Do Imaging Markers Of Cerebral Small Vessel Disease Predict Hematoma Volume And Outcome In Acute Intracerebral Hemorrhage?

Background and Purpose: Cerebral small vessel disease (CSVD) markers have not been widely studied in relation to hematoma volume and growth in hypertensive intracerebral hemorrhage (ICH). The objectives to assess the relationship of white matter hyperintense lesions (WMHL), microbleeds (MBs), and cortical siderosis (CSS) with hematoma volume, hematoma expansion (HE), and 3 months outcome in patients with hypertensive ICH. Methods: All consecutive acute hypertensive supratentorial ICH presenting to the emergency were prospectively recruited. Baseline and 24 hours computed tomography (CT) to assess hematoma volume and magnetic resonance imaging (MRI) for CSVD markers were performed in all subjects. WMHL (graded using Fazekas's scale), MBs, and CSS were assessed and compared with baseline variables and outcomes. All the images were assessed by an experienced stroke neurologist/neuroradiologist. Results: One hundred and fifty-seven patients were screened and 60 were included. Mean age was 54.08 +/- 11.57 years and 47 (78%) were males. Of 60, 19 (28.1%) had HE, 31 (51.6%) had major bleed (>30 ml), and 28 (47.46%) had poor 3 month outcome (mRS 4-6). On univariate analysis, high grade WMHL was associated with greater HE [odds ratio (OR): 2.65, confidence interval (CI) 1.48-4.72, P = 0.001), greater proportion with volume >30 ml (OR: 7.16, CI: 1.09-47.13, P = 0.001) and poor outcome (OR: 2.1, CI: 0.05-3.27, P = 0.001). MBs were associated with poor outcome (P = 0.029) but not with HE/volume. CSS was related to HE (P = 0.031), a large volume bleed (P = 0.023), and poor outcome (P = 0.021). On multivariate model, only WMHL independently predicted HE (P = 0.034), greater proportion with bleed volume >30 ml (P = 0.041), and poor outcome (P = 0.042). Conclusions: WMHL in MRI serves as a predictor of hematoma expansion, a large volume bleed, and poor outcome in hypertensive ICH and may be incorporated into existing prediction models.