Multiple indicators model of long-term mortality in traumatic brain injury

Abstract

Objective: To examine the prognostic ability of protein S100B, neuron-specific enolase (NSE) and glial fibrillary acid protein (GFAP) for prediction of 1-year mortality in patients with traumatic brain injury (TBI) in relation to clinical and radiological characteristics of TBI.

Methods: Brain injury was quantified in 84 patients (Glasgow Coma Scale [GCS] ≤ 12) using clinical (GCS, pupils), radiological (computed tomography [CT] classification and individual CT characteristics) and biochemical (S100B, NSE and GFAP) data at admission and in the acute post-injury period.

Results: Initial and peak S100B, NSE and GFAP concentrations were higher in non-survivors (n = 26) than in survivors (p-value range: <0.001–0.018). Cox regression showed that GFAP and S100B concentration and the temporal profile of S100B were more powerful independent predictors of mortality than baseline clinical and radiological characteristics or clinical and radiological indicators of neurological deterioration. The prognostic models containing admission variables and those available during the subsequent clinical course showed the same discrimination ability (area under receiver characteristic curve 0.92), but the model based on variables available in the acute post-injury period calibrated better (p = 0.428).

Conclusion: Mortality at 1-year post-TBI is accurately predicted by the combination of GFAP and S100B concentration and clinical and radiological characteristics at admission or in the acute post-injury period.

• S100B
• NSE
• GFAP
• Glasgow Coma Scale
• CT
• prognostic model