A serum protein biomarker panel improves outcome prediction in human traumatic brain injury

Journal of Neurotrauma
|

Eric Peter Thelin, Faiez Al Nimer, Arvid Frostell , Henrik Zetterberg, Kaj  Blennow, Harriet Nyström, Mikael Svensson, Bo‐Michael Bellander, Fredrik Piehl, David W Nelson

J. Neurotrauma. 2019 May 9.

DOI: 10.1089/neu.2019.6375

 

ABSTRACT

Brain-enriched protein biomarkers of tissue fate are being introduced clinically to aid in traumatic brain injury (TBI) management. The aim of this study was to determine how concentrations of six different protein biomarkers, measured in samples collected during the first weeks after TBI, relate to injury severity and outcome. We included neuro-critical care TBI patients that were prospectively enrolled from 2007 to 2013, all having 1 to 3 blood samples drawn during the first two weeks. The biomarkers analyzed were S100B, neuron-specific enolase (NSE), glial fibrillary acidic protein (GFAP), ubiquitin carboxy-terminal hydrolase-L1 (UCH-L1), tau and neurofilament-Light (NF-L). Glasgow Outcome Score (GOS) was assessed at 12 months. In total, 172 patients were included. All serum markers were associated with injury severity as classified on computed tomography scans at admission. Almost all biomarkers outperformed other known outcome predictors with higher levels the first five days, correlating with unfavorable outcomes, and UCH-L1 (0.260 pseduo-R2) displaying the best discrimination in univariate analyses. After adjusting for acknowledged TBI outcome predictors, GFAP and NF-L added most independent information to predict favorable/unfavorable GOS, improving the model from 0.38 to 0.51 pseudo-R2. A correlation matrix indicated substantial co-variance, with the strongest correlation between UCH-L1, GFAP and tau (r=0.827 to 0.880). Additionally, the principal component analysis exhibited clustering of UCH-L1 and tau, as well as GFAP, S100B and NSE, which was separate from NF-L. In summary, a panel of several different protein biomarkers, all associated with injury severity, with different cellular origin and temporal trajectories, improve outcome prediction models.