Mapping Longitudinal Biomarker Relationships in Alzheimer’s Disease with Latent Time Joint Mixed Effects Modeling

Through the recent progress in development of Alzheimer’s Disease therapies it has become evident that fluid and imaging biomarkers hold the key to unlocking further advances in the clinic. Biomarker-driven trial designs hold the potential to improve statistical power by enabling the targeting of specific subpopulations of patients and the use of predictive endpoints. However, cross-sectional correlations between these biomarkers can be non-linear and trends hidden in longitudinal data may be obscured by variability in disease status at enrollment time. Latent Time Joint Mixed Effects Modeling (LTJM) aligns biomarker and clinical endpoint data across a common disease timescale – time from amyloid positivity (TFAP). We present a LTJM disease chronicity analysis integrating data across fluid, imaging, and clinical markers, and estimate the longitudinal sequencing and variability in biomarker progression. We also demonstrate the utility of the model in making longitudinal (up to 7 years post-baseline) predictions of subject-level progression profile conditioned on baseline data. This capability supports the refinement of clinical trial parameters (enrollment criteria, duration, endpoint, etc.) through clinical trial simulation. Additionally, it provides a platform for high-throughput screening of emerging biomarkers which may hold clinical and pharmacological relevance, improving the prospects of precision medicine approaches in Alzheimer’s Disease.