Abstract
The objective of this study was to develop new risk classifications for conversion to Alzheimer's disease (AD) by comparing the relative reliability of classifiers in patients with mild cognitive impairment (MCI). The 397 MCI subjects and all baseline data, including characteristics, neuropsychological tests, cerebrospinal fluid biomarkers and MRI findings in Alzheimer's Disease Neuroimaging Initiative (ADNI), were used for analysis by Cox proportional hazard regression, bootstrap sampling, and c-index. Multivariate Cox regression analysis revealed the following factors to be associated with increased risk of conversion from MCI to AD during the 53-month follow-up period: AVLT 30-minute delayed recall, AVLT trial 1, Boston naming, logical delayed recall, trail-making B, CDR-sob, ADAS13, the cortical thickness of the right inferior temporal lobe (st91ta), and the left hippocampus volume. The combinations of ADAS13 at a cutoff point of 15.67 with CDR-sob at 1.5 or with the cortical thickness of the right inferior temporal lobe at 2.56 mm3 produced high conversion rates of 92.7% (82.4%–100.0%) and 88.8% (77.3%–100.0%), respectively, at 48 months. The discriminative ability based on c-index for the proposed combination was 0.68. The sample size was estimated as 504 in the group with a combination of ADAS13 and CDR-sob whose conversion rate is highest. The combination of ADAS13 with CDR-sob at an optimal cutoff point has a high reliability in classifying the MCI patients into high- and low-risk conversion to AD and will be benefit for patients' assessment and potentially facilitate the clinical development of novel therapeutics.