Prognosis of conversion of mild cognitive impairment to Alzheimer's dementia by voxel-wise Cox regression of FDG PET scans

Aim: The value of FDG PET for the prognosis of conversion from mild cognitive impairment (MCI) to Alzheimer’s dementia (AD) is controversial. In this work, the identification of cerebral metabolic patterns with significant prognostic value for conversion of MCI patients is investigated with voxel-based Cox regression, which in contrast to common categorical comparisons also utilizes time information. Methods FDG PET data of 544 MCI patients from the ADNI database were randomly split into two equally-sized datasets. Within a median follow-up duration of 47 months (95% CI: 46-84) 181 patients developed AD. In dataset A, voxel-wise Cox regressions were used to find regions associated with MCI to AD conversion and were compared to regions identified with an ANCOVA (as implemented in SPM 8) between converters and non-converters (both methods were adjusted for APOE genotype, MMSE score, age and sex). In dataset B, normalized FDG uptake within the clusters from voxel-wise Cox- and ANCOVA analyses (Cox- and ANCOVA-ROI, respectively) and clinical variables APOE status and MMSE score were tested in different Cox models (adjusted for age and sex) including: (1) only clinical variables, (2) clinical variables plus FDG uptake from ANCOVA-ROI, (3) clinical variables plus FDG uptake from Cox-ROI, (4) only normalized FDG uptake from ANCOVA-ROI, (5) only normalized FDG uptake from Cox-ROI. Results Conversion-related regions with relative hypometabolism comprise parts of the temporo-parietal and posterior cingulate cortex/precuneus (p<0.01, FDR corrected) for voxel-wise ANCOVA, plus partially the frontal lobe for voxel-wise Cox regression. The clinical-only model (1) significantly predicted conversion to AD (Wald Test (WT): p<0.001; Harrell’s c = 0.7) and was significantly improved by adding imaging information in model (2) (WT: p<0.001, c = 0.74, LR-test (1) to (2): p<0.001) and model (3) (WT: p<0.001, c = 0.74, LR-test (1) to (3): p<0.001). The models based on normalized FDG uptake from Cox-ROI-only (4)- and ANCOVA-ROI-only (5) also significantly predicted conversion (c=0.72 and c=0.76, respectively; both WT: p<0.001). Conclusions Prediction of conversion to AD in this dataset was improved by adding information from FDG PET - either based on regions identified by classical group contrasts or by voxel-wise Cox regressions.