HIV bio-behavioural surveillance surveys conducted at regular intervals are critical for monitoring of, and informing a targeted response to the HIV pandemic. We used Respondent-driven Sampling in 2006, 2008 and 2010 to recruit men who have multiple female sexual partners. We performed several logistic regression analyses to compare HIV sexual risk behaviours, and HIV infection over time. Decreases in inconsistent condom use with main partners were not sustained in 2010. Inconsistent condom use with non-main partners, partner numbers and having one-time partners continued to decrease over time. Levels of alcohol consumption in 2010 reverted to a level higher than in 2006. Non-significant increases in HIV prevalence and reporting a symptom of a sexually transmitted infection (STI) were found. The decrease in numbers of and one-time sexual partners, and in inconsistent condom use with non-main partners augers well for decreasing HIV incidence among men in the study community, but might be offset by decreases in consistent condom use with main partners, and increases in alcohol consumption and STIs.