Treatment with highly selective cyclooxygenase-2 inhibitors is associated with significantly fewer serious adverse gastrointestinal events than is treatment with non-selective NSAIDs, provided that the drug employed inhibits COX-2 but not COX-1 at therapeutic plasma levels. Several factors might influence the gastrointestinal (GI) safety of a COX-2 inhibitor administered to an individual patient. These factors include pharmacokinetic and pharmacodynamic variables (e.g. COX-2 selectivity), the interaction of these features with preexisting risk factors for drug-dependent adverse effects, as well as the variability in the individual response. Biochemical selectivity is one of the determinants of the risk of experiencing a serious GI complication during long-term NSAID therapy. The wider the separation between the COX-2 and COX-1 dose-response curves of the inhibitor (an index of biochemical selectivity), the lower the probability of experiencing a clinically relevant inhibition of platelet COX-1 due to an unusually high drug level or intense pharmacodynamic response to a normal drug level. The clinical relevance of biochemical selectivity has to be studied in large GI outcome trials with adequate statistical power to detect realistic differences in these relatively rare events.