Diagnostic accuracy of blood biomarkers for excluding head CT abnormalities after mild traumatic brain injury in adults: a systematic review and meta-analysis

Résumé fourni par la source

Traumatic brain injury (TBI) is a major cause of mortality and long-term disability, and mild TBI accounts for more than 90% of hospital-presenting cases. Diagnostic management of mild TBI relies on clinical assessment and neuroimaging, with non-contrast head computed tomography (CT) representing the diagnostic standard for detecting acute intracranial abnormalities. Nevertheless, approximately 90% of CT scans performed in mild TBI show no intracranial injury. Blood biomarkers have emerged as objective measures of brain injury for identifying patients at low probability of CT-detectable injury. We conducted a systematic review and diagnostic test accuracy meta-analysis to evaluate blood biomarkers measured using automated analytical platforms for excluding CT-detectable intracranial abnormalities in adults with mild TBI. The review was conducted according to PRISMA-DTA guidelines and registered in PROSPERO (CRD420251087456). PubMed/MEDLINE, Embase, Web of Science, Cochrane Library, Scopus, Science Citation Index and Google Scholar were searched for studies published from January 2010 to December 2025. Eligible studies enrolled adults with mild TBI (Glasgow Coma Scale score 13-15), biomarkers measured within 24 h of injury using commercially available automated assays and non-contrast head CT used as the reference standard. Methodological quality was assessed using QUADAS-2. Pooled sensitivity and specificity were estimated using bivariate random-effects models, and heterogeneity was explored using meta-regression. Thirty studies including more than 15,000 patients were included. Eligible evidence was available only for S100B, glial fibrillary acidic protein (GFAP), ubiquitin C-terminal hydrolase-L1 (UCH-L1), and combined biomarkers GFAP and UCH-L1. No eligible studies evaluating neuron-specific enolase, neurofilament light chain or other biomarkers met the predefined criteria. S100B showed high pooled sensitivity of 92% (95% CI 89-94%) but low specificity of 31% (95% CI 27-35%). GFAP alone showed pooled sensitivity of 89% (95% CI 83-93%) and specificity of 37% (95% CI 27-48%), whereas UCH-L1 alone showed lower sensitivity of 72% (95% CI 62-80%) and specificity of 53% (95% CI 47-59%). The combination of GFAP and UCH-L1 demonstrated the highest pooled sensitivity of 94% (95% CI 92-95%) but lowest specificity of 28% (95% CI 23-33%). Specificity showed substantial heterogeneity across biomarkers. Meta-regression suggested that earlier sampling improved S100B sensitivity, while age was associated with specificity for the combined biomarkers GFAP and UCH-L1. Sensitivity analyses excluding high risk of bias studies did not substantially change the pooled estimates, and Deeks' funnel plot did not indicate significant small-study effects. Blood biomarkers measured using automated platforms showed clinically meaningful rule-out potential for CT-detectable intracranial abnormalities in adults with mild TBI. Specifically, the results of our diagnostic test accuracy meta-analysis support the clinical utility of S100B and the combination of GFAP and UCH-L1 as adjunctive triage tools in acute care settings for identifying adult patients with mild TBI (GCS 13-15) and low probability head CT abnormalities. Their safest and most effective application is likely to be within validated clinical algorithms that account for patient characteristics, timing of sampling, assay-specific cutoffs, high-risk clinical features and local acute care practice.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.