Comparative incidence of upper gastrointestinal bleeding associated with individual non-steroidal anti-inflammatory drugs.

OBJECTIVE: The existence of two isoforms of cyclo-oxygenase (COX), COX-1 and COX-2, is now well established. Because inhibition of COX-1 by nonsteroidal anti-inflammatory drugs (NSAIDs) is linked to gastrointestinal (GI) damage, agents with a better COX-2/COX-1 inhibition ratio may have less GI toxicity. The aim of this study was to compare the incidence of upper gastrointestinal bleeding (UGIB) in association with specific NSAIDs including the new "COX-2 preferential" inhibitors. PATIENTS Y METHOD: Individual incidence of NSAID-related UGIB in a health-authority area was estimated, based on cases of UGIB, sales of non-aspirin NSAIDs, and data of used doses in community subjects, during a four-year period. Comparisons were made by calculating individual rate-ratio (RR) and 95% confidence interval (CI) taking as a reference the specific NSAID with the lowest incidence of UGIB. RESULTS: The incidence of UGIB associated with thirteen specific NSAIDs included in the study varied greatly, from 1.7 per 1,000 person-years for aceclofenac to 25.8 per 1,000 person-years for ketorolac. The use of "COX-1 preferential" inhibitors versus "COX-2 preferential" inhibitors was associated with a RR of 5.3 (95% CI, 2.78-10.04), and between NSAIDs with "COX-1-COX-2 mixed" inhibition RR was 2.2 (95% CI, 1.13-4.28). CONCLUSIONS: Our results show that there are differences in GI toxicity according to specific NSAIDs. A substantial reduction in number of cases of UGIB could result from the use of NSAIDs with a "COX-2 preferential" inhibition.