Abstract W P16: Comparing Outcome Prediction Scoring Systems in the Interventional Management of Stroke (IMS) III Trial: No Clear Winner

Abstract

Introduction: Several prognostic scores have been developed to predict patient outcome after ischemic stroke. We aimed to compare eight scores in the IMS -III dataset. Hypothesis: Prediction of good and poor outcome using different scores is possible in IMS-III. Methods: We retrospectively calculated area under the curve (AUC), sensitivity, and specificity for poor (mRS 4-6) and good (mRS 0-2) 90-day outcome in all IMS-III participants with complete data. The scoring systems analyzed were: Houston Intra-arterial Therapy 2 (HIAT2), Ischemic Stroke Predictive Risk Score (iScore), Totaled Health Risks in Vascular Events (THRIVE), DRAGON, Hemorrhage after Thrombolysis (HAT), Alberta Stroke Programme Early CT Score (ASPECTS), Stroke-Thrombolytic Predictive Instrument (TPI), and Simple Variables Model (SVM). Those receiving IV rtPA followed by endovascular therapy (treatment) and those receiving IV rtPA only (control) were analyzed separately. Results: Among 656 IMS III subjects, 651 had complete data, including 429 treatment and 222 control subjects. Overall, all scoring systems showed good prediction abilities (AUC range 0.60 to 0.78). AUC comparisons did not clearly favor a single scoring system. Treatment group: For predicting poor outcome, the TPI score outperformed ASPECTS (AUC 0.73 vs 0.61; P <0.05). For predicting good outcome, the HAT score was significantly better than ASPECTS (0.73 vs. 0.64; P <0.05). Control group: For poor outcome, the TPI score was significantly better than ASPECTS and THRIVE (0.78 vs. 0.62 and 0.71, respectively; P <0.05). For good outcome, TPI performed better than ASPECTS and THRIVE (0.73 vs. 0.61 and 0.67, respectively; P <0.05). Conclusions: Outcome prediction using different scores in IMS III is possible. HIAT2, iSCORE, DRAGON, HAT, TPI, and SVM all performed comparably. Prognostic performances were comparable in both treatment arms.