In London, HIV-positive gay men appear to meet casual UAI partners of the same status through the Internet. This presents a risk for STI transmission. However, gay men were no more likely to meet casual UAI partners of unknown or discordant HIV status - which presents a risk for HIV transmission - online rather than offline.
Gay men are increasingly using the Internet to look for sex . Numerous studies in the USA, Europe and Australia have found elevated levels of high risk sexual behaviour among gay men who seek - and meet - sexual partners through the Internet. Gay men who look for sex through the Internet are more likely to have had a sexually transmitted infection (STI) than other men and more likely to report sexual behaviour which presents a risk for HIV/STI transmission . In London, HIV-positive men who looked for sex through the Internet were also more likely to say they had had unprotected anal intercourse (UAI) with a casual partner who, like themselves, was HIV positive . This does not present a risk of HIV transmission to an uninfected person. However, for HIV-positive men themselves it could have a negative impact on their prognosis due to co-infection with an STI or a drug-resistant strain of HIV .
The limitation of all the studies to date is that they were not able to establish whether the excess risk for HIV and STI actually occurred with sexual partners whom gay men met through the Internet. Unravelling the nature of the association between seeking sex on the Internet and high risk sexual behaviour is essential for developing effective online interventions .
The objective of this study was to examine whether the excess risk for HIV and STI seen among gay men who look for sex through the Internet actually occurs with the men they meet online.
Research among hard-to-reach groups such as gay men is usually based on convenience rather than probability samples . This may introduce selection bias that can be partially overcome by sampling from more than one source. Consequently we recruited four samples of London gay men - one online, three offline. The four samples were: (1) gay men using Internet chatrooms and profiles; (2) HIV-positive gay men attending an outpatients clinic; (3) gay men seeking an HIV test; and (4) gay men using central London gyms. All men were asked to complete a self-administered, confidential, anonymous questionnaire either online (Internet sample) or by pen-and-paper (offline samples). The methods have been described in detail elsewhere .
Recruitment
All respondents were surveyed in 2002 or 2003. Gay men using UK chatrooms or personal profiles on gaydar and gay.com in May-June 2002, and again in May-June 2003, were invited, by means of banner advertisements and pop ups, to complete a questionnaire online. Between October 2002 and May 2003, gay men diagnosed with HIV infection attending an NHS (i.e. public) hospital outpatient treatment clinic in London were invited by a trained member of staff to complete a questionnaire in the clinic. Since everyone who is diagnosed HIV positive in the UK is offered free treatment and care through the NHS, a clinic sample is broadly representative of people diagnosed with HIV. Gay men seeking an HIV antibody test in an NHS HIV testing clinic or sexual health clinic in London between October 2002 and November 2003 were invited to complete a questionnaire in the clinic before having the test. Their HIV test result was linked to the questionnaire without breaching confidentiality. In February-March 2002 and again in February-March 2003 gay men using any one of seven central London gyms during a 1-week period were asked to complete a questionnaire either in the gym or at home. The socio-demographic characteristics of the men surveyed in the gyms are broadly similar to those of other gay community samples in London .
Questionnaires
The questions in the online and pen-and-paper questionnaires were worded identically to ensure direct comparability between the four samples. Men were asked to provide information on their age, ethnicity, education (higher education was defined as having had 3 or more years education after the age of 16 years), employment, socio-economic status (based on occupation for those currently employed) and relationship status. They were also asked to classify their sexual orientation as either gay or bisexual, and were asked about HIV test history (date and result of last test), HIV treatment optimism , recreational drug use and history of STI in the previous 12 months. HIV-positive men were asked whether or not they were currently taking HAART (highly active antiretroviral therapy), about their viral load and disclosure of their HIV status to sexual partners. We asked all men whether they had completed a questionnaire at more than one site, for example, in the HIV treatment clinic and in a gym.
Internet
Men were asked (1) whether they had access to the Internet at home or work; (2) whether they had used the Internet to look for sexual partners in the previous year; and (3) whether they had looked for sexual partners in bars, clubs, saunas, etc in the previous year (i.e. anywhere except the Internet).
Sexual behaviour
Men were asked whether they had had unprotected anal intercourse (UAI) in the previous 3 months and if so, type (main or casual) and HIV status of their UAI partner(s). UAI was classified as concordant (with a partner of the same HIV status) or non-concordant (with a partner of unknown or discordant HIV status). Non-concordant UAI presents a risk for HIV transmission.
Concordant UAI does not present a risk for HIV transmission to an uninfected person. Men in a relationship can establish concordance with their main partner if both men agree to have an HIV test together . It is also possible for HIV-positive, but not for HIV-negative men to establish concordance with a casual partner through mutual disclosure. A number of researchers have examined concordant UAI with a casual partner among HIV-positive men while recognizing the limitations of mutual disclosure .
Where men met their casual UAI partners
Men who said they had used the Internet to look for sex were asked whether they had had UAI with casual partners whom they had met through the Internet in the previous 3 months and if so, with how many. UAI with casual partners they met through the Internet (i.e. with 'online partners') was classified as non-concordant for HIV-positive,-negative and never-tested men or as concordant (HIV-positive men only).
All men were asked whether they had had UAI with casual partners whom they had met in places such as bars, pubs, clubs, saunas, etc (i.e anywhere except the Internet) and if so, with how many. UAI with such men ('offline partners') was classified as non-concordant for HIV-positive,-negative and never-tested men or as concordant (HIV-positive men only).
Statistical analysis
Data from the four samples were analysed using SPSS for Windows 11.5. Analyses were conducted separately for HIV-positive, HIV-negative and never-tested men. Chi squared, Fisher's exact, and t-tests were used for examining differences in proportions and means between the samples (Tables 1 and 2). HIV-negative men who reported concordant UAI with a casual partner were excluded from the analysis (Tables 2-4) since HIV-negative seroconcordance cannot be reliably established in a casual encounter. For HIV-positive men, however, those reporting non-concordant or concordant UAI with a casual partner were included.
In Table 3, men who used the Internet to look for sex were compared with men who did not seek sex in this way. The univariate association between seeking sex through the Internet and UAI was initially explored in a logistic model. Significant univariate associations between UAI and seeking sex through the Internet were further explored in a multivariate logistic model. To identify potential confounding factors in the model, variables known to be associated with UAI in this sample of gay men were compared between men who did and did not seek sex through the Internet . The odds ratios were adjusted by simultaneously entering the independent variable (seeking sex through the Internet), the dependent variable (UAI) and appropriate confounding factors in the multivariate model. Separate analyses were conducted for concordant and non-concordant UAI broken down by type of partner (main only or casual) and HIV status of respondent. HIV-positive men reporting concordant as well as non-concordant UAI with casual partners were initially included in both behavioural groups.
We then examined at an individual, rather than group level, the sexual behaviour of men who used the Internet to look for sex. Men who looked for sex through the Internet were classified as having had UAI (1) only with online partners; (2) only with offline partners; (3) with both online and offline partners; (4) with neither (i.e. no UAI with casual partners). These categories were mutually exclusive. The null hypothesis was that Internet sex seekers were just as likely to have UAI with online as with offline casual partners. The p-values generated by McNemar and Wilcoxon Z tests assessed the probability of obtaining the observed difference in online and offline casual UAI partners if the null hypothesis were true; P < 0.05 provided evidence against the hypothesis. The McNemar test examined whether there was a significant difference between the percentage of men reporting non-concordant UAI with casual partners they met online rather than offline, taking into account men who reported UAI with both. The Wilcoxon test also examined this differential but was more robust since it took into account the percentage of respondents reporting UAI with online or offline partners as well the number of UAI partners. Separate analyses were conducted for non-concordant UAI according to HIV status of respondent and for concordant UAI (HIV-positive men only).
HIV status was laboratory-confirmed for HIV-positive men attending the treatment clinic and for men seeking an HIV antibody test but was self-reported for the gym and Internet samples. Men seeking an HIV test who tested positive (n = 15) were excluded from the analysis; the remaining men in the testing clinic sample were all HIV-negative.
Completed questionnaires were returned by 4225 London gay men in 2002-2003; 528 HIV-positive gay men in the treatment clinic (66% response rate); 404 gay men seeking an HIV test (72% response rate); 921 gay men in London gyms in 2002, 543 in 2003 (50-60% response rate); 1250 gay men using Internet chatrooms and profiles in 2002, 579 in 2003. While it is difficult to estimate a response rate for the Internet sample, electronic monitoring of respondents revealed that 75% of those who started the questionnaire completed it. In each sample, the majority of men (90-98%) said they had only completed the questionnaire at that location .
The analysis was restricted to 4015 men who provided complete information on HIV status, UAI, type and HIV status of UAI partner and use of the Internet for seeking sex. Data for the treatment clinic, testing clinic, 2003 Internet and 2003 gym samples are presented in Tables 1-4. Data for the 2002 Internet (n = 1223) and 2002 gym (n = 892) samples are available from the authors on request.
There were significant differences between the samples in their socio-demographic and behavioural characteristics (Tables 1 and 2). Because of these statistically significant differences, the data for each sample were analysed separately.
Internet
In all samples the majority of men (86-99%) had access to the Internet either at home or work. Not surprisingly, men in the Internet sample were most likely to have Internet access (P < 0.01) (Table 1).
Forty-three percent of HIV-positive men in the clinic sample, 46% of HIV-negative men in the testing clinic sample and just over half the men in the gym sample (52%; 280 of 538) had used the Internet to look for sex in the previous 12 months (gym sample: HIV-positive men, 75%; HIV-negative men, 51%; never-tested men, 36%, P < 0.001). Over 90% of men in the Internet sample said they had used the Internet in this way. In each sample, the majority of men (76-97%) who looked for sex through the Internet said they had also looked for sex in bars, clubs, saunas, etc.
UAI with a casual partner
Among HIV-positive men in the clinic and gym samples, those who used the Internet to look for sex were significantly more likely to report concordant UAI with a casual partner than other men (Table 3). This association remained significant after controlling for confounding factors (P < 0.05).
In the clinic and gym samples, in multivariate analysis, HIV-positive and -negative men who used the Internet to look for sex were significantly more likely to report non-concordant UAI with a casual partner than other men (P < 0.05, Table 3). A univariate association between seeking sex on the Internet and UAI with a non-concordant casual partner was also seen for HIV-negative men in the Internet sample (P < 0.01).
For never-tested men there was no significant association between seeking sex on the Internet and non-concordant UAI with a casual partner (P ≥ 0.5).
A similar pattern was seen for STIs (data available from authors on request).
In all samples there was no significant association between seeking sex on the Internet and either concordant or non-concordant UAI with a main partner alone (P ≥ 0.1) (data available from the authors on request).
Where Internet-sex seekers met their casual UAI partners
In all samples, HIV-positive men who looked for sex through the Internet were significantly more likely to report concordant UAI with a casual partner they met online rather than offline (P < 0.05) (Table 4). For example, 9.9% of HIV-positive men in the clinic reported concordant UAI with casual partners they met online only, 3.8% offline only, 7.7% both (McNemar P < 0.05). Repeating the analysis for HIV-positive men who only reported concordant UAI with a casual partner (i.e. excluding those who reported both concordant and non-concordant UAI) did not substantially alter this finding (data available from authors on request). In the clinic sample, HIV-positive men were more likely to say they disclosed their HIV status to sexual partners they met online rather than offline. Among men who had met sexual partners both online and offline (n = 175), 24% had disclosed their HIV status to all their online partners compared with 14% who disclosed to all their offline partners (P < 0.01).
For HIV-positive,-negative and never-tested men who looked for sex through the Internet, there was no evidence to suggest they were more likely to meet their casual non-concordant UAI partners online rather than offline (Table 4). For example, among HIV-negative men in the Internet sample, 9.7% reported non-concordant UAI with a casual partner they met online only, 11.1% offline only, 6.3% both (McNemar P = 0.6). In fact, for HIV-negative men in the clinic and gym samples, the reverse pattern was seen; they were more likely to report non-concordant UAI with a casual partner they met offline.
Repeating the analyses in Tables 1-4 using 2002 data for the Internet and gym samples yielded similar findings to those described above (data available from authors on request).
Discussion
Among 4000 gay men living in London, surveyed both online and offline, there was a clear association between seeking sex on the Internet and high risk sexual behaviour. HIV-positive men who looked for sex through the Internet were more likely to report unprotected anal intercourse (UAI) with a casual partner who, like themselves, was HIV positive. Furthermore, HIV-positive and -negative men who looked for sex through the Internet were more likely to report UAI with a casual partner of unknown or discordant HIV status than other men. In the clinic and gym samples between 40 and 50% of men said they looked for sex on the Internet.
What is new about this study is that we can establish whether the excess risk for HIV and STI seen among gay men who looked for sex through the Internet actually occurred with men they met online. Previous studies were not able to do this .
In our study, HIV-positive men who looked for sex through the Internet were more likely to meet online, rather than offline, other HIV-positive men with whom they had (concordant) UAI. For men reporting UAI with casual partners of unknown or discordant HIV status, however, we saw a different pattern. Men who looked for sex through the Internet were no more likely to meet their non-concordant UAI partners online than offline. This was seen for HIV-positive, HIV-negative and never-tested men alike.
Why might HIV-positive men be more likely to meet online rather than offline other HIV-positive men with whom they have unprotected sex? One explanation may be that the Internet provides HIV-positive men with a relatively safe environment where they can disclose their HIV status. Compared with offline venues such as bars and clubs, HIV-positive respondents found the Internet to be less stigmatizing. Through a process of 'filtering' and 'serosorting' on the Internet, HIV-positive men were able to establish concordance in a way that could not happen easily offline .
As a group, men who looked for sex through the Internet were more likely to report UAI with a non-concordant casual partner than other men. One explanation for this may be that high risk men are more likely to use the Internet to look for sex than other men. Yet Internet-sex-seekers were no more likely to report high risk sex with their online than with their offline partners. How can we account for this finding? The explanation seems to lie in the fact that most of the Internet-sex-seekers in our study looked for sex both online and offline. This is not entirely surprising since there are numerous bars, clubs and other venues in London where it is possible to meet casual partners. So at an individual level, Internet-sex-seekers were just as likely to meet their non-concordant UAI partners offline, in bars, clubs, saunas, etc, as on the Internet itself.
It has been suggested that because the Internet is such an efficient way of meeting sexual partners this could create more opportunities for sex with online rather than offline partners, including UAI with a non-concordant casual partner . However, in our study we found no evidence that this was more likely to occur with online rather than offline partners. Our data do not therefore support the suggestion that the Internet, per se, creates a risk for HIV transmission.
Our study has several strengths: we sampled from more than one site to overcome some of the inherent biases of convenience sampling ; the HIV status of men in the clinic samples was laboratory-confirmed; HIV-positive men in the treatment clinic were broadly representative of men diagnosed with HIV; we undertook qualitative research which helped us understand some of the processes underlying the quantitative findings . However, data on sexual behaviour was self-reported which is a limitation of much behavioural research . Furthermore, not all the men provided complete information on where they had met their UAI partners so they could not be included in the individual level analysis (Table 4). Nonetheless, despite these limitations, similar associations and patterns emerged across all samples in this analysis.
Our findings have allowed us to better understand the association between seeking sex on the Internet and sexual behaviour among gay men. While men who seek sex through the Internet are more likely to engage in non-concordant UAI with casual partners than other men, they are no more likely to meet these partners online than offline. Internet-based interventions among this group of men must therefore address risks with offline as well as online partners. On the other hand, HIV-positive men are more likely to meet casual UAI partners of the same status online rather offline. The Internet may therefore lend itself to targeted interventions among HIV-positive men around the health risks of 'positive-positive' sex . Our study provides a foundation on which to build and evaluate Internet-based interventions among gay men in London and other major metropolitan areas with large gay communities .
Acknowledgements
This research was funded by the Medical Research Council (grant number GO 100159). The authors would like to thank: gaydar, gay.com, Royal Free Hampstead NHS Trust hospital, Barts and The London NHS Trust hospital, central London gyms; all the men who completed a questionnaire or were interviewed one-to-one; all those who distributed and collected questionnaires; Teresa Allan for statistical advice.