Glial fibrillary acidic protein as a biomarker in severe traumatic brain injury patients: a prospective cohort study

Study population

This was a prospective cohort study that examined adult severe TBI patients admitted
to a local neurotrauma center (Shanghai Institute of Head Trauma) in Shanghai, China,
from March 2011 to September 2014. The institute, located in the Department of Neurosurgery
of a tertiary teaching hospital (Renji Hospital), admits more than 400 TBI patients
per year and provides 24-hour neurosurgery facilities and specialized intensive care
for all patients.

The patients were included in the study based on established criteria, including older
than 18 years, an acute TBI less than 4 hours before admission, a GCS score of 3–8
after resuscitation, and abnormal head CT scan on admission. Exclusion criteria were
severe combined injury, unstable vital signs after resuscitation such as systolic
blood pressure less than 90 mmHg or oxygen saturation less than 94 %, and pregnant
women. Additionally, a total of 135 healthy blood donors who had no prior reports
of TBI or neurological diseases were enrolled as the control group, with serum samples
collected and analyzed for comparison.

This study was approved by the ethics committee of Renji Hospital, Shanghai Jiaotong
University School of Medicine, and written informed consent was obtained from the
control subjects and each severely brain-injured patient’s legal authorized representative
owing to their comatose state within 24 hours of being deemed eligible.

Treatment protocol

In Renji Hospital, nearly all injured patients were transported by emergency medical
services (EMS) professionals to the emergency department (ED) shortly after trauma
without intubation or sedation. Thus, we were able to evaluate the patients without
the interference of prehospital interventions. On their arrival at the ED, patients
with indications of TBI were neurologically assessed, received a computed tomography
(CT) scan of the head, and then were screened for enrollment in this study. Large
intracranial hematomas and hemorrhagic contusions, once confirmed by head CT, were
evacuated immediately and the patients then transferred into the neurosurgical intensive
care unit (NICU).

All patients were monitored and managed according to a standardized institutional
protocol including hypothermia and other intensive care treatments. An intraventricular
catheter or an intraparenchymatous microtransducer was used for the continuous measurement
of intracranial pressure (ICP). Mild hypothermia (34–35 °C) was rapidly induced using
cooling blankets (MTRE Advanced Technologies, Trevose, PA, USA) placed below and above
the patients, which was maintained for as long as 3 days. The cooled patients were
tracheotomized for ventilation, continuously sedated (chlorpromazine, 5–10 mg/h) and
paralyzed (atracurium, 10–40 mg/h). During hypothermia, the systolic blood pressure
of the patient was maintained above 90 mmHg, peripheral oxygen saturation greater
than 95 %, ICP less than 25 mmHg, and cerebral perfusion pressure at a level of 50–70 mmHg.
After a 3-day maintenance period, the patients were passively rewarmed to 36–37 °C
at a rate no greater than 0.5 °C every 4 hours, carefully individualized by the patients’
ICP level. If the rebound of ICP was indicated, then a much slower rewarming rate
(0.5 °C every 8 hours) and the extensive use of other ICP-reducing measures were
employed.

Intracranial hypertension during the whole hospitalization course was resolved progressively
with a set of stepwise strategies including head elevation, ventilation control, sedation,
hyperosmolar therapy, and ventricular drainage. If the raised ICP was resistant to
these therapies, then decompressive craniotomy and/or further surgical operation were
adopted. In addition, serum glucose, blood gases, and serum electrolytes were assessed
regularly, with treatment provided for any abnormal findings. After hypothermia was
discontinued, any patients with fever, defined as body temperature more than 38 °C,
were examined for infectious or noninfectious causes and treated with ice packs around
the neck and limbs, sponging the body with alcohol, use of paracetamol, or a combination.

Biomarker determination

Blood samples (3–5 ml) were taken by the study investigators (JL and GG) using red
top tubes at the time of admission (within 4 hours post injury) to the hospital and
then each morning (6:00–8:00 am) for the first 5 days in the NICU. After collection,
specimens were maintained at room temperature for 30–60 minutes and centrifuged for
10 minutes at 4000 rpm; sera were stored at ?80 °C until analysis.

Serum GFAP concentration was measured using an enzyme-linked immunosorbent assay (ELISA)
via a commercial kit according to the manufacturer’s protocol (Biovendor, Candler,
NC, USA). Briefly, standards, quality controls (QC-Low and QC-High), and serum samples
were incubated in microplate wells precoated with polyclonal anti-human GFAP antibody
for 2 hours, followed by 1 hour incubation with a biotin-labeled monoclonal anti-human
GFAP antibody solution and 1 hour incubation with a streptavidin–horseradish peroxidase
conjugate. Between each step, the plate was thoroughly washed four times with washing
buffer. After the last washing step, the remaining conjugate was allowed to react
with the substrate solution for an additional 15 minutes. The reaction was stopped
by the addition of an acidic solution, and the absorbance of the resulting yellow
product was measured by reading the ELISA plate at 450 nm. The absorbance was proportional
to the concentration of GFAP. A standard curve (0.25–25 ng/ml) was constructed by
plotting absorbance values against concentrations of standards, and concentrations
of unknown samples were determined using this standard curve. The lower limit for
detection in the assay was 0.045 ng/ml. All concentrations below this limit of detection
were reported as zero. The coefficients of variation for inter- and intra-assay variability
were 5.7 % and 5.1 %, respectively. All samples were assayed in duplicate, and average
results were used for analysis.

Data collection and outcome assessment

Demographic and clinical data were collected from patient medical records including
sex, age, and cause of injury, CT scan, admission GCS score, pupil reactions, and
surgery. Initial head CT scans were analyzed and classified according to the criteria
raised by Marshall and associates 19] to determine brain injury severity (Diffuse injury I: no visible intracranial pathology;
Diffuse injury II: cisterns present with midline shift of 0–5 mm, no focal lesion
of??25 ml; Diffuse injury III: cistern compressed or absent with midline shift of
0–5 mm, no focal lesion of??25 ml; Diffuse injury IV: midline shift of??5 mm, no
focal lesion of??25 ml; Evacuated mass lesion: any lesion surgically evacuated; Non-evacuated
mass lesion: lesion of??25 ml, not surgically evacuated). Since the recent trial
by Clifton et al. 20] reported possible differential effects of hypothermia among patients with surgical
lesion compared with diffuse injury, the surgical lesion and diffuse injury groups
were further defined based on the above CT classification in our study to examine
the potential correlation between serum GFAP concentration and the pathological types
of brain injury and the role of hypothermia.

Neurological outcome assessment was performed by a blinded investigator (CW) who was
unaware of the patient’s clinical history and biomarker data using the Glasgow Outcome
Scale (GOS) at 6 months after injury with the use of a structured interview by direct
patient contact or via telephone 21]. The GOS is a five-category scale used for assessing the neurological outcome after
brain injury as follows: 1, death; 2, vegetative state—unable to interact with the
environment; 3, severe disability—unable to live independently but able to follow
commands; 4, moderate disability—capable of living independently but unable to return
to work or school; and 5, good recovery—able to return to work or school. For statistical
analyses, the outcome was further dichotomized in death (GOS 1) versus survival (GOS
2–5) and unfavorable (GOS 1–3) versus favorable (GOS 4–5).

Statistical analysis

Basic descriptive statistics were used to describe the data, including standard measures
of central tendency and dispersion for continuous data (mean, standard deviation (SD),
median, and range) and frequencies and proportions for categorical variables. Comparisons
were made using a chi-squared test, Fisher’s exact test, unpaired Student t test, or Wilcoxon rank-sum test, if appropriate. Significance was set at P??0.05.

To analyze the association of serum GFAP levels during the first 6-day period and
neurological outcomes at 6-month post injury, we use the longitudinal multiple linear
regression models (also called cross-sectional time series) that controlled for confounding
clinical parameters, which have the ability to provide efficient statistical power
to analyze continuous outcomes (biomarker levels) that are measured at regular time
intervals, which are the same for all subjects 22]. Serum GFAP levels were logarithmically transformed to achieve approximately normal
distribution and treated as the dependent variable. Outcome group (i.e., death or
unfavorable outcome, as defined above) and time of blood sampling were included in
each model as independent variables. Additional variables, including age, sex, cause
of injury, time from injury to admission, surgery, baseline GCS score, Marshall CT
classification, and pupil reactions were evaluated as potential confounders and included
in the models. These models allowed us to determine whether the injured patients from
dichotomized outcome groups had differing serum GFAP levels over time.

Receiver operating characteristic (ROC) curve analysis was performed to determine
the discriminatory characteristics of serum GFAP during the course from the day of
admission to day 5. ROC curves were secondarily developed along with the respective
diagnostic parameters of the coordinates of the curve (i.e., sensitivity, specificity,
positive and negative predictive values), the area under the curve (AUC), and the
optimal cutoff point for the best overall predictive ability (i.e., the sum of the
sensitivity and the specificity for prediction of outcome was maximal). The AUC value
was used to compare the ability of admission serum GFAP to predict 6-month lethal
or unfavorable outcomes compared with the clinical predictors in the IMPACT core model
(i.e., age, GCS score, or pupil reactions).