Integrative analysis for patient-centered outcomes and time-to-event data in Alzheimers disease

Objectives: To develop and validate analytical models that integrate NIAGADS-hosted genetic data with NACC longitudinal clinical and biomarker profiles to characterize the progression of the continuum of Alzheimer’s disease from preclinical stages. The goals are to reconstruct individualized biomarker dynamics under sparse and irregular sampling, relate these dynamics to the timing of clinical hallmarks, and identify genetic variants associated with quantitative trajectory-based phenotypes.

Study design: The project focuses on new statistical and ML model development, as well as on retrospective observational analyses of de-identified NACC participants with linked genotype data and repeated assessments. Methods will be developed and internally validated using the observed NACC follow-up schedules and available longitudinal history.

Analysis plan: We will build statistical and machine learning models tailored to irregular longitudinal data to estimate person-specific trajectories from available biomarkers, including multi-modal imaging biomarkers, and from clinical and cognitive measures. From fitted trajectories, we will derive composite trajectory traits that summarize individualized marker dynamics and their relationship to event timing, including onset and progression to MCI and Alzheimer’s disease when available. We will test associations between genetic variants and these quantitative phenotypes using appropriate covariate adjustment for sex and genetic ancestry components, apply multiple testing control, and assess stability with resampling-based inference where appropriate. We will develop dynamic prediction models that update risk over time using subject-specific landmark times defined by each participant’s assessment schedule, and evaluate discrimination and calibration by cross-validation. We will also develop time-varying composite measures and examine how they modify predicted risk trajectories.

We will develop new analytical methods to learn how Alzheimer’s disease develops before symptoms appear. Using de-identified NACC health and biomarker data linked to NIAGADS genetics, we will build models that reconstruct how key biomarkers change over time even when measurements are infrequent, irregular, or missing. These models will produce individualized trajectory scores and other quantitative summaries, and we will study how genetic differences relate to those early patterns. The long-term aim is to improve understanding of early disease biology and support earlier identification of vulnerability.