Comparison of MRI-Based Traumatic Axonal Injury Grading Scores to Predict Patient Outcome

Background/Objectives: To compare the predictive performance of seven previously published MRI-based traumatic axonal injury (TAI) scoring systems for four key outcomes: in-hospital mortality, coma behavioral profile (disorders of consciousness [DoC] vs. clinical cognitive-motor dissociation [cCMD]) at early phase of hospitalization, functional outcome at discharge and functional outcome (Glasgow Outcome Scale Extended [GOSE]) at 6 months from hospital discharge. Methods: The Gentry, Firsching, Abu Hamdeh, Stockholm, Trondheim, Lausanne logistic regression (L-LR) and support vector machine (L-SVM) scores were retrospectively and independently assessed by two board-certified neuroradiologists, from MRI patients with TBI admitted to our single center Acute Neurorehabilitation Unit. Receiver operating characteristic curves and logistic regression analyses adjusted for age, Marshall CT score, and initial Glasgow Coma Score (GCS) were performed. p-value corrected for multiple testing was set <0.007. Results: Of the 107 patients, 91 who underwent CT and MRI were included (median age: 40 [IQR:25–63], 72 males, GCS: 6 [4–8]), 15 died during hospitalization and 17 had DoC. Only the L-LR (OR: 1.38 [1.10–1.74], p = 0.006) and L-SVM (OR: 1.31 [1.08–1.59], p = 0.005) scores significantly predicted in-hospital mortality. The Trondheim score (OR: 2.68 [1.46–4.94], p = 0.002), L-LR (OR:1.30 [1.11–1.51], p = 0.001) and L-SVM (OR: 1.23 [1.09–1.39], p = 0.001) scores significantly predicted the DoC status. The Stockholm grading score (OR: 3.37 [1.53–7.40], p = 0.003), L-LR (OR:1.28 [1.11–1.47], p = 0.001) and L-SVM (OR: 1.15 [1.05–1.26], p = 0.002) scores were significantly related to poor functional outcome at 6 months. Conclusions: Both Lausanne scores were the only models significantly associated with in-hospital mortality, DoC phenotype classification, and functional outcome at discharge and at 6 months after TBI.