S100B outperforms clinical decision rules for the identification of intracranial injury on head CT scan after mild traumatic brain injury

Objective: To compare the classification accuracy of S100B to two clinical decision rules- Canadian CT Head Rule (CCHR) and New Orleans Criteria (NOC)-for predicting traumatic intracranial injuries (ICI) after mild traumatic brain injury (mild TBI). Methods: A secondary analysis of a prospective observational study of mild TBI patients was performed. The diagnostic performance of S100B for predicting ICI on head CT was compared to both the CHRR and NOC. Area under receiver operator characteristic (AUC) curves were used and multivariable analysis was used to create a new decision rule based on a combination of S100B and decision rule-related variables. Results: S100B had the highest negative predictive value (97.3%), positive predictive value (7.21%), specificity (33.6%) and positive likelihood ratio (1.3), and the lowest negative likelihood ratio (0.5). The proportion of mild TBI subjects with potentially avoidable head CT scans was highest using S100B (37.7%). The addition of S100B to both clinical decision rules significantly increased AUC. A novel decision rule adding S100B to three decision rule-related variables significantly improved prediction (p < 0.05). Conclusion: Serum S100B outperformed clinical decision rules for identifying mild TBI patients with ICI. Incorporating clinical variables with S100B maximized ICI prediction, but requires validation in an independent cohort.