Child's Nervous System

, 26:205

Clinical applications of biomarkers in pediatric traumatic brain injury


  • Simon J. I. Sandler
    • Division of Neurosurgery, School of Child and Adolescent Health, Red Cross War Memorial Children’s HospitalUniversity of Cape Town
    • Division of Neurosurgery, School of Child and Adolescent Health, Red Cross War Memorial Children’s HospitalUniversity of Cape Town
    • Red Cross Children’s Hospital
  • P. David Adelson
    • Phoenix Children’s Neuroscience Institute
Focus Session

DOI: 10.1007/s00381-009-1009-1

Cite this article as:
Sandler, S.J.I., Figaji, A.A. & Adelson, P.D. Childs Nerv Syst (2010) 26: 205. doi:10.1007/s00381-009-1009-1



The diagnosis, treatment, and prediction of outcome in pediatric traumatic brain injury (TBI) present significant challenges to the treating clinician. Clinical and radiological tools for assessing injury severity and predicting outcome, in particular, lack sensitivity and specificity. In patients with mild TBI, often there is uncertainty about which patients should undergo radiological imaging and who is at risk for long term neurological sequelae. In severe TBI, often there is uncertainty about which patients will experience secondary insults and what the outcome for individual patients will be. In several other clinical specialties, biomarkers are used to diagnose disease, direct treatment, and prognosticate. However, an ideal biomarker for brain injury has not been found.


In this review, we examine the various factors that must be taken into account in the search for a reliable biomarker in brain injury. We review the important studies that have investigated common biomarkers of structural brain injury, in particular S100B, neuron-specific enolase, myelin basic protein, and glial fibrillary acid protein.


The potential uses and limitations of these biomarkers in the context of TBI are discussed.


BiomarkersTraumatic brain injuryChildrenS100B


Traumatic brain injury (TBI) is a leading cause of death and disability worldwide. Management is challenging—both minor and severe TBI create significant difficulties for treating clinicians. Several typical questions confront the clinician regarding extent of injury and prognosis. Because there is such marked heterogeneity in TBI, predicting outcome is difficult, as is deciding on optimal treatment. For patients who survive severe TBI, some make a good recovery while others are left with severe neurological disability. In children, uncertainties about predicting outcome are even greater than in adults and are multifactorial [23]. There are increased difficulties in assessing both initial injury severity and outcome. Objective tools to better understand the degree of injury and to prognosticate are urgently needed.

Difficulties in assessing injury severity and outcome

Clinical examination has limited ability to assess injury severity or predict outcome. Patients with mild head injury usually have few clinical symptoms but may be at risk of later deterioration. Even in the absence of neurological deterioration, many patients suffer long term sequelae. For patients with severe TBI, various determinants of primary and secondary injury confound the relationship between the initial clinical examination and outcome. Sedation and ventilation in the intensive care unit further reduce the usefulness of the clinical examination. Imaging of the brain fares little better. Computerized tomography (CT) scanning of the brain has a low sensitivity for detecting diffuse brain damage, and although magnetic resonance imaging (MRI) is more sensitive, it is usually difficult to scan ventilated, often hemodynamically unstable patients, and MRI will not likely be used widely in the assessment of patients with mild TBI. Monitoring the brain may provide more information about pathophysiological disturbances after TBI but have limited sensitivity and specificity. These include monitoring for intracranial pressure (ICP), cerebral perfusion pressure, brain oxygenation, cerebral metabolism, and cerebral hemodynamics [13, 15, 16, 29, 70]. None of these predict outcome or clinical deterioration with sufficiently high accuracy to be reliable on their own.

A role for biomarkers in neurotrauma

In medicine today, several specialties employ biomarker blood tests to diagnose, direct treatment, and prognosticate. Commonly used biomarkers include troponin T/CK-Mb in cardiology, procalcitonin and ESR in sepsis, amylase and lipase in pancreatic disease, etc. Many of these markers of organ function and injury have long been used to guide management, yet similar biomarkers have not existed in routine care of TBI. In comparison with the coronary care unit, neurointensive care units have fewer markers or monitors of organ function and fewer therapies for intervention.

In 1983, the properties for an optimal brain biomarker were theorized by Bakay and Ward [5]. The authors suggested that such a biomarker should (1) show high specificity and sensitivity for brain, (2) have a rapid appearance in serum, (3) be released only after irreversible destruction of brain tissue, (4) be released in a time-locked sequence with the injury, (5) show low age- and sex-related variability, and (6) have reliable assays for immediate analysis available. Most importantly, the brain biomarker should show clinical significance.

Broadly, the ideal biomarker has several potential benefits in the management of patients with mild and severe TBI. In mild TBI, particularly when there is little clinical or radiological evidence of injury, biomarker levels may more accurately predict the degree of brain injury and identify patients who are most likely to develop long term sequelae. These patients may benefit the most by being targeted for rehabilitation therapy. High biomarker levels may also identify patients who are at highest risk for secondary deterioration and who would benefit from repeat imaging, monitoring, and increased surveillance. It may also be useful for sports medicine (after concussion), worker’s compensation, and malingering [8]. For patients with severe TBI, a biomarker may be helpful to predict which patients are likely to experience secondary insults, such as raised ICP and help prognosticate. For these reasons, the last decade and, in particular, the last few years have witnessed an increasing interest in biomarkers for TBI [36].

Finding such a biomarker, though, has proved difficult for several reasons. The brain is a hugely complex organ and is protected by a selective blood brain barrier. Its functions are both qualitative and quantitative, while most biomarkers are purely quantitative. For example, lobar injury has different consequences for outcome compared with the same volume of tissue injury in the brainstem. Furthermore, several factors, including secondary brain insults, may occur independently of the primary brain injury and contribute to outcome, thereby reducing the sensitivity of biomarkers to predict outcome. Extracranial sources of the biomarker also may limit its specificity. The level of a biomarker in the serum may also reflect both the degree of cellular injury and/or the degree of blood brain barrier disruption. Finally, as in most aspects of TBI, less is known about the various clinical and pathophysiological factors that affect outcome in children when compared to their adult counterparts. Lack of precision in assessing outcome may further reduce the relationship between biomarker levels and outcome in pediatric studies.

Which biomarkers are used?

Biomarkers can be measured in cerebrospinal fluid (CSF), serum, and urine [24]. In broad terms, useful markers may be specific to brain injury or may be markers of inflammation or other biochemical and physiological processes. Brain-specific markers may be released in response to oxidative stress, inflammation, cerebral blood flow dysregulation, excitotoxicity, regeneration and repair, apoptosis, and cell death [63]. Although several potential biomarkers have been assessed in the last decade, for the purposes of this review, we will focus on the markers of structural brain injury that have received the most attention in recent clinical series. The neurospecific biomarkers that have been most thoroughly investigated are S100B, neuron-specific enolase (NSE), myelin basic protein (MBP), and glial fibrillary acid protein (GFAP). Arguably, the most promising marker, and the one most extensively studied in TBI, is S100B.


S100 calcium binding protein B, or S100B, is a protein of the S100 family. S100 protein is a type of low molecular weight protein found in vertebrates and is characterized by two calcium binding sites of the helix-loop (EF-hand type) conformation. The protein was first identified in 1965 [44]. It has two subunits, alpha and beta, with the αα found in striated muscle, heart, and kidney, αβ in glial cells, and ββ in high concentrations in astroglia [26, 33]. Although predominantly astroglial in origin, S100B is found in a range of neural cells [69]. Expression has also been found in white and brown fat, skin, skeletal muscle, melanoma, and glioblastoma cells [24, 27, 42, 78]. The name is derived from the fact that the protein is 100% [75] soluble in ammonium sulfate at neutral pH [41, 65]. It regulates homeostasis and enzyme activity and inhibits protein phosphorylation. Experimental evidence suggests that S100B secreted from glial cells may play neurotrophic or neurotoxic roles depending on its extracellular concentration [75]. At lower levels (nanomolar), it plays a physiologic role in promoting neuronal development, repair, regeneration, plasticity, and survival. At higher (micromolar) concentrations, it may be harmful to the brain.

The mechanism of release of S100B is still uncertain. It may be released by cell damage but is also actively secreted into the extracellular space by activated glial cells. It enters the serum through transient disruption of the blood brain barrier or via CSF circulation, is metabolized by the kidneys, and is excreted in the urine. It can be measured in serum and remains stable for hours without centrifugation and freezing. Elimination half-life is 30 min, biological half-life is 2 h, and levels rise immediately after primary trauma and decline rapidly over the next few hours [32]. Levels usually normalize within 24 h; therefore, persistent elevation of S100B may reflect ongoing or secondary cellular injury [59, 60]. A recent review by Ingebrigtsen et al. [30], with an emphasis on clinical utility in TBI, summarized that serum S100B levels correlated to both clinical measures of injury severity, neuroradiological findings, and outcomes, and they concluded that S100B protein was the most promising marker for evaluation of TBI in patients with mild head injury.

Immunoassays and reference ranges

Assay kits from several different manufacturers are available to measure S100B. Various immunoassays can be used: chemiluminescence immunoassay, electrochemiluminescence, immunoradiometric assay, enzyme-linked immunosorbent assay (ELISA), and immunofluorometric assay. Studies have examined S100B in CSF, serum, urine, and even by MR spectroscopy [10, 11, 35, 66].

In healthy adults (18–65 years), the median serum concentration of S100B is 0.052 μg/l (10th percentile = 0.023 μg/l; 90th percentile = 0.097 μg/l), and there are no sex-related differences [76]. Commonly recommended cut-off thresholds for detection of abnormally elevated S100B in adult TBI range from 0.2 to 0.5 ng/ml [8]. Higher thresholds may be less sensitive but more specific. Assessment of S100B in children may be more challenging though, and age within the pediatric group has to be considered in determining whether a S100B level is elevated. Spinella et al. [67] found that S100B levels in 136 healthy children had a mean of 0.3 μg/l (90% confidence interval, 0.03–1.47) and was correlated inversely with age (r = −0.32, p < 0.001). Median values for healthy children in the study by Portella et al. [56] were similar. Berger et al. [9] examined S100B in 64 controls in their study and found a median value for serum S100B of 0.016 ng/ml (using an ELISA). In a larger study of 394 children up to 18 years of age admitted for elective surgery (without brain injury), S100B had an upper reference range of 0.16 micrograms/l (using a Roche electrochemiluminescence kit) [14]. A reference range for children younger than 3 years of age was not determined in this study due to reduced numbers in this group. Although the authors of this study considered it to be the largest study of S100B in healthy children, an earlier study by Gazzolo et al. [22] had in fact recruited 1,004 healthy children between the ages of 1 month and 15 years and found some differing results. Their study showed in greater detail the age dependency of S100B. There was a varying pattern of S100B in normal children with age: The highest values were found in the first year of life, with a second peak between 7 and 13 years old and a decline thereafter. The authors speculated that high values in the first year of life may relate to early neurodevelopment or differences in blood brain barrier permeability. The later changes may also relate to the neurotrophic role of the protein and may correlate with height growth velocities and nerve elongation. Values were somewhat higher than in the other studies, with a median S100B in the first year of 0.81–0.9 μg/l (males–females) and 0.6–0.86 μg/l at age 6. The authors used a commercially available Sangtec Medical immunoluminometric assay. Of additional note in this study was the significant sex-related differences found in this study in contrast with others: S100B concentrations were generally higher in females, which may relate to sex-related differences in brain maturation. Some of the differences in reference values between these studies may be explained by the different methods used to assay S100B.

Clinical studies of S100B in adults

S100B and outcome

Woertgen et al. [77] examined S100B taken within 6 h and showed it to be a promising marker of injury severity and neuronal damage in severe head injury. S100B had a better predictive value for outcome than CT imaging and the GCS, having a sensitivity of 75% and specificity of 82%. Raabe et al. [58] sampled S100B early after admission and every 24 h for a maximum of 10 consecutive days. Patients who died had considerably higher S100B levels compared to those who survived: 2.7 versus 0.54 μg/l, respectively. They also showed a strong correlation between S100B levels and CT findings.

Townend et al. [72] found that a serum S100B concentration of greater than 0.32 μg/l predicted severe disability with a sensitivity of 93%, a specificity of 72%, and a negative predictive value of 99%. Lomas et al. [39] reviewed 200 studies of S100B to assess its value as a predictor of long term disability after head injury, of which 12 presented the best evidence to answer their clinical question. They concluded that a raised level of S100B is a marker of poorer long-term outcome after both major and minor head injury.

S100B and imaging

The association between S100B levels and imaging findings has been examined in several clinical studies [31, 45, 62, 73, 74]. If S100B was a highly sensitive marker of injury and radiological abnormalities, it might have a role in determining indications for CT scans after minor TBI. Ingebrigtsen et al. [31] examined the relationship between serum S100B levels and MRI findings following minor TBI. They concluded that S100B serum levels provided a valid measure of the presence and severity of TBI if performed within the first hours after minor head injury. The proportion of patients with detectable serum levels of S100B was significantly higher when MRI revealed a brain contusion. In a study of minor TBI based on six prospective studies, the sensitivity and negative predictive value of S100B for CT findings were 98.2% and 99.5%, respectively, and both were 100% for clinically relevant intracranial complications [74]. The authors estimated that the need for CT scanning could be reduced by 30% by integrating S100B with existing management guidelines. Romner et al. [62] also found that normal S100B levels predicted normal CT scan findings. However, in a later study [73], these authors found that extradural hematomas may be missed by normal S100B levels, presumably because these were little direct brain tissue injury, an important finding that may limit S100B as a screening tool to determine whether CT is indicated or not.

S100B may also have a role in predicting or assessing secondary insults. Raabe et al. [57] showed that secondary insults were predicted by late peaks in S100B levels prior to a rise in ICP or clinical deterioration. Hergenroeder et al. [28] also indicated that serum biomarkers may have utility for predicting secondary pathologies (e.g., elevated ICP) associated with TBI.

S100B in pediatric TBI

Elevation of S100B has been described in several pediatric neurological disorders, including TBI, meningitis, intraventricular hemorrhage of prematurity, and hypoxic-ischemic injury [10, 1721, 47, 68]. Most studies have focused on TBI. S100B in pediatric trauma requires separate consideration from the adult literature because pediatric brain injury differs from that of adults in several ways, and different reference ranges exist for children. However, fewer studies have been performed in pediatric trauma.

In a large study of 152 children who had sustained TBI, Berger et al. [10] examined initial and peak levels of S100B, NSE, and MBP. S100B correlated well with the 6-month Glasgow Outcome Scale (GOS)-Extended Pediatric score. In fact, most measures of these markers correlated well with each other. The study showed a negative and positive predictive value of 97% and 75%, respectively, for the simultaneous affect of all biomarkers on outcome. Spinella et al. [67] showed that a serum S100B level of 2.0 μg/l or greater within 12 h was associated with poor outcome with a sensitivity of 86% and a specificity of 95%.

Serum biomarkers have also been used to assess neurocognitive outcomes in inflicted versus noninflicted traumatic brain injury in young children. Beers et al. [7] examined S100B, NSE, and MBP using the GOS, Vineland Adaptive Behavior Scale, and an intelligence quotient measure to assess outcome. The inflicted TBI group performed more poorly compared to the noninflicted TBI group. Significant differences between groups were found for time-to-peak NSE, S100B, and MBP. Time-to-peak concentrations were significantly correlated with outcome measures. The time-to-peak concentrations were longer in inflicted TBI, perhaps representing delayed neuronal death.

Shore et al. [66] examined the issue of age in the comparison of biomarkers with assessment of injury severity. This study examined CSF S100B and NSE in the first 24 h after severe TBI (accidental and inflicted) and their relationship to GCS and GOS. Both markers showed overall significant, inverse correlation with GCS and GOS scores. However, when subgroup analysis was performed, it was found that correlation with GCS and GOS was only significant for children with non-inflicted TBI and >4 years of age.

Not all reports of S100B have been positive, however. Pickering et al. [55] studied emergency department measurement of urinary S100B in children less than 13 years old in 20 TBI patients and 15 controls with extracranial trauma. Urinary samples were taken within 12 h of injury. Despite detecting measurable S100B levels in the urine following head injury, similar levels occurred following extracranial trauma. Therefore, the authors concluded that urinary S100B is not useful as an early biochemical marker following head injury in children in this clinical scenario. Piazza et al. [54] also did not find a correlation between S100B and outcome in children with TBI. The sample sizes in both studies, however, were small.

S100B has also been used to measure brain injury in preterm neonates, hypoxic ischemic encephalopathy, and intraventricular hemorrhage [18, 19, 21]. For example, Gazzolo et al. [19] examined S100B in the urine of preterm newborns that subsequently died. S100B at first urination had a sensitivity of 100% and a specificity of 97.8% for predicting death, with a positive predictive value 78.6% and a negative predictive value of 100%. S100B (and NSE) has also been shown to correlate negatively with near-infrared spectroscopy recordings in critically ill children [71].

In summary, most studies agree that S100B is of prognostic value in pediatric TBI; however, this is not universal. As in adult TBI, some variability in these studies exists due to differences in sample size, range of ages studied, severity of TBI, timing of samples, degree of extracranial injury, and method of assessing outcome.

Extracranial sources of S100B

Although commonly thought of as a neurospecific biomarker, the specificity of S100B as a marker for brain injury may be limited by the fact that extracranial sources of S100 exist. Several studies in recent years have addressed this concern. Pelinka et al. [53] showed that bilateral femoral fractures in rats are associated with increased serum levels of S100B. S100B reached a peak 30–120 min after fracture (P < 0.001). They concluded that bone marrow is a potential extracerebral source of S100B. Others have confirmed elevated S100B in trauma patients without head injury and that this most commonly occurs with bone fractures (range, 2–10 μg/l) and thoracic contusions (range, 0.5–4 μg/l) [4]. Even burns produced an increase of S100B in this study (range, 0.8–5 μg/l). Increased S100B levels may also be related to ischemia and reperfusion of intra-abdominal organs [48].

Other studies report various clinical scenarios in which S100B is elevated without apparent brain injury, such as critical illness [64], cardiac ischemia [42], and hemorrhage [43], although it could be argued that each of these may be associated with occult brain injury.

The short half-life and potential extracranial sources of S100B may act as competing factors in choosing an optimal time for S100B sampling in the polytrauma TBI patient. On one hand, an early sample is much more likely to be elevated in keeping with the severity of the head injury, and so early sampling would be more sensitive for detecting brain injury. On the other hand, the contribution of extracranial sources also peaks in the early period, so a later sample may be more specific for brain injury.

Neuron-specific enolase

Neuron-specific enolase is a glycolytic isoenzyme located in central and peripheral neurons and neuroendocrine cells [34]. NSE has shown potential in both serum and CSF as a marker of TBI. Guzel et al. [25] showed that serum levels of NSE were significantly higher in patients who died. Berger et al. [9, 12] examined NSE concentrations in the CSF of infants and children after inflicted and noninflicted TBI and showed that NSE was increased following TBI and that levels were higher than in adult TBI. They hypothesized that this may be related to an increased susceptibility of the developing brain to cellular death after TBI. The study also showed later peak levels of NSE in patients with inflicted versus noninflicted TBI, thought to be due to delayed neuronal death in abuse victims secondary to an imbalance between pro- and antiapoptotic factors. This latent peak may allow for some discrimination between inflicted versus noninflicted TBI. This was again confirmed by a later study that examined serum levels of NSE [9]. Bandyopadhyay et al. [6] found that a serum NSE level of 21.2 ng/dl was 86% sensitive and 74% specific in predicting poor outcome in children with closed traumatic brain injury.

However, like all biomarkers, NSE has limitations. NSE may be limited by the occurrence of false positives with hemolysis [61] and extracranial sources [50]. Some studies suggest that NSE has limited utility as a prognostic index in TBI [49].

Myelin basic protein

Myelin basic protein is an abundant protein in central nervous system (CNS) white matter and is an essential structural component of CNS myelin. It is hypothesized that TBI-mediated axonal injury causes secondary structural damage to the adjacent myelin membrane, resulting in MBP degradation. This potentially could initiate myelin sheath instability and demyelination, which may further promote axonal vulnerability [37].

Serum levels appear to have good specificity but poor sensitivity for TBI. In a study of inflicted and noninflicted TBI in a pediatric population, initial peak MBP concentrations in the serum had a specificity of 96% but sensitivity of only 44% [9]. Children with inflicted TBI had later peak concentrations of MBP, similar to that seen with NSE, and higher MPB levels on admission compared to patients with noninflicted TBI.

Glial fibrillary acid protein

Glial fibrillary acid protein is an intermediate filament protein found in the cytoskeleton of astroglia. It is released after CNS damage and, thus, may serve as a marker of TBI. Lumpkins et al. [40] showed that a GFAP threshold of 1 pg/ml yielded 100% specificity and 50–60% sensitivity for TBI. Persistent elevation on day 2 was predictive of increased mortality. They also showed good specificity for CT-documented brain injury. This association between GFAP and clinical outcome in TBI has been confirmed by others [46]. GFAP may be particularly useful, especially in combination with other markers because the extracranial contribution to GFAP in multiple trauma appears to be minimal [52].

GFAP has been compared with S100B in TBI [51]. Both markers were higher in patients with elevated ICP and in patients who died. Both were also related to CT imaging findings, although slightly differently. The authors concluded that both biomarkers were useful noninvasive tests for identifying brain damage with some differences based on pattern of TBI and accompanying multiple trauma and/or shock.

Using paired or multiple biomarkers

Given the extracranial sources of biomarkers such as S100B and the failure to find a single biomarker that satisfies the criteria for reliable use as an accurate screening tool, some investigators have examined combinations of biomarkers to improve outcome prediction [9, 13, 38]. For example, Lo et al. [38] examined the predictive capacity of multiple biomarkers from different mediator families to determine whether combinations of two serum biomarkers could achieve better outcome prediction than individual biomarker levels in 28 children with TBI. Eight different neurospecific and inflammatory biomarkers (S100B, NSE, interleukin (IL)-6, IL-8, IL-10, SICAM, l-selectin, and endothelin) were quantified using ELISA on day 1 and compared with outcome assessed at 6 months after injury. None of the eight biomarkers assessed individually achieved an area under the ROC curve (AUC) of more than 0.95 for predicting unfavorable outcome, but five of the 20 biomarker pairs assessed achieved this high degree of outcome predictability. Two combinations of biomarkers, using S100B as the “screening marker” and either l-selectin or IL-6 as the “varying marker,” achieved an AUC of 0.98, and their specificity and sensitivity for unfavorable outcome prediction were 96% and 100%, respectively. They concluded that prognostic pairs combining serum levels of two biomarkers (inflammatory mediators and brain specific proteins) improved prediction for unfavorable outcome after childhood brain trauma compared to single markers. New technologies may create opportunities for evaluating current and future biomarkers of injury. One such emerging technique is multiplex bead technology, which allows quantification of multiple proteins simultaneously in very small amounts of biological samples [13, 36].


Over the last decade, although extensive research has not found an ideal brain biomarker, several biomarkers have shown clinical utility for prediction of imaging abnormalities and outcome after TBI. Choice of an appropriate biomarker may require selection according to the clinical context. The choice of a suitable biomarker in a particular pediatric clinical scenario requires knowledge of the origin of the biomarker, and in particular its extracerebral origins, and the relationship between age and biomarker levels in the normal population. Given the limitations of individual markers and the heterogeneity of various factors that influence outcome in TBI, combinations of markers may be better suited to guide management, warn of secondary injury, and help prognosticate in TBI.


This work was supported in part by a grant from the South African-Swedish Links Program (GUN 2072790).

Copyright information

© Springer-Verlag 2009