A Prediction Model for Alzheimer's Disease Severity Using Gut Microbiota and Neuroimaging

Abstract

Background: Alzheimer's disease (AD) is a progressive neurodegenerative disorder whose severity must be assessed accurately to guide clinical decisions. Conventional evaluation is often subjective, and single-domain biomarkers inadequately capture the complex pathophysiology of AD. The emerging "gut-brain axis" concept suggests that combining multidimensional indicators may improve prediction. This study developed and validated a model integrating gut microbiota profiles with neuroimaging features to predict AD severity. Methods: In this cross-sectional study, 340 AD patients were randomly assigned to training (n=238) or validation (n=102) sets at a 7:3 ratio. Demographic, gut microbiota, neuroimaging, and fluid biomarker data were collected. Severity was graded by the Clinical Dementia Rating (CDR) scale: early-stage AD was CDR 0-0.5, and moderate-to-severe AD was CDR ≥1.0. Predictors were screened by univariate analysis, Least Absolute Shrinkage and Selection Operator (LASSO) regression, and multivariate logistic regression. Logistic Regression, Random Forest, and Support Vector Machine models were built and evaluated by the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). Results: Seven variables differed significantly between groups (P<0.05): gut microbiota α-diversity, Faecalibacterium and Escherichia-Shigella relative abundance, plasma p-tau181 and NfL concentrations, hippocampal volume, and mean whole-brain cortical thickness. Multivariate logistic regression identified α-diversity, Faecalibacterium abundance, hippocampal volume, and cortical thickness as independent protective factors, and Escherichia-Shigella abundance, plasma p-tau181, and NfL as independent risk factors for severity. The Random Forest model showed moderate-to-good discrimination, with AUCs of 0.758 and 0.714 in the training and validation sets. Calibration curves showed acceptable agreement between predicted and observed outcomes, and DCA indicated favorable net clinical benefit across a 20%-60% risk threshold. Conclusion: A model integrating gut microbiota and neuroimaging features was developed and validated for AD severity assessment, showing good discrimination, calibration, and clinical utility. It offers an objective, reliable quantitative tool for severity classification and individualized intervention in AD.