A CT imaging-based prediction model of functional outcome and benefit of endovascular thrombectomy for ischemic stroke

OBJECTIVE: To investigate the value of baseline CT imaging for the prediction of functional outcome and benefit of endovascular thrombectomy (EVT) for anterior large vessel occlusion (LVO). MATERIALS AND METHODS: We used individual patient data from seven randomized EVT trials and included patients with available baseline CT imaging and outcome data. We developed a model to predict functional outcome and benefit of EVT, including baseline stroke-related and brain frailty CT imaging features alone. We compared the discriminative performance of our model for predicting good functional outcome (modified Rankin Scale [mRS] 0-2) and treatment benefit (difference between the probability of mRS 0-2 with vs without EVT) with MR PREDICTS by calculating the difference in C-statistics (delta C and delta C-for-benefit). RESULTS: We included 1391 patients (median age, 67 years, interquartile range 59-76; 53% male). Discrimination of the model based on CT imaging alone was substantial for the prediction of good functional outcome (C-statistic 0.700, 95% CI: 0.666-0.731) and treatment benefit (C-for-benefit 0.640, 95% CI: 0.586-0.690). After adding the known strongest clinical predictors namely age and National Institutes of Health Stroke Scale score, discrimination improved to slightly lower than MR PREDICTS for prediction of good functional outcome (C-statistic 0.733 vs 0.750; delta C, -0.017 [95% CI: -0.037 to 0.003]) and treatment benefit (C-for-benefit 0.675 vs 0.692; delta C-for-benefit -0.017 [95% CI: -0.084 to 0.050]). CONCLUSIONS: Baseline CT imaging holds considerable predictive value with regard to functional outcome and treatment benefit, but a combination of clinical and imaging features offers the best predictive performance. KEY POINTS: Question The predictive value of baseline CT imaging for the prediction of functional outcome and benefit of EVT for anterior LVO stroke is uncertain. Findings Discrimination of a model based on CT imaging alone is substantial, but can further be improved by the addition of limited clinical characteristics. Clinical relevance Baseline CT imaging holds considerable predictive value with regard to functional outcome and treatment benefit. The addition of limited clinical information is needed to achieve predictive performance similar to an established prediction model.