Metformin (MET) has been reported to have antidepressant effects in animal models and in diabetic patients with depression, owing to its anti-inflammatory, antioxidant, and neuroprotective activity. Accordingly, we proposed that MET would show antidepressant effects in patients with major depressive disorder (MDD) without other comorbidities. In this double-blind placebo-controlled study, 80 adult outpatients with MDD (DSM-IV criteria) and a Hamilton Depression Rating Scale (HAM-D) score >18 were randomized to receive fluoxetine 20 mg once daily plus placebo (n = 40) or fluoxetine 20 mg once daily plus MET 1000 mg once daily for 12 weeks. Patients were assessed by HAM-D score (weeks 0, 4, 8, and 12). The serum levels of TNF-α, IL-1β, IL-6, IGF-1, MDA, CRP, BDNF, and serotonin were measured before and after therapy. Mixed-effects model repeated-measures analysis of covariance was used to compare the HAM-D scores and the biological markers between the two groups. After 4, 8 and 12 weeks, patients in the MET group showed a statistically significant decline in HAM-D score relative to the placebo group (least squares mean difference [LSMD] -2.347, p = 0.000, LSMD -3.369, p = 0.000, and LSMD -3.454, p = 0.000, respectively). Response and remission rates were significantly higher in the MET group (89% and 81%, respectively) than in the placebo group (59% and 46%, respectively). Moreover, the MET group was superior in conserving the measured biological markers compared with the placebo group. Our findings suggest MET as a promising, effective, and safe short-term adjunctive approach in nondiabetic MDD patients. Trial registration ID: NCT04088448.