OBJECTIVE: To better define the role of nonsteroidal antiinflammatory drugs (NSAID) in cognitive decline of the elderly. METHODS: Population based inception cohort of the rural elderly. NSAID user status was characterized as high dose, low/medium dose, or nonuser at 2 successive in-person interviews 3 years apart (F3 and F6). Respondents were from the Iowa 65+ Rural Health Study, one of the 4 cohorts of the National Institute on Aging's Established Populations for Epidemiologic Studies of the Elderly. Memory decline was assessed by a change in immediate word recall between F3 and F6. Multivariable logistic regression models were created to determine important predictors of the F3 to F6 recall memory decline in groups with poor, average, and good word recall at the F3 interview. Specific NSAID were compared to assess which drugs, if any, were associated with memory decline. RESULTS: The 2 factors most strongly associated with a significant immediate word recall decline between F3 and F6 among individuals in the average F3 word recall group were limitation in functional status [odds ratio (OR) = 2.31, 95% Confidence Interval (CI) (1.51, 3.52)] and high dose NSAID use [OR = 2.04, 95% CI (1.07, 3.89)]. No single NSAID agent was significantly more strongly associated with word recall decline than the others. However, in exploratory analyses use of high dose NSAID of the proprionic acid family neared significance for recall decline [OR = 3.17, 95% CI (0.92, 10.9). CONCLUSION: In elderly respondents with average baseline recall memory, high dose NSAID were a significant risk factor for longitudinal memory decline in this community based cohort. Although a large scale clinical trial is needed to definitely address this issue, we provide further evidence that NSAID play a role in cognitive dysfunction in the elderly.