BACKGROUND: Prior authorizations (PA) are intended to promote safe and cost-effective medication use. Unwanted outcomes may occur, however, such as a patient forgoing drug therapy after a PA. The label for rosiglitazone was revised in November 2007 to include the warning of contraindicated use with nitrates or insulin, creating an opportunity for a PA directed at safe use. OBJECTIVE: To evaluate antidiabetic drug utilization after the implementation of an electronic PA that denied a claim for rosiglitazone if the patient had a history of either insulin or nitrate supply in the previous 60 days. METHODS: This quasi-experimental study used pharmacy claims for 1.4 million commercially insured members who were exposed to a rosiglitazone PA beginning on January 1, 2009, compared with a group of approximately 2 million commercially insured members who did not have this safety PA intervention. Continuously enrolled members were identified who had a rejected (intervention group) or paid (comparison group) claim for rosiglitazone during the period from January 1, 2009, through June 30, 2009. Pharmacy claims were assessed for the presence of nitrates, insulin, rosiglitazone, other antidiabetic therapy, or no antidiabetic therapy supply on days 30, 60, 90, and 180 after the rejected/paid claim. A time-series analysis using rosiglitazone claims for all health plan members from January 2008 through December 2009 was used to evaluate the impact of the PA on rosiglitazone utilization overall. RESULTS: At 30 days, there were 134 patients (60.4% of 222) in the comparison group with concurrent supply of rosiglitazone with insulin and/or nitrates versus 4 patients (2.4% of 168, P qqqlt; 0.001) in the PA intervention group, and the utilization rate remained significantly higher at 180 days in the comparison group (37.8%, n = 84) versus the PA group (2.4%, n = 4, P qqqlt; 0.001). Beginning at 60 days, there was no significant difference in the percentage of members with no antidiabetic therapy in the comparison and PA intervention groups (9.9% vs. 15.5%, respectively, P = 0.133), and the rates remained similar through 180 days (15.3% vs. 13.7%, respectively, P = 0.760). The PA was associated with an absolute decrease of 5.1 average monthly rosiglitazone claims per day per million members (P qqqlt; 0.001). CONCLUSIONS: This PA, intended to reduce known cardiovascular event risks among health plan members with type 2 diabetes, was associated with a significant reduction in concurrent use of rosiglitazone with nitrates or insulin.