Background Early initiation of combination antiretroviral treatment (ART) is being considered as a HIV prevention strategy. Thus, it is important to examine the association between ART, viral load (VL) and the sexual behaviour of HIV+ve patients. We compared the reporting of unprotected anal sex (UAI) in 2010 and 2000 among HIV+ve men who have sex with men (MSM) and its association with VL.

Methods HIV+ve MSM attending a central London HIV clinic were recruited in cross-sectional surveys in 2000 (411 men) and 2010 (423 men). Data on recent plasma VL (detectable (DVL) or undetectable (UVL)); ART status (treatment naïve or receiving ART); UAI with a boyfriend (main partner) and casual partners in the last 6 and 12 months respectively; boyfriend's HIV status (unknown/negative, or positive) were collected. Nonconcordant UAI (ncUAI) was defined as UAI with unknown/negative HIV status partner(s) and concordant UAI (cUAI) as UAI only with HIV+ve partner(s). Men reporting cUAI and ncUAI were treated as engaging in ncUAI. OR for 2010 compared to 2000 for UAI with casual partners were adjusted for age and DVL/UVL. Analysis for UAI with a boyfriend was adjusted also for boyfriend's HIV status and is restricted to those reporting a boyfriend.

Results Compared to 2000, 2010 respondents were more likely to be on ART with UVL (76% vs 40%), but less likely to be on ART with DVL (6% vs 32%) or treatment naïve (19% vs 28%), p<0.001. They were older (age ≥45 years: 46% vs 22%, p<0.001), more likely to report an HIV+ve boyfriend (41% vs 25%; p=0.001). An increase in cUAI with a boyfriend (27% vs 14%) remained significant after adjustment, (OR: 2.84, 95% CI: 1.59% to 5.06%; p<0.001). ncUAI with a boyfriend did not change (12% vs 11%; p=0.63). With regards to casual partners, cUAI decreased over time (OR: 0.50, 95% CI: 0.29% to 0.86%; p=0.01) but ncUAI increased (34% vs 17%; OR: 2.79, 95% CI: 1.93% to 4.04%; p<0.001). UVL was not associated with ncUAI with casual partners or boyfriend, but was associated with increased cUAI with casual partners (OR: 1.83, 95% CI: 1.08% to 3.13%; p=0.02).

Conclusions In comparison with 2000, cUAI with a boyfriend has increased, perhaps due to increased serosorting. In casual partnerships cUAI decreased and ncUAI increased. The latter has implications for onward HIV transmission and needs to be addressed by health promotion programmes. As UVL was not associated with ncUAI, these data do not suggest that early initiation of effective ART will increase HIV transmission.