Predictive Value of Clinical, CSF and Vessel Wall MRI Variables in Diagnosing Primary Angiitis of the CNS.

Abstract

Background and objectives: Without brain biopsy, there are limited diagnostic predictors to differentiate primary angiitis of the CNS (PACNS) from intracranial atherosclerotic disease (ICAD). We examined the utility of clinical, CSF, and quantitative vessel wall magnetic resonance imaging (VWMRI) variables in predicting PACNS from ICAD.

Methods: In this cross-sectional design, observational study, we reviewed electronic medical records to identify patients (18 years and older) who presented to our medical center between January 2015 and December 2021 for ischemic stroke due to intracranial vasculopathy. Patients with biopsy-proven PACNS, probable PACNS, or ICAD were included. Patients with secondary CNS vasculitis or no VWMRI data were excluded. On VWMRI, for each patient, a total of 20 vessel wall segments were analyzed for percent concentricity, percent irregularity, and concentricity to eccentricity (C/E) ratios. We also collected several clinical and CSF variables. Using logistic regression models, we assessed the diagnostic value of VWMRI, CSF, and clinical variables in predicting PACNS in patients with biopsy-proven disease. We then performed a sensitivity analysis to assess predictors of biopsy-proven and probable PACNS.

Results: Thirty-two patients with ICAD (54.2%) and 27 patients with PACNS (45.8%) were included. Of the patients with PACNS, 21 (77.8%) were not biopsied and considered probable PACNS. Twenty-four patients with ICAD (75%) and 6 biopsy-proven patients with PACNS (22.2%) showed large vessel involvement and were included in the primary analysis. Encephalopathy (odds ratio [OR], 7.60; 95% CI 1.07-54.09) and seizure (OR 23.00; 95% CI 1.77-298.45) were significantly associated with PACNS. All patients were included in the sensitivity analysis, in which headache significantly predicted PACNS (OR 7.60; 95% CI 1.07-54.09). In the primary analysis, for every 1 white blood cell/µL increase in CSF, there was a 47% higher odds of PACNS (OR 1.47; 95% CI 1.04-2.07). On VWMRI, a C/E ratio >1 (OR 115.00; 95% CI 6.11-2165.95), percent concentricity ≥50% (OR 55.00; 95% CI 4.13-732.71), and percent irregularity <50% (OR 55.00; 95% CI 4.13-732.71) indicated significantly higher odds of PACNS compared with ICAD.

Discussion: Our results suggest that quantitative VWMRI metrics, CSF pleocytosis, and clinical features of encephalopathy, seizure, and headache significantly predict a diagnosis of probable PACNS when compared with ICAD.