TY - JOUR
T1 - Fluid-Based Protein Biomarkers in Traumatic Brain Injury
T2 - The View from the Bedside
AU - Agoston, Denes V.
AU - Helmy, Adel
N1 - Publisher Copyright:
© 2023 by the authors.
PY - 2023/11
Y1 - 2023/11
N2 - There has been an explosion of research into biofluid (blood, cerebrospinal fluid, CSF)-based protein biomarkers in traumatic brain injury (TBI) over the past decade. The availability of very large datasets, such as CENTRE-TBI and TRACK-TBI, allows for correlation of blood- and CSF-based molecular (protein), radiological (structural) and clinical (physiological) marker data to adverse clinical outcomes. The quality of a given biomarker has often been framed in relation to the predictive power on the outcome quantified from the area under the Receiver Operating Characteristic (ROC) curve. However, this does not in itself provide clinical utility but reflects a statistical association in any given population between one or more variables and clinical outcome. It is not currently established how to incorporate and integrate biofluid-based biomarker data into patient management because there is no standardized role for such data in clinical decision making. We review the current status of biomarker research and discuss how we can integrate existing markers into current clinical practice and what additional biomarkers do we need to improve diagnoses and to guide therapy and to assess treatment efficacy. Furthermore, we argue for employing machine learning (ML) capabilities to integrate the protein biomarker data with other established, routinely used clinical diagnostic tools, to provide the clinician with actionable information to guide medical intervention.
AB - There has been an explosion of research into biofluid (blood, cerebrospinal fluid, CSF)-based protein biomarkers in traumatic brain injury (TBI) over the past decade. The availability of very large datasets, such as CENTRE-TBI and TRACK-TBI, allows for correlation of blood- and CSF-based molecular (protein), radiological (structural) and clinical (physiological) marker data to adverse clinical outcomes. The quality of a given biomarker has often been framed in relation to the predictive power on the outcome quantified from the area under the Receiver Operating Characteristic (ROC) curve. However, this does not in itself provide clinical utility but reflects a statistical association in any given population between one or more variables and clinical outcome. It is not currently established how to incorporate and integrate biofluid-based biomarker data into patient management because there is no standardized role for such data in clinical decision making. We review the current status of biomarker research and discuss how we can integrate existing markers into current clinical practice and what additional biomarkers do we need to improve diagnoses and to guide therapy and to assess treatment efficacy. Furthermore, we argue for employing machine learning (ML) capabilities to integrate the protein biomarker data with other established, routinely used clinical diagnostic tools, to provide the clinician with actionable information to guide medical intervention.
KW - biomarkers
KW - brain
KW - clinical
KW - injury
KW - mechanistic
KW - phenotypes
KW - utility
UR - http://www.scopus.com/inward/record.url?scp=85177747641&partnerID=8YFLogxK
U2 - 10.3390/ijms242216267
DO - 10.3390/ijms242216267
M3 - Review article
C2 - 38003454
AN - SCOPUS:85177747641
SN - 1661-6596
VL - 24
JO - International Journal of Molecular Sciences
JF - International Journal of Molecular Sciences
IS - 22
M1 - 16267
ER -