Modelling the impact of HIV prevention and treatment for men who have sex with men on HIV epidemic trajectories in low- and middle-income countries

Abstract

Little is known about the impact of combination HIV prevention interventions for men who have sex with men (MSM) and the impacts on the wider epidemics. Modelling analyses of MSM-specific interventions across varied HIV epidemics may inform evidence-based responses. The Goals model was adapted to project the impacts of providing HIV interventions for MSM and access to expanded coverage of antiretroviral therapy (ART) for adults to measure the effects on the MSM and adult epidemics in Peru, Ukraine, Kenya and Thailand. Positive impacts were observed in all four countries. Across epidemics, 14–25% of infections among MSM may be averted between 2012 and 2016 when MSM interventions are brought to scale and MSM have equal access to expanded ART for adults. Among adults, MSM interventions may avert up to 4000 new infections, in addition to the benefits associated with increased ART. Greatest impacts from expanded interventions were observed in countries where same sex transmission contributes significantly to the HIV epidemic. While significant benefits are observed among the adult and MSM populations with expansion of ART, consideration should be given to the synergies of combining ART expansion with targeted interventions to reach hidden, high-risk populations for HIV testing and counselling and linkages to care.

Keywords

HIV AIDS homosexuality MSM HIV infection Goals model transmission prevention

INTRODUCTION

The HIV epidemics among men who have sex with men (MSM) have received growing attention from researchers and medical communities, donors and political stakeholders. Baral and colleagues’ systematic review of HIV prevalence data reported by low- and middle-income countries (LMICs) from 2000 to 2006 demonstrated a markedly higher risk for HIV among MSM when compared with the general population, with a pooled odds ratio (OR) of 19.3 (95% confidence interval [CI] 18.8 to 19.8) across all countries when comparing MSM to all reproductive age adult males. HIV risks were 33 times greater for MSM in the Americas (OR 33.3, 95% CI 32.3 to 34.2), 19 times greater for MSM in Asia (OR 18.7, 95% CI 17.7 to 19.7) and four times for MSM in Africa (OR 3.8, 95% CI 3.3 to 4.3) compared with other men.¹ Responses have varied across countries; ranging from countries where stakeholders have actively sought to understand and increase prevention strategies to respond to the epidemic among MSM compared to other countries where response is challenged by a lack of data and/or sociopolitical issues that inhibit care seeking. Same-sex behaviour among men remains criminalized in over 30 LMICs globally, including approximately half of the countries in Africa, and is punishable by death in at least four countries.² Stigma in countries where same sex practices have been decriminalized still inhibits MSM from seeking HIV-related care or prevention services. These barriers compromise and limit understanding of HIV epidemics among MSM and, hence, of the wider populations with which they interact.

We recently reported on emerging MSM epidemics, developing a method to classify HIV epidemic scenarios among MSM and to incorporate the concepts of concentrated or generalized epidemics in a more distinctive approach that is inclusive of MSM and other at-risk groups.³ These scenarios situate MSM transmission within wider epidemic contexts. Based on a systematic review of the existing data we found that HIV epidemics in LMICs could be described with four country scenarios: (1) the predominant mode of transmission of HIV cases is related to MSM anal sex transmission; (2) HIV transmission among MSM occurs in the context of injecting drug use (IDU) epidemics; (3) HIV transmission among MSM occurs in the context of widespread and generalized HIV transmission among heterosexuals; and (4) MSM, heterosexual and parenteral modes of transmission all contribute to HIV spread.³ These epidemic scenarios allow for more nuanced understandings of the interplay of transmission related to MSM and the other components of the HIV epidemics that exist at population levels. Acknowledging, for example, that some MSM may also inject drugs or that in many epidemic contexts a majority of MSM may be married to women and acquisition and transmission risks, therefore, exist for both MSM and women.

Evidence-based responses to prevent HIV transmission among MSM require an understanding of the impacts of interventions available for MSM. Through clinical and outreach services, interventions can be provided to MSM for individual prevention of HIV transmission and acquisition. Limited data on population-level impacts with HIV endpoints exist for these interventions for MSM. Mathematical models can fill this gap to inform national and local programming by estimating the impacts of alternative combinations of interventions. We used the Goals model to assess the impact of combinations of HIV prevention interventions and antiretroviral therapy (ART) treatment access and uptake for MSM on overall country-level HIV epidemic scenarios, through 2016, represented by Peru, Ukraine, Kenya and Thailand.

METHODS

Modification of Goals

The Goals model was adapted and used to estimate HIV incidence, with and without one or more interventions delivered at different levels of coverage. The Goals model (Spectrum software, Futures Institute, Glastonbury, CT, USA) is a deterministic model that uses data in three main areas to project new HIV infections: demography, sexual behaviour and HIV/sexually transmitted infection (STI) epidemiology.⁴ Prior to this effort, Goals considered five different risk groups in the general population: MSM, IDU, high-risk heterosexuals, medium-risk heterosexuals and low-risk heterosexuals. Recognizing that risk may vary among MSM as well, we adapted the Goals model within the Spectrum software to allow for the input of epidemiological data relating to four MSM risk categories (Table 1). The MSM risk categories were defined as high-risk MSM or male sex workers, medium risk or MSM with more than one sex partner in last 12 months, low-risk MSM or men with 0–1 male sex partner in last 12 months (stable relationship) and MSM IDU or MSM who also inject drugs. Further details of the Goals model are available in the appendix or the Goals manual.⁵

Table 1

Major updates to the Goals model and rationale

Addition of MSM risk categories	Based on literature reviews followed by expert and community group review and consensus High-risk MSM: male sex workers Medium risk: MSM with more than one sex partner in last 12 months Low-risk MSM: men with 0–1 male sex partner in last 12 months (stable relationship) MSM-IDU: MSM who also inject drugs
Addition of MSM-specific interventions and impacts	Based on systematic review of literature and analysis for interventions impacts Individual level with promotion and distribution of condoms and water-based lubricants (acknowledging that water-based lubricants must be provided with condoms to maintain condom efficacy) Community-level behavioural change interventions based on evidence of higher efficacy of community over individual level interventions for MSM¹
Epidemiological inputs: infectiousness when on ART	Based on the most recent finding reported by Eshleman and colleagues who reported an 86% reduction in HIV transmission after ARV treatment was initiated in discordant couples

MSM = men who have sex with men; ARV = antiretroviral

The Goals model was further adapted to include selected MSM-specific interventions with evidence for efficacy. These included promotion and distribution of condoms with water-based lubricants; community-level behavioural change interventions which include HIV counselling and testing (selected over individual level behavioural change as a recent Cochrane review showed higher efficacy of community over individual level interventions for MSM⁶); and earlier initiation of ART (Table 1).^7,8

Selection of countries and inputs for Goals

Four countries with HIV prevalence and transmission risk data for MSM were selected, each representing an epidemic scenario. MSM specific data for each country were collected through a systematic review (methods previously published in Beyrer et al. ³) and augmented by communications with country experts and review of gray literature and country reports. Key data used to parameterize the Goals model are presented in Table 2 and described briefly below. Once parameterized, the Goals models were calibrated to the median UNAIDS estimations for the prevalence of HIV infection among the adult population and fit to within 95% CIs around the recent projected UNAIDS estimates (unpublished). The HIV prevalence trends among MSM that were estimated by the models were cross-checked against surveillance or other published estimates of HIV prevalence among MSM in each country.^9–19

Table 2

Epidemiological assumptions and behavioural inputs and sources per MSM risk category for each country model

	Ukraine				Kenya				Peru				Thailand
	Low	Med	High	MSM-IDU	Low	Med	High	MSM-IDU	Low	Med	High	MSM-IDU	Low	Med	High	MSM-IDU
Proportion of adult male population
Value in model (%)	0.41	1.40	0.55	0.15	0.53	2.01	0.42	0.4	0.62	1.86	0.91	N/A	0.68	1.78	0.59	0.35
Reference(s) for estimates	2,3	2,3	3,4	3,5	3,6	3,6,7	3,8	3,8	3,9	3,9	3,9	3,9	3,10	3,10	3,11	3,12
HIV Prevalence among MSM
Value in model (%)	N/D	10.6*	16.0	43.0	N/D	24.8	16.7	N/D	20.2	23	25.2	N/A	N/D	19.2	15.6	31.8
Reference(s)		5,13	14	15		16	7,16		9	9	9			10	10	12
Ulcerative STI prevalence among MSM
Value in model (%)	10.5^†	10.5	12.0	10.5^†		31.8	19.7	N/D	7.0	15.0	20.0	N/A	7.0	7.0	7.0	7.0
Reference(s)	17	17	14	17	16	16	16		9	9	9		18	18	18	18
Percent married
Value in model (%)	10.3	10.3	10.3	10.3	0	12.4	4.5	N/D	1.6	9.5	9.5	N/A	14.1	14.1	14.1	24.2
Reference(s)	2	2	2	2	16	16	16		9	9	9		19	19	19	19
Consistent condom use (adjustment factor – 33% for use at last contact)
Value in model (%)	25.0	27.1	26.5	27.1	25.0	38.6	32.1	N/D	16.0	28.0	42.0	N/A	15.0	45.0	80.0	45.0
Reference(s)	2	2	2	2	7,16	7,16	7,16		9	9	9		19	10,19	19,20	12
Mean sexual partners per year
Value in model	1	4	12	4^†	1	3	20	N/D	1	3	20	N/A	1	4.7	28	15
Reference(s)	t	2	21	2	t	16	16		9	9	9		t	22	22	19

MSM = men who have sex with men; STI = sexually transmitted infection

*Pooled HIV prevalence used for all risk groups in the absence of group-specific data

^†Medium-risk group estimates used as a proxy when low or MSM-IDU data are unavailable; t number of partners by definition of low risk group; N/D = no data; N/A = not applicable

Scenario 1 countries, where MSM transmission predominantly contributes to the HIV epidemic, were represented by Peru. General population prevalence data and projected ART coverage estimates were available from DemProj and AIM of the Spectrum suite, as well as behavioural inputs, and intervention coverage among the general population. Personal communication with research experts from Peru (Konda and Caceres CC, personal communication) provided data specific to the MSM population, and a 2009 paper by Aldridge and colleagues,²⁰ who conducted similar modelling work using an Excel version of the Goals model, was used to supplement any missing information and confirm inputs.

Scenario 2 countries, where HIV transmission among MSM occurs in the context of IDU epidemics, were represented by Ukraine. General population data and ART-projected coverage were populated using data available in DemProj and AIM of the Spectrum suite and further updated with UNAIDS and Alliance reports;^12,21,22 HIV and STI prevalence of HIV among IDU and MSM and behavioural data were derived from UNGASS reports²¹ and published papers; and^9,10,23 data from Russia were used as a proxy for behavioural and epidemiological data specific to the MSM IDU category.^24–30

Scenario 3 countries, where HIV transmission among MSM occurs in the context of well-established HIV transmission among heterosexuals, were represented by Kenya. General prevalence data and ART-projected coverage were provided from DemProj and AIM of the Spectrum suite, population prevalence of MSM were calculated using data from peer-reviewed publications;^15,31 prevalence of HIV and STI and MSM behavioural data were provided by experts from the Kenya Medical Research Institute (KEMRI, unpublished); and coverage of MSM specific interventions were obtained from UNGASS/MoH reports.^13,32,33

Scenario 4 countries, where both sexual and parenteral modes of transmission contribute significantly to HIV transmission, were represented by Thailand. General population data, including incidence and prevalence trends and national projections for ART coverage, were populated using data available in DemProj and AIM of the Spectrum suite. IDU and MSM epidemiological and behavioural data were collected with assistance from the Bangkok MSM Cohort research experts from data published in the Asian Epidemic Model³⁴ and other publications from the CDC, the Thai Red Cross, Thailand's Ministry of Public Health and other regional experts.^{11,18,19,35–38} Intervention coverage estimates were obtained from UNGASS/MoH reports³⁹ and expert opinion.

For all country models, we updated infectiousness when on ART with the finding reported by Eshleman et al. ⁸ that demonstrated reduction in HIV transmission after ART treatment was initiated in discordant couples. Given the importance of recent findings from the HPTN 052 study, we modelled the impact of earlier initiation of treatment among all adults in need of ART, according to the recent change in standards.⁴⁰ These models utilized an initiation of ART at a CD4 count of 350 cells/mm³ and the demonstrated effectiveness value of 88.6% reduction in risk associated with ART at this initiation.⁸ For this study, we used a slightly conservative estimate of 87% (0.13 reduction in transmission when on ART).

The behavioural intervention impact matrix used in the Goals model records the impact of exposure to a prevention intervention on the reduction in non-use of condoms and the reduction in number of partners per risk group (See Goals Description, Appendix, for further details).⁴ Because two new MSM interventions were added, the impacts of exposure to MSM prevention interventions were derived through the systematic review of HIV prevention interventions for MSM.^3,41 The results of these reviews were then used to develop summary estimates of the reductions of condom non-use (unprotected anal intercourse) and reduction in number of partners, stratified for each MSM risk group (Table 3). These estimates provided the data for the impacts of the MSM specific interventions.

Table 3

Country intervention coverage: baseline 2011 coverage levels and expansion scenarios used for projections

	Outreach for MSM			Condoms and lubricants for MSM
		Scenario 1	Scenario 2		Scenario 1	Scenario 2	ART for all adults who meet criteria
Country of interest	Status quo	+20% of status quo	+40% of status quo	Status quo	+20% of status quo	+40% of status quo	Status quo (2011)	Expanded (2016)
Peru	21%²³	41%	61%	45%²³	65%	85%	43%	57%
Ukraine	33%⁵	53%	73%	40%⁵	60%	80%	12%	24%
Kenya	15%⁷	35%	55%	5%¹⁶	25%	45%	63%	85%
Thailand	20%²⁴	40%	60%	20%²⁴	40%	60%	53%	59%

MSM = men who have sex with men; ART = antiretroviral therapy

Estimating the fraction of HIV transmission attributable to MSM

To understand how MSM transmission is situated within each country's epidemic, we estimated the total fraction of new HIV-positive cases attributable to same sex behaviour of MSM. This was estimated through a calculation nested within the overall Goals model analysis of each of the case countries. The baseline number of new HIV infections was estimated using the ‘current intervention scenario’ of each country, in which no additional coverage of existing or new interventions is added to what is currently in place in 2011. Coverage levels from 2011 were selected as most data and reports were available from this year and provided the most coverage information for all four selected countries. The number of new HIV infections attributable to MSM is then calculated as a residual by subtracting the number of new HIV infections not attributable to MSM from the total number of new HIV infections. To estimate the number of new HIV infections not attributable to MSM, the HIV transmission probability for MSM was assumed to be zero, and the number of new HIV infections was then calculated. The numerator was calculated as the number of new HIV cases not attributed to MSM subtracted from the total number of new HIV cases; the denominator was the number of total new HIV cases. The percentage is presented for year 2016 for each case country to understand how the epidemic would progress without further expansion of intervention coverage.

Projections of MSM intervention scenarios

The impact of MSM focused interventions on the HIV epidemic can be assessed by varying levels of intervention coverage among MSM and coverage of ART among the adult population, including and assuming equitable distribution to MSM in need of treatment. Estimated coverage levels of each intervention modelled are detailed in Table 3. The modelled scenarios include (1) Status quo: current coverage of MSM-specific prevention interventions are maintained from 2011 to 2016 while all other interventions remain constant. (2) Scenario 1: all MSM interventions increase from current, 2011 levels incrementally to reach an additional 20% by 2016 (coverage changes from 2012 to 2016), while all interventions are held constant from 2011 levels. (3) Scenario 2: MSM interventions increase from current, 2011 levels incrementally to reach an additional 40% by 2016 (coverage changes from 2012 to 2016), while all interventions are held constant from 2011 levels.

All three scenarios were analysed with ART coverage of the adult population held constant from 2011 levels forward. The scenarios were then replicated with the planned expansion of ART among the adult population. These scenarios assume that the increase in ART coverage among adults applies to MSM as well as other adult risk groups, and also assumes a supportive environment in which MSM interventions are fully provided and HIV-positive MSM have full access to HIV counselling and testing and ART. We investigate the impacts of increased coverage of MSM interventions and ART expansion among adults on new infections among the MSM population from 2012 to 2016. We investigate the combined impacts and the impacts that are attributable to the increased coverage of the MSM interventions. We also investigate the impacts on the new infections among the adult population that are attributable to expansion of the MSM interventions, in the environments of either maintained or expanded ART.

Sensitivity analyses were conducted to assess the uncertainty surrounding the impact coefficients for the disaggregated MSM groups. These analyses were performed using upper and lower bounds of impacts of optimal intervention scenario for MSM-specific interventions (Table 4) and sensitivity analyses were conducted for second scenario that included ART coverage among the adult population. The results are presented as the percent reduction in new infections among MSM, compared with the status quo estimates without expansion of ART.

Table 4

Transmission impacts of MSM-specific interventions on condom use and number of partners per MSM risk categories: results of a systematic review for inputs to the Goals impact matrix

	Reduction in condom non-use			Reduction in number of partners
	High-risk MSM	Medium-risk MSM	Low-risk MSM	High-risk MSM	Medium-risk MSM	Low-risk MSM
Intervention	% reduction (range)	% reduction (range)	% reduction (range)	% reduction (range)	% reduction (range)	% reduction (range)
Community-based outreach	−19.6	−19.6	−19.6	−19.0	−19.0	−19.0
	(−8.1 to −29.9)	(−8.1 to −29.9)	(−8.1 to −29.9)	(−0.7 to −25.0)	(−0.7 to −25.0)	(−0.7 to −25.0)
Individual with condoms and lubricants	−22.8	−17.4	−6.9	−18.3	−0.8	0.0
	(−0.6 to −26.0)	(−2.0 to −58.8)	(−4.0 to −9.0)	(−1.1 to −25.0)	(0 to −3.8)	0.0

MSM = men who have sex with men

Source: Systematic review of HIV prevention interventions for MSM²⁰

RESULTS

The proportion of new HIV cases attributed to male anal sex transmission from 2012 through 2016 varies across the four countries when coverage levels were held constant (data not displayed). Figure 1 shows the fraction of new HIV cases attributed to MSM by 2016. Ongoing MSM incidence would contribute to 8% of new infections in Kenya, to a high of 96.3% of new infections in Peru, by 2016. The attributable fraction of transmission related to MSM is increasing in Peru and, to a lesser degree, in Thailand. The prevalence of HIV among IDUs and sex workers in Thailand and Ukraine may account for a significant proportion of the remaining 53% and 80% of HIV that is attributed to other modes of transmission, respectively. Finally, the attributable fraction of HIV infections related to MSM may be increasing in Kenya. Estimating these proportions allows us to account for behaviour and transmission between MSM and bridging with other populations (e.g. MSM who inject drugs and female sex partners of bisexual men).

Figure 1

Attributable fraction of new HIV infections from MSM behaviour in selected countries by 2016. MSM = men who have sex with men

Figures 2–5 display the modelled impact of the interventions on new infections among MSM. Figure 6 displays the sensitivity analysis around the maximal intervention – increasing coverage of MSM interventions by 40% by 2016 and allowing equal access to expanded ART for MSM – compared with a status quo scenario in which there is no increase in coverage of MSM interventions or ART. The error bars display the high and low values of the sensitivity analysis of the impact matrix. Tables 5–8 depict the numbers of new HIV infections among adults for each country stratified by intervention combination without and with ART expansion among the adult populations.

Figure 2

Trajectories of the estimated number of new HIV infections among MSM in Peru with implementation of three intervention scenarios for MSM without (a) and with ART expansion among the adult population (b) (2011–2016). y-Axes do not start at zero new infections for the purpose of ensuring differences in new infections are visible over time. MSM = men who have sex with men; ART = antiretroviral therapy

Figure 3

Trajectories of the estimated number of new HIV infections in Ukraine with implementation of three intervention scenarios for MSM without (a) and with ART expansion among the adult population (b) (2011–2016). y-Axes do not start at zero new infections for the purpose of ensuring differences in new infections are visible over time. MSM = men who have sex with men; ART = antiretroviral therapy

Figure 4

Trajectories of the estimated number of new HIV infections in Kenya with implementation of three intervention scenarios for MSM without (a) and with ART expansion among the adult population (b) (2011–2016). y-Axes do not start at zero new infections for the purpose of ensuring differences in new infections are visible over time. MSM = men who have sex with men; ART = antiretroviral therapy

Figure 5

Trajectories of the estimated number of new HIV infections in Thailand implementation of three intervention scenarios for MSM without (a) and with ART expansion among the adult population (b) (2011–2016). y-Axes do not start at zero new infections for the purpose of ensuring differences in new infections are visible over time. MSM = men who have sex with men; ART = antiretroviral therapy

Figure 6

Sensitivity analysis of reduction in new infections among MSM with increased coverage of interventions for MSM and equal access to expanded ART, 2012–2016. MSM = men who have sex with men; ART = antiretroviral therapy

Table 5

Projected estimates of the number of new adult HIV infections in Peru with implementation of three intervention scenarios for MSM (2011–2016)

	Status quo	Scenario 1	Scenario 2	Status quo	Scenario 1	Scenario 2
	(without expansion of ART)			(with expansion of ART)
2011	6583	6583	6583	6666	6666	6666
2012	6636	6542	6450	6494	6402	6312
2013	6710	6517	6329	6310	6126	5948
2014	6764	6469	6187	6122	5849	5589
2015	6798	6400	6026	5934	5573	5236
2016	6816	6313	5850	6038	5572	5146
Cumulative	40,307	38,824	37,425	37,564	36,188	34,897
% reduction	(ref)	3.7	7.2	(ref)	3.7	7.1
Infections averted		1483	2882		1376	2667

MSM = men who have sex with men

Table 6

Projected estimates of the number of new adult HIV infections in Ukraine with implementation of three intervention scenarios for MSM (2011–2016)

	Status quo	Scenario 1	Scenario 2	Status quo	Scenario 1	Scenario 2
	(without expansion of ART)			(with expansion of ART)
2011	45,589	45,589	45,589	45,589	45,589	45,589
2012	45,914	45,841	45,769	44,544	44,474	44,404
2013	46,001	45,859	45,720	43,530	43,395	43,264
2014	45,873	45,666	45,467	42,553	42,361	42,176
2015	45,596	45,328	45,075	41,513	41,269	41,039
2016	45,211	44,886	44,584	39,841	39,555	39,289
Cumulative	274,184	273,169	272,204	257,570	256,643	255,761
% reduction	(ref)	0.4	0.7	(ref)	0.4	0.7
Infections averted		1015	1980		927	1809

MSM = men who have sex with men

Table 7

Projected estimates of the number of new adult HIV infections in Kenya with implementation of three intervention scenarios for MSM (2011–2016)

	Status quo	Scenario 1	Scenario 2	Status quo	Scenario 1	Scenario 2
	(without expansion of ART)			(with expansion of ART)
2011	102,756	102,756	102,756	102,756	102,756	102,756
2012	103,347	103,213	103,081	92,435	92,312	92,191
2013	105,643	105,357	105,082	81,619	81,384	81,157
2014	108,160	107,715	107,294	75,500	75,159	74,839
2015	110,685	110,074	109,508	74,216	73,762	73,344
2016	113,120	112,337	111,630	72,504	71,939	71,432
Cumulative	643,711	641,452	639,351	499,030	497,312	495,719
% reduction	(ref)	0.4	0.7	(ref)	0.3	0.7
Infections averted		2259	4360		1718	3311

MSM = men who have sex with men

Table 8

Projected estimates of the number of new adult HIV infections in Thailand with implementation of three intervention scenarios for MSM (2011–2016)

	Status quo	Scenario 1	Scenario 2	Status quo	Scenario 1	Scenario 2
	(without expansion of ART)			(with expansion of ART)
2011	25,636	25,636	25,636	25,636	25,636	25,636
2012	25,342	25,178	25,016	22,653	22,503	22,356
2013	24,735	24,397	24,070	20,384	20,098	19,821
2014	23,927	23,416	22,929	18,816	18,400	18,004
2015	23,014	22,331	21,692	17,522	16,980	16,475
2016	22,063	21,208	20,425	16,255	15,595	14,993
Cumulative	144,717	142,166	139,768	121,266	119,212	117,285
% reduction	(ref)	1.8	3.4	(ref)	1.7	3.3
Infections averted		2551	4949		2054	3981

MSM = men who have sex with men; ART = antiretroviral therapy

Projections from Peru show that much higher coverage of interventions for MSM will be needed to change the trajectory of the HIV epidemic among MSM in Peru, if ART is maintained at current coverage levels (Figure 2). Increasing MSM-specific interventions and providing HIV-positive MSM equal access to and treatment with ART may also lead to a 15% reduction in new infections (4600 averted) among MSM. Beyond ART, MSM interventions may avert 2600–2800 infections between 2012 and 2016, almost a 10% reduction in new infections among MSM. The expansion of MSM-specific interventions with and without additional expansion of ART for adults, may also contribute an additional 2700–2900 infections averted among the adult population (a reduction of 7%) within the five-year time span (Table 5).

In Ukraine, increasing coverage of MSM-specific interventions demonstrates an impact on the HIV epidemic among the general population in later years and may avert between 1800 and 2000 new adult infections between 2012 and 2016, a 1% reduction in addition to the benefits observed with expansion of ART (Table 6). Among MSM, there is general decline in new infections; this decline is augmented by the expansion of ART and further enhanced through scale-up of MSM interventions (Figure 3). MSM interventions, with and without ART expansion, may lead to a cumulative 1400–1600 infections averted among MSM or a reduction of 8% within five years beyond benefits provided by ART. The combined effect of increasing MSM interventions and allowing equal access to expanded ART may lead to a 14% reduction in new infections, or 2700 infections averted, among MSM within five years.

There has been a great expansion of ART among the adult population in Kenya and this expanded coverage is planned to continue into the future and, thus far, has provided obvious impact for the adult and MSM populations. The expansion of ART among adults also shows demonstrable impacts among the MSM population (Figure 4) and, in combination with the MSM interventions may result in a 25% reduction in new infections among MSM, or almost 10,000 infections averted within five years. MSM specific interventions may avert 3000 new MSM infections (9% reduction) over the five-year period, in addition to the benefits provided by ART. Almost 4000 new infections may be averted if ART does not increase beyond current coverage levels (10% reduction in new infections). MSM-specific interventions positively impact the general population in Kenya and, while relative impact may be modest, the numbers of infections averted are considerable averting almost 3300 new adult infections when ART is expanded and more (almost 4400) when ART coverage is maintained at 2011 levels (a reduction of approximately 1% of new infections) (Table 7).

The epidemic projections of new infections in Thailand show an increase over time, particularly when MSM-specific interventions and ART coverage is maintained at 2011 levels. Like Peru, the increase in coverage of MSM-specific interventions could change the trajectory of the epidemic among MSM by 2016 compared (Figure 5). Among MSM, approximately 3600–4500 infections may be averted, a 9–10% reduction in new infections, by MSM interventions when they are scaled-up without and with expansion of ART coverage among adults. The combined impact of ART expansion and MSM interventions may result in a cumulative 10,300 infections averted, or a 22% reduction in new infections among MSM from 2012 to 2016. Among adults, in addition to the benefits observed with the expansion of ART, almost 4000–5000 new infections may be averted (3% reduction) within five years when MSM interventions are scaled up (Table 8).

DISCUSSION

These modelling projections show the combined benefit for MSM of providing equal access to ART as it expands according to national strategies and increasing coverage of targeted MSM interventions. The greatest impacts among MSM that are associated with the combined access to ART and expanded MSM interventions are observed in Kenya and Thailand, likely a result of greater coverage of ART. The projections also demonstrate additional impact of MSM interventions among both the MSM and the general populations when such interventions are applied within epidemic scenarios. Greatest impacts of these MSM interventions on the adult epidemics are observed among countries such as Peru and Thailand, where the attributable fraction of HIV transmission is predominantly related to male same sex behaviour. While the percent reductions may appear to be low, likely due to the short five-year time frame of the projections, the absolute numbers of infections averted are significant and are likely to continue increasing as the interventions continue in the future.

The attributable fraction can be used to allocate HIV prevention resources and target programmes on the basis of the contribution of risk groups to local HIV incidence; however, it is important that these estimates be examined in relation to all other risk groups contributing to the total number of HIV. Countries that have actively sought to understand the epidemic and prevent transmission among MSM within the overall epidemic scenario, such as Thailand and Peru, demonstrate benefits for MSM and the general population.⁴² Even in countries where positive social changes are observed, research suggests that national health strategies need to follow and provide appropriate resources for prevention among MSM. Caceres’ work in Peru, for example, shows that despite evidence that HIV transmission occurs predominantly among MSM, the majority of funding has been allocated to prevention for other vulnerable populations among which the prevalence of HIV is lower than that observed among MSM.^32,42

Though Kenya has a generalized epidemic, the estimated proportion attributable to MSM suggests that HIV infection among MSM in Kenya is emerging as a significant contributor. Sanders et al. ¹⁵ recently reported an HIV prevalence of 43.0% among exclusive MSM and 12.3% among men who have sex with men and women, highlighted the vulnerability of these populations. It is important to note that the projections in this study also rely on available epidemiological and behavioural data of MSM; in a stigmatizing environment that currently exists for MSM in Africa, data are particularly difficult to obtain and, thus, these projections may underestimate the attributable fraction and the potential impacts of HIV interventions for MSM. Addressing this HIV epidemic among MSM at an early stage may contribute to reductions in incidence for this particularly hidden group.

The role of injecting drug use in the HIV epidemic cannot be ignored, and IDU may face both parenteral and sexual risk exposures. Ukraine, with an estimated HIV prevalence of 1.1% among the adult population (2010), now has the greatest prevalence of HIV in Eastern Europe Central Asia⁴³ and effective strategies for HIV prevention among the general population and at-risk groups are needed. Strathdee et al. ⁴⁴ estimated that the risk of HIV infection attributable to unprotected sex among IDUs ranged from 15% to 45% in Odessa, Ukraine. This estimate, however, is based on heterosexual transmission and more data are needed to understand the roles and interactions among IDU and MSM. Our projections take into consideration the overlapping sexual and parenteral risks among MSM-IDU. The projections demonstrate the impact of MSM risk reduction interventions on this population, as a subgroup of the MSM population, and highlight the importance of MSM interventions to also access other networked risk group and suggest that, in a similar manner, interventions targeted for IDU may also impact a portion of the MSM population. These findings agree upon the need for developing strategies that effectively target and prevent HIV transmission among at-risk populations.

Critical to effectively decreasing transmission of HIV among MSM and the general population in all countries is enabling full access to ART for HIV positive MSM. Since the HPTN 052 results reported by Cohen et al. ⁷ have confirmed observational study findings, the importance of ART as prevention further supports the modelling projections reported here. The impacts observed with expansion of ART, however, should not be interpreted as an argument for expanding ART only; rather, the synergies of ART expansion in combination with MSM interventions are important. MSM interventions, particularly outreach, are a method to access hidden MSM, or MSM-IDU, populations for HIV counselling and testing and linkage to ART care.

Both assets and methodological challenges are associated with the Goals model, as highlighted by a recent assessment of mathematical models to evaluate health programmes.⁴⁵ While Goals is unique in its ability to project impacts based on varying coverage rates, it assumes that the intervention effects are additive and does not currently accommodate interaction effects. This is acceptable if there is agreement that the interaction effects are multiplicative (because then the model estimates would be lower bound estimates of the true impact); but may be problematic if interventions have synergistic effects. A key component of the Goals model is an impact matrix that indicates how behavioural changes when a group of people are exposed to specific prevention interventions. We did not include oral chemoprophylaxis with Truvada, as this intervention has only been tested in a single randomized control trial at the time of writing; however, given recent results,⁴⁶ further analysis is warranted to understand this use and its impact on general and at-risk populations. As the Goals model continues to be refined, estimates of impact should be updated and efforts to validate the projections should be made. As with any model, these projections rely on data availability and quality, and is particularly challenging where underreporting may occur due to pervasive stigma towards MSM. This work may be subject to reporting and publication bias.

Finally, more understanding of the costs associated with such interventions are required before policy-makers can make informed decisions, particularly as LMICs are often faced with limited financial resources and have to make difficult decisions about how to spend these resources among competing priorities.

Taken together, our findings suggest that responding to HIV among MSM is an important aspect of overall country-level responses, and this is true across diverse epidemic scenarios. Whether such levels can be achieved is a question of commitment, political will, and a willingness to provide resources for MSM. Continuing to provide current inadequate levels of services to these men will mean that HIV epidemic control will be difficult or impossible to achieve.

Footnotes

Appendix

References

Baral

, Sifakis

, Cleghorn

, Beyrer

. Elevated risk for HIV infection among men who have sex with men in low and middle income countries 2000–2006. PLoS Med 2007;4:e339

International Lesbian Gay Bisexual Trans and Intersex Association. Map: Male-to-Male Relationship, Punishments 2010. 2010. See http://ilga.org/ilga/en/index.html (cited 20 June 2010)

Beyrer

, Baral

, Walker

, Wirtz

, Johns

, Sifakis

. The expanding epidemics of HIV-1 among men who have sex with men in low and middle income countries: diversity and consistency. Epidemiol Rev 2010;32:137–51

Stover

, Bollinger

, Cooper-Arnold

, The Futures Group International. Goals Model: For estimating the Effects of Resource Allocation Decisions on the Achievement of the Goals of the HIV/AIDS Strategic Plan. Glastonbury, CT: SAGE Publications, 2003

Futures Institute. Goals manual: a model for estimating the effects of interventions and resource allocation on HIV infections and deaths. Glastonbury, CT, 2011

Johnson

, Diaz

, Flanders

, Behavioral interventions to reduce risk for sexual transmission of HIV among men who have sex with men. Cochrane Database Syst Rev 2008;3:CD001230

Cohen

, Chen

, McCauley

, Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med 2011

Eshleman

, Hudelson

, Redd

, Analysis of genetic linkage of HIV from couples enrolled in the HIV Prevention Trials Network 052 trial. J Infect Dis 2011;204:1918–26

Amjadeen

, Kaschenkova

, Konoplytska

, Report on the survey “Monitoring behaviours of MSM as a component of second-generation surveillance”. International AIDS Alliance. Kiev; 2005. Available at: http://www.aidsalliance.org.ua/cgi-bin/index.cgi?url=/en/library/our/index.htm

10.

Balakiryeva

, Bondar

, Kasyanczuk

. Report on the survey ‘Monitoring behaviours of men having sex with men as a component of second generation surveillance’. Kiev: SAGE Publications, 2008

11.

Centers for Disease Control and Prevention (CDC). HIV prevalence among populations of men who have sex with men – Thailand, 2003 and 2005. Morb Mortal Wkly Rep 2006;55:844–8

12.

International HIV/AIDS Alliance in Ukraine. Monitoring behaviour of men having sex with men: analytical report on the results of the research. Kyiv: SAGE Publications, 2008

13.

MINSA. Estadistica: Indicadores de salud. Ministerio de Salud del Peru, 2009

14.

Sanchez

, Lama

, Peinado

, High HIV and ulcerative sexually transmitted infection incidence estimates among men who have sex with men in Peru: awaiting for an effective preventive intervention. J Acquir Immune Defic Syndr 2009;51(Suppl. 1):S47–51

15.

Sanders

, Graham

, Okuku

, HIV-1 infection in high risk men who have sex with men in Mombasa, Kenya. AIDS 2007;21:2513–20

16.

Hierholzer

, Montano

, Hoelscher

, Molecular epidemiology of HIV type 1 in Ecuador, Peru, Bolivia, Uruguay, and Argentina. AIDS Res Hum Retroviruses 2002;18:1339–50

17.

Angala

, Parkinson

, KIlonzo

, Natecho

, Taegtmeyer

. Men who have sex with men (MSM) as presented in VCT data in Kenya (Abstract no. MOPE0581). AIDS 2006-XVI International AIDS Conference; 2006; Toronto, 2006

18.

Van Griensven

, Thanprasertsuk

, Jommaroeng

, , Bangkok MSM Study Group. Evidence of a previously undocumented epidemic of HIV infection among men who have sex with men in Bangkok, Thailand. AIDS 2005;19:521–6

19.

van Griensven

, Varangrat

, Wimonsate

, Trends in HIV prevalence, estimated HIV incidence, and risk behavior among men who have sex with men in Bangkok, Thailand, 2003–2007. J Acquir Immune Defic Syndr 2009

20.

Aldridge

, Iglesias

, Caceres

, Miranda

. Determining a cost effective intervention response to HIV/AIDS in Peru. BMC Public Health 2009;9:352

21.

UNAIDS, Ministry of Health of Ukraine. Ukraine: National report on monitoring progress toward the UNGASS Declaration of Commitment on HIV/AIDS. Reporting period: January 2006–December 2007. Kiev, 2008

22.

International HIV/AIDS Alliance in Ukraine. Analytical report: estimation of the size of populations most-at-risk for HIV infection in ukraine in 2009. Kyiv: SAGE Publications, 2010

23.

Kruglov

, Kobyshcha

, Salyuk

. The most severe HIV epidemic in Europe: Ukraine's national HIV prevalence estimates for 2007. Sex Transm Infect 2008;84(Suppl. 1):i37–41

24.

Baral

, Kizub

, Masenior

, Male sex workers in Moscow, Russia: a pilot study of demographics, substance use patterns, and prevalence of HIV-1 and sexually transmitted infections. AIDS Care 2010;22(1):112–8

25.

WHO Regional Office for Europe. HIV prevalence and risks among men who have sex with men in Moscow and Saint Petersburg. Copenhagen, 2007

26.

Amirkhanian

, Kelly

, Issayev

. AIDS knowledge, attitudes, and behaviour in Russia: results of a population-based, random-digit telephone survey in St Petersburg. Int J STD AIDS 2001;12:50–7

27.

Sifakis

, Peryskina

, Sergeev

. Rapid Assessment of HIV infection and associated risk behaviors among men who have sex with men in Russia. [Abstract MOPE0396]. XVII International AIDS Conference. Mexico City, Mexico, 2008

28.

Kelly

, Amirkhanian

, McAuliffe

, HIV risk characteristics and prevention needs in a community sample of bisexual men in St. Petersburg, Russia. AIDS Care 2002;14:63–76

29.

PSI Russia. HIV/AIDS TRaC Study of Risk, Health-seeking Behaviors, and Their Determinants, Among Men Who Have Sex with Men in Eight Regions of the Russian Federation. Second Round. Moscow: SAGE Publications, 2008

30.

Tyusova

, Verevochkin

, Heimer

. HIV in population groups recruited by respondent-driven sampling in St. Petersburg. [Abstract]. III Eastern Europe and Central Asia AIDS Conference. Moscow, Russia, 2009

31.

Onyango-Ouma

, Birungi

, Geibel

. Understanding the HIV/STI Risks and prevention needs of men who have sex with men in Nairobi, Kenya. Nairobi, 2005

32.

Ugarte Ubilluz

, Arce Rodriguez

, Acosta Saal

, Suarez Ognio

. Informe nacional sobre los progresos realizados en la aplicacion del ungass periodo: Enero 2008 - Diciembre 2009. Lima, 2010 30 Marzo

33.

UNAIDS, WHO. Epidemiological fact sheets on HIV and AIDS, 2008 Update – Peru. Report. Geneva: UNAIDS/WHO Working Group on Global HIV/AIDS and STI Surveillance, 2008 December 2008

34.

Family Health International, Bureau of AIDS TB and STIs, Department of Disease Control MOPH Thailand. The Asian Epidemic Model (AEM) Projections for HIV/AIDS in Thailand: 2005–2025: The Thai Working Group on HIV/AIDS Projections (2005), 2008

35.

Wimonsate

, Chaikummao

, Tongtoyai

, Kittinunvorakoon

, Sriporn

, Varangrat

. Successful start of a preparatory HIV cohort study among men who have sex with men (MSM) in Bangkok, Thailand: preliminary baseline, follow-up and HIV incidence data (Abstract MOAC0105). XVII International AIDS Conference Mexico City, 2008

36.

Beyrer

, Sripaipan

, Tovanabutra

, High HIV, hepatitis C and sexual risks among drug-using men who have sex with men in northern Thailand. AIDS 2005;19:1535–40

37.

Kunawararak

, Beyrer

, Natpratan

, The epidemiology of HIV and syphilis among male commercial sex workers in northern Thailand. AIDS 1995;9:517–21

38.

Chariyalertsak

, Kosachunhanan

, Saokhieo

, HIV incidence, risk factors, and motivation for biomedical intervention among gay, bisexual men, and transgender persons in Northern Thailand. PLoS One 2011;6:e24295

39.

National AIDS Prevention and Alleviation Committee (Thailand). UNGASS Country Progress Report: Thailand (Period January 2008- December 2009). Bangkok: National AIDS Prevention and Alleviation Committee, 2010

40.

WHO. Antiretroviral therapy for HIV infection in adults and adolescents: 2010 revision. Geneva, 2010

41.

Beyrer

, Wirtz

, Walker

, Johns

, Sifakis

, Baral

. The global HIV epidemics among men who have sex with men. Washington, DC, 2011. The World Bank. Available at: http://siteresources.worldbank.org/INTHIVAIDS/Resources/375798-1103037153392/MSMReport.pdf

42.

Caceres

, de Zalduondo

, Hallett

, Avila

, Clayton

. Combination HIV prevention: Crafting a new standard for the long-term response to HIV. International AIDS Conference; 2010 July 2010; Vienna, 2010

43.

UNAIDS. UNAIDS Report on the Global AIDS Epidemic: 2010. Geneva, 2010

44.

Strathdee

, Hallett

, Bobrova

, HIV and risk environment for injecting drug users: the past, present, and future. Lancet 2010;376:268–84

45.

Garnett

, Cousens

, Hallett

, Steketee

, Walker

. Mathematical models in the evaluation of health programmes. Lancet 2011;378:515–25

46.

Grant

, Lama

, Anderson

, Preexposure chemoprophylaxis for HIV prevention in men who have sex with men. N Engl J Med 2010;363:2587–99

47.

Bollinger

. How can we calculate the ‘E’ in ‘CEA’? AIDS 2008;22(Suppl. 1):S51–7