OBJECTIVES: The gastrointestinal (GI) risks associated with selective cyclooxygenase-2 inhibitors (COX-2s) versus non-selective non-steroidal anti-inflammatory drugs (NSAIDs) among arthritis patients are well documented in clinical trials. This study is to estimate the major GI risks among elderly chronic users of COX-2s versus NSAIDs, with/without aspirin (ASA), in clinical practice. METHODS: A cohort study was conducted using secondary data from the GE logician database (Centricity EMR), which contained medical records of 3 million patients seen by 5,000 physicians across 27 states. Inclusion criteria: chronic use (2 or more medication mentions) of COX-2s or NSAIDs within 60 days between 1/1/1999 and 6/30/2003, 65 or older, no switch between COX-2s and NSAIDs during one-year follow-up or before a major GI event, defined as GI hemorrhage including melena (ICD-9 codes: 578.xx). Descriptive and multivariate logistic analyses were conducted to determine how major GI risks differed across chronic users of COX-2s alone, NSAIDs alone, COX-2s+ASA, and NSAIDs+ASA. The logistic analysis controlled for gender, age, pre- or post-index GI-harmful drug use, major and minor GI events in the year prior to index date, and prior GI-protective drug use. RESULTS: The number of patients and the percent having major GI events during one-year follow-up period were as follows: COX-2s-alone 7,338 (1.73%); NSAIDs-alone 3,826 (2.06%); COX-2s+ASA 963 (1.77%); and NSAIDs+ASA 602 (2.66%). The multivariate logistic results showed that compared to COX-2s-alone users, NSAIDs-alone and NSAIDs+ASA users had higher major GI risks (OR=1.35, p=0.04, 95% CI: 1.01-1.80; and OR=1.68, p=0.06, 95% CI: 0.99-2.86 respectively). COX-2s+ASA users had similar risks (OR=0.96, p=0.88, 95% CI: 0.57-1.61) to COX-2s-alone users. CONCLUSION: The major GI risk was highest among elderly chronic users of NSAIDs+ASA, followed by NSAIDs-alone. Only NSAIDs-alone users had a statistically significant higher risk than COX-2s-alone users. The addition of ASA did not significantly increase major GI risk among COX-2 users.