Tau‐Neurodegeneration mismatch reveals vulnerability and resilience in Alzheimer’s continuum and Non‐Alzheimer’s pathophysiology

Abstract

AbstractBackgroundIn previous work (Das et al., 2021), we exploited the non‐specific nature of neurodegeneration by investigating the amount and spatial pattern of deviation from the level of neurodegeneration (N) expected for a given level of tau (T). Using a data‐driven clustering approach based on T‐N mismatch, we found phenotypes that had greater N than expected for T (“vulnerable”) and phenotypes with less N than expected for T (“resilient”) in specific patterns. Here we extend this work to include amyloid positive (A+) and negative (A‐) participants. The logic of inclusion of A‐ individuals is that areas of mismatch in A+ individuals is likely driven by factors outside of AD pathology, which can be present in the absence of AD pathology.MethodWe employed cortical thickness to represent N and 18F‐flortaucipir tracer uptake as T for 104 bilateral cortical regions of interest (ROI) in 343 symptomatic individuals including both A+ and A‐ from ADNI. We established T‐N robust regression and partitioned participants based on thresholded residuals. Longitudinal trajectories of cognitive scores were assessed with linear mixed‐effects model effect models using CDRSB as the dependent variable.ResultWe replicated prior work and obtained clusters with distinct T‐N mismatch patterns. We defined the largest group as “canonical”, as it displayed typical T‐N relationships (N∼T, low residuals). The limbic vulnerable cluster displayed N>T (negative residuals) in limbic regions, while diffuse vulnerable group displayed N>T throughout the cortex. Alternatively, the resilient (N<T, positive residuals) clusters were classified as limbic resilient or prefrontal resilient based on the pattern of this relationship. Note the highly similar regional patterns of the A+ and A‐ subgroups, supporting our hypothesis that this residual reflects non‐AD related phenomenology which could be present irrespective of amyloid status. Phenotypes differed in longitudinal cognitive trajectories. The vulnerable clusters declined significantly faster than the canonical cluster regardless of amyloid status likely attributable to the presence of non‐AD pathologies. The resilient groups tended to progress less than the canonical group.ConclusionOur findings suggest T‐N mismatch depicts vulnerability and resilience likely attributable to non‐AD factors. This may provide an approach for detecting co‐pathologies in AD continuum and benefiting therapeutic management.