To establish the potential effect of intervention packages at a population level, we used the stochastic, agent-based simulation model of HIV transmission that was developed for the Prevention Umbrella for MSM in the Americas Project, and used by Beyrer and colleagues (their appendix
describes the modelling framework in detail).68
We used a more fully parameterised model to represent the MSM transmission network for four case studies of epidemic patterns: an MSM-focused epidemic in a developed country (USA), an MSM-focused epidemic in a developing country (Peru), a widespread epidemic mainly in heterosexual people with some transmission in MSM (Kenya), and a mixed epidemic in heterosexuals, MSM, and injecting drug users (India). Details of this parameterisation, including data sources and selected parameter values for the US and Peru are in Beyrer’s paper’s appendix68
and for India and Kenya in our appendix
. Setting-specific aspects of the model included sexual behaviour with main and casual partners; role versatility (ie, insertive vs
versatile); patterns of testing and treatment; patterns of ageing, birth, and death; prevalence of circumcision and sex with women (in MSM); and HIV prevalence in women. Viral load trajectories on and off treatment and transmission probabilities by viral load were constant across models. Baseline models assumed 2011 levels of treatment coverage and condom use for anal sex and no use of pre-exposure prophylaxis. Baseline models were calibrated against independent sources of prevalence data (appendix
We simulated three prevention packages based on condoms, oral pre-exposure prophylaxis, and improved antiretroviral initiation (ie, more people taking antiretrovirals and more prompt initiation of antiretroviral therapy), respectively and applied them to the baseline models. We based uptake, adherence, and response on findings from our systematic review (appendix
). If few data were available, we did sensitivity analyses across several variables selected on the basis of expert opinion. To show the need for adequate resources for scale-up, we modelled the effect of prevention at varying levels of intervention coverage. We investigated the idea that packaging of complementary interventions together increases their effect with the oral pre-exposure prophylaxis package model at three levels of adherence and corresponding estimated efficacy to simulate implementation with and without supportive adherence interventions. Although not modelled here, adherence is important for the antiretroviral-therapy-based package. We assessed the effect of each intervention scenario as the proportionate reduction in the number of new infections within 10 years after rollout compared with that at baseline, and the effect of varying degrees of coverage.
The results of our modelling show that, if oral pre-exposure prophylaxis and antiretroviral treatment coverage were assumed to be 40%, and 20% of unprotected anal intercourse encounters were replaced with condom-protected intercourse, between 7–29% of incident HIV infections would be averted during 10 years (). Increasing pre-exposure prophylaxis coverage from 20% to 80% increased the estimated cumulative proportions of infections averted (). Additional data about varying coverage of antiretroviral therapy and condom provision are included in the appendix
. Increasing the frequency of pre-exposure prophylaxis adherence sufficient to achieve the high efficacy (ie, a 73% reduction in acquisition of HIV infection) in men from 50% to 75% resulted in higher estimates of averted HIV incidence in all countries compared with baseline. However, the number of infections averted did not rise when we increased the modelled proportion of men with sufficient adherence from 75% to 90% ().
Modelled estimation of proportion of infections averted in 10 years after provision of HIV prevention packages to men who have sex with men
Stochastic simulation estimating the proportion of HIV infections averted in 10 years by an oral pre-exposure prophylaxis prevention package in men who have sex with men, by degree of coverage
Stochastic simulation estimating the proportion of HIV infections averted in 10 years by an oral pre-exposure prophylaxis prevention package in men who have sex with men with 40% coverage, by degree of adherence
The results of the modelling show several key notions of HIV prevention for MSM. First, packages of interventions with sufficient coverage can have pronounced effects on the incidence of HIV infection in MSM worldwide. Second, a high degree of coverage for efficacious interventions is important to increase their effect. In many cases, coverage will be poor irrespective of funding until men can safely access care, comfortably discuss their sexual risks for HIV with health-care providers, receive referrals for appropriate services, and confidently use prevention methods and services that will reduce their risks of acquisition or transmission of HIV infection. Finally, packaging of complementary inter ventions—eg, adherence support—can increase the effect of primary biomedical interventions.
Between-country variations in the proportion of infections averted could be because of stochasticity, differences in underlying baseline conditions, or a combination of both. For example, the proportion of infections averted through use of the antiretroviral therapy package was higher in countries that had less treatment and lower baseline CD4 counts at the beginning of treatment than it was in countries with more treatment and higher CD4 counts. The estimated number of infections averted by the condom package was higher in India than in other countries, which could be partly because of higher estimates of baseline condom use in India.