There are currently no effective treatments for the prevention of cognitive decline [3] leaving the healthcare system vulnerable to the strain of a growing elderly population. This gap illustrates the need for improving our understanding of the determinants of cognitive decline so future interventions can be successful in alleviating the burden of Alzheimer?s disease and related dementias (ADRD). Statistical challenges in aging research, which are typically ignored, lower the power of studies and can lead to spurious and biased associations [2]. Therefore, there is an urgent need to improve current longitudinal modeling techniques to produce more accurate estimates and to help correctly identify the underlying mechanisms driving cognitive decline. One statistical challenge in modeling cognitive decline is informative missingness. Such missingness results from the adverse health outcomes and mortality that have been closely linked to declining cognitive function [6-9]. Ignoring the association between mortality and cognitive decline can lead to spurious associations or mask true associations [4], a phenomenon known as ?survival bias?. Further, cognitive trajectories contain many sources of variability that current methods fail to capture due to rigid model structures, resulting in significant loss of power to detect associations [2,12]. The goal of this proposed study is to develop novel statistical methods for longitudinal data that better account for survival bias and variability of cognitive trajectories in aging research. Joint models combine the traditional mixed effects model for longitudinal data with time-to-event data and assume missingness is informative. However, simulation studies have generated concerns regarding how robust joint models are in handling various causal mechanisms of survival bias. Therefore, aim 1 will extend the joint model to additionally estimate the missing data mechanism directly making the joint model more robust under various causal mechanisms of survival bias.
Aim 2 will focus on capturing more variability in cognitive trajectories by allowing the static random effects of the joint model to vary with time. The proposed study will develop novel statistical methods to address important limitations of longitudinal studies in ADRD research; leading to a better understanding of cognitive decline and improved identification of subjects at a higher risk of ADRD for clinical trials of preventive regimens. The interdisciplinary training environment, expert mentorship, and use of cutting edge methods will provide the applicant an opportunity to develop unmatched subject matter, technical, and methodological expertise for her career as a future independent researcher in aging studies.
Cognitive decline is associated with debilitating health outcomes placing high burden on the healthcare system. Further, the determinants of cognitive decline remain elusive due in part to the statistical challenges in aging research that are commonly ignored. This study employs novel statistical methods to address these important barriers to uncovering the mechanisms of cognitive decline; leading to better identification of subjects at risk of dementia and more targeted clinical trials.
