Objective: Mild cognitive impairment (MCI)
in non-clinic samples often have high prevalence and unclear predictive validity for dementia risk. We used longitudinal data from a community-based cohort to evaluate whether MCI stability over two years reliably identified people who would develop dementia.

Participants and Methods: We evaluated data from the Adult Changes in Thought study. Participants were >65, not demented at baseline, and seen every two years. Cognition was assessed with a screening test and participants with low scores were evaluated to identify incident dementia cases. At baseline 1,721 participants were evaluated for MCI using Petersen et al. criteria. To examine MCI stability over the first two biennial visits, a convenience sub-sample of 708 was also evaluated. We evaluated conversion to dementia for the entire sample and the convenience sub-sample.

Results: At baseline, 738 of the 1,721 participants were identified with MCI (43%). Over a mean of 5.4 years of follow-up, 292 people (17%) developed dementia (200 with MCI, 68%; 92 without MCI, 32%). Baseline MCI was associated with increased dementia risk (Hazard ratio [HR] 3.1, 95% CI 2.5-4.0). For the convenience sub-sample, 254 participants had normal cognition at both time points (36%), 87 participants had MCI then normal cognition (12%), 121 participants had normal cognition then MCI (17%), and 246 participants had MCI at both time points (35%). Over a mean of 5.6 years of follow-up, 178 participants were diagnosed with dementia (25%). Compared to participants with normal cognition at both time points, participants with MCI then normal cognition had dementia HR 1.8 (95% CI 0.9-3.7), which was not significant. Participants with normal cognition then MCI had dementia HR 2.7 (95% CI 1.5-4.8) and people with MCI at both time points had dementia HR 7.1 (95% CI 4.4-11.3), both of which were significant.

Conclusions: Stability of MCI may be useful in identifying a small subset of people with very high dementia risk.