Peer education has no significant impact on HIV risk behaviours among gay men in London

In the USA, peer education among gay men living in small towns has brought about significant reductions in high-risk sexual behaviour 1–5. However, this approach to HIV prevention, often focusing on men visiting gay bars, may not transfer directly to other countries or settings. Consequently, peer-led interventions need to be evaluated in well-designed trials among British gay men 6 in the light of their continuing high-risk sexual behaviour 7–11. The prevalence of HIV is greater in London than anywhere else in Britain 12,13. London gay bars do not, however, lend themselves to a controlled trial because movement between venues would lead to the ‘contamination’ of intervention and control groups. We therefore evaluated a peer-led, community-level intervention among gay men using central London gyms. Individuals tend to go regularly to one gym, so that an intervention could be introduced into some gyms whereas others could serve as a control (91% of men said they used only one of the five gyms in the study). A gym-based project could also focus on the injecting behaviour of men who used anabolic steroids 14. One gym was exclusively gay whereas the others estimated that gay men comprised 40–90% of their male membership 15. Peer education draws on a ‘diffusion of innovation’ model 16, whereby popular opinion leaders engage in conversation with gay men to promote and endorse HIV risk reduction. Gym managers were asked to identify potential peer educators according to defined criteria 1–3. After training, peer educators were asked to talk to at least 20 gay men in their gym over the next 5 months about HIV risk reduction. Between September 1997 and April 1999 the intervention was introduced into the gyms in three 6 month phases: phase 1, gyms A and B; phase 2, gym C; phase 3, gyms D and A (`top-up'). Gym E remained a control throughout. For administrative reasons the randomization of gyms was not feasible, but this should not have introduced systematic bias. Further details about the intervention have been published elsewhere 15. Outcome variables were the proportion of gay men: (i) reporting status-unknown unprotected anal intercourse (UAI) in the previous 3 months; (ii) ever-tested for HIV; (iii) who had shared needles or syringes in the previous 6 months (current steroid injectors only). Information was also collected on the number of men who spoke to a peer educator in their gym. Self-administered questionnaires were distributed in all gyms at baseline, then at 6, 12 and 18 months follow-up 15. Changes in outcome were initially examined for the whole study group. When significance was found, the analysis was repeated excluding at follow-up men who had completed questionnaires at an earlier time point (approximately 50% of respondents) to ensure the independence of samples. Only significant P values (< 0.05) from the independent sample analysis are reported. To evaluate the impact of the intervention after 6 months follow-up, the change in the proportion of men reporting an outcome in the intervention group (gyms A, B; two clusters) was compared with the corresponding change in the control group (gyms C, D, E; three clusters) in a logistic model using the individual as the unit of analysis 3,5, controlling for age and relationship 17. When significant differences were found, the analysis was repeated using independent samples at follow-up as described above, modelling for the number of clusters 3. Similar procedures were used at 12 and 18 months follow-up. In a power calculation using the individual as the unit of analysis, we estimated that to detect a decline in status-unknown UAI from 15% at baseline to 10% after the intervention (i.e. a 33% reduction 3,5) would require a total sample size of 1200 men (alpha = 0.05, beta = 0.2). To detect an increase in the percentage of men ever-tested for HIV from 65% at baseline 7 to 75% after the intervention required a total sample of 800 men. Just over 1000 gay men returned completed questionnaires at baseline and 6 months, 916 at 12 months and 720 at 18 months; response rates were (minimum estimate) 51, 47, 49 and 39%, respectively (full data available from authors). At each time-point the majority of men were young (median age 33 years), white (90%), employed (90%), and university educated (70%) 15,17. In the baseline survey men were not asked for their HIV status. In subsequent surveys 14–17% of respondents reported being HIV positive 18,19. Overall, the proportion of men reporting status-unknown UAI varied little between baseline 15 and 18 months follow-up (13.9%, 14.2%) (P = 0.5). The change in the percentage of men reporting status-unknown UAI did not differ significantly between intervention and control gyms (P ≥ 0.1) (Fig. 1). The non-significant increase in the intervention group was predominantly seen in one gym alone.Fig. 1.: Percentage of men ever-tested for HIV and reporting status unknown unprotected anal intercourse in the previous 3 months. C, Control group; I, intervention group; UAI, unprotected anal intercourse; 0, baseline. P values from a logistic model testing whether the change in outcome in the intervention group was significantly different from the corresponding change in the control group. 0–6 months intervention gyms A, B (472, 445); control gyms C, D, E (532, 573). 0–12 months intervention gyms A, B, C (764, 651); control gyms D, E (240, 265). 0–18 months intervention gyms A–D (879, 612); control gym E (115, 108). Brackets contain the number of men surveyed at baseline and follow-up, respectively.The overall percentage of men ever-tested for HIV increased from 73.0% at baseline 15 to 79.6% at 18 months follow-up (P = 0.002). This increase was seen equally in both the intervention and control gyms (P ≥ 0.5) (Fig. 1). The percentage of men who injected steroids increased from 8.1% at baseline 20 to 12.6% at 18 months (P = 0.02). Similar increases were seen in intervention and control gyms (P ≥ 0.2). In none of the gyms did steroid injectors report sharing needles or syringes at baseline 20 or 6 months. At 12 and 18 months follow-up, one injector in the intervention group (1%) reported sharing needles or syringes, whereas none (0%) did so in the control group (P = 0.7). It appears that this peer-led intervention among London gay men had no significant impact on their risk behaviours. Why not? Whereas most men (80%) thought it was useful to have peer educators in the gym to talk about risk reduction and half the men were aware of their presence, only 3% (19/612) said they had spoken to a peer educator during the intervention period. Consequently, the critical mass required for diffusion was not established 16. Although the sample size declined at follow-up, the lack of diffusion rather than statistical power almost certainly accounted for the absence of any significant effect. It seems that rather than peer education not working in this trial, it simply did not occur. Why was this so? Process evaluation offers some explanation 21. First, recruiting and retaining peer educators was more difficult than anticipated. Of the 144 popular opinion leaders identified by gym staff, 46 (32%) underwent training but only 27 (19%) remained with the project throughout the intervention period. Second, peer educator feedback revealed difficulties in talking to complete strangers about sex, and relatively few conversations with gym members occurred (average 10). Third, some peer educators felt that the intervention period (5 months) was not long enough to establish a rapport. Finally, some felt that an American model of health promotion did not transfer easily to London. Barriers to peer education may have been specific to the gym environment (less than a third of the men said they came to the gym to meet friends) or may have reflected differences between small towns 1–5 and large cities. But there also appeared to be important cultural differences between UK and US gay men in their interaction with peers. These findings highlight the importance of incorporating process evaluation into controlled trials of behavioural interventions in order to interpret outcome 22. To conclude, peer education did not appear to be an effective tool for HIV prevention among gay men in central London gyms. Models for HIV prevention shown to be effective among gay men in one country may not necessarily transfer successfully to another country. Acknowledgements The authors would like to thank the managers and members of the gyms for their support and participation in the project; Mark Maguire, Fraser Serle, Andrew Billington, Will Devlin and Glen Monks, Health Promotion Service, Camden and Islington Community Health Services NHS Trust, for developing and running the peer-led intervention; Richard Morris and Fiona Lampe for statistical advice; all those who distributed and collected questionnaires. Jonathan Elford Graham Bolding Lorraine Sherr