Multi-analyte Serum Biomarker Panels Improve Specificity For Accurate TBI Diagnosis in the Acute Phase After Injury.

Timothy Van Meter1, Nazanin Mirshahi1, Hayley Falk2, Ramon Diaz-Arrastia3, Vani Rao4, Haris Sair5, W. Frank Peacock6, and Frederick Korley2,7

1Program for Neurological Diseases, ImmunArray, Inc., Richmond, VA. 2Department of Emergency Medicine, Johns Hopkins School of Medicine, Baltimore, MD, 3Department of Neurology, University of Pennsylvania, Philadelphia, PA. 4 Departments of Psychiatry, and 5Radiology, Johns Hopkins University School of Medicine, Baltimore, MD. 6Department of Emergency Medicine, Baylor College of Medicine, Houston, TX. 7Department of Emergency Medicine, University of Michigan, Ann Arbor, MI.

References Cited1. Korley FK, Kelen GD, Jones CM, Diaz-Arrastia R. Emergency Department Evaluation of Traumatic Brain Injury in the United States, 2009-2010. J. Head Trauma Rehabil. (Sept 2015, Epub ahead of print) 2. Yang J, Korley FK, et al. Serum neurogranin measurement as a biomarker of acute traumatic brain injury. Clinical Biochemistry. 48(13-14):843-8, 2015. 3. Korley FK, Diaz-Arrastia R, et al. Circulating Brain-Derived Neurotrophic Factor Has Diagnostic and Prognostic Value in Traumatic Brain Injury. J Neurotrauma. 33(2):215-25, 2016. 4. Zafonte RD et al. Effect of citicoline on functional and cognitive status among patients with traumatic brain injury: Citicoline Brain Injury Treatment Trial (COBRIT). JAMA. 308(19):1993-2000, 2012.

Table 3: Selected panels of 3 marker combinations were analyzed in R using multivariate logistic regression.

Introduction• Head injury brings nearly 5 million patients into emergency departments per year in the US1. The majority of patients receiving a CT (90%)

have a negative CT result. Structural MRI scanning reveals structural abnormalities in up to one third of those patients, and advancedneuroimaging methods, such as diffusion tensor imaging (DTI), detects abnormalities in an even larger fraction. There is a great need toidentify TBI in patients using objective laboratory tests, as these patients are at risk for persistent post-concussive symptoms that may affecttheir overall quality of life. These patients could be eligible for clinical trials with novel therapies. Blood based diagnostic tests measuringchanges in physiological levels of circulating biomarkers may aid in identifying patients at risk for long-term symptoms of TBI, and allowstratification of patients for more effective treatment planning.

• Several protein biomarkers discovered in serum and CSF have been detected in TBI, including Brain Derived Neurotrophic Factor (BDNF),Neurogranin (NRGN) and Glial Fibrillary Acidic Protein (GFAP), as well as other proteins and their break-down products2,3. However, none ofthe antigens tested to date have been useful as single biomarkers to help confirm a diagnosis of brain injury in ACRM+ patients. The currentstudy evaluated 6 brain-specific serum protein biomarkers, tested individually and in combination, to diagnose brain injury in mild tomoderately injured patients.

Results• No single protein analyte demonstrated sufficient combined

sensitivity and specificity for clinical utility (i.e., where both >0.9).However, significant improvements in the combined sensitivity andspecificity for the classification of TBI patients as brain-injured areobserved with panels of least three analytes.

• Using ROC analysis, several multi-analyte panels demonstrated areasunder the curve (AUCs) greater than 0.95, with sensitivities andspecificities of over 90%. (e.g., BDNF, GFAP and SNCB).

• These results were obtained using multivariate logistic regressionanalysis (model building with 100 bootstrap replicates, in RStudioversion 0.99.896, pROC package, Table 3).

Figure 3. Comparison of Results from Biomarker Assays for ACRM+ TBI Patients and Healthy Controls. Box plot analyses show the datadistribution for individual serum biomarkers in TBI patients (ACRM+) and healthy controls (HC). Wilcoxon Rank Sum tests were used tocompare the medians between TBI and healthy control groups, for GFAP (p<0.0001), BDNF (p=0.757), and SNCB (p<0.0001). Outliers wereremoved from the analysis (Less than 5 values removed overall; defined as values > 150% of the interquartile distance).

Discussion• An objective biomarker-based test for diagnosis of brain injury requires a combination of both high sensitivity and specificity for the

successful identification of brain-injured patients who are at risk following suspected brain injury. Three-biomarker panels were able toclassify patients as brain-injured with high sensitivity and specificity (> 0.90).

• Additional cohorts and time points post-injury should be studied to ascertain the optimal sampling time post-injury for each biomarkerstudied. These studies are ongoing in our laboratories in multiple patient cohorts.

• Further validation of these biomarker assay panels is warranted, in order to evaluate whether testing can provide benefit to clinical decisionmaking and patient stratification. Studies currently underway, highlighted below in Future Directions, show the potential clinical utility ofImmunArray biomarkers to predict recovery and to stratify patients. Use of algorithms that consider clinical symptoms and biomarker datatogether, particularly neuropsychological symptoms, may better predict long term disability due to brain injury.

MethodsSamples:

Biomarker assays were performed on 249 brain-injured patients from 2 independent studies and compared with 250 healthy control serumsamples. All studies were conducted under IRB-approved study protocols at each respective institution. HeadSMART, a prospective studybeing conducted at Johns Hopkins University, was the main TBI cohort used in the study. Median baseline blood draw was 4.2 hours frominjury. Subsequently, patients were evaluated at 7 additional time points out to 6 months post injury. Additional samples (n=35) wereobtained from the COBRIT study4. Healthy controls (n= 250) were recruited from Baylor College of Medicine (patients not presenting forhealth assessment of any kind, recruited from companions of ED patients). For TBI patients clinical data was compiled, including detailedneurocognitive and neuroimaging results, consistent with NIH Common Data Elements (CDE). Patient demographics are described in Tables 1and 2. All TBI patients used in the study met ACRM criteria for TBI diagnosis (215/249 patients). ACRM+ TBI is defined as: An alteration inbrain function, or other evidence of brain pathology caused by an external force. Including 1 of the following clinical signs: (i) loss ofconsciousness; (ii) any loss of memory for events immediately before (retrograde amnesia) or after the injury (PTA); (iii) neurologic deficits(weakness, loss of balance, vision, paralysis, sensory loss, aphasia, etc.); (iv) any alteration in mental state (injury related AMS).

Biomarker Assays: Serum biomarker concentrations for Brain-Derived Neurotrophic Factor (BDNF), Glial Fibrillary Acidic Protein (GFAP), Metallothionein 3 (MT3),

Neurogranin (NRGN), Neuron Specific Enolase (NSE), and Beta-Synuclein (SNCB) were assessed in duplicate tests using high sensitivitysandwich ELISA tests across replicate assays. Detection technologies were either Meso Scale Discovery (MSD) electro-chemiluminescence orperoxidase-mediated colorimetric detection with 3,3’,5,5’-tetramethylbenzidine (TMB). Biomarker values were subjected to statistical(multivariate logistic regression, performed in R).

CT positive (42) CT negative (207) All subjects (249)

Age 54.8 ± 17.9 48.7 ± 21.1 49 ± 20.7

Gender

Female 15 (35.7%) 92 (44.4%) 107 (43%)

Male 27 (64.3%) 115 (55.6%) 142 (57%)

Race

African American 15 (35.7%) 81 (39.1%) 96 (38.6%)

Caucasian 23 (54.8%) 118 (57%) 141 (56.6%)

Other 4 (9.5%) 8 (3.9%) 12 (4.8%)

Mechanism

Assault 12 (28.6%) 24 (11.6%) 36 (14.4%)

Fall from elevation ( > 3ft or >5 stairs)

8 (19%) 42 (20.3%) 50 (20.1%)

Motorized cycle 1 (2.4%) 6 (2.9%) 48 (19.3%)

MVC patient not ejected 4 (9.5%) 40 (19.3%) 3 (1.2%)

Other fall 10 (23.8%) 49 (23.7%) 59 (23.7%)

Pedal cycle 1 (2.4%) 3 (1.4%) 4 (1.6%)

Pedestrian Struck by motor vehicle 6 (14.3%) 32 (15.5%) 38 (15.3%)

Struck by/against 0 (0%) 11 (5.3%) 11 (4.4%)

Table 1. HeadSMART Study Patient Demographics and Clinical Data.

CT positive (42) CT negative (207) All subjects (249)

Loss of Consciousness (LOC)

No 16 (38.1%) 105 (50.7%) 121 (48.6%)

Not Sure 0 (0%) 3 (1.5%) 3 (1.2%)

Yes 26 (61.9%) 99 (47.8%) 12 (50.2%)

Amnesia

No 15 (35.7%) 114 (55.1%) 129 (51.8%)

Yes 27 (64.3%) 93 (44.9%) 120 (48.2%)

Prior Concussion

None 35 (83.3%) 147 (71%) 182 (73.1%)

One or more 7 (16.7%) 60 (29%) 67 (26.9%)

Altered Mental Status (AMS)

No 15 (35.7%) 108 (52.2%) 123 (49.4%)

Yes 27 (64.3%) 99 (47.8%) 126 (50.6%)

GCS (based on 246 out of 249 total cases)

15 29 (70.7%) 188 (91.7%) 217 (88.2%)

14 5 (12.2%) 13 (6.3%) 18 (7.3%)

13 2 (4.9%) 3 (1.5%) 5 (2%)

3-12 5 (12.2%) 1 (0.5%) 6 (2.5%)

All subjects (250)

Age

36.13 ± 11.52

Gender

Female 160 (64%)

Male 90 (36%)

Race

African American 72 (28.8%)

Caucasian 66 (26.4%)

Other 112 (44.8%)

Table 3: ACRM+ TBI Vs. Healthy Control

Biomarkers AUC Sensitivity Specificity# of HC

# of ACRM+

NRGN, GFAP, BDNF 0.9373 0.919 0.918 200 173

NSE, GFAP, SNCB 0.9945 0.982 0.970 199 144

MT3, BDNF, SNCB 0.9608 0.895 0.950 199 114

MT3, GFAP, SNCB 0.9869 0.970 0.956 199 114

BDNF, GFAP, SNCB 0.9904 0.976 0.962 199 172

Future Directions: Use of Blood Biomarkers to Predict Recovery

TBI: The Continuum of Injury and Recovery

First 30 Days (approx.)Neurological dysfunction occurs with initial insult and as the

injured brain rebuilds itself.

Beyond 30 Days (approx.)Neurological dysfunction may re-emerge with the potential

induction of autoantibodies against circulating signaling molecules and brain structures.

Figure 2. Conceptual View of the TBI Continuum of Care. Following brain injury, short term damage may be observed in the first few days followinginitial injury. Patient samples from the first blood draw post injury (HeadSMART, median 4.2 hours post-injury; range of 1.5 -24 hours post injury)were analyzed by ELISA for independent detection of 6 biomarkers, of the more than 20 that are being developed by ImmunArray, Inc.

Figure 1. Comparison of Trial Designs and Time of Blood Sampling from TBI Cohorts Used in the Study.

Serum Biomarker Trial Design

0 4 24 72 168 1 month 3 month 6 month

HeadSMART COBRIT X

Time point: (hour/month)

6 M o n t h S t u d y T i m e C o u r s e

Time 0 Day 7 Day 30 Year 30

Acute Event

Long-term Damage/ImpactShort-term Damage

Day 1 Circuit dismantling and acute cell injury and death

Axonal injury and degeneration Vascular injury and inflammation

Analytes/Autoantigens Autoantibodies

> Day 30 Auto-antibody attachmentDay 7-30 Circuit reorganizationAxonal regeneration and remodeling

Vascular reestablishmentAutoimmunity develops

Table 2. Healthy Control Demographics.

TBI Biomarkers Correlate with Outcome Measures: Glasgow Outcome Scale- Extended (A-B).

Serum TBI Biomarkers and neuropsychological symptoms may predict recovery 6 months after brain injury (C-E).

• Upper panels: A number of outcome metrics, including Glasgow Outcome Scale-Extended (GOSE), have been examined to better understand therelationships between brain injury biomarkers detectable in serum and differences in patient recovery. For example, BDNF levels are lowerimmediately after injury in patients with incomplete recovery (GOSE <8) at 1 month (A; p=0.028) and 6 months (B; p= 0.036). Comparison of medianlevels was performed using Wilcoxon Rank Sum Tests. Brain image courtesy of www.braininjuryhub.co.uk.

• Lower panels: Several of ImmunArray’s biomarker assays show promise in predicting global outcomes in patient clusters based on neuropsychologicalsymptoms during the acute phase after injury. Incomplete recovery at 6 months (GOSE) is associated with lower serum levels of MT3, BDNF, andSNCB when detected immediately after injury. Significant differences were found in patients stratified according to having a history of depression (C),LOC at injury and history of depression (D), or LOC and memory deficit (E). (p<0.05; Wilcoxon Rank Sum Tests). Use of these biomarkers in panels canadditionally enhance the overall diagnostic and prognostic performance of these tests.

A B

C D E

RecoveryInjury

Blood draw