Monitoring Biomarkers of Cellular Injury and Death in Acute Brain Injury

Abstract

Background

Molecular biomarkers have revolutionalized diagnosis and treatment of many diseases, such as troponin use in myocardial infarction. Urgent need for high-fidelity biomarkers in neurocritical care has resulted in numerous studies reporting potential candidate biomarkers.

Methods

We performed an electronic literature search and systematic review of English language articles on cellular/molecular biomarkers associated with outcome and with disease-specific secondary complications in adult patients with acute ischemic stroke (AIS), intracerebral hemorrhage (ICH), subarachnoid hemorrhage (SAH), traumatic brain injury (TBI), and post-cardiac arrest hypoxic ischemic encephalopathic injuries (HIE).

Results

A total of 135 articles were included. Though a wide variety of potential biomarkers have been identified, only neuron-specific enolase has been validated in large cohorts and shows 100 % specificity for poor outcome prediction in HIE patients not treated with therapeutic hypothermia. There are many promising candidate blood and CSF biomarkers in SAH, AIS, ICH, and TBI, but none yet meets criteria for routine clinical use.

Conclusion

Current studies vary significantly in patient selection, biosample collection/processing, and biomarker measurement protocols, thereby limiting the generalizability of overall results. Future large prospective studies with standardized treatment, biosample collection, and biomarker measurement and validation protocols are necessary to identify high-fidelity biomarkers in neurocritical care.

Introduction

Cellular and molecular biomarkers play important roles in critical care—they may help monitor disease progression, probe underlying physiology, and improve prognostic accuracy. The presence of the blood brain barrier (BBB) and its injury introduce the following unique complexities in the interpretation and the use of molecular biomarkers in neurocritical care:

1. Biomarker origin—whether it is primarily synthesized in the central nervous system (CNS) or elsewhere.

2. The anatomical source of the biosample—whether it is serum, plasma, CSF, urine, or other. Molecular biomarkers have significantly different concentrations, biological half-lives, and time-course variations in different anatomical compartments.

3. The timing of biomarker measurement relative to disease onset—The sensitivity and specificity of biomarkers vary depending on measurement time relative to disease onset.

4. Association with potential injury mechanisms—Whether a molecule is part of a pathogenic pathway or a final result of cellular injury (reversible or irreversible) affects the clinical interpretation of this molecule as a biomarker of disease.

Here, we review current acute brain injury (ABI) biomarker data in light of these special considerations.

Methods

Search Criteria

This systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. Using the PICO approach, we searched PubMed database for English language articles from January 1990 to August 2013 based on the following criteria:

1. Patient population Adult patients ≥18 years of age with acute ischemic stroke (AIS), intracerebral hemorrhage (ICH), spontaneous subarachnoid hemorrhage (SAH), traumatic brain injury (TBI), and post-cardiac arrest (CA) hypoxic ischemic encephalopathic injuries (HIE).

2. Intervention Cellular/molecular biomarkers from biological fluids such as serum, plasma, cerebrospinal fluid (CSF), and urine.

3. Controls Patients without ABI.

4. Outcome endpoints Primary outcomes of interest are mortality and long-term neurological outcome. Secondary outcomes of interest are prediction of disease-specific secondary deteriorations and complications such as hemorrhagic transformation of AIS and delayed cerebral ischemia (DCI) and vasospasm following SAH.

Study Selection and Data Collection

We excluded unpublished data or congress presentations/abstracts, review articles, case reports or case series, studies with sample size ≤10 patients for CSF biomarkers and studies with sample size ≤30 for blood biomarkers, pediatric ICU studies, studies not conducted on ICU patients, and studies dealing with brain death. We also excluded microbiological markers of infection, biomarkers from microdialysis, and metabolites. One hundred and thirty-five studies were included. Both authors using pre-defined criteria systematically extracted data from the included studies, and the levels of evidence were classified and practical recommendations were developed according to the GRADE system. We included 28 studies for HIE following cardiac arrest (Tables 1, 2), 25 for SAH (Table 3), 25 for AIS (Table 4), 29 for ICH (Table 5), and 28 for TBI (Table 6). Available biomarkers that predict overall prognosis and secondary complications in ABI patients were reviewed to address specific endpoints.

Table 1.

Biomarkers for outcome following cardiac arrest without therapeutic hypothermia treatment

Authors/year/Ref	Population	N	Bio-marker	Sample source	Findings
Molecules of CNS origin
Zandbergen/2006/[1]	Post cardiac arrest, unconscious >24 h after CPR	407	NSE, s100β	Serum	100 % of patients with NSE >33 μg/L at any time had a poor outcome (40 % PPV; 0 % FPR) s100β >0.7 μg/L at 24–72 h post cardiac arrest predicts poor outcome (47 % PPV; 2 % FPR) Performance of clinical tests was inferior to SSEP and NSE in predicting outcome
Meynaar/2003/[3]	Post cardiac arrest, comatose post CPR	110	NSE	Serum	NSE at 24 and 48 h after CPR was significantly higher in patients who did not regain consciousness vs. those who did No one with NSE >25 μg/L at any time regained consciousness (100 % specificity)
Pfeifer/2005/[2]	Post cardiac arrest within 12 h of ROSC, survived >48 h	97	NSE, s100β	Serum	NSE >65 μg/L predicted increased risk of death and persistent vegetative state at 28 days post CPR (97 % PPV) s100β >1.5 μg/L predicts poor outcome (96 % PPV)
Rosen/2001/[6]	Out-of-hospital cardiac arrest	66	s100β, NSE	Serum	s100β >0.217 μg/L and NSE >23.2 μg/L at 2 days post cardiac arrest predicted poor 1-year outcome (100 % PPV)
Bottiger/2001/[8]	Non-traumatic out-of-hospital cardiac arrest	66	s100β	Serum	Significant differences in s100β level between survivors and non-survivors after cardiac arrest were observed from 30 min to 7 days post cardiac arrest s100β >1.10 μg/L at 48 h post cardiac arrest predicted brain damage (100 % specificity)
Martens/1998/[4]	Post cardiac arrest, unconscious and ventilated for >24 h	64	NSE, s100β	Serum	s100β and NSE were significantly higher in patients who did not regain consciousness compared to those who did s100β >0.7 μg/L is a predictor of not regaining consciousness after cardiac arrest (95 % PPV; 96 % specificity) NSE >20 μg/L predicted poor outcome (51 % sensitivity; 89 % specificity)
Hachimi-Idrissi/2002/[9]	Post cardiac arrest	58	s100β	Serum	s100β >0.7 μg/L at admission predicted not regaining consciousness (85 % specificity; 66.6 % sensitivity; 84 % PPV; 78 % NPV; 77.6 % accuracy)
Schoerkhuber/1999/[5]	Non-traumatic out-of-hospital cardiac arrest	56	NSE	Serum	NSE was significantly higher in patients who had poor 6-month outcome at 12, 24, 48, and 72 h after ROSC NSE cutoffs for poor outcome were as follows: NSE > 38.5 μg/L at 12 h, NSE > 40 μg/L at 24 h, NSE >25.1 μg/L at 48 h, and NSE >16.4 μg/L at 72 h (100 % specificity) NSE >27.3 μg/L at any time predicted poor outcome (100 % specificity)
Molecules of non-CNS origin
Nagao/2004/[10]	Age >17 years, out-of-hospital cardiac arrest of presumed cardiac origin	401	BNP	Blood	Rate of survival to hospital discharge decreased in dose-dependent fashion with increasing quartiles of BNP on admission BNP >100 pg/mL predicted lack of survival until hospital discharge (83 % sensitivity; 96 % NPV)
Kasai/2011/[12]	Post cardiac arrest	357	Ammonia	Blood	Elevated ammonia on ER arrival is associated with decreased odds for good outcome at hospital discharge (OR = 0.98 [0.96–0.99]) Ammonia >192.5 μg/dL had 100 % NPV for good outcome at discharge 61 patients were treated with TH
Sodeck/2007/[11]	Post cardiac arrest, comatose	155	BNP	Blood	Highest quartile BNP on admission is associated with poor outcome as compared to lowest quartile BNP >230 pg/mL predicts unfavorable neurological outcome (OR = 2.25 [1.05–4.81]) and death at 6 months (OR = 4.7 [1.27–17.35])
Shinozaki/2011/[13]	Non-traumatic out-of hospital cardiac arrest with ROSC	98	Ammonia, Lactate	Blood	Elevated ammonia and lactate on admission were associated with poor outcome Ammonia >170 μg/dL predicted poor outcome (90 % sensitivity; 58 % specificity). Lactate >12 mmol/L predicted poor outcome (90 % sensitivity; 52 % specificity)
CSF biomarkers
Roine/1989/[13]	Out-of hospital VF arrest who survived >24 h	67	NSE, CKBB	CSF	NSE and CKBB at 20–26 h post CPR were elevated in patients who did not regain consciousness compared with those who did All patients with NSE > 24 μg/L remained unconscious or died at 3 months (74 % sensitivity; 100 % specificity) CKBB >17 μg/L predicted poor outcome (52 % sensitivity; 98 % specificity)
Sherman/2000/[17]	Comatose cardiac arrest patients with SSEP studies	52	CKBB	CSF	CKBB >205 U/L predicted non-awakening (49 % sensitivity; 100 % specificity) CSF sampling time not standardized
Martens/1998/[4]	Post cardiac arrest, unconscious and ventilated for >48 h	34	NSE, s100β	CSF	s100β and NSE were both significantly higher in patients who did not regain consciousness compared to those who did NSE >50 μg/L (89 % sensitivity; 83 % specificity) and s100β >6 μg/L (93 % sensitivity; 60 % specificity) predicted death or vegetative state CSF sampling time is not standardized
Rosen/2004/[19]	Post cardiac arrest, survive >12 days post ROSC	22	NFL	CSF	CSF sampled at 12–30 days after cardiac arrest NFL >18,668 μg/L predicted dependency in ADL at 1 year (100 % specificity; 46 % sensitivity)
Karkela/1993/[16]	VF or asystolic arrest	20	CKBB, NSE	CSF	Case controlled CSF collected at 4, 28, and 76 h after resuscitation Elevated CKBB at 4 and 28 h, and elevated NSE at 28 and 76 h after cardiac arrest were associated with not regaining consciousness
Oda/2012/[14]	Out-of-hospital cardiac arrest of presumed cardiac	14	HMGB1, s100β	CSF	CSF sampled at 48 h after ROSC HMGB1 and s100β were significantly higher in poor outcome group compared to good outcome group and to normal controls
Tirschwell/1997/[18]	Post cardiac arrest with CSF CKBB CKBB measured	351	CSF	CSF	Retrospective study CSF sampling time not standardized CKBB >205 U/L predicted non-awakening at hospital discharge (100 % specificity; 48 % sensitivity). Only 9 patients with CKBB >50U/L awakened and none regained independent ADLs

All studies are prospective observational unless otherwise noted

NPV negative predictive value, PPV positive predictive value, FPR false positive rate, OR odds ratio, ROSC return of spontaneous circulation, SSEP somatosensory evoked potential, TH therapeutic hypothermia, VF ventricular fibrillation

Table 2.

Biomarkers for outcome following cardiac arrest with therapeutic hypothermia treatment

Authors/year/Ref	Study design	Population	N	Bio-marker	Sample source	Findings
Tiainen/2003/[27]	RCT	Witnessed VF or VT arrest, ≤60 min between collapse to ROSC	70	NSE, s100β	Serum	NSE levels were lower in TH compared to normothermia NSE did not reach 100 % specificity in TH, whereas it does in normothermia TH: NSE >31.2 μg/L at 24 h, >26 μg/L at 36 h, and >25 μg/L at 48 h predicted poor outcome (96 % specificity) Normothermia: NSE >13.3 μg/L at 24 h, >12.6 μg/L at 36 h, and >8.8 μg/L at 48 h had 100 % specificity for poor outcome TH: s100β >0.21 μg/L at 24 h (100 % specificity), s100β >0.21 μg/L at 36 h, and s100β >0.23 μg/L at 48 h (96 % specificity) predicted poor outcome
Cronberg/2011/[29]	Pro	Post cardiac arrest with GCS <8 after ROSC	111	NSE	Serum	Elevated NSE was associated with worse outcome, DWI changes on MRI, and worse neuropathology All patients with NSE >33 μg/L at 48 h died without regaining consciousness NSE >27 μg/L predicted poor outcome at 6 months (100 % specificity)
Rundgren/2009/[25]	Pro	In- or out-of-hospital cardiac arrest, GCS ≤7	107	NSE, s100β	Serum	NSE >28 μg/L at 48 h predicted poor 6-month outcome (100 % specificity; 67 % sensitivity). s100β >0.51 μg/L at 24 h predicted poor 6-month outcome (96 % specificity; 62 % sensitivity)
Daubin/2011/[24]	Pro	In- or out-of-hospital cardiac arrest, comatose >48 h	97	NSE	Serum	Elevated NSE correlated with worse outcome at 3 months. NSE >47 μg/L predicted poor 3-month outcome (84 % specificity; 72 % sensitivity) NSE >97 μg/L predicted poor outcome (100 % PPV)
Shinozaki/2009/[23]	Pro	In- or out-of-hospital non-traumatic cardiac arrest with ROSC >20 min, with GCS ≤8	80	NSE, s100μ	Serum	s100β and NSE are both elevated in poor outcome group. s100β had better predictive performance than NSE s100β cutoff for poor outcome are as follows: s100β >1.41 μg/L at admission, s100β >0.21 μg/L at 6 h, and s100β >0.05 μg/L at 24 h post cardiac arrest (100 % specificity)
Stammet/2013/[28]	Pro	Post cardiac arrest	75	NSE, s100μ	Serum	Elevated s100β and NSE levels are associated with poor outcome at 6 months Adding s100β to Bispectral index improved predictive value for poor outcome
Rossetti/2012/[21]	Pro	Post cardiac arrest, comatose	61	NSE	Serum	5 cardiac arrest survivors, including 3 with good outcome, had NSE >33 μg/L
Mortberg/2011/[30]	Pro	Post cardiac arrest, SBP >80 mmHg x > 5 min, GCS ≤7, <6 h following ROSC	31	NSE, s100β, BDNF, GFAP	Serum	No association between BDNF and GFAP levels and outcome NSE >4.97 μg/L at 48 h and NSE >3.22 μg/L at 96 h post cardiac arrest predicted poor outcome at 6 months (93 % specificity) s100β >1.0 μg/L at 2 h (93 % specificity), and s100β >0.18 μg/L at 24 h (100 % specificity) post-cardiac arrest predicted poor outcome

PPV positive predictive value, Pro prospective observational, RCT randomized controlled trial, ROSC return of spontaneous circulation, TH therapeutic hypothermia, VF ventricular fibrillation, VT ventricular tachycardia

Table 3.

Biomarkers for subarachnoid hemorrhage

Author/year/Ref	Study design	Population	N	Bio-marker	Sample source	Findings
Molecules of CNS origin
Weismann/1997/[31]	Pro	Aneurysmal SAH within 3 days of ictus	70	s100β	Serum	s100β is higher at 24 h, 3 and 7 days post SAH compared to controls Higher s100β levels correlate with worse HH grade. higher s100β in the first week after SAH correlate with worse 6 month outcome
Stranjalis/2007/[32]	Pro	Spontaneous SAH within 48 h of ictus	52	s100β	Serum	Admission s100β >0.3 μg/L predicted unfavorable outcome and is an independent predictor of short-term survival (HR 2.2) (77.8 % sensitivity; 76 % specificity) s100β correlates positively with HH and Fisher scores s100β decreased after EVD insertion
Oertel/2006/[33]	Pro	Aneurysmal SAH	51	s100β, NSE	Serum	s100β during first 3 days of SAH is higher in those who died compared to survivors all patients with s100β >1.0 μg/L had unfavorable outcome NSE had no association with outcome s100β is lower in patients with vasospasm (by transcranial doppler)
Coplin/1999/[34]	Pro	Aneurysmal SAH	27	CKBB	CSF	CKBB >40μ/L is associated with poor outcome at hospital discharge (100 % specificity)
Inflammatory markers
Pan/2013/[65]	Pro	Aneurysmal SAH	262 SAH, 150 CTRL	pGSN	Blood	pGSN were lower in SAH compared with controls pGSN was an independent predictor of poor functional outcome (OR = 0.957) and death (OR = 0.953) at 6 months Adding pGSN improved predictive performance of WFNS and Fisher scores for functional outcome but not for mortality
Frijns/2006/[48]	Pro	SAH within 72 h of ictus, exclude perimesencephalic SAH	106	vWF	Serum	vWf >94.5 nmol/L was independently associated with increased odds for poor outcome at 3 months (OR 1.1–9.8) sICAM-1, sP-selectin, sE-selectin, vWf propeptide and ED 1-fibronectin were not independently associated with outcome
Mack/2002/[47]	Pro	SAH, excluding those with pro-inflammatory disease process	80	sICAM-1	Serum	sICAM-1 was elevated in SAH (293.3 ± 15 μg/L) compared with controls sICAM-1 on post-SAH days 8, 10, and 12 were significantly elevated in those with unfavorable mRS at discharge
Beeftink/2011/[46]	Pro	Aneurysmal SAH	67	TNFα, Leukocytes, CRP	Serum	Neither TNFα nor TNF-α genotype were associated with DCI or with SAH outcome at 3 months High leukocyte count and high CRP are not associated with DCI or SAH outcome
Chou/2011/[38]	Pro	Spontaneous SAH, within 96 h of ictus	55	MMP-9	CSF	Elevation of MMP-9 on post-SAH days 2–3 is associated with poor outcome (mRS 3–6) at 3 months
Chou/2011/[38]	Pro	Spontaneous SAH, within 96 h of ictus	55	Neutrophil, WBC	Blood	Elevated neutrophil count on post-SAH day 3 is associated with poor 3-month outcome Elevated WBC count throughout post-SAH days 0–14 is associated with angiographic vasospasm
Chou/2012/[50]	Pro	Spontaneous SAH, within 96 h of ictus	52	TNFα, IL-6	Serum	Elevated TNFα over post-SAH days 0–14 is independently associated with poor long-term outcome IL-6 is not associated with SAH outcome Neither TNFα nor IL-6 was associated with angiographic vasospasm
Chou/2011/[64]	Pro	Spontaneous SAH, within 96 h of ictus	42	pGSN	CSF, Serum	Serum pGSN is decreased in SAH compared to controls and decreases over time in SAH CSF pGSN is decreased in SAH compared to controls. Novel pGSN fragments found in SAH CSF but not in controls
Fassbender/2001/[52]	Pro	Aneurysmal SAH within 48 h of ictus	35	IL-1β, IL-6, TNFα	CSF, Serum	IL-lβ and IL-6 are significantly higher in CSF than in serum in SAH CSF IL-6 on post-SAH day 5 is significantly elevated in poor outcome group CSF TNFα did not show significant association with outcome
Mathiesen/1997/[53]	Pro	SAH patients with EVD	22	IL-1Rα, TNFα	CSF	IL-IRα was higher in poor grade SAH (HH 3–4; 318 vs. 82 pg/mL) Elevated IL-IRα and TNFα on post-SAH days 4–10 were associated with poor outcome
Weir/1989/[43]	Retro	Aneurysmal SAH with vital signs and CBC data (76 % missing data)	173	WBC	Blood	admission WBC >15 × 109/L shows 55 % mortality vs. 25 % mortality in the lower WBC group
Niikawa/1997/[39]	Retro	Fisher grade 3 SAH treated with aneurysm clipping within 24 h of ictus	103	WBC	Blood	WBC counts during days 3–5, 6–8, 9–11, and 12–14 after onset of SAH were significantly higher in patients with than in patients without symptomatic vasospasm
Other biomarkers
Niskakangas/2001/[79]	Case control	Aneurysmal SAH	108	ApoE4	Blood	Presence of ApoE4 was associated with unfavorable outcome (OR = 2.8 [1.18–6.77])
Juvela/2009/[76]	Case control	SAH within 48 h of ictus	105	ε2, ε4–containing genotypes	Blood	Apolipoprotein E ε2 or ε4-containing genotypes were not associated with outcome or occurrence of cerebral infarction
Lanterna/2005/[78]	Case control	SAH HH grade 1–3	101	ApoE4 genotype	Blood	Presence of Apo E4 genotype is associated with negative overall outcome Apo E4 genotype is associated with development of DIND
Leung/2002/[77]	Case control	Spontaneous SAH	72	ApoE4 genotype	Blood	ApoE4 genotype is associated with poor 6-month outcome (OR = 11.3 [2.2–57.0])
Kay/2003/[81]	Case Control	Spontaneous SAH requiring EVD	19	s100β, ApoE	CSF	s100β is significantly higher in SAH compared to controls ApoE is significantly lower in SAH compared to controls. Lower ApoE was associated with better clinical outcome
Lanterna/2005/[78]	Meta-analysis	Consecutive SAH, with 3-month follow-up data	696	ApoE4 genotype	Blood	Apo E4 genotype is associated with negative outcome (OR = 2.558 [1.610–4.065]) and delayed ischemia (OR 2.044 [1.269–3.291])
Moussouttas/2012/[88]	Pro	SAH with EVD, HH grade 3–5, endovascular aneurysm treatment	102	Epinephrine	CSF	Elevated CSF epinephrine within 48 h of admission is independently associated with mortality at 15 days (OR = 1.06 [1.01–1.10]) and with death and disability at 30 days (OR = 1.05 [1.02–1.09])
Yarlagadda/2006/[84]	Pro	Spontaneous SAH, >21 years	300	BNP, cTI	Serum	Initial BNP >600 pg/mL is associated with death (OR = 37.7 [5.0–286.2]) cTI >0.3 mg/L (on post-SAH day 9 ± 4) is associated with death (OR = 4.9 [2.1–26.8]) No standardized time of biosample collection
Naidech/2005/[82]	Pro	Spontaneous non-traumatic SAH	253	cTI	Serum	Peak cTI was independently predictive of death or severe disability at hospital discharge (OR = 1.4 [1.1–1.9]) cTI not independently predictive of 3-month outcome by mRS
Ramappa/2008/[83]	Retro	SAH diagnosed by CT scan or CSF, SAH ICD-9 code, with cTI measured	83	cTI	Blood	Peak cTI and GCS on presentation independently predicted in-hospital mortality

Pro prospective observational, Retro retrospective, CTRL control subjects, CBC complete blood count, HH grade hunt and hess grade, WFNS World Federation of Neurosurgeons Classification, DIND delayed ischemic neurological deficit, DCI delayed cerebral ischemia, mRS modified Rankins score, OR odds ratio

Table 4.

Biomarkers for acute ischemic stroke

Authors/year/Ref	Study design	Population	N	Bio-marker	Sample source	Findings
Molecules of CNS origin
Kazmierski/2012/[123]	Pro	AIS	458	s100β, OCLN, CLDN5, ZO1	Serum	Patients with clinical deterioration due to hemorrhagic transformation had higher s100β, OCLN, and CLDN/ZO1 ratio
Foerch/2004/[93]	Pro	AIS within 6 h of onset with proximal MCA occlusion	51	s100β	Serum	Mean s100β were higher in patients with malignant cerebral edema defined. s100β >1.03 μg/L at 24 h post AIS predicted malignant infarction (94 % sensitivity; 83 % specificity)
Missler/1997/[89]	Pro	AIS diagnosed by CT	44	s100β, NSE	Serum	s100β correlated with infarct volume and with 6 month outcome NSE correlated with infarct volume but not with clinical outcome Did not adjust for stroke subtype or tPA treatment
Foerch/2005/[91]	Pro	AIS within 6 h of onset	39	s100β	Serum	s100β at 48–72 h post AIS correlated with 6-month outcome and with infarct volume s100β ≤0.37 μg/L at 48 h post stroke predicted functional independence at 6 months (87 % sensitivity; 78 % specificity)
Herrmann/2000/[90]	Pro	Anterior circulation AIS	32	s100β, GFAP	Serum	s100β and GFAP correlated with total infarct volume and neurologic status at hospital discharge Did not adjust for stroke subtype or tPA treatment
Foerch/2003/[92]	Pro	AIS ≤5 h of onset with Ml occlusion	23	s100β	Serum	s100β <0.4 μg/L at 48–96 h post AIS predicted MCA recanalization within 6 h (86 % sensitivity; 100 % specificity)
Biomarkers of inflammation and blood brain barrier
Den Hertog/2009/[100]	RCT	AIS ≤12 h onset, no liver disease, prior mRS <2	561	CRP	Serum	From RCT for paracetamol for ischemic stroke. CRP measured within 12 h of stroke onset CRP >7 mg/L is associated with poor outcome (OR = 1.6 [1.1–2.4]) and death (OR = 1.7 [1.0–2.9])
Idicula/2009/[101]	Nested Pro	AIS ≤24 h onset	498	CRP	Serum	CRP >10 mg/L is independently associated with high NIHSS and high long-term mortality at 2.5 years
Montaner/2006/[99]	Pro	AIS in MCA territory treated with IV tPA within 3 h; exclude inflammatory disease or infection	143	CRP	Serum	CRP measured before tPA administration. CRP was higher in those who died after thrombolysis compared with survivors (0.85 vs. 0.53 mg/dL) CRP is independently associated with mortality at 3 months (OR = 8.51 [2.16–33.5]).
Winbeck/2002/[102]	Pro	AIS ≤12 h onset, NOT treated with IV tPA	127	CRP	Serum	CRP >0.86 mg/dL 24 h and at 48 h post-stroke are associated with death and lower likelihood of event-free survival at 1 year
Topakian/2008/[103]	Pro	AIS in MCA territory treated with IV tPA ≤6 h of onset, exclude CRP >6 mg/dL	111	CRP	Serum	CRP measured before tPA administration CRP level was not associated with NIHSS within 24 h or outcome at 3 months
Shantikumar/2009/[98]	Pro	AIS surviving >30 days	394	CRP	Serum	CRP higher in subject who died compared to survivors CRP is independently predictive of mortality after adjusting for conventional risk factors
Elkind/2006/[96]	Retro	Age >40, reside in northern Manhattan >3 months	467	hs-CRP	Serum	Highest quartile of hs-CRP is associated with increased risk of stroke recurrence (HR 2.08 [1.04–4.18]) and with combined outcome of stroke, MI, or vascular death (HR = 1.86 [1.01–3.42])
Huang/2012/[97]	Retro	Age >40, reside in northern Manhattan >3 months	741	hs-CRP	Serum	hs-CRP >3 mg/L was associated with higher mortality at 3 months and all-cause mortality (HR = 6.48 [1.41–29.8])
Castellanos/2003/[116]	Pro	Hemispheric AIS within 7.8 ± 4.5 h of onset	250	MMP-9	Plasma	MMP-9 ≥140 μg/L predicted hemorrhagic transformation (61 % PPV; 97 % NPV)
Castellanos/2007/[113]	Pro	AIS ≤3 h treated with IV tPA	134	c-Fn, MMP-9	Serum	MMP-9 ≥140 μg/L predicted hemorrhagic transformation (92 % sensitivity; 74 % specificity; 26 % PPV; 99 % NPV) c-Fn ≥3.6 μg/mL predicted hemorrhagic transformation (100 % sensitivity; 60 % specificity; 20 % PPV; 100 % NPV)
Moldes/2008/[119]	Pro	AIS treated with IV tPA	134	ET-1, MMP-9, c-Fn	Serum	ET-1, MMP-9, and c-Fn measured upon admission before tPA bolus ET-1 and c-Fn significantly higher in those with severe cerebral edema ET-1 >5.5 fmol/mL before tPA was independently associated with severe brain edema in multivariate analysis
Serena/2005/[118]	Case control	Malignant MCA infarction, <70 years	40 AIS, 35 CTRL	c-Fn, MMP-9	Plasma	c-Fn and MMP-9 were significantly higher in patients with malignant MCA infarcts c-Fn >16.6 μg/mL predicted malignant infarction (90 % sensitivity; 100 % specificity; 89 % NPV; 100 % PPV)
Montaner/2003/[114]	Pro	AIS in MCA territory treated with IV tPA within 3 h	41	MMP-9	Plasma	Higher baseline (pre-tPA) MMP-9 was associated with hemorrhagic transformation in dose-dependent fashion MMP-9 was predictive of hemorrhagic transformation in multivariate model (OR = 9.62)
Montaner/2001/[115]	Pro	Cardioembolic AIS in MCA territory	39	MMP-9	Plasma	Elevated baseline MMP-9 was associated with late-hemorrhagic transformation in multivariate regression (OR = 9)
Castellanos/2004/[93]	Pro	AIS treated with IV tPA by ECASS II criteria	87	c-Fn	Plasma	c-Fn was independently associated with hemorrhagic transformation in multivariate analysis (OR = 2.1) 71 of the patients were treated within 3 h of AIS onset. Similar results were found in these patients
Guo/2011/[57]	Pro	First onset AIS	172 AIS, 50 CTRL	pGSN	Plasma	Samples from first 24 h of stroke onset obtained pGSN decreased in AIS compared to controls pGSN was independent predictor for 1-year mortality pGSN >52 mg/L predicted 1-year mortality (73 % sensitivity; 65.2 % specificity)
Yin/2013/[106]	Pro	AIS	186 AIS, 100 CTRL	Visfastin	Plasma	Visfatin was higher in AIS than in controls. Visfatin was independent predictor of 6-month clinical outcome Adding visfatin did not improve predictive performance of NIHSS
Other biomarkers
Haapaniemi/2000/[122]	Case control	AIS	101 AIS, 101 CTRL	ET-1	Plasma	No difference in ET-1 levels between stroke and controls
Lampl/1997/[126]	Pro	AIS within 18 h from onset	26	ET-1	CSF, Plasma	CSF ET-1 correlated with volume of the lesion and higher in cortical infarcts compared to subcortical infarcts Plasma ET-1 was not elevated
Chiquete/2013/[124]	Pro	AIS	463	Uric acid	Serum	Uric acid ≤4.5 mg/dL at hospital admission was associated with very good 30 day outcome (OR = 1.76 [1.05–2.95]; 81.1 % NPV)
Matsumoto/2013/[125]	Retro	AIS from non-valvular AF within 48 h of onset	124	d-dimer	Plasma	d-dimer level at hospital admission is independently associated with infarct volume Highest d-dimer tertile group had worse outcome compared to middle and lowest tertiles

AF atrial fibrillation, NPV negative predictive value, PPV positive predictive value, Pro prospective observational, RCT randomized controlled trial, Retro retrospective, CTRL control subjects, NIHSS NIH stroke scale, OR odds ratio

Table 5.

Biomarkers for intracerebral hemorrhage

Authors/year/Ref	Study design	Population	N	Bio-marker	Sample source	Findings
Molecules of CNS origin
Hu/2012/[132]	Pro	Basal ganglia ICH within 6 h of onset	176	Tau	Serum	Tau >91.4 pg/mL predicted poor 3-month outcome (83.6 % sensitivity; 75.8 % specificity) Addition of tau improved prognostic value of NIHSS for outcome but not for mortality
Hu/2010/[127]	Pro	Basal ganglia ICH	86 ICH, 30 CTRL	s100β	Plasma	s100β was significantly associated with IVH, GCS scores, and ICH volumes s100β is independently associated with mortality at 1 week (OR = 1.046) s100β >192.5 pg/mL predicted 1-week mortality (93.8 % sensitivity; 70.4 % specificity)
Delgado/2006/[126]	Pro	ICH	78	s100β	Blood	s100β was higher in patients who deteriorated early and in patients with a poor neurological outcome
Brea/2009/[129]	Pro	ICH and AIS	44 ICH, 224 AIS	NSE	Blood	NSE elevation at 24 h post ICH was independently associated with poor outcome (OR = 2.6 [1.9–15.6])
James/2009/[128]	Pro	ICH	28	s100β, BNP	Blood	s100β and BNP levels correlated with outcome at hospital discharge Inclusion of biomarkers added little to the predictive power of ICH score
Cai/2013/[133]	Case control	Basal ganglia ICH	112 ICH, 112 CTRL	pNF-H	Plasma	pNF-H is higher in ICH compared to controls pNF-H is an independent predictor of 6-month mortality (OR = 1.287), 6-month unfavorable outcome (OR = 1.265) and early neurological deterioration (OR = 1.246) Addition of pNF-H did not improve predictive value of NIHSS
Biomarkers of inflammation
Leira/2004/[137]	Pro	ICH within 12 h of onset	266	Neutrophils, fibrinogen	Blood	Higher neutrophil count (OR = 2.1) and fibrinogen >523 mg/dL (OR = 5.6) on admission were independently associated with early neurological deterioration
Di Napoli/2011/[135]	Pro	ICH	210	WBC, CRP, Glucose	Blood	Higher WBC, CRP, and glucose were significantly related to mortality Only CRP remained significantly related to mortality when adjusted for ICH score and the combination of ICH score and CRP had the best predictive ability
Agnihotri/2011/[136]	Retro	Spontaneous ICH	423	WBC	Blood	Change in WBC (difference between max WBC in first 72 h and WBC on admission) correlated with worse discharge disposition and decline in modified Barthel Index at 3 months
Zhao/2013/[149]	Pro	Basal ganglia ICH within 6 h of onset	132 ICH, 68 CTRL	pGSN	Plasma	pGSN was lower in ICH compared to controls pGSN is an independent predictor of 6-month mortality and unfavorable outcome in multivariate analysis pGSN improved prognostic value of NIHSS for poor outcome but not for mortality
Castillo/2002/[134]	Pro	ICH within 24 h of onset	124	Glutamate, TNFα	Blood	Glutamate level was an independent predictor of poor outcome TNFα correlated with volume of peri-hematoma edema
Wang/2011/[139]	Pro Posthoc analysis	ICH within 24 h of onset	60	sICAM-1, sE-selectin	Plasma	Higher levels of sICAM-1 and sE-selectin were found in patients who had a poor outcome at hospital discharge.
Li/2013/[148]	Pro	ICH within 24 h of onset	59	MMP-3, MMP-9	Plasma	Elevated MMP-3 was independently associated with peri-hematoma edema volume MMP-3 >12.4 μg/L and MMP-9 >192.4 μg/L were associated with poor outcome in multivariate analysis
Hernandez-Guillamon/2012/[140]	Pro	ICH within 48 h of onset	66 ICH, 58 CTRL	VAP-1/SSAO	Plasma	VAP-1/SSAO activity <2.7 pmol/min·mg was independent predictor of neurological improvement after 48 h (OR = 6.8)
Fang/2005/[141]	Pro	ICH	43	IL-11	Plasma	samples collected in first 4 days of ICH plasma IL-11 higher in non-survivors compared to survivors
Diedler/2009/[138]	Retro	Supratentorial ICH	113	CRP	Blood	CRP is independent predictor of poor long-term functional outcome
Gu/2013/[142]	Pro	Basal ganglia ICH within 6 h of onset	85 ICH, 85 CTRL	Visfatin	Plasma	Visfatin was higher in ICH compared to controls Visfatin level was independent predictor of hematoma growth. (OR = 1.154 [1.046–3.018]) and of early neurological deterioration (OR = 1.195 [1.073–3.516])
Huang/2013/[143]	Case control	Basal ganglia ICH	128 ICH, 128 CTRL	Visfatin	Plasma	ICH patients had higher visfatin compared to controls Visfatin correlated with NIHSS and is independent predictor for 6-month mortality and unfavorable outcome
Zhang/2013/[144]	Pro	Basal ganglia ICH	92 ICH, 50 CTRL	Leptin	Plasma	Leptin higher in ICH compared to controls Leptin on admission is independent predictor of 6-month mortality and unfavorable outcome
Other biomarkers
Chiu/2012/[150]	Pro	ICH within 24 h of onset, >16 years old	170	d-dimer	Serum	d-dimer is independently associated with 30-day mortality (OR = 2.72)
Delgado/2006/[126]	Pro	ICH	98	d-dimer	Plasma	d-dimer levels were associated with presence of IVH or SAH extension d-dimer >1,900 μg/L is independently associated with early neurological deterioration (OR = 4.5) and with mortality (OR = 8.75)
Rodriguez-Luna/2011/[151]	Pro	Supratentorial ICH within 6 h of onset	108	LDL-C	Serum	Lower LDL-C levels were associated with hematoma growth, early neurological deterioration and 3-month mortality but not with NIHSS or ICH volume
Ramirez-Moreno/2009/[152]	Pro	ICH within 12 h of onset	88	LDL-C	Serum	Lipid profile measured in first hour after admission Low LDL-C levels were independently associated with death after ICH in multivariate analysis (HR = 3.07) LDL-C correlated with NIHSS, GCS, and ICH volume
Hays/2006/[153]	Retro	ICH	235	cTn1	Blood	Elevated cTn1 was independent predictor of in-hospital mortality
Chen/2011/[158]	Pro	ICH	64 ICH, 114 CTRL	Oxidative markers	Blood	Blood collected within 3 days of ICH Measured 8-OHdG, G6PD, GPx, MDA, vitamin E, vitamin A 8-OHdG elevation was independently associated with 30-day lower Barthel index but not with outcome by mRS
Wang/2012/[155]	Pro	ICH within 24 h of onset	60 ICH, 60 CTRL	Nuclear DNA	Plasma	Nuclear but not mitochondrial DNA correlated with GCS and ICH volume on presentation Nuclear DNA >18.7 μg/L on presentation was associated with poor outcome at discharge (63.6 % sensitivity; 71.4 % specificity)
Huang/2009/[154]	Pro	Basal ganglia ICH	36 ICH, 10 CTRL	Microparticles	Plasma, CSF	Plasma and CSF microparticles levels were associated with GCS score, ICH volume, IVH, and survival Controls have suspected SAH
Zheng/2012/[156]	Case control	ICH	79	miRNAs	Blood	Patients with hematoma expansion had different expression pattern of miRNAs (19 with increased expression, 7 with decreased expression)
Zhang/2012/[157]	Pro	Basal ganglia ICH	89 ICH, 50 CTRL	Copeptin	Plasma	Copeptin level is an independent predictor for 1-year mortality, poor outcome, and early neurological deterioration Copeptin did not improve prognostic value of NIHSS

Pro prospective observational, RCT randomized controlled trial, Retro retrospective, CTRL control subjects, OR odds ratio

Table 6.

Biomarkers for traumatic brain Injury

Authors/year	Study design	Population	N	Bio-marker	Sample source	Findings
Molecules of CNS origin
Okonkwo/2013/[180]	Pro	Mild, moderate & severe TBI	215	GFAP-BDP	Blood	Levels of GFAP-BDP were related to number of CT scan lesions and to neurological recovery A level of 0.68 μg/L was associated with a 21.61 OR for a positive CT and a 2.07 OR for failure to return to pre-injury baseline
Metting/2012/[170]	Pro	Mild TBI	94	s100β. GFAP	Blood	Levels of GFAP but not s100β were related to outcome, but the PPV was not high (<50 %)
Vos/2010/[178]	Pro	Moderate & severe TBI	79	s100β, GFAP	Blood	Levels of si00P and GFAP on admission were associated with poor outcome at 6 months and with mortality at 6 months even after adjusting for injury severity
Vos/2004/[172]	Pro	Severe TBI	85	s100β, NSE, GFAP	Blood	s100β, NSE, and GFAP were all higher in non-survivors and in those with poor 6-month outcome. s100β >1.13 μg/L predicted death with 100 % discrimination.
Wiesmann/2009/[167]	Pro	Mild, moderate, & severe TBI	60	s100β, GFAP	Blood	levels of s100β and GFAP were correlated with 6 month GOS Levels of s100β at 24 h post injury had the highest correlation
Pelinka/2004/[168]	Pro	TBI within 12 h	92	s100β, GFAP	Blood	GFAP and s100β were higher in non-survivors and predicted mortality
Nylen/2008/[167]	Pro	Severe TBI	59	s100β, s100a1b, s100βb	Blood	Levels of s100β, s100a1b, and s100βb were all related to 1 year GOS
Nylen/2006/[179]	Pro	Severe TBI	59	GFAP	Blood	Levels of GFAP were independently associated with 1-year outcome
Olivecrona/2009/[171]	Pro	Severe TBI	48	s100β, NSE	Blood	Levels of NSE and s100β were not significantly related to outcome at 3 or 12 months
Topolovec-Vranic/2011/[175]	Pro	Mild TBI within 4 h	141	s100β, NSE	Blood	s100β predicted poor cognitive outcome at 1 week NSE is independently associated with poor cognitive outcome at 6 weeks post injury
Rainey/2009/[165]	Pro	Severe TBI within 24 h	100	s100β	Blood	s100β at 24 h post injury were higher in patients with unfavorable outcome. s100β >0.53 μg/L predicted poor outcome (>80 % sensitivity; 60 % specificity)
Thelin/2013/[166]	Retro	Severe TBI	265	s100β	Blood	Levels of s100β between 12 and 36 h of injury were correlated with 6–12 month GOS and remained significantly related to outcome after adjustment for injury severity factors
Rodriguez-Rodriguez/2012/[159]	Pro	Severe TBI	55	s100β	Blood urine	Blood and urine s100β at 24 h postTBI were significantly higher in non-survivors Serum s100β >0.461 μg/L (88.4 % specificity) and urine s100β >0.025 μg/L (62.8 % specificity) predicted mortality
Kay/2003/[81]	Case control	TBI with GCS < 8	27 TBI, 28 CTRL	ApoE, s100β	CSF	s100β is elevated and ApoE is decreased in TBI compared with controls
Mondello/2012/[183]	Case control	severe TBI	95	UCH-L1	Blood, CSF	Blood and CSF levels of UCH-L1 were higher in patients with lower GCS, in patients who died, and in patients with unfavorable outcome. Levels at 6 h had the highest correlation Cumulative serum UCH-L1 >5.22 μg/L predicted death with OR = 4.8
Brophy/2011/[184]	Pro	Severe TBI GCS ≤8	86 (blood), 59 (CSF)	UCH-L1	Blood, CSF	Non-survivors had higher median serum and CSF UCH-L1 levels in the first 24 h
Papa/2009/[181]	Pro	TBI GCS ≤8 with EVD	41 TBI, 25 CTRL	UCH-L1	CSF	UCH-L1 was higher in TBI compared with controls at all time points up to 168 h Levels of UCH-L1 were higher in patients with a lower GCS at 24 h, post-injury complications, in those died within 6 weeks, and in those with poor outcome at 6 months
Papa/2012/[185]	Pro	Mild & moderate TBI GCS 9–15	96 TBI, 199 CTRL	UCH-L1	Blood	UCH-L1 within 4 h of injury distinguished TBI from uninjured controls (AUC = 0.87 [0.82–0.92]) UCH-L1 was associated with severity of injury in TBI
Liliang/2010/[177]	Pro	severe TBI	34	Tau	Blood	Tau levels were significantly higher in patients with a poor outcome Remained significant when adjusted for injury severity factors
Pineda/2007/[186]	Pro	Severe TBI	41	SBDP145, SBDP150	CSF	SBDP145 and 150 levels were significantly related to outcome at 6 months
Brophy/2009/[187]	Case control	Severe TBI	38	SBDP145, SBDP150	CSF	SBDP145 and 150 levels were higher in patients with worse GCS and longer ICP elevation
Mondello/2010/[188]	Pro	Severe TBI	40 TBI, 24 CTRL	SBDP145, SBDP120	CSF	SBDP145 >6 μg/L (OR = 5.9) and SBDP 120 > 17.55 μg/L (OR = 18.34) predicted death SBDP145 within 24 h of injury correlated with GCS score
Inflammatory markers
Schneider Soares/2012/[195]	Pro	mild, moderate, & severe TBI	127	IL-10, TNFα	Blood	Levels of IL-10 but not TNFα were related to mortality, even when adjusted for injury severity characteristics
Stein/2012/[194]	Pro	severe TBI	68	IL-8, TNFα	Serum	High levels of both IL-8 and TNFα predicted subsequent development of intracranial hypertension (specificity was high but sensitivity was low)
Tasci/2003/[196]	Pro	mild, moderate, & severe TBI	48	IL-1	Blood	IL-1 levels within 6 h correlated with the initial injury severity (GCS) and with GOS, but timing of the GOS is not described
Antunes/2010/[197]	Pro	TBI with hemorrhagic contusions	30	IL-6	Blood	IL-6 levels at 6 h were higher in patients who would subsequently clinically deteriorate due to evolving contusions
Combinations of markers
Diaz-Arrastia/2013/[189]	Pro	mild, moderate, & severe TBI	206	UCH-L1, GFAP	Blood	Levels of UCH-L1 were higher with moderate-severe than with mild TBI UCH-L1 levels were poorly predictive of complete recovery but better at predicting poor outcome For predicting complete recovery, UCH-L1 in combination with GFAP was not better than GFAP alone. For predicting favorable vs. unfavorable outcome, UCH-L1 is marginally better than GFAP and both together are better than either alone
Czeiter/2012/[190]	Pro	severe TBI	45	GFAP, UCH-L1, SBDP145	Serum, CSF	GFAP, UCH-L1, and SBDP145 all had at least one measure that was significantly related to unfavorable outcome When included in a model with IMPACT predictors of outcome, serum GFAP during first 24 h and the first CSF UCH-L1 value obtained were significantly related to mortality and only serum GFAP during first 24 h was significantly related to unfavorable outcome In combination, the IMPACT core model with the first CSF GFAP value, the first serum GFAP value, and the first CSF SBDP145 value performed the best

Pro prospective, PPV positive predictive value, Retro retrospective, CTRL control subjects, GOS Glasgow outcome scale, OR odds ratio, PPV positive predictive value

Review End-Points

1. (a) Are there cellular/molecular biomarkers that help predict long-term neurological prognosis in comatose cardiac arrest patients either treated or not treated with therapeutic hypothermia (TH)?

2. Are there cellular/molecular biomarkers that help predict (a) long-term outcome and (b) development of vasospasm and/or DCI after SAH?

3. Are there cellular/molecular biomarkers that help predict (a) long-term outcome and (b) incidence of malignant cerebral edema or hemorrhagic transformation following AIS?

4. Are there cellular/molecular biomarkers that help predict (a) long-term outcome and (b) hematoma expansion and cerebral edema following ICH?

5. Are there cellular/molecular biomarkers that help predict (a) long-term outcome and (b) cerebral edema and intracranial pressure (ICP) elevation after TBI?

Literature Summary

Cardiac Arrest HIE: Without TH

We found 20 prospective observational studies specific to these patients (Table 1). Outcome measures examined include regaining consciousness and Cerebral Performance Category (CPC) at 3 or 6 months after cardiac arrest.

Molecules of CNS Origin

The most extensively studied biomarkers in this population are s100β and neuron-specific enolase (NSE); both primarily originate from the central nervous system (CNS). The largest prospective study (n = 407) found serum NSE >33 μg/L at 24, 48, or 72 h after CA had 0 % false positive rate (FPR) for poor outcome (death or persistent unconsciousness) at 1 month [1]. Serum NSE >33 μg/L at 24–72 h post CA predicts death or persisting unconsciousness after 1 month with 100 % specificity in these patients. Several other studies with sample sizes between 56 and 110 supported serum NSE as a prognostic biomarker for HIE though optimal NSE cutoff values differed (20–65 μg/L) [2–6]. Additional confounds arise from the poor reproducibility of NSE assays used [7]. Studies with serial biomarker measurements over time found that serum NSE increased over the first 72 h following CA, reaching maximal between-group difference at 24–72 h, while serum s100β had maximal between-group difference immediately after CA and then levels decreased over time [2, 4, 8]. Though elevated serum s100β also predicts poor outcome [4, 6, 8, 9], it never achieved 100 % specificity in any study and appears to be inferior to somatosensory evoked potentials (SSEP) and serum NSE to predict CA outcome [1].

Molecules of Non-CNS Origin

Four prospective studies examined biomarkers from the first blood sample taken upon hospital arrival in CA patients, and found that elevated brain natriuretic peptide (BNP), ammonia, and lactate are associated with in-hospital mortality [10] and with poor 6-month outcome in comatose survivors [11–13]. These biomarkers lack replication and well-defined cutoff values.

CSF Biomarkers

There are few prospective studies on CSF biomarkers in CA; the largest sample size has 67 subjects. These studies suggest that elevation of CSF s100β [4, 14], NSE [15], creatinine kinase brain isoenzyme (CKBB) [15–18], and neurofilament (NFL) [19] are associated with non-awakening and poor outcome following CA. There was no standardized timing of CSF collection in these studies, and sampling time in these studies ranged from 4 h to 30 days post CA.

Cardiac Arrest HIE: With TH (Table 2)

Many of the clinical, electrophysiological, and molecular predictors of poor outcome following CA have decreased accuracy in patients treated with TH [20–22]. Several studies that either in part [23–25] or entirely [20, 26–28] examined TH-treated cardiac arrest patients showed that TH is associated with lower serum NSE levels and that it no longer predicts poor outcome after CA with 100 % specificity after TH [27]. Eight prospective observational studies examined this patient population. The largest studies consisted of 97–111 subjects and found that all patients with NSE >28–97 μg/L at 48 h post CA had poor outcome [24, 25, 29] and had MRI evidence of extensive brain injury [29]. Several other studies supported that higher serum s100β and NSE are associated with poor outcome, but none found 100 % specificity [23, 27, 28]. The latest prospective cohort study (n = 66) showed that 5 out of 28 CA survivors had NSE >33 μg/L at 24–48 h post CA, including 3 subjects who survived with good outcome (CPC 1–3) [21]. This study raised caution that NSE >33 μg/L does not have 100 % specificity for poor outcome after cardiac arrest in the era of TH.

s100β may have better predictive value than NSE for poor outcome in this patient population, but studies reported a wide range of optimal cutoff values for this biomarker [23, 28, 30]. Biomarker combinations such as s100β plus NSE [26] or molecular biomarkers plus physiological markers [28] may improve prediction precision for outcome after cardiac arrest following TH, but all require further validation.

Subarachnoid Hemorrhage (Table 3)

Molecules of CNS Origin

Several prospective studies found that serum s100β is elevated in SAH compared to healthy controls [31] and that elevated blood s100β is associated with initial neurologic severity and long-term outcome [31–33]. However, studies to date are relatively small and have wide standard deviation in s100β levels. Furthermore, external ventricular drain (EVD) insertion in SAH is associated with decreased blood s100β [32], which is an important confounding factor. One study suggested that lower serum s100β is associated with vasospasm while higher serum s100β is associated with mortality [33], which introduces contradicting associations between s100β and SAH. Elevation of CSF CKBB, a marker of astrocytic cell death, has been linked to poor short-term SAH outcome [34].

Biomarkers of Inflammation

Inflammation may play a role in secondary brain injury and outcome after SAH [35, 36]. Several prospective cohort studies have found that blood leukocyte elevations are associated with vasospasm [37–39] and poor SAH outcome [37–45], though there are reports to the contrary [46]. Elevation of blood CRP and adhesion molecules, e.g., soluble ICAM-1, are associated with DCI [47] and poor outcome in SAH [48], but other moderately sized studies report the contrary [48]. Elevation of serum von Willebrand factor (vWf), a marker of endothelial cell activation, is associated with DCI [48, 49] and worse outcome after SAH [48], but these results have yet to be validated.

Pro-inflammatory cytokines have been implicated in DCI and brain injury after SAH. Human studies found that elevated serum TNFα [50] is associated with poor outcome and elevated blood soluble endoglin (sEng), blood transforming growth factor-β (TGFβ), and elevated CSF TNFα, IL-6, and IL-1Ra are associated with vasospasm [51–54]. However, some studies report contrary results with blood TNFα [46, 52] and IL-6 [50]. Studies examining CSF cytokine profiles in SAH are all small. While all studies suggest elevated CSF proinflammatory cytokines such as TNFα, IL-1Rα, and IL-6 and soluble TNFα receptor I (sTNFR-1) which are associated with initial SAH clinical severity, there are conflicting reports on the associations between CSF cytokines and SAH outcome [52, 53, 55]. An important consideration is that intracranial hypertension causes CSF IL-6 elevation [56], and this may confound studies of CSF IL-6 levels in SAH and other conditions associated with increased ICP.

Metalloproteinases (MMPs) can disrupt neuron-extracellular matrix interaction leading to brain injury [57, 58], and elevated blood and CSF MMP-9 levels have been linked with vasospasm [59] and poor SAH outcome [38]. MMPs can cleave plasma-type gelsolin (pGSN) [60], which is thought to mitigate pro-inflammatory effects of cytokines [61, 62] and may be neuroprotective [63]. SAH is associated with decrease pGSN in blood and CSF; decreased blood pGSN may be associated with poor outcome [64, 65], and novel pGSN breakdown fragments have been identified in SAH CSF [64].

Vasoactive Biomarkers

Activity of endopeptidases such as MMPs [66, 67] produce Endothelin-1 (ET-1), the strongest vasoconstrictor in the CNS. ET-1 is implicated in the pathogenesis of vasospasm [68, 69]. Several human studies link elevated CSF ET-1 levels with vasospasm [70–72], though other studies do not [73, 74]. A randomized clinical trial showed that ET-1_A antagonist reduced the incidence of angiographic vasospasm but not DCI or poor outcome in SAH [75], raising new questions about the role of ET-1 in SAH.

Other Biomarkers

ApoE genotype is associated with outcome after brain injury. Several moderately sized studies found conflicting associations between ApoE4 genotype and SAH outcome [76–79]. A meta-analysis of eight studies (n = 696) found ApoE4 is associated with worse SAH outcome [80]. One small study found decreased CSF ApoE protein is associated with higher SAH severity and less favorable outcome [81].

SAH is associated with cardiac dysfunction. Epidemiologic studies describe a positive association between elevated cardiac troponin-I (cTI) and initial SAH clinical severity as well as death or severe disability at discharge [82, 83], though one large study did not support this conclusion [84]. Elevation of another cardiac-derived peptide, BNP, has been linked to hyponatremia [85, 86], DCI [87], and mortality after SAH [84]. It is not known whether the association between cTI and BNP and SAH outcome reflect the effect of cardiac injury on SAH outcome. CSF epinephrine elevation in the first 48 h of SAH has also been independently linked to higher SAH mortality and disability [88].

Acute Ischemic Stroke (Table 4)

Advances in AIS therapies have created the need for fast and accurate biomarkers that can reflect extent of real-time brain tissue injury to guide revascularization therapy and to predict the risk of secondary hemorrhagic transformation—a “brain troponin”.

Molecules of CNS Origin

Small (n < 50) prospective studies found that blood s100β, NSE [89], and GFAP [90]—all correlate positively with total infarct volume and outcome after AIS [89, 90], but these biomarkers did not add to the prognostic accuracy of existing clinical predictors. The strongest AIS biomarker studies to date studied patients with middle cerebral artery (MCA) territory infarcts and found that blood s100β at 48–96 h after AIS had the highest predictive value for functional outcome and for total infarct volume [91], and that low serum s100β was associated with early MCA recanalization [92]. Elevated s100β is also associated with the malignant cerebral edema after AIS [93, 94].

Biomarkers of Inflammation and BBB Injury

C-reactive protein (CRP) is an acute-phase response protein which may itself have pro-inflammatory effects and increase secondary brain injury [95]. Elevated serum CRP in acute and subacute phases (0–15 days) of AIS is independently associated with mortality or stroke recurrence in large cohort studies [96–98]. One large (n > 100) study found that hyperacute serum CRP elevation was associated with death after IV thrombolysis, regardless of vessel recanalization, and that serum CRP correlated with 3-month modified Rankin scores (mRS) in a dose-dependent fashion [99]. Several other studies [100–102], but not all [103] supported the correlation of hyperacute serum CRP and AIS mortality. The cutoff values for CRP in these studies vary widely. Measurement imprecision and confounding by infection and non-specific inflammation are important limitations in its utility as an AIS biomarker. General leukocyte elevations [104, 105] and elevation of visfatin, a pro-inflammatory factor, [106] also are associated with poor AIS outcome.

Leukocytes are a major source of MMP release following ABI [107]. MMPs are implicated in numerous pathogenic mechanisms in AIS including BBB disruption [108, 109], progression of cerebral edema, and worsening of cerebral ischemic injury [110]. In AIS, higher baseline blood MMP-9 levels are associated with thrombolysis failure [111] and with increased risk of parenchymal hematoma either with [112–114] or without IV tPA [115, 116]. Higher MMP-9 mRNA level is associated with poor outcome and mortality after AIS [117]. Elevated serum MMP-9, c-fibronectin (cFn), and ET-1 have been associated with malignant cerebral edema after AIS [118, 119], though the data on MMP-9 is inconsistent [119]. The association of blood c-Fn and malignant cerebral edema after AIS is consistent across two moderately sized studies [118, 119]. Plasma c-Fn level also is associated with hemorrhagic transformation in tPA-treated AIS cohort. Pre-tPA blood MMP-9 ≥140 ng/mL with c-Fn ≥3.6 μg/mL predicts hemorrhagic transformation with 87 % specificity and PPV of 41 % [94].

Several studies have found contradictory results regarding whether ET-1 levels, in plasma or CSF, are elevated after AIS [120, 121]. The largest study to date found no association between plasma ET-1 levels and AIS or its outcome [122]. Other potential biomarkers of BBB injury include tight junction proteins such as occludin (OCLN), claudin 5 (CLDN5), and zonula occudens 1 (ZO1). Elevated plasma levels of these proteins are associated with hemorrhagic transformation of AIS [123].

Other Biomarkers

A large prospective registry found that low serum uric acid level in the first 7 days of AIS is independently associated with good short-term outcome [124]. Admission level of serum d-dimer has also been linked with total infarct volume and outcome at hospital discharge in AIS from atrial fibrillation [125].

Intracerebral Hemorrhage (Table 5)

Molecules of CNS Origin

Four prospective observational studies (total of 236 subjects) examined s100β in ICH. Three of the four studies found that elevated serum s100β is associated with early deterioration and poor ICH outcome at hospital discharge [126–128]. One found that serum NSE level was associated with poor 3-month ICH outcome while s100β was not [129]. In these studies, serum s100β levels were found to correlate with Glasgow Coma Scale (GCS), ICH volume, and the presence of intraventricular hemorrhage (IVH). Elevated serum myelin basic protein (MBP) has been reported in TBI and ICH [130, 131]. Elevated serum tau protein [132], phosphorylated axonal neurofilament subunit H (pNF-H) [133], and blood glutamate [134] were also found to predict poor outcome in ICH. However, it is unclear whether the addition any of the above biomarkers improved predictive value of known clinical predictors of ICH outcome [126, 128, 133].

Biomarkers of Inflammation

Elevated leukocyte or neutrophil count [135–137], CRP [135, 138], and adhesion molecules such as soluble ICAM-1, soluble E-selectin [139], and vascular adhesion protein-1 (VAP-1) [140] have been associated with increased ICH mortality and worse ICH outcome. Elevated cytokines such as TNFα and IL-11 have also been associated with perihematoma edema and ICH mortality [134, 141]. Blood adipokines such as visfatin and leptin are found to be elevated in basal ganglia ICH and are associated with hematoma expansion, early neurological deterioration, and poor long-term outcome [142–144]. Plasma MMP-9 levels have been associated with ICH peri-hematoma edema [145, 146] and ICH enlargement [148], while MMP-3 is associated with poor ICH outcome [146, 148]. Decreased anti-inflammatory marker pGSN has also been associated with poor ICH outcome and mortality [149].

Other Markers

Molecules from the coagulation cascade such as d-dimer [126, 150] and fibrinogen [137] have been associated with increased mortality and early deterioration after ICH. Decreased serum LDL-C levels also correlate with hematoma expansion, ICH volume, and ICH mortality [151, 152]. Elevated cTn-I has been associated with increased in-hospital ICH mortality [153]. Emerging data suggest some novel biomarkers such as blood and CSF microparticles [154], blood nuclear DNA [155], blood microRNA expression patterns [156], plasma copeptin [157], and oxidative marker blood leukocyte 8-hydroxy-2′-deoxyguanosine (8-OHdG) [158] may be associated with ICH volume, ICH expansion, mortality, and outcome.

Of the above candidate biomarkers of ICH outcome and secondary deterioration, none have been validated or proven to improve the predictive performance of existing clinical predictors of ICH outcome.

Traumatic Brain Injury (Table 6)

Molecules of CNS Origin

Elevated blood and urine levels [159] of s100β and its isoforms s100A1B and s100BB are associated with poor outcome in TBI [160–169], though this may not hold true in the case of mild TBI [170] or severe TBI patients treated with ICP-targeted therapy [171]. In a study of 265 patients, blood s100β levels drawn between 12 and 36 h after injury were associated with neurological outcome. The area under curve (AUC) of s100β levels during the first 48 h after injury had the strongest relationship to neurological outcome [166]. Elevated blood NSE levels have been linked to poor outcome in severe [172–174] as well as mild TBI [175], though results have been inconsistent [163, 171, 176]. Elevations in serum and CSF MBP, CKBB [131], and tau protein [177] were also found in TBI.

Several studies found elevated blood GFAP levels to predict TBI outcome [167, 168, 172, 178, 179], including mild TBI [170]. GFAP is not elevated in polytrauma without TBI and therefore may have more CNS-specificity compared to s100β or NSE. Limitations in the utility of GFAP as a biomarker include high variability in reported blood GFAP levels after TBI and the rapid decline of blood GFAP levels after TBI requiring measurements within the first hours of TBI [167]. GFAP breakdown products (GFAP-BDP) are emerging as a potential TBI biomarker. In a study of 215 patients with mild to severe TBI, GFAP-BDP >0.68 μg/L within 24 h of injury was associated with acute traumatic lesions on the CT and with unfavorable 6-month outcome [180].

Two highly neuron-specific protein biomarkers, ubiquitin c-terminal hydrolase L1 (UCH-L1) and alpha-II spectrin breakdown products (SBDP), are emerging as promising candidate TBI biomarkers. UCH-L1 is involved in the ubiquitination of abnormal proteins for proteasome degradation and is present in almost all neurons. UCH-L1 elevation is detected soon after TBI in both CSF and serum [181], and elevated UCH-L1 levels are associated with lower GCS and poor outcome after TBI [181–185]. Alpha-II spectrin is a protein primarily expressed in neurons but not in glia and is most concentrated in axons. It is broken down by the calpain and caspase-3 cysteine proteases, and its breakdown products SBDP120, SBDP145, and SBDP150 in the CSF of TBI patients are associated with worse GCS, longer ICP elevation, and poor outcome following TBI [186–188]. The high neuronal specificity and early detection of these markers make them promising future candidate biomarkers for TBI.

Combinations of biomarkers may have better prognostic value than individual biomarkers. In a study of 206 subjects, GFAP and UCH-L1 together had better sensitivity and specificity to discriminate between TBI patients and healthy controls than either biomarker alone. The combination also had better sensitivity and specificity for predicting 3-month outcome [189]. Other combinations of biomarkers found to have prognostic value for TBI outcome include GFAP, UCH-L1, and SBDP145 [190], and NSE and s100β [191].

Biomarkers of Inflammation

Leukocyte elevation is associated with poor outcome after TBI [192, 193]. Elevation of blood cytokines TNFα and IL-8 levels are associated with impending intracranial hypertension and cerebral hypoperfusion in TBI [194], while elevation of IL-10 is independently linked to TBI mortality [195]. Other proinflammatory cytokines such as IL-1 and IL-6 have been linked with initial injury severity and with deterioration following TBI [196, 197], though there are conflicting reports for IL-1 [198]. Decreased levels of pGSN also are linked with mortality and worse outcome in TBI [199, 200].

Though numerous candidate biomarkers have a positive association to TBI outcome, it is not clear that they add significant prognostic value to existing clinical predictors. The heterogeneity of TBI pathophysiology will require large numbers of patients to confirm the clinical usefulness of any of these biomarkers. Combining biomarkers into a panel may provide more information than individual biomarkers.

Limitations

Studies to date have identified numerous candidate molecular biomarkers for prognostication and prediction of secondary complications in ABI, but few have been validated in large cohorts and none have translated into routine clinical use in neurocritical care. Major barriers to validate existing biomarkers include high biological and treatment heterogeneity and the lack of standardization in phenotype definition and in methods of sample collection, processing, storage, and biomarker assay. The formation of large collaborative consortiums and multidisciplinary research teams are vital towards the successful identification and validation of high-fidelity biomarkers, particularly in the era of multiplex technology allowing high-throughput and multiplex screening of potential biomarkers [201].

Recommendations (and see Summary Statement)

1. In comatose post-cardiac hypoxic-ischemic encephalopathy (HIE) patients not treated with TH, we suggest the use of serum NSE in conjunction with clinical data for neurologic prognostication (Weak recommendation, Moderate quality of evidence).

2. We recommend against the use of serum NSE for prognostication in HIE treated with TH (Strong Recommendation, Moderate quality of evidence).

3. We recommend against the routine use of molecular biomarker for outcome prognostication in AIS, SAH, ICH, or TBI (Strong Recommendation, Low quality of evidence).

Additional Conclusions

• Routine use of CSF biomarkers for prognostication in comatose post-cardiac hypoxic-ischemic encephalopathy (HIE) patients not treated with TH does not appear to provide valuable information. (Low quality of evidence)

• There is a limited role for routine use of blood or CSF molecular biomarkers to predict vasospasm and DCI in SAH. (Low quality of evidence)

• Plasma MMP-9 and c-Fn can be used in conjunction with clinical data to support prediction of hemorrhagic transformation in AIS patients treated with IV tPA within 3 hours of onset. (Low quality of evidence)

• Routine use of molecular biomarkers does not help predict secondary deterioration after ICH or TBI. (Low quality of evidence)

Appendix 1: Participants in the International Multi-disciplinary Consensus Conference on Multimodality Monitoring

Peter Le Roux, MD, FACS,

Brain and Spine Center,

Suite 370, Medical Science Building,

Lankenau Medical Center,

100 East Lancaster Avenue, Wynnewood, PA 19096, USA.

Tel: +1 610 642 3005

Fax: 610 642 3057

lerouxp@mlhs.org

David K Menon MD PhD FRCP FRCA FFICM FMedSci,

Head, Division of Anaesthesia, University of Cambridge

Consultant, Neurosciences Critical Care Unit

Box 93, Addenbrooke’s Hospital

Cambridge CB2 2QQ, UK

dkm13@wbic.cam.ac.uk

Paul Vespa, MD, FCCM, FAAN, FNCS

Professor of Neurology and Neurosurgery

Director of Neurocritical Care

David Geffen School of Medicine at UCLA

Los Angeles, CA 90095 USA

PVespa@mednet.ucla.edu

Giuseppe Citerio, MD

Director NeuroIntensive Care Unit,

Department of Anesthesia and Critical Care

Ospedale San Gerardo, Monza.

Via Pergolesi 33, Monza 20900, Italy

g.citerio@hsgerardo.org

Mary Kay Bader RN, MSN, CCNS, FAHA, FNCS

Neuro/Critical Care CNS

Mission Hospital

Mission Viejo CA 92691, USA

Marykay.Bader@stjoe.org

Gretchen M. Brophy, PharmD, BCPS, FCCP, FCCM

Professor of Pharmacotherapy & Outcomes Science and

Neurosurgery

Virginia Commonwealth University

Medical College of Virginia Campus

410 N. 12th Street

Richmond, Virginia 23298–0533 USA

gbrophy@vcu.edu

Michael N. Diringer, MD

Professor of Neurology, Neurosurgery & Anesthesiology

Chief, Neurocritical Care Section

Washington University

Dept. of Neurology, Campus Box 8111

660 S Euclid Ave

St Louis, MO 63110 USA

diringerm@neuro.wustl.edu

Nino Stocchetti, MD

Professor of Anesthesia and Intensive Care

Department of physiopathology and transplant

Milan University

Director Neuro ICU

Fondazione IRCCS Cà Granda Ospedale Maggiore

Policlinico

Via F Sforza, 35 20122 Milan Italy

stocchet@policlinico.mi.it

Walter Videtta, MD

ICU Neurocritical Care

Hospital Nacional ‘Prof. a. Posadas’

El Palomar - Pcia. de Buenos Aires

Argentina

wvidetta@ar.inter.net

Rocco Armonda, MD

Department of Neurosurgery

MedStar Georgetown University Hospital

Medstar Health, 3800 Reservoir Road NW

Washington DC 20007

USA

Rocco.Armonda@gmail.com

Neeraj Badjatia, MD

Department of Neurology

University of Maryland Medical Center,

22 S Greene St

Baltimore, MD, 21201

USA

nbadjatia@umm.edu

Julian Boesel, MD

Department of Neurology

Ruprect-Karls University

Hospital Heidelberg, Im Neuenheimer Feld 400,

D-69120 Heidelberg,

Germany

Julian.Boesel@med.uni-heidelberg.de

Randal Chesnut, MD, FCCM, FACS

Harborview Medical Center,

University of Washington Mailstop 359766

325 Ninth Ave,

Seattle WA 98104–2499

USA

chesnutr@u.washington.edu

Sherry H-Y. Chou, MD, MMSc, FNCS

Assistant Professor of Neurology, Harvard Medical School

Department of Neurology, Brigham and Women’s Hospital

75 Francis Street

Boston MA, 02115

USA

schou1@partners.org

Jan Claassen, MD, PhD, FNCS

Assistant Professor of Neurology and Neurosurgery

Head of Neurocritical Care and Medical Director of the

Neurological Intensive Care Unit

Columbia University College of Physicians & Surgeons

177 Fort Washington Avenue, Milstein 8 Center room 300,

New York, NY 10032

USA

jc1439@cumc.columbia.edu

Marek Czosnyka, PhD

Department of Neurosurgery

University of Cambridge,

Addenbrooke’s Hospital, Box 167

Cambridge, CB20QQ

United Kingdom

mc141@medschl.cam.ac.uk

Michael De Georgia, MD

Professor of Neurology

Director, Neurocritical Care Center

Co-Director, Cerebrovascular Center

University Hospital Case Medical Center

Case Western Reserve University School of Medicine

11100 Euclid Avenue

Cleveland, Ohio 44106

michael.degeorgia@uhhospitals.org

Anthony Figaji, MD, PhD

Head of Pediatric Neurosurgery

University of Cape Town

617 Institute for Child Health

Red Cross Children’s Hospital

Rondebosch, 7700 Cape Town,

South Africa

anthony.figaji@uct.ac.za

Jennifer Fugate, DO

Department of Neurology,

Mayo Clinic,

200 First Street SW

Rochester, MN 55905

Fugate.Jennifer@mayo.edu

Raimund Helbok, MD

Department of Neurology, Neurocritical Care Unit

Innsbruck Medical University,

Anichstr.35, 6020

Innsbruck,

Austria

raimund.helbok@uki.at

David Horowitz, MD

Associate Chief Medical Officer

University of Pennsylvania Health System,

3701 Market Street

Philadelphia, PA, 19104

USA

david.horowitz@uphs.upenn.edu

Peter Hutchinson, MD

Professor of Neurosurgery

NIHR Research Professor

Department of Clinical Neurosciences

University of Cambridge

Box 167 Addenbrooke’s Hospital

Cambridge CB2 2QQ

United Kingdom

pjah2@cam.ac.uk

Monisha Kumar, MD

Department of Neurology

Perelman School of Medicine, University of Pennsylvania,

3 West Gates

3400 Spruce Street

Philadelphia, PA, 19104

USA

monisha.kumar@uphs.upenn.edu

Molly McNett, RN, PhD

Director, Nursing Research

The MetroHealth System

2500 MetroHealth Drive,

Cleveland, OH 44109

USA

mmcnett@metrohealth.org

Chad Miller, MD

Division of Cerebrovascular Diseases and Neurocritical

Care

The Ohio State University

395 W. 12th Ave, 7th Floor

Columbus, OH 43210

ChadM.Miller@osumc.edu

Andrew Naidech, MD, MSPH

Department of Neurology

Northwestern University Feinberg SOM 710

N Lake Shore Drive, 11th floor

Chicago, IL 60611

ANaidech@nmff.org

Mauro Oddo, MD

Department of Intensive Care Medicine

CHUV University Hospital, BH 08–623

Faculty of Biology and Medicine University of Lausanne

1011 Lausanne, Switzerland

Mauro.Oddo@chuv.ch

DaiWai Olson, RN, PhD

Associate Professor of Neurology, Neurotherapeutics and

Neurosurgery

University of Texas Southwestern

5323 Harry Hines Blvd

Dallas, TX 75390–8897

USA

daiwai.olson@utsouthwestern.edu

Kristine O’Phelan M.D.

Director of Neurocritical Care

Associate Professor, Department of Neurology

University of Miami, Miller School of Medicine

JMH, 1611 NW 12th Ave, Suite 405

Miami, FL, 33136

USA

kophelan@med.miami.edu

Javier Provencio, MD

Associate Professor of Medicine

Cerebrovascular Center and Neuroinflammation Research

Center

Lerner College of Medicine

Cleveland Clinic,

9500 Euclid Ave, NC30

Cleveland, OH 44195

USA

provenj@ccf.org

Corina Puppo, MD

Assistant Professor, Intensive Care Unit,

Hospital de Clinicas, Universidad de la República,

Montevideo

Uruguay

coripuppo@gmail.com

Richard Riker, MD

Critical Care Medicine

Maine Medical Center,

22 Bramhall Street

Portland, Maine 04102–3175

USA

RRiker@cmamaine.com

Claudia Robertson, MD

Department of Neurosurgery

Medical Director of Center for Neurosurgical Intensive

Care,

Ben Taub Hospital

Baylor College of Medicine,

1504 Taub Loop, Houston, TX 77030

USA

claudiar@bcm.tmc.edu

J. Michael Schmidt, PhD, MSc

Director of Neuro-ICU Monitoring and Informatics

Columbia University College of Physicians and Surgeons

Milstein Hospital 8 Garden South, Suite 331

177 Fort Washington Avenue,

New York, NY 10032

USA

mjs2134@columbia.edu

Fabio Taccone, MD

Department of Intensive Care, Laboratoire de Recherche

Experimentale

Erasme Hospital,

Route de Lennik, 808

1070 Brussels

Belgium

ftaccone@ulb.ac.be