OBJECTIVE: To evaluate the effectiveness of a pilot peer education STI prevention programme with male sex workers. DESIGN: A process and outcome evaluation of the pilot programme undertaken in three London male escort agencies, using a quasi-experimental design. SUBJECTS: Workers in three London escort agencies, including 88 who completed a questionnaire, five peer educators, and a further 16 men (including management) working in two of these agencies. METHODS: A peer education STI prevention programme run by the Working Men Project (WMP), a specialist sexual health service for male sex workers, was piloted in two London escort agencies. Five male sex workers participated in a 2 day peer education training programme. They then returned to their respective agencies to disseminate information and condoms, in an attempt to influence norms of behaviour. An outcome evaluation aimed to assess changes in STI related knowledge, high risk sexual behaviour, and attendance at a sexual health service. A pre-intervention questionnaire assessing variables such as STI related knowledge, sexual behaviour, and demographic information was administered in both agency A and agency B and a third agency, C, which acted as a control. Ten weeks after the peer educators returned to their agencies, the same questionnaire was administered in the same agencies. Peer educator referrals to the WMP were also recorded over this time period. The process evaluation involved interviews and focus groups with peer educators, and the completion of diaries about their experiences in the role. A further 16 men working in the agencies (including managers and an owner) were interviewed about their experience of the programme. Participant observation was also undertaken through regular outreach work to the agencies. RESULTS: 57 men completed the questionnaire at time 1 and 44 at time 2. Unfortunately, only 13 of these were matched, precluding any meaningful analysis of change in STI related knowledge and sexual behaviour. The questionnaire provided a profile of the men working in the agencies. Of the 88 men who completed the questionnaire at least once, the majority were homosexual, and in their late teens/early 20s. Most were of a "white" ethnic group, though there was some range within these categories. Most preferred to speak English and education levels were high. Relative STI knowledge revealed a high understanding of HIV and hepatitis B, moderate understanding of gonorrhoea, syphilis, genital warts and herpes, and little knowledge of non-specific urethritis (NSU) or chlamydia. Sexual behaviour suggested a highly sexually active population with both male and female paying and non-paying partners. Condom use was highest for paying partners, particularly for anal sex. Condom use for oral sex with all partners was less consistent, and condom use for all types of sex with regular partners was lower than with other partners. The small number of men engaging in vaginal sex with paying and regular partners were less likely to use condoms. 26 new patients registered at the WMP as a result of peer educator referrals, representing 65% of all new contacts over the study period. The process evaluation revealed that while the training programme was considered adequate and while peer educators felt the programme and their roles to be a success, their experience of the role was difficult. The role of management support was crucial in supporting the programme. The assumption that "peers" are particularly effective educators was not borne out by the results. While peers were considered suitable to discuss some aspects of the industry, many preferred to consult "professionals" about health related matters. The concept of "peers" was problematic with most of the men drawing "peers" from subgroups within the agencies. Other constraints on behaviour such as a lack of power, particularly with regard to a lack of management support, or poverty, had a substantial impact on behaviour which were not influenced by the peer educators. CONCLUSIONS: The study illustrated the difficulties of utilising quasi-experimental evaluation methodology with this client group. It also demonstrated the limitations of peer education based on information provision health education models which focus on individual behaviour change. Suggestions are given for future interventions.