In this analysis of over 12,000 North American MSM with more than 60,000 six-month intervals of follow-up and 663 seroconversions, we found that both pure and condom serosorting (having UAI with multiple HIV-negative partners, but no UAI with HIV-positive partners), and condom seropositioning (reporting condom use in all receptive anal sex with an HIV-positive or unknown-serostatus partner, and unprotected insertive anal sex with these partners) placed men at increased risk for HIV acquisition, compared with those reporting no UAI with partners of any serostatus. However, these and all other seroadaptive practices were significantly less risky than reporting receptive anal sex with HIV-positive or unknown-serostatus partners, with or without a condom. Men reporting only one HIV-negative partner or always being the insertive partner, regardless of condom use, were at the lowest risk.
It should be noted that all sexual practices in these cohorts of MSM carried some risk, as men in each of the categories, whether or not condom use was taken into account, had seroconversions. In fact, men reported no unprotected anal sex in the majority of intervals, and 173 of the 663 seroconversions occurred during the intervals when men reported no unprotected anal sex. This is likely a result of under-reporting of unprotected anal sex, in part due to social desirability bias, under-reporting of intermitted condom use and/or failure, and potentially some episodes of HIV acquisition from oral sex.
Definitions of various seroadaptive practices continue to evolve in the literature. Our pure 
and condom 
serosorting definitions have both been used previously. We did not find meaningful differences between these two categorizations. Our estimates of the risks associated with both definitions are similar to a recent report by Kennedy et. al. 
of cohorts of MSM from the US and Australia 
. In that meta-analysis, the summary odds-ratio for serosorting was 1.80 (95% CI: 1.21–2.70) relative to consistent condom use, and 0.46 (95% CI: 0.25–0.83) relative to no condom use and no partner selection.
In our study, having only insertive sex (even with a potentially serodiscordant partner) appeared to be safer than no UAI. Other studies have found similar risk for this practice, sometimes also called strategic positioning 
, as for no UAI 
. Reasons for these findings are likely two-fold. First, per-contact risk for insertive anal sex is an order of magnitude lower than for receptive anal sex 
. Second, some intervals were almost surely misclassified as no UAI, because of the under-reporting of UAI and unrecognized condom failure.
Having UAI with a single negative partner was associated with an even lower risk of HIV seroconversion than reporting no UAI. This finding underscores the importance of minimizing number of partners as a primary HIV prevention strategy. The cohorts we studied did not collect the data needed to assess another related seroadaptive practice called negotiated safety, defined as UAI with a steady seroconcordant partner, but only protected sex outside of this relationship. An Australian study found reduced risk among MSM reporting strict adherence to negotiated safety when compared to having no seroadaptive strategy at all 
The effectiveness of seroadaptive practices is dependent on accurate knowledge of serostatus and proper disclosure of serostatus, both of which are in turn dependent on frequent HIV testing. In the United States, an estimated 20% of those infected with HIV do not know their status 
. Expanded testing initiatives, including home testing, may be key to improving testing frequency, and thus improving the effectiveness of seroadaptive practices.
Black MSM are disproportionately affected by HIV. Although they are equally likely to report seroadaptive behavior as White MSM, there is concern that seroadaptive behavior may not be as effective in this group 
because of the high rates of undiagnosed HIV infection, lower reported serodisclosure, and assortative mixing in a community with high background HIV prevalence 
. In our study, we did not find any evidence of effect modification by race. Encouragingly, Marks et. al
, also found that serosorting was an effective risk reduction strategy among Black MSM, when compared to UAI with no seroadaptive strategies 
. Further research on the prevalence and effectiveness of seroadaptive practices in this vulnerable population is needed.
We did not examine the impact of seroadaptive practices on sexually transmitted infections (STIs) in our cohort. However, other groups have show that there is an increased risk of bacterial STI among men who report engaging in seroadaptive practices compared to those who report no UAI at all 
. STI risk should also be taken into consideration when weighing the advantages and disadvantages of seroadaptive practices.
There are several limitations to this study. The cohorts included in this analysis were followed over a decade, between 1995 and 2004, when HIV testing and treatment were changing rapidly. However, calendar time did not modify our findings for seroadaptive categories. In addition, our results were obtained using self-reported behavior, which despite the use of ACASI in one of the included cohorts, is subject to desirability bias as well as random errors; this probably explains the sizable number of seroconversions in intervals where no UAI was reported. Finally, because there was a 6–12 week window between HIV infection and seroconversion with the earlier generations of HIV antibody testing used in these studies, some seroconversions may have been misattributed to a later six-month interval, with potentially a different risk category. However, results were unchanged in analyses categorizing behaviors over the last 12 rather than last 6 months.
Another limitation is that we did not have the data to determine whether participants intentionally adopted one or more seroadaptive practices. Among the 25% of men classified into the same seroadaptive category across all of five or more follow up visits, we may have some indirect evidence of intention, but this pattern was uncommon. In general, individuals made multiple transitions between the six behavioral categories – variability which prevention interventions and counseling messages need to take into account. MSM intentionally adopting seroadaptive practices may be different than those reporting apparently seroadaptive behavior adopted for other reasons, making our results harder to interpret. The links between seroadaptive intention, uptake, maintenance, and risk are a much-needed area of study.
Our study also had several strengths. The four included cohorts were all large, at moderate to high risk, and followed for 18–48 months, providing good statistical power. In addition the sample was recruited in several US cities, enhancing representativeness. Finally, the questionnaires used were similar, enabling us to categorize behavior uniformly.
Given the increased risk we observed for some seroadaptive practices, compared to no UAI, we conclude that condom use and limiting number of partners should continue to be advocated as first-line HIV prevention strategies. Seroadaptive practices may be considered as harm-reduction strategies for MSM engaging in extremely high-risk behavior, but should not be recommended as primary prevention strategies. Finally, behavioral interventions need to be ongoing and tailored since individual risk behavior appears to vary substantially over time.