Behavioral interventions continue be an essential component of HIV prevention in MSM [4] but targeting evidence based interventions to achieve maximum benefit is challenging. MSM attending clinics exhibit variable levels of sexual risk behavior and the differences in risk behavior, when assessed through routine clinical questioning, between men testing HIV+ve and HIV-ve are small [5]. Clinic and community based behavior change interventions that reduce UAI and partner number in MSM are resource intensive [6] and the criteria for the selection of men who will benefit most from such interventions are to date unclear. Reducing onwards transmission by improving condom use amongst those who use them presents an underexplored area of sexual health research. Moreover, increasing condom skills may well reduce negative condom experiences and lead to an overall increase in condom use.

The effectiveness of condoms in preventing sexually transmitted infections (STI) including HIV depends on consistent and correct use [7]. The frequency and correlates of condom use problems and errors have been reported in a range of populations [8] and in men engaging in vaginal sex and anal sex with female partners [9] and anal sex with male partners [10]. Recent research also indicates that condom use problems are common among Scotland’s MSM. A 2008 cross-sectional study of MSM recruited from the gay bars of Glasgow and Aberdeen, or men receiving condoms through a Scottish postal distribution scheme, found that 27% of men reported condom breakage and 40% condom slippage in the last year [11]. To date condom errors have been associated with the likelihood of sexually transmitted infections in heterosexual STI clinic attendees [12] but not in MSM [13] and direct evidence of a link between condom problems and STI acquisition in MSM has so far been lacking.

Pilot questionnaires were trialed on 15 clinic attendees to assess suitability of questionnaire length and to identify potential difficulties with comprehension or presentation of questions within this particular population. The pilot study indicated clinic attendees were happy to participate and most were able to fully complete the questionnaire within 10minutes. Following the pilot, minimal adjustments were made to the wording of some questions in order to increase clarity. The questionnaire recorded data on demographics, sexual activity, condom use for anal sex and condom problems and assessed condom knowledge and proficiency. Self-reported data on STIs diagnosed in the last 12months were also collected. For the purpose of this study ‘STIs’ included gonorrhea, syphilis, chlamydia, HIV, Hepatitis B and Hepatitis C only; diagnoses of genital warts and herpes simplex virus were excluded due to difficulty distinguishing between incident and recurrent infections.

In clinic attendees additional consent was requested to allow the questionnaire to be linked anonymously to laboratory STI results by attaching a unique patient identifier to each questionnaire and to laboratory request forms. The unique identifier was used to obtain results of STI tests from each participant at that clinic visit. This enabled the matching of self-report data with clinical data for a subset of the sample (n=333), while avoiding the exclusion of men who were not happy for their clinical records to be linked to the questionnaire.

Information leaflets describing the study were available in both clinic and community settings. In clinics, questionnaires were distributed by receptionists, nurses, health advisors and doctors to all men identified as MSM attending the clinic during the study period. In community settings an established time and location sampling methodology [14] was used to recruit a representative sample of men visiting the gay bars and saunas of Edinburgh (6 different settings). Trained volunteers distributed all questionnaires in community settings, offering questionnaires to all men attending each venue during the sampling period. Response rates were recorded for both clinic and community arms of the study.

The mean age of men was 36 (range 17–74) in the clinic sample and 38 in the community sample (range 17–73). The majority of participants in both samples were of white UK ethnicity; 79% of clinic and 81% of community sample; and most were employed; 72% of clinic and 76% of community sample. Men in both samples tended to be highly educated; 58% of the clinic and 55% of the community sample were educated to undergraduate degree level or above. 36% of the clinic sample and 7% of the community sample self-reported as being HIV positive. None of these demographic characteristics were associated with self-reported STI acquisition over the previous 12months (Table ​(Table33).

Number of UAI partners over the previous year was used as a measure of men’s STI risk. The mean number of UAI partners in the previous 12months was 1.5 in the clinic sample and 3.3 in the community sample. The range of UAI partners was wide in both samples (range 0–100 in the clinic sample and 0–245 in the community sample). In order to test associations between number of UAI partners and STI diagnosis, and to account for the possible influence of the few men reporting very high numbers of UAI partners, number of UAI partners was categorised as 0, 1, 2–4, 5–10, or >10. Number of UAI partners was found to be the strongest predictor of self-reported STI acquisition over the previous 12months (Table ​(Table33).

The condom proficiency score was significantly associated with self-reported STI diagnosis over the previous 12months in both clinic and community samples. The condom problem score was associated with self-reported STI diagnosis in the clinic but not the community sample (Table ​(Table4).4). Number of UAI partners (i.e. neither condom proficiency nor problem score) was the only variable to be significantly associated with laboratory confirmed STIs among the clinic sample.

All variables found to be associated with self-reported STI diagnosis in univariate analysis (number of UAI partners, condom proficiency score and condom problem score) were entered into a logistic regression model. Number of UAI partners in both samples and condom problem score among the clinic sample remained significant (Table ​(Table5)5) in explaining the variance in self-reported STI diagnosis. That is, among the clinic sample, condom problem score remained significantly associated with self-reported STI diagnosis after adjusting for UAI risk. In the community sample condom problem and proficiency scores were not associated with self-reported STI diagnosis after adjusting for UAI risk.

Number of UAI partners was strongly associated with self-reported STI diagnosis over the previous 12months as well as with prevalent STI. This is unsurprising and is consistent with other behavioral community based surveys of self-reported STIs in MSM [19] and with clinic based surveys of HIV positive MSM [20]. With evidence of ongoing high-risk behavior in some MSM there is clearly a continuing need for interventions to promote condom use to successfully reduce STI acquisition.

Men at high risk of HIV in Scotland are in contact with prevention services [21] and interventions are likely to be acceptable [22]. A UK study to determine whether condom thickness affects condom breakage found that user characteristics, including lack of confidence in condom use, rather than condom characteristics were associated with condom breakage among MSM [23]. Although clinic based behavioral interventions have been proven to reduce the risk of STIs in a large randomized controlled trial [24] it is recognised that these interventions are resource intensive and recent work has focused on reducing the duration and frequency of interventions. Recent studies have shown a reduction in repeat STIs following a single session of condom skills training delivered to young heterosexual African American men by a lay health adviser [25]. A small pilot study of home based condom skills training showed some improvement in measures of condom self-efficacy and reduction in condom errors in 30 young heterosexual men [26]. The condom self-efficacy and error scores used in that study had high internal consistency but have not been shown to be indicators of STI risk. There is good evidence that intensive multi-session clinic based brief behavior change interventions with MSM can reduce the frequency of UAI with partners of unknown HIV status [27] but there is a lack of evidence on less intensive single session interventions or interventions designed to reduce condom errors in MSM. It is possible that the lower frequency of condom problems in MSM might reduce the impact of condom skills training – although interventions to reduce problems may increase overall levels of use. Such interventions could focus on a number of condom skills, for example, appropriate sized condoms, skills relating to appropriate use or skills relating to condom communication and negotiation.

Strengths of our study included use of both clinical and community samples, the high proportion of HIV positive men included in the sample and the matching of self-report risk behavior to clinical records. Limitations of the study include convenience sampling of populations at selected venues. This could limit the transferability of study findings to MSM populations in other settings and locations. Epidemiological studies show that MSM recruited from community settings have higher rates of undiagnosed HIV infection than those recruited from STI clinics [14] and men in different community samples report widely differing rates of UAI [28,29]. The effect of convenience sampling was mitigated by recruiting from both STI clinics and community venues. All ‘gay identified’ commercial venues in Edinburgh were included. All MSM attending STI clinics and all men attending each community venue during the sampling period; which included different weekdays and different sample times over the 4-month study period; were asked to participate in an attempt to obtain a representative sample. Response rates were high for both clinic and community samples although it is acknowledged that high response rates do not necessarily mean a representative samples. As the study was designed to inform the development of prevention interventions which are likely to be delivered through the same clinics and commercial venues and to the same population, this methodology also carries some advantages. A further limitation is the reliance of this study on subjective reporting of condom use, condom problems, sexual behavior and STI diagnoses. This may have resulted in misclassification of both exposure and outcome. MSM participating in the study may have under-reported high-risk sexual behavior in an attempt to satisfy perceived social norms, but the reported levels of UAI suggest that this occurred to no greater extent than in other published studies and surveys. The use of self-completed anonymous questionnaires may have reduced under-reporting by maintaining anonymity, although further improvements may have been seen through the use of computer–assisted survey technology. Any under-reporting of high-risk sexual behavior is likely to have been evenly distributed throughout the samples, resulting in non-differential bias. This would have lead to an under-estimation of any association between risk behavior and STI diagnosis and is unlikely therefore to change any of our conclusions. It was also not possible to estimate the incidence of STIs using our cross-sectional methodology or to establish any temporal relationship. Another limitation is the moderate internal consistency within the condom problem score. The condom problem score included a large number (23) of relatively varied and potentially unrelated issues, which may happen in isolation or in different circumstances. This limits its value as a global measure and suggests that a refinement of the score, informed by other recent publications [8-10] might be possible.