Both glial fibrillary acidic protein (GFAP) and S100β are found in glial cells and are released into serum following a traumatic brain injury (TBI), however, the clinical utility of S100β as a biomarker has been questioned because of its release from bone. This study examined the ability of GFAP and S100β to detect intracranial lesions on computed tomography (CT) in trauma patients and also assessed biomarker performance in patients with fractures and extracranial injuries on head CT. This prospective cohort study enrolled a convenience sample of adult trauma patients at a Level I trauma center with and without mild or moderate traumatic brain injury (MMTBI). Serum samples were obtained within 4 h of injury. The primary outcome was the presence of traumatic intracranial lesions on CT scan. There were 397 general trauma patients enrolled: 209 (53%) had a MMTBI and 188 (47%) had trauma without MMTBI. Of the 262 patients with a head CT, 20 (8%) had intracranial lesions. There were 137 (35%) trauma patients who sustained extracranial fractures below the head to the torso and extremities. Levels of S100β were significantly higher in patients with fractures, compared with those without fractures (p<0.001) whether MMTBI was present or not. However, GFAP levels were not significantly affected by the presence of fractures (p>0.05). The area under the receiver operating characteristics curve (AUC) for predicting intracranial lesions on CT for GFAP was 0.84 (0.73-0.95) and for S100β was 0.78 (0.67-0.89). However, in the presence of extracranial fractures, the AUC for GFAP increased to 0.93 (0.86-1.00) and for S100β decreased to 0.75 (0.61-0.88). In a general trauma population, GFAP out-performed S100β in detecting intracranial CT lesions, particularly in the setting of extracranial fractures.

Objectives This study compared the clinical performance of the Canadian CT Head Rule (CCHR) and the New Orleans Criteria (NOC) for detecting any traumatic intracranial lesion on computed tomography (CT) in patients with a Glasgow Coma Scale (GCS) score of 15. Also assessed were ability to detect patients with “clinically important” brain injury and patients requiring neurosurgical intervention. Additionally, the performance of the CCHR was assessed in a larger cohort of those presenting with GCS of 13 to 15. Methods This prospective cohort study was conducted in a U.S. Level I trauma center and enrolled a consecutive sample of mildly head-injured adults who presented to the emergency department (ED) with witnessed loss of consciousness, disorientation or amnesia, and GCS 13 to 15. The rules were compared in the group of patients with GCS 15. The primary outcome was prediction of “any traumatic intracranial injury” on CT. Secondary outcomes included “clinically important brain injury” on CT and need for neurosurgical intervention. Results Among the 431 enrolled patients, 314 patients (73%) had a GCS of 15, and 22 of the 314 (7%) had evidence of a traumatic intracranial lesion on CT. There were 11 of 314 (3.5%) who had “clinically important” brain injury, and 3 of 314 (1.0%) required neurosurgical intervention. The NOC and CCHR both had 100% sensitivity (95% confidence interval [CI] = 82% to 100%), but the CCHR was more specific for detecting any traumatic intracranial lesion on CT, with a specificity of 36.3% (95% CI = 31% to 42%) versus 10.2% (95% CI = 7% to 14%) for NOC. For “clinically important” brain lesions, the CCHR and the NOC had similar sensitivity (both 100%; 95% CI = 68% to 100%), but the specificity was 35% (95% CI = 30% to 41%) for CCHR and 9.9% (95% CI = 7% to 14%) for NOC. When the rules were compared for predicting need for neurosurgical intervention, the sensitivity was equivalent at 100% (95% CI = 31% to 100%) but the CCHR had a higher specificity at 80.7% (95% CI = 76% to 85%) versus 9.6% (95% CI = 7% to 14%) for NOC. Among all 431 patients with a GCS score 13 to 15, the CCHR had sensitivities of 100% (95% CI = 84% to 100%) for 27 patients with clinically important brain injury and 100% (95% CI = 46% to 100%) for five patients requiring neurosurgical intervention. Conclusions In a U.S. sample of mildly head-injured patients, the CCHR and the NOC had equivalently high sensitivities for detecting any traumatic intracranial lesion on CT, clinically important brain injury, and neurosurgical intervention, but the CCHR was more specific. A larger cohort will be needed to validate these findings.