Development and validation of a predictive model to forecast the cerebrovascular risk burden in older people using the NACC and ROSMAP dataset.

Résumé

INTRODUCTION: The integrity of the brain's vascular system is vital for neuronal health. Cerebrovascular microbleeds, microinfarcts, and infarcts (defined as cerebrovascular disease burden) contribute to stroke and cognitive impairment. Here, we developed and validated the first model, the Cerebrovascular Disease Burden Risk Score (CDBRS) for individualized risk prediction.METHODS: Leveraging the AutoScore and AutoScore-Ordinal algorithms, the CDBRS was developed using National Alzheimer's Coordinating Center (NACC) data and externally validated on Religious Orders Study and Memory and Aging Project (ROSMAP) data. The CDBRS was evaluated using mean area under the receiver operating characteristic curve (mAUC), Harrell's generalized c-index, sensitivity, and specificity.RESULTS: The CDBRS achieved a promising predictive performance and outperformed the respective baseline models. The performance of the CDBRS remains good on external validation.DISCUSSION: CDBRS can possibly aid in identifying individuals with high risks of cerebrovascular microbleeds, microinfarcts, and infarcts, to enable closer monitoring and personalized management. Further validation of CDBRS in a diverse population is warranted.HIGHLIGHTS: Microbleeds, microinfarcts, and infarcts often precede clinical stroke and vascular dementia. We developed and validated the Cerebrovascular Disease Burden Risk Score (CDBRS), a new and interpretable model for stratifying the risk of microbleeds, microinfarcts, and infarcts in older adults. This tool developed using National Alzheimer's Coordinating Center (NACC) participants data achieved promising predictive performance in Religious Orders Study and Memory and Aging Project (ROSMAP) participants. The robust predictive performance warrants its clinical validation is prospective studies.