A Systematic Review of Behavioral Interventions to Prevent HIV Infection and Transmission among Heterosexual, Adult Men in Low-and Middle-Income Countries
- First Online:
- Cite this article as:
- Townsend, L., Mathews, C. & Zembe, Y. Prev Sci (2013) 14: 88. doi:10.1007/s11121-012-0300-7
- 694 Views
Prevention of new HIV infections needs to move to the forefront in the fight against HIV and AIDS. In the current economic crisis, low- and middle-income countries (LMICs) should invest limited resources to amass reliable evidence-based information about behavioral prevention efforts, and on behaviors that are driving the epidemic among people who are engaging in those behaviors. This paper aims to provide a systematic review and synthesis of behavioral interventions among a group of people in high HIV-burden countries: heterosexual men in LMICs. The review includes articles published between January 2001 and May 2010 that evaluated behavioral prevention interventions among heterosexual males aged 18+ years in LMICs. The studies were evaluated using the quality assessment tool for quantitative studies developed by the Effective Public Health Practice Project. The review identified 19 articles that met the review's inclusion criteria. Most studies were conducted in South Africa (n=6); two each in Uganda and Thailand; and one in each of Angola, Brazil, Bulgaria, India, Nigeria, the Philippines, Russia, Ukraine and Zimbabwe. Eight of 19 interventions increased condom use among their respective populations. Those interventions that sought to reduce the number of sexual partners had little effect, and those that addressed alcohol consumption and intimate partner violence had mixed effects. There was no evidence for any specific format of intervention that impacted best on any of the targeted risk behaviors. The paucity of evaluated interventions for heterosexual men in LMICs suggests that adult men in these countries remain underrepresented in HIV prevention efforts.
KeywordsSystematic reviewHeterosexual menHIV behavioral interventionsLow- and middle-income countries
For every two people who started HIV treatment in 2007, five were newly infected with the virus globally (UNAIDS 2009a). In 2008 the world saw 2.7 million people becoming newly infected with HIV, the majority of which occurred in the low-and middle-income countries (LMICs) of sub-Saharan Africa (1.9 million) (UNAIDS 2009a). If the global community has any chance of reversing the global HIV epidemic, a concerted effort needs to be made to stop all new infections, or at least, decrease them sharply (Global HIV Prevention Working Group 2007).
“By delivering comprehensive HIV prevention to those who need it—the right interventions focused on the right people at the right scale—half of all new infections projected to occur by 2015 could be averted” (Global HIV Prevention Working Group 2007: p. 1). However, the challenge in the current economic crisis is to allocate HIV prevention resources appropriately and thus get the best possible impact from any investment in HIV prevention (UNAIDS 2009b). Another challenge is to amass reliable evidence-based information about behavioral prevention efforts that have had an effect, or not. Prevention intervention practitioners and policy makers could use this information to make informed decisions about implementation of behavioral prevention interventions that have been shown to work, and relinquish or modify those that have had limited or no effect. More importantly, resource-constrained LMICs need to invest the limited prevention resources on behaviors that are driving the epidemic, and among the people who are engaging in those behaviors. For many LMICs, where HIV prevalence is highest (for example, in sub-Saharan Africa where 71 % of all new HIV infections occurred in 2008 (UNAIDS 2009a)), one such group of people is heterosexual men.
More than 10 years ago, Exner and colleagues published a review of risk reduction interventions conducted among heterosexual men in the United States (Exner et al. 1999). At the time, Exner noted the strong focus of risk-reduction on heterosexual women and the limited focus on heterosexual men, suggesting they comprise a “forgotten group.” These authors argued that “heterosexually active men be included in strategic efforts to reduce heterosexual transmission because sexual behavior is dyadic and men are the [sexual] partners of women” (Exner et al. 1999: p. 348). A recent systematic review of behavior change interventions among women living in LMIC was motivated by the complete lack of a summary of behavioral intervention for HIV prevention in the developing world (McCoy et al. 2010). This paper aims to fill important gaps in the review literature by providing a systematic review and synthesis of behavioral interventions among an underrepresented group in high HIV-burden countries: heterosexual men in LMIC. The review also aims to provide a forum for cross-national and cross-discipline sharing of information that will have the potential to inform and improve HIV prevention, research and collaboration efforts in other parts of the world. To satisfy these aims the review will present strong science derived from consistent, objective rating criteria applied to the studies included in the review and will also be accessible to professional service providers who are not necessarily academics, but who are at the forefront of HIV prevention efforts internationally.
We searched the following databases for articles reporting on a HIV/AIDS and/or STI, behavioral prevention interventions that fit the review criteria: PubMed, PsycInfo, PsycArticles, SocINDEX, Academic Search Premier, Science Direct, The Centers for Disease Control and Prevention’s (CDC) HIV/AIDS Prevention Research Synthesis Project’s compendium of HIV Prevention Interventions with Evidence of Effectiveness (Centers for Disease Control and Prevention 2001); and the Cochrane HIV/AIDS Group reviews and protocols. Keywords for the search and permutations thereof included HIV/AIDS; sexually transmitted infections/STI/sexually transmitted diseases/STD; prevention; intervention; adults; men/male; control; comparison; trial. The Boolean phrase “not” was included where possible to exclude articles reporting on studies conducted among men who have sex with men (MSM)/gay/transgender; women/females; and youth/adolescents/children. We hand-searched the reference lists of all sourced and eligible articles. We contacted study authors when clarification was needed to decide eligibility of studies for the review; for example, the composition of samples with respect to sexual orientation and/or gender.
Inclusion and Exclusion Criteria
The review includes studies that evaluated behavioral prevention interventions aiming to reduce HIV and/or STI risk behaviors among heterosexual adult males residing in LMICs globally, according to the World Bank income criteria (World Bank 2010). The review includes studies that: 1) included heterosexual males aged 18+ years; or a majority (≥75 %) of 18+ year old males if younger aged males were included in the sample; or a majority (≥75 %) of males if females were included in the sample; 2) were conducted in LMICs; 3) that evaluated an intervention group/condition relative to a control or comparison group/condition and included between-group analysis of the data; 4) reported outcome data on at least one HIV-related risk behavior; and 5) were published in English-language, peer-reviewed journals from January 2001 to May 2010.
We excluded studies that did not report intervention outcome data separately for gender or control for gender in the analysis if > 25 % females were included in the sample. Similarly, we excluded studies that had ≥15 % of men who reported having male sexual partners within their sample of males.
Studies that met the inclusion criteria
First author Intervention setting
Assessment [compensation paid]
EPHPP Global Rating Weakness/es
Outcomes: Intervention effects (control/comparison vs intervention)
ONE-ON-ONE, INDIVIDUAL LEVEL COUNSELLING INTERVENTIONS
Cornman et al. (2008)
Standard care (n=49): 2, HIV, ARV’s, adherence & nutrition counselling sessions during routine clinic care every ±3 months. Free condoms available nationally.
Izindlela Zokuphila/Options for Health (n=103): 15-minute patient-centered discussions with counsellor during routine clinic care every ±3 months. Based on IMB principles, including motivational interviewing. Free condoms available nationally.
Pre-intervention, 6 month follow-up
No gender differences
KwaZulu Natal, South Africa
152, 18+ year old men (43%) and women (mean age: 34y) in HIV clinic care
[US$20 per month for all visits]
Weakness/es: Validity/reliability of assessment tools not reported
Condom use: Estimated mean number of unprotected sex acts in previous 3 months among men (baseline: 2.32 vs 2.19; 6 months: 3.85 vs 0.11, time X condition interaction: p<.05).
Kalichman, et al. (2007a)
HIV Education (n=74):
HIV & alcohol risk reduction
Pre-intervention, 3 & 6 months follow-ups
Controlling for gender:
Cape Town, South Africa
143 men (n=122) & women, mean age 29.3y, attending a STI clinic
single, 20-minute counselling session. Free condoms available nationally.
(n=69): single, 60-minute HIV & alcohol risk reduction behavioral skills counselling session. Based on IMB principles. Free condoms available nationally.
[US$12 for baseline; US$15 each follow-up]
Condom use: 1) % in previous month (baseline: 58.6 vs 64.8; 3 months: 78.4 vs 90.5, p<.05; 6 months: 76.4 vs 87.8, p<.05). 2) % at last sex (baseline: 71 vs 74; 3 months: 84 vs 98, OR 10.5; CI 1.2, 90.4; p<.01; 6 months: 82 vs 96, OR 5.3; CI 1.0,2.9; ns)
Number of partners: Mean number in previous month (baseline: 2.8 vs 2.5; 3 months: 2.0 vs 1.9, ns; 6 months: 2.5 vs 1.6, ns).
Alcohol use: Mean number of times used in sexual contexts in previous month (baseline: 4.5 vs 3.7; 3 months: 3,4 vs 1.5, p<.01; 6 months: 1.2 vs 2.1, ns).
Pechansky et al. (2007)
Standard (n=62): NIDA standard pre-& post HIV counselling. No condoms provided.
“Thought mapping” (n=57): NIDA standard pre-& post HIV counselling + thought mapping & structured stories. Grounded in transtheoretical model of health behavior change. No condoms provided.
Pre-intervention, 2 & 8 week follow-ups
Condom use: During vaginal sex. No estimates reported (adjusted IRR 1.19; CI 1.01,1.40; p=.002).
Porto Alegre in Southern Brazil
119 men, 18+ years
Weakness/es: Pilot, quasi-experimental design; potential for selection bias
Number of partners: In the previous 30 days. No estimates reported (adjusted IRR 1.01; CI 0.78,1.31; p=.923).
Samet et al. (2008)
Narcology hospital patients:
Standard addiction treatment including HIV testing & counselling (n=87). Condoms provided.
PREVENT (n=94): 2-session counselling including HIV testing, + 3, monthly telephone booster sessions. Based on RESPECT harm-reduction model. 30 condoms provided at baseline.
Pre-intervention, 3 (by telephone) & 6 months follow-ups
Condom use: No estimates reported. 1) Median difference in % of safe sex episodes in previous 3 months. (unadjusted: 12.7%, p=.01; adjusted: 23%, p=.07). 2) Any condom use in previous 3 months (unadjusted OR 2.5, CI 1.1,5.5, p=.03; adjusted OR 3.7; CI 1.5,8.9; p=0.004).
St. Petersberg, Russia
181, 18+ year old men (75%) and women (median age: 30y)
[US$5 at baseline & 3 m, US$30 at 6 months]
Simbayi et al. (2004)
STI clinic patients:
Single, 20-minute, didactic HIV information/education session (n=114). Free condoms available nationally.
Single, 60-minute, HIV information/education, motivational, & communication skills-building session (n=114). Based on IMB model. Free condoms available nationally.
Pre-intervention, 1 & 3 month follow-ups
Controlling for gender:
Cape Town, South Africa
228 men (n=151) & women with repeat STD diagnoses (mean age: 27.5 years)
[US$10 at baseline; US$12 each follow-up]
Weakness/es: Validity/reliability of assessment tools not reported
Condom use: 1) Mean number unprotected sex acts in past 3 months (baseline: 7.0 vs 6.6; 3 months: 4.6 vs 1.5, p<.05). 2) % unprotected in past 3 months (baseline: 40.9 vs 31.6; 3 months: 23.6 vs 13.3, p<.01).
Number of partners: Mean number in past 3 months (baseline: 2.2 vs 2.3; 3 months: 1.4 vs 2.2, ns).
SMALL GROUP INTERVENTIONS
Bing et al. (2008)
HIV prevention (n=280): 5 daily sessions including
Condom use: % who increased condom use during vaginal sex (3 months: 20 vs 27; p<.01; 6 months: 30 vs 20, ns).
568 men, aged 18–51 years (mean: 29y), from 12 military bases
(n=288): 5 daily sessions including 1-hour HIV prevention. Condoms provided.
1-hour malaria prevention + 5 optional, monthly, 1-hour booster sessions. Based on IMB principles. Condoms provided.
3 & 6 month follow-ups
Number of partners: Mean number in past 3 months (baseline: 0.4 vs 0.3; 3 months: 0.3 vs 0.2, ns; 6 months: 0 .2 vs 0.1, ns).
Alcohol use: Mean frequency before sex in past month (baseline: 1.3 vs 1.3; 3 months: 1.3 vs 1.2, ns; 6 months: 1.2 vs 1.2, ns).
Cornman et al. (2007)
Information-only (n=125): Single, didactic HIV/AIDS prevention information workshop. Condoms provided.
Truckers’ Health Project (n=125): Single, 4-hour, IMB-based group workshop. Based on IMB principles. Condoms provided.
Pre- & post-intervention (IMB constructs); 10 month follow-up (sexual behaviors)
Condom use: Mean number of times in previous 4 months with marital partners (baseline: 0.59 vs 0.77;
250, 18+ years men
[US$6–7 for each assessment]
Weakness/es: No assessment of potential, confounders
10 months: 0.56 vs 3.40, p<.001); with non-marital partners (baseline: 2.31 vs 2.17; 10 months: 2.01 vs 4.6, p<.001).
Number of partners: Mean number of non-marital part-ners in previous 4 months (10 months: 18.81 vs 15.91, p<.001).
Exner et al. (2009)
Urban community members:
(n=132): 1, ½-day didactic workshop. No condoms provided.
(n=149): 2, 5-hour group workshops & monthly, 2-hour check-in sessions. Based on participatory curriculum development. No condoms provided.
Pre-intervention, 3 month follow-up
Condom use: 1) % of any unprotected sex in past
281, 18–73 year old men (median age: 38y)
[US$2 for screening]
Weakness/es: Quasi-experimental design
3 months (3 months: 53.9 vs 28.3, controlling for baseline and controlling for age and baseline, p<.01). 2) at last sex with main partner (3 months: 47.8 vs 78.5, controlling for baseline, p<.001 and controlling for age and baseline, p<.01).
Jewkes et al. (2008)
Rural & peri-urban community members:
Control intervention (35 clusters; n=1467): a single 3 hour session on HIV, safer sex & condoms. Free condoms available nationally.
Stepping Stones (35 clusters; n=1409):13 single-sex, 3-hour peer group sessions, 3 mixed sex meetings over 6–8 weeks. Based on adult education theory. Free condoms available nationally.
Pre-intervention, 1 & 2 year follow-ups
Eastern Cape Province, South Africa
2776, 15–26 year old men (n=1360; >66%, 18+ years) & women in 70 villages & townships
[US$3 for each follow-up assessment]
Condom use: Mean % at last sex (baseline: 48.6 vs 44.6; 12 months: 68.9 vs 73.5, p=.16; 24 months: 75.1 vs 73.2, p=.43).
Number of partners: Mean number in past 12 months (baseline: not reported; 12 months: 2.51 vs 2.28, p=.06; 24 months: 2.39 vs 2.15, p=.12).
Alcohol use: % problem use (baseline: 25.7 vs 24.5; 12 months: 26.5 vs 19.8, p=.021; 24 months: 25.7 vs 26.6, p=.56).
IPV: % of >1 episode since last visit (baseline: 14.5 vs 14.5; 12 months: 14.9 vs 11.4, p=.099; 24 months: 9.6 vs 6.2, p=.054).
Kalichman et al. (2008)
Township community members:
HIV-alcohol education (n=183): 1, 1-hour alcohol information/education group. Free condoms available nationally.
HIV-alcohol risk reduction (n=170): 1, 3-hour group consisting of HIV-alcohol education + motivation & behavioral skills. Based on social cognitive model of behavior change & IMB principles. Free condoms available nationally.
Pre-intervention, 3 & 6 month follow-ups
No differences X gender
Cape Town, South Africa
353, 18+ year old men (n=117) & women (mean: 34.1y) who drink alcohol
[US$12 for baseline; US$24 at 6 month follow-up]
Condom use: 1) % in previous month among heavier drinkers (baseline: 57.7 vs 52.8; 3 months: 66.6 vs 59.3, p<.05; 6 months: 63.2 vs 58.7, ns) 2) % in previous month among lighter drinkers (baseline: 56.7 vs 43.8; 3 months: 52.0 vs 77.5, p<.05; 6 months: 65.7 vs 72.2, ns).
Number of partners: 1) % reporting 2+ in prior 3 months among heavier drinkers (baseline: 26 vs 20; 3 months: 12 vs 18, p<.05; 6 months: 14 vs 14, ns). 2) % reporting 2+ in prior 3 months among lighter drinkers (baseline: 15 vs 12; 3 months: 16 vs 6, p<.05; 6 months: 15 vs 8, ns).
Alcohol use: 1) Mean number of times before sex in previous month among heavier drinkers (baseline: 531 vs 5.1; 3 months: 2.3 vs 3.5, p<.05; 6 months: 2.8 vs 2.4, p<.05). 2) Mean number of times before sex in previous month among lighter drinkers (baseline: 2.9 vs 3.9; 3 months: 2.1 vs 1.4, p<.05; 6 months: 2.5 vs 1.9, p<.05).
Kalichman et al. (2009)
Township community members:
Alcohol/HIV prevention intervention (n=233): Single, 3-hour group session. Free condoms available nationally.
Gender-based violence/HIV (GBV/HIV) prevention intervention (n=242): 5, integrated gender-based violence and HIV risk-reduction group sessions. Based on social cognitive theory & IMB principles. Free condoms available nationally.
Pre-intervention, 1, 3 & 6 month follow-ups
Condom use: Mean number unprotected sex acts in past 3 months (baseline: 4.6 vs 7.2; 3 months: 1.7 vs 3.9, p<.01; 6 months: 3.1 vs 2.5, ns).
Cape Town, South Africa
475 (mean age: 30.2y) men
[US$15 at each assessment]
Weakness/es: Potential for selection bias
Number of partners: Mean number in past 3 months (baseline: 1.5 vs 1.6; 1 month: 1.4 vs 1.5, ns; 3 months: 1.4 vs 1.5, ns; 6 months: 1.4 vs 1.6, p<.05)
Alcohol use: Mean number of times before sex in past 3 months (baseline: 4.5 vs 6.6; 3 months: 2.6 vs 3.6, ns; 6 months: 3.6 vs 3.4, ns).
GBV: 1) % lost temper with woman among those who had (1 month: 13 vs 23, p<.01; 3 months: 15 vs 18, ns; 6 months: 14 vs 23, p<.01). 2) Hit a sex partner (1 month: 19 vs 23; 3 months: 29 vs 36, ns; 6 months: 61 vs 31, p<.01).
Kajubi et al. (2005)
Peri-urban community members:
Comparison (n=201): increase condom availability & brief AIDS informational presentation. Condoms provided via coupon redemption.
Condom promotion (n=297): increase condom availability & 1 of 8, 3-hour technical use condom skills workshops. Condoms provided via coupon redemption.
Pre-intervention & 6 month follow-up
Condom use: 1) % consistent in past 6 month with any partner (baseline: 46.7 vs 40.9: 6 months: 60.6 vs 59.6, p=.57). 2) % consistent in past 6 months with casual partner (baseline: 43.0 vs 35.7: 6 months: 40.6 vs 40.4, p=.33).
498, 18–30 year old men residing in 2 peri-urban communities
Weakness/es: Reliability/validity of assessment tool not reported
Number of partners: Mean number in past 6 months (baseline: 2.20 vs 2.13; 6 months: 2.03 vs 2.44, p=.004).
PEER EDUCATION/INDIGENOUS LEADER INTERVENTIONS
Booth et al. (2009)
Injecting drug users:
HIV counseling & education (C&E) (n=900): pre-& post test HIV counselling. No condoms provided.
Indigenous leader outreach model (ILOM) (n=898): HIV counseling & education + individualized risk reduction over 5m by outreach workers. Based on peer education principles. No condoms provided.
Pre-intervention, 6 month follow-up
Condom use: % unprotected sex in the past 30 days (baseline: 49.6 vs 56.4, p<.05: 6 months: 40.0 vs 43.5, ns).
Kiev, Odessa & Makeevka/Donesk, Ukraine
1798 (76% men), 18+ years (mean: 29.5y)
[US$5 at baseline; US$6 at follow-up]
Weakness/es: Cross-over experimental study design
Number of partners: % reporting multiple partners in the past 30 days (baseline: 26.6 vs 24.6; 6 months: 18.9 vs 21.1, p<.05).
Kelly et al. (2006)
Impoverished Roma (gypsy) community members:
Control (26 networks; n=137): 15-minute individual counselling on HIV risk reduction. Condoms provided.
Social Network Intervention (26 networks; n=145): 15-minute individual counselling on HIV risk reduction + counselling and advice on reducing HIV risk behavior for network leaders. An indigenous leader outreach model. Condoms provided.
Pre-intervention, 3 & 12 month follow-ups
Condom use: 1) % unprotected vaginal sex with casual partners (baseline: 43 vs 51; 3 months: 37 vs 26, p=.08; 12 months: 40 vs 28, p=.07). 2) % unprotected vaginal sex with >1 partner (baseline: 30 vs 32; 3 months: 26 vs 15, p=.03; 12 months: 29 vs 15, p=.01). 3) % unprotected vaginal sex with multiple partners (baseline: 39 vs 43; 3 months: 29 vs 16, p=.03; 12 months: 26 vs 13, p=.01).
286 men (mean age 19.7 years) in 52 social networks
Latkin et al. (2009)
Injecting drug users: (Thailand only) 427 men (84%; n=357) & women recruited by 182 eligible network indexes
Individual HIV testing & counselling (n=213 including 91 indexes) with risk reduction counselling at each follow-up. No condoms provided.
Peer education (n=214, + 91 indexes): Individual HIV testing & counselling with risk reduction counselling at each follow-up + 6, 2-hour group peer educator sessions + 2 booster sessions to indexes only. Based on peer education principles. No condoms provided.
Pre-intervention, 6, 12,18, 24 & 30 month follow-ups
In Thailand only
Chiang Mai, Thailand (& Philadelphia, U.S.A.)
Weakness/es: Validity/reliability of assessment tools not reported
Condom use: No estimates reported.1) Any unprotected sex in past week (OR 1.03; CI 0.81;1.30, p=.83). 2) No condom use past week (OR 1.06; CI 0.82,1.35, p=.67).
Number of partners: No estimates reported. Multiple partners in past month (OR 1.02; CI 0.77,1.35, p=.90).
Morisky et al. (2004)
No expose (n=1570). [cross-over study design: control participants received intervention after intervention group]. No condoms provided.
Peer counsellor intervention (n=1819): STI/HIV/AIDS prevention, distribution of collaboratively-developed IEC materials. Based on participatory curriculum development. No condoms provided.
Pre- & post-intervention, 6 month follow-up
Condom use: 1) % ever used (baseline: 32.63 vs 36.10; post-test: 33.79 vs 38.70, p<.01; 6 months: 38.72 vs 46.31, p<.01). 2) % at last sex (baseline: 10.71 vs 10.62; post-test: 11.08 vs 13.36, p<.05; 6 months: 11.81 vs 18.65, p<.01).
3389 men, in 6 sites (mean age: 34.7y)
Weakness/es: Quasi-experimental design; no assessment of potential confounders
Sherman et al. (2009)
Life skills (n=488 including 206 indexes): 7, 2-hour, bi-weekly, small group sessions for indexes only. Condoms provided.
Peer education (n=495 including 209 indexes): 7, 2-hour, bi-weekly, small group sessions + 2 booster sessions for indexes only. Based on peer education principles. Condoms provided.
Pre-intervention, 3, 6, 9 & 12 month follow-ups
No differences by gender:
Chiang Mai, Thailand
983 men (75%) & women, 18–25 years (median age 19y),
[US$5 each assessment; US$5 each session]
Weakness/es: Validity/reliability of assessment tools not reported
Condom use: % always in past 3 months (baseline: 28.8 vs 34.2; 3 months: 45.3 vs 47.7, ns; 6 months: 40.1 vs 44.9, ns; 9 months: 40.0 vs 44.8, ns; 12 months: 39.5 vs 40.5, ns).
LARGE-SCALE, COMMUNITY OUTREACH INTERVENTIONS
Gregson et al. (2007)
Rural community members:
Standard Government services (46%=male): syndromic STI management, condom distribution and social marketing, home based care, limited HIV-focused IEC activities.
Interventions (45%=male): Community & clinic-based: workplace & community peer education; condom distribution among CSW & clients; income-generating projects for women; strengthening management of STIs & health center open-days. Based on peer education principles & included IEC model of behavior change.
Pre-intervention, 3 year follow-up
Manicaland Province, Eastern Zimbabwe
11980 men aged 17–54 & women aged 15–44 residing in 12 communities (from a population-based, closed cohort survey)
Weakness/es: High dropout/withdrawal
Condom use: 1) % unprotected sex with casual partners in past 3 years (follow-up: 73 vs 79, p=.039). 2) % unprotected sex with regular partners in past 3 years (follow-up: 86 vs 88, p=.975).
Number of partners: 1) % with multiple new partners in past year (follow-up: 10 vs 11, p=.513). 2) % with >2 regular partners in past 3 years (follow-up: 10 vs 8, p=.025).
Kamali et al. (2003)
Rural community members
Comparison: Routine government health services & community development activities (6 communities; n=6742). Social marketing of male condoms.
1) IEC: Information, education & community activities (6 communities; n=6918)
Pre-intervention, 2 follow-ups during a median of 3.6 years
No differences by gender:
Control vs IEC
Masaka district in SW Uganda
±15000 men (45%) & women, 13+ years (± 75% were 20+ years) in 18 communities
2) IEC & STI: Information, education & community activities + improved STI management (6 communities; n=6856). Based on IEC model of behavior change & social marketing of male condoms.
Condom use: % with last casual partner (baseline: 23 vs 21; 1st follow-up: 49 vs 60; 2nd follow-up: 66 vs 73, for both follow-ups: p=.057).
Number of partners: % with >2 in past year (baseline: 11 vs 12; 1st follow-up: 15 vs 17; 2nd follow-up: 14 vs 16, for both follow-ups: ns).
IEC vs IEC & STI:
Condom use: % with last casual partner (baseline: 21 vs 20; 1st follow-up: 60 vs 50; 2nd follow-up: 73 vs 56, for both follow-ups: p=.036).
Number of partners: % with >2 in past year (baseline: 12 vs 12; 1st follow-up: 17 vs 17; 2nd follow-up: 19 vs 16, for both follow-ups: ns).
Each of the studies was evaluated and rated using the quality assessment tool for quantitative studies developed by the Effective Public Health Practice Project (EPHHP) (Effective Public Health Practice Project 2009). This tool provides pertinent questions with which to evaluate studies in the following domains: 1) The potential for selection bias was assessed by means of two questions about the representativeness of the sample to the target population and refusals to participate in the intervention. 2) Study designs considered strong include randomized controlled trials and controlled clinical trials. Those rated moderate are cohort analytic (two group pre-and post-intervention assessment), case–control, cohort (one group pre-and post-intervention assessment) and interrupted time series studies. Any other form of study design is considered “weak” according to EPHHP. 3) The potential for confounding was identified through questions about differences between groups prior to the intervention and how many confounders were controlled. 4) The adequacy of blinding referred to whether outcome assessor/s were aware of the intervention or exposure status of participants and/or whether participants were aware of the research question. 5) Data collection methods and instruments were evaluated according to whether the validity and/or reliability was reported. 6) The potential for bias as a result of withdrawal and/or dropout was determined by the percentage of participants who withdrew and/or dropped out of the intervention or assessments. 7) The integrity of the intervention was evaluated from reports on the percentage of participants who received the intervention, whether fidelity to the intervention protocol was measured and whether participants may have received an unintended or co-intervention (i.e., contamination). 8) Analyses were assessed by providing information about the unit of allocation, the unit of analysis, the appropriateness of statistical methods and whether the analysis followed an intention-to-treat method.
With the aid of a quality assessment tool dictionary (available from http://www.ephpp.ca/Tools.html), ratings (strong, moderate, weak) were allocated to studies for each of the first six domains from which a global rating was then obtained. The global ratings are “strong” if none of the domains were rated “weak;” “moderate” if one of the domains was rated as “weak;” and “weak” if two or more of the domains were rated “weak.” Table 1 provides the EPHHP-derived global rating, and the domain/s identified as “weak.”
The review identified 19 HIV/AIDS/STI studies that fit the inclusion criteria. Most of studies were conducted in South Africa (n = 6), with two each in Uganda and Thailand, and one in each of Angola, Brazil, Bulgaria, India, Nigeria, the Philippines, Russia, Ukraine and Zimbabwe. The majority reported on community-based interventions (n = 9) where participants were drawn from the general population. Specifically, people resided in peri-urban, urban and rural communities in Nigeria, South Africa, Zimbabwe, Uganda and the Philippines (Exner et al. 2009; Gregson et al. 2007; Jewkes et al. 2008; Kajubi et al. 2005; Kamali et al. 2003; Morisky et al. 2004), townships in South Africa (Kalichman et al. 2008, 2009), and a settlement in Bulgaria (Kelly et al. 2006). Six studies reported on interventions conducted with people who were alcohol or drug users, or in drug treatment programs in Ukraine, South Africa, Thailand, Brazil and Russia (Booth et al. 2009; Kalichman, et al. 2007a; Latkin et al. 2009; Pechansky et al. 2007; Samet et al. 2008; Sherman et al. 2009), and the remaining five included clinic patients in South Africa (n = 2) (Cornman et al. 2008; Simbayi et al. 2004), military personnel in Angola (n = 1) (Bing et al. 2008) and truck drivers in India (n = 1) (Cornman et al. 2007). A combined total of 39,887 people were included in the interventions, of whom 24,113 (60.5 %) were men. Of these men, 0.2 % (n = 58) were men who had sex with men and women in the 3 months prior to baseline assessment (Kelly et al. 2006). There were no reports of participants who were men who had sex with men only. Just under half of the studies reported on interventions implemented among heterosexual men only (8 of 19).
Intervention exposure ranged from 15-minute single sessions with individual participants to 13, single- and mixed-sex, peer-group sessions. Six comparison conditions comprised standard of care (for example, HIV testing and pre-, post-counselling); nine comprised the same topic as the intervention conditions albeit implemented with less intensity; three consisted of different intervention topics (for example, an alcohol/HIV prevention intervention vs. a gender-based violence/HIV prevention intervention); and one had no comparison intervention. This last study was a cross-over design and comparison participants received the intervention at a later stage. Three studies’ final follow-up assessments were conducted less than 5 months post intervention; ten were conducted between 6 and 11 months post intervention; and six were conducted 12 or more months post intervention.
All interventions reported on changes in condom use between participants in the two conditions at follow-up/s; 13 reported on numbers of sexual partners; 5 assessed problem alcohol use or alcohol use in contexts of sexual acts; and 2 examined intimate partner violence (IPV). All studies provided a detailed and comprehensive description of both their respective intervention and control conditions. Two studies had “weak” global EPHPP ratings (Morisky et al. 2004; Pechansky et al. 2007); seven were rated as “strong” (Bing et al. 2008; Jewkes et al. 2008; Kamali et al. 2003; Kalichman et al. 2007a, 2008; Kelly et al. 2006; Samet et al. 2008); and the remainder were rated “moderate.”
One-On-One, Individual Level Counselling Interventions
Five studies evaluated one-on-one, individual-level counselling interventions. Three articles reported on randomized clinical trials in South Africa. One intervention was conducted among HIV-infected hospital outpatients (Cornman et al. 2008); and two were conducted among sexually transmitted infection (STI) clinic patients (Kalichman et al. 2007a; Simbayi et al. 2004). One article reported on a quasi-experimental clinical trial conducted in Brazil among cocaine injectors and crack smokers (Pechansky et al. 2007). Another article reported on a randomized controlled trail among patients with alcohol and/or heroin dependence in two substance abuse treatment, narcology hospitals in St. Petersburg (Samet et al. 2008).
Condoms were either supplied or were freely available to study participants in four of the five interventions. Monetary compensation was provided to all participants for baseline and follow-up assessments in four studies.
These studies used different measures of condom use: mean number of unprotected sex acts in the prior 3 months (Cornman et al. 2008; Simbayi et al. 2004); percentage of participants using condoms in the previous 30 days and at last sex (Kalichman et al. 2007b); median difference in percentage of safe sex acts and any condom use in the past 3 months (Samet et al. 2008); percentage of participants not using condoms in the previous 3 months (Simbayi et al. 2004); and condom use during vaginal sex at time of assessment (Pechansky et al. 2007). Despite the differing measures of condom use and length of follow-up assessment times and irrespective of the number of or interval between follow-up assessments, all five interventions were successful in increasing condom use among the respective intervention participants.
Measures of partner numbers in the three studies that measured this behavior were varied. One reported the mean number of sexual partners in the previous 30 days (Kalichman et al. 2007a); another reported the mean number of sexual partners in the previous 3 months (Simbayi et al. 2004); and Pechansky et al. (2007) reported median numbers in the past 30 days. None of the three interventions was successful in decreasing sexual partner numbers (Kalichman et al. 2007a; Pechansky et al. 2007; Simbayi et al. 2004).
One intervention that aimed to impact on alcohol use showed short-term positive effects at 3 months that were not sustained at 6 months (Kalichman et al. 2007a).
Small Group Interventions
Seven studies reported on interventions that were conducted in small groups. Four articles reported on randomized controlled trials among military personnel in Angola (Bing et al. 2008), long distance truck drivers in India (Cornman et al. 2007), rural and peri-urban community members in South Africa (Jewkes et al. 2008), and township community members in South Africa (Kalichman et al. 2008). One article reported on a quasi-experimental, proof-of-concept study among urban community members in Nigeria (Exner et al. 2009); another reported on a quasi-experimental field trial among township community members in South Africa (Kalichman et al. 2008); and a controlled community trail was conducted among peri-urban community members in Uganda (Kajubi et al. 2005).
Condoms were supplied or freely available to study participants in six of the seven interventions. Monetary compensation was provided for assessments in five interventions.
There was no consistency in condom use indicators among these seven interventions: the percentage of participants who increased use (Bing et al. 2008); mean number of times used with marital and non-marital partners in the previous 4 months (Cornman et al. 2007); the percentage of any non-use in the last 3 months and condom use at last sex with a main partner (Exner et al. 2009); the percentage of participants who used condoms at last sex (Jewkes et al. 2008); the percentage of light or heavy alcohol drinkers who used condoms with any partner in the previous month (Kalichman et al. 2008); the mean number of unprotected sex acts in the past 3 months (Kalichman et al. 2009); and the percentage of participants who used condoms consistently with any partner in the past 6 months (Kajubi et al. 2005).
Intervention impacts were mixed. Two interventions were able to show positive intervention effects at a 10-month follow-up among truck drivers in India (Cornman et al. 2007), and at a 3-month follow-up among urban community members in Nigeria (Exner et al. 2009). Three interventions showed short-term (at 3 months) but not long-term (at 6 months) positive effects among military personnel in Angola and urban township residents in South Africa (Bing et al. 2008; Kalichman et al. 2008; 2009); and two had no effect on condom use at 12 and 24 months among rural and peri-urban community members in South Africa (Jewkes et al. 2008) and at 6 month follow-up among peri-urban community members in Uganda (Kajubi et al. 2005).
Five of the seven small-group interventions sought to impact on sexual partner numbers. Partner number indicators were not consistent across the studies. For example, three reported on mean number of any type of partner in the past 3 months (Bing et al. 2008; Kalichman et al. 2009), past 12 months (Jewkes et al. 2008); one reported on the mean number of non-marital partners in the past 4 months (Cornman et al. 2007); and another on two or more partners in the prior 3 months (Kalichman et al. 2008). Intervention effects with respect to partner numbers were varied. Two studies showed no effects at 3-and 6-month follow-ups (Bing et al. 2008) and at 12-month follow-up (Jewkes et al. 2008); another showed positive effects at 10-month follow-up (Cornman et al. 2007); one showed short-term (at 3 months) but not long-term (at 6 months) effects (Kalichman et al. 2008) and long-term (at 6 months) but not short-term (at 1 and 3 months) effects (Kalichman et al. 2009).
Four interventions aimed to impact on alcohol use. Three studies reported on participants’ use of alcohol before sex (Bing et al. 2008; Kalichman et al. 2008; 2009) and one measured problem alcohol use by means of the AUDIT scale (Jewkes et al. 2008). Only one of these interventions showed positive intervention effects that were sustained at two follow-up assessments at 3 and 6 months (Kalichman et al. 2008). Two studies reported no intervention effects (Bing et al. 2008; Kalichman et al. 2009) and one showed short-term (at 12-month follow-up) but not long-term (at 24-month follow-up) effects (Jewkes et al. 2008).
One study showed long-term effects on gender-based violence (i.e. at 6-month follow-up) (Kalichman et al. 2009) and another showed positive changes in measures of intimate partner violence that were strengthened over time (i.e., at 24-month follow-up) (Jewkes et al. 2008).
Peer Education/Indigenous Leader Interventions
Five studies reported on peer education/indigenous leader formats for their interventions. Two studies had cross-over experimental designs, one of which was conducted among injecting drug users (IDU) in Ukraine (Booth et al. 2009) and the other among heterosexual male clients of commercial sex workers (CSW) in the southern part the Philippines (Morisky et al. 2004). The remaining three were randomized control trails conducted in Bulgaria among impoverished Roma (Gypsy) men (Kelly et al. 2006), in Thailand among IDU (Latkin et al. 2009) and methamphetamine users (Sherman et al. 2009).
Two interventions provided condoms to study participants. Compensation was provided for baseline and follow-up assessments in two interventions. In addition, participants in Sherman et al’s (2009) intervention were provided compensation for each of the seven intervention sessions attended.
With respect to condom use, this format for interventions was successful among Roma gypsies (Kelly et al. 2006) and the clients of sex workers (Morisky et al. 2004). Continued increases in condom use over follow-up assessments was found in these two interventions despite differing measures of condom use: percentage of participants reporting unprotected sex with casual partner, with more than one partner or with multiple partners (Kelly et al. 2006), and percentage of participants who had ever used condoms (Morisky et al. 2004). The same cannot be said for the interventions among drug users (Booth et al. 2009; Latkin et al. 2009; Sherman et al. 2009). None of these studies reported positive effects on condom use among IDU in Ukraine (Booth et al. 2009) and among IDU (Latkin et al. 2009) and methamphetamine users (Sherman et al. 2009) in Thailand. These studies also used different measures of condom use: percentage of participants reporting unprotected sex in the past 30 days (Booth et al. 2009); any unprotected sex in the past week (Latkin et al. 2009) and percentage of participants reporting consistent condom use in the past 3 months (Sherman et al. 2009).
While Booth et al. (2009) reported significant reductions in the percentage of participants reporting multiple partners in the past month among their intervention IDU, Latkin et al. (2009) reported no effects on this, similarly-measured behavior.
Large-Scale, Community Outreach Interventions
Intervention effects in the rural communities in Uganda showed an increase in condom use with a most recent casual sexual partner; but no significant reduction in sexual partner numbers in the past year (Kamali et al. 2003). The intervention conducted in Zimbabwe showed increased condom use with casual, but not regular partners, and a reduction in the number of people having more than two sexual partners in the past 3 years (Gregson et al. 2007).
This review sourced articles from 19 LMICs where a variety of behavioral HIV prevention interventions were conducted among diverse populations or communities of heterosexual men. On the whole, the majority of the interventions reported positive behavior changes with respect to condom use among their respective populations. However, those interventions that sought to reduce the number of sexual partners had little effect, and those that addressed alcohol consumption had mixed effects. The two interventions that sought to impact on IPV were largely successful.
Correct and consistent use of condoms is one of the most reliable methods to prevent sexual transmission of HIV (Hearst and Chen 2004; Shelton 2006). The effectiveness of condoms in preventing HIV transmission or acquisition has been estimated to be approximately 90 % (Hearst and Chen 2004). It is not unexpected that all of the 19 studies included in this review measured the effects of their respective interventions on condom use among their populations.
Despite the heterogeneity of settings and target populations, the variety of ways in which condom use was measured and irrespective of whether condoms were made available to intervention participants or not, five of nine interventions that reported positive intervention effects on condom use comprised individual-level counselling strategies (Cornman et al. 2008; Kalichman et al. 2007a; Pechansky et al. 2007; Samet et al. 2008; Simbayi et al. 2004). Four of these interventions were based on IMB principles. Two of the other four were small group interventions; one incorporating IMB principals (Cornman et al. 2007) and the other involving community participation in intervention curriculum development (Exner et al. 2009). The other two were an indigenous leader intervention (Kelly et al. 2006) and a peer education intervention that included the peer educators in IEC material development (Morisky et al. 2004). Evidence from these studies suggests that individual-level counselling strategies based on IMB principles might be most appropriate to ensure positive effects on condom use behavior. Although based on evidence from two studies only, involving peer educators and/or the community at large in intervention content and/or curriculum development might also be advantageous and could be used to design interventions in other settings and among other risk populations. Both these strategies enhance the acceptability and applicability of an intervention among the target population that likely impacts positively on retention in an intervention and persistent behavior change. Furthermore, the use of indigenous leaders to implement interventions might mitigate the distrust that participants often feel towards outsiders.
Three interventions that reported short-term positive effects on condom use that were not sustained at later follow-up, irrespective of partner type comprised small group formats based on IMB principles (Bing et al. 2008; Kalichman et al. 2008; 2009). While this evidence suggests that small-group workshops based on IMB principles might not be effective in impacting on condom use behaviors in the longer term, some individual study limitations may have accounted for the deterioration in effects over time in these studies. Simultaneous improvements in the control groups’ condom use behavior due to a national HIV prevention radio program airing at the time of the intervention and up-scaled HIV prevention activities in the study areas (Bing et al. 2008), and the similarity in HIV prevention activities in the control and interventions arms in Kalichman et al’s (2009) study, likely contributed to the apparent deterioration in behavior change over time among intervention participants in these two studies. One other intervention’s positive effects depended on partner type and comprised a large-scale, community outreach intervention (Gregson et al. 2007). The authors of this study suggested a number of limitations that may have impacted on the intervention effects: possible contamination between intervention communities; another intervention being simultaneously implemented in some sites; negative role-modelling of peer educators; and secular trends characteristic of a mature, stabilizing HIV epidemic in Zimbabwe (Gregson et al. 2007). Despite these limitations, condom use with casual sexual partners increased significantly over time, while condom use with regular partners did not. The fact that this effect was found 3 years after intervention activities took place suggests that large-scale, community-based interventions may be appropriate for impacting on condom use in the long term with casual sexual partners. Future such interventions should consider stronger messaging about condom use with regular partners including addressing (mis)perceptions of faithfulness and trust, and thus safety that underpin non-condom use in regular partner relationships (Halperin and Epstein 2004; McPhail and Campbell 2001; Richards et al. 2008).
Three of the six interventions that showed no effect on condom use used peer education strategies among injecting drug users (Booth et al. 2009; Latkin et al. 2009) and methamphetamine users (Sherman et al. 2009). While this evidence suggests that peer education strategies might not be suitable for promoting condom use, particularly among drug users, some possible alternative explanations for the negative findings must be noted. Firstly, the use of peer educators in intervention activities is based on the assumption that people who have similar experiences, concerns, lifestyles and cultural backgrounds are likely to be more effective teachers than outsiders. Thus while peer educators may be completely at ease when discussing and attempting to impact on risky drug-use behaviors, they may not find discussing sensitive and personal issues involving sexual behavior/s as easy. Thus peer educators might have avoided or limited discussions about condom use. It would be prudent for future interventions to provide peer educators with the necessary skills to openly discuss topics involving sexual encounters, irrespective of the intervention’s primary objectives. Secondly, of the 13 interventions that reported positive effects on condom use, 12 supplied or had condoms available to study participants. All three interventions among drug users that had no effect on condom use did not supply or have condoms available to study participants. This suggests either that this omission, or that drug users face particular difficulties in accessing condoms may have affected condom use outcomes. Future interventions with drug users should consider supplying participants with condoms and perhaps address possible personal and/or structural barriers to condom access. Thirdly, a national drug policy known as the “war on drugs” was operating in Thailand while Latkin et al’s (2009) and Sherman et al’s (2009) interventions were being implemented, which likely negatively affected the willingness of individuals to report drug use and possibly other risk behaviors such as condom use to study personnel whom they may not have trusted.
Three other interventions that showed no effect on condom use despite having condoms provided as part of the intervention or freely available, were conducted among peri-urban (Kajubi et al. 2005) and rural populations (Jewkes et al. 2008; Kamali et al. 2003). This finding suggests that peri-urban and rural populations might face particular challenges when attempting to use condoms consistently despite having easy access to condoms. However, given the evidence of condom use behavior change deterioration noted above, the finding of no intervention effects in two studies (Jewkes et al. 2008; Kamali et al. 2003), might have more to do with the long post-intervention intervals (between 12 and 36 months) to follow-up assessments of condom use than poor intervention performance. Future interventions among rural populations should consider having shorter intervals between assessments of condom use to determine whether short-term effects are evident. This information could inform future intervention content and/or the possible use of “booster” sessions to maintain positive behavior change. They should also address personal and/or structural barriers to consistent condom use. Furthermore, while condom skills are necessary for correct and thus safe condom use, findings from Kajubi et al’s study (2005) suggests that skills alone do not translate into greater use of condoms.
Despite Hearst and Chen’s (2004) conclusion that there was no evidence, anywhere in the world, to suggest that a generalized HIV epidemic has been reversed as a result of increased/consistent condom use, these studies’ interventions to increase correct condom use were largely effective. Thirteen of the 19 studies were able to demonstrate some positive impact on condom use. Similarly, a number of other reviews of studies conducted internationally found positive intervention effects for condom use in a variety of settings (Lyles et al. 2007) and heterosexual populations (Neumann et al. 2002), STI clinic attenders in the United Kingdom (Ward et al. 2005), PLWHA (Crepaz et al. 2006), Hispanics (Herbst et al. 2007), Black and Hispanic STI patients (Crepaz et al. 2007), African Americans (Darbes et al. 2008) in the U.S. and people living in and sub-Saharan Africa and Asia (Foss et al. 2007). There does, however, need to be sensitivity to the challenges some populations may face in accessing condoms and/or consistently using them. Irrespective of the target population, findings from this review suggests that the supply of condoms should be considered as a component of any intervention. Notably, three of the four interventions that did not supply or have condoms available, did not demonstrate positive effects on condom use. Furthermore, in South Africa men who had multiple sexual partners were found to use condoms when they were readily available and easily accessible (Townsend et al. 2010). Even if we accept that increased condom use may not reverse an epidemic, in combination with other (positive) behavior changes, this behavioral change is likely to impact on the spread of HIV, and should continue to be the cornerstone of HIV prevention efforts both globally and in LMICs.
Multiple Sexual Partners
Although not incontrovertible (Lurie and Rosenthal 2010), one of the risk behaviors thought to be a significant factor in the spread and persistence of HIV in southern Africa is multiple, concurrent sexual partners (MCP): a behavior that has received a great deal of recent attention and debate (Epstein 2008, 2010; Green et al. 2009; Halperin and Epstein 2004; Lurie and Rosenthal 2010; Mah and Halperin 2010a, b; Morris 2010). MCP has been inconsistently defined and measured in the literature to date (Nelson et al. 2007), but is generally considered to be instances where an individual has two or more sexual relationships that overlap in time (UNAIDS 2009c). Although having multiple sexual partners does not necessarily mean that they are also concurrent, it does point to high partner turnover, the possibility of concurrency, and concomitant risk for HIV.
Findings from the 13 studies whose interventions targeted and measured numbers of sexual partners (not necessarily concurrent), suggest that they had very little effects on this behavior. Of the four studies that reported positive intervention effects with respect to partner numbers, two were small group interventions (Cornman et al. 2007; Kajubi et al. 2005), one a peer education intervention (Booth et al. 2009), and another a large-scale community outreach intervention (Gregson et al. 2007). This suggests that no particular format of intervention delivery is any more effective than another in impacting on this risk behavior.
Two studies reported mixed intervention effects with one reporting short-term (at 3 months) but not longer-term effects (at 6 months) (Kalichman et al. 2008), and the other reporting long-term effects (at 6 months) that were not evident at first follow-up (at 3 months) (Kalichman et al. 2009). Kalichman et al’s (2008) intervention comprised a single group session based on IMB principles, and differences between the intervention and control group were moderated by levels of alcohol consumption among participants. Compared to heavy drinkers, light drinkers showed greater intervention effects. Kalichman et al’s (2009) intervention consisted of five group sessions that was also based on IMB principles. This evidence suggests that group interventions that comprise multiple sessions and based on IMB principles show promise for impacting on partner numbers. More intense efforts among heavy alcohol consumers might be necessary to impact partner numbers and other risk behaviors.
The remaining seven interventions that failed to impact on sexual partner numbers are heterogeneous with respect to setting, target populations, and intervention content and format. It is thus difficult to provide any conclusions about why they may have failed, and to make any recommendations for future interventions focusing on this behavior. As noted above, few differences between intervention and comparison groups in Bing et al’s (2008) study might be due to other HIV prevention efforts operating in the study communities during intervention implementation, rather than the intervention itself. Peer educators’ possible avoidance of matters relating to sexual encounters, also noted above, might have accounted for Latkin et al’s (2009) negative findings. Of note, however, is that all three of the one-on-one, individual-level counselling interventions that sought to impact on partner numbers failed suggesting that this format might not be effective in reducing sexual partner numbers among people in LMICs.
Examining intervention effects across the different countries, of note is that five of the six interventions implemented in South Africa that aimed to impact on partner numbers showed no intervention effects. In South Africa where multiple (often concurrent) sexual partners is normative, implicitly acknowledged and tolerated (Leclerc-Madlala 2008; Mah 2008; Mah and Maughan-Brown 2009) men might be resistant to changing this behavior, irrespective of the intervention format. Future interventions in this country or any setting where multiple sexual partners is normative should consider addressing social norms with regard to partner numbers rather than, and/or while simultaneously trying to impact on individual behavior.
Of the five studies that reported on their respective interventions’ effects on alcohol use, findings are mixed. In one intervention, alcohol use in the context of sexual encounters sustained a decrease over time among local South African community members (Kalichman et al. 2008). Two interventions demonstrated short-term reductions in alcohol use in sexual contexts (Kalichman et al. 2007a) and problem alcohol use (Jewkes et al. 2008) that were not sustained at later follow-ups. Two other interventions reported no reductions in alcohol use before last sex (Bing et al. 2008; Kalichman et al. 2009). Because the control condition in this last-mentioned intervention also focused on alcohol reduction, it is perhaps not unexpected that their findings did not show any significant reductions in alcohol use.
Given these mixed findings and the variety of ways in which alcohol use/problem alcohol consumption was measured, recommendations for future interventions focusing on this behavior is difficult. Four of the five interventions were small-group interventions. Of these one showed positive effects, two found no effect, and one demonstrated short-term, but not long-term effects on alcohol use. It is unlikely that these varying effects can be attributed to poor quality studies as none had a weak EPHPP rating. Given the plethora of research evidence that links problem alcohol use and alcohol use in the context of sexual encounters (Baliunas et al. 2010; Fisher et al. 2007; Kalichman et al. 2007b; Shuper et al. 2009), the paucity of interventions targeting this risky behavior and the inconclusive evidence of their effectiveness is lamentable. Furthermore, the fact that light drinkers showed greater reductions in risk behaviors compared to heavy drinkers in Kalichman et al’s (2008) study suggests that more intensive efforts to reduce alcohol consumption are needed.
Other Risk Behaviors
There is increasing and convincing evidence that intimate partner violence (IPV)/gender-based violence (GBV) is a major risk factor for HIV (Jewkes and Morrell 2010; Jewkes et al. 2010). It is thus not unexpected that addressing violence against women is one of the nine priority areas identified by the UNAIDS Outcome Framework for 2009–2011 (UNAIDS, 2009d). Only two studies’ interventions sought to address IPV/GBV among rural and semi-rural community members (Jewkes et al. 2008) and local community members in South Africa (Kalichman et al. 2009). Only one of these interventions was successful, showing a long-term rather than short-term effect on this behavior: losing one’s temper with and hitting a woman (Kalichman et al. 2009). The other showed a sustained (albeit not significant) decrease over time (Jewkes et al. 2008). These interventions were conducted among relatively disparate populations (rural and semi-rural, and urban informal populations), and the theoretical grounding of each was quite different. Interventions targeting IPV and GBV may have wide acceptability in different geographical locations and thus populations, irrespective of the type of intervention, particularly in South Africa.
Provision of Compensation to Study Participants
Because 11 interventions paid varying amounts of cash compensation to their participants at baseline and follow-up assessments, and 1 intervention paid participants for each of seven intervention workshops (Sherman et al. 2009) some comment on the possible effects this may have had on intervention effects is appropriate. The payment of compensation to study/intervention participants may impact on motivations to participate in interventions and the social desirability of responses to assessment questions (Noguchi et al. 2007; Talbot 1999). Participants who enrol in interventions may be motivated by the prospect of receiving monetary “reward” rather than a desire to benefit from the intervention or assist with the intervention research (Talbot 1999). These concerns are particularly relevant in LAMICs and especially in sub-Saharan Africa, where large portions of intervention participants are most likely poor, and where monetary “rewards” might impact most on motivations and responses.
If payment of compensation biased samples and/or responses to behavioral questions to those motivated more by monetary reward than a desire to benefit from the intervention or assist with the study, then one would have expected that those interventions that did provide compensation would have reported largely positive intervention effects across all outcomes. On the contrary, intervention effects were not consistent across all outcomes as a function of whether compensation was paid or not. Furthermore, there were very few differences in condom use or partner number outcome measures as a function of whether compensation had been paid to participants or not. While recognizing the differences in study populations, intervention content and format, and measures of outcomes, this evidence suggests that among heterosexual adult males in LMICs, the payment of compensation is empirically acceptable.
This review has provided a synthesis of behavioral interventions among heterosexual men in LMICs. It has also provided an evaluation of the interventions and studies so that interventionists looking to impact on their respective populations’ high-risk behaviors are aware of the strengths and weaknesses of those in this review. A limitation which is not unique to this review is the possibility of inadvertently excluding some reports on interventions that may not have been included in the resources that were searched.
Several limitations among the studies that were included in the review need mention. First, all had short post-intervention, follow-up assessments rarely stretching beyond 12 months, and longer-term post-intervention assessments among high-risk groups may have shown more positive effects. Indeed it has been suggested that assessments at 5 years or more are more likely to show significant population-level effects (Hallett et al. 2007). Second, contamination between intervention participants, which may have helped explain limited, negative or contrary results was not addressed by the majority of studies. Third, many studies did not report on refusals to participate in, and dropout from intervention session/s. These studies might have had a high risk of bias due to refusals and attrition, which was not addressed and which might have contributed to weak evidence. Fourth, behavioral interventions in many of the studies required interviewer-administered questionnaires and are therefore at risk of detection bias. Taken together, these limitations indicate that conclusions drawn should be viewed with some caution. A further limitation that has implications for the possible replicability of effective interventions in resource-constrained LMICs is that none of the studies provided a cost analysis of their respective interventions. Strengths of the studies are the inclusiveness with which they reported the content of their respective intervention components, the detailed description of the ways in which participants were allocated to the intervention conditions, a clear description of attrition during post-intervention assessments among those studies that provided this information, and the reporting of baseline data.
While we accept that there is a possibility of inadvertently excluding some reports on interventions that may not have been included in the resources that were searched, we were able to source only 19 appropriate studies reported over a period of almost a decade. Clearly more interventions for heterosexual males in LMICs are needed. Even if we acknowledge that there might be interventions being implemented among heterosexual men in LMICs, but not reported or reported in the largely difficult to access “grey” literature, it would seem that heterosexual men remain an under-served group (Exner et al. 1999).
Further, there is a distinct need for interventions to address multiple sexual partners and alcohol use. The interventions seeking to change these latter two risk behaviors showed limited effectiveness that in many instances was not sustained at follow-ups. Future studies evaluating such interventions would be well advised to strive for consistent partner number indicators and alcohol use outcomes. For example, numbers of sexual partners during a consistently applied and specific time period is needed, as is the need to disaggregate numbers of partners into main partners (spouses or steady girlfriends), casual long-term partners outside of main relationships, and one-night stands. Similarly, a consistent measure of alcohol use needs to be adopted; for example, either an overall problem alcohol consumption measure or a measure of alcohol use in the context of sex. A great deal more effort needs to go into developing interventions to address IPV. Although based on very limited evidence from this review of studies, interventions addressing IPV may be welcomed in a range of target populations. It is recommended that, irrespective of the target population, HIV risk reduction interventions should include the supply of condoms to all participants.
In addition to these recommendations, some tentative conclusions can be drawn from the review’s findings. First, individual-level counselling strategies based on IMB principles might be most appropriate when attempting to impact on condom use. Peer education strategies might not be an appropriate intervention delivery format for impacting on this behavior among drug users. There appears to be some advantage to involving peer educators and/or the community at large in the development of intervention materials and curricula. Second, the IMB model implemented in group settings, and large-scale and community-based interventions show promise for interventions aiming to reduce sexual partner numbers. Individual-level counselling strategies might not be effective in impacting this behavior. Third, interventions that incorporate the IMB principles of risk reduction—particularly in group settings—may be effective in reducing risky alcohol consumption. However, there is a clear need for intervention innovation in programs to reduce alcohol-related HIV risk among heterosexual men in LMICs. Finally, payment of compensation to study participants does not seem to bias either the samples or responses, and may be empirically acceptable.
Research efforts need to focus on rigorous evaluation of HIV prevention interventions to provide a strong evidence base for prevention efforts among heterosexual males in LMICs. Rather than relying on behavioral measures as indicators of intervention effectiveness, it may be prudent for intervention evaluation strategies to include biological measures in research designs; namely, HIV prevalence or incidence rates (Padian et al. 2010). That being said, due consideration needs to be given to the additional expense required for evaluations that include biological measures. While randomized controlled trials are the gold standard for providing an evidence base for prevention programs (Padian et al. 2010), perhaps good quality, consistent behavioral data collected by means of methodologically sound, experimental or quasi-experimental studies are the best that can be striven for in the current economic climate (Hallett et al. 2007).
There is validity in the argument that no magic bullet, single-level HIV prevention is likely to be effective in reversing the HIV/AIDS epidemic, and that combination prevention (a combination of behavioral, biomedical and structural actions) is more likely to have the strongest and most enduring impact on reducing HIV infection (Merson et al. 2008). However, given the poor performance of the two combination prevention interventions included in this review, prevention practitioners in LMICs might be well-advised to replicate combination prevention interventions in other settings before adopting a combination prevention approach. Evidence from this review suggests that prevention practitioners in LAMICs could continue to implement combinations of those behavioral interventions (e.g., individual-level interventions for condom use and small-group interventions for partner numbers) that have shown promise and/or evidence of effectiveness.
Loraine Townsend and Yanga Zembe were supported by the Columbia University-Southern African Fogarty AIDS International Training and Research Program (AITRP) funded by the Fogarty International Center, National Institutes of Health (grant X D43TW00231) (May to July 2010).