Response to correspondence from Hastings et al. entitled ‘Will ART rollout in Africa drive an epidemic of drug-resistant HIV?’

Abstract

Hastings et al. comment on our recent review paper of empirical data and theoretical models in which we assessed the likely outcome of the antiretroviral (ART) rollout on the transmission of drug-resistant HIV in Africa [1]. Our central conclusion was that – in the next few years – high levels of transmitted drug-resistant HIV are unlikely to evolve in Africa as a result of the rollout. Hastings et al. [2] state that they disagree with our conclusion, and they do this on the basis of conducting four scenarios of a simple spreadsheet model. We would like to stress that our central conclusion is based upon an extensive review of the literature of the experiences with ART in Europe and North America [1]. In these places ART has been used fairly extensively for almost 20 years; however, ART durably suppresses viral replication only with the combination of at least three medications and combination therapy has been common practice for less than a decade. Furthermore, in certain communities with a high prevalence of HIV (such as communities of men who have sex with men) levels of risky behavior have increased. We stress that it is important to note that even under these conditions (that are extremely conducive to the evolution of transmitted resistance) relatively low rates of transmitted resistance have occurred [1]. The relatively low rates of transmitted resistance that have evolved in Europe and North America can be understood in terms of population dynamics and evolutionary biology [3,4]. Drug-resistant strains have evolved, and have been transmitted, in communities that already have a high prevalence of HIV; exponential growth of transmitted drug resistance would only occur if these strains were introduced into a community that was HIV-naive. In a high-prevalence community the vast majority of transmitters are transmitting wild-type strains; only relatively few individuals are transmitting resistant strains. Competitive dynamics occur between the wild-type and the drug-resistant strains. These competitive dynamics ensure that the rise of transmitted drug resistance is limited and stabilizes, as has been predicted by previous modeling [5–7], and has been found to occur in practice [1,8]. If the drug-resistant strains that evolve are less transmissible than the wild-type strains then the level of transmitted resistance stabilizes at a moderate to low level [5]. However, if the drug-resistant strains that evolve are more transmissible than the wild-type strains then the transmitted drug-resistant strains will – eventually – out-compete the wild-type strains [5]. The length of time that this out-competition process would take would depend upon the transmissibility of the drug-resistant strains, the rate of acquired resistance, and the usage rate of ART. In Africa only a low usage of ART is expected (only 5 to 10% of HIV-infected individuals are expected to receive treatment), and drugs that are effective in viral suppression will be used [1]. The rollout is just beginning. Therefore, based upon our experience of ART in Europe and North America we conclude that relatively little transmitted resistance will evolve in the next few years. In our review paper we also use mathematical models to illustrate the concepts of population biology and evolutionary dynamics of transmitted resistance. Hastings et al. [2] state that they have repeated our mathematical analyses, but they have not done so. Firstly, the minor technical points that they raise about our model are incorrect, and are based upon their misunderstanding of our modeling work. Secondly, they have used difference equations and spreadsheet modeling rather than (as we did) differential equations and numerical integration, which has resulted in numerical errors in their calculations. Thirdly, they examined only four scenarios, they did not conduct a full uncertainty analysis as we did (thus they did not examine the full range of outcomes). Finally, and most importantly, they assumed that 40% of HIV-infected individuals would receive treatment whereas we assumed that only 5–10% would. If they had repeated our analysis with the correct parameters for Africa treatment rates (as we did) then they would have obtained similar results to ours. Mathematical models are useful tools for health policy modeling, but it is always important to examine the current, and the historical, epidemiological data. Often the best prediction for the future is based upon an evaluation as to what has occurred in the past. Thus, we conclude that, in the next few years, relatively low levels of transmitted resistance will evolve in Africa. Past observation and our modeling provide encouragement that roll-out efforts in Africa can continue without unwarranted fears of large-scale transmitted ART resistance and that surveillance resources to track the emergence and spread of primary resistant strains can be efficiently targeted to the few sentinel sites where they are likely to appear first.