We compared the effect of dapagliflozin versus glibenclamide on the ratio of lean-to total mass in patients with type 2 diabetes mellitus, carotid subclinical atherosclerosis, HbA1c 7.0-9.0% and 40-70 years-old. Ninety-eight patients (61% male; mean age 57 ± 7 years) were randomized into dapagliflozin 10 mg/day or glibenclamide 5 mg/day on top of metformin. Body composition was measured by Dual Energy X-Ray at randomization and after 12 weeks of treatment. Glycemic control was equivalent in both groups. Dapagliflozin decreased total body mass (-2741 g [95% CI: -3360 to 1945]; p < 0.001) and lean mass (-347 g [95% CI: -761 to -106]; p < 0.001), while glibenclamide increased total body mass (1060 g [95% CI: 140 to 1836]; p < 0.001) and lean mass (929 g [95% CI: 575 to 1283]; p < 0.001) for the differences between arms. The lean-to-total mass ratio increased by 1.2% in the dapagliflozin group and 0,018% in the glibenclamide group (p < 0.001). Dapagliflozin reduced the risk of a negative balance in the lean-to total mass ratio [OR: 0.16 (95% CI: 0.05 to 0.45); p < 0.001] even after adjustment for baseline lean-to total mass ratio, waist circumference, HOMAIR, HbA1c, mean of the two hands handgrip strength and gait speed [OR: 0.13 (95% CI: 0.03-0.57); p < 0.007]. In conclusion, under equivalent glycemic control, dapagliflozin reduced total body mass but increased the ratio of lean-to-total mass when compared with glibenclamide.