Preliminary Evaluation of the Scandinavian Guidelines for Initial Management of Minimal, Mild, and Moderate Head Injuries with Glial Fibrillary Acidic Protein

Glial fibrillary acidic protein (GFAP) has become the most promising biomarker for detecting traumatic abnormalities on head computed tomography (CT) in patients with traumatic brain injury (TBI), but most studies have not addressed the potential added value of combining the biomarker with clinical variables that confer risk for intracranial injuries. The Scandinavian Guidelines for Initial Management of Minimal, Mild, and Moderate Head Injuries in Adults were the first clinical decision rules in the field with an incorporated biomarker, the S100 astroglial calcium-binding protein B (S100B), which is used in the Mild (Low Risk) group defined by the guidelines. Our aim was to evaluate the performance of the guidelines when S100B was substituted with GFAP. The sample (N = 296) was recruited from the Tampere University Hospital's emergency department between November 2015 and November 2016, and there were 49 patients with available GFAP results who were stratified in the Mild (Low Risk) group (thus patients undergoing biomarker triaging). A previously reported cutoff of plasma GFAP ≥140 pg/mL was used. Within the Mild (Low Risk) group (n = 49), GFAP sensitivity (with 95% confidence intervals in parentheses) for detecting traumatic CT abnormalities was 1.0 (0.40–1.00), specificity 0.34 (0.19–0.53), the negative predictive value (NPV) 1.0 (0.68–1.00), and the positive predictive value (PPV) 0.16 (0.05–0.37). The sensitivity and specificity of the modified guidelines with GFAP, when applied to all imaged patients (n = 197) in the whole sample, were 0.94 (0.77–0.99) and 0.20 (0.15–0.28), respectively. NPV was 0.94 (0.80–0.99) and PPV 0.18 (0.13–0.25). In the Mild (Low Risk) group, none of the patients with GFAP results below 140 pg/mL had traumatic abnormalities on their head CT. These findings were derived from a small patient subgroup. Future researchers should replicate these findings in larger samples and assess whether GFAP has added or comparable value to S100B in acute TBI management.


Introduction
Traumatic brain injury (TBI) is among the most common reasons for seeking emergency department (ED) care and a leading cause of morbidity globally. 1,2ead computed tomography (CT) is the imaging modality of choice for identifying patients with acute traumatic intracranial pathology in the ED. 3,4However, there are reasons to be judicious in referrals for head CT, such as reducing exposure to radiation 5 and reducing healthcare costs. 6Researchers are exploring blood-based biomarkers as an alternative approach to diagnose intracranial traumatic lesions.These efforts aim to discover and validate the most promising blood-based biomarker for detecting acute intracranial traumatic lesions in patients with TBI. 7,8he S100 astroglial calcium-binding protein B (S100B) was the first blood biomarker to be combined into the risk stratification for triaging patients with TBIs for referral for head CT in the 2013 Scandinavian Guidelines for Initial Management of Minimal, Mild, and Moderate Head Injuries in Adults 9 (i.e., Scandinavian Guidelines).][12] Glial fibrillary acidic protein (GFAP) has emerged as the leading blood biomarker candidate for TBI, 13 it has been included in the U.S. Food and Drug Administration (FDA)-approved biomarker kit for identifying traumatic CT abnormalities in patients with mild traumatic brain injury (mTBI) in 2018, [14][15][16] and it was also included in a FDA-cleared rapid test for the same use. 17Per a meta-analysis of nine studies on GFAP prediction of CT abnormalities, a threshold of 22 pg/mL maximized sensitivity at 93% with a specificity of 36%. 8However, the differences in assay choices and sample types (i.e., serum or plasma) between studies make it difficult to assess the threshold-performance relationship of GFAP. 8lthough studies have shown GFAP to outperform S100B when studied as a sole predictor for traumatic CT abnormalities, 18,19 GFAP has recently been incorporated into clinical guidelines, the French Recommendations for the management of patients sustaining mTBI by the French Society of Emergency Medicine and the French Society of Anaesthesiology and Critical Care Medicine. 20GFAP is detectable within 1 h post-TBI and peaks within 20-24 h with a half-life of 24-48 h. 21 clinically reliable sampling time for GFAP has not been determined, but a longer half-life allows it to be determined in a wider time window than S100B, theoretically making it more useful in various clinical situations.
The Scandinavian Guidelines are designed to guide the initial management of minimal, mild, and moderate head and brain injuries.After the initial clinical evaluation, a decision is made about whether to send the patient for head CT.According to the Scandinavian Guidelines, injuries that are Moderate, Mild (High Risk), and Mild (Medium Risk) are referred for CT.Injuries that are deemed Mild (Low Risk) are triaged by the examination of S100B levels (if the injury is within 6 h), and if S100B is elevated, they are sent for head CT.In this study, we evaluated GFAP instead of S100B and extended the sampling time to 24 h in the Mild (Low Risk) group as defined by the 2013 Scandinavian Guidelines for the emergency management of TBIs. 9 This study relied on data collected from a prospective cohort study designed to validate the Scandinavian Guidelines. 22We hypothesized that the modified guidelines with GFAP would have high sensitivity and low-medium specificity in detecting acute intracranial abnormalities.We examined whether the modified guidelines, with GFAP, could safely reduce the frequency of CT scanning of mTBI patients without acute intracranial abnormalities.

Participants
The study sample was recruited from the Tampere University Hospital ED (Tampere, Finland) between November 2015 and November 2016.All consecutive adults ( ‡18 years) with acute (£24 h) TBI were eligible for inclusion.Minimum criteria for TBI were determined as either blunt injury to the head or acceleration/deceleration-type injury resulting in a witnessed loss of consciousness, disorientation, or amnesia and a Glasgow Coma Scale (GCS) score of 13-15 as assessed 30 min after injury. 23During the study period, 3067 adult patients with TBIs (mean age = 56.9years, standard deviation [SD] = 23.2,median = 58.0,range = 18-103; women = 46.8%)were treated in the ED and 325 (10.6%) consented to participate in the study.After excluding patients with ED admission >24 h post-injury, a total of 296 patients (mean age = 61.1 years, SD = 22.7, median = 67.0,range = 18-100; women 5 49.0%) were enrolled in a prospective study designed to validate the Scandinavian Guidelines. 22njury mechanism, post-injury signs and symptoms, and the findings of the physical examination performed in the ED were described by the on-call physician in detailed case reports of each patient.All enrolled patients provided informed written consent according to the Declaration of Helsinki.The study was approved by the Ethics Committee of Pirkanmaa Hospital District, Tampere, Finland (ethical code: R15045).

Outcomes
An acute traumatic lesion on head CT was determined to be the primary outcome.Secondary outcomes were delayed complications resulting from the head or brain injury, including ED or hospital readmission, repeat head CT, or death within a week after injury.At the 1-week follow-up, a study nurse identified possible complications by a phone call and medical record review.

Head computed tomography imaging
Referral to a head CT was based on an evaluation by the on-call physician and applying the Scandinavian Guidelines. 9Non-contrast head CT was performed with a 64-row CT scanner (Lightspeed VCT; GE, Waukesha, WI).The findings were systematically coded by a neuroradiologist (K.B.) based on the National Institute of Neurological Disorders and Stroke Common Data Elements. 24omarker analyses Venous blood samples were originally drawn from all enrolled patients, and a part of the blood was analyzed at Tampere University Hospital (Tampere, Finland) for a previous study. 22The remaining serum and plasma samples were immediately frozen at À70°C for future use.Samples were sent to the Sahlgrenska University Hospital, Mo ¨lndal, Sweden (transfer in 20 kg of dry ice by a courier) for further analyses.On September 14-15, 2019, the samples were analyzed using the GFAP Discovery Kit (Quanterix, Billerica, MA) on an HD-1 Simoa instrument to determine plasma GFAP levels.The lower limit of detection and lower limit of quantification were 0.211 and 0.686 pg/mL, respectively.The laboratory technicians performing the analyses were blinded to the clinical data.The details of our GFAP analytics are described more comprehensively in a previous publication and in the Supplementary Material. 25e Mild (Low Risk) group According to the Scandinavian Guidelines, patients are divided into severity classifications for triaging as follows: 1) Minimal; 2) Mild (Low Risk); 3) Mild (Medium Risk); 4) Mild (High Risk); and 5) Moderate (see Supplementary Fig. S1).S100B levels are used to determine the need for head CT in the Mild (Low risk) group, which is comprised of either 1) patients with GCS of 15 who have suspected/confirmed loss of consciousness, repeated vomiting, or both or 2) patients with GCS of 14. Patients who are ‡65 years of age and who are on antiplatelet medication cannot be classified as Low Risk; they are classified as Medium Risk and are sent for head CT.Patients cannot be in the Low Risk classification if they present with post-traumatic seizures, focal neurological deficits, clinical signs of depressed or basal skull fracture, shunt-treated hydrocephalus, therapeutic anticoagulation, or a coagulation disorder (i.e., patients with any of those features are referred for head CT). 9 In this study, we examined the performance of plasma GFAP in the Mild (Low Risk) group using a cut-off level of 140 pg/mL.There are no validated cut-off scores for the research-use assays used in this study.This cutoff level was previously derived from another study cohort that was analyzed for GFAP in the same research laboratory using the Human Neurology 4-Plex A assay on a Simoa HD-1. 26We opted to use plasma GFAP levels in this study instead of serum because of the lack of previously derived cutoffs with a similar biomarker assay for serum GFAP.The time limit for the interval between injury and blood sampling was moved from the guideline's original 6-h limit to 24 h for the purposes of this study because of the slower release of GFAP after TBI. 21

Statistical analyses
Statistical analyses were performed with the Statistical Package for Social Sciences software program (IBM SPSS Statistics for Windows, Versions 22.0-25.0;IBM Corp., Armonk, NY).Distributions of sample characteristics and the plasma GFAP (P-GFAP) in the Mild (Low Risk) group were examined using histograms and Shapiro-Wilk tests of normality, and nearly all distributions were non-normal.Descriptive statistics (frequency [n], percentage, median, interquartile range, and range) were used to describe the sample characteristics.Bivariate associations between the biomarkers were calculated using Spearman rho correlation coefficients.A p < 0.05 was indicative of a statistically significant finding.
Sensitivities were calculated by dividing the number of patients with a head CT indication (indicated by either the P-GFAP result in the Mild [Low Risk] group or the modified guidelines as presented in Supplementary Fig. S1) and a positive head CT result by the total number of positive head CT results, and the specificities by dividing the number of patients without a head CT indication and a negative head CT result by the total number of negative head CT results.
Positive predictive values (PPVs) were calculated by dividing the number of patients with a head CT indication and a positive head CT result by the total number of patients with a head CT indication, and the negative predictive values (NPVs) by dividing the number of patients without a head CT indication and a negative head CT result by the total number of patients without a head CT indication.Confidence intervals (CIs) were calculated by Clinical Calculator 1 of the VassarStats website, 27 using the continuity-corrected Newcombe-Wilson score method. 28

Results
Patients (N = 296) were divided into subgroups according to the Scandinavian Guidelines (see Supplementary Fig. S1).Sample characteristics of the total sample (N = 296) are presented in a previous study. 22Of the 296, 197 underwent head CT (see row 3 of Supplementary Fig. S1).Sample characteristics for the total sample of 197 patients who underwent head CT are provided in Table 1. Figure 1 presents the sample selection for this study, and the characteristics of the patients with available GFAP results in the Mild (Low Risk) group (n = 49) are presented in Table 1.Of these 49 patients, 36 (73.5%) underwent head CT scanning and 4 (11.1%) had acute CT abnormalities.The most common CT abnormality was traumatic subarachnoid hemorrhage (8.3%; n = 3).
Figure 2 presents the outcomes of using P-GFAP in the Scandinavian Guideline for the Mild (Low Risk) group (n = 49).Within the group, P-GFAP levels were elevated in 33 of 49 patients (67.3%).Among those patients, 25 (75.8%)underwent CT imaging.Among the 16 patients (32.7%) with P-GFAP levels <140 pg/mL, 11 (68.8%)still underwent CT imaging, and none of the scans showed traumatic abnormalities.
For the 26 (34.7%) patients in the Mild (Low Risk) group who did not have P-GFAP data, a head CT was indicated because the guidelines recommend imaging patients in the Mild (Low Risk) group if biomarker results are not attainable.Of those patients, 19 (73.1%) were imaged and two had traumatic abnormalities on their head CT.Characteristics of patients with traumatic CT abnormalities in the Mild (Low Risk) group are presented in Table 2.
Within the guideline-specified Minimal TBI group, there were 64 patients with available P-GFAP results.The biomarker results and brain imaging results of these patients are presented in the online Supplementary Table S5.When these patients were combined with those in the Mild (Low Risk) group, there were a total of 55 patients with available P-GFAP and brain imaging results.Within this combined group (n = 55), the sensitivity of P-GFAP in detecting intracranial CT abnormalities would have been 1.0 (95% CI, 0.52-1.00),specificity 0.39 (95% CI, 0.26-0.54), the NPV 1.0 (95% CI, 0.79-1.00),and the PPV 0.17 (95% CI, 0.07-0.33),as presented in Supplementary Table S6.
In the 197 patients who underwent imaging, there were 2 patients with traumatic CT abnormalities who did not have a guideline-based indication to undergo a head CT but who underwent a head CT because the physician ordered one.These patients were in the minimal TBI group.Thus, according to the guidelines, they would not need blood sampling or a head CT.Both, however, underwent blood sampling and both patients exceeded the cutoff used in this study  (P-GFAP, 886 and 2860 pg/mL).The modified guidelines' sensitivity for detecting traumatic CT abnormalities, calculated within the 197 imaged patients in the whole sample, was 0.94 (0.77-0.99) when using GFAP in the Mild (Low Risk) group.The specificity was 0.20 (0.15-0.28),PPV was 0.18 (0.13-0.25), and NPV was 0.94 (0.80-0.99; see Supplementary Table S1).

Discussion
There is a growing body of research supporting the use of blood biomarkers for detecting intracranial CT abnormalities in patients with acute TBIs, but their integration into clinical practice has only just begun. 7,29The ALERT-TBI study that led to the first FDA-approved biomarker test for the acute management of patients with mTBIs, although considered pivotal, has received criticism for its lack of comparison with clinical decision rules, especially the Scandinavian Guidelines with S100B. 15,30We addressed this gap in the literature by combining GFAP with the clinical variables included in the Scandinavian Guidelines and examining the sensitivity and specificity of the Scandinavian Guidelines for detecting ''low-risk'' TBI patients who have traumatic CT abnormalities.FIG. 2. Outcomes of applying P-GFAP in the Scandinavian Guideline for the Mild (Low Risk) group.CT+/-= presence or absence of acute traumatic lesion on head computed tomography (CT).Complication+/-= presence or absence of hospital or ED readmission, repeat head CT, or death within 1 week after injury.P-GFAP, plasma glial acidic fibrillary protein.The modified Scandinavian guidelines with GFAP had a sensitivity of 0.94 (0.77-0.99) and a specificity of 0.20 (0.15-0.28) in detecting traumatic CT abnormalities within the imaged patients in the whole sample (n = 197).PPV and NPV were 0.18 (0.13-0.25) and 0.94 (0.80-0.99), respectively.In comparison, the performance of the Scandinavian Guidelines with S100B was almost identical in our previous publication (sensitivity 0.94, specificity 0.19, and PPV and NPV 0.18 and 0.94, respectively)-albeit with a larger sample. 22These findings suggest that the Scandinavian Guidelines perform in comparable ways using S100B and GFAP-at least in the present sample.Within the Mild (Low Risk) group (n = 49), a GFAP level pg/mL was 1.0 (0.40-1.00) sensitive and 0.34 (0.19-0.53) specific for traumatic CT abnormalities, with an NPV and PPV of 1.0 (0.68-1.00) and 0.16 (0.05-0.37), respectively.Incorporating GFAP into a guideline-based subgroup of patients may have led to lowering the number of false-positive test results, compared to relying solely on the biomarker concentrations in a broad spectrum of TBI patients.2][33][34] By using elevated GFAP levels as an indication for imaging only in the pre-specified subgroup of patients and excluding the patients less likely to require a head CT from biomarker measurements, the probability of a falsepositive test result is reduced.
In our subanalyses, we tried extending the use of biomarker measurements to the lowest risk group, which is the Minimal TBI group as defined by the Scandinavian Guidelines.According to the guidelines, patients in this group could be discharged home without imaging or blood sampling.However, our previous results revealed that 2 patients in this group had traumatic CT abnormalities on their head scans.If P-GFAP had been used in this group, it would have indicated the need for imaging in these cases.Nevertheless, this expansion of biomarker usage would have led to an additional 24 patients (8% of the total sample of 296) undergoing imaging if the elevated GFAP levels had been used in a clinical pathway within the Minimal TBI group.The ALERT-TBI study reported greater sensitivity and specificity for the FDA-approved biomarker test compared to our results, but the study sample only consisted of patients with TBIs who had undergone a head CT. 14 This may have resulted in an enhancement of specificity, because the patients pre-determined not to need a head CT had been excluded.In contrast, our study design was prospective and S100B values may have influenced the clinicians' decision whether to refer a patient for a head CT, possibly resulting in the inclusion of participants with higher biomarker results who would have been excluded in a study design similar to ALERT-TBI.
S100B is the first biomarker to be included in a clinical decision rule for the acute management of TBI patients in the ED. 9 The guideline is used in Europe, but the use of S100B is not recommended by either the American College of Emergency Physicians 3 or the Eastern Association for the Surgery of Trauma 35 for pre-head CT screening in patients with TBIs.GFAP has been shown to outperform S100B in several studies when studied as a sole predictor for traumatic CT abnormalities in patients with mTBIs, 15,18,19,36 but a CENTER-TBI study was the first to assess the added value of combining the biomarkers with a wide range of clinical variables. 15That study used a previously developed clinical decision rule variable combination that was not based on the Scandinavian Guidelines; therefore, it differed somewhat from the variables used in our study by including additional risk factors, such as high-risk injury mechanisms, retrograde and post-traumatic amnesia, headache, intoxication, significant injury above clavicles, and deterioration of the GCS score.Unlike the Scandinavian Guidelines, antiplatelet medication in older adults or shunt-treated hydrocephalus were not considered as risk factors.The results of that study showed comparable discriminative ability of GFAP compared to S100B (area under the curve of 0.92 vs. 0.90, respectively).Their study did not implement any cut-off levels in the analyses; and thus, the biomarkers' performances could not be assessed further.
We extended the time limit for blood sampling from 6 h post-injury to 24 h post-injury because of the slower kinetics of GFAP compared to S100B in mTBI. 21,36Theoretically, the longer half-life of GFAP could make its use more efficient than S100B in the acute management of mTBI, allowing a longer measurement time window in EDs.On the other hand, the slightly slower kinetics of GFAP may influence its accuracy in very early measurement.There may be value in each biomarker at different times post-injury.
Our study has limitations, the most notable being the small sample size and the limited number of abnormal head CTs among patients with P-GFAP results in the Mild (Low Risk) group.This limitation is evident statistically, for example, in the CI for sensitivity for P-GFAP in the Mild (Low Risk) group.Future research is needed to replicate these results in a larger sample.Additionally, we did not test any other cutoffs for GFAP in our analyses in addition to the pre-selected cutoff.The cut-off selection was based on a study using a biomarker assay on the same instrument in the same laboratory that reported exploratory cutoffs for GFAP. 26The cutoff used was much higher than that used in another study combining serum GFAP with clinical decision rules (i.e., 30 pg/mL). 37However, the differences in biomarker assays between the study and ours make the lower cutoff inapplicable in our sample.Serum and plasma GFAP levels are highly correlated, but there are assay-dependent differences between the overall plasma and serum levels of GFAP in the same patients. 25,38This has resulted in plasma GFAP levels being much higher than serum GFAP levels in our study sample (see Supplementary Tables S7 and S8).

Conclusion
GFAP is often considered to be the best emerging blood biomarker for head CT abnormalities in patients with head and brain injuries, 16 and our findings suggest that when included in the subgroup of patients defined by the Scandinavian Guidelines, the biomarker can identify patients with CT abnormalities with a low but reasonable level of specificity.Future researchers should seek to replicate these findings and continue exploring the potential utility of the biomarkers.

FIG. 1 .
FIG.1.Sample selection.Because of the pragmatic study design, plasma GFAP was unavailable for 26 subjects, because storage samples were not always collected despite being required by the study protocol.ED, emergency department; P-GFAP, plasma glial fibrillary acidic protein; CT, computed tomography.

Table 1 .
Study Sample Characteristics for the Mild (Low Risk) Group (n = 49) and the Patients With Available CT Image Results (n = 197) in the Total Sample CT, computed tomography; ED, emergency department; GFAP, glial fibrillary acidic protein; GCS, Glasgow Coma Scale; IQR, interquartile range; Mdn, median; TBI, traumatic brain injury.

Table 2 .
Characteristics of the Mild (Low Risk) Group Patients With Traumatic Abnormalities on Head CT Blood was not stored for two of the patients.GCS, Glasgow Coma Scale; ED, emergency department; GFAP, plasma glial fibrillary acidic protein; CT, computed tomography; GLF, ground-level fall; LOC, loss of consciousness; PTA, post-traumatic amnesia; SAH, subarachnoid hemorrhage; SDH, subdural hematoma; TAI, traumatic axonal injury.