longer-term predictions. If only 10% of HIV-infected individuals are treated (as seems likely) then even 10 years after the rollout begins the prevalence will not have decreased (blue data shown in Fig. 2a), and the incidence rate will have decreased by less than 5% (blue data shown in Fig. 2b). Therefore, these modeling analyses indicate that changes in the incidence rate in Africa as a result of the rollout will be undetectable. However, as approximately 3 million people are infected each year in sub-Saharan Africa (Table 3) then even a modest reduction in incidence rates will translate into a substantial number of infections prevented. Modeling analyses suggest that (if 10% are treated) then 143 296-473 598 infections would be prevented simply as a direct result of treating infected individuals in the first 5 years after the rollout (Table 3). However, if the treatment rate was 25% then between 434 745 and 1 342 887 infections would be prevented over 5 years (simply as a result of treating infected individuals) (Table 3).

Table 3

Fig. 2

The modeling analyses also suggest that (if 10% are treated) that a total of 22 688-300 293 cases of transmitted resistance can be expected to occur in the next 5 years (Table 3). The modeling reveals the evolution of the expected epidemic of transmitted resistance [red data Fig. 3a (1 year after the rollout), Fig. 3b (5 years after the rollout), Fig. 3c (10 years after the rollout)]. The level of transmitted resistance that evolves over time is directly dependent upon the level of treatment ^[7,9,16]. Fig. 3a-c shows the evolution of the quantitative relationship between the percentage of the infected population on treatment (varying between 0 and 40%) and the level of transmitted resistance (red data). The World Health Organization (WHO) is calling for wide-scale surveillance and monitoring of transmitted resistance during the rollout, with a threshold criteria of 5% (Fig. 3a-c). This threshold criterion means sample sizes would be set in order to detect transmitted resistance if it exceeds 5% of new cases. However, the modeling analyses clearly show that if 10% (or less) of the HIV-infected receive treatment then, even after 10 years, the levels of transmitted resistance will be below the surveillance threshold of 5% set by the WHO (Fig. 3c). These results suggest that large-scale surveillance systems for detecting and monitoring drug-resistant HIV in Africa will be unnecessary. However, as treatment coverage is likely to be heterogeneous, we propose that a sentinel surveillance system for drug resistance monitoring would be most practical. We suggest that sentinel surveillance sites should be based in major urban areas where treatment coverage rates are likely to exceed 10%. Modeling analyses ^[9,16] predict that the prevalence of drug resistance will rise (Fig. 3d-f) and (if antiretroviral therapy coverage rates are 10%) will be 10% or less after a decade (Fig. 3f). However, a prevalence of 10% translates into 2.5-2.8 million individuals infected with drug-resistant strains. The majority of these cases of drug resistance will occur as the result of acquired resistance (Fig. 4a), and not transmitted resistance. A high proportion (30-80%) of treated cases can be expected to become drug resistant, with the proportion increasing over time (Fig. 4b). As the proportion of the treated cases increases, transmitted resistance will increase (Fig. 4b). However, it is essential to realize that it is possible for a high percentage of the treated cases to become resistant, but for the levels of transmitted resistance to remain low (Fig. 4b). We suggest that monitoring drug resistance in treated cases would be an effective strategy for both monitoring the effectiveness of the program for therapeutic purposes and could also provide some indication (albeit imperfect) of the level of transmitted resistance (Fig. 4b).

Fig. 3

Fig. 4

The primary purpose of the antiretroviral therapy rollout is to provide direct therapeutic benefits for treated individuals by increasing the length and quality of their lives. A high usage of antiretroviral therapy in developed countries has generated high levels of drug resistance. However, in Africa usage rates of antiretroviral therapy will be substantially lower than they have been in the developed world. By evaluating the empirical data from the developed world, and using mathematical models we predict that over the next decade in Africa: (i) the impact of the rollout on reducing HIV transmission (and prevalence) is likely to be undetectable (however the number of infections prevented will be substantial); (ii) the transmission rate of drug-resistant HIV will be below the surveillance threshold of 5% set by the WHO (except in a few urban centers where antiretroviral therapy usage is greater than 10%); and (ii) the majority of cases of drug-resistant HIV that will occur will be the result of acquired resistance and not transmitted resistance. We conclude that large-scale surveillance systems for detecting and monitoring transmitted drug-resistant HIV in Africa will be unnecessary. However, implementing sentinel surveillance in a few urban centers, where treatment coverage is high (> 10%), would be very useful for monitoring the emergence of transmitted resistance. We recommend that monitoring the prevalence of drug resistance in treated cases would be an effective strategy both for monitoring program effectiveness in achieving durable suppression of viral replication and for assessing the public health impact of the rollout on drug-resistant HIV. The epidemic-level impact of antiretroviral agents will mainly be determined by how widely they are made available and how well they are used. The effect of antiretroviral agents on the HIV epidemic is complex, because these therapies produce both a beneficial and a potentially detrimental effect at the epidemic level. However, any health policy decisions as to the availability of antiretroviral therapy should be made on the basis of therapeutic goals and ethical standards. It is clear that large-scale prevention efforts for both HIV-negative and HIV-positive individuals are needed and that there is an urgent need to develop, evaluate and implement effective HIV prevention programs to alter the current, tragic trajectory of the epidemic in Africa. The need for antiretroviral agents will continue to increase; only a vaccine (even if imperfect) is likely to be able to control the ever-expanding pandemic ^[12,105-107].