BACKGROUND: Long-term hormone therapy alone is standard care for metastatic or high-risk, non-metastatic prostate cancer. STAMPEDE-an international, open-label, randomised controlled trial-uses a novel multiarm, multistage design to assess whether the early additional use of one or two drugs (docetaxel, zoledronic acid, celecoxib, zoledronic acid and docetaxel, or zoledronic acid and celecoxib) improves survival in men starting first-line, long-term hormone therapy. Here, we report the preplanned, second intermediate analysis comparing hormone therapy plus celecoxib (arm D) with hormone therapy alone (control arm A). METHODS: Eligible patients were men with newly diagnosed or rapidly relapsing prostate cancer who were starting long-term hormone therapy for the first time. Hormone therapy was given as standard care in all trial arms, with local radiotherapy encouraged for newly diagnosed patients without distant metastasis. Randomisation was done using minimisation with a random element across seven stratification factors. Patients randomly allocated to arm D received celecoxib 400 mg twice daily, given orally, until 1 year or disease progression (including prostate-specific antigen [PSA] failure). The intermediate outcome was failure-free survival (FFS) in three activity stages; the primary outcome was overall survival in a subsequent efficacy stage. Research arms were compared pairwise against the control arm on an intention-to-treat basis. Accrual of further patients was discontinued in any research arm showing safety concerns or insufficient evidence of activity (lack of benefit) compared with the control arm. The minimum targeted activity at the second intermediate activity stage was a hazard ratio (HR) of 0·92. This trial is registered with ClinicalTrials.gov, number NCT00268476, and with Current Controlled Trials, number ISRCTN78818544. FINDINGS: 2043 patients were enrolled in the trial from Oct 17, 2005, to Jan 31, 2011, of whom 584 were randomly allocated to receive hormone therapy alone (control group; arm A) and 291 to receive hormone therapy plus celecoxib (arm D). At the preplanned analysis of the second intermediate activity stage, with 305 FFS events (209 in arm A, 96 in arm D), there was no evidence of an advantage for hormone therapy plus celecoxib over hormone therapy alone: HR 0·98 (95% CI 0·90-1·06). 2-year FFS was 51% (95% CI 46-56) in arm A and 51% (95% CI 43-58) in arm D. There was no evidence of differences in the incidence of adverse events between groups (events of grade 3 or higher were noted at any time in 123 [23%, 95% CI 20-27] patients in arm A and 64 [25%, 19-30] in arm D). The most common grade 3-5 events adverse effects in both groups were endocrine disorders (55 [11%] of patients in arm A vs 19 [7%] in arm D) and musculoskeletal disorders (30 [6%] of patients in arm A vs 15 [6%] in arm D). The independent data monitoring committee recommended stopping accrual to both celecoxib-containing arms on grounds of lack of benefit and discontinuing celecoxib for patients currently on treatment, which was endorsed by the trial steering committee. INTERPRETATION: Celecoxib 400 mg twice daily for up to 1 year is insufficiently active in patients starting hormone therapy for high-risk prostate cancer, and we do not recommend its use in this setting. Accrual continues seamlessly to the other research arms and follow-up of all arms will continue to assess effects on overall survival. FUNDING: Cancer Research UK, Pfizer, Novartis, Sanofi-Aventis, Medical Research Council (London, UK).