Objective: To understand the impact of the October 2004 withdrawal of rofecoxib on prescription analgesic use for arthritis patients who had been taking this medication. Methods: Patients were selected from the MarketScan databases who, during January-September 2004, had a diagnosis of osteoarthritis on a medical claim and who filled prescriptions for at least 90 days of therapy with rofecoxib, an alternative COX-2 inhibitor (celecoxib), or a branded, non-selective, nonsteroidal, anti-inflammatory drug (NSAID) (meloxicam). For each cohort, use of prescription analgesics was tracked pre-withdrawal (January-September 2004) and post-withdrawal (2005-2006). The impact of the withdrawal on drug utilization by former rofecoxib users was assessed by comparing changes in their drug consumption with that of celecoxib or meloxicam users during the same time period. Results: The study cohorts included: 29,438 rofecoxib , 34,937 celecoxib, and 7131 meloxicam users. The groups were demographically similar, except for a higher proportion of women in the meloxicam cohort. In 2005, former rofecoxib users primarily substituted other COX-2 inhibitors (22% celecoxib, 10% valdecoxib) or non-selective NSAIDS (which increased to 48% from 14% pre-withdrawal). There was no substantial change in use of opioid /other analgesics. The percentage of former rofecoxib users continuing to receive any prescription analgesic was 83% (2005) and 77% (2006). By comparison, overall utilization of celecoxib and meloxicam declined to 87% and 91% respectively in 2005 and 80% and 84% respectively in 2006. By 2006, 59% of celecoxib users and 53% of meloxicam users had discontinued these medications. Conclusion: By 2006, the proportion of former rofecoxib users who had discontinued prescription analgesics altogether was substantially higher than among patients who not previously receiving a COX-2 inhibitor. This suggests the potential harm from unsafe drugs may not be limited to side effects, but also includes the disruption of care due to withdrawal of effective treatment.