Predicting the reversion from mild cognitive impairment to normal cognition based on magnetic resonance imaging, clinical, and neuropsychological examinations.

Reversion from mild cognitive impairment (MCI) to normal cognition (NC) is not uncommon and indicates a better cognitive trajectory. This study aims to identify predictors of MCI reversion and develop a predicting model. A total of 391 MCI subjects (mean age = 74.3 years, female = 61 %) who had baseline data of magnetic resonance imaging, clinical, and neuropsychological measurements were followed for two years. Multivariate logistic analyses were used to identify the predictors of MCI reversion after adjusting for age and sex. A stepwise backward logistic regression model was used to construct a predictive nomogram for MCI reversion. The nomogram was validated by internal bootstrapping and in an independent cohort. In the training cohort, the 2-year reversion rate was 19.95 %. Predictors associated with reversion to NC were higher education level (p = 0.004), absence of APOE 4 allele (p = 0.001), larger brain volume (p < 0.005), better neuropsychological measurements performance (p < 0.001), higher glomerular filtration rate (p = 0.035), and lower mean arterial pressure (p = 0.060). The nomogram incorporating five predictors (education, hippocampus volume, the Alzheimer's Disease Assessment Scale-Cognitive score, the Rey Auditory Verbal Learning Test-immediate score, and mean arterial pressure) achieved good C-indexes of 0.892 (95 % confidence interval [CI], 0.859–0.926) and 0.806 (95 % CI, 0.709–0.902) for the training and validation cohort. Observational duration is relatively short; The predicting model warrant further validation in larger samples. This prediction model could facilitate risk stratification and early management for the MCI population. • The 2-year reversion rate from MCI to NC was 19.95%. • Predictors of reversion included education, APOE4 allele, brain volume, neuropsychological tests performance, eGFR, and MAP. • The predicting model achieved good performance for the training and validation cohort. [ABSTRACT FROM AUTHOR]