Background: Endocrine treatment is recommended by clinical guidelines as the preferred treatment option for premenopausal as well as postmenopausal women with hormone receptor-positive, HER2-negative metastatic breast cancer. In real-world clinical practice, however, a substantial number of patients are treated with chemotherapy. We aimed to compare the clinical antitumour activity and safety of palbociclib plus endocrine therapy with that of capecitabine chemotherapy in premenopausal women with hormone receptor-positive, HER2-negative metastatic breast cancer. Methods: This multicentre, open-label, randomised, phase 2 study was done in 14 academic institutions in South Korea. Premenopausal women aged 19 years or older with hormone receptor-positive, HER2-negative breast cancer that had relapsed or progressed during previous tamoxifen therapy and with an Eastern Cooperative Oncology Group performance status of 0–2 were included. One line of previous chemotherapy for metastatic breast cancer was allowed. Patients were randomly assigned, using a random permuted block design (with a block size of two), to receive palbociclib plus combination endocrine therapy (oral exemestane 25 mg per day for 28 days and oral palbociclib 125 mg per day for 21 days every 4 weeks plus leuprolide 3·75 mg subcutaneously every 4 weeks) or chemotherapy (oral capecitabine 1250 mg/m2 twice daily for 2 weeks every 3 weeks). Randomisation was stratified by previous chemotherapy for metastatic breast cancer and visceral metastasis. The primary endpoint was progression-free survival. All analyses were done in a modified intention-to-treat population that excluded patients who did not receive study medication. This study is registered with ClinicalTrials.gov, NCT02592746, and is ongoing for follow-up of overall survival. Findings: Between June 15, 2016, and Dec 10, 2018, 189 patients were enrolled, of whom 184 were randomly assigned to the palbociclib plus endocrine therapy group (n=92) or the capecitabine group (n=92). Six patients in the capecitabine group withdrew from the study before drug administration; therefore, 92 patients in the palbociclib plus endocrine therapy group and 86 patients in the capecitabine group were included in the modified intention-to-treat analyses. 46 (50%) of 92 patients in the palbociclib plus endocrine therapy group and 45 (51%) of 92 in the capecitabine group were treatment naive for metastatic breast cancer. During a median follow-up of 17 months (IQR 9–22), median progression-free survival was 20·1 months (95% CI 14·2–21·8) in the palbociclib plus endocrine therapy group versus 14·4 months (12·1–17·0) in the capecitabine group (hazard ratio 0·659 [95% CI 0·437–0·994], one-sided log-rank p=0·0235). Treatment-related grade 3 or worse neutropenia was more common in the palbociclib plus endocrine therapy group than in the capecitabine group (69 [75%] of 92 vs 14 [16%] of 86 patients). 2 (2%) patients in the palbociclib plus endocrine therapy group and 15 (17%) patients in the capecitabine group had treatment-related serious adverse events. No treatment-related deaths occurred. Interpretation: Exemestane plus palbociclib with ovarian function suppression showed clinical benefit compared with capecitabine in terms of improved progression-free survival in premenopausal patients with hormone receptor-positive, HER2-negative metastatic breast cancer. Palbociclib plus exemestane with ovarian suppression is an active treatment option in premenopausal patients with hormone receptor-positive, HER2-negative metastatic breast cancer who have been pretreated with tamoxifen. Funding: Pfizer, Shinpoong, and Daewoong Korea and Takeda.