An exploratory investigation of potential biomarkers of inhibitory control in patients with mild traumatic brain injury

Background Mild traumatic brain injury (mTBI) often leads to persistent deficits in executive functioning, including inhibitory control. We evaluated the potential association between a panel of blood-based inflammatory, vascular and neuroglial markers and performance in inhibitory control tests in a cohort of patients 30 days after mTBI. Methods Executive function was assessed with Five-Digit Test (FDT), Psychology Experiment Building Language Continuous Performance Test (PEBL CPT/PCPT), and Hayling Test in 20 patients with mTBI, yielding composite scores for inhibitory control, flexibility, reaction time, and error rates. A three-stage machine learning framework was applied: (1) outlier identification using standard deviation and interquartile range methods; (2) permutation-based feature importance using Random Forest regression; and (3) multiple linear regression with backward elimination of predictors with nominal p-values ≥ 0.05. A cross-validated LASSO analysis was additionally performed as a conservative sensitivity analysis. Results Proinflammatory mediators, such as soluble CD40 ligand (sCD40L), and growth-regulated oncogene (GRO/CXCL1), were associated with poorer inhibitory control in specific measures. In contrast, the regulatory cytokine interleukin-4 (IL-4), extracellular matrix remodeling enzymes, including matrix metalloproteinase-9 (MMP-9), lipocalin-2, and fibroblast activation protein (FAP), and growth factors, including platelet-derived growth factor-AA (PDGF-AA) and neuregulin-1-β1 (NRG1-β1), were associated with better performance in selected outcomes. Copeptin exhibited context-dependent associations within the PCPT. Repeated cross-validation indicated limited out-of-sample predictive performance, and the cross-validated LASSO sensitivity analysis retained no non-zero predictors. Conclusions This study leverages machine learning methods to link blood‐based biomarkers to inhibitory control in mTBI, identifying hypothesis-generating associations that require validation in larger independent cohorts.