Abstract Metformin is associated with lower breast cancer risk in epidemiologic studies and showed decreased proliferation in HER2-positive breast cancer in a presurgical trial. To provide insight into its preventive potential, we measured proliferation by Ki-67 labeling index (LI) of intraepithelial lesions surrounding breast cancer. We randomly assigned 200 nondiabetic patients diagnosed with invasive breast cancer in core biopsies to metformin, 1,700 mg or placebo once daily for 28 days before surgery. Upon surgery, five to seven specimens of cancer adjacent (≤1 cm) and distant (>1 cm) tissue were screened for LCIS, ductal carcinoma in situ (DCIS), and ductal hyperplasia (DH). The prevalence of LCIS, DCIS, and DH was 4.5% (9/200), 67% (133/200), and 35% (69/200), respectively. Overall, metformin did not affect Ki-67 LI in premalignant disorders. The median posttreatment Ki-67 LI (IQR) in the metformin and placebo arm was, respectively, 15% (5–15) versus 5% (4–6) in LCIS (P = 0.1), 12% (8–20) versus 10% (7–24) in DCIS (P = 0.9), and 3% (1–4) versus 3% (1–4) in DH (P = 0.5). However, posttreatment Ki-67 in HER2-positive DCIS lesions was significantly lower in women randomized to metformin especially when ER was coexpressed: 22% (11–32) versus 35% (30–40) in HER2-positive DCIS (n = 22, P = .06); 12% (7–18) versus 32% (27–42) in ER-positive/HER2-positive DCIS (n = 15, P = .004). Eight of 22 (36%) HER2-positive DCIS were adjacent to HER2-negative invasive breast cancer. In tissue samples obtained following 4 weeks of study drug, proliferation was lower in HER2-positive DCIS for women randomized to metformin versus placebo. An adjuvant trial incorporating metformin in HER2-positive DCIS is warranted. Cancer Prev Res; 8(10); 888–94. ©2015 AACR.