Identifying individuals with non‐AD co‐pathologies: A precision medicine approach to clinical trials in sporadic AD. (December 2021)

• Record Type: Journal Article
• Authors: Tosun, Duygu; Ellis, Ryan; Weiner, Mike W.; Yardibi, Ozlem; Simen, Arthur; Schwarz, Adam J.
• Is Part Of: Alzheimer's & dementia. Volume 17(2021)Supplement 4
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12
• 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.053246
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: The presence of co‐pathologies such as TDP‐43, Lewy bodies (LB), and cerebral amyloid angiopathy (CAA) is common in older individuals with AD, even those pathologically confirmed to have amyloid‐β and tau pathology, and contributes to cognitive and functional decline. Treatments targeting just one pathology will thus have variable and reduced overall measurable efficacy compared with an assumed population with no co‐pathologies. Tools to identify individuals with non‐AD co‐pathologies would be enabling technologies for a precision medicine approach to clinical trials in sporadic AD. In the absence of biomarkers to measure these co‐pathologies directly, our aim was to develop AI models to identify autopsy‐confirmed structural MRI‐based signatures of non‐AD degenerative brain pathologies, to facilitate participant selection for AD trials. This builds on the recent development of models to impute amyloid and tau positivity from structural MRI scans and associated clinical data. Method: ADNI and NACC participants with ante‐mortem MRI and histopathologic assessments for scoring of neuritic plaques, staging of neurofibrillary tangles, LB, CAA, and TDP inclusions were included (Table 1). A multilabel classifier was trained on demographics, AD pathology, and MRI to jointly model positivity for TDP‐43, LB, and CAA. The leave‐one‐out validated model was applied to a separate ADNI cohort without neuropathologic assessments to assess the imputed prevalence of co‐pathologies at baseline and variance in cognitive decline and atrophy explained by baseline AD and non‐AD pathologies. Result: When optimized for 90% overall NPV (Fig. 1), the multi‐label classifier identified 16.0% of ADNI Dementia participants as TDP‐43+, 54.8% LB+, and 17.8% CAA+ (Fig. 2). While co‐pathology positivity was rare in cognitively unimpaired, LB and CAA positivity were common in MCI. Although dependent upon clinical disease stage, imputed probability of being TDP‐43+ explained a significant percentage of variance in clinical outcomes measures in cognitively impaired participants (Fig. 3). Conclusion: These initial results provide promising evidence that imputed co‐pathology burden via widely available imaging and clinical data can be used to impute the presence of non‐AD co‐pathologies and their contribution to cognitive decline in vivo . The primary limitation of this approach is the limited sample size in the autopsy cohorts for model development.