Nociception evoked prostaglandin (PG) release in the spinal cord considerably contributes to the induction of hyperalgesia and allodynia. To evaluate the relative contribution of cyclooxygenase-1 (COX-1) and COX-2 in this process we assessed the effects of the selective COX-1 inhibitor SC560 and the selective COX-2 inhibitor celecoxib on formalin-evoked nociceptive behaviour and spinal PGE(2) release. SC560 (10 and 20 mg/kg) significantly reduced the nociceptive response and completely abolished the formalin-evoked PGE(2) raise. In contrast, celecoxib (10 and 20 mg/kg) was ineffective in both regards, i.e. the flinching behaviour was largely unaltered and the formalin-induced PGE(2) raise as assessed using microdialysis was only slightly, not significantly reduced. This suggests that the formalin-evoked rapid PG release was primarily caused by COX-1 and was independent of COX-2. Mean free spinal cord concentrations of celecoxib during the formalin assay were 32.0 +/- 4.5 nM, thus considerably higher than the reported IC50 for COX-2 (3-7 nM). Therefore, the lack of efficacy of celecoxib is most likely not to be a result of poor tissue distribution. COX-2 mRNA and protein expression in the spinal cord were not affected by microdialysis alone but the mRNA rapidly increased following formalin injection and reached a maximum at 2 h. COX-2 protein was unaltered up to 4 h after formalin injection. The time course of COX-2 up-regulation suggests that the formalin-induced nociceptive response precedes COX-2 protein de novo synthesis and may therefore be unresponsive to COX-2 inhibition. Considering the results obtained with the formalin model it may be hypothesized that the efficacy of celecoxib in early injury evoked pain may be less than that of unselective NSAIDs.