Modeling patient‐specific tau spreading patterns in Alzheimer's disease: Towards precision medicine: Person‐centered prediction of tau spreading in Alzheimer's disease and related disorders. (7th December 2020)

• Record Type: Journal Article
• Authors: Franzmeier, Nicolai; Frontzkowski, Lukas; Rubinski, Anna; Neitzel, Julia; Smith, Ruben; Strandberg, Olof; Ossenkoppele, Rik; Hansson, Oskar; Ewers, Michael
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 5
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.040587
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: In Alzheimer's disease (AD), tau pathology spreads from the temporal lobe throughout the brain, ensuing cognitive decline. PET‐assessed tau‐spreading patterns are, however, heterogeneous across patients, posing challenges for assessing patient‐specific or group‐level longitudinal tau‐changes. Yet, this may become critical for disease management and evaluating tau‐targeting treatments. Using tau‐PET and resting‐state fMRI, we reported previously that tau spreads preferentially across functionally connected regions, supporting a trans‐neuronal tau‐spreading hypothesis (Franzmeier et al., Brain, 2019;NatComms, 2020). Here, we propose a novel connectivity‐based model to predict individual tau‐spreading patterns, which improves quantifying patient‐specific longitudinal tau‐changes over conventional staging approaches. Methods: We included two samples with cross‐sectional AV1451‐tau‐PET of amyloid‐negative controls (Aβ‐;ADNI/BioFINDER, n=231/16) and AD spectrum patients (Aβ+;ADNI/BioFINDER, n=213/41), plus longitudinal tau‐PET in Aβ+ (ADNI/BioFINDER, n=83/41, ∼1.3‐2yrs follow‐up). To determine tau‐abnormality, we transformed tau‐PET SUVRs of 200 regions of interest (ROIs, Fig.1A) to tau‐positivity probabilities using two‐component gaussian mixture models (Fig.1B&C). Multi‐leg connectivity‐based distance between 200‐ROIs was assessed on resting‐state fMRI of 1000 human connectome project participants (Fig.1D). In Aβ+, we rank‐ordered cross‐sectional tau‐abnormality probabilities across subjects and ROIs to determine hierarchical tau‐abnormality sequences (Fig.2A‐D). For "epicenters" with earliest tau‐abnormality (top 10% of rank‐ordered ROIs), connectivity‐based distance between epicenters and remaining ROIs (Fig.2E&F) was tested as a predictor of tau‐abnormality sequences. For subject‐level prediction of tau‐spreading, subject‐specific tau‐epicenters were defined as 10% ROIs with highest baseline tau‐abnormality and rates of longitudinal tau‐changes were determined in connectivity‐based distance to these epicenters. Results: Connectivity‐based distance of tau epicenters predicted cross‐sectionally‐assessed tau‐positivity sequences in Aβ+ (ADNI: R 2 =0.52, p<0.0001; BioFINDER: R 2 =0.50, p<0.0001, Fig.2G&H). For subject‐specific tau‐spreading, we found strongest tau‐changes in those 25% ROIs (i.e. first Quartile, Q1‐ROI in Fig.3A) in closest connectivity‐based proximity to subject‐specific epicenters (Fig.3B&C). Tau‐changes in closest proximity to the epicenters were higher than in Braak‐stage‐specific or whole‐grey‐matter ROIs (Fig.3D&E). Sample size estimation for treatments targeting tau‐accumulation showed that using subject‐tailored ROIs in closest connectivity‐based proximity to tau epicenters for assessing tau‐changes decreases sample sizes by ∼40% when compared to using whole‐brain or Braak‐stage‐specific ROIs. Conclusions: Connectivity‐based distance of tau epicenters predicts subject‐level tau accumulation patterns, which improves detecting longitudinal tau‐changes.