Multi-Omic Admission-Based Biomarkers Predict 30-day Survival and Persistent Critical Illness in Trauma Patients after Injury

Abstract

Introduction: Admission-based biomarkers for the prediction of outcomes in trauma patients could be useful for clinical-decision support. The extent to which biomarkers from different molecular classes might contribute to predictive modeling is unknown. Here, we analyzed a large multi-omic database (proteomics, metabolomics, and lipidomics, over 9,000 circulating variables) to identify prognostic biomarkers for adverse outcomes, including mortality and slow recovery, in severely injured patients. Methods: Admission plasma samples from patients (n = 148) enrolled in the PAMPer (Prehospital Plasma during Air Medical Transport in Trauma Patients at Risk for Hemorrhagic Shock) trial were analyzed using mass-spectrometry (metabolomics and lipidomics) and aptamer-based (proteomics) assays. Patients were stratified by 30-day survival or prolonged ICU LOS (>7 days defined as slow resolving). Biomarkers were selected via LASSO-regression modeling and machine-learning analysis. Results: Admission biomarkers derived from all three classes of molecules (n = 9 out of 9000) contributed to the best predictive model for 30-day survival, while a set of 5 proteins yielded the best prognostic model for slow resolution of critical illness (p < 0.5 with FDR < 5%). Both models output had mean AUC’s and accuracies above 70%. Correlation network analysis revealed multiple direct positive associations (r>0.8) between selected features within each model (from online GO database). Conclusion: These findings indicate that multi-omic analyses are likely to be useful to the identification of early prognostic biomarkers in trauma and that biomarkers are likely to differ based on the outcome. High dimensional proteomics may be especially useful for predicting recovery from critical illness.