Postoperative magnetic resonance imaging may predict poor outcome in children with severe traumatic brain injuries who undergo cranial surgery.

Objective: Multiple studies have evaluated the use of MRI for prognostication in pediatric patients with severe traumatic brain injury (TBI) and have found a correlation between diffuse axonal injury (DAI)-type lesions and outcome. However, there remains a limited understanding about the use of MRI for prognostication after severe TBI in children who have undergone cranial surgery.

Methods: Children with severe TBI who underwent craniectomy or craniotomy at Primary Children's Hospital in Salt Lake City, Utah, between 2010 and 2019 were identified retrospectively. Of these 92 patients, 43 underwent postoperative brain MRI within 4 months of surgery. Susceptibility-weighted imaging (SWI) and FLAIR sequences were used to designate areas of hemorrhagic and nonhemorrhagic cerebral lesions related to DAI. Patients were then stratified based on the location of the DAI as read by a neuroradiologist as superficial, deep, or brainstem. The location of the DAI and other variables associated with poor outcome, including Glasgow Coma Scale (GCS) score, pediatric trauma score, mechanism of injury, and time to surgery, were analyzed for correlation with poor outcome. Outcomes were reported using the King's Outcome Scale for Childhood Head Injury (KOSCHI).

Results: In the 43 children with severe TBI who underwent postoperative brain MRI, the median GCS score on arrival was 4. The most common cause of injury was falls (14 patients, 33%). The most common primary intracranial pathology was subdural hematoma in 26 patients (60%), followed by epidural hematoma in 9 (21%). Fifteen patients (35%) had cerebral herniation and 31 (72%) had evidence of contusion. Variables associated with poor outcome included cerebral herniation (r = 0.338, p = 0.027) and location of DAI (r = 0.319, p = 0.037). In a separate analysis, brainstem DAI was shown to predict poor outcome, whereas location (no, superficial, or deep DAI) did not. Logistic regression showed that brainstem DAI (OR 22.3, p = 0.020) had a higher odds ratio than cerebral herniation (OR 10.5, p = 0.044) for poor outcome. Thirty-six children (84%) had a satisfactory outcome at last follow-up; 3 (7%) children died.

Conclusions: The majority of children in this series who presented with a severe TBI and underwent craniectomy or craniotomy made a satisfactory recovery. In patients in whom there is a concern for poor outcome, the location of DAI-type lesions with SWI and FLAIR may assist in prognostication. The authors' results revealed that DAI-type lesions in the brainstem and evidence of cerebral herniation may indicate a poorer prognosis; however, more studies with larger cohorts are needed to make definitive conclusions.