Biomarker-Based Prediction of Progression to Dementia: F-18 FDG-PET in Amnestic MCI

Background: Metabolic patterns on brain F-18 fluorodeoxyglucose (FDG) positron emission tomography (PET) can predict the decline in amnestic mild cognitive impairment (aMCI) to Alzheimer's disease dementia (AD) or other dementias. Objective: This study was undertaken to evaluate the diagnostic accuracy of baseline F-18 FDG-PET in aMCI for predicting conversion to AD or other dementias on follow-up. Patients and Methods: A total of 87 patients with aMCI were enrolled in the study. Each patient underwent a detailed clinical and neuropsychological examination and FDG-PET at baseline. Each PET scan was visually classified based on predefined dementia patterns. Automated analysis of FDG PET was performed using Cortex ID (GE Healthcare). The mean follow-up duration was 30.4 +/- 9.3 months (range: 18-48 months). Diagnosis of dementia at follow-up (obtained using clinical diagnostic criteria) constituted the reference standard, and all the included aMCI patients were divided into two groups: the aMCI converters (MCI-C) and MCI nonconverters (MCI-NC). Diagnostic accuracy of FDG PET was calculated using this reference standard. Results: There were 23 MCI-C and 64 MCI-NC. Of the 23 MCI-C, 19 were diagnosed as probable AD, 1 as frontotemporal demetia (FTD), and 3 as vascular dementia (VD). Of the 64 MCI-NC, 9 had subjective improvement in cognition, and 55 remained stable. The conversion rate for all types of dementia in our series was 26.4% (23/87) and for Alzheimer's type dementia was 21.8% (19/87). The of PET-based visual interpretation was 91.9%. Sensitivity, specificity, positive predictive value, and negative predictive value for FDG-PET-based prediction of dementia conversion were 86.9% [confidence interval (CI) 66.4%-97.2%)], 93.7% (CI 84.7%-98.2%), 83.3% (CI 65.6%-92.9%), and 95.2% (CI 87.4%-98.9%), respectively. Kappa for agreement between visual and Cortex ID was 0.94 indicating excellent agreement. In the three aMCI patients progressing to VD, no specific abnormality in metabolic pattern was noted; however, there was marked cortical atrophy on computed tomography. Conclusion: FDG-PET-based visual and cortex ID classification has a good accuracy in predicting progression to dementia including AD in the prodromal aMCI phase. Absence of typical metabolic patterns on FDG-PET can play an important exclusionary role for progression to dementia. Vascular cognitive impairment with cerebral atrophy needs further studies to confirm and uncover potential mechanisms.