PURPOSE: To compare treatment changes after the rofecoxib withdrawal with changes occurring normally and to re-assess 12 months afterwards. METHODS: The PHARMO database comprised medication and hospital discharge records of over 3 million inhabitans in the Netherlands. The Study cohort included chronic coxib users with a coxib prescription on 30th September 2004; the Reference cohort others with a coxib prescription on 1st June 2004. Initial treatment changes were based on first new prescription since cohort entry. Twelve-month changes were studied within the Study cohort only. RESULTS: The Study cohort (n = 6974) and Reference cohort (n = 5393) had similar demographics, stratified on type of coxib. In the Study cohort, 3341 (48%) initially stopped coxibs, of whom 1121 (16%) stopped all analgesic, versus 13 and 5% in the Reference cohort (p < 0.001). Among 'other coxib' users 32% stopped coxibs, and 15% stopped all analgesics, versus 14% and 4%, p < 0.001 in the Reference cohort. Among those whostopped coxibs, 34% switched to non-selective non-steroidal anti-inflammatory drug (nsNSAID) without PPI, 21% to nsNSAID with PPI, and 45% stopped NSAID treatment (Reference cohort: 35, 20, and 44%, respectively). These rates for 'other coxib users' were: switching to nsNSAID without PPI 23% (Study Cohort) versus 35% (Reference Cohort), 13 versus 28%, and 64 versus 37% respectively (p < 0.001). Twelve months later, stopping NSAID increased to 43%, stopping all analgesics to 32%. Rheumatologists continued coxibs more frequently than other caregivers (87, 65, 54%, respectively). CONCLUSIONS: The rofecoxib withdrawal resulted in a large proportion of patients who discontinued analgesic treatment altogether regardless of original coxib therapy. Copyright(c) 2007 John Wiley & Sons, Ltd.