Abstract TP40: Prediction of a Small DWI Lesion Using Single Phase Collateral Score and Noncontrast CT ASPECT Score

Abstract

Introduction and Hypothesis: Small initial DWI lesion volume (≤70 mL) is a strong predictor of favorable outcome post i IA thrombectomy and has been used for patient selection in multiple clinical trials. We sought to investigate if collateral and/or CT ASPECT scores could predict a small DWI lesion volume. Methods: We included 130 consecutive patients with CTA showing ICA and/or proximal MCA occlusions who underwent DWI within 8 hours of stroke onset. DWI lesion volumes were dichotomized to small (≤70mL) and large (>70ml). Collateral vessels were categorized to 5 groups: 0=none, 1=minimal, 2=less than 50%, 3=more than 50%, 4=equal and 5=more than the contralateral side. Logistic regression, ROC curve analyses and weighted Kappa test were performed. Results: In 130 patients, 62 female (48%), the average values (mean±SD) were: age 70±17 years, NIHSS 16±6, DWI volume 59±65 mL, time after stroke onset 4.4±1.8 hours, CT ASPECT score 6±2 and collateral score 3±1. Time, age, gender, or occlusion site (ICA vs MCA) were not independent predictors of a small DWI volume with p values all >0.05. Collateral score (OR 6.32, P<0.0001), CT ASPECT score (OR 2.37, P<0.0001), and NIHSS (OR 0.87, P<0.0003) significantly contributed to prediction of a small DWI volume. ROC curve analysis with both ASPECT and collateral scores contributing to predicted probabilities showed an AUC of 0.93. The AUC was 0.88 and 0.85 for collateral and ASPECT scores alone respectively. Collateral score ≥3 with ASPECT ≥9 yields specificity of 97.4% and sensitivity of 48.9% while collateral score ≥4 alone provides specificity of 97.4% and sensitivity of 38.9%, missing approximately 16% and 43% of potential candidates for IA therapy based on a DWI lesion volume < 70 mL respectively. Inter-rater agreements between two neuroradiologists were k=0.53 (41.5% agreement) for collateral and k=0.53 (32.3% agreement) for ASPECT scores. Conclusion: Collateral score alone or combined with ASPECTS can predict a small DWI lesion volume with high specificity. However, sensitivity is relatively low and inter-rater agreement is fair. These findings should be considered when DWI is not available to select patients for IA therapy.