Long-acting local anesthetics (LAs) are used to provide continuous perioperative pain relief by blocking the nociceptive input from the surgical site. An anti-inflammatory action of LAs has been suggested. The present study describes the effect of two commonly used LAs (lidocaine and bupivacaine) on local PGE2 production and cyclooxygenase (COX) gene expression as an indicator of their anti-inflammatory action and its relation to postoperative pain. Subjects (n = 114) undergoing the surgical removal of impacted third molars, a clinical model of acute inflammation, were randomly assigned to receive either 2% lidocaine or 0.5% bupivacaine before surgery and either rofecoxib 50 mg or placebo orally 90 minutes before surgery and for the following 48 hours. Oral mucosal biopsies were taken before surgery and 48 hours after surgery. After extraction, a microdialysis probe was placed at the surgical site for prostaglandin E₂ (PGE₂) and thromboxane B₂ (TXB₂) measurements. Pain was assessed over 4 hours and at 24 and 48 hours using visual analog scale. The bupivacaine/rofecoxib group reported significantly less pain compared to the other 3 treatment groups during the immediate postoperative period but not at 24 or 48-hours. On the other hand, the bupivacaine/placebo group reported significantly more pain at 24 hours. PGE₂ levels in the bupivacaine/placebo group were significantly higher than the other three treatment groups. Moreover, bupivacaine significantly increased COX-1 and COX-2 gene expression following tissue injury. The COX-2 gene expression was significantly higher in the bupivacaine/placebo group as compared to the lidocaine/placebo group. TXB₂ concentrations at the surgical site showed no significant differences among the four groups. This suggests that, in the absence of general anesthesia, bupivacaine stimulates COX-2 gene expression after tissue injury, which is associated with higher PGE₂ production and pain after the local anesthetic effect dissipates.