Purpose: Both isoforms of cyclo-oxygenase, COX-1 and COX-2, are inhibited to varying degrees by all of the available nonsteroidal anti-inflammatory drugs (NSAIDs). Because inhibition of COX-1 by NSAIDs is linked to gastrointestinal ulcer formation, those drugs that selectively inhibit COX-2 may have less gastrointestinal toxicity. We measured the extent to which NSAIDs and other anti-inflammatory or analgesic drugs inhibit COX-1 and COX-2 in humans.

Subjects and methods: Aliquots of whole blood from 16 healthy volunteers were incubated ex vivo with 25 antiinflammatory or analgesic drugs at six concentrations ranging from 0 (control) to 100 microM (n = 5 for each). Blood was assayed for serum-generated thromboxane B2 synthesis (COX-1 assay) and for lipopolysaccharide-stimulated prostaglandin E2 synthesis (COX-2 assay). In addition, gastric biopsies from the same volunteers were incubated with each drug ex vivo and mucosal prostaglandin E2 synthesis measured.

Results: Inhibitory potency and selectivity of NSAIDs for COX-1 and COX-2 activity in blood varied greatly. Some NSAIDs (eg, flurbiprofen, ketoprofen) were COX-1 selective, some (eg, ibuprofen, naproxen) were essentially nonselective, while others (eg, diclofenac, mefenamic acid) were COX-2 selective. Inhibitory effects of NSAIDs on gastric prostaglandin E2 synthesis correlated with COX-1 inhibitory potency in blood (P < 0.001) and with COX-1 selectivity (P < 0.01), but not with COX-2 inhibitory potency. Even COX-2 "selective" NSAIDs still had sufficient COX-1 activity to cause potent inhibitory effects on gastric prostaglandin E2 synthesis at concentrations achieved in vivo.

Conclusion: No currently marketed NSAID, even those that are COX-2 selective, spare gastric COX activity at therapeutic concentrations. Thus, all NSAIDs should be used cautiously until safer agents are developed.