Few data have been published on the patterns of use COX-2 inhibitors (coxibs)and non-selective NSAIDs in "real-life" setting; a major determinant of effectiveness. OBJECTIVES: 1) To assess the rate of persistency of coxibs over the first 3 months of treatment in the elderly population; 2) To determine the rate of switching; 3) To identify characteristics associated with persistency. METHODS: A retrospective cohort study was conducted in a cumulative incidence random sample of members (age 65 +) of the Quebec drug plan who initiated a treatment with celecoxib (n = 14,396), rofecoxib (n = 6,120) or non-selective NSAIDs (n = 1,202) between January 1st and August 31st 2000. All medical services and prescriptions received by these patients during the previous year were obtained through linkage with the Quebec health services databases. Cox proportional hazard was used to identify factors associated with persistency. RESULTS: Over the first 3 months of treatment, the median duration of use was 22 days for celecoxib, 21 days for rofecoxib, and 12 days for non-selective NSAIDs. On average, 24.2% of incident celecoxib users and 17.3% of rofecoxib users were treated for more than 3 months, compared to 10.2% of non-selective NSAID users. Switches were not frequent: 4.1% of celecoxib users switched to rofecoxib, and 2.3% to non-selective NSAIDs. Of rofecoxib users, 2.9% switched to celecoxib and 2.4% to non-selective NSAIDs. Factors significantly associated with persistency were: celecoxib (OR = 1.19; 1.11-1.27), use of gastroprotective agents at treatment initiation (OR = 1.22; 1.16-1.27) or during follow-up (OR = 1.18; 1.13-1.22), high chronic disease scores (CDS 10+: OR = 1.15; 1.10-1.20; CDS 5-9: OR = 1.13 (1.09-1.17). Previous use of gastroprotective agents, indicative of a history of gastropathy, was negatively associated with persistency: (OR = 0.92; 0.88-0.95). Age, gender, dosage at initiation, prescriber's specialty, and income level did not influence persistency. CONCLUSION: Persistency beyond three months was higher for coxibs than for non-selective NSAIDs.