Metformin may exert anti-cancer effects through indirect (insulin-mediated) or direct (insulin-independent) mechanisms. We report results of a neoadjuvant "window of opportunity" study of metformin in women with operable breast cancer. Newly diagnosed, untreated, non-diabetic breast cancer patients received metformin 500 mg tid after diagnostic core biopsy until definitive surgery. Clinical (weight, symptoms, and quality of life) and blood [fasting serum insulin, glucose, homeostasis model assessment (HOMA), C-reactive protein (CRP), and leptin] attributes were compared pre- and post-metformin as were terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) and Ki67 scores (our primary endpoint) in tumor tissue. Thirty-nine patients completed the study. Mean age was 51 years, and metformin was administered for a median of 18 days (range 13-40) up to the evening prior to surgery. 51 % had T1 cancers, 38 % had positive nodes, 85 % had ER and/or PgR positive tumors, and 13 % had HER2 overexpressing or amplified tumors. Mild, self-limiting nausea, diarrhea, anorexia, and abdominal bloating were present in 50, 50, 41, and 32 % of patients, respectively, but no significant decreases were seen on the EORTC30-QLQ function scales. Body mass index (BMI) (-0.5 kg/m(2), p < 0.0001), weight (-1.2 kg, p < 0.0001), and HOMA (-0.21, p = 0.047) decreased significantly while non-significant decreases were seen in insulin (-4.7 pmol/L, p = 0.07), leptin (-1.3 ng/mL, p = 0.15) and CRP (-0.2 mg/L, p = 0.35). Ki67 staining in invasive tumor tissue decreased (from 36.5 to 33.5 %, p = 0.016) and TUNEL staining increased (from 0.56 to 1.05, p = 0.004). Short-term preoperative metformin was well tolerated and resulted in clinical and cellular changes consistent with beneficial anti-cancer effects; evaluation of the clinical relevance of these findings in adequately powered clinical trials using clinical endpoints such as survival is needed.