Abstract
The aim of the study was to examine the ability of eight protein biomarkers and their combinations in discriminating computed tomography (CT)-negative and CT-positive patients with traumatic brain injury (TBI), utilizing highly sensitive immunoassays in a well-characterized cohort. Blood samples were obtained from 160 patients with acute TBI within 24 h of admission. Levels of β-amyloid isoforms 1–40 (Aβ40) and 1–42 (Aβ42), glial fibrillary acidic protein (GFAP), heart fatty-acid binding protein (H-FABP), interleukin 10 (IL-10), neurofilament light (NF-L), S100 calcium-binding protein B (S100B), and tau were measured. Patients were divided into CT-negative (n = 65) and CT-positive (n = 95), and analyses were conducted separately for TBIs of all severities (Glasgow Coma Scale [GCS] score 3–15) and mild TBIs (mTBIs; GCS 13–15). NF-L, GFAP, and tau were the best in discriminating CT-negative and CT-positive patients, both in patients with mTBI and with all severities. In patients with all severities, area under the curve of the receiver operating characteristic (AUC) was 0.822, 0.817, and 0.781 for GFAP, NF-L, and tau, respectively. In patients with mTBI, AUC was 0.720, 0.689, and 0.676, for GFAP, tau, and NF-L, respectively. The best panel of three biomarkers for discriminating CT-negative and CT-positive patients in the group of all severities was a combination of GFAP+H-FABP+IL-10, with a sensitivity of 100% and specificity of 38.5%. In patients with mTBI, the best panel of three biomarkers was H-FABP+S100B+tau, with a sensitivity of 100% and specificity of 46.4%. Panels of biomarkers outperform individual biomarkers in separating CT-negative and CT-positive patients. Panels consisted mainly of different biomarkers than those that performed best as an individual biomarker.
Keywords: biomarkers, computed tomography, traumatic brain injury
Introduction
Traumatic brain injury (TBI) is a major global health problem with more than 50 million new cases annually, and the incidence is rising among both young and elderly people.1–3 Mild TBI (mTBI) represents 80–90% of all TBIs.4 Some patients meeting the clinical diagnostic criteria for mTBI may have significant traumatic intracranial findings on head computed tomography (CT), requiring neurosurgery.5 To determine the need for a head CT is sometimes challenging due to commonly occurring confounding factors.
The clinical significance of blood-based biomarkers in TBI for detecting patients with traumatic intracranial findings is still unclear. It is likely that instead of focusing on the use of a single biomarker, optimized combinations of biomarkers should be sought for different clinical questions due to the complexity of the brain and heterogeneity of TBIs.
Guidelines on when to perform a head CT scan have been introduced into clinical practice to help in screening patients who may have significant intracranial injuries.6–8 Still, the majority of patients scanned following these recommendations show a negative CT.9,10 Thus, improved regimens for decision making regarding CT scanning are warranted to decrease radiation load and costs. The Scandinavian guideline for management of mild head injury recommends the use of biomarker S100 calcium-binding protein B (S100B) in patients with mTBI who are admitted to hospital within 6 h after the injury.11 However, S100B is expressed in multiple extra-cerebral tissues and its levels increase e.g., after extracranial injuries12 and physical exercise13.
Recent research has found several novel protein biomarkers with more brain-specific origin, which thus could be more suitable for assessing the need for a CT following TBI. Glial fibrillary acidic protein (GFAP), which is expressed in the cytoskeleton of glial cells,14 has been studied widely in detecting acute intracranial injuries after a TBI, with promising results.15–18 Both S100B and the combination of GFAP and ubiquitin C-terminal hydrolase L1 (UCH-L1) have showed promise as biomarkers in screening for CT-positivity/negativity in patients with acute TBI.19–21 Heart fatty-acid binding protein (H-FABP), a cytosolic trafficking protein,22 is expressed in the heart but also in the brain, and has been shown to predict TBI-related intracranial pathologies.23 Anti-inflammatory mediator interleukin 10 (IL-10) also has shown promise in differentiating CT-positive from CT-negative patients with mTBI.24
Other brain-related protein biomarkers that have been studied in the diagnostics of TBI are β-amyloid isoforms 1–40 (Aβ40)19,20 and 1–42 (Aβ42)17,25 reflecting amyloid precursor protein metabolism, neurofilament light chain (NF-L) being abundant in the long myelinated subcortical axons,26,27 and microtubule-associated protein tau located in the axonal cytoskeleton.17,28 All these proteins have been mainly studied in the subacute/chronic stage of TBI and their utility in predicting intracranial pathologies on acute CT after TBI is poorly known.
We investigated the ability of Aβ40 and Aβ42, GFAP, H-FABP, IL-10, NF-L, S100B, and tau and their combinations in discriminating CT-negative (CT−) and CT-positive (CT+) patients with TBI, utilizing modern highly sensitive immunoassays23,24,29 in a well-characterized study cohort.15,30–32
Methods
Study population
This prospective study was part of the EU-funded TBIcare (Evidence-based Diagnostic and Treatment Planning Solution for Traumatic Brain Injuries) project, where we recruited patients with TBIs of all severities at Turku University Hospital, Finland, during November 2011 to October 2013 as described elsewhere.15 All patients were treated according to local guidelines based on existing international guidelines and recommendations.33 All patients were examined and classified for the presence of extracranial injuries using the Injury Severity Score (ISS).
Biomarker analyses
Blood samples for Aβ40, Aβ42, GFAP, H-FABP, IL-10, NF-L, S100B, and tau were obtained within 24 h of admission. Plasma H-FABP and IL-10 were analyzed using the K151HTD and K151QUD kits, respectively, from Meso Scale (Meso Scale Diagnostics, Rockville, MD) and S100B was measured using EZHS100B-33K kit from Millipore (Millipore, Billerica, MA) according to the manufacturers' recommendations. For H-FABP, the lower limit of detection (LLoD) was 0.103 ng/mL and the calibration range was 0.137-100 ng/mL. The test has not yet been fully validated by Meso Scale and therefore there is no established lower limit of quantification (LLoQ). For IL-10, the LLoD was 0.04 pg/mL, with the LLoQ at 0.298 pg/mL with a calibration range between 0.0774 and 317.0 pg/mL. For the S100B, the LLoD was 2.7 pg/mL and the calibration range went from 2.7 to 2000.0 pg/mL. One patient was below detection range of the S100B and therefore the concentration of 1 pg/mL was attributed to this patient permitting statistical analysis. This applied concentration does not impact the statistics obtained.
Plasma GFAP, NF-L, and tau concentrations were measured using the Human Neurology 4-Plex A assay (N4PA) on an HD-1 Single molecule array (Simoa) instrument according to instructions from the manufacturer (Quanterix, Lexington, MA). For GFAP, the LLoD was 0.221 pg/mL, while the LLoQ was 0.467 pg/mL and the calibration range was 0.987 pg/mL to 725.0 pg/mL. The corresponding figures for NF-L were 0.104 pg/mL (LLoD), 0.241 pg/mL (LLoQ), and a calibration range between 0.533 pg/mL and 453.0 pg/mL. The corresponding figures for tau were 0.024 pg/mL (LLoD), with a calibration range between 0.053 pg/mL (LLoQ) and 0.136 pg/mL to 112.0 pg/mL. Plasma Aβ40 and Aβ42 concentrations were measured using a duplex Simoa immunoassay (Quanterix). For Aβ40, the LLoD was 0.045 pg/mL and the LLoQ was 0.142 pg/mL, with a calibration range between 0 pg/mL and 90.0 pg/mL. For Aβ42, the LLoD was 0.142 pg/mL and the LLoQ was 0.69 pg/mL, with a calibration range between 0 pg/mL and 11.0 pg/mL.
Excluding the single patient with S100B level below the detection range, there were no samples below the LLoDs and LLoQs. The measurements were performed by board-certified laboratory technicians who were blinded to clinical data.
TBI severity and CT scan grading
TBI severity assessment was solely based on the lowest Glasgow Coma Scale (GCS) score before intubation, either at the scene of accident or emergency department. A GCS value of 13–15 was considered mild, 9–12 moderate, and 3–8 severe TBI. For the analysis and taking into account clinical relevance, we analyzed the results in both the whole patient group and in the mTBI group separately. In addition, the groups were further divided into non-isolated (i.e., concomitant extracranial injuries) and isolated TBI subgroups.
Inclusion criteria were as follows: age ≥18 years, clinical diagnosis of TBI, and indications for acute head CT according to the National Institute for Health and Care Excellence (NICE) criteria (http://www.nice.org.uk/guidance/cg176). Exclusion criteria were blast-induced or penetrating injury, chronic subdural hematoma, inability to live independently as a result of pre-existing brain disease, TBI or suspected TBI not needing head CT, >2 weeks from the injury, not speaking the local language, and no consent obtained. CT scans were classified according to the Marshall grading system.34 Diffuse injury/grade I (no visual pathology) was considered CT-, whereas the other grades (II-VI) were regarded as CT+. Neuroradiologists at the Turku University Hospital and a senior neurosurgeon (JPP) double-read the CT scans.
Statistical analysis
Demographics of the subjects and time elapse from injury to blood sampling are presented as mean ± standard deviation. Normality of distribution of biomarkers levels was assessed with the Kolmogorov-Smirnov test and by visually inspecting data histograms. The levels of biomarkers were not normally distributed, and data are presented as medians and 25th and 75th percentiles. The GCS scores were not normally distributed. Mann-Whitney U test was used to compare the biomarker levels between the groups and to compare GCS score between patient groups. The ability of biomarkers in discriminating CT+ and CT− patients is presented with area under receiver operating characteristic curve (AUC) and Youden's Index (J). J captures the maximum performance of a dichotomous diagnostic test when equal importance is given to sensitivity and specificity (J = sensitivity+specificity-1).35 Partial AUC (pAUC) was used to compare only a portion of the biomarkers AUC curves, which here was set to the clinically relevant range 90–100% sensitivity. Panels were developed by the iterative combination of biomarkers and thresholds method using the toolbox Panelomix.36 For each biomarker, several cut-offs are selected and the best combination of markers and threshold is selected to give the best panel performance. The size of panels was set to maximum three biomarkers and was evaluated when sensitivity was set at 90–100% and at 100%. The biomarker levels in different patient groups have been presented as medians and interquartile ranges. The correlations between the biomarker levels were analyzed with Pearson correlation coefficient.
Data availability statement
De-identified clinical, imaging, and biochemical data not published within the article can be shared with a qualified investigator by request.
Results
One-hundred-sixty (160) patients were enrolled. There were 117 males (73.1%) and 43 females (26.9%), with a mean age of 47.2 ± 19.6 years. In 94 patients (58.7%), the TBI was isolated, and 66 patients (41.3%) had TBI with other concomitant extracranial injuries. There were 93 patients with mTBI. Among all patients, a negative CT was found in 65 patients (40.6%) and a positive CT in 95 (59.4%). Demographic data are presented in Table 1. Blood samples of all patients were obtained within 24 h of admission. In patients for whom the exact time of injury was available, the time elapse from injury to blood sampling was 15.2 ± 11.5 h (n = 70). Among patients in whom the exact injury time was unavailable, 34 patients were sampled within 24 h and 56 patients were sampled after 24 h from the injury.
Table 1.
Demographics of the Different TBI Subgroups
| All severities (n = 160) | Isolated all severities (n = 94) | mTBI (n = 93) | Isolated mTBI (n = 55) | |
|---|---|---|---|---|
| Age (mean ± SD) | 47.22 ± 19.59 | 48.59 ± 19.15 | 42.78 ± 18.60 | 42.76 ± 18.39 |
| Sex (male/female) | 117/43 | 62/32 | 60/33 | 30/25 |
| GCS (mean ± SD) | 12.45 ± 3.91 | 12.70 ± 3.41 | 14.65 ± 0.58 | 14.60 ± 0.63 |
| CT-negative | 65 (40.6%) | 39 (41.5%) | 56 (60.2%) | 36 (65.5%) |
| CT-positive | 95 (59.4%) | 55 (58.5%) | 37 (39.8%) | 19 (34.5%) |
| Marshall II | 27 (16.9%) | 11 (11.7%) | 21 (22.6%) | 9 (16.4%) |
| Marshall III | 3 (1.9%) | - | 1 (1.1%) | - |
| Marshall IV | 1 (0.6%) | 1 (1.1%) | 1 (1.1%) | 1 (1.8%) |
| Marshall V | 38 (23.8%) | 23 (24.5%) | 5 (5.4%) | 2 (3.6%) |
| Marshall VI | 26 (16.3%) | 20 (21.3%) | 9 (9.7%) | 7 (12.7%) |
TBI, traumatic brain injury; mTBI, mild traumatic brain injury; SD, standard deviation; GCS, Glasgow Coma Scale; CT, computed tomography.
As the need for CT was an inclusion criterion and the imaging was done rapidly after deciding the need for CT, the blood samples were drawn after the CT scan with few exceptions. The ability of individual biomarkers to distinguish CT− from CT+ patients is shown in Figures 1 and 2 and Table 2A and 2B (patients with mTBI and TBIs of all severities), as well as in Figures 3 and 4 and Table 2C and 2D (isolated mTBIs and isolated TBIs of all severities). Combinations of biomarkers for discriminating CT− and CT+ patients are presented in Table 3A and 3B (patients with mTBI and TBIs of all severities), and in Table 3C and 3D (isolated mTBIs and isolated TBIs of all severities). The biomarker levels in different patient groups are presented in Table 4.
FIG. 1.
Ability of the individual biomarkers in discriminating computed tomography (CT)-negative and CT-positive patients with traumatic brain injuries of all severities. Within each receiver operating characteristics curve, the uppermost value in parentheses are specificity at 100% sensitivity, the centermost values are specificity at sensitivity set the range of 90–100%, and the undermost values are specificity and sensitivity from Youden's index; a value before parentheses is a cut-off value of a protein at a particular specificity/sensitivity; all cut off values are presented in pg/mL except H-FABF in ng/mL; AUC, area under receiver operating characteristic curve; Aβ40, β-Amyloid isoform 1–40; Aβ42 β-Amyloid isoform 1–42; GFAP, glial fibrillary acidic protein; H-FABP, heart fatty-acid binding protein; IL-10, interleukin 10; NF-L neurofilament light; S100B, S100 calcium-binding protein B.
FIG. 2.
Ability of the individual biomarkers in discriminating computed tomography (CT)-negative and CT-positive patients with mild traumatic brain injury. Within each receiver operating characteristics curve, the uppermost value in parentheses are specificity at 100% sensitivity, the centermost values are specificity at sensitivity set the range of 90–100%, and the undermost values are specificity and sensitivity from Youden's index (exception: Aβ40 same dot for 100% and 90–100%); a value before parentheses is a cut-off value of a protein at a particular specificity/sensitivity; all cut off values are presented in pg/mL except H-FABF in ng/mL; AUC, area under receiver operating characteristic curve; Aβ40, β-Amyloid isoform 1–40; Aβ42 β-Amyloid isoform 1–42; GFAP, glial fibrillary acidic protein; H-FABP, heart fatty-acid binding protein; IL-10, interleukin 10; NF-L neurofilament light; S100B, S100 calcium-binding protein B.
Table 2A.
Ability of the Individual Biomarkers in Discriminating CT-Negative and CT-Positive Patients in TBIs of All Severities (All, n = 160; CT-Negative, n = 65; CT-Positive, n = 95)
| Mann U CT+ vs. CT− | AUC (95%CI) | J (%SP/%SE; cut off) | %SP @ 100%SE (cut off) | pAUC (95%CI) | %SP/ 90–100%SE (cut off) | |
|---|---|---|---|---|---|---|
| GFAP | < 0.001 | 0.822 (0.757–0.881) | 0.52 (92.3/60; 2690) | 13.8 (66.62) | 3.26 (2.06–4.77) | 41.5/93.7 (243) |
| NF-L | < 0.001 | 0.817 (0.748–0.877) | 0.55 (83.1/71.6; 20.25) | 6.2 (4.43) | 2.20 (1.06–4.01) | 40/90.5 (8.79) |
| tau | < 0.001 | 0.781 (0.702–0.852) | 0.51 (80/70.5; 3.09) | 4.6 (0.29) | 1.93 (1.0–3.20) | 23.1/94.7 (0.91) |
| Aβ40 | < 0.001 | 0.680 (0.594–0.760) | 0.36 (72.3/63.2; 19.9) | 0 (-) | 1.17 (0.51–2.3) | 12.3/98.9 (6.72) |
| IL-10 | < 0.001 | 0.676 (0.592–0.761) | 0.30 (76.9/52.6; 0.87) | 4.6 (0.14) | 1.82 (0.91–3.06) | 20.0/96.8 (0.21) |
| H-FABP | < 0.001 | 0.666 (0.582–0.751) | 0.34 (50.8/83.2; 4480) | 1.5 (1625.05) | 1.33 (0.38–2.90) | 30.8/90.5 (3760) |
| Aβ42 | 0.018 | 0.610 (0.519–0.692) | 0.29 (76.9/51.6; 19.4) | 1.5 (2.92) | 0.54 (0.09–1.13) | 6.2/96.8 (4.90) |
| S100B | 0.072 | 0.584 (0.491–0.667) | 0.20 (81.5/38.9; 141) | 0 (-) | 0.57 (0.12–1.35) | 4.6/97.9 (11.5) |
Table 2B.
Ability of the Individual Biomarkers in Discriminating CT-Negative and CT-Positive Patients with Mild TBI (All, n = 93; CT-Negative n = 56; CT-Positive n = 37)
| Mann U CT+ vs. CT− | AUC (95%CI) | J (%SP/%SE; cut off) | %SP @ 100%SE (cut off) | pAUC (95%CI) | %SP/ 90–100%SE (cut off) | |
|---|---|---|---|---|---|---|
| GFAP | < 0.001 | 0.720 (0.616–0.820) | 0.33 (62.5/70.3; 540) | 16.1 (66.62) | 2.86 (1.47–4.46) | 32.1/97.3 (132) |
| tau | 0.002 | 0.689 (0.576–0.797) | 0.37 (75.0/62.2; 2.42) | 14.3 (0.52) | 2.23 (1.09–3.66) | 26.8/94.6 (0.93) |
| NF-L | 0.004 | 0.676 (0.563–0.780) | 0.33 (67.9/64.9; 13.7) | 7.1 (4.43) | 1.62 (0.52–3.34) | 26.8/91.9 (6.67) |
| H-FABP | 0.021 | 0.642 (0.525–0.750) | 0.31 (55.4/75.7; 4490) | 1.8 (1709.61) | 1.1 (0.20–3.48) | 14.3/97.3 (2520) |
| Aβ40 | 0.066 | 0.613 (0.493–0.730) | 0.23 (71.4/51.4; 20.0) | 14.3 (6.78) | 1.51 (0.71–2.86) | 14.3/100 (6.78) |
| S100B | 0.265 | 0.569 (0.445–0.693) | 0.18 (60.7/56.8; 76.0) | 0 (-) | 0.33 (0–1.65) | 12.5/91.9 (168) |
| IL-10 | 0.177 | 0.583 (0.463–0.703) | 0.17 (82.1/35.1; 0.86) | 5.4 (0.14) | 1.42 (0.42–2.83) | 19.6/94.6 (0.21) |
| Aβ42 | 0.350 | 0.557 (0.428–0.683) | 0.19 (78.6/40.5; 19.6) | 1.8 (2.92) | 0.51 (0–1.43) | 8.9/94.6 (5.83) |
FIG. 3.
Ability of the individual biomarkers in discriminating computed tomography (CT)-negative and CT-positive patients with isolated traumatic brain injuries of all severities. Within each receiver operating characteristics curve, the uppermost value in parentheses are specificity at 100% sensitivity, the centermost values are specificity at sensitivity set the range of 90–100%, and the undermost values are specificity and sensitivity from Youden's index (exceptions: same dot for S100B 100% and 90–100%, Aβ42 100% and 90–100%, Aβ40 100% and 90–100%); a value before parentheses is a cut-off value of a protein at a particular specificity/sensitivity; all cut off values are presented in pg/mL except H-FABF in ng/mL; AUC, area under receiver operating characteristic curve; Aβ40, β-Amyloid isoform 1–40; Aβ42 β-Amyloid isoform 1–42; GFAP, glial fibrillary acidic protein; H-FABP, heart fatty-acid binding protein; IL-10, interleukin 10; NF-L neurofilament light; S100B, S100 calcium-binding protein B.
FIG. 4.
Ability of the individual biomarkers in discriminating computed tomography (CT)-negative and CT-positive patients with isolated mild TBI. Within each receiver operating characteristics curve, the uppermost value in parentheses are specificity at 100% sensitivity, the centermost values are specificity at sensitivity set the range of 90–100%, and the undermost values are specificity and sensitivity from Youden's index (exceptions: same dot for S100B 90–100% as Youden's Index, Aβ40 100% and 90–100%, GFAP 90–100% as Youden's Index); a value before parentheses is a cut-off value of a protein at a particular specificity/sensitivity; all cut off values are presented in pg/mL except H-FABF in ng/mL; AUC, area under receiver operating characteristic curve; Aβ40, β-Amyloid isoform 1–40; Aβ42 β-Amyloid isoform 1–42; GFAP, glial fibrillary acidic protein; H-FABP, heart fatty-acid binding protein; IL-10, interleukin 10; NF-L neurofilament light; S100B, S100 calcium-binding protein B.
Table 2C.
Ability of the Individual Biomarkers in Discriminating CT-Negative and CT-Positive Patients in Isolated TBIs of All Severities (All, n = 94; CT-Negative, n = 39; CT-Positive n = 55)
| Mann U CT+ vs. CT− | AUC (95%CI) | J (%SP/%SE; cut off) | %SP @ 100%SE (cut off) | pAUC (95%CI) | %SP/ 90–100%SE (cut off) | |
|---|---|---|---|---|---|---|
| GFAP | < 0.001 | 0.859 (0.786–0.933) | 0.57 (89.7/67.3; 1381.6) | 17.9 (66.6) | 4.03 (2.33–6.29) | 59.0/90.9 (244) |
| NF-L | < 0.001 | 0.848 (0.771–0.925) | 0.59 (84.6/74.5; 17.7) | 5.1 (4.18) | 3.29 (1.26–5.83) | 53.8/90.9 (8.79) |
| tau | < 0.001 | 0.789 (0.699–0.879) | 0.52 (92.3/60.0; 3.09) | 7.7 (0.29) | 2.52 (1.10–4.31) | 35.9/92.7 (0.91) |
| H-FABP | < 0.001 | 0.721 (0.613–0.830) | 0.45 (61.5/83.6; 4480) | 2.6 (1625.05) | 2.49 (0.75–5.11) | 43.6/90.9 (3760) |
| IL-10 | < 0.001 | 0.721 (0.616–0.825) | 0.39 (82.1/56.4; 0.66) | 2.6 (0.14) | 1.38 (0.35–2.98) | 17.9/96.4 (0.21) |
| Aβ40 | 0.001 | 0.695 (0.588–0.803) | 0.39 (71.8/67.3; 19.9) | 12.8 (6.72) | 1.31 (0.40–2.94) | 12.8/100 (6.72) |
| Aβ42 | 0.253 | 0.570 (0.452–0.687) | 0.27 (79.5/47.3; 20.9) | 2.6 (2.92) | 0.49 (0.00–1.28) | 2.6/100 (2.92) |
| S100B | 0.573 | 0.466 (0.583–0.348) | 0.09 (61.5/47.3; 76.0) | 0 (-) | 0.02 (0.00–0.42) | 0/100 (-) |
Table 2D.
Ability of the Individual Biomarkers in Discriminating CT-Negative and CT-Positive Patients with Isolated Mild TBI (All, n = 55; CT-Negative, n = 36; CT-Positive, n = 19)
| Mann U CT+ vs. CT− | AUC (95%CI) | J (%SP/%SE; cut off) | %SP @ 100%SE (cut off) | pAUC (95%CI) | %SP/ 90–100%SE (cut off) | |
|---|---|---|---|---|---|---|
| GFAP | 0.003 | 0.749 (0.614–0.883) | 0.39 (44.4/94.7; 140) | 19.4 (66.6) | 3.13 (1.11–5.95) | 44.4/94.7 (140) |
| H-FABP | 0.016 | 0.699 (0.559–0.839) | 0.43 (63.9/78.9; 4490) | 19.4 (2520) | 3.0 (1.11–6.33) | 41.7/94.7 (3370) |
| S100B | 0.022 | 0.689 (0.833–0.544) | 0.36 (41.7/94.7; 95.2) | 11.1 (179) | 2.56 (0.56–5.56) | 41.7/94.7 (95.2) |
| tau | 0.044 | 0.667 (0.515–0.818) | 0.31 (88.9/42.1; 2.46) | 22.2 (0.52) | 2.6 (1.23–4.99) | 30.6/94.7 (0.63) |
| NF-L | 0.049 | 0.662 (0.512–0.812) | 0.34 (50.0/84.2; 8.16) | 5.6 (4.18) | 1.61 (0.00–5.28) | 27.8/94.7 (6.34) |
| Aβ40 | 0.124 | 0.627 (0.472–0.783) | 0.25 (72.2/52.6; 19.95) | 13.9 (7.00) | 1.39 (0.56–4.17) | 13.9/100 (7.0) |
| Aβ42 | 0.608 | 0.542 (0.362–0.723) | 0.23 (80.6/42.1; 20.8) | 2.8 (2.92) | 0.54 (0.00–1.67) | 8.3/94.7 (5.83) |
| IL-10 | 0.860 | 0.515 (0.347–0.683) | 0.17 (63.9/52.6; 0.40) | 2.8 (0.14) | 0.67 (0.00–2.78) | 11.1/94.7 (0.18) |
Biomarkers are presented in order according to their AUC. Statistically significant p values are in bold. All cut off levels are presented in pg/mL.
CT, computed tomography; TBI, traumatic brain injury; Mann U, Mann-Whitney U-test; AUC, area under the curve; CI, confidence interval; J, Youden's Index; %SP @ 100%SE, specificity at 100 % sensitivity; pAUC%, partial area under the curve in the range of 90–100% sensitivity; %SP/90–100%SE, specificity / sensitivity in the range of 90–100 %; Aβ40, β-Amyloid isoform 1–40; Aβ42 β-Amyloid isoform 1–42; GFAP, glial fibrillary acidic protein; H-FABP, heart fatty-acid binding protein; IL-10, interleukin 10; NF-L neurofilament light; S100B, S100 calcium-binding protein B.
Table 3A.
Panels of the Best Biomarker Combinations in Discriminating CT-Negative and CT-Positive Patients in TBIs of All Severities
| Sensitivity | Number of biomarkers | Biomarkers (threshold to be classified as positive, pg/mL) | Number of biomarkers needed to be classified as + | %Specificity (95%CI) | %Sensitivity (95%CI) | ||
|---|---|---|---|---|---|---|---|
| 90–100% | 2 | Aβ40 (> 19.9) | NF-L (> 17.7) | - | 1 | 61.5 (49.2–73.8) | 91.6 (85.3–96.8) |
| 3 | Aβ40 (> 19.3) | IL-10 (> 0.21) | NF-L (>17.7) | 2 | 69.2 (58.5–80.0) | 90.5 (84.2–95.8) | |
| 100% | 2 | IL-10 (> 0.39) | GFAP (> 467) | - | 1 | 35.4 (24.6–47.7) | 100 (100–100) |
| 3 | GFAP (> 467) | H-FABP (> 2520) | IL-10 (>0.39) | 2 | 38.5 (26.2–50.8) | 100 (100–100) | |
Table 3B.
Panels of the Best Biomarker Combinations in Discriminating CT-Negative and CT-Positive Patients with Mild TBI
| Sensitivity | Number of biomarkers | Biomarkers (threshold to be classified as positive, pg/mL) | Number of biomarkers needed to be classified as + | %Specificity (95%CI) | %Sensitivity (95%CI) | ||
|---|---|---|---|---|---|---|---|
| 90–100% | 2 | H-FABP (> 4620) | tau (> 2.56) | - | 1 | 50.0 (37.5–62.5) | 91.9 (83.8–100) |
| 3 | H-FABP (> 4490) | S100B (< 105) | tau (> 2.42) | 2 | 60.7 (48.2–73.2) | 91.9 (83.8–100) | |
| 100% | 2 | IL-10 (> 0.39) | GFAP (> 468) | - | 1 | 37.5 (25.0–50.0) | 100 (100–100) |
| 2 | H-FABP (> 4170) | tau (> 2.53) | - | 1 | 37.5 (25.0–50.0) | 100 (100–100) | |
| 3 | H-FABP (> 4170) | S100B (< 136) | tau (> 2.42) | 2 | 46.4 (33.9–58.9) | 100 (100–100) | |
Table 3C.
Panels of the Best Biomarker Combinations in Discriminating CT-Negative and CT-Positive Patients in Isolated TBIs of All Severities
| Sensitivity | Number of biomarkers | Biomarkers (threshold to be classified as positive, pg/mL) | Number of biomarkers needed to be classified as + | %Specificity (95%CI) | %Sensitivity (95%CI) | ||
|---|---|---|---|---|---|---|---|
| 90–100% | 2 | Aβ42 (> 20.9) | NF-L (> 17.7) | - | 1 | 69.2 (53.8–82.1) | 90.9 (81.8–98.2) |
| 3 | GFAP (> 4510) | H-FABP (> 4480) | S100B (< 110) | 2 | 79.5 (66.7–92.3) | 90.9 (83.6–98.2) | |
| 100% | 2 | GFAP (> 468) | IL-10 (> 0.39) | - | 1 | 48.7 (33.3–64.1) | 100 (100–100) |
| 3 | GFAP (> 468) | S100B (< 110) | IL-10 (> 0.39) | 2 | 66.7 (51.3–82.1) | 100 (100–100) | |
Table 3D.
Panels of the Best Biomarker Combinations in Discriminating CT-Negative and CT-Positive Patients with Isolated Mild TBI
| Sensitivity | Number of biomarkers | Biomarkers (threshold to be classified as positive, pg/mL) | Number of biomarkers needed to be + | %Specificity (95%CI) | %Sensitivity (95%CI) | ||
|---|---|---|---|---|---|---|---|
| 90–100% | 2 | H-FABP (> 4490) | tau (> 2.46) | - | 1 | 58.3 (41.7–75.0) | 100 (100–100) |
| 3 | GFAP (> 540) | H-FABP (> 3880) | IL-10 (> 0.40) | 2 | 69.4 (55.6–83.3) | 94.7 (84.2–100) | |
| 100% | 2 | H-FABP (> 4490) | tau (> 2.46) | - | 1 | 58.3 (41.7–75.0) | 100 (100–100) |
| 3 | H-FABP (> 4490) | S100B (< 141) | tau (> 2.46) | 2 | 66.7 (50.0–80.6) | 100 (100–100) | |
CT, computed tomography; CI, confidence interval; Aβ42, β-Amyloid isoform 1–42; GFAP, glial fibrillary acidic protein; H-FABP, heart fatty-acid binding protein; IL-10, interleukin 10; NF-L neurofilament light; S100B, S100 calcium-binding protein B.
Table 4A.
Biomarker Levels in Patients with TBIs of All Severities
| Non-isolated TBI CT+, median level (IQR) pg/mL (n = 40, GCS 11.18 ± 4.94) | Non-isolated TBI CT-, median level (IQR) pg/mL (n = 26, GCS 13.50 ± 3.46) | Isolated TBI CT+, median level (IQR) pg/mL (n = 55, GCS 11.58 ± 3.91) | Isolated TBI CT-, median level (IQR) pg/mL (n = 39, GCS 14.28 ± 1.52) | |
|---|---|---|---|---|
| Aβ40 | 21.4 (14.2–29.8) | 15.7 (10.5–21.3) | 24.7 (15.3–32.2) | 17.0 (11.9–21.4) |
| Aβ42 | 20.0 (14.9–26.1) | 15.9 (11.6–19.0) | 18.5 (11.5–29.0) | 15.7 (11.7–20.1) |
| GFAP | 5890 (1830–32700) | 1140 (435–2210) | 6840 (585–47600) | 204 (82.0–530) |
| H-FABP | 16800 (5570–42500) | 8440 (4150–28000) | 6720 (4980–12800) | 4080 (2880–6770) |
| IL-10 | 1.10 (0.47–4.04) | 0.79 (0.23–1.60) | 0.86 (0.41–2.46) | 0.37 (0.24–0.55) |
| NF-L | 48.4 (19.1–85.7) | 13.6 (10.4–22.7) | 57.6 (15.9–116) | 8.66 (6.26–16.8) |
| S100B | 122 (63.2–259) | 81.1 (45.3–143) | 83.5 (49.9–209) | 85.9 (50.4–120.3) |
| tau | 11.0 (4.00–36.1) | 2.35 (1.52–4.82) | 6.83 (1.52–32.8) | 1.53 (0.55–2.21) |
Table 4B.
Biomarker Levels in Patients with Mild TBI
| Non-isolated mTBI CT+, median level (IQR) pg/mL (n = 18, GCS 14.61 ± 0.61) | Non-isolated mTBI CT-, median level (IQR) pg/mL (n = 20, GCS 14.80 ± 0.41) | Isolated mTBI CT+, median level (IQR) pg/mL (n = 19, GCS 14.53 ± 0.61) | Isolated mTBI CT-, median level (IQR) pg/mL (n = 36, GCS 14.64 ± 0.64) | |
|---|---|---|---|---|
| Aβ40 | 17.1 (11.14–27.53) | 14.2 (9.68–23.4) | 20.1 (14.8–27.9) | 16.9 (11.9–21.3) |
| Aβ42 | 17.4 (11.50–22.99) | 15.9 (11.9–18.9) | 17.6 (9.26–26.2) | 15.7 (11.7–19.1) |
| GFAP | 1830 (822.28–2600.97) | 1140 (449–2040) | 604 (214–2290) | 186.64 (75.9–500) |
| H-FABP | 15400 (4220–37900) | 7250 (4150–24800) | 5670 (4550–9640) | 3880.90 (2860–6380) |
| IL-10 | 1.10 (0.33–5.97) | 0.73 (0.22–1.45) | 0.41 (0.23–0.56) | 0.34 (0.23–0.53) |
| NF-L | 19.1 (12.7–41.1) | 13.0 (8.63–18.4) | 14.0 (8.59–19.5) | 8.23 (6.11–15.3) |
| S100B | 92.3 (45.9–164) | 81.1 (48.1–120) | 51.4 (38.4–87.0) | 83.46 (48.8–118) |
| tau | 4.29 (3.08–7.61) | 2.44 (1.52–4.40) | 1.79 (1.19–2.92) | 1.51 (0.55–2.08) |
GCS is presented as group mean ± standard deviation.
TBI, traumatic brain injury; CT, computed tomography; IQR, interquartile range; GCS, Glasgow Coma Scale; Aβ40, β-Amyloid isoform 1–40; Aβ42 β-Amyloid isoform 1–42; GFAP, glial fibrillary acidic protein; H-FABP, heart fatty-acid binding protein; IL-10, interleukin 10; NF-L neurofilament light; S100B, S100 calcium-binding protein.
In patients with both mTBI and TBIs of all severities, there were no significant differences in the GCS scores between patients with isolated versus non-isolated TBI (Table 5A). A comparison of biomarker levels between isolated and non-isolated TBIs was made both for all patients and patients with mTBI. Non-isolated TBI patients with all severities showed higher levels of H-FABP than isolated TBI in the CT+ group of all severities (p = 0.023), and CT+ patients with non-isolated mTBI had higher levels of IL-10 (p = 0.014), NF-L (p = 0.036), S100B (p = 0.019), and tau (p = 0.005). The biomarker levels and their differences in different patient groups are presented in Table 5B. Table 6 shows the correlations between the various biomarkers used in this study.
Table 5A.
Comparison GCS Scores between Different Patient Groups
| Non-isolated all severities TBI CT+ vs. isolated all severities TBI CT+ | Non-isolated all severities TBI CT− vs. isolated all severities TBI CT− | Non-isolated mTBI CT+ vs. isolated mTBI CT+ | Non-isolated mTBI CT− vs. isolated mTBI CT- | |
|---|---|---|---|---|
| n | 40 vs. 55 | 26 vs. 39 | 18 vs. 19 | 20 vs. 36 |
| Mean GCS ± SD | 11.18 ± 4.94 vs. 11.58 ± 3.91 | 13.50 ± 3.46 vs. 14.28 ± 1.52 | 14.61 ± 0.61 vs. 14.53 ± 0.61 | 14.80 ± 0.41 vs. 14.64 ± 0.64 |
| p value | 0.981 | 0.567 | 0.460 | 0.367 |
Table 5B.
Differences in the Biomarker Levels in Different Patient Groups
| Mann U, non-isolated all severities TBI CT+ vs. isolated all severities TBI CT+ (median levels, pg/mL) | Mann U, non-isolated all severities TBI CT− vs. isolated all severities TBI CT− (median levels, pg/mL) | Mann U, non-isolated mTBI CT+ vs. isolated mTBI CT+ (median levels, pg/mL) | Mann U, non-isolated mTBI CT− vs. isolated mTBI CT− (median levels, pg/mL) | |
|---|---|---|---|---|
| Aβ40 | 0.449 (21.4 vs. 24.7) | 0.533 (15.7 vs. 17.0) | 0.775 (17.0 vs. 20.1) | 0.521 (14.2 vs. 16.9) |
| Aβ42 | 0.438 (20.0 vs. 18.5) | 0.888 (15.9 vs. 15.7) | 0.753 (17.4 vs. 17.6) | 0.898 (15.9 vs. 15.7) |
| GFAP | 0.916 (5890 vs. 6840) | 0.001 (1140 vs. 204) | 0.118 (1830 vs. 604) | 0.001 (1140 vs. 186) |
| H-FABP | 0.023 (16800 vs. 6720) | 0.002 (8440 vs. 4080) | 0.298 (15400 vs. 5670) | 0.004 (7250 vs. 3880) |
| IL-10 | 0.386 (1.10 vs. 0.86) | 0.024 (0.79 vs. 0.37) | 0.014 (1.10 vs. 0.41) | 0.090 (0.73 vs. 0.34) |
| NF-L | 0.871 (48.4 vs. 57.6) | 0.034 (13.6 vs. 8.66) | 0.036 (19.1 vs. 14.0) | 0.070 (13.0 vs. 8.23) |
| S100B | 0.107 (122 vs. 83.5) | 0.851 (81.1 vs. 85.9) | 0.019 (92.3 vs. 51.4) | 0.824 (81.1 vs. 83.5) |
| tau | 0.216 (11.0 vs. 6.83) | 0.002 (2.35 vs. 1.53) | 0.005 (4.29 vs. 1.79) | 0.001 (2.44 vs. 1.51) |
Statistically significant p values are in bold.
GCS, Glasgow Coma Scale; TBI, traumatic brain injury; CT, computed tomography; SD, standard deviation; Mann U, p value from Mann-Whitney U-test; Aβ40, β-Amyloid isoform 1–40; Aβ42 β-Amyloid isoform 1–42; GFAP, glial fibrillary acidic protein; H-FABP, heart fatty-acid binding protein; IL-10, interleukin 10; NF-L neurofilament light; S100B, S100 calcium-binding protein B; Mann U, p value from Mann-Whitney U-test.
Table 6.
Correlations between the Biomarker Levels in All Patients (n = 160)
| Biomarker | GFAP | NF-L | tau | Aβ40 | IL-10 | H-FABP | Aβ42 | S100B | |
|---|---|---|---|---|---|---|---|---|---|
| GFAP | Pearson's r | 1 | 0.638 | 0.948 | 0.146 | 0.057 | -0.017 | 0.154 | 0.399 |
| p value | p < 0.001 | p < 0.001 | 0.065 | 0.470 | 0.831 | 0.052 | p < 0.001 | ||
| NF-L | Pearson's r | 0.638 | 1 | 0.640 | 0.184 | 0.114 | 0.135 | 0.218 | 0.302 |
| p value | p < 0.001 | p < 0.001 | 0.020 | 0.151 | 0.089 | 0.006 | p < 0.001 | ||
| tau | Pearson's r | 0.948 | 0.640 | 1 | 0.148 | 0.097 | 0.080 | 0.167 | 0.475 |
| p value | p < 0.001 | p < 0.001 | 0.062 | 0.223 | 0.313 | 0.034 | p < 0.001 | ||
| Aβ40 | Pearson's r | 0.146 | 0.184 | 0.148 | 1 | 0.111 | 0.108 | 0.386 | 0.296 |
| p value | 0.065 | 0.020 | 0.062 | 0.163 | 0.174 | p < 0.001 | p < 0.001 | ||
| IL-10 | Pearson's r | 0.057 | 0.114 | 0.097 | 0.111 | 1 | 0.384 | 0.012 | 0.383 |
| p value | 0.470 | 0.151 | 0.223 | 0.163 | p < 0.001 | 0.882 | p < 0.001 | ||
| H-FABP | Pearson's r | -0.017 | 0.135 | 0.080 | 0.108 | 0.384 | 1 | 0.138 | 0.413 |
| p value | 0.831 | 0.089 | 0.313 | 0.174 | p < 0.001 | 0.082 | p < 0.001 | ||
| Aβ42 | Pearson's r | 0.154 | 0.218 | 0.167 | 0.386 | 0.012 | 0.138 | 1 | 0.117 |
| p value | 0.052 | 0.006 | 0.034 | p < 0.001 | 0.882 | 0.082 | 0.141 | ||
| S100B | Pearson's r | 0.399 | 0.302 | 0.475 | 0.296 | 0.383 | 0.413 | 0.117 | 1 |
| p value | p < 0.001 | p < 0.001 | p < 0.001 | p < 0.001 | p < 0.001 | p < 0.001 | 0.141 |
Statistically significant p values are in bold.
GFAP, glial fibrillary acidic protein; NF-L neurofilament light; Aβ40, β-Amyloid isoform 1–40; IL-10, interleukin 10; H-FABP, heart fatty-acid binding protein; Aβ42, β-Amyloid isoform 1–42; S100B, S100 calcium-binding protein B.
Individual biomarkers in CT− and CT+ patients with TBIs of all severities
In patients with TBIs of all severities (Table 2A), the biomarker levels in CT+ and CT− groups were significantly different (p < 0.001) for all other studied biomarkers, except Aβ42 and S100B. The AUC values varied from 0.584 to 0.822, with GFAP and NF-L showing the highest values (0.822 and 0.817, respectively; Fig. 1; Table 2A). J's varied from 0.20 to 0.55, again with NF-L and GFAP having the best indices (0.55 and 0.52, respectively). When sensitivity was set to 100%, all biomarkers showed poor specificities, ranging from 0 to 13.8%; the best ones shown by GFAP and NF-L (13.8% and 6.2%, respectively). If the pAUC in the range of 90–100% sensitivity or the best specificity at 90–100% sensitivity was used to compare the biomarkers, GFAP and NF-L were again the best (Table 2A).
In patients with isolated TBIs of all severities, all biomarkers except for Aβ42 and S100B, were significantly different (p ≤ 0.001) between the CT+ and CT− groups. The AUC values were slightly higher in isolated TBIs of all severities, ranging from 0.466 to 0.859, with NF-L and GFAP having the best AUCs (0.848 and 0.859, respectively). Also, the J's and pAUCs were higher in the isolated TBIs of all severities, with NF-L and GFAP showing the best values (NF-L: 0.59 and 3.29 and GFAP: 0.57 and 4.03, respectively). When sensitivity was set to 100%, GFAP and Aβ40 showed the best specificities (17.9% and 12.8%, respectively; Table 2C).
Individual biomarkers in CT− and CT+ patients with mTBI
In cases with mTBI (Table 2B), the biomarker levels in the CT+ and CT− groups were significantly different for tau, GFAP, NF-L, and H-FABP (p < 0.05). The AUC values varied from 0.557 to 0.720, with GFAP and tau showing the highest values (Fig. 2; Table 2B). The J's varied from 0.17 to 0.37, with tau and GFAP having the best values (0.37 and 0.33, respectively). At 100% sensitivity, the specificity varied from 0 to 16.1%, with GFAP, Aβ40, and tau having the highest specificities (16.1%, 14.3%, and 14.3%, respectively). Using the pAUC in the range of 90–100% sensitivity or the best specificity at 90–100% sensitivity as measures, GFAP and tau showed the best results (Table 2B).
In patients with isolated mTBIs, the levels of H-FABP, GFAP, S100B, tau, and NF-L (p < 0.05) were significantly different between the CT+ and CT− groups. The biomarkers with the highest AUC values were slightly different in isolated mTBIs compared with all mTBIs. The AUCs varied from 0.515 to 0.749, with GFAP and H-FABP having the best values (0.749 and 0.699, respectively). The J's and pAUCs were generally higher in the isolated mTBIs, with H-FABP and GFAP showing the best values (H-FABP: 0.43 and 3.0 and GFAP: 0.39 and 3.13, respectively). When sensitivity was set to 100%, tau, GFAP, and H-FABP showed the best specificities (22.2%, 19.4%, and 19.4%, respectively; Table 2D).
Combinations of biomarkers in CT− and CT+ patients with TBIs of all severities
We studied if various combinations of biomarkers could improve the ability to discriminate patients with intracranial CT abnormalities from those without. In patients with TBIs of all severities, the optimal combinations varied slightly depending on if the sensitivity was set to 100% or to 90–100%. With 100% sensitivity, the best specificity (38.5%) was reached with a combination of GFAP+H-FABP+IL-10. The best combination of two biomarkers was IL-10+GFAP, which reached 35.4% specificity. Using the sensitivity range from 90–100%, the best specificity was shown by a combination Aβ40+IL-10+NF-L (69.2% specificity with 90.5% sensitivity), whereas the best combination of two biomarkers was Aβ40+NF-L (61.5% specificity with 91.6% sensitivity; Table 3A).
A corresponding analysis was conducted for patients with isolated TBIs of all severities. When sensitivity was set to 100%, the best combination of three biomarkers was GFAP+S100B+IL-10 with a specificity of 66.7%. A similar result was obtained for two-biomarker combination, where IL-10+GFAP was still the best, but with improved specificity of 48.7%. With a sensitivity of 90–100%, the best panel of three biomarkers was totally different from the whole TBI group, namely GFAP+H-FABP+S100B with 79.5% specificity and 90.9% sensitivity. The best combination of two biomarkers was Aβ42+NF-L with 69.2% specificity and 90.9% sensitivity. The results are demonstrated in Table 3C.
Combinations of biomarkers in CT− and CT+ patients with mTBI
In patients with mTBI, with 100% sensitivity, the best combination of three biomarkers was H-FABP+S100B+tau (with 46.4% specificity), and of two biomarkers either H-FABP+tau or IL-10+GFAP (both with 37.5% specificity). If the sensitivity range of 90–100% was used, the best specificity was reached by a combination of H-FABP+S100B+tau (60.7% specificity with 91.9% sensitivity), and the best combination of two biomarkers was H-FABP+tau (50.0% specificity with 91.9% sensitivity; Table 3B).
In isolated mTBIs with sensitivity set to 100%, the best panel of three biomarkers was a combination H-FABP+S100B+tau (specificity of 66.7%), and the best two-biomarker combination was H-FABP+tau (specificity of 58.3%). With a sensitivity of 90–100%, the best panel of three biomarkers was GFAP+H-FABP+IL-10 (specificity of 69.4% with sensitivity of 94.7%), and the aforementioned combination H-FABP+tau remained as the best option for two biomarkers combined. The results are shown in Table 3D.
Discussion
This prospective, observational study of patients with acute TBIs investigated the ability of eight protein biomarkers in discriminating CT+ and CT− patients, utilizing modern highly sensitive immunoassays in a well-characterized cohort. NF-L, GFAP, and tau exhibited the best abilities in discriminating CT+ and CT− patients, both in patients with mTBI and TBIs of all severities. In patients with isolated TBIs of all severities, NF-L, GFAP, and tau again performed again the best, but in patients with isolated mTBI, H-FABP, GFAP, and S100B showed the best results. Overall, single biomarkers had very low specificities (0–22.2%) when sensitivity was set to 100%.
Hence, we studied whether panels of biomarkers would give better specificity/sensitivity. In the whole group, a combination of GFAP+H-FABP+IL-10 yielded the best specificity in separating CT+ and CT− patients when sensitivity was set to 100%. In patients with mTBI, a panel of H-FABP+S100B+tau showed the best specificity when sensitivity set to 100%.
Next, we hypothesized that in the case of isolated TBIs, the optimal biomarker combinations may be different, because none of the proteins are apparently 100% brain specific. In isolated TBIs of all severities, when sensitivity was set to 100%, the best specificity was reached with a panel of GFAP+S100B+IL-10. For patients with isolated mTBI, a panel of H-FABP+S100B+tau showed the best specificity when sensitivity set to 100%. The panel is the same as in all mTBI patient group, with relatively similar thresholds for the panel to be classified as positive.
These results suggest that the best diagnostic value in discriminating CT+ and CT− patients can be achieved by utilizing biomarkers that do not necessarily perform best when applied alone. In the current study, NF-L, GFAP, and tau exhibited the best AUCs and J's when studied individually, but H-FABP, IL-10, and S100B appeared in several of the best panel options, along with GFAP. Intriguingly, S100B alone showed 0% specificity and statistically nonsignificant difference between the CT+ and CT− groups in TBIs of all severities, isolated TBIs of all severities, and patients with mTBI, yet it appeared in some of the best combinations of biomarkers. The levels of S100B in CT+ patients were significantly higher in the non-isolated vs. isolated mTBIs, suggesting extracerebral release of the protein at the time of injury, which is in line with the previous literature.12
H-FABP23 and IL-1024 have earlier been studied in screening patients with mTBI for a need for head CT. In this study, they perform well also in TBIs of all severities, whether isolated or not, although they both appear to be released also from extracerebral sources, because their levels were higher in non-isolated CT−positive patients compared with isolated ones, without a difference in the severity of TBI between the groups. The AUCs of H-FABP and IL-10 were relatively similar as reported earlier, but the specificities were lower both in mTBIs and TBIs of all severities in the current study.23,24
GFAP is the second most studied biomarker of TBI, after S100B. In the current study, AUCs were slightly lower compared with earlier studies in patients with mTBI.17,37–39 We have previously analyzed the levels of GFAP using a less sensitive assay from a cohort of patients with TBIs with all severities, including also the patients of the current study.15 An ultrasensitive Simoa method29 was used in the current study to analyze GFAP levels, and slightly higher AUCs were observed compared with our earlier study, but the patient cohorts were not identical.
Aβ40, Aβ42, tau, and NF-L have been less studied in acute diagnostics of TBI. In this study, NF-L and tau exhibited very good AUCs and J's in the group of all patients with TBI. However, in patients with isolated mTBIs, H-FABP, GFAP, and S100B all outperformed NF-L and tau, suggesting that the less brain-specific biomarkers H-FABP and S100B are useful in cases of isolated mTBI. Levels of NF-L and tau were higher in patients with non-isolated CT+ mTBI than in those with isolated CT+ mTBI, suggesting possible extracerebral release. The levels of Aβ42 did not differ between CT+ and CT− patients in any of the analyzed groups, while the levels of Aβ40 levels were significantly higher in CT+ patients, both in TBIs of all severities and isolated TBIs of all severities.
Both S100B and the combination of GFAP and UCH-L1 have been used as biomarkers to screen for CT-positivity/negativity in patients with acute TBI.19–21 We analyzed also the levels of UCH-L1 from our samples, but the coefficients of variation were at a level where we could not consider the results sufficiently reliable, and therefore they have not been included in the analyses. For S100B, the publications have yielded an AUC of 0.69–0.78, and with 98–100% sensitivity and specificity from 5% to 22.9%.19,20 FDA recently approved the combination of GFAP+UCH-L1 to be used to screen the need for a CT scan in acute mTBI and supported by a study reporting 36.5% specificity with 97.5% sensitivity for patients with GCS of 9–15 and 36.7% specificity with 97.3% sensitivity for a subset of patients with GCS of 14–15 (AUC values not given).21 Compared with the above-mentioned results, the best panels in this study suggest that clearly improved specificities might be reached with 100% sensitivity using optimal biomarker combinations for targeted patient groups.
In this study, the NICE criteria8 for head CT scanning were used. In a validation study including several international guidelines for indications of head CT, the NICE criteria yielded a sensitivity of 82.1% and a specificity of 46.1% for detecting traumatic intracranial findings in patients with GCS of 13–15.40 The biomarker panels for both mTBI groups in the current study outperform the NICE criteria, but a proper comparison for diagnostic accuracy is not possible because of the study design and because most blood samples were drawn after the head CT scans.
There are both considerable strengths and limitations of this study. Strengths are the use of several different biomarkers in the same cohort, use of sensitive advanced analytics, and a prospectively collected well-characterized study population. Our results are comparable to recent studies utilizing the same methodologies, but it is uncertain how studies utilizing different assays give comparable results, especially at low biomarker levels. The main shortcomings include variable delays between the injury and the sampling, and the timing of the sampling after the CT scan. Due to the latter, we are not able to present our results as predictors for CT findings. This may affect the results especially in case of those biomarkers, which have a fairly short half-life, such as H-FABP, IL-10, and S100B. Indeed, most of our blood samples have not been collected within the 6-h time window recommended for the use of S100B. For other biomarkers (e.g., NF-L), the sampling time-point may have been too close to the injury; NF-L is a slow marker that reaches its maximum more than 10 days after the injury.21 It may thus not be the optimal biomarker for acute injury detection. Earlier sampling might have either improved or attenuated the diagnostic capabilities of the biomarker panels.
When interpreting our results, it has to be noticed that especially what comes to patients with mTBI, our series cannot be considered to represent cases with mTBI at large, since the mildest cases were often discharged before possibility for recruitment, and a fairly large percentage showed traumatic intracranial abnormalities in CT and required hospital admission. In addition, as the inclusion criteria was based on using the NICE criteria for head CT, the results are not necessarily applicable for other head CT rules. Pre-selection using any head CT rule also gives different sensitivities and specificities than using biomarkers for screening the whole population of patients with head trauma attending emergency care.
This study analyzed only the associations of different biomarkers with visible intracranial abnormalities in CT. Biomarkers or biomarker panels that are needed to separate patients with TBI from patients with acute injuries without a TBI, or to predict the outcome of TBI, may well be different from those found in this study. In addition, the optimal biomarkers do not depend only on the indication and patient population, but also on timing, which is why these results should be replicated and widened in further studies.
Conclusions
The main finding was that panels of protein biomarkers perform better in discriminating CT+ and CT− patients than individual biomarkers. A panel including GFAP+H-FABP+IL-10 yielded the best specificity (38.5%) in separating CT+ and CT− patients with 100% sensitivity within 24 h of admission in TBIs of all severities. In patients with mTBI, a panel of H-FABP+S100B+tau gave the best specificities with 100% sensitivity not depending on whether TBI was isolated or not. The true diagnostic value of these biomarker panels compared with existing head CT rules should be addressed in further studies. Our results also suggest that different biomarkers may be needed when searching for optimal diagnostic tools for different types of patients.
Acknowledgments
This work was partially funded by the European Commission under the 7th Framework Programme (FP7-270259-TBIcare), Government's Special Financial Transfer tied to academic research in Health Sciences (Finland; JPP), Emil Aaltonen Foundation sr (JPP), Finnish Brain Foundation sr (JPP), Maire Taponen Foundation (JPP), Integra EANS Research Grant (IH), University of Turku Graduate School funding (MM), National Institute for Health Research (NIHR) Research Professorship and the NIHR Cambridge BRC (PJH), NIHR Research UK (through a Senior Investigator Award and the Cambridge Biomedical Research Centre; DKM); Academy of Medical Sciences/Health Foundation Clinician Scientist Fellowship (VFN); Wallenberg Academy Fellowship and grants from the Swedish and European Research Councils (HZ), Torsten Söderberg Professorship in Medicine, award by the Royal Swedish Academy of Sciences, grants from the Swedish Research Council (KB). The authors thank our research nurses Patricia Bertenyi and Satu Honkala for their valuable contribution to this study.
Funding: This work was partially funded by the European Commission under the 7th Framework Programme (FP7-270259-TBIcare), Government's Special Financial Transfer tied to academic research in Health Sciences (Finland) (JPP), Emil Aaltonen Foundation (JPP), Finnish Brain Foundation (JPP), Integra EANS Research Grant (IH), University of Turku Graduate School funding (MM), NIHR Research Professorship and the NIHR Cambridge BRC (PJH), NIHR Research UK (through a Senior Investigator Award and the Cambridge Biomedical Research Centre) (DKM), Academy of Medical Sciences/The Health Foundation Clinician Scientist Fellowship (VFN); Wallenberg Academy Fellowship and grants from the Swedish and European Research Councils (HZ), Torsten Söderberg Professorship in Medicine, award by the Royal Swedish Academy of Sciences, grants from the Swedish Research Council (KB).
Author Disclosure Statement
Jussi P. Posti has no competing financial interests. JPP has received speaker's fees from Orion corporation and Finnish Medical Association.
Riikka S.K. Takala has no competing financial interests. RSKT has received speakers' fees from Abbott, Fresenius-Kabi, Orion corporation and UCB, conference funding from Pfizer and Steripolar and is stockholder of Orion.
Janek Frantzén is a Member of Board at Bonalive Ltd., JF has received travel support from Abbot, Teva, and Medtronic. He has received speaker's fees from Teva, Medtronic and Bonalive Ltd.
David K. Menon reports collaborative research or consultancy agreements with GlaxoSmithKline Ltd; Ornim Medical; Shire Medical; Calico Inc; Pfizer Ltd; Pressura Ltd; Glide Pharma Ltd; NeuroTraumaSciences LLC; Lantasman AB.
Kevin Hrusovsky is an employee and stockholder at Quanterix Corp.
David H. Wilson is an employee and stockholder at Quanterix Corp.
Henrik Zetterberg has served at advisory boards for Roche Diagnostics, Wave, Samumed and CogRx, and is a co-founder of Brain Biomarker Solutions in Gothenburg AB, a GU Ventures-based platform company at the University of Gothenburg.
Kaj Blennow has served as a consultant or at advisory boards for Alzheon, BioArctic, Biogen, Eli Lilly, Fujirebio Europe, IBL International, Merck, Novartis, Pfizer, and Roche Diagnostics, and is a co-founder of Brain Biomarker Solutions in Gothenburg AB, a GU Ventures-based platform company at the University of Gothenburg.
For the other authors, no competing financial interests exist.
References
- 1. Kasper C.E. (2015). Traumatic Brain Injury. Springer: New York [Google Scholar]
- 2. Peeters W., van den Brande R., Polinder S., Brazinova A., Steyerberg E., Lingsma H., and Maas A.I.R. (2015). Epidemiology of traumatic brain injury in Europe. Acta Neurochir. (Wien). 157, 1683–1696 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Maas A.I.R., Menon D.K., Adelson D., Andelic N., Bell M.J., Belli A., Bragge P., Brazinova A., Büki A., Chesnut R.M., Citerio G., Coburn M., Cooper J., Crowder T., Czeiter E., Czosnyka M., Diaz-Arrastia R., Dreier J.P., Duhaime A.-C., Ercole A., Van Essen T.A., Feigin V.L., Gao G., Giacino J., Gonzalez-Lara L.E., Gruen R.L., Gupta D., Hartings J.A., Hill S., Jiang J.-Y., Ketharanathan N., Kompanje E.J.O., Lanyon L., Laureys S., Lecky F., Levin H., Lingsma H.F., Maegele M., Majdan M., Manley G., Marsteller J., Mascia L., Mcfadyen C., Mondello S., Newcombe V., Palotie A., Parizel P.M., Peul W., Piercy J., Polinder S., Puybasset L., Rasmussen T.E., Rossaint R., Smielewski P., Söderberg J., Stanworth S.J., Stein M.B., Von Steinbüchel N., Stewart W., Steyerberg E.W., Stocchetti N., Synnot A., Ao B., TeTenovuo O., Theadom A., Tibboel D., Videtta W., Wang K.K.W., Williams W.H., Wilson L., and Yaffe K.; InTBIR Participants and Investigators. (2017). Traumatic brain injury: integrated approaches to improve prevention, clinical care, and research. Lancet Neurol. 16, 987–1048 [DOI] [PubMed] [Google Scholar]
- 4. Levin H., and Diaz Arrastia R. (2015). Diagnosis, prognosis, and clinical management of mild traumatic brain injury. Lancet Neurol. 14, 506–517 [DOI] [PubMed] [Google Scholar]
- 5. Sweeney T.E., Salles A., Harris O.A., Spain D.A., and Staudenmayer K.L. (2015). Prediction of neurosurgical intervention after mild traumatic brain injury using the national trauma data bank. World J. Emerg. Surg. 10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Haydel M.J., Preston C.A., Mills T.J., Luber S., Blaudeau E., and DeBlieux P.M. (2000). Indications for computed tomography in patients with minor head injury. N. Engl. J. Med. 343, 100–105 [DOI] [PubMed] [Google Scholar]
- 7. Stiell I.G., Wells G.A., Vandemheen K., Clement C., Lesiuk H., Laupacis A., McKnight R.D., Verbeek R., Brison R., Cass D., Eisenhauer M.E., Greenberg G., and Worthington J. (2001). The Canadian CT Head Rule for patients with minor head injury. Lancet 357, 1391–1396 [DOI] [PubMed] [Google Scholar]
- 8. Hodgkinson S., Pollit V., Sharpin C., and Lecky F.; National Institute for Health and Care Excellence (NICE) Guideline Development Group. (2014). Early management of head injury: summary of updated NICE guidance. BMJ 348, g104. [DOI] [PubMed] [Google Scholar]
- 9. Brenner D. and Hall E. (2007). Computed tomography—an increasing source of radiation exposure. N. Engl. J. Med. 357, 2277–2284 [DOI] [PubMed] [Google Scholar]
- 10. Isokuortti H., Luoto T., Kataja A., Brander A., Siironen J., Liimatainen S., Iverson G., Ylinen A., and Ohman J. (2014). Necessity of monitoring after negative head CT in acute head injury. Injury 45, 1340–1344 [DOI] [PubMed] [Google Scholar]
- 11. Undén J, Ingebrigtsen T, and Romner B; Scandinavian Neurotrauma Committee (SNC). (2013). Scandinavian guidelines for initial management of minimal, mild and moderate head injuries in adults: an evidence and consensus-based update. BMC Med. 11,50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Ohrt Nissen S., Friis Hansen L., Dahl B., Stensballe J., Romner B., and Rasmussen L. (2011). How does extracerebral trauma affect the clinical value of S100B measurements? Emerg. Med. J. 28, 941–944 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Hasselblatt M., Mooren F.C., von Ahsen N., Keyvani K., Fromme A., Schwarze Eicker K., Senner V., and Paulus W. (2004). Serum S100beta increases in marathon runners reflect extracranial release rather than glial damage. Neurology 62, 1634–1636 [DOI] [PubMed] [Google Scholar]
- 14. Missler U., Wiesmann M., Wittmann G., Magerkurth O., and Hagenstrom H. (1999). Measurement of glial fibrillary acidic protein in human blood: analytical method and preliminary clinical results. Clin. Chem. 45, 138–141 [PubMed] [Google Scholar]
- 15. Posti J.P., Takala R.S.K., Runtti H., Newcombe V.F., Outtrim J., Katila A.J., Frantzén J., Ala-Seppälä H., Coles J.P., Iftakher Hossain M., Kyllönen A., Maanpää H.-R., Tallus J., Hutchinson P.J., Van Gils M., Menon D.K., and Tenovuo O. (2016). The levels of glial fibrillary acidic protein and ubiquitin C-terminal hydrolase-L1 during the first week after a traumatic brain injury: correlations with clinical and imaging findings. Neurosurgery 79. [DOI] [PubMed] [Google Scholar]
- 16. Welch R., Ayaz S., Lewis L., Unden J., Chen J., Mika V., Saville B., Tyndall J., Nash M., Buki A., Barzo P., Hack D., Tortella F., Schmid K., Hayes R., Vossough A., Sweriduk S., and Bazarian J. (2016). Ability of serum glial fibrillary acidic protein, ubiquitin C-terminal hydrolase-L1, and S100B to differentiate normal and abnormal head computed tomography findings in patients with suspected mild or moderate traumatic brain injury. J. Neurotrauma 33, 203–214 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Bogoslovsky T., Wilson D., Chen Y., Hanlon D., Gill J., Jeromin A., Song L., Moore C., Gong Y., Kenney K., and Diaz-Arrastia R. (2017). Increases of plasma levels of glial fibrillary acidic protein, tau, and amyloid beta up to 90 days after traumatic brain injury. J. Neurotrauma 34, 66–73 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Luoto T.M., Raj R., Posti J.P., Gardner A.J., Panenka W.J., and Iverson G.L. (2017). A systematic review of the usefulness of glial fibrillary acidic protein for predicting acute intracranial lesions following head trauma. Front. Neurol. 8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Bazarian J.J., Blyth B.J., He H., Mookerjee S., Jones C., Kiechle K., Moynihan R., Wojcik S.M., Grant W.D., Secreti L.M., Triner W., Moscati R., Leinhart A., Ellis G.L., and Khan J. (2013). Classification accuracy of serum Apo A-I and S100B for the diagnosis of mild traumatic brain injury and prediction of abnormal initial head computed tomography scan. J. Neurotrauma 30, 1747–1754 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Papa L., Silvestri S., Brophy G.M., Giordano P., Falk J.L., Braga C.F., Tan C.N., Ameli N.J., Demery J.A., Dixit N.K., Mendes M.E., Hayes R.L., Wang K.K., and Robertson C.S. (2014). GFAP out-performs S100beta in detecting traumatic intracranial lesions on computed tomography in trauma patients with mild traumatic brain injury and those with extracranial lesions. J. Neurotrauma 31, 1815–1822 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Bazarian J.J., Biberthaler P., Welch R.D., Lewis L.M., Barzo P., Bogner-Flatz V., Gunnar Brolinson P., Büki A., Chen J.Y., Christenson R.H., Hack D., Huff J.S., Johar S., Jordan J.D., Leidel B.A., Lindner T., Ludington E., Okonkwo D.O., Ornato J., Peacock W.F., Schmidt K., Tyndall J.A., Vossough A., and Jagoda A.S. (2018). Serum GFAP and UCH-L1 for prediction of absence of intracranial injuries on head CT (ALERT-TBI): a multicentre observational study. Lancet Neurol. 17, 782–789 [DOI] [PubMed] [Google Scholar]
- 22. Furuhashi M. and Hotamisligil G. (2008). Fatty acid-binding proteins: role in metabolic diseases and potential as drug targets. Nat. Rev. Discov. 7, 489–503 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Lagerstedt L., Egea Guerrero J., Bustamante A., Montaner J., Rodríguez Rodríguez A., El Rahal A., Turck N., Quintana M., García Armengol R., Prica C., Andereggen E., Rinaldi L., Sarrafzadeh A., Schaller K., and Sanchez J.C. (2017). H-FABP: a new biomarker to differentiate between CT-positive and CT-negative patients with mild traumatic brain injury. PLoS One 12, e0175572. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Lagerstedt L., Egea Guerrero J., Rodríguez-Rodríguez A., Bustamante A., Montaner J., El Rahal A., Andereggen E., Rinaldi L., Sarrafzadeh A., Schaller K., and Sanchez J.C. (2018). Early measurement of interleukin-10 predicts the absence of CT scan lesions in mild traumatic brain injury. PLoS One 13, e0193278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Franz G., Beer R., Kampfl A., Engelhardt K., Schmutzhard E., Ulmer H., and Deisenhammer F. (2003). Amyloid beta 1–42 and tau in cerebrospinal fluid after severe traumatic brain injury. Neurology 60, 1457–1461 [DOI] [PubMed] [Google Scholar]
- 26. Ljungqvist J., Zetterberg H., Mitsis M., Blennow K., and Skoglund T. (2017). Serum neurofilament light protein as a marker for diffuse axonal injury: results from a case series study. J. Neurotrauma 34, 1124–1127 [DOI] [PubMed] [Google Scholar]
- 27. Shahim P., Zetterberg H., Tegner Y., and Blennow K. (2017). Serum neurofilament light as a biomarker for mild traumatic brain injury in contact sports. Neurology 88, 1788–1794 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Rubenstein R., Chang B., Yue J., Chiu A., Winkler E., Puccio A., Diaz Arrastia R., Yuh E., Mukherjee P., Valadka A., Gordon W., Okonkwo D., Davies P., Agarwal S., Lin F., Sarkis G., Yadikar H., Yang Z., Manley G., Wang K.K.W., Cooper S., Dams O'Connor K., Borrasso A., Inoue T., Maas A.I.R., Menon D., Schnyer D., and Vassar M. (2017). Comparing plasma phospho tau, total tau, and phospho tau-total tau ratio as acute and chronic traumatic brain injury biomarkers. JAMA Neurol. 74, 1063–1072 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Wilson D.H., Rissin D.M., Kan C.W., Fournier D.R., Piech T., Campbell T.G., Meyer R.E., Fishburn M.W., Cabrera C., Patel P.P., Frew E., Chen Y., Chang L., Ferrell E.P., von Einem V., McGuigan W., Reinhardt M., Sayer H., Vielsack C., and Duffy D.C. (2016). The simoa HD-1 analyzer: a novel fully automated digital immunoassay analyzer with single-molecule sensitivity and multiplexing. J. Lab. Autom. 21, 533–547 [DOI] [PubMed] [Google Scholar]
- 30. Ala-Seppälä H., Heino I., Frantzén J., Takala R.S.K., Katila A.J., Kyllönen A., Maanpää H.-R., Posti J.P., Tallus J., and Tenovuo O. (2016). Injury profiles, demography and representativeness of patients with TBI attending a regional emergency department. Brain Inj. 30. [DOI] [PubMed] [Google Scholar]
- 31. Takala R.S.K., Posti J.P., Runtti H., Newcombe V.F., Outtrim J., Katila A.J., Frantzén J., Ala-Seppälä H., Kyllönen A., Maanpää H.-R., Tallus J., Hossain M.I., Coles J.P., Hutchinson P., Van Gils M., Menon D.K., and Tenovuo O. (2016). Glial fibrillary acidic protein and ubiquitin C-terminal hydrolase-L1 as outcome predictors in traumatic brain injury. World Neurosurg. 87. [DOI] [PubMed] [Google Scholar]
- 32. Posti J.P., Hossain I., Takala R.S.K., Liedes H., Newcombe V., Outtrim J., Katila A.J., Frantzén J., Ala-Seppälä H., Coles J.P., Kyllönen A., Maanpää H.R., Tallus J., Hutchinson P.J., Van Gils M., Menon D.K., and Tenovuo O. (2017). Glial fibrillary acidic protein and ubiquitin C-terminal hydrolase-L1 are not specific biomarkers for mild CT-negative traumatic brain injury. J. Neurotrauma 34. [DOI] [PubMed] [Google Scholar]
- 33. Brain Trauma Foundation; American Association of Neurological Surgeons; Congress of Neurological Surgeons; Joint Section on Neurotrauma and Critical Care AA.NS/CNS, Carney NA, and Ghajar J. (2007). Guidelines for the management of severe traumatic brain injury. Introduction. J. Neurotrauma 24 Suppl 1, S1–S2 [DOI] [PubMed] [Google Scholar]
- 34. Marshall L.F., Marshall Klauber M.R., S.B., van Berkum Clark M., Eisenberg H.M., Jane J.A., Luerssen T.G., Marmarou A., and Foulkes M.A. (1991). A new classification of head injury based on computerized tomography. J. Neurosurg. 75, S20 [Google Scholar]
- 35. Ruopp M.D., Perkins N.J., Whitcomb B.W., and Schisterman E.F. (2008). Youden Index and optimal cut-point estimated from observations affected by a lower limit of detection. Biometrical J. 50, 419–430 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Robin X., Turck N., Hainard A., Tiberti N., Lisacek F., Sanchez J.C., and Müller M. (2013). PanelomiX: A threshold-based algorithm to create panels of biomarkers. Transl. Proteomics 1, 57–64 [Google Scholar]
- 37. Diaz-Arrastia R., Wang K.K., Papa L., Sorani M.D., Yue J.K., Puccio A.M., McMahon P.J., Inoue T., Yuh E.L., Lingsma H.F., Maas A.I., Valadka A.B., Okonkwo D.O., and Manley G.T.; TRACK-TBI Investigators. (2014). Acute biomarkers of traumatic brain injury: relationship between plasma levels of ubiquitin C-terminal hydrolase-L1 and glial fibrillary acidic protein. J. Neurotrauma 31, 19–25 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Papa L., Brophy G., Welch R., Lewis L., Braga C., Tan C., Ameli N., Lopez M., Haeussler C., Mendez Giordano D., Silvestri S., Giordano P., Weber K., Hill Pryor C., and Hack D. (2016). Time course and diagnostic accuracy of glial and neuronal blood biomarkers GFAP and UCH-L1 in a large cohort of trauma patients with and without mild traumatic brain injury. JAMA Neurol. 73, 551–560 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Welch R., Ellis M., Lewis L., Ayaz S., Mika V., Millis S., and Papa L. (2017). Modeling the kinetics of serum glial fibrillary acidic protein, ubiquitin carboxyl-terminal hydrolase-L1, and S100B concentrations in patients with traumatic brain injury. J. Neurotrauma 34, 1957–1971 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Smits M., Dippel D.W.J., de Haan G.G., Dekker H.M., Vos P.E., Kool D.R., Nederkoorn P.J., Hofman P.A.M., Twijnstra A., Tanghe H.L.J., and Hunink M.G.M. (2007). Minor head injury: cuidelines for the use of CT—a multicenter validation study. Radiology 245, 831–838 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
De-identified clinical, imaging, and biochemical data not published within the article can be shared with a qualified investigator by request.