Available evidence does not support serosorting as an HIV risk reduction strategy

Serosorting is an increasingly common practice among gay men whereby the selection of sexual partners is based on concordance of HIV serostatus [1–3]. Serosorting occurs in order to facilitate unprotected anal intercourse without risk of HIV transmission. Evidence suggests that disclosure of HIV serostatus is increasingly influential not only on the formation of casual partnerships among gay men (serosorting), but also on the sexual position that each man is likely to take in a coital act (known as strategic positioning) and the likelihood of using a condom (known as negotiated safety) [3–9]. Although these inter-related strategies appear to be sophisticated measures for reducing risk, in some circumstances they may lead to an increased risk of HIV transmission in a population [10,11]. Serosorting, when coupled with actual positioning and condom behavior, may result in an increased risk of HIV acquisition if a moderate proportion of the HIV-infected population is undiagnosed. This undiagnosed proportion may lead to the formation of partnerships that are thought to be concordant but are actually discordant. In partnerships that are thought to be concordant, HIV-negative men are more likely to take a receptive role (which is more risky in a discordant partnership) and less likely to use condoms. Cassels et al. [12] developed a detailed compartmental mathematical transmission model in order to demonstrate that ‘under realistic scenarios of sexual behavior and testing for men who have sex with men (MSM) in the USA, serosorting can be an effective harm reduction strategy’ [12]. However, they failed to consider how strategic positioning and negotiated safety behaviors change with or without disclosure of HIV status and therefore their assertions about the benefits of serosorting are overstated. Additionally, they assumed that all men test for HIV 1–2 times per year. This is much greater than the best available data on actual testing rates in the USA [13]. It is estimated that between 25% [14–16] and 48% [17] of all HIV infections in the USA are undiagnosed, which implies that testing rates are considerably lower than 1–2 times per year for all men. If Cassel's et al. [12] used more conservative testing rates in line with rates that are consistent with levels of undiagnosed infections then based on their model-derived Figure 4, serosorting would have been shown to result in higher equilibrium prevalence. The authors also state that, with serosorting, 22% of contacts are apparently serodiscordant, and 50% without. With a prevalence of 16%, their model would require partner acquisition rates among HIV-positive men to be much greater than among HIV-negative men. Lastly, we question the relevance of using equilibrium prevalence (that is, no epidemiological or behavioral change) to gain insight into the effect of a recent phenomenon on an epidemic where there is increasing incidence. Using more realistic assumptions, the model of Cassel et al. [12] would yield very different conclusions, such that serosorting is not likely to reduce risk of HIV acquisition among gay men in the USA but in fact is likely to increase risk in most contexts. Serosorting, strategic positioning and negotiated safety cannot be decoupled. It has been demonstrated that accounting for behavior change is integral because the risk reduction strategies adopted by negative men vary with the level of knowledge of their partner's serostatus [11,18]. There is insufficient evidence to measure the true effectiveness of serosorting as a harm-reduction strategy. Modeling can be useful to inform our understanding when realistic behavioral inputs are used. The benefit of serosorting is highly context specific and depends upon the frequency of testing, accurate disclosure of serostatus and behavior. Our position, based on this study and others in the literature, is that serosorting has a real potential to increase risk and should not be promoted as a public-health strategy.