Inflammatory mechanisms have been implicated in Alzheimer's disease (AD) and might be mediated via the COX-2 enzyme. Previous studies with the selective COX-2 inhibitors, rofecoxib and celecoxib, have shown that they do not alter the progression of AD. We conducted a double-blind study to investigate whether rofecoxib could delay a diagnosis of AD in patients with mild cognitive impairment (MCI), a group with an expected annual AD diagnosis rate of 10-15%. MCI patients > or =65 years were randomized to rofecoxib 25 mg (N=725) or placebo (N=732) daily for up to 4 years. The primary end point was the percentage of patients with a clinical diagnosis of AD. The estimated annual AD diagnosis rate was lower than the anticipated 10-15%: 6.4% in the rofecoxib group vs 4.5% in the placebo group (rofecoxib : placebo hazard ratio=1.46 (95% CI: 1.09, 1.94), p=0.011). Analyses of secondary end points, including measures of cognition (eg the cognitive subscale of the AD Assessment Scale (ADAS-Cog)) and global function (eg the Clinical Dementia Rating (CDR)), did not demonstrate differences between treatment groups. There was also no consistent evidence that rofecoxib differed from placebo in post hoc analyses comparing ADAS-Cog and CDR-sum of boxes scores in overlapping subgroups of patients who had Mini Mental State Exam scores of 24-26 in the present MCI study and in a previous AD treatment study with a similar design. The results from this MCI study did not support the hypothesis that rofecoxib would delay a diagnosis of AD. In conjunction with the lack of effects observed in previous AD studies, the findings suggest that inhibition of COX-2 is not a useful therapeutic approach in AD.