Background: Alzheimer’s disease is characterized by multiple pathological processes involving amyloid-β (Aβ) and tau. It remains uncertain which marker most reliably reflects the degree of hippocampal atrophy. We aimed to identify which biomarker best predicts hippocampal atrophy.
Methods: We conducted a retrospective cross-sectional study using Alzheimer’s Disease Neuroimaging Initiative data from 576 participants who underwent MRI, amyloid PET/CT, and CSF testing. Two independent neuroradiologists rated medial temporal atrophy (MTA) in each hippocampus using the Scheltens MTA scale. Aβ Centiloids, meta-temporal tau-PET SUVR, CSF total tau (t-tau), phosphorylated tau (p-tau), Aβ42, and Aβ40 were the biomarkers evaluated. Multivariable logistic regression models adjusted for age, sex, and APOE ε4 carrier status were used to assess associations with hippocampal atrophy, with MTA ≥ 2 as the primary outcome. Discrimination was assessed using ROC curves and AUC, based on covariate-adjusted models.
Results: Participants with MTA ≥ 2 were older (74.5 ± 6.9 years) and had worse cognitive status, with higher prevalence of mild cognitive impairment (MCI) (99; 40.4%) and Alzheimer’s disease dementia (61; 24.9%). They also showed higher amyloid burden (Aβ Centiloids: 45.3 ± 50.3) and higher tau-PET uptake (meta-temporal tau-PET SUVR: 1.4 ± 0.3), along with lower CSF Aβ40 (16599.1 ± 5180.5) and Aβ42 (939.7 ± 510.3) (Table 1). Meta-temporal tau-PET SUVR (AUC 0.784, 95% CI 0.745-0.823, n=523) showed the best discrimination for MTA ≥ 2 among all biomarkers evaluated. Aβ42 (AUC 0.759, 95% CI 0.709-0.809, n=360) demonstrated the highest discrimination among CSF biomarkers (Figure 1). Meta-temporal tau-PET SUVR (OR per 1 SD increase = 1.919, 95% CI 1.519-2.425, p <0.0001) showed the strongest association with hippocampal atrophy in covariate-adjusted logistic regression, while Aβ42 (OR per 1 SD increase = 0.495, 95% CI 0.361-0.677, p <0.0001) showed the strongest association among CSF biomarkers (Figure 2).
Conclusion: In this cohort, meta-temporal tau-PET SUVR was the single biomarker most associated and most discriminative for hippocampal atrophy. Aβ42 performed best for identifying participants with hippocampal atrophy among CSF biomarkers. PET-based markers, particularly tau-PET, appeared more informative than CSF biomarkers for estimating hippocampal atrophy severity, supporting tau-PET as the main biomarker to differentiate hippocampal atrophy.