Recent studies have shown that cyclooxygenase exists in two isozyme forms. Since differences in the pharmacological profiles of nonsteroidal anti-inflammatory drugs (NSAIDs) might be accounted for by varying degrees of selectivity for these isozymes, cyclooxygenase-1 and -2, the relative potency of various NSAIDs in inhibiting their activities was examined in intact human cells. We used human platelets cyclooxygenase-1 and interleukin-1beta-stimulated human synovial cell cyclooxygenase-2 for measuring cyclooxygenase selectivity. The presence of the enzymes was confirmed by immunoblotting and immunoprecipitation analysis, and by the reverse transcriptase-polymerase chain reaction. Mean IC50 values (microM) for human platelet cyclooxygenase-1 and interleukin-1beta-stimulated human synovial cell cyclooxygenase-2 and cyclooxygenase-1/-2 IC50 ratio of various NSAIDs were as follows: aspirin, 3.2, 26, 0.12; diclofenac, 0.037, 0.00097, 38; etodolac, 122, 0.68, 179; ibuprofen, 3.0, 3.5, 0.86; indomethacin, 0.013, 0.044, 0.30; loxoprofen (active metabolite), 0.38, 0.12, 3.2; NS-398, 12, 0.0095, 1263; oxaprozin, 2.2, 36, 0.061; zaltoprofen, 1.3, 0.34, 3.8; respectively. Our bioassay system employing intact human cells to assess the cyclooxygenase selectivity of NSAIDs may provide clinically useful information.