To determine whether the cyclooxygenase-2 (COX-2) inhibitor celecoxib affects cardiovascular thrombotic risk, we analyzed the incidence of cardiovascular events for celecoxib, placebo, and nonsteroidal anti-inflammatory drugs (NSAIDs) in the entire controlled, arthritis clinical trial database for celecoxib. The primary analysis used the Antiplatelet Trialists' Collaboration end points, which include: (1) cardiovascular, hemorrhagic, and unknown deaths, (2) nonfatal myocardial infarction, and (3) nonfatal stroke. Other secondary thrombotic events were also examined. Separate analyses were performed for all patients and for those not taking aspirin. Data from all controlled, completed arthritis trials of > or =4 weeks duration, including 13 new drug application studies and 2 large post-marketing trials (CLASS and SUCCESS) were included for analyses. Patients were randomized to celecoxib at doses from 100 to 400 mg twice daily (18,942 patients; 5,668.2 patient-years of exposure), diclofenac 50 to 75 mg twice daily, ibuprofen 800 mg thrice daily, naproxen 500 mg twice daily (combined NSAID exposure of 11,143 patients; 3,612.2 patient-years), or placebo (1,794 subjects; 199.9 subject-years). Data from a long-term uncontrolled trial with 5,209 patients (6,950 patients-years) treated with celecoxib were included in a supplemental analysis. The entire 15-trial database was searched for possible serious thrombotic events as well as to identify all deaths. For these patients, detailed clinical data were obtained and reviewed by 2 of the investigators (WBW and JSB), who were independently and blinded to exposure, to classify the event as primary, secondary, or neither. All analyses were done using the intent-to-treat population, and time-to-event analyses were performed using per-patient data. To examine heterogeneity of results among studies, tests of interaction were performed using the Cox model. Incidences of the primary and secondary events were not significantly different between the celecoxib and placebo groups, nor for the celecoxib group compared with the NSAIDs group, regardless of aspirin use and NSAID type. The relative risks comparing celecoxib with the NSAIDs for the primary events were 1.06 (95% confidence interval 0.70 to 1.61, p = 0.79) for all patients, and 0.86 (95% confidence interval 0.48 to 1.56, p = 0.62) for the subgroup not taking aspirin. Similarly, for secondary cardiovascular end points, all relative risks were < or =1 for celecoxib compared with either placebo or NSAIDs. These comparative analyses demonstrate no evidence of increased risk of cardiovascular thrombotic events associated with celecoxib compared with either conventional NSAIDs or placebo.