Temporoparietal cortical thickness outperforms hippocampal volume as a biomarker for atypical and early-onset Alzheimer's disease

Abstract

Hippocampal volume (HV) and temporoparietal cortical (TPC) integrity have been proposed as MRI biomarkers of neuronal injury in operationalizing new diagnostic criteria for Alzheimer's disease (AD). HV performs well in distinguishing late age-of-onset AD patients (LO-AD) from normal controls (NC), however it may be less sensitive in patients with early age-of-onset (EO-AD) or non-amnestic presentations. We compared the sensitivity of HV and TPC in patients with EO-AD, LO-AD, and two non-amnestic variants: logopenic-variant primary progressive aphasia (lvPPA) and posterior cortical atrophy (PCA). We identified 45 patients seen at UCSF who met NIA-AA criteria for probable AD dementia (mean age=64.4±9.0, MMSE=20.4±7.0), underwent T1-MPRAGE MRI on 1.5T (N=24) or 3T (N=21) scanners, and were PIB-PET positive. Patients included 12 EO-AD (age at onset ≤ 65, mean=59.3±5.9), 7 LO-AD (mean age-of-onset=79.6±5.0), 13 lvPPA, and 13 PCA. HV (adjusted for intracranial volume) and mean TPC thickness were measured using FreeSurfer 5.1. Thresholds for classifying HV and TPC in the AD range were derived by applying receiver operator characteristic (ROC) analyses to independent samples of ADNI NC (n=166, age at study=75.3±6.6) and AD patients (n=97, age=76.3±7.3, MMSE=22.9±2.4). Thresholds were derived separately for ADNI 1.5 T (N=89) and 3T (N=174) data, and applied separately to UCSF 1.5T and 3T data to compare the sensitivity of HV and TPC across AD variants. HV and TPC performed well in discriminating ADNI NC and AD subjects per MRI field strength (FIGURE). For 1.5T, HV had an area under the ROC curve (AUC)=0.929, sensitivity=90%, and specificity=90%, and for 3T, AUC=0.936, sensitivity=85%, specificity=86%. For TPC: AUC=0.813, sensitivity=82%, and specificity=69% for 1.5T; AUC=0.839, sensitivity=72%, specificity=87% for 3T. Overall, both HV and TPC correctly classified 6/7 UCSF LO-AD subjects (sensitivity=86%). However, HV was poorly sensitive in early-onset and atypical AD variants (overall sensitivity=32%: 6/12 EO-AD, 2/13 lvPPA, 4/13 PCA, FIGURE), while TPC was significantly more sensitive in these patients (overall sensitivity=82%: 10/12 EO-AD, 11/13 lvPPA, 10/13 PCA; χ 2(2)=14.4, p<0.001). Figure. ROC-derived thresholds discriminating Alzheimer's disease patients from normal controls. Plots are separated by MR field strength and region of interest: (A) 1.5T- derived temporoparietal cortical (TPC) thickness; (B) 1.5T-derived hippocampal to intracranial volume ratio (HC/ICV); (C) 3T-derived TPC thickness; and (D) 3T-derived HC/ICV. ADNI subjects were either normal controls (ADNINC) or Alzheimer's disease patients (ADNI AD). UCSF patients were diagnosed with either late-onset AD (LO-AD), early onset AD (EO-AD), logopenic-variant primary progressive aphasia (1vPPA), or posterior cortical atrophy (PCA).