Numerous in vitro assays have been developed for testing and comparing the relative inhibitory activities of non-steroidal anti-inflammatory drugs against cyclooxygenase (COX)-1 and COX-2. Despite variability among these systems, which precludes direct comparison of data, analysis of the ratio of inhibition of COX-1 to COX-2 by non-steroidal anti-inflammatory drugs, suggests inhibitors can be classified based on their COX selectivity. Standard non-steroidal anti-inflammatory drugs can be considered nonselective; compounds such as meloxicam and nimesulide can be classified as COX-2 preferential; and compounds such as SC 58125 and L-754,337 are selective for COX-2. Although in vitro systems are important for characterizing COX-1 and COX-2 inhibitory activity, the clinical relevance of these data should be considered carefully. The level of inhibition of COX-1 and COX-2, in vivo at a given dose in patients, cannot be predicted from in vitro data alone. The pharmacokinetic properties of each compound, including plasma levels, distribution and binding to plasma proteins, have to be taken into account. Human pharmacology studies concentrating on the inhibition of prostanoid synthesis in target tissues are of paramount importance in determining the clinical relevance of COX-2 selectivity.