STUDY OBJECTIVE To evaluate the effects of a cyclooxygenase (COX)-2 inhibitor prior authorization (PA) program on direct medical and pharmacy costs. DESIGN Prospective, pre- and postimplementation cohort study with reference group. SETTING Large corporation in the Midwest. PATIENTS Of 26,375 continuously enrolled members, 737 used a COX-2 inhibitor in the 3 months before January 1, 2003, when the PA program was implemented. MEASUREMENT AND MAIN RESULTS The PA program limits coverage for a COX-2 inhibitor to members with a documented risk for a nonselective nonsteroidal antiinflammatory drug (NSAID)-induced gastrointestinal adverse event. All pharmacy and medical claims and costs were analyzed from the payer's perspective for a 15-month period. Separate pharmacy cost comparisons and medical cost comparisons were made between the 3-month quarter before PA program implementation and each follow-up quarter after PA program implementation. In the 3 months after PA program implementation, 620 (84.1%) of 737 members had no claims for a COX-2 inhibitor, and during this period their pharmacy and medical costs initially declined 40.0% (p < 0.001) and 18.7% (p < 0.001), respectively, and remained significantly lower. Among a subgroup of 156 members who tried to fill a COX-2 inhibitor prescription but were denied coverage, pharmacy and medical costs initially declined, 48.1% (p < 0.001) and 10.3% (p < 0.001), respectively, with pharmacy costs remaining significantly lower; however, overall medical expenditures increased, then returned to baseline. No change was noted in physician outpatient encounters, and two members had an emergency department visit for abdominal pain with no gastrointestinal ulcerations or bleeds during the 12-month follow-up. CONCLUSION Among members denied coverage for a COX-2 inhibitor after implementation of a PA program, pharmacy costs declined without a medical cost increase associated with gastrointestinal diagnoses.