Traumatic brain injury (TBI) is associated with substantial consumption of health care resources. No clinical or paraclinical examination can reliably predict neurological evolution. In this study, we evaluated the ability of a combined clinical and MRI approach to predict outcome.
Methods: This prospective study took place between June 2001 and March 2005 in a Neurosurgical Intensive Care Unit in Paris, France. Inclusion criteria were TBI patients still mechanically ventilated and without clinical signs of awareness after 2 weeks. Four clinical signs were assessed after cessation of sedation: grasping, yawning, chewing and paroxysmal sympathetic storm. FLAIR and T2* acquisitions on MRI were used in order to localize brain lesions. Statistically linked regions (clusters) were defined. Outcome was assessed at one year by Glasgow Outcome Scale (GOS).
Findings: 73 patients were included: 41 had poor outcome (GOS 1-3) and 32 had good outcome (GOS 4-5). Lesions in the clusters "right upper pons and right lower midbrain"," hypothalamus and basal forebrain","left parietal, left temporal, left occipital lobes and left insula" and the presence of grasping or chewing were associated with poor outcome in multivariate analysis. This combined clinical and MRI approach gives a much better prediction than MRI approach only (P < 0.009), with an area under the ROC curve of 0.94 (95 % CI, 0.89-1.00).
Interpretation: These data suggest that MRI associated with clinical assessment improves outcome prediction in severe TBI patients.