Serum lipidomics identifies outcome signatures in patients with traumatic brain injury, presenting with Glasgow Coma Scale score of 13-15

Patients presenting with a Glasgow Coma Scale (GCS) score of 13–15, conventionally labelled mild traumatic brain injury (mTBI), account for most hospital-presenting TBI; a substantial proportion have not recovered to pre-injury health at six months, fewer than one in ten receives structured follow-up, and at presentation they are largely indistinguishable from those who recover. Prognostic models based on clinical variables discriminate poorly here, and established protein biomarkers add little. Lipidomics reflects membrane turnover and repair rather than injury magnitude, offering a potentially complementary perspective.

Blood collected within 24 h of injury from 929 patients with mTBI in the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) cohort was processed to serum and analysed by Ultra-high performance liquid chromatography – Quadrupole-Time-of-flight mass spectrometry (UHPLC-QTOF). After quality-control, 278 lipid features were retained and clustered by Gaussian mixture modelling. Associations between lipid features, six protein biomarkers, Corticosteroid Randomisation After Significant Head injury (CRASH) clinical variables and 6-month Glasgow Outcome Scale–Extended (GOSE) outcome were assessed by Spearman correlation and Jonckheere–Terpstra trend tests. Variables were selected by ordinal elastic-net and modelled by proportional odds logistic regression, comparing CRASH-only with CRASH-plus-marker models, built in a development set (n = 411 with complete covariates) and assessed in a test set (n = 367).

Lipid clusters rich in lysophosphatidylcholines (LPCs), sphingomyelins/ceramides and ether-linked phosphatidylcholines (PC-O) correlated positively with favourable outcome and inversely with the injury proteins. Summed choline-containing lipid intensity rose across GOSE (p < 0.0001). Elastic-net selected three proteins (GFAP, NfL, S100B) and seven lipid features, three annotated (LPC(18:2), LPC(20:5), PC(O-34:3)). Adding lipids and proteins increased the area under the receiver operating characteristic curve (AUC) at every threshold, reaching nominal significance at GOSE ≥ 7 (AUC 0.728, 95% CI 0.671–0.785, p = 0.022), not surviving correction for four comparisons. The averaged c-index rose from 0.720 to 0.751 and Nagelkerke R ² from 0.125 to 0.242. The three proteins and three annotated lipids showed opposing trends across GOSE (all p < 0.001).

The acute serum lipidome is associated with 6-month functional outcome after mTBI. LPCs and PC-Os moved opposite to the injury proteins and may index membrane repair rather than injury burden, offering pathophysiological insight and possible therapeutic targets. Adding them to clinical variables gave modest gains, greatest at the upper end of the outcome scale, where existing models perform worse. The gain is small and exploratory; replication with targeted assays is needed first.