OBJECTIVE Most longitudinal studies of bone mineral density (BMD) in HIV-infected cohorts have been of short duration, typically 1-2 years. Some studies, especially of cohorts treated with highly active antiretroviral therapy (HAART), report short-term stable or increasing BMD, but other studies, often in cohorts initiating HAART, report short-term losses in BMD. We assessed BMD changes over the medium term in HIV-infected men already established on HAART at baseline. METHODS Six-year, prospective, longitudinal study. METHODS Forty-four HIV-infected men treated with HAART and 37 uninfected, healthy controls. METHODS Participants had measurements of BMD at baseline, 2 and 6 years. RESULTS In the HIV-infected men at baseline, the mean age was 49 years, the mean duration of infection was 8 years, and the mean duration of HAART was 50 months. Over 6 years of follow-up, there was a greater increase in lumbar spine BMD (5·3%, 95% CI 3·8-6·5%) in the HIV-infected men compared with controls (0·3%, 95% CI -1·0 to 1·6%), P < 0·001. There was no difference between the groups in the change in BMD over time at the total hip (HIV group: -0·6%, 95% CI -1·7 to 0·4%, controls -1·0%, 95% CI -2·2 to 0%, P = 0·8) or at the total body (HIV group, 0·3%, 95% CI -0·3 to 1·0%; controls, 0·5%, 95% CI -0·2 to 1·1%, P = 0·15). Lean mass increased in the HIV group, but not in the controls. CONCLUSIONS There was no evidence of accelerated bone loss over 6 years in middle-aged, HIV-infected men treated with HAART. For such patients, routine monitoring of BMD is not necessary over the short/medium term.