Determining Acute-Phase Biomarkers for Mild Traumatic Brain Injury Through Exploratory Data Analysis: A Preliminary...

The purpose of this study is to identify acute-phase serum biomarkers for mild traumatic brain injury (mTBI) within 24 h of trauma and evaluate their discriminatory value versus orthopedic trauma and healthy status using a leakage-safe machine-learning framework. We enrolled 60 patients with computed tomography (CT)-negative mTBI (World Health Organization [WHO] criteria; Glasgow Coma Scale [GCS] 13-15; age 18-59) within 24 h of injury, 17 orthopedic injury controls, and 24 healthy controls without mTBI in the prior 5 years. Fifty-nine serum biomarkers were quantified by Luminex®. The group comparisons used ANOVA or Kruskal-Wallis tests with corrected post hoc contrasts. Random Forest models were trained under leakage-safe nested cross-validation (outer 5-fold; inner 3-fold) with biomarker-only and biomarker-plus-demographics feature sets. SHAP values were computed on held-out folds and stability was assessed across resamples. Twenty-three biomarkers differed across groups after Bonferroni correction for post hoc comparisons (BDNF, EGF, Fracktalkine, G-CSF, GRO, IL15, IL-1α, IL-1β, IL-2, IL-3, IL-4, IL-5, IL-9, LIGHT, MMP-9, MPO, NCAM, NRG1-β1, S100B, TGF- α, VEGF A, sICAM-1). In nested cross-validation, three-class balanced accuracy ranged ~0.48-0.70, and mTBI one-vs-rest receiver operating characteristic area under the curve (ROC-AUC) frequently exceeded 0.85; Control vs mTBI classification achieved balanced accuracy ~0.70-0.90 with precision-recall area under the curve (PR-AUC) consistently >0.82. Sex and age added minimal incremental value. Stable SHAP contributors (top-5 in ≥4/5 folds) were LIGHT, IL-5, sCD40L, MMP-9, EGF, and fractalkine. Acute mTBI is associated with a reproducible serum signature spanning inflammatory, vascular, and growth-factor pathways.