Boston Criteria 2.0 stratifies hemorrhage location within mixed intracerebral hemorrhage: external validation of CADMUS small vessel disease classification.

BACKGROUND: CADMUS is an MRI-based classification for small vessel disease (SVD) related intracerebral hemorrhage (ICH), providing distinct SVD phenotypes: cerebral amyloid angiopathy (CAA), deep perforator arteriopathy (DPA), Mixed CAA-DPA, and Undetermined SVD. We evaluated construct validity in an external cohort by examining whether CADMUS subtypes demonstrate biologically coherent anatomical distributions, and whether lobar hematoma location is associated with probable CAA within the mixed ICH subtypes: Mixed CAA-DPA and Undetermined SVD. METHODS: Cross-sectional analysis of patients with first-ever spontaneous ICH admitted to a tertiary hospital (2008-2024). Location was classified as lobar supratentorial, deep/non-lobar supratentorial, or infratentorial. Differences in hematoma location across CADMUS subtypes were assessed by chi-square with Bonferroni-corrected pairwise comparisons. Within the mixed ICH subtypes, we assessed whether lobar hemorrhage location was associated with Probable CAA according to the Boston Criteria 2.0. RESULTS: There were 843 patients with sufficient neuroimaging for CADMUS classification: 164 (19.5%) with CAA, 185 (18%) with DPA and 495 (62.5%) with one of the mixed ICH subtypes Location differed markedly across subtypes (χ2 = 394.2, p < 0.001): CAA was exclusively lobar and DPA exclusively deep or infratentorial. Among patients with a mixed ICH subtype, those with lobar hemorrhage location were more likely to have Probable CAA criteria than those with a non-lobar hemorrhage: 63/262 (24.0%) of lobar patients were Boston-probable compared with 1/232 (0.4%) of non-lobar patients OR 73.1 (95% CI 10-534, p < 0.001). CONCLUSIONS: CADMUS subtypes demonstrate near-complete anatomical segregation in an independent external cohort. Among patients with a mixed ICH subtype, lobar hemorrhage location was strongly associated with Probable CAA, supporting its use as a potential marker of CAA in this population.