Patterns of sex partner recruitment are diverse; have changed over the past few decades; are associated with rates of formation and dissolution of sex partnerships, and may be associated with differential risk for acquiring sexually transmitted infections (STIs) and HIV infection. During earlier times, people met sex partners at work, school and church; in the neighbourhood; through friends and acquaintances; or on the street and/or in the brothel, massage parlour, etc.1 More recently, under the influence of technological developments, formation of new social ties has become faster and more diverse; hundreds of meet-ups are announced in local areas every week. Formation of sexual ties followed a similar pattern. Recruitment of sex partners through personal ads in newspapers,2 which was a relatively new mode of meeting partners in the late 1980s and early 1990s, has mostly been replaced by internet-based searches. In January 2009, a Google search based on the words “find sex partner on web” yielded 1 050 000 results; options included “local moms looking for younger men”, “one night stand” and “find a naughty partner for tonight”.

In theory, to the extent that a particular partner recruitment approach results in: (1) faster formation of new ties; (2) greater likelihood that infected individuals hook-up with uninfected individuals; and (3) greater numbers of individuals participating in sex-partner recruitment, holding everything else constant, it would be associated with faster spread of infection in the population. At the individual level, a particular sex-partner recruitment approach would be associated with a higher risk of STI acquisition if it increases the likelihood of “hooking-up” with an infected partner. Infection status of individuals is difficult to ascertain; therefore, we use high risk behaviours as proxies for infection status. Thus, if partners one is likely to meet through a particular recruitment venue tend to engage in other high-risk behaviours, one would expect that recruitment venue to be associated with high risk of STI acquisition for the individual. In reality, everything else is not held constant; there are interdependencies among rates of partnership formation, probability of infected individuals having sex with uninfected individuals, and numbers of individuals participating in sex-partner recruitment; furthermore, most changes carry with them unintended and unanticipated consequences. Hence, it is important to conduct empirical studies of partner recruitment patterns and their association with rate of spread of infection in the population and infection acquisition risk for the individual.

In this issue of the journal, Al-Tayyib and colleagues3 present data on the association between sexual encounters with partners met through the internet and current Chlamydia trachomatis and Neisseria gonorrhoeae infections. There are several methodological strengths to this study. The investigators evaluated a large number of individuals (n = 14 955); they assessed the relation between objectively measured STI and online sex seeking rather than relying on self-report of previously diagnosed infections; and they evaluated men who have sex with women (MSW) and women, in addition to men who have sex with men (MSM), around whom most of the previous research has focused. Furthermore, they compared individuals with recent internet sex-seeking experiences (for example, within the past 4 months) to those with more distant or no reported online sex-seeking experiences. This restriction allowed them to assess a more proximal relation between internet sex-seeking and acquisition of bacterial STI.

That Al-Tayyib and colleagues found no increased risk for bacterial STI among MSW and women and, in fact, observed decreased risk for gonorrhoea and/or chlamydia among MSM, is consistent with several previous reports of no increase in high-risk sexual behaviours among internet sex-seekers.4^–6 However, most of the previous studies have been conducted among MSM, who have somewhat different sexual behaviours and partner identification patterns than heterosexuals. And, as the authors themselves noted, even the population evaluated in the Al-Tayyib study is specialised. STD clinic attendees are not like the general population as a whole.

Previous work has demonstrated that the number and type of risk factors identified in a population-based sample can differ from those identified in an STD clinic sample.7 This is probably because individuals who attend STD clinics have a high overall level of risk behaviours—the majority of them are there precisely because they have engaged in a high-risk behaviour. This often results in an inability to differentiate between infected and uninfected individuals on the basis of risk behaviours. In a comparative study of risk factors for gonococcal infection, this effect manifested itself in two distinct ways: (1) only a subset of characteristics identified as significant risk factors in a general population sample were also identified as risk factors in an STD clinic sample, and (2) the magnitude of the association for the risk factors commonly identified in both study populations was substantially higher in the general population than it was in the STD clinic population (for example, adjusted odds ratio (AOR) of 5.4 for having spent a night in jail in the general population vs AOR of 3.1 in the STD clinic population).7

These observations suggest that a study designed similarly to the Al-Tayyib study, but performed in the general population, might yield different results regarding the association between internet sex-seeking and objectively measured STI. We currently lack such general population studies, making the need for them paramount. General population studies are difficult to undertake; however, the infrastructure for some population-based studies that collect clinical specimens, such as NHANES, Add Health and NATSAL 2010, currently exists, and future waves of data collection could add questions on internet sex-seeking.

Random digit dialling has been a mainstay in our toolbox of methods for obtaining population-level data and identifying control groups for case-control studies. Indeed, random digit dialling studies have provided important data on population-level risk factors for STI in a number of cases.8^–12 However, the days of using random digit dialling to obtain a representative sample of the general population have likely met their end. In this era of cell phones and caller identity, the likelihood of obtaining a reasonable response rate from a truly representative sample is slim. Furthermore, without an objective measure of STI infection, such studies are limited to using self-reported, previous STI diagnosis as an outcome, which can substantially underestimate true STI prevalence.

Internet surveys are probably the wave of the future for obtaining relatively representative samples of the general population, although the extent to which marginalised segments of the population are able to regularly access and use the internet is not clear. Once hampered by the same limitation as RDD surveys in terms of relying on self-report measures of previous STI diagnosis, today such innovative programmes as “I Want the Kit” (http://www.iwantthekit.org/) provide a model for mailing home sampling kits to study participants. Methods such as these could be used to obtain objective measures of laboratory-diagnosed STI and link test results to individual behavioural data obtained in a web-based survey.

Whatever method is used to gather population-based data on risk factors for STI, the need for such data could not be clearer. A small proportion of the population attends STD clinics, they are clearly not representative of the general population, and our ability to identify risk factors for STI that are generalisable to the general population is limited. Thus, it may be important to wait for population-based evidence before we declare a lack of association between particular, theoretically high-risk sexual behaviours and STI acquisition.