Some statistical issues on progression of Alzheimer's disease
Abstract
Alzheimer’s disease (AD) is a major public health problem among elderly adults. According to the 2015 Alzheimer’s disease facts and figures, about one-in-nine Americans aged over 65 years has AD, but only 45% of them are diagnosed with the disease. In addition, the disease is very costly to the society and individuals. It requires a high level of commitment from the caregivers whose emotional stress, health, and financial stability are adversely affected. In this research, two statistical methods will be proposed to address the research issues raised in Alzheimer’s disease studies. One is to predict the pre-clinic symptom duration of an AD patient based on the longitudinal observations of neuropsychological variables and the other is to investigate the natural history of AD disease progression in the framework of the continuous time Markov chain (CTMC) with censors. The longitudinal mixed effect model will be adopted to predict the pre-clinic symptom duration using EM algorithm in which complicated algebraic derivations are required. To examine the natural history of AD disease progression and its determinants, explicit likelihood function of the CTMC transition probability with or without a censor will be derived and the model estimates are to be obtained using a frequentist approach. The significance of this research can be viewed in two aspects. In AD research, the pre-clinic symptom duration is an important measure that is unobservable and even unlikely for a physician to estimate accurately. This measure will provide important information for understanding rate of AD progression and hence the natural history of the disease. Ultimately, it will lead to a more accurate estimation of, the financial and health facility needs of a patient. The stage progression of AD is an essential dynamic process that can predict the disease duration at each stage crucial for developing intervention and health service strategy. From statistical methodology viewpoint, to predict the pre-clinic symptom duration requires a novel approach in development of a prediction algorithm under a longitudinal mixed effect model. For studying AD stage progression, the estimation of CTMC model parameters demands a novel estimation procedure when censorship occurs because the current literature only allows estimation without censorship.
Subject Area
Biostatistics
Recommended Citation
Peng, Ho-Lan, "Some statistical issues on progression of Alzheimer's disease" (2016). Texas Medical Center Dissertations (via ProQuest). AAI10182183.
https://digitalcommons.library.tmc.edu/dissertations/AAI10182183