Predicting severity of cerebral amyloid angiopathy neuropathology: A modeling approach using NACC and ROSMAP data.

INTRODUCTION

Cerebral amyloid angiopathy (CAA) is associated with an increased risk of amyloid-related imaging abnormalities (ARIA) in patients with Alzheimer's disease using anti-amyloid beta (Aβ) monoclonal antibody drugs. Here, we developed a tool, CAA risk score (CAARS) to predict the severity of CAA neuropathology.

METHODS

The National Alzheimer's Coordinating Center (NACC) data were used to develop the CAARS, which was then externally validated using the Religious Orders Study and Memory and Aging Project (ROSMAP) data.

RESULTS

The CAARS-4 model achieved a mean area under the receiver-operating characteristic (ROC) curve (AUC-ROC) of 0.71 (95% confidence interval [CI]: 0.69-0.72) and a Harrell's generalized C-index of 0.69 (95% CI: 0.68-0.71) in the NACC cohort validation. It outperformed the baseline models, and the promising performance was validated on ROSMAP participants, demonstrating the robustness and generalizability of the model.

DISCUSSION: CAARS has the potential to predict CAA severity; however, its clinical utility should be evaluated in follow-up studies.

HIGHLIGHTS: Cerebral amyloid angiopathy (CAA) severity can only be confirmed postmortem. CAA is associated with an elevated risk of amyloid-related imaging abnormalities (ARIA). The CAA risk score (CAARS) can stratify CAA risks in living patients. Hypertension and apolipoprotein E (APOE) ε4 are risk factors for CAA. The CAARS has the potential to predict risk of ARIA.