Selective cyclooxygenase (COX)-2 inhibitors have been associated with an increased risk of thromboembolic cardiovascular (CV) events. Prior studies have found that rofecoxib has a destabilizing effect on blood pressure; however, whether this translates to an increased risk of thromboembolic CV events is unknown. The objective of this study was to evaluate risk of thromboembolic CV events among hypertensive and nonhypertensive patients treated with rofecoxib or celecoxib, nonselective nonsteroidal anti-inflammatory drugs (ns-NSAIDs), or no NSAIDs (nonusers). This was a retrospective cohort study of 31,743 adult arthritis patients enrolled in a Blue Cross/Blue Shield health insurance plan in the northeastern United States. The main outcome measure was incident acute myocardial infarction and stroke. A clinically significant channeling effect was observed where selective COX-2 inhibitor users had a more severe CV risk profile. Among normotensive patients, the hazard ratio (HR) of CV events for ns-NSAIDs, rofecoxib, or celecoxib versus nonusers was 0.91 (95% confidence interval, 0.68-1.21), 1.05 (0.61-1.80), and 1.19 (0.86-1.66), respectively. Among hypertensive patients, the risk of CV events for ns-NSAIDs users was not significantly different versus nonusers (HR=1.21; 0.88-1.67). However, rofecoxib was associated with a significant 2-fold increase in CV risk versus nonusers of NSAIDs (HR=2.16; 1.51-3.09), whereas celecoxib was not (HR=1.18; 0.89-1.57). These data support the hypothesis that elevated CV risk is not a drug class effect of selective COX-2 inhibitors. That this effect was specific to hypertensive patients indicates that blood pressure destabilization is likely an important contributing mechanism.