BACKGROUND Nonsteroidal anti-inflammatory drugs are the most widely used agents in the symptomatic treatment of osteoarthritis (OA). No data are presently available on the medium-term management of this disease with an on-demand treatment regimen, which nevertheless reflects medical practice. OBJECTIVE The aim of this study was to compare nimesulide-beta-cyclodextrin and naproxen in terms of short-term (2 weeks) pain control with scheduled dosing and medium-term (5.5 months) pain control with on-demand dosing in patients with OA. METHODS In this multicenter, randomized, double-blind, controlled study, we compared 2 weeks of scheduled treatment plus 5.5 months of on-demand treatment in patients with OA of the hip and/or knee and moderate to severe pain, with no important concomitant disorders. Treatment consisted of nimesulide-beta-cyclodextrin (400 mg BID, orally = 100 mg nimesulide BID) or naproxen (500 mg BID). The primary outcome measures for scheduled dosing were pain on movement (measured by visual analog scale), morning stiffness score, Lequesne index, and adverse events. For on-demand dosing, the measures were the same as for scheduled dosing, plus duration of treatment and global assessment of efficacy and tolerability by patient and physician. RESULTS After 2 weeks, there was equivalent reduction from baseline in pain on movement in the 2 treatment groups (nimesulide-beta-cyclodextrin, -41.5%; naproxen, -40.5%); the reduction was significant after 1 week (P < 0.001). The findings were also similar for the morning stiffness score and Lequesne index. There were no significant differences in mean duration of on-demand treatment (nimesulide-beta-cyclodextrin, 163.03 days; naproxen, 166.3 days) or in mean consumption of study drug (nimesulide-beta-cyclodextrin, 0.85 +/- 0.61 sachets/d; naproxen, 0.74 +/- 0.42 sachets/d). Withdrawal due to intolerance occurred in 8 patients given nimesulide-beta-cyclodextrin and 13 patients given naproxen, with no significant difference between groups; 3 and 12 patients, respectively, withdrew due to gastrointestinal intolerance, a finding that was significantly different between groups (P < 0.01). Global assessment of efficacy by patient and physician was similar for both drugs. Assessment of tolerability significantly favored nimesulide-beta-cyclodextrin on the physician assessments (P < 0.05) but was similar for the 2 drugs on the patient assessments (physicians, 46.9% vs 30.9%; patients, 43.5% vs 33.3%). CONCLUSIONS The results suggest that nimesulide-beta-cyclodextrin provides similar pain relief to naproxen in the management of OA of the hip and/or knee and is associated with fewer gastrointestinal adverse reactions. On-demand dosing may be an effective and well-tolerated low-dose regimen of nonsteroidal anti-inflammatory drugs for the maintenance of pain control in OA in the medium term.