A Statistical Framework for Assessing the Relationship between Biomarkers and Clinical Endpoints in Alzheimer’s Disease

Abstract

Changes in biomarker levels of Alzheimer’s disease (AD) reflect underlying pathophysiological changes in the brain and can provide evidence of direct and downstream treatment effects linked to disease modification. Recent results from clinical trials of anti–amyloid β (Aβ) treatments have raised the question of how to best characterize the relationship between AD biomarkers and clinical endpoints. Consensus methodology for assessing such relationships is lacking, leading to inconsistent evaluation and reporting. In this review, we provide a statistical framework for reporting treatment effects on early and late accelerating AD biomarkers and assessing their relationship with clinical endpoints at the subject and group levels. Amyloid positron emission tomography (PET), plasma p-tau, and tau PET follow specific trajectories during AD and are used as exemplar cases to contrast biomarkers with early and late progression. Subject-level correlation was assessed using change from baseline in biomarkers versus change from baseline in clinical endpoints, and interpretation of the correlation is dependent on the biomarker and disease stage. Group-level correlation was assessed using the placebo-adjusted treatment effects on biomarkers versus those on clinical endpoints in each trial. This correlation leverages the fundamental advantages of randomized placebo-controlled trials and assesses the predictivity of a treatment effect on a biomarker or clinical benefit. Harmonization in the assessment of treatment effects on biomarkers and their relationship to clinical endpoints will provide a wealth of comparable data across clinical trials and may yield new insights for the treatment of AD.