Nonsteroidal anti-inflammatory drugs have been a mainstay in the treatment of inflammatory diseases such as rheumatoid arthritis. However, these agents can result in severe and occasionally life-threatening adverse effects that can limit therapeutic benefit. Progress toward safer anti-inflammatory therapy was aided by the discovery that cyclooxygenase (COX) exists as two isozymes, COX-1 and COX-2. Both isozymes form prostaglandins that support physiologic functions; however, the formation of proinflammatory prostaglandins is catalyzed by COX-2. Inhibition of COX-2 accounts for the anti-inflammatory and analgesic action of NSAIDs; however, concurrent inhibition of COX-1 inhibits prostaglandin-dependent mechanisms such as gastroduodenal mucosal defense and platelet aggregation. This inhibition is the basis of the gastrointestinal toxicity and bleeding characteristic of these drugs. These findings led to the hypothesis that agents that selectively inhibit COX-2 would possess anti-inflammatory and analgesic action but would spare COX-1, thereby avoiding adverse effects in the gastrointestinal tract and platelets. Selective COX-2 inhibitors are now available. The novelty of these agents has raised questions in the medical community as to what constitutes selectivity for COX-2. This review outlines the criteria that must be met to characterize a compound as COX-2-specific. Clinical evidence of clear improvement in gastrointestinal tolerability and safety must be demonstrated in addition to complementary evidence of COX-2 selectivity obtained from enzyme, biochemical, and clinical pharmacology evaluations.