Abstract
282
Aim: Amyloid PET is an established biomarker for the prognosis of conversion from mild cognitive impairment (MCI) to Alzheimer’s dementia (AD), but the optimal evaluation method is still under debate. In this work, three methods for the evaluation of amyloid PET are compared amongst each other and with the amyloid-β ratio Aβ42/Aβ40 in cerebrospinal fluid (CSF).
Methods: AV45 PET data of 319 MCI patients from the ADNI database were randomly split into two datasets. In the derivation dataset A (N = 160), voxel-wise Cox regressions were used to find regions associated with MCI to AD conversion (p < 0.01 FEW corrected; all voxels with a hazard ratio (HR) > 80% of maximum HR were combined into a Cox-VOI). In addition, an AD conversion-related pattern on amyloid PET (Aβ-ADCRP) was calculated via principle component analysis in a previous work (1). In validation dataset B (N=129), four Cox models were used to test the value of the following predictors: (I) mean normalised AV45 uptake (SUVR; reference: cerebellar cortex) in AD-typical brain regions (anatomical VOI), (II) HR-weighted, normalised SUVR in Cox-VOI, (III) pattern expression score (PES) of the Aβ-ADCRP and (IV) Aβ42/Aβ40 from CSF.
Results: Within a median follow-up duration of 47 months (95% CI: 35-54 months) 23% of the patients developed AD. In the derivation dataset A, conversion-related regions identified with voxel-wise Cox regressions (Cox-VOI) comprise parts of the frontal and temporal cortex, the precuneus and the striatum. All four Cox models significantly predicted conversion within the validation dataset B (Wald-test p < 0.001). Prediction accuracies (Harrell’s c) were: model (I), c = 0.73 (HR = 2.2 per unit z-score increase (95% C.I. 1.5 - 3.2)); model (II), c = 0.71 (HR = 1.8 per unit z-score increase (95% C.I. 1.3 - 2.6)); model (III), c = 0.77 (HR = 2.7 per unit z-score increase (95% C.I. 1.8 - 4.1)); and model (IV), c = 0.75 per unit z-score increase (HR = 3.9 (95% C.I. 2.1-7.2)).
Conclusions: For amyloid PET, all tested methods of image evaluation significantly predict conversion of MCI to AD. The PES of Aβ-ADCRP has the highest predictive value and is slightly superior to the amyloid-β-ratio from CSF. Literature: (1) Blazhenets G, Ma Y, Sörensen A, et al. Predictive Value of (18)F-Florbetapir and (18)F-FDG PET for Conversion from Mild Cognitive Impairment to Alzheimer Dementia. J Nucl Med. 2019. Figure 1: Surface projection of the hazard ratios from the voxel-wise Cox model in the training dataset. Hazard ratios above one correspond to one-unit increase in the normalized AV45 uptake (SUVR; reference: cerebellar cortex).