Data-Driven Clustering Approach to Identify Different Phenotypes of Primary Central Nervous System Vasculitis

Area of Science:

• Neurology
• Immunology
• Radiology

Background:

• Primary CNS vasculitis (PCNSV) is a rare condition affecting the brain and spinal cord vasculature.
• Understanding PCNSV phenotypes is crucial for accurate diagnosis and treatment.
• Previous classifications have not fully captured the heterogeneity within PCNSV.

Purpose of the Study:

• To investigate if hierarchical unsupervised cluster analysis can distinguish phenotypic differences in adult PCNSV patients.
• To identify distinct patient groups based on clinical and radiological characteristics.

Main Methods:

• Utilized agglomerative hierarchical cluster analysis (Ward method) on 153 adult PCNSV patients from the COVAC registry.
• Analyzed baseline phenotypic data including neurological deficits, neuroimaging findings, and diagnostic work-up results.

Main Results:

• Two distinct clusters emerged: Cluster 1 (71%) with motor deficits and acute infarcts, and Cluster 2 (29%) with seizures and parenchymal lesions.
• Cluster 2 showed a higher rate of positive CNS biopsies (95%) compared to Cluster 1 (28%).
• Functional independence, relapse, and mortality rates did not significantly differ between clusters at 12 months.

Conclusions:

• Unsupervised analysis revealed two distinct clinical and radiological phenotypes in PCNSV.
• These findings underscore the importance of considering vessel size in differentiating PCNSV phenotypes and guiding diagnostic approaches.