Purpose Adherence to aromatase inhibitor (AI) therapy for early-stage breast cancer is limited by AI-associated musculoskeletal symptoms (AIMSS). Duloxetine is US Food and Drug Administration approved for treatment of multiple chronic pain disorders. We hypothesized that treatment of AIMSS with duloxetine would improve average joint pain compared with placebo. Methods This randomized, double-blind, phase III trial included AI-treated postmenopausal women with early-stage breast cancer and who had average joint pain score of ≥ 4 out of 10 that developed or worsened since AI therapy initiation. Patients were randomly assigned 1:1 to duloxetine or placebo for 13 weeks. The primary end point was average joint pain through 12 weeks, examined using multivariable linear mixed models, adjusted for stratification factors (baseline pain score of 4 to 6 v 7 to 10 and prior taxane use). Clinically significant change in average pain was defined as a ≥ 2-point decrease from baseline. Results Of 299 enrolled patients, 127 patients treated with duloxetine and 128 who received placebo were evaluable for the primary analysis. By 12 weeks, the average joint pain score was 0.82 points lower for patients who received duloxetine compared with those who received placebo (95% CI, -1.24 to -0.40; P = .0002). Similar patterns were observed for worst joint pain, joint stiffness, pain interference, and functioning. Rates of adverse events of any grade were higher in the duloxetine-treated group (78% v 50%); rates of grade 3 adverse events were similar. Conclusion Results of treatment with duloxetine for AIMSS were superior to those of placebo among women with early-stage breast cancer, although it resulted in more frequent low-grade toxicities.