Non-steroidal antiinflammatory drugs (NSAIDs) inhibit the cyclooxygenase (COX) enzyme and so they are effective analgesic, antiinflammatory and antipyretic drugs. The discovery of COX-2 led to the search for new NSAIDs with a selective action over this isoenzyme. The experiments performed to date have shown either more, less or no different efficacy of new COX-2 selective NSAIDs when compared to the non-selective inhibitors, probably because the comparison has not been performed under similar conditions. We have therefore compared the analgesic activity of six NSAIDs with different selectivity for the COX isoenzymes. The experiments were performed using the recording of spinal cord nociceptive reflexes in anaesthetised rats and in awake mice. The non-selective COX inhibitors, such as dexketoprofen trometamol, were effective in reducing nociceptive responses both in normal and monoarthritic rats (ED50s: 0.31 and 3.97 micromol/kg, respectively), and in mice with paw inflammation (12.5 micromol/kg, p < 0.01). The COX-1 selective inhibitor SC-58560 showed efficacy in normal rats (ED50: 0.8 micromol/kg) and in mice with paw inflammation (15 micromol/kg, p < 0.05), but not in monoarthritic rats. The COX-2 selective inhibitors celecoxib (105 micromol/kg) and rofecoxib (128 micromol/kg) however, were not effective in any of the groups studied. We conclude that inhibition of both COX isoenzymes is needed to achieve an effective analgesia in inflammation.