Cyclo-oxygenase-2 selective inhibitors are frequently used to manage osteoarthritis. We compared the analgesic efficacy of the novel cyclo-oxygenase-2 selective inhibitor lumiracoxib (Prexige®) versus placebo and celecoxib in patients with knee osteoarthritis. This seven day, double-blind, placebo and active comparator controlled, parallel group study included 364 patients aged ≥50 years with moderate-to-severe symptomatic knee osteoarthritis. Patients received lumiracoxib 400 mg/day (four times the recommended chronic dose in osteoarthritis; n = 144), placebo (n = 75), or celecoxib 200 mg twice daily (n = 145). The primary variable was actual pain intensity difference (100 mm visual–analogue scale) between baseline and the mean of three hour and five hour assessments after the first dose. Actual pain intensity difference, average and worst pain, pain relief and functional status (Western Ontario and McMaster Universities Osteoarthritis Index [WOMAC™]) were measured over seven days. Patients also completed a global evaluation of treatment effect at study end or premature discontinuation. For the primary variable, the superiority of lumiracoxib versus placebo, the noninferiority of lumiracoxib versus celecoxib, and the superiority of lumiracoxib versus celecoxib were assessed by closed test procedure adjusting for multiplicity, thereby maintaining the overall 5% significance level. In addition, celecoxib was assessed versus placebo in a predefined exploratory manner to assess trial sensitivity. Lumiracoxib provided better analgesia than placebo 3–5 hours after the first dose (P = 0.004) through to study end. The estimated difference between lumiracoxib and celecoxib 3–5 hours after the first dose was not significant (P = 0.185). Celecoxib was not significantly different from placebo in this analysis (P = 0.069). At study end 13.9% of lumiracoxib-treated patients reported complete pain relief versus 5.5% and 5.3% of celecoxib and placebo recipients, respectively. WOMAC™ total and subscales improved for both active treatments versus placebo except for difficulty in performing daily activities, for which celecoxib just failed to achieve significance (P = 0.056). In the patient's global evaluation of treatment effect, 58.1% of patients receiving lumiracoxib rated treatment as 'excellent' or 'good', versus 48.6% of celecoxib and 25.3% of placebo patients. Lumiracoxib was well tolerated. The overall incidence of adverse events was similar across treatment groups.