BACKGROUND Postmenopausal status and type 2 diabetes mellitus (T2DM) are independent risk factors for fractures. An increased fracture risk has been observed with rosiglitazone (RSG), a thiazolidinedione, in patients with T2DM. METHODS This was a randomized, double-blind study in postmenopausal women with T2DM. A 52-week double-blind phase (RSG or metformin [MET]) was followed by a 24-week open-label phase, during which time all patients received MET. METHODS The primary endpoint was to assess the mean percentage change in bone mineral density (BMD) at the femoral neck (FN) by dual-energy x-ray absorptiometry from baseline to week 52 in the RSG treatment group. Key secondary objectives included assessment of changes in BMD at the total hip, trochanter, and lumbar spine and to evaluate RSG effects on bone turnover markers. RESULTS From baseline to week 52, RSG was associated with a reduction in FN BMD by dual-energy x-ray absorptiometry (-1.47%). During the open-label phase (weeks 52-76), no further loss in FN BMD was observed. A decrease in BMD occurred at the total hip during RSG or MET treatment at 52 weeks (-1.62 and -0.72%, respectively). Total hip BMD loss by RSG was attenuated after switching to MET and was similar between treatment groups at the end of the open-label phase. From baseline to week 52, bone turnover markers significantly increased with RSG compared with MET, but decreased significantly during the open-label phase. CONCLUSIONS RSG for 52 weeks in postmenopausal women with T2DM was associated with small reductions in FN, total hip, and lumbar spine BMD and increased bone turnover markers. These effects are attenuated after cessation of RSG treatment.