PURPOSE: To evaluate the relationship of nonsteroidal antiinflammatory drug (NSAID) use to level of cognitive function in community-dwelling elderly persons. METHODS: The prospective cohort study included 2765 nonproxy subjects from the Duke University Established Populations for Epidemiologic Studies of the Elderly who were cognitively intact at baseline (1986-1987) and alive at follow-up three year later. Cognitive function was assessed by the Short Portable Mental Status Questionnaire (i.e., intact vs. impaired and change in score) and by the individual domains of the Orientation-Memory-Concentration Test (i.e., number of errors). NSAID use, determined from in-home interviews, was coded for chronicity, dose, frequency of use, and prescription status. RESULTS: After controlling for demographic factors as well as health status and behavior, continuous, regularly-scheduled, prescription use of NSAID was associated with preservation of one aspect of cognitive functioning: concentration (beta coefficient, 0.29; 95% confidence interval [CI] -0.54 to -0.04, indicating fewer errors). However, no consistent dose-response relationship was found. Current and prior NSAID use was unrelated to level of cognitive functioning across all five measures; among current users, those taking moderate or high doses (beta coefficient, 0.41; 95% CI, 0.08 to 0.74) made more errors on the memory test compared with those taking low doses (beta coefficient 0.03; 95% CI, -.85 to 0.91). CONCLUSIONS: These results suggest no substantial or consistent protective effect of prescription NSAID use on cognitive function in community-dwelling elderly. However, recent use at higher doses may be associated with memory deterioration in this population.