Identifying Progression Pattern of Alzheimer’s Disease Using Longitudinal Clinical and Neuroimaging Biomarkers

Alzheimer’s disease is one of the most common neurodegenerative disorders worldwide. Therapeutics to treat or prevent Alzheimer’s disease progression has not been successful due to the possible heterogeneity of progression and response to treatment. The goal of this study was to determine whether there is a common pattern in Alzheimer’s Disease progression that can be used to guide the clinical trials and treatment. Longitudinal clinical and neuroimaging biomarkers data from Open Access Series of Imaging Studies were analyzed using an approach based on Mixture of Gaussian Processes Model to identify clusters that has similar pattern in progression. The approach enables non-parametric analyses with no assumptions on linearity and number of clusters. It was demonstrated that normalizing the patient’s own onset value yields better accuracy in stratifying patients by progression rate because it removes the systematic variations and imbalance induced by environmental and genetic differences among the patients. The clustering result showed that the Alzheimer’s Disease progression is nonlinear, suggesting that the initial progression rate should not be used to evaluate the future progression trajectory. It was demonstrated that the progression is heterogeneous among the patients within each biomarker and among biomarkers of each patient, indicating the importance of personalized treatment for individual patient. Additional analyses with more data to ensure the robustness of the clustering will provide neurologists a powerful tool to estimate the progression trajectory of each patient, with which a personalized treatment becomes possible.