Systematic review and cumulative meta-analysis of the diagnostic accuracy of glial fibrillary acidic protein vs. S100 calcium binding protein B as blood biomarkers in observational studies of patients with mild or moderate acute traumatic brain injury

• Published in: Diagnosis (Berlin, Germany) Vol. 9; no. 1; pp. 18 - 27
• Main Authors: Mastandrea, Paolo, Mengozzi, Silvia, Bernardini, Sergio
• Format: Journal Article
• Language: English
• Published: Germany De Gruyter 01.02.2022
• Subjects: Biomarkers; Brain Injuries - diagnosis; Brain Injuries, Traumatic - diagnosis; cumulative meta-analysis; diagnostic odds ratio; diagnostic test accuracy; Glial Fibrillary Acidic Protein; Humans; Middle Aged; mild traumatic brain injury; prediction interval; risk of bias; S100 calcium binding protein B; S100 Calcium Binding Protein beta Subunit; test sensitivity; test specificity; diagnostic odds ratio; prediction interval; test sensitivity; S100 calcium binding protein B; risk of bias; diagnostic test accuracy; cumulative meta-analysis; glial fibrillary acidic protein; mild traumatic brain injury; test specificity
• ISSN: 2194-8011; 2194-802X; 2194-802X
• DOI: 10.1515/dx-2021-0006

Traumatic brain injuries (TBIs) and sports-related concussions (SRCs) are the leading causes of hospitalization and death in subjects <45 years old in the USA and Europe. Some biomarkers (BMs) have been used to reduce unnecessary cranial computed tomography (CCT). In recent years, the astroglial S100 calcium-binding B protein (S100B) has prevented approximately 30% of unnecessary CCTs. Glial fibrillary acidic protein (GFAP) has also been studied in direct comparison with S100B. The aim of our cumulative meta-analysis (cMA) is to compare – in the context of hospital emergency departments or SRC conditions – the differences in diagnostic accuracy (DA), sensitivity (Se) and specificity (Sp) of GFAP and S100B. The main cMA inclusion criterion was the assessment of both BMs in the included subjects since 2010, with blood samples drawn 1–30 h from the suspected TBI or SRC. The risk-of-bias (RoB) score was determined, and both the publication bias (with the Begg, Egger and Duval trim-and-fill tests) and sensitivity (with the box-and-whiskers plot) were analyzed for outliers. Seven studies with 899 subjects and nine observations (samples) were included. The diagnostic odds ratios (dORs) with their prediction intervals (PIs), Se and Sp (analyzed with a hierarchical model to respect the binomial data structure) were assessed, and a random-effects MA and a cMA of the difference in the BMs dOR natural logarithms (logOR(G-S)) between the BMs were performed. The cMA of dOR(G-S) was significant (5.78 (CI 2–16.6)) probably preventing approximately 50% of unnecessary CCTs. Further work is needed to standardize and harmonize GFAP laboratory methods.