OBJECTIVE: To investigate the impact of a long-acting injectable progestin, depo-medroxyprogesterone acetate (DMPA), compared with combination oral contraceptives (COCs) on the risk of diabetes in Latino women with prior gestational diabetes mellitus (GDM). RESEARCH DESIGN AND METHODS: An observational cohort study of 526 Hispanic women with prior GDM who were not diabetic in their postpartum visit during January 1987 to October 1997 and who elected DMPA (n = 96) or COCs (n = 430) as initial contraception were followed for a maximum of 9.2 years with a median follow-up of approximately 12 months. Oral glucose tolerance tests were performed and choice of contraception method was recorded at each visit as part of routine clinical care. RESULTS: Annual diabetes incidence rates were 19% in the DMPA group and 12% in the COC group, with an unadjusted hazard ratio (HR) of 1.58 (95% CI 1.00-2.50; P = 0.05) for DMPA compared with COCs. Adjustment for baseline imbalances reduced the HR to 1.18 (0.67-2.28; P = 0.57). Additional adjustment for weight gain during follow-up, which was on average 1.8 kg higher in the DMPA group (P < 0.0001), reduced the HR to 1.07. DMPA interacted with baseline serum triglyceride levels and, separately, with breast-feeding to increase the diabetes risk. CONCLUSIONS: DMPA use was associated with an increased risk of diabetes that appeared to be explained by three factors: 1) use in women with increased baseline diabetes risk, 2) weight gain during use, and 3) use with high baseline triglycerides and/or during breast-feeding.