A head-to-head comparison of S100B and GFAP/UCH-L1 for detection of traumatic intracranial lesions in a Scandinavian trauma cohort

Background
Few countries recommend glial fibrillary protein (GFAP) and ubiquitin C-terminal hydrolase-L1 (UCH-L1) as a substitute for S100 astroglial calcium-binding protein B (S100B) in early detection of traumatic intracranial lesions in mild TBI (mTBI). This study aims to evaluate the classification agreement between S100B and GFAP/UCH-L1 in a Scandinavian trauma cohort, to evaluate the performance characteristics of S100B and GFAP/UCH-L1 for detection of traumatic intracranial lesions, and lastly to evaluate the laboratory performance of the GFAP/UCH-L1 assay.

Methods
Prospectively collected data from an unselected cohort of 379 adult trauma patients admitted to a level I trauma center at Aarhus University Hospital, Denmark, were retrospectively analyzed. Analyses were performed in the unselected cohort, in sub-cohort 1 (n = 218) i.e. patients with any evidence of TBI in their chart as well as in sub-cohort 2 (n = 105) i.e. patients with mTBI defined as Glasgow Coma Scale score ≥ 14, an injury severity score ≤ 15, and blood sampling within 6 h or 12 h after trauma. Plasma-samples were used for GFAP/UCH-L1 measurement and serum-samples were used for S100B measurement. Data analysis involved agreement analysis using Cohens kappa and sensitivity, specificity, positive predictive value and negative predictive value for each biomarker in each of the three cohorts. Lastly, levels of GFAP/UCH-L1 measured by the Alinity-I platform and the Simoa platform were compared.

Results
Classification agreement between GFAP/UCH-L1 and S100B was high in all three cohorts, but Cohens kappa improved with increasing proximity to clinical biomarker use and reached an almost perfect identity in sub-cohort 2 (0.70, 95% CI 0.62–0.92). S100b had a sensitivity and negative predictive value of 100% in sub-cohort 2, while GFAP/UCH-L1 reached 100% across all three cohorts. The specificities for both S100B and GFAP/UCH-L1 were relatively low. Comparing GFAP/UCH-L1 levels between platforms revealed a low concordance with the Alinity-I platform measuring GFAP levels on average 65% lower and UCH-L1 levels 84% higher than the Simoa platform.

Conclusions
In this study, S100B and GFAP/UCH-L1 had an almost perfect agreement for classification of mTBI patients and comparable diagnostic performances for detecting traumatic intracranial lesions. Our results therefore support GFAP/UCH-L1 as a feasible alternative to S100B for detecting traumatic intracranial lesions in mTBI.