The data come from a cross-sectional Integrated Biological and Behavioural Assessment (IBBA), which was conducted in urban Bangalore, Karnataka in 2006 as part of the evaluation of Avahan 5 7 11 . Respondents provided blood samples for HIV and STI testing, and provided behavioural information through a structured face-to-face questionnaire. In Bangalore, since the Avahan intervention is delivered by local NGO Sangama, all questions about contact with Avahan refer to contact with Sangama. MSM were recruited by time-location sampling at local ‘cruising sites’ (locations where men are known to look for male sexual partners) 11 . Three hundred and twenty out of 554 MSM approached (58%) completed the behavioural survey 20 and are included in this analysis. Standardized weights were assigned to the data to account for differences in selection probabilities between different clusters of MSM arising from the sampling design 7 11 ; clusters were also taken into account, using the Complex Samples module in SPSS.

This research received approval from the Ethics Committee of the London School of Hygiene and Tropical Medicine. The Bangalore data collection methods were approved by the Ethics Review Boards of St Johns Medical College and Hospital, Bangalore, the Centre hospitalier affilié universitaire de Québec (CHA), Québec, Canada, and the University of Manitoba, Canada. For the monitoring and evaluation of the Avahan intervention in India, ethical approval was also obtained from the Ethics Review Board of the Centre hospitalier affilié universitaire de Québec, Québec, Canada.

The outcome variables were: (1) reporting CCU (study participants responded that in general they used condoms “every time” rather than “most of the time”, “sometimes”, or “never”); and (2) reporting using a condom at last sex act, with each of the following types of partner: (i) all male sexual partners; (ii) main (regular) male sexual partner; (iii) casual (“new” or “unknown”) male sexual partners. Only MSM who reported having sex with new or unknown partners in the past week were asked about condom use with casual partners in the survey. Although many MSM reported selling sex, the data collected in this survey did not differentiate between commercial and non-commercial partners, and so both ‘all’ and ‘casual’ partners are likely to reflect a mixture of commercial and non-commercial partnerships.

Three binary variables reflecting exposure to the Avahan intervention were investigated: whether or not MSM had ever been contacted by community mobilizers/staff from Sangama; whether or not they had ever been given condoms by Sangama staff; and whether they had ever seen a demonstration of correct condom use by a Sangama worker. Four continuous exposure variables were also investigated: duration since first contact by Sangama staff (for those ever contacted); number of times contacted by Sangama staff in the last month (for those ever contacted); number of condoms received on the most recent occasion that they were given condoms by Sangama staff (for those who had ever received condoms); and number of condom demonstrations seen in the last month (for those who had ever seen a condom demonstration).

A number of additional variables which might influence the relationship between intervention exposure and condom use were also examined for any relationship with either condom use or intervention exposure. All of the continuous variables were also considered as categorical variables, which were derived by dividing the data into quartiles, giving four groups of roughly equal sizes. The variables considered were: age (continuous variable or categorised into 16-22/23-25/26-31/32-60 years of age); duration as an MSM (calculated as time since first had sex with a man; continuous variable or categorised into 0-4/5-7/8-13/14+ years); ever married to a woman; ever paid a female for sex; religion (Hindu or other (Muslim or Christian)); circumcision status; literacy (whether or not they could read and write); school grade reached for those literate (categorised into 0-8/9-10/11-12/13-16); had ever sold sex; sex work being their main source of income; three sexual identity groups (Kothis + Hijras/Panthis + bisexuals/double deckers (Kothis and Hijras mainly take the receptive role in anal sex, Panthis and bisexuals mainly take the insertive role, and double deckers take both roles 21 22 23 ); location where sex usually occurs with male partners (public or private, where public = bar/nightclub, public garden, public toilet, railway station, bus stop/stand, cinema hall/theatre or other public place, and private = home, rented room, vehicle, hammam or lodge); number of MSM partners having had anal sex with in the last week (continuous variable or categorised into 0/1-2/3-4/5+); number of times having had anal sex with known partners in the last week (continuous variable or categorised into 0/1-2/3/4+); and the number of times having had anal sex with new partners in the last week (continuous variable or categorised into 0/1-2/3-5/6+).

Logistic regression analyses were used to identify the factors associated with each of the outcome variables (condom use with each partner type, either at last act or consistently). First, we looked separately at the relationship between each outcome variable and each of the intervention exposure variables in univariate regression analyses. Second, we identified any socio-demographic and behavioural variables which were associated with either the outcome or the intervention exposure variables (i.e. potential confounding variables) using univariate regression – logistic regression for assessing associations with condom use variables and binary exposure variables, and linear regression for assessing associations with continuous exposure variables.

Separate multivariable logistic regression models were constructed for each outcome variable and each exposure variable, which included relevant confounding variables. Socio-demographic or behavioural variables were considered for inclusion in the multivariable models if the p-value for their association with either the exposure or the outcome variable was less than 0.1. Since the variables reflecting sexual activity (number of partners in the last week, or number of sex acts with regular or new partners in the last week) were highly correlated, we identified the one which was most often associated with condom use – number of sex acts with new male partners in the last week (as a continuous variable) - and included this (and none of the other sexual activity variables) in all multivariable models. Since ‘sex work as main income source’ is a subset of ‘ever sold sex’, when both of these variables were associated with the outcome, the one most closely associated was included in multivariable models. All variables were entered simultaneously into the multivariable models.

Results were considered to be statistically significant if p < 0.05. All associations of the main exposure and outcome variables with socio-demographic and behavioural variables with a p-value < 0.1 are reported, as these variables were considered for inclusion in multivariate models.

All analyses were performed using IBM SPSS Statistics version 20.0.0.

The MSM surveyed had a mean age of 28 years old (range 16-60), and had been having sex with men for an average of 10 (range 0-37) years. Forty-one percent reported sex work as their main source of income, 76% were literate and 82.5% were Hindu. Fifty-one percent of the sample self-identified as Kothi or Hijra, 22% as Panthi or bisexual and 27% as double deckers. One hundred and thirty-one MSM (41%) reported a main male sexual partner and 238 (76%) declared casual male partners. MSM reported an average of three (range 0-50) male sex partners in the past week. Twenty-four percent had ever been married to a woman, and 16% had ever paid a female for sex.

Seventy-two percent of MSM reported ever being contacted by Sangama staff, 67% had received condoms from Sangama, and 56% had witnessed a condom demonstration. There was considerable overlap between these exposures; all of those who received condoms or seen a condom demonstration had also been contacted by the intervention, and 55% of MSM reported all three of these exposures (they had been contacted, received condoms and seen a condom demonstration). Almost all (98%) of those who had seen a condom demonstration also reporting receiving condoms. Among those contacted, there was a positive correlation between duration since first contact with Sangama and number of contacts in the last month (r = 0.185, p = 0.005); this association persisted when only those who had first had contact with Sangama within the previous 11 months (since Sangama began offering Avahan services) were considered (r = 0.211, p = 0.03, n = 106). No association was seen between the other continuous intervention exposure variables.

Seventy-one percent of MSM reported CCU with all male partners, 74% reported CCU with their main male partner, and 74% reported CCU with casual male partners.

The percentage of MSM reporting condom use at last sex act was 79% for sex with any male partner, 78% with the main partner and 86% with casual partners.

Our study used self-reported data, which may be subject to social desirability bias. Social desirability bias is likely to have led to an overestimate of the true impact of the intervention upon condom use in this study, since those exposed to the intervention should have heard more messages promoting condom use than those not exposed, and so may feel a greater pressure to report high condom use. On the other hand, the study may have underestimated the true impact of the intervention upon condom use, since those MSM not directly exposed to the intervention may be having sex with those who have been exposed and are using condoms more frequently. There may have been a selection effect by the intervention, i.e. those reached by Avahan may have been more likely to use condoms anyway. We attempted to control for this by including in our multivariate models measured factors associated with exposure to the intervention, but there may have been other unmeasured factors which could account for both a higher likelihood of intervention exposure and higher condom use. The sampling methods used to gather the data (designed to capture high-risk MSM), and relatively low participation rate, implies that the data may not be representative of the wider MSM population. The analysis used cross-sectional data, and so we cannot demonstrate whether exposure to Avahan preceded increased levels of condom use.

In light of our findings, we recommend that the Avahan program, and other programs following this model, focus upon providing high-quality contacts with MSM, which wherever possible should include demonstrations of correct condom use and distribution of condoms, as well as communication about risk reduction.