BACKGROUND: There is a paucity of effective systemic therapy options for patients with advanced, chemotherapy-refractory colorectal cancer. We aimed to evaluate the efficacy and safety of fruquintinib, a highly selective and potent oral inhibitor of vascular endothelial growth factor receptors (VEGFRs) 1, 2, and 3, in patients with heavily pretreated metastatic colorectal cancer. METHODS: We conducted an international, randomised, double-blind, placebo-controlled, phase 3 study (FRESCO-2) at 124 hospitals and cancer centres across 14 countries. We included patients aged 18 years or older (≥20 years in Japan) with histologically or cytologically documented metastatic colorectal adenocarcinoma who had received all current standard approved cytotoxic and targeted therapies and progressed on or were intolerant to trifluridine-tipiracil or regorafenib, or both. Eligible patients were randomly assigned (2:1) to receive fruquintinib (5 mg capsule) or matched placebo orally once daily on days 1-21 in 28-day cycles, plus best supportive care. Stratification factors were previous trifluridine-tipiracil or regorafenib, or both, RAS mutation status, and duration of metastatic disease. Patients, investigators, study site personnel, and sponsors, except for selected sponsor pharmacovigilance personnel, were masked to study group assignments. The primary endpoint was overall survival, defined as the time from randomisation to death from any cause. A non-binding futility analysis was done when approximately one-third of the expected overall survival events had occurred. Final analysis occurred after 480 overall survival events. This study is registered with ClinicalTrials.gov, NCT04322539, and EudraCT, 2020-000158-88, and is ongoing but not recruiting. FINDINGS: Between Aug 12, 2020, and Dec 2, 2021, 934 patients were assessed for eligibility and 691 were enrolled and randomly assigned to receive fruquintinib (n=461) or placebo (n=230). Patients had received a median of 4 lines (IQR 3-6) of previous systemic therapy for metastatic disease, and 502 (73%) of 691 patients had received more than 3 lines. Median overall survival was 7·4 months (95% CI 6·7-8·2) in the fruquintinib group versus 4·8 months (4·0-5·8) in the placebo group (hazard ratio 0·66, 95% CI 0·55-0·80; p<0·0001). Grade 3 or worse adverse events occurred in 286 (63%) of 456 patients who received fruquintinib and 116 (50%) of 230 who received placebo; the most common grade 3 or worse adverse events in the fruquintinib group included hypertension (n=62 [14%]), asthenia (n=35 [8%]), and hand-foot syndrome (n=29 [6%]). There was one treatment-related death in each group (intestinal perforation in the fruquintinib group and cardiac arrest in the placebo group). INTERPRETATION: Fruquintinib treatment resulted in a significant and clinically meaningful benefit in overall survival compared with placebo in patients with refractory metastatic colorectal cancer. These data support the use of fruquintinib as a global treatment option for patients with refractory metastatic colorectal cancer. Ongoing analysis of the quality of life data will further establish the clinical benefit of fruquintinib in this patient population. FUNDING: HUTCHMED.