OBJECTIVES: Little is known about the association of length of exposure to COX-2 inhibitors (COX-2s) and the risk of cardiac events. This study determines the impact of extended exposure to COX-2s among high risk Medicaid patients, and is based on a previous propensity adjusted model that showed no added risk of cardiac events in COX-2 versus NSAID users, in this Medicaid population. METHODS: Selecting COX-2 users alone, we analyzed all medical and prescription claims of all continuously enrolled Medicaid patients, with at least one prescription for a COX-2 between January 1, 2000 and January 1, 2003, and no such prescriptions in the first six months. We used both direct adjustment and propensity score methods, and assessed length of exposure to COX-2s as a risk factor for post-use cardiac events, defining risk as a categorical variable (<30, 30-59, 60-89, 90-119 and >120 days), then as a continuous variable (divided by 30). The models are adjusted for age, gender, race and location (urban/suburban/rural), and clinical risk factors. RESULTS: A total of 1784 patient used COX-2s, 25% for less than 30 days. From the categorical analysis, there are significant increases in the likelihood of a post-use cardiac event given increased use compared to less than 30 days of cumulative exposure. From the analysis of exposure to COX-2s as a continuous variable, in the propensity-adjusted model, each 30-day increase in exposure corresponds to a concomitant but non-significant 2% increase in risk of cardiac events. For the direct adjusted model, there is a concomitant 5.5%, significant increase in post-use cardiac events. CONCLUSIONS: Among Medicaid COX-2 users, the risk of cardiac events is associated with longer exposure to COX-2s only when exposure is categorized in 30-days increments, but not when used as a continuous variable, suggesting a nonlinear relationship between exposure and events.