Diagnostic Accuracy of the Boston Criteria v2.0 in Memory Clinic Patients: An MRI-Neuropathology Validation Study.

Abstract (English)

BACKGROUND AND OBJECTIVES: Cerebral amyloid angiopathy (CAA) is common in older adults and frequently contributes to cognitive impairment and dementia. Existing in vivo diagnostic criteria for CAA (Boston Criteria) were developed primarily in patients with intracerebral hemorrhage, and their performance in memory clinic populations remains uncertain. The updated Boston Criteria v2.0 incorporate nonhemorrhagic MRI markers intended to improve case detection. We evaluated the diagnostic accuracy of the Boston Criteria v1.5 and v2.0 against neuropathologically confirmed CAA in memory clinic patients. METHODS: We performed a retrospective diagnostic accuracy study of participants from the Alzheimer's Disease Neuroimaging Initiative and National Alzheimer's Coordinating Center, selected based on availability of required brain MRI and autopsy-based neuropathology data. Patients were classified as no, possible, or probable CAA according to the Boston Criteria v1.5 and v2.0. The primary reference standard was moderate-to-severe neuropathologic CAA; analyses using any neuropathologic CAA were secondary/exploratory. Diagnostic performance was assessed using sensitivity, specificity, predictive values, likelihood ratios, F1 scores, accuracy, and area under the (receiver-operating characteristic) curve (AUC), with formal comparisons between criteria versions. RESULTS: Eighty patients were included (mean age: 81 years, interquartile range 74-86 years; 36.2% female, ∼80% with dementia and Alzheimer disease). Using moderate-to-severe CAA as the neuropathologic reference standard, probable CAA by Boston Criteria v1.5 had a sensitivity of 32% (95% CI 15%-50%), specificity 87% (77%-95%), and AUC 0.59 (0.49-0.69). Using the Boston Criteria v2.0, the corresponding values were 43% (25%-62%), 83% (72%-92%), and 0.63 (0.52-0.74), respectively. No overall performance measures were significantly different between the criteria versions. Secondary analyses using any neuropathologic CAA showed similar patterns. DISCUSSION: In memory clinic patients, both Boston Criteria versions showed only modest overall diagnostic performance against neuropathology. Compared with v1.5, Boston Criteria v2.0 showed a numerical shift toward greater sensitivity at the expense of specificity but no clear overall gain in accuracy. These findings support cautious, context-dependent interpretation of MRI-based CAA criteria in memory clinic settings and highlight the need for additional biomarkers to improve in vivo diagnosis in nonhemorrhagic populations. CLASSIFICATION OF EVIDENCE: This study provides Class II evidence that, in memory clinic populations, Boston Criteria v2.0 show a trade-off between sensitivity and specificity for probable CAA diagnosis, with only modest overall diagnostic accuracy for identifying moderate-to-severe neuropathologically defined CAA.