Summary Background Bone metastases are a major cause of morbidity and mortality in men with prostate cancer. Preclinical studies suggest that osteoclast inhibition might prevent bone metastases. We assessed denosumab, a fully human anti-RANKL monoclonal antibody, for prevention of bone metastasis or death in non-metastatic castration-resistant prostate cancer. Methods In this phase 3, double-blind, randomised, placebo-controlled study, men with non-metastatic castration-resistant prostate cancer at high risk of bone metastasis (prostate-specific antigen PSA ≥8·0 μg/L or PSA doubling time ≤10·0 months, or both) were enrolled at 319 centres from 30 countries. Patients were randomly assigned (1:1) via an interactive voice response system to receive subcutaneous denosumab 120 mg or subcutaneous placebo every 4 weeks. Randomisation was stratified by PSA eligibility criteria and previous or ongoing chemotherapy for prostate cancer. Patients, investigators, and all people involved in study conduct were masked to treatment allocation. The primary endpoint was bone-metastasis-free survival, a composite endpoint determined by time to first occurrence of bone metastasis (symptomatic or asymptomatic) or death from any cause. Efficacy analysis was by intention to treat. The masked treatment phase of the trial has been completed. This trial was registered at ClinicalTrials.gov , number NCT00286091. Findings 1432 patients were randomly assigned to treatment groups (716 denosumab, 716 placebo). Denosumab significantly increased bone-metastasis-free survival by a median of 4·2 months compared with placebo (median 29·5 95% CI 25·4–33·3 vs 25·2 22·2–29·5 months; hazard ratio HR 0·85, 95% CI 0·73–0·98, p=0·028). Denosumab also significantly delayed time to first bone metastasis (33·2 95% CI 29·5–38·0 vs 29·5 22·4–33·1 months; HR 0·84, 95% CI 0·71–0·98, p=0·032). Overall survival did not differ between groups (denosumab, 43·9 95% CI 40·1–not estimable months vs placebo, 44·8 40·1–not estimable months; HR 1·01, 95% CI 0·85–1·20, p=0·91). Rates of adverse events and serious adverse events were similar in both groups, except for osteonecrosis of the jaw and hypocalcaemia. 33 (5%) patients on denosumab developed osteonecrosis of the jaw versus none on placebo. Hypocalcaemia occurred in 12 (2%) patients on denosumab and two (<1%) on placebo. Interpretation This large randomised study shows that targeting of the bone microenvironment can delay bone metastasis in men with prostate cancer. Funding Amgen Inc.