A
A
Neurofilament Light Protein: a marker for injury severity, clinical course and outcome following moderate-severe Traumatic Brain InjurySpyridoulaTsetsou1,2
LeahE.McQuillan3
DavidJ.Barton4
JohnWilliamson5,6
DamonG.Lamb5,7
FirasKobeissy8
GuangzhengCai8
RichardRubenstein10
KevinK.W.Wang8,9
ClaudiaS.Robertson
MD
2✉Emailclaudiar@bcm.eduAmyK.Wagner3,11,13
1Department of NeurologyBaylor College of MedicineHoustonTXUSA
2Department of NeurosurgeryBaylor College of MedicineOne Baylor Plaza770030HoustonTXUSA
3Department of Physical Medicine and RehabilitationUniversity of PittsburghPittsburghPAUSA
4Department of Emergency MedicineUniversity of Pittsburgh Medical CenterPittsburghPAUSA
5Department of PsychiatryUniversity of FloridaGainesvilleFLUSA
6
A
Brain Rehabilitation Research CenterMalcom Randall VAMCGainesvilleFL7Department of Neuroscience and Biomedical EngineeringUniversity of FloridaGainesvilleFLUSA
8Department of Neurobiology, Neuroscience InstituteMorehouse School of MedicineAtlantaGAUSA
9Center for Visual and Neurocognitive RehabilitationAtlanta VA Health Care SystemDecaturGAUSA
10Department of NeurologySUNY Downstate Health Sciences UniversityBrooklynNYUSA
11Department of NeuroscienceUniversity of PittsburghPittsburghPAUSA
12Safar Center for Resuscitation ResearchUniversity of PittsburghPittsburghPAUSA
13Clinical and Translational Science InstituteUniversity of PittsburghPittsburghPAUSA
Spyridoula Tsetsou1,2, Leah E. McQuillan3, David J. Barton4, John Williamson5,6, Damon G. Lamb5–7, Firas Kobeissy8, Guangzheng Cai8, Richard Rubenstein10, Kevin K. W. Wang8,9, Claudia S. Robertson2, Amy K. Wagner3,11–13
Affiliations:
1. Department of Neurology, Baylor College of Medicine, Houston, TX, USA
2. Department of Neurosurgery, Baylor College of Medicine, Houston, TX, USA
3. Department of Physical Medicine and Rehabilitation, University of Pittsburgh, Pittsburgh, PA, USA
4. Department of Emergency Medicine, University of Pittsburgh Medical Center, Pittsburgh, PA, USA
5. Department of Psychiatry, University of Florida, Gainesville, FL, USA
6. Brain Rehabilitation Research Center, Malcom Randall VAMC, Gainesville FL
7. Department of Neuroscience and Biomedical Engineering, University of Florida, Gainesville, FL, USA
8. Department of Neurobiology, Neuroscience Institute, Morehouse School of Medicine, Atlanta, GA, USA
9. Center for Visual and Neurocognitive Rehabilitation, Atlanta VA Health Care System, Decatur, GA, USA
10. Department of Neurology, SUNY Downstate Health Sciences University, Brooklyn, NY, USA
11. Department of Neuroscience, University of Pittsburgh, Pittsburgh, PA, USA
12. Safar Center for Resuscitation Research, University of Pittsburgh, Pittsburgh, PA, USA
13. Clinical and Translational Science Institute, University of Pittsburgh, Pittsburgh, PA, USA
Keywords:
traumatic brain injury
critical illness
intracranial pressure
cerebral perfusion pressure
brain tissue oxygenation
neurofilament light
Word count: 4571
No. of figures and tables
7
Corresponding Authorship
Claudia S. Robertson, MD
Department of Neurosurgery, Baylor College of Medicine
One Baylor Plaza, Houston, TX, 770030, USA
Email: claudiar@bcm.edu
A
Authors contributions:
Conceptualization: ST, CR, LM, KKW, AKW
Data collection: CR, GC, KKW
Data visualization, analysis, and interpretation: ST, LM, CR, DB, KKW, AKW
Manuscript writing: ST, LM, CR, AKW
Manuscript review and critical revision: CR, DB, JW, DL, FK, GC, RR, KKW, AKW
A
Funding:
CR, JW, RR, KKW AKW
A
A
All coauthors have seen and agree with the contents of the manuscript; the manuscript complies with all instructions to authors; requirements for authorship have been met; each author believes that the manuscript represents honest work; the STROBE checklist was used. This work has not been published elsewhere and is not under consideration by another journal and is not redundant with previous work. We used archived blood samples that were originally collected as part of our clinical trial of erythropoietin (clinicaltrials.gov NCT00313716). Patients were offered the option to have research blood samples banked for future studies analyses after clinical trial analysis were completion. Baylor College of Medicine Institutional Review Board approved waived consent for the use of these archived, de-identified blood samples (protocol H-44131).Funding
This project was funded by Department of Defense grant W81XWH-19-2-0012 (samples analysis, data analysis) and Department of Defense grant Gap-Based Milieu Biomarkers for Traumatic Brain Injury (GAMBIT- TBI) W81XWH2211089 (data analysis, manuscript preparation). The clinical trial of erythropoietin that originally collected the samples used in this study was pre-registered at clinicaltrials.gov (NCT00313716). Further support received from VA research grant IK2RX002490 and the Brain Rehabilitation Research Center, Malcom Randall VAMCGainesville FL.
A
Disclosures:KKW is a share-holder of Gryphon Bio, Inc. The other authors have no relevant disclosures related to this work.
Abstract (300 words)
Background
Recent studies demonstrate increased Neurofilament Light (NfL) levels across all types of Traumatic Brain Injury (TBI). We aimed to evaluate the relationship between acute temporal NfL with ICU physiology and outcome.
Methods
Stored serum from 97 patients with severe TBI were analyzed. Samples were collected every 6h for the first 24h and then daily through day 10. Temporal NfL patterns were examined using group-based trajectory (TRAJ) modeling, and means were calculated over day 0, 1–5, 6–10 and 0–10. ICU injury burden scores were generated using ridge regression anchored to Glasgow Outcome Scale-Extended (GOSE) over the same time-bins. We used both univariable and multivariable regression to assess NfL TRAJ and means with respect to ICU injury burden, demographic data, clinical factors, and outcome. NfL TRAJ and day 0–10 ICU injury burden were added to a baseline model to evaluate outcome discrimination. Mediation analysis was applied to assess the mediation effect of ICU injury burden between NfL and GOSE.
Results
Patients demonstrated increasing levels over the monitoring period. Sixty-three patients had low NfL temporal profile (low-TRAJ), and 34 patients had higher profile (high-TRAJ). Individuals in the high-TRAJ group had lower GCS (p = 0.0009), more frequently required barbiturate coma (p = 0.01) or decompressive surgery (p = 0.02) and had worse 6-months outcome (p = 0.001). NfL means and ICU injury burden both negatively correlated with GCS. Multivariate logistic regression showed a 2.24X increased odds of unfavorable GOSE per unit increase in ICU injury burden (p = 0.0447). When modeled with ICU injury burden, the NfL high-TRAJ group relationship to outcome was attenuated, with ICU burden being a significant mediator of the NfL relationship to GOSE.
Conclusion
Our results suggest that ICU injury burden contributes to serum NfL associations with GOSE. Temporal NfL profiles provide novel information about intracranial pathophysiology as a causal mediator of axonal injury burden following severe TBI.
A
A
A
A
Introduction
Traumatic brain injury (TBI) remains a leading cause of death and major disability, with increasing incidence and health-care costs1–3. The discovery of blood-based biomarkers, as potential prognostic, monitoring and diagnostic tools, has led to increasing interest and research in this field4–7. Glial fibrillary acidic protein (GFAP) and ubiquitin C-terminal hydrolase L1 protein (UCH-L1) levels in blood during the acute phase of TBI, have been approved as an aid in the decision to obtain Computed Tomography (CT) imaging among individuals clinically assessed as having mild TBI8,9. The neurotrauma field has had a rising interest in exploring the utility of neurofilament light (NfL), a marker showing high sensitivity for axonal injury that maps to subacute/chronic brain injury and MRI brain atrophy after moderate/severe TBI (msTBI) 10,11. However, less is known about the utility this marker may have in reflecting ongoing secondary brain injury among critically ill in the Intensive Care Unit (ICU) with msTBI.
Neurofilament (Nf) proteins are neuron-specific intermediate filaments abundant in axons. During axonal injury, they dissociate from the cytoskeleton, cause cellular membrane integrity loss and accumulate in extracellular fluid, cerebrospinal fluid (CSF) and blood. They are highly specific to axonal injury12,13. NfL, Nf-heavy (NfH) and neurofilament-medium form the three subunits of Nf proteins. NfL is considered to be a biomarker reflecting neurodegeneration and is associated with cognitive impairment in patients with mild cognitive impairment and dementia (Alzheimer’s disease, frontotemporal dementia)14–17. NfL remains high in CSF and serum following TBI10,11,18 and is associated with progressive neurodegeneration in this population19. Prior work20 involving patients with msTBI explored the potential for Nf proteins to predict 6-months cognitive outcome in this patient group. Interestingly, the NfL levels were measured later post-injury (16–90 days post injury) did not correlate with the early NfL levels (0–15 days post injury) or 6-months cognitive performance20.
However, NfL has been linked with 6- or 12-months functional outcome, in all severities of TBI10,21. Unlike other biomarkers, blood NfL levels continue to rise up to 6 weeks post msTBI22, with the peak NfL level linked to poor 6-months functional outcome and to brain atrophy on MRI22. In related work23, we recently conducted NfL temporal kinetic profile characterizations from the time of injury through subacute and chronic recovery to find that serum NfL level peak ~ 30 days post-injury; also, both acute and chronic serum NfL levels are associated with global recovery at 6- and 12-months post-injury.
We previously screened multiple serum biomarkers, for associations with clinical physiological variables measured over time in the acute injury phase24. This work highlighted that GFAP levels in the first 10 days were associated with intracranial pressure (ICP) and cerebral perfusion pressure (CPP) summary variables, injury type presentation on CT scans, mortality and recovery, and demonstrated rapidly declining GFAP levels after initial peak correlated with secondary insults. Also, averaged NfL serum levels in the first 10 days were associated with acute physiological parameters, specifically with brain hypoxia (longer duration of PbtO2 < 10 mmHg) and ICP treatment24. However, the role of NfL within the ICU clinical course remains poorly understood. Evidence supporting its utility in ICU-based clinical decision-making is limited, and the potential causal relationship between ICU injury burden profiles and outcomes has not been explored.
Thus, we evaluated how acute phase, serial NfL levels correlate with the ICU clinical course and outcomes in patients with msTBI. We hypothesized that NfL levels reflect axonal injury sensitive to the extent of secondary brain injury, as indicated by ICU-associated clinical data, and 6-months global outcomes.
Methods:
Study design and participants
A
We used archived blood samples that were originally collected as part of our clinical trial of erythropoietin (clinicaltrials.gov NCT00313716). Inclusion criteria for this study were: 1) age at least 15 years old, 2) motor GCS ≤ 5 after resuscitation related to a closed TBI and 3) ability to enroll within 6 hours from trauma. Exclusion criteria included 1) patients with GCS of 3 with fixed and dilated pupils, 2) penetrating trauma, 3) pregnancy (positive urine test) and 4) life-threatening injuries or severe pre-existing comorbidities. Patients enrolled in the trial were randomly assigned to treatment with erythropoietin or placebo and to a transfusion threshold of 7 or 10 g/dL using a 2x2 factorial design. Blood samples were collected every 6 hours for the first 24 hours after injury and then once daily through day 10 post-injury. Samples were centrifuged for 15 minutes at 1500g, and serum was removed and stored at -80o C until analyzed. Patients were offered the option to have research blood samples banked for future studies analyses after clinical trial analysis were completion. Of the 200 trial participants, 114 opted to allow future use of the samples and their associated clinical data. Samples were available on 103 patients but only 97 had at least two sample biomarker measurements available during the biosample collection period, and included in this analysis (Fig. 1A).Fig. 1
A) Consort diagram of patients included for analysis. B) Serum NfL trajectory
groups. Two subgroups with high (n = 34) vs low (n = 63) NfL levels over time. Both groups
exhibited increasing levels over the 10-day monitoring period. Data shown are mean ± standard
deviation
Baylor College of Medicine Institutional Review Board approved waived consent for the use of these archived, de-identified blood samples (protocol H-44131).
Biomarker assays
As described in prior work24, we assayed NfL with an ultrasensitive immunoassay using digital array technology (Single Molecule Arrays SiMoA)-based Human Neurology 4-Plex B assay (N4PB; Item 103345). All SiMoA assays were run on the SR-X benchtop assay platform (Quanterix Corp., Lexington MA) at the University of Florida (Gainesville, FL) according to manufacturer’s instructions. The lower limit of quantification (LLOQ), lower limit of detection (LLOD), and dynamic range were 0.625 pg/mL, 0.097 pg/mL, and 0.0971-10,000 pg/mL for NfL, respectively. Inter-assay and intra-assay % CV were 4.6–4.9% and 3.5–7.5%, respectively for NfL.
Outcome measures and other ICU variables
Demographics and injury severity variables included age, sex, race, GCS sum and motor component (pre- and post-resuscitation), injury severity scale (ISS), clinical exam including pupillary reactivity, prehospital hypotension and hypoxia, and acute CT findings (initial and worst Marshall CT category and a CT burden score). International Mission for Prognosis and Analysis of Clinical Trials in TBI (IMPACT) risk of poor outcome and mortality were also calculated25. Glasgow Outcome Scale-Extended (GOSE) was used to determine 6-months outcome. For these analyses, GOSE was dichotomized as favorable (score 5–8) and unfavorable outcome (score 1–4).
CT reports for the admission scan were reviewed by neurointensivist and/or neurosurgeon attendings and categorized for pre-specified brain lesion types: intraventricular hemorrhage (IVH), intraparenchymal hemorrhage (IPH), subdural hematoma (SDH), subarachnoid hemorrhage (SAH), epidural hematoma (EDH) or diffuse brain injury (DBI), contusion and midline shift.
Details about the acute care, clinical events and treatment were extracted from the electronic medical record (EMR), prepared, and summarized for data analysis. Data including ICP, mean arterial pressure (MAP), CPP, jugular venous oxygen saturation (SjvO2), and brain tissue pO2 (PbtO2) were collected and analyzed. The physiological variables were recorded hourly for as long as the ICP was being monitored over the first 10 days of injury. We used average ICP, duration of time that ICP was higher than 25 mmHg and also higher than 30 mmHg, duration of time of CPP was lower than 50 mmHg, duration of time that PbtO2 was less than 10 mmHg, and duration of time that SjvO2 less than 50%. Average values of other parameters and duration of time of abnormal values were collected as well. Treatments to control ICP (e.g. mannitol, barbiturate coma, decompressive craniectomy) were also collected and analyzed. Sustained increases in ICP were treated according to local protocols and the three-tiered algorithm, including but not limited to sedative medications, CSF drainage, hyperventilation, hyperosmotic therapy, barbiturate induced coma, surgical decompression. Physiological variables data and clinical events were binned into 24-h increments to examine changes over time.
ICU injury burden score
The ICU injury burden score was calculated as described in our prior work24. To calculate the ICU injury burden score, the physiological data were also binned into 24-h increments, and the following conditions known to be associated with poor outcome were quantified: duration of time that ICP > 25 mmHg, CPP < 50 mmHg, and PbtO2 < 10 mmHg; mannitol dose administration, implementation of barbiturate coma and/or decompressive craniectomy. Frequency of out-of-range values was evaluated over the full 10-day time course post-injury and in three separate time epochs: the first 24 h, days 1–5, and days 6–10.
Given the inter-related nature of acute care physiology data, and inherent collinearity associated with these multiple ICU factors included in the modeling approach, we generated a composite score reflecting cumulative ICU related physiological injury burden rather than assessing individual factors with respect to biomarker levels and long-term outcome. Generating an aggregate ICU injury burden score through data reduction techniques reduces noise and provides a cumulative measure of relevant injury factors that reflect ICU practice patterns over time. A series of simple logistic regressions were conducted considering ICU injury factors over three-time epochs (first 24h-day 0, days 1–5, and days 6–10) with respect to 6-months GOSE and 6-months mortality.
ICU factors associated with outcome with a threshold p-value of < 0.1 were identified using unadjusted linear regression and included in the respective ICU injury burden score derivation. A ridge regression that included each significant ICU factor stated above was then applied to create the ICU injury burden scores for each respective time epoch. This method yielded β-weight estimates for each factor in the model, with smaller β-weights reflecting smaller effect sizes. Appropriate to our research question and analytical approach, this method also penalizes multicollinearity when present, which leads to more reliable β-weight estimates and reduced standard error (SE). This penalty term reduces the impact of multicollinearity by shrinking correlated predictors’ coefficients toward zero in a balanced way, thereby producing more stable and interpretable estimates. The most stable and frequent ridge penalty parameter was selected after iterating a sixfold cross validation 1000 times using the R package glmnet26. The coefficients generated were then multiplied to each individual’s set of relevant ICU injury factors and summed to produce the ICU injury burden score in each time epoch.
CT burden score
An admission CT injury burden score was also generated using ridge regression as described previously24. CT reports for the admission scan were reviewed and categorized as mentioned above. The injury types associated with GOSE in univariate logistic regression (midline shift, SAH, SDH, EDH, IPH, IVH, DBI and contusion) were selected for inclusion in the weighted sum reflecting CT injury burden.
Group-based trajectory analysis
We previously reported temporal profiles for days 0–10 NfL23 using group-based trajectory (TRAJ) analysis27 for subjects with ≥ 2 sample timepoints in the first 10 days post-injury. Polynomial orders were assigned to group data from each cohort based on minimizing Bayesian Information Criterion (BIC) and ensuring posterior probability > 80% using CNORM distribution (SAS software, Proc TRAJ). These TRAJ group memberships were leveraged in this study for further analysis of relationships to ICU injury burden score, demographic and clinical factors, and global outcome (GOSE).
Statistical analysis
Descriptive analyses for continuous variables used mean, standard error (SE) of the mean, median, and interquartile range (IQR). Frequencies and percentages were used to quantify categorical variables. Spearman rank correlation tests were applied to compare continuous variables that were not normally distributed. Univariate and multivariate linear and binary logistic regression models were used, where applicable, to assess associations between NfL means, NfL TRAJ group membership, ICU injury burden score, and demographic and clinical covariates with 6-months outcomes (GOSE). ICU injury burden score was evaluated, using the Baron and Kenny approach28 for a possible mediating relationship between NfL mean levels—and also TRAJ group membership—and 6-month outcome. The area under the receiver operating characteristic curve (AUC) was used to evaluate the discriminative performance of each model. DeLong’s test was employed to compare the AUCs between models, assessing whether differences in model performance were significant. Multivariate models assessing ICU injury burden as a mediator were adjusted for age and sex. Results were reported when the total effect was greater than the direct effect and a p-value threshold less than 0.1 was met. Mediation percentage was calculated as the proportion of the indirect effect to the total effect, shown below. Statistical analyses were conducted using SAS Version 9.4 (Cary, North Carolina), R and GraphPad Prism Version 10 (Boston, Massachusetts). For all analyses, a p-value ≤ 0.05 was considered statistically significant.
Results
Demographics, clinical, ICU variables and outcomes by NfL TRAJ
Ninety-seven (97) patients had samples and data available and were used for analysis. TRAJ analysis identified two distinct subgroups with high (n = 34) vs low (n = 63) NfL levels; both groups exhibited increasing levels over the 10-day monitoring period (Fig. 1B). Table 1A shows the demographics, clinical and radiological characteristics of the two groups. The high TRAJ group had worse admission GCS (p = 0.0009), higher CT burden scores (p = 0.0086), more frequently compressed basal cisterns on admission brain CT (p = 0.03), and prehospital hypotension (p = 0.046). Higher IMPACT score for risk of poor outcome and risk of mortality was also observed for those in high TRAJ group (p < 0.0001, p < 0.0001 respectively). Similarly, high TRAJ group membership was associated with an increased mortality rate (p = 0.001) and worse clinical outcome with GOSE assessment at 6 months (p = 0.001).
Table 1B shows the ICU physiological variables for patients of the two TRAJ groups. Average ICP was higher for in the high TRAJ group (p = 0.01), and this group had also longer duration of time of ICP > 25mmHg (p = 0.01) and > 30mmHg (p = 0.0069). High TRAJ group members also was associated with more prolonged cerebral hypoperfusion (7.58hrs vs 2.88hrs, p = 0.01), though with only trend-level differences observed in the average CPP value. Patients in the high TRAJ group more frequently required barbiturate coma (p = 0.01) or decompressive surgery (p = 0.02) for controlling ICP.
Cross-sectional NfL and ICU injury burden relationships
Day 0, Day 1–5, Day 6–10, and Day 0–10 means for CSF and serum NfL were all screened against their respective ICU injury burden by time epoch (not shown). Serum NfL, but not CSF, levels were associated with ICU injury burden at all time points.
Figure 2A presents a cross-leg framework of univariate linear regression models that show the consistent interrelatedness between serum NfL and ICU injury burden factors over the first 10 days post-injury. Day 0 NfL means were significantly associated with Day 1–5 NfL means, and subsequently Day 1–5 NfL means to Day 6–10 NfL means (p < 0.001 both comparisons). ICU injury burden demonstrated a similar relationship between adjacent time epochs (p < 0.001 both comparisons). Additionally, mean NfL and ICU injury burden scores were significantly related to each other within time epoch (Day 0 NfL to Day 0 ICU injury burden) and across time epochs (day 0 to Day 1–5, and Day 1–5 to Day 6–10).
Fig. 2
A) Cross-Leg regression schematic using standardized biomarker means and 6-
month GOSE anchored ICU injury burden score. Each arrow represents an unadjusted linear
regression with the corresponding beta-value and p-value. Bolded p-values signify significance
lower than p < 0.05 threshold. Factors included in each ICU injury burden anchored to GOSE, by
time epoch are listed below:
•Day 0: Duration (h) CPP < 50, Duration (h) PbtO2 < 10, Duration (h) ICP > 25, mannitol
dose required, and decompressive craniectomy and/or barbiturate coma dose required.
•Day 1–5: Duration (h) PbtO2 < 10, and decompressive craniectomy and/or barbiturate coma
dose required.
•Day 6–10: Mannitol dose required, and decompressive craniectomy and/or barbiturate
coma dose required.
NfL, ICU injury burden, demographic and clinical factors relationship
A correlational heatmap (Fig. 2B) presents pairwise comparisons (by Spearman rank correlation) between NfL means and ICU injury burden scores at all time epochs (Days: 0, 1–5, 6–10, 0–10) with GCS, age and CT burden. Each cell shows the correlation coefficient (r) for the pairwise comparison. Mean NfL levels were positively associated with ICU injury burden and CT burden, and negatively associated with GCS. ICU injury burden was also negatively associated with GCS and positively associated with CT burden at all time epochs. Age had a stronger relationship to ICU injury burden at Day 0 and 1–5 than later timepoints. NfL means and ICU injury burden scores were strongly associated levels and scores at other time epochs.
NfL means at all time points were associated, in univariate regression, with 6-months GOSE, with higher NfL levels being associated with unfavorable outcome (Table 2A). When adjusting for the ICU injury burden at the corresponding time point (Table 2B), the NfL mean relationships to outcome were variably attenuated, while ICU injury burden scores were significantly related to outcome (p < 0.05 all comparisons). Here, Day 0 mean NfL was no longer significant in relationship to GOSE when Day 0 ICU injury burden was introduced (p = 0.2497); however, Day 1–5 and Day 6–10 remained significant (p = 0.0362 and p = 0.0223, respectively). Together, the discriminative performance improved when ICU injury burden score was added to outcome models that included mean NfL levels. Each time epoch model showed a significant improvement in AUC (NfL only model vs. NfL + ICU injury burden model) when comparing ROC curves using Delong’s Test (p < 0.05 all comparisons).
In preparation for multivariate model development, demographic (age, sex) and clinical injury factors (ER GCS, best in 24h GCS, CT burden) were investigated for NfL TRAJ and mean NfL measured over different time bins, with ICU injury burden and global outcome modeled as dependent variables (Table 3). Older age was associated with higher Day 0 ICU injury burden and increased odds of unfavorable outcome. Injury factors (ER and best in 24h GCS, CT burden) were associated with both NfL TRAJ and unfavorable outcomes. Those with more severe injuries had an increased odds of having high NfL TRAJ profiles over the first 10 days post-injury and unfavorable 6-months outcome. CT burden was the factor most strongly associated with all ICU injury burden time epochs (p < 0.0001 all comparisons).
Serum NfL TRAJ GROUP | |||
|---|---|---|---|
Low NfL | High NfL | p-value | |
Number of patients: | 63 | 34 | |
Sex - Male - Female | - 51 - 12 | - 28 - 6 | 0.86 |
Age (mean) | - 32.3 | 34.6 | 0.6884 |
Admission Motor GCS - 1–3 - 4–6 | - 17 (27%) - 46 (73%) | - 22 (64%) - 12 (36%) | 0.0005 |
Admission Sum GCS - 3–5 - 6–8 - > 8 | - 15 (24%) - 30 (48%) - 18 (28%) | - 21 (61%) - 7 (21%) - 6 (18%) | 0.0009 |
GCS (best in 24h), median [IQR] | - 6 [4–7] | - 4.5 [4–7] | 0.0646 |
Pupils: - Both reactive - One reactive - Neither reactive | - 46 (73%) - 6 (10%) - 11 (17%) | - 17 (50%) - 6 (18%) - 11 (32%) | 0.76 |
Pre hospital hypotension - Absent - Present | - 59 (94%) - 4 (6%) | - 27 (80%) - 7 (20%) | 0.046 |
Pre hospital Hypoxia - Absent - Present | - 51 (81%) - 12 (19%) | - 22 (65%) - 12 (35%) | 0.08 |
Admission Marshall CT - Diffuse injury 2 - Diffuse injury 3 - Evacuated Mass Lesion - Unevacuated Mass Lesion | - 31 - 14 - 16 - 2 | - 8 - 13 - 12 - 1 | 0.09 |
Subarachnoid Hemorrhage - Present - Absent | - 41 (65%) - 22 (35%) | - 27 (79%) - 7 (21%) | 0.16 |
Basal Cisterns - Normal - Compressed/absent | - 37 (59%) - 26 (41%) | - 12 (35%) - 22 (65%) | 0.03 |
Midline shift - None - < 5 mm - 5–15 mm - > 15 mm | - 44 (70%) - 8 (13%) - 9 (14%) - 2 (3%) | - 16 (47%) - 6 (18%) - 10 (29%) - 2 (6%) | 0.15 |
Diffuse Brain Injury (DBI) - Present - Absent | - 45 (71.4%) - 18 (28.6%) | - 22 (64.7%) - 12 (35.3%) | 0.4943 |
CT Burden, median [IQR] | 0.51 [0.5–1.62] | 1.62 [0.87–1.97] | 0.0086 |
Ventricles - Normal - Small - Enlarged | - 29 (46%) - 33 (52%) - 1 (2%) | - 12 (35%) - 22 (65%) - 0 (0%) | 0.53 |
IMPACT risk of poor outcome (mean) | 0.33 | 0.58 | < 0.0001 |
IMPACT risk of mortality (mean) | 0.20 | 0.39 | < 0.0001 |
GOSE at 6 mo - GR/MD - SD/V/D - Missing/lost to f/u | - 30 (48%) - 30 (48%) - 3 (4%) | - 4 (12%) - 26 (76%) - 4 (12%) | 0.001 |
Mortality at 6 mo - Survived - Died - Missing/lost to f/u | - 57 (90%) - 3 (4%) - 3 (4%) | - 21 (62%) - 9 (26%) - 4 (12%) | 0.001 |
| Bold denotes p ≤ 0.05; italic denotes p ≤ 0.1. GCS: Glasgow Coma Scale. GOSE GR/MD/SD/V/D: Good Recovery/Moderate Disability/Severe Disability/Vegetative State/Dead. IQR: interquartile range. IMPACT: International Mission for Prognosis and Analysis of Clinical Trials in TBI. | |||
(A) NfL Mean → 6-month GOSE outcome (individual regressions) | ||||
|---|---|---|---|---|
OR | CI | p-value | AUC | |
Serum NfL Mean (day 0) | 1.007 | 1.001–1.014 | 0.0326 | 0.646 |
Serum NfL Mean (days 1–5) | 1.005 | 1.001–1.008 | 0.0108 | 0.660 |
Serum NfL Mean (days 6–10) | 1.002 | 1.00-1.004 | 0.0111 | 0.664 |
(B) NfL Mean (Adjusted for ICU injury burden score) → 6-month GOSE outcome | ||||
OR | CI | p-value | AUC | |
Serum NfL (day 0) | 1.004 | 0.99–1.01 | 0.2497 | 0.779 |
ICU injury burden (day 0) | 3.418 | 1.72–6.79 | 0.0004 | |
OR | CI | p-value | AUC | |
Serum NfL (days 1–5) | 1.004 | 1.0-1.07 | 0.0362 | 0.733 |
ICU injury burden (days 1–5) | 3.814 | 1.35–10.75 | 0.0013 | |
OR | CI | p-value | AUC | |
Serum NfL (days 6–10) | 1.002 | 1.0-1.004 | 0.0223 | 0.770 |
ICU injury burden (days 6–10) | 2.97 | 1.25–7.07 | 0.0141 | |
| Each row represents and individual univariate binary logistic regression in A) with the corresponding odds ratio (OR), 95% confidence interval (CI), p-value, and area under the curve (AUC). Each pair of NfL and ICU injury burden by time epoch represent an individual multivariate binary logistic regression in B) with the corresponding OR, CI, p-value, and AUCs. Model results odds of an unfavorable GOSE = 1–4. Bold denotes p ≤ 0.05. | ||||
A) 6-month GOSE (dichotomized) | B) Day 0–10 NfL TRAJ | C) Day 0 ICU injury burden | D) Day 1–5 ICU injury burden | E) Day 6–10 ICU injury burden | F) Day 0–10 ICU injury burden | |
|---|---|---|---|---|---|---|
OR [CI] p-value | OR [CI] p-value | B (SE) p-value | B (SE) p-value | B (SE) p-value | B (SE) p-value | |
Age | 1.034 [1.0-1.07] p = 0.0382 | 1.012 [0.98–1.04] p = 0.4316 | 0.01 (0.009) p = 0.0889 | 0.0038 (0.004) p = 0.2896 | 0.0001 (0.004) p = 0.9860 | 0.001 (0.005) p = 0.8994 |
Sex | 0.826 [0.29–2.34] p = 0.7179 | 1.098 [0.37–3.24] p = 0.8656 | -0.54 (0.35) p = 0.1259 | -0.088 (0.13) p = 0.4979 | -0.15 (0.05) p = 0.3212 | -0.19 (0.19) p = 0.3334 |
ER GCS Sum | 0.74 [0.63–0.87] p = 0.0003 | 0.785 [0.663–0.93] p = 0.0049 | -0.09 (0.05) p = 0.0454 | -0.008 (0.02) p = 0.6365 | -0.03 (0.02) p = 0.1274 | -0.04 (0.03) p = 0.1681 |
GCS (best in 24h) | 0.725 [0.58–0.9] p = 0.0037 | 0.779 [0.62–0.98] p = 0.0357 | -0.05 (0.06) p = 0.4336 | -0.01 (0.02) p = 0.6485 | -0.01 (0.03) p = 0.7074 | -0.03 (0.04) p = 0.4540 |
CT burden | 3.73 [1.99–6.98] p = < 0.0001 | 2.24 [1.19–4.24] p = 0.0130 | 0.72 (0.16) p < 0.0001 | 0.34 (0.06) p < 0.0001 | 0.42 (0.06) p < 0.0001 | 0.56 (0.08) p < 0.0001 |
| Day 0 ICU injury burden factors: Duration (h) CPP < 50, Duration (h) PbtO2 < 10, Duration (h) ICP > 25, mannitol dose required, and decompressive craniectomy and/or barbiturate coma dose required | ||||||
| Day 1–5 ICU injury burden factor: duration (h) PbtO2 < 10, decompressive craniectomy and/or barbiturate coma dose required. | ||||||
| Day 6–10 ICU injury burden factors: mannitol dose required, decompressive craniectomy and/or barbiturate coma dose required | ||||||
| Day 0–10 ICU injury burden factors: duration (h) CPP < 50, Duration (h) PbtO2 < 10, and decompressive craniectomy and/or barbiturate coma dose required. | ||||||
| Bold denotes p ≤ 0.05; italic denotes p ≤ 0.1. GCS: Glasgow Coma Scale. CT: Computed Tomography. ICU: Intensive Care Unit. OR: Odds Ratio. CI: Confidence interval. SE: Standard Error | ||||||
6-month GOSE (dichotomized) | A) Baseline Characteristic Model (AUC = 0.787) | B) NfL Risk Model (AUC = 0.803) | C) NfL Risk + ICU injury burden Model (AUC = 0.819) | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
Independent variables | OR | CI | p-value | OR | CI | p-value | OR | CI | p-value | |
Age | 1.01 | 0.97–1.05 | 0.5852 | 1.008 | 0.97–1.05 | 0.6828 | 1.08 | 0.97–1.06 | 0.4267 | |
Gender (Male vs. Female) | 1.38 | 0.39–4.87 | 0.6217 | 1.303 | 0.36–4.68 | 0.6853 | 1.10 | 0.28–4.33 | 0.8921 | |
GCS, best in 24h | 0.74 | 0.58–0.93 | 0.0104 | 0.74 | 0.57–0.96 | 0.0241 | 0.69 | 0.58–0.92 | 0.0106 | |
CT burden score | 3.60 | 1.76–7.35 | 0.0004 | 3.02 | 1.42–6.42 | 0.0040 | 2.04 | 0.87–4.78 | 0.1007 | |
NfL risk TRAJ | - | - | - | 2.57 | 0.89–7.42 | 0.0823 | 1.95 | 0.65–5.92 | 0.2361 | |
Overall Day 0–10 ICU injury burden score (GOSE specific) | - | - | - | - | - | - | 2.24 | 1.02–4.92 | 0.0447 | |
| Bold denotes p ≤ 0.05. Italic denotes p ≤ 0.1. GCS: Glasgow Coma Scale. CT: Computed Tomography. ICU: Intensive Care Unit. OR: Odds Ratio. CI: Confidence interval. | ||||||||||
NfL relationship to outcome in the context of ICU clinical course and demographic factors
Multivariate logistic regression models were developed to examine relationships between the pattern of serum NfL levels over time and outcome, after adjusting for other injury severity variables known to be associated with outcome (age, sex, best in 24hr GCS and CT burden score). The baseline characteristic model is shown in Table 4A, with NfL TRAJ and ICU injury burden added incrementally to demonstrate the relative influence of model factors and the change in outcome discrimination by AUC (from 0.787 to 0.803 and 0.819, respectively). Here, lower GCS (p = 0.0104) and greater CT burden (p = 0.004) were associated with increased odds of unfavorable GOSE (Table 4A). When adding Day 0–10 NfL TRAJ to the model, these relationships were attenuated (GCS: p = 0.0241, CT burden score: p = 0.004) yet remained significant. There was a trend level association with high NfL TRAJ membership being associated with 2.57 increased odds of unfavorable GOSE versus the low TRAJ membership (Table 4B). Lastly, when factoring in Day 0–10 ICU burden into the model (Table 4C), the CT burden and NfL TRAJ relationships were fully attenuated (p > 0.1), and the GCS association with outcome was attenuated to a lesser degree, remaining significantly associated with outcome (p = 0.0106). For every unit increase in Day 0–10 ICU burden, there was 2.24 increased odds of unfavorable GOSE (p = 0.0447). The AUC increased with the addition of NfL TRAJ and ICU injury burden; however, pairwise comparisons of ROC curves using DeLong’s test revealed no statistically significant differences in AUC between models (all p-values > 0.23), indicating similar discriminative performance.
The dynamic role of ICU burden as a mediator of NfL associations over time with 6-months outcome was tested (Fig. 3) while adjusting for age and sex as covariates. The mediating effects of ICU burden on day 0 mean and day 1–5 mean NfL relationship to GOSE scores were tested subsequent ICU burden time epochs (Fig. 3A), and Day 6–10 ICU burden was assessed as a mediator for the day 1–5 NfL mean association with outcome (Fig. 3B). Day 1–5 ICU injury burden mediated 28% of the day 0 NfL mean association with unfavorable GOSE. Similarly, Day 6–10 ICU burden mediated 39.9% of the day 1–5 NfL mean association with unfavorable GOSE; however, the direct effect remained trend-level (p = 0.0861). Figure 3C demonstrates Day 0–10 ICU injury burden as a mediator of the NfL TRAJ relationship to GOSE. Here, the overall burden score mediated 28.9% of the NfL TRAJ association with GOSE resulting in a trend-level direct effect (p = 0.0501). Together these models suggest that ICU injury burden is a partial mediator of the NfL biomarker associations with global outcome.
Fig. 3
A) Mediation model of Day 0 serum NfL mean and Day 1–5 ICU injury burden, B)
Day 1–5 serum NfL mean and Day 6–10 ICU injury burden, and C) Serum Day 0–10 NfL
TRAJ and Day 0–10 ICU injury burden to dichotomized 6-month unfavorable GOSE (GOSE = 1–4 vs. GOSE = 5–8). Covariates included age and sex for all individual models. B(SE) or
OR[CI] and p-values are shown for each model. The model labeled direct includes ICU injury
burden in the model to test for the mediation effect. Mediation percentage is shown. P < 0.05 are
bolded.
Discussion
This study highlights temporal and overall associations between NfL and ICU physiological parameters over the first 10 days after msTBI, and their collective impact on global outcome, while considering initial brain injury. The main strengths of the study include the serial biomarker measurements over the first 10 days post-injury along with the detailed clinical, ICU and radiological variables available for analyses.
Previous studies have examined serial NfL measurements post-TBI, correlating increasing NfL serum or CSF levels over time with worse clinical outcome. Nimer et al29 studied 182 patients with all TBI severities and collected 1 to 3 samples per patient over the course of 2 weeks. NfL levels showed an upward trend over time and correlated to outcome. Shahim et al10, evaluated 72 severe TBI patients where daily NfL measurements were obtained from day 0 to 12 post-injury. NfL levels were increased at admission, continued to rise daily, and also were associated with clinical outcome. In a longer time-course study we showed that serum NfL levels rise from the acute phase and peak at 30 days post-injury and both acute serum and CSF NfL levels discriminated 6- and 12-months outcome23. While that study data included a portion of this cohort23, we similarly highlight that serum NfL levels keep rising in the acute phase post-TBI. This linear increase in NfL likely reflects ongoing axonal injury after the initial insult and the importance of serial measurements assessing longitudinal biomarker patterns rather than single time point measurement strategies.
The TRAJ analysis approach27 demonstrated temporal NfL dynamics by identifying two groups of individuals with unique serum NfL pattern over the first 10 days post-injury. A major advantage of TRAJ is that it is a clustering algorithm for assessing longitudinal patterns rather than single-point, averaged, or peak biomarker level for diagnostic or prognostic models. Its capacity for patient clustering for potential TBI patient subtype identification has been demonstrated for a variety of CNS derived23,30–33 and systemic biomarkers34–38. Validation studies are required to address the minimum time needed for patient assigned to a TRAJ group in real time based on their initial NfL levels; however, this methodology, and the dynamic relationships between NFL and ICU burden, shows potential utility for early clinical and prognostic management.
NfL levels demonstrated a linear increase over the first 10 days post-injury, contrary to GFAP, which presented with an early peak followed by variable decay patterns over the acute period24. Rising NfL elevations over time likely reflect higher degree of evolving secondary axonal injury, as NfL means are closely associated with ICU injury burden over the full 10-day time course post-injury. The regression-based cross lag analysis showed mean NfL and ICU injury burden scores were significantly related to each other within time epoch from day 0 (Day 0 NfL to Day 0 ICU injury burden) and across subsequent time epochs (day 0 to Day 1–5, and Day 1–5 to Day 6–10). In contrast, Day 0 GFAP wasn’t associated with Day 0 ICU injury burden when using a similar analysis construct, highlighting the strong correlation of NfL with initial injury severity.
NfL levels were strongly correlated with GCS and CT burden. NfL means also positively correlated with both initial GCS, and best in 24h GCS, across all time epochs, with CT burden scores primarily associated with day 0 and day 1–5 NfL levels. Similarly, higher NfL TRAJ groups were significantly associated with worse GCS and CT burden. This contrasts with the GFAP analysis24, where the CT characteristics but not the initial GCS and baseline injury severity factors were associated to the GFAP TRAJ group. Notably, other studies have shown GFAP levels within the first 24 hours correlated with admission GCS39. Both GFAP and NfL rise with aging in the general population40, 41,42. We found no age differences between low and high TRAJ groups. This work contrasts with our GFAP findings, where individuals in the high TRAJ GFAP group were significantly older24.
To further evaluate the NfL association with ICU factors and outcome, age and sex were introduced as covariates in mediation analysis. GCS and CT burden scores were excluded due to associations between NfL TRAJ, ICU injury burden and GOSE, allowing NfL to be treated as the biological readout of initial injury severity and allowing it to be effectively modeled as a temporally dynamic variable in pathway analysis without the absence of confounding covariates with shared variance. Mediation analyses suggested that the ICU injury burden is a partial mediator, indicating that NfL biomarker associations with functional outcome occur via secondary CNS physiological injury response in response to neurocritical illness.
Taken together, these data suggest NfL is a robust blood-based biomarker representation of initial neurological injury severity, and interestingly, tracks closely with ICU factor time course. Independent of these correlative factors, higher NfL TRAJ and mean NfL levels measurement over the 10-day time epochs were each significantly associated with worse outcome. When introducing ICU injury burden into the modeling framework, these relationships were attenuated, suggesting ICU burden is in the causal pathway explaining a significant amount of the relationship between acute NfL levels and global outcome. These findings align with prior studies that have confirmed elevated NfL after a TBI of any severity is associated with degree of axonal injury11,43, and thus worse outcome at 6 months; however, the fact that acute ICU appears to have a causal effect on outcome suggests that there may be modifiable targets within ICU care best practice pathways where serial NfL monitoring may be a useful adjunct to clinical management and assessing clinical treatment effectiveness.
To the best of our knowledge this is the first time NfL has been linked with acute ICU physiological parameters and acute treatment. Patients in the high TRAJ group had more frequent refractory intracranial hypertension and cerebral hypoperfusion requiring advanced and invasive treatments. These findings build on previous work linking NfL to axonal injury and outcome post-TBI by providing an additional lens through ICU-related physiology to present opportunities for acute management and treatment.
Limitations
Our study has some limitations. Our sample size is relatively small, representing a single Level 1 trauma center’s experience. The clinical information available was exported from EMR and collected prospectively as part of a previous treatment trial. We did not assess the impact of each ICP treatment component on NfL levels, as in clinical practice multiple treatment measures are done simultaneously. Future larger, multicenter trials are necessary to validate the NfL role in acute ICU setting. Samples were only collected daily up to day 10, with no chronic samples available. However, for the purposes of this study, having serial samples in the first 10 days of ICU stay was sufficient to show the importance of longitudinal biomarker patterns in their relationship with ICU metrics over time. This work presents novel approaches to handling CT injury profiles and longitudinal ICU factors in time-specific load score derivations. However, the granular time course data and the presence of multicollinearity among NfL, clinical factors, and outcome limit the effectiveness of traditional regression approaches and may necessitate more sophisticated analytical techniques (e.g. machine learning) in larger cohorts to establish NFL as a biomarker for real-time clinical decision making. Future research with more frequent serial measurements over a longer time course, with multiple biomarker inclusion and advanced analysis planning, would further refine the role of NfL in individualized ICU care and prognostication, and provide treatment trial readouts for monitoring effectiveness.
Conclusion
Serum NfL levels correlate with worse clinical outcomes after msTBI and are associated with ICU clinical course. The linear increase of NfL levels and strong correlation with ICU burden and 6-months outcome provide information about the association with secondary brain injury following msTBI. Future research using a high-resolution sampling approach in assessing extended longitudinal patterns, as well as multi-marker panels for treatment triaging and clinical outcome is necessary in preparation for large scale multisite trials focused on assessing complex biological patterns and relationships to CNS physiology co-occurring critical illness associated with msTBI.
A
References:
1.
Taylor CA, Bell JM, Breiding MJ, Xu L. Traumatic Brain Injury-Related Emergency Department Visits, Hospitalizations, and Deaths - United States, 2007 and 2013. MMWR Surveill Summ. 2017;66(9):1–16. 10.15585/mmwr.ss6609a1.
2.
Hyder AA, Wunderlich CA, Puvanachandra P, Gururaj G, Kobusingye OC. The impact of traumatic brain injuries: a global perspective. NeuroRehabilitation. 2007;22(5):341 – 53. (https://www.ncbi.nlm.nih.gov/pubmed/18162698).
3.
Maas AIR, Menon DK, Manley GT, et al. Traumatic brain injury: progress and challenges in prevention, clinical care, and research. Lancet Neurol. 2022;21(11):1004–60. 10.1016/S1474-4422(22)00309-X.
4.
Shahim P, Politis A, van der Merwe A, et al. Time course and diagnostic utility of NfL, tau, GFAP, and UCH-L1 in subacute and chronic TBI. Neurology. 2020;95(6):e623–36. 10.1212/WNL.0000000000009985.
5.
Mouhieddine TH, El Houjeiri L, Sabra M, Hayes RL, Mondello S. CNS Trauma Biomarkers and Surrogate Endpoints Pipeline from Bench to Bedside: A Translational Perspective. In: Kobeissy FH, ed. Brain Neurotrauma: Molecular, Neuropsychological, and Rehabilitation Aspects. Boca Raton (FL)2015.
6.
Mondello S, Hayes RL, Biomarkers. Handb Clin Neurol. 2015;127:245–65. 10.1016/B978-0-444-52892-6.00016-7.
7.
Nielson JL, Cooper SR, Yue JK, et al. Uncovering precision phenotype-biomarker associations in traumatic brain injury using topological data analysis. PLoS ONE. 2017;12(3):e0169490. 10.1371/journal.pone.0169490.
8.
Korley FK, Jain S, Sun X, et al. Prognostic value of day-of-injury plasma GFAP and UCH-L1 concentrations for predicting functional recovery after traumatic brain injury in patients from the US TRACK-TBI cohort: an observational cohort study. Lancet Neurol. 2022;21(9):803–13. 10.1016/S1474-4422(22)00256-3.
9.
Bazarian JJ, Biberthaler P, Welch RD, et al. Serum GFAP and UCH-L1 for prediction of absence of intracranial injuries on head CT (ALERT-TBI): a multicentre observational study. Lancet Neurol. 2018;17(9):782–9. 10.1016/S1474-4422(18)30231-X.
10.
Shahim P, Politis A, van der Merwe A, et al. Neurofilament light as a biomarker in traumatic brain injury. Neurology. 2020;95(6):e610–22. 10.1212/WNL.0000000000009983.
11.
Gao W, Zhang Z, Lv X, et al. Neurofilament light chain level in traumatic brain injury: A system review and meta-analysis. Med (Baltim). 2020;99(38):e22363. 10.1097/MD.0000000000022363.
12.
Yuan A, Rao MV, Veeranna, Nixon RA. Neurofilaments and Neurofilament Proteins in Health and Disease. Cold Spring Harb Perspect Biol. 2017;9(4). 10.1101/cshperspect.a018309.
13.
Gaetani L, Blennow K, Calabresi P, Di Filippo M, Parnetti L, Zetterberg H. Neurofilament light chain as a biomarker in neurological disorders. J Neurol Neurosurg Psychiatry. 2019;90(8):870–81. 10.1136/jnnp-2018-320106.
14.
Mattsson N, Cullen NC, Andreasson U, Zetterberg H, Blennow K. Association Between Longitudinal Plasma Neurofilament Light and Neurodegeneration in Patients With Alzheimer Disease. JAMA Neurol. 2019;76(7):791–9. 10.1001/jamaneurol.2019.0765.
15.
Benedet AL, Ashton NJ, Pascoal TA, et al. Plasma neurofilament light associates with Alzheimer's disease metabolic decline in amyloid-positive individuals. Alzheimers Dement (Amst). 2019;11:679–89. 10.1016/j.dadm.2019.08.002.
16.
Rohrer JD, Woollacott IO, Dick KM, et al. Serum neurofilament light chain protein is a measure of disease intensity in frontotemporal dementia. Neurology. 2016;87(13):1329–36. 10.1212/WNL.0000000000003154.
17.
Olsson B, Lautner R, Andreasson U, et al. CSF and blood biomarkers for the diagnosis of Alzheimer's disease: a systematic review and meta-analysis. Lancet Neurol. 2016;15(7):673–84. 10.1016/S1474-4422(16)00070-3.
18.
Ljungqvist J, Zetterberg H, Mitsis M, Blennow K, Skoglund T. Serum Neurofilament Light Protein as a Marker for Diffuse Axonal Injury: Results from a Case Series Study. J Neurotrauma. 2017;34(5):1124–7. 10.1089/neu.2016.4496.
19.
Shahim P, Pham DL, van der Merwe AJ, et al. Serum NfL and GFAP as biomarkers of progressive neurodegeneration in TBI. Alzheimers Dement. 2024;20(7):4663–76. 10.1002/alz.13898.
20.
Trifilio E, Bottari S, McQuillan LE, et al. Temporal Profile of Serum Neurofilament Light (NF-L) and Heavy (pNF-H) Level Associations With 6-Month Cognitive Performance in Patients With Moderate-Severe Traumatic Brain Injury. J Head Trauma Rehabil. 2024. 10.1097/HTR.0000000000000932.
21.
Shahim P, Gren M, Liman V, et al. Serum neurofilament light protein predicts clinical outcome in traumatic brain injury. Sci Rep. 2016;6:36791. 10.1038/srep36791.
22.
Graham NSN, Zimmerman KA, Moro F, et al. Axonal marker neurofilament light predicts long-term outcomes and progressive neurodegeneration after traumatic brain injury. Sci Transl Med. 2021;13(613):eabg9922. 10.1126/scitranslmed.abg9922.
23.
Wang KKW, Barton DJ, McQuillan LE, et al. Parallel Cerebrospinal Fluid and Serum Temporal Profile Assessment of Axonal Injury Biomarkers Neurofilament-Light Chain and Phosphorylated Neurofilament-Heavy Chain: Associations With Patient Outcome in Moderate-Severe Traumatic Brain Injury. J Neurotrauma. 2024;41(13–14):1609–27. 10.1089/neu.2023.0449.
24.
Robertson CS, Martinez FS, McQuillan LE, et al. Serial Measurements of Serum Glial Fibrillary Acidic Protein in Moderate-Severe Traumatic Brain Injury: Potential Utility in Providing Insights into Secondary Insults and Long-Term Outcome. J Neurotrauma. 2024;41(1–2):73–90. 10.1089/neu.2023.0111.
25.
Steyerberg EW, Mushkudiani N, Perel P, et al. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med. 2008;5(8):e165. 10.1371/journal.pmed.0050165. discussion e165.
26.
Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw. 2010;33(1):1–22. https://www.ncbi.nlm.nih.gov/pubmed/20808728.
27.
Niyonkuru C, Wagner AK, Ozawa H, Amin K, Goyal A, Fabio A. Group-based trajectory analysis applications for prognostic biomarker model development in severe TBI: a practical example. J Neurotrauma. 2013;30(11):938–45. 10.1089/neu.2012.2578.
28.
Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol. 1986;51(6):1173–82. 10.1037//0022-3514.51.6.1173.
29.
Al Nimer F, Thelin E, Nystrom H, et al. Comparative Assessment of the Prognostic Value of Biomarkers in Traumatic Brain Injury Reveals an Independent Role for Serum Levels of Neurofilament Light. PLoS ONE. 2015;10(7):e0132177. 10.1371/journal.pone.0132177.
30.
Rubenstein R, McQuillan L, Wang KKW, et al. Temporal Profiles of P-Tau, T-Tau, and P-Tau:Tau Ratios in Cerebrospinal Fluid and Blood from Moderate-Severe Traumatic Brain Injury Patients and Relationship to 6–12 Month Global Outcomes. J Neurotrauma. 2024;41(3–4):369–92. 10.1089/neu.2022.0479.
31.
Kumar RG, Diamond ML, Boles JA, et al. Acute CSF interleukin-6 trajectories after TBI: associations with neuroinflammation, polytrauma, and outcome. Brain Behav Immun. 2015;45:253–62. 10.1016/j.bbi.2014.12.021.
32.
Wagner AK, Amin KB, Niyonkuru C, et al. CSF Bcl-2 and cytochrome C temporal profiles in outcome prediction for adults with severe TBI. J Cereb Blood Flow Metab. 2011;31(9):1886–96. 10.1038/jcbfm.2011.31.
33.
Goyal A, Failla MD, Niyonkuru C, et al. S100b as a prognostic biomarker in outcome prediction for patients with severe traumatic brain injury. J Neurotrauma. 2013;30(11):946–57. 10.1089/neu.2012.2579.
34.
Munoz MJ, Kumar RG, Oh BM, et al. Cerebrospinal Fluid Cortisol Mediates Brain-Derived Neurotrophic Factor Relationships to Mortality after Severe TBI: A Prospective Cohort Study. Front Mol Neurosci. 2017;10:44. 10.3389/fnmol.2017.00044.
35.
Santarsieri M, Kumar RG, Kochanek PM, Berga S, Wagner AK. Variable neuroendocrine-immune dysfunction in individuals with unfavorable outcome after severe traumatic brain injury. Brain Behav Immun. 2015;45:15–27. 10.1016/j.bbi.2014.09.003.
36.
Santarsieri M, Niyonkuru C, McCullough EH, et al. Cerebrospinal fluid cortisol and progesterone profiles and outcomes prognostication after severe traumatic brain injury. J Neurotrauma. 2014;31(8):699–712. 10.1089/neu.2013.3177.
37.
Barton DJ, Kumar RG, Schuster AA, Juengst SB, Oh BM, Wagner AK. Acute Cortisol Profile Associations With Cognitive Impairment After Severe Traumatic Brain Injury. Neurorehabil Neural Repair. 2021;35(12):1088–99. 10.1177/15459683211048771.
38.
Wagner AK, McCullough EH, Niyonkuru C, et al. Acute serum hormone levels: characterization and prognosis after severe traumatic brain injury. J Neurotrauma. 2011;28(6):871–88. 10.1089/neu.2010.1586.
39.
Okonkwo DO, Puffer RC, Puccio AM, et al. Point-of-Care Platform Blood Biomarker Testing of Glial Fibrillary Acidic Protein versus S100 Calcium-Binding Protein B for Prediction of Traumatic Brain Injuries: A Transforming Research and Clinical Knowledge in Traumatic Brain Injury Study. J Neurotrauma. 2020;37(23):2460–7. 10.1089/neu.2020.7140.
40.
Khalil M, Pirpamer L, Hofer E, et al. Serum neurofilament light levels in normal aging and their association with morphologic brain changes. Nat Commun. 2020;11(1):812. 10.1038/s41467-020-14612-6.
41.
Tybirk L, Hviid CVB, Knudsen CS, Parkner T. Serum GFAP - reference interval and preanalytical properties in Danish adults. Clin Chem Lab Med. 2022;60(11):1830–8. 10.1515/cclm-2022-0646.
42.
Disanto G, Barro C, Benkert P, et al. Serum Neurofilament light: A biomarker of neuronal damage in multiple sclerosis. Ann Neurol. 2017;81(6):857–70. 10.1002/ana.24954.
43.
Shahim P, Zetterberg H, Tegner Y, Blennow K. Serum neurofilament light as a biomarker for mild traumatic brain injury in contact sports. Neurology. 2017;88(19):1788–94. 10.1212/WNL.0000000000003912.
List of captions for figures.
Serum NfL TRAJ GROUP | |||
|---|---|---|---|
Low NfL | High NfL | p-value | |
Number of patients | 63 | 34 | |
Average ICP (mean, mmHg) | 14.75 | 18.65 | 0.01 |
ICP - Never > 25 - > 25mmHg at least once | - 16 (25%) - 47 (75%) | - 10 (29%) - 24 (71%) | 0.81 |
ICP duration > 25 mmHg (h) | 9.66 | 20.94 | 0.01 |
ICP - Never > 30 - > 30 mmHg at least once | - 25 ( 40%) - 38 (60%) | - 15 (44%) - 19 (56%) | 0.14 |
ICP duration > 30 mmHg (h) | 2.49 | 9.94 | 0.0069 |
Required mannitol: - No - Yes | - 24 (38%) - 39 (62%) | - 11 (33%) - 23 (67%) | 0.66 |
Required barbiturate coma: - No - Yes | - 57 (91%) - 6 (9%) | - 24 (71%) - 10 (29%) | 0.01 |
Required decompressive craniectomy - No - Yes | - 46 (73%) - 17 (27%) | - 17 (50%) - 17 (50%) | 0.02 |
Average MAP (mmHg) | 90.65 | 91.38 | 0.60 |
MAP: - Never < 70 mmHg - < 70 mmHg at least once | - 26 (41%) - 37 (59%) | - 8 (24%) - 26 (76%) | 0.11 |
MAP duration < 70 mmHg (h) | 3.4 | 4.7 | 0.34 |
Average CPP (mmHg) | 75.58 | 72.36 | 0.06 |
CPP: - Never < 50mmHg - < 50mmHg at least once | - 23 (37%) - 40 (63%) | - 9 (26%) - 25 (74%) | 0.37 |
CPP duration < 50mmHg (h) | 2.88 | 7.58 | 0.01 |
Average PbtO2 (mmHg) | 33.06 | 26.5 | 0.05 |
PbtO2 : - Never < 10 mmHg - < 10mmHg at least once | - 38 (62%) - 23 (38%) | - 18 (55%) - 15 (45%) | 0.5 |
PbtO2 duration < 10mmHg (mean, hours) | 4.75 (missing 2) | 14.6 (missing 1) | 0.08 |
Average Sjv02 | 76.93 Missing 14 | 74.83 Missing 6 | 0.20 |
SjvO2 - Never < 50% - < 50% at least once | - 38 (78%) - 11 (22%) | - 17 (61%) - 11 (39%) | 0.12 |
SjvO2 duration < 50% (h) | 0.75 | 1.17 | 0.23 |
Delayed hematoma - Absent - Present | - 44 (70%) - 19 (30%) | - 22 (65%) - 12 (35%) | 0.65 |
Worst Marshall CT - Diffuse injury 2 - Diffuse injury 3 - Diffuse injury 4 - Evacuated Mass Lesion - Unevacuated Mass Lesion | - 28 - 14 - 1 - 17 - 3 | - 7 - 13 - 1 - 11 - 2 | 0.0014 |
| Bold denotes p ≤ 0.05; italic denotes p ≤ 0.1.ICP: intracranial pressure. MAP: Mean arterial pressure. CPP: Cerebral perfusion pressure. PbtO2: Brain tissue oxygen partial pressure. SjvO2: Jugular venous oxygen saturation. CT: Computed Tomography | |||