Enhancing Outcomes and Efficiency in the AMARANTH Alzheimer’s Disease Trial Through AI-Driven Patient Stratification

The Personalized Prognostic Model (PPM) utilizes Generalized Metric Learning Vector Quantization (GMLVQ) to distinguish between clinically stable and declining individuals in early Alzheimer’s stages. Trained on baseline data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) with an accuracy of 91.1%, PPM highlights the key features affecting predictions, primarily β-Amyloid burden. Through the AMARANTH trial, the PPM identified slow vs. rapid progressive patients using multimodal data, which enhanced treatment outcome measurements. Results revealed that lanabecestat treatment significantly reduced β-Amyloid levels and cognitive decline (CDR-SOB scores) in the slow progressive group. Importantly, PPM-guided stratification markedly decreased the required sample size for detecting treatment effects, emphasizing its clinical relevance and potential application in future trials for Alzheimer’s disease management. This innovative approach not only aids in fine-grained classification but also improves treatment assessment efficacy at critical neurodegenerative stages, setting a new standard for dementia research methodology.

Source link