Abstract
The goal of this study was to identify the optimal combination of magnetic resonance imaging (MRI), [18F]-fluorodeoxyglucose positron emission tomography (FDG-PET), and cerebrospinal fluid (CSF) biomarkers to predict conversion from amnestic mild cognitive impairment (aMCI) to Alzheimer's disease (AD) dementia within two years, for enriching clinical trial populations. Data from 63 subjects in the Alzheimer's Disease Neuroimaging Initiative aMCI cohort who had MRI and FDG-PET imaging along with CSF data at baseline and at least two years clinical follow-up were used. A Bayesian classification method was used to determine which combination of 31 variables (MRI, FDG-PET, CSF measurements, apolipoprotein E (ApoE) genotype, and cognitive scores) provided the most accurate prediction of aMCI to AD conversion. The cost and time trade-offs for the use of these biomarkers as inclusion criteria in clinical trials were evaluated. Using the combination of all biomarkers, ApoE genotype, and cognitive scores, we achieved an accuracy of 81% in predicting aMCI to AD conversion. With only ApoE genotype and cognitive scores, the prediction accuracy decreased to 62%. By comparing individual modalities, we found that MRI measures had the best predictive power (accuracy = 78%), followed by ApoE, FDG-PET, CSF, and the Alzheimer's disease assessment scale-cognitive subscale. The combination of biomarkers from different modalities, measuring complementary aspects of AD pathology, provided the most accurate prediction of aMCI to AD conversion within two years. This was predominantly driven by MRI measures, which emerged as the single most powerful modality. Overall, the combination of MRI, ApoE, and cognitive scores provided the best trade-off between cost and time compared with other biomarker combinations for patient recruitment in clinical trial.