Cyclooxygenase-2 (COX-2) after induction peripherally, and within the CNS, plays an important role in producing inflammatory pain. However, its role in neuropathic pain models is controversial. Recently a robust and persistent model of partial nerve injury pain, the spared nerve injury (SNI) model, has been developed. The aim of the present study was to examine the regulation of COX-2 in the rat SNI model and to evaluate the effectiveness of the selective COX-2 inhibitor rofecoxib in preventing neuropathic allodynia and hyperalgesia. RNase protection assays revealed only a very small and transient increase in COX-2 mRNA in the dorsal horn of the spinal cord in the SNI model with a maximum change at 24 h. Immunohistochemical analysis showed a small increase in COX-2 protein in the deep layers of the dorsal horn 10 h following SNI surgery. Rofecoxib (100 microM) did not affect spontaneous excitatory postsynaptic currents or alpha-amino-3-hydroxy-5-methyl-4-isoxazole propanoic acid (AMPA) and N-methyl-d-aspartate (NMDA) responses in lamina II neurons from spinal cords of animals with SNI indicating no detectable action on transmitter release or postsynaptic activity. Furthermore, rofecoxib treatment (1 and 3.2 mg/kg for 5 and 3 days respectively starting on the day of surgery) failed to modify the development of allodynia and hyperalgesia in the SNI model. However, rofecoxib significantly reduced inflammatory hypersensitivity evoked by injection of complete Freund's adjuvant into one hindpaw, indicating that the doses used were pharmacologically active. The pain hypersensitivity produced by the SNI model is not COX-2-dependent.