A recent clinical trial in patients with Mild Cognitive Impairment (MCI) found an increased rate of possible or probable Alzheimer's disease (AD) diagnoses in patients assigned to rofecoxib compared to placebo. This unexpected finding was difficult to interpret due to methodological issues and a lack of confirmation on secondary endpoints, as well as a lack of confirmation in trials in related populations. We performed additional post hoc analyses to explore explanations for the finding based on possible neuropathological, cardiovascular/cerebrovascular, or cognitive effects of rofecoxib. 1) Neuropathological hypothesis: Of the 189 incident cases of possible or probable AD, 154 were probable AD. In probable AD patients, the treatment hazard ratio was reduced compared to the primary analysis -- a concordant finding would have strengthened a conclusion that rofecoxib accelerated the underlying neuropathology of AD. The treatment hazard ratio was increased in the remaining 35 patients with less certain diagnoses, but there was no single predominant reason for the reduced certainty of diagnosis. 2) Cardiovascular hypothesis: Neither cardiovascular risk status nor mean arterial blood pressure had an overall effect on AD diagnosis or modified the treatment difference. 3) Cognitive side-effects hypothesis: The percentages of patients with non-specific NSAID-type central nervous system adverse events were similar between the treatment groups. In summary, the present analyses are limited by their post hoc nature but provided little support for any of the possible explanations explored. The significance of the observation that rofecoxib increased the rate of conversion from MCI to AD remains uncertain.