Introduction There are currently two prognostic tools available for predicting outcome in traumatic brain injury (TBI): prognostic models combining clinico-demographic characteristics of patients for outcome prediction and brain injury biomarkers (eg, S100B).
Objective To identify which method has better prognostic strength.
Methods We analysed data from 100 TBI patients, all of whom were admitted to the intensive care unit and had venous S100B levels recorded at 24-h after injury. TBI prognostic models A and B, constructed in Trauma Audit and Research Network (TARN)1, were run on the dataset and then S100B was added as an independent predictor to each model. Furthermore, another model was developed containing only S100B and subsequently, other important TBI predictors were added to assess their ability to enhance the predictive power of this model. The outcome measures were survival and favourable outcome at 3 months.
Results Among all the prognostic variables (including age, cause of injury, Glasgow Coma Scale, pupillary reactivity, Injury Severity Score (ISS) and CT classifications); S100B has the highest predictive strength on multivariate analysis. No difference between performance of prognostic models or S100B in isolation was observed. Addition of S100B to the prognostic models improves the performance (eg, area under the roc curve (AUC), R2 Nagelkerke and classification accuracy of TARN model A to predict survival increase from 0.64, 0.08 and 71% to 0.72, 0.20 and 74.7% respectively). Similarly, the predictive power of S100B increases by adding other predictors to S100B (eg, AUC (0.69 vs 0.78), R2 Nagelkerke (0.15 vs 0.30) and classification accuracy (73% vs 77%) for survival prediction).
Conclusion S100B appears to be the strongest prognostic variable in TBI. A better prognostic tool than those which are currently available may be a combination of both clinicodemographic predictors with S100B.