Skip to main content
. 2017 Dec 4;8:652. doi: 10.3389/fneur.2017.00652

Table 1.

Summary of study results.

Controls
 TBI
Reference Method (GFAP) Lower level of detection (ng/mL) N, type Age, years Male (%) GFAP level,
ng/mL		 N,
severity		 Age,
years		 Male (%) Time of blood sampling (h after injury) CT-positive,
GFAP level,
ng/mL		 CT-negative,
GFAP level,
ng/mL		 Acute traumatic lesions on head CT,
n (%)		 GFAP related to lesions Extracranial injuries accounted for
Adult studies, mostly MTBI
Bogoslovsky et al. (28) Quanterix Corporation 0.0008 n = 69,
healthy volunteers		 Md = 45,
IQR = 31–52		 n = 35 (51%) Md = 0.0008,
IQR = 0.0008–0.00107		 n = 34,
mild-severe		 Md = 39,
IQR = 23–52		 n = 29 (85%) Admission,
≤22		 Md = 0.0176,
IQR = 0.00388–0.1296		 N/A n = 34 (100%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; trauma center; USA, eight sites; 2007–2011
Main findings: CT-positive TBIs had significantly higher GFAP levels than controls. GFAP was able to discriminate CT-positive TBIs from controls
AU-ROC: 0.94 (CT+ vs. controls); cutoff: N/A
Buonora et al. (29) Meso Scale discovery 0.21 n = 74,
healthy volunteers		 M = 47,
SD = 19		 n = 15 (30%) <0.3 n = 260,
mild-severe		 Mild-moderate:
M = 47,
SD = 19,
moderate-severe:
M = 47,
SD = 21		 n = 188 (72%) At admission,
no details provided		 N/A N/A n = 132 (51%) No Yes
Study design, setting, country, and number of sites; study year(s): case–control; trauma center; USA, two sites; Canada, four sites; N/A
Main findings: no relation between GFAP and CT findings in the cohort with mild-moderate TBI. Higher levels of GFAP in moderate-severe TBI patients compared to controls
AU-ROC: N/A; cutoff: N/A
Diaz-Arrastia et al. (19)a Banyan biomarkers 0.1 n = 175,
healthy volunteers		 M = 34,
SD = 14		 n = 93 (53%) N/A n = 206,
mild-severe		 M = 42,
SD = 18		 n = 150 (73%) M = 10.9,
SD = 6.4,
min = 0.5,
max = 24.3		 N/A N/A n = 106 (51%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; trauma center; USA, three sites; N/A
Main findings: GFAP was able to discriminate: (i) CT-positive TBIs from CT-negative TBIs and (ii) TBIs from controls.
AU-ROC: 0.88 (CT+ vs. CT−) and 0.91 (TBI vs. controls); cutoff: N/A
Honda et al. (15) BioVendor 0.1 N/A N/A N/A N/A n = 34,
mild-severe		 CT-positives:
Md = 72,
IQR = 54–85;
CT-negatives:
Md = 41,
IQR = 30–59		 n = 22 (65%) At admission,
≤3		 N/A N/A n = 18 (53%) Yes No
Study design, setting, country, and number of sites; study year(s): cohort; trauma center, Japan, one site; 2006–2007
Main findings: GFAP was able to discriminate CT-positive TBIs from CT-negative TBIs.
AU-ROC: 0.98 (CT+ vs. CT−); cutoff: N/A
Lumpkins et al. (30) BioVendor N/A N/A N/A N/A N/A n = 51,
mild-severe		 M = 43,
SD = 21		 n = 37 (73%) At admission, no details provided M = 0.00677,
SD = 0.01005		 M = 0.00007,
SD = 0.00018		 n = 39 (76%) Yes Yes
Study design, setting, country, and number of sites; study year(s): cohort; trauma center, USA, one site; 2005–2006
Main findings: CT-positive TBIs had significantly higher GFAP levels than CT-negative TBIs. GFAP could discriminate CT-positive TBIs from CT-negative TBIs. Also, patients with surgical CT lesions had significantly higher GFAP levels than patients with diffuse lesions.
AU-ROC: 0.90 (CT+ vs. CT−); cutoff: 0.001 ng/mL, sensitivity = 62%, specificity = 100%
McMahon et al. (20)a Banyan Biomarkers 0.01 N/A N/A N/A N/A n = 215,
mild-severe		 M = 42,
SD = 18		 n = 156 (73%) At admission, ≤24 M = 2.86,
SD = 3.74		 M = 0.26,
SD = 0.4		 n = 110 (51%) Yes Yes
Study design, setting, country, and number of sites; study year(s): cohort; trauma center, USA, one site; N/A
Main findings: CT-positive TBIs had significantly higher GFAP levels than CT-negative TBIs. GFAP could discriminate CT-positive TBIs from CT-negative TBIs.
AU-ROC: 0.87 (CT+ vs. CT−); cutoff: 1.66 ng/mL, sensitivity = 45%, specificity = 99%, Brier score = 0.29
Metting et al. (31) BioVendor 0.045 N/A N/A N/A N/A n = 94,
mild		 M = 34.3,
SD = 13.9		 N/A M = 2.4,
SD = 2.1		 M = 1.20,
SD = 1.65		 M = 0.05,
SD = 0.17		 n = 19 (20%) Yes Yes
Study design, setting, country, and number of sites; study year(s): cohort; University hospital, The Netherlands, one site; 2005–2007
Main Findings: CT-positive TBIs had significantly higher GFAP levels than CT-negative TBIs
AU-ROC: N/A; cutoff: N/A
Okonkwo et al. (18)a Banyan Biomarkers 0.1 N/A N/A N/A N/A n = 215,
mild-severe		 M = 42,
SD = 18		 n = 157 (73%) M = 10.9,
SD = 6.4,
min = 0.5,
max = 23.4		 M = 2.86,
SD = 3.74		 M = 0.26,
SD = 0.41		 n = 109 (51%) Yes No
Study design, setting, country, and number of sites; study year(s): cohort; trauma center, USA, three sites; N/A
Main Findings: GFAP was able to discriminate CT-positive TBIs from CT-negative TBIs
AU-ROC: 0.88 (CT+ vs. CT−); cutoff: 0.68 ng/mL, sensitivity = 73%, specificity = 89%, positive predictive value = 87%
Papa et al. (16)d Banyan Biomarkers 0.000008 n = 188,
non-TBI trauma controls		 M = 40,
SD = 16		 n = 103 (55%) N/A n = 209,
mild-moderate		 M = 40,
SD = 16		 n = 131 (63%) M = 3.1,
95% CI = 3.0–3.3		 N/A N/A n = 20 (10%) Yes Yes
Study design, setting, country, and number of sites; study year(s): case–control; trauma center, USA, one site; N/A
Main Findings: GFAP was able to discriminate CT-positive TBIs from CT-negative TBIs
AU-ROC: 0.84 (CT+ vs. CT−); cutoff: 0.067 ng/mL, sensitivity = 100%, specificity = 55%, negative predictive value = 100%, positive predictive value = 20%
Papa et al. (32) Banyan Biomarkers 0.02 n = 199,
no injuries and trauma controls		 No injuries:
M = 37,
SD = 14;
trauma controls:
M = 44,
SD = 17		 n = 109 (55%) (no injuries: n = 93; 53%; trauma controls: n = 16; 70%) M = 0.057,
95% CI = 0.044–0.071		 n = 108,
mild-moderate		 M = 39,
SD = 15		 n = 70 (65%) M = 2.6,
95% CI = 2.4–2.9		 N/A N/A n = 32 (30%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; Trauma center, USA, 3 sites; N/A
Main findings: GFAP could discriminate CT-positive TBIs from CT-negative TBIs. GFAP was more reliable in discriminating: (i) TBIs from controls and (ii) TBIs with surgical CT lesions from non-surgical lesions
AU-ROC: 0.79 (CT+ vs. CT−) and 0.90 (TBI vs. controls); cutoff: 0.035 ng/mL, sensitivity = 97%, specificity = 18%, negative predictive value = 94%
Papa et al. (23) Banyan Biomarkers 0.008 n = 259,
trauma controls		 M = 41,
SD = 16,
range = 18–83		 n = 150 (58%) Md = 0.008,
IQR = 0.008–0.030; range = 0.008–0.773		 n = 325,
mild-moderate		 M = 39,
SD = 16,
range = 18–78		 n = 212 (65%) M = 3.0,
SD = 0.9		 Md = 0.588,
IQR = 0.140–2.014,
range = 0.008–8.078		 Md = 0.033,
IQR = 0.008–0.189,
range = 0.008–7.785		 n = 35 (11%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; trauma center, USA, one site; 2010–2004
Main findings: CT-positive TBIs had significantly higher GFAP levels than CT-negative TBIs. GFAP could discriminate CT-positive TBIs from CT-negative TBIs
AU-ROC: 0.86 (CT+ vs. CT−); cutoff: N/A
Posti et al. (21) The Evidence Investigator Cerebral Custom Array IV N/A n = 81,
orthopedic controls		 M = 44.9,
SD = 18.8		 n = 35 (43%) N/A n = 324,
mild-severe		 M = 45.3,
SD = 19.2		 n = 238 (74%) At admission,
<24		 N/A N/A n = 200 (69%) Yes No
Study design, setting, country, and number of sites; study year(s): cohort; University hospital; Finland, one site; the United Kingdom, one site; 2011–2003
Main findings: CT-positive TBIs had significantly higher GFAP levels than CT-negative TBIs. GFAP could discriminate CT-positive TBIs from CT-negative TBIs. Also, GFAP levels were significantly higher in patients with mass lesions than with non-mass lesions
AU-ROC: 0.74 (CT+ vs. CT−); cutoff: N/A
Shehab and Nassar (33) ELISA assay, not otherwise specified N/A n = 20,
healthy volunteers		 N/A N/A M = 0.0015,
SD = 0.00037		 n = 70,
mild-severe		 M = 40.8,
SD = 8,
range = 22–64		 n = 52 (74%) At admission,
no details available		 M = 0.1029,
SD = 0.0471		 M = 0.0668,
SD = 0.0224		 n = 43 (61%) Yes Yes
Study design, setting, country, and number of sites; study year(s): case–control; University hospital, Egypt, one site; N/A
Main findings: CT-positive TBIs had significantly higher GFAP levels than CT-negative TBIs. Also, TBIs had significantly higher GFAP levels than controls
AU-ROC: N/A; cutoff: N/A
Welch et al. (17)e Banyan Biomarkers 0.02 N/A N/A N/A N/A n = 231,
mild-moderate		 M = 45.6,
SD = 18.4		 n = 151 (60%) At admission,
≤6		 Md = 0.1105,
IQR = 0.0204–0.4318		 Md = 0.0078,
IQR = 0.0027–0.0221		 n = 36 (14%) Yes No
Study design, setting, country, and number of sites; study year(s): cohort; trauma center; USA, five sites; Hungary, two sites; N/A
Main findings: GFAP could discriminate CT-positive TBIs from CT-negative TBIs
AU-ROC: 0.79 (CT+ vs. CT−); cutoff: 0.015 ng/mL, sensitivity = 81%, specificity = 67%
Welch et al. (22)e Banyan Biomarkers 0.02 N/A N/A N/A N/A n = 167,
mild-moderate		 M = 46.0,
SD = 17.8		 n = 102 (61%) Multiple time points:
0–6,
>6–12,
>12–18,
and >18–24 h		 Md = 0.122,
IQR = 0.020–0.437		 Md = 0.010,
IQR = 0.004–0.031		 n = 33 (20%) Yes No
Study design, setting, country, and number of sites; study year(s): cohort; trauma center; USA, five sites; Hungary, two sites; N/A
Main findings: GFAP could discriminate CT-positive TBIs from CT-negative TBIs
AU-ROC: 0.84–0.94 (CT+ vs. CT−); cutoff: N/A
Adult studies, moderate and severe TBI
Lei et al. (34) BioVendor 0.045 n = 135,
healthy blood donors		 M = 39.2,
SD = 15.3,
range = 18–65		 n = 88 (65%) Md = 0,
IQR = 0–0,
range = 0.048–0.076		 n = 67,
severe		 M = 37.2,
SD = 14.3		 n = 51 (76%) At admission,
≤4		 Md = 1.924,
IQR = 0.891–3.126		 N/A n = 64 (100%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; trauma center, China, one site; 2011–2004
Main findings: TBIs had significantly higher GFAP levels than controls. Also, patients with surgical CT lesions had significantly higher GFAP levels than patients with diffuse lesions
AU-ROC: N/A; cutoff: N/A
Mondello et al. (35)f BioVendor N/A n = 167,
healthy blood donors		 M = 36.9,
SD = 14.1		 n = 95 (57%) M = 0.07,
SD = 0.03		 n = 81,
severe		 M = 47.9,
SD = 20.4		 n = 65 (80%) At admission,
no details provided		 N/A N/A n = 80 (99%) Yes Yes
Study design, setting, country, and number of sites; study year(s): case–control; trauma center; USA, two sites; Hungary, two sites; N/A
Main findings: TBIs had significantly higher GFAP levels than controls. Also, patients with mass lesions on CT had significantly higher GFAP levels than patients with diffuse lesions
AU-ROC: N/A; cutoff: N/A
Mondello et al. (36)f BioVendor N/A N/A N/A N/A N/A n = 59,
severe		 M = 46.7,
range = 19–89		 n = 46 (78%) M = 9,
SEM = 1		 N/A N/A n = 58 (98%) Yes No
Study design, setting, country, and number of sites; study year(s): cohort; trauma center; USA, two sites; Hungary, two sites; N/A
Main findings: TBI patients with mass lesions on CT had significantly higher GFAP levels than patients with diffuse lesions. GFAP was able to discriminate TBIs with mass lesions from TBIs with diffuse lesions
AU-ROC: 0.72 (mass lesions vs. diffuse lesions); cutoff: N/A
Pelinka et al. (37)b LIAISON® GFAP and S100B assay 0.03 N/A N/A N/A N/A n = 92,
moderate-severe		 Md = 39,
IQR = 28–55		 n = 67 (73%) At admission,
<12		 N/A N/A n = 92 (100%) Yes No
Study design, setting, country, and number of sites; study year(s): cohort; trauma center, Austria, three sites; 1999–2002
Main Findings: GFAP levels were positively related to the severity of traumatic CT findings (Marshall grade)
AU-ROC: N/A; cutoff: N/A
Pelinka et al. (38)b LIAISON® GFAP and S100B assay 0.03 n = 13,
polytrauma patients		 Md = 39,
IQR = 28–48		 n = 7 (54%) N/A n = 101,
moderate-severe		 Md = 39,
IQR = 27–55		 n = 76 (75%) At admission,
<12		 N/A N/A n = 101 (100%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; trauma center, Austria, three sites; 1999–2003
Main findings: GFAP levels were positively related to the severity of traumatic CT findings (Marshall grade)
AU-ROC: N/A; cutoff: N/A
Vos et al. (39) Future diagnostics N/A N/A N/A N/A N/A n = 79,
moderate-severe		 M = 47.0,
range = 18–91		 n = 57 (72%) Md = 1,
IQR = 0.5–5		 Md = 0.1–2.17 Md = 0.02,
95% CI = 0.02–1		 n = 64 (84%) Yes No
Study design, setting, country, and number of sites; study year(s): cohort; trauma center, The Netherlands, one site; 2004–2006
Main findings: GFAP levels were significantly related to the severity of traumatic CT findings (Marshall grade)
AU-ROC: N/A; cutoff: N/A
Vos et al. (40) ELISA assay,
not otherwise specified		 N/A N/A N/A N/A N/A n = 85,
severe		 Md = 32,
range = 15–81		 n = 61 (72%) Md = 2.5,
range = 0.25–30		 N/A N/A n = 82 (96%) Yes No
Study design, setting, country, and number of sites; study year(s): Cohort; University hospital, The Netherlands, one site, 1999–2000
Main findings: GFAP levels were significantly related to the severity of traumatic CT findings (Marshall grade)
AU-ROC: N/A; cutoff: N/A
Pediatric studies
Fraser et al. (41) Ridascreen Risk Material 10/5 N/A N/A N/A N/A N/A n = 27,
severe		 M = 10.6,
SD 0.9,
range = 2.4–17		 n = 14 (52%) At admission,
no details available		 N/A N/A n = 27 (100%) No Yes
Study design, setting, country, and number of sites; study year(s): cohort; pediatric intensive care unit, Canada, four sites; N/A
Main findings: only severe CT-positive TBI cases included in the study. GFAP failed to correlate with traumatic CT abnormalities
AU-ROC: N/A; cutoff: N/A
Mondello et al. (42) Meso Scale Discovery N/A n = 40,
patients treated for trivial reason other than head injury		 M = 3.9,
SD = 3.8		 n = 23 (58%) Md = 0.01,
IQR = 0.00–0.05		 n = 45,
mild-severe		 M = 3.8,
SD = 3.8		 n = 28 (62%) Md = 4.7,
range = 0.5–20.6		 Md = 0.73,
IQR = 0.15–2.28		 Md = 0.21,
IQR = 0.08–1.37		 n = 29 (64%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; trauma center, USA, one site; N/A
Main findings: CT-positive TBIs had significantly higher GFAP levels than controls. GFAP could discriminate mild TBIs from controls. However, GFAP could not discriminate between CT-positive and CT-negative TBIs
AU-ROC: 0.81 (mild TBI vs. control); cutoff: N/A
Zurek and Fedora (43) BioVendor N/A N/A N/A N/A N/A n = 59,
severe		 M = 8.9 n = 36 (61%) At admission,
<3		 N/A N/A N/A No No
Study design, setting, country, and number of sites; study year(s): cohort; University hospital, The Czech Republic, one site; 2007–2009
Main findings: GFAP failed to correlate with traumatic CT abnormalities
AU-ROC: N/A; cutoff: N/A
Papa et al. (44)c Banyan Biomarkers 0.000008 n = 60,
non-TBI trauma controls		 M = 12,
SD = 6,
range = 0.1–21		 n = 39 (65%) Md = 0.03,
IQR = 0.01–0.05		 n = 197,
mild-moderate		 M = 11.5,
SD = 7,
range = 0.1–21		 n = 131 (66%) M = 3.3,
95% CI = 3.1–3.5		 Md = 1.01,
95% CI = 0.59–1.48		 Md = 0.18,
95% CI = 0.06–0.47		 n = 18 (12%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; trauma center, USA, three sites; N/A
Main findings: CT-positive TBIs had significantly higher GFAP levels than CT-negative TBIs. GFAP could discriminate CT-positive TBIs from CT-negative TBIs
AU-ROC: 0.82 (CT+ vs. CT−); cutoff: 0.15 ng/mL, sensitivity = 94%, specificity = 47%, negative prediction value = 98%
Papa et al. (45)c Banyan Biomarkers 0.000008 n = 42,
non-TBI orthopedic trauma controls		 M = 13,
SD = 5		 n = 24 (59%) Md = 0.03,
IQR = 0.01–0.06		 n = 114,
mild-moderate		 M = 13,
SD = 7		 n = 76 (67%) At admission,
<6 h		 Md = 1.19,
IQR = 0.78–5.13		 Md = 0.25,
IQR = 0.10–0.63		 n = 8 (9%) Yes No
Study design, setting, country, and number of sites; study year(s): case–control; trauma center, USA, three sites; N/A
Main findings: head injury patients had significantly higher GFAP levels than controls. GFAP could discriminate CT-positive TBIs from CT-negative TBIs
AU-ROC: 0.85 (CT+ vs. CT−); cutoff: 0.15 ng/mL, sensitivity = 100%, specificity = 36%, likelihood ratio = 1.6

M, mean; Md, median; GFAP, glial fibrillary acidic protein; GFAP-BDP, glial fibrillary acidic protein breakdown product; CT, computed tomography; CDE, common data elements; AU-ROC, area under the receiver operating curve; IQR, interquartile range; 95% CI, 95% confidence interval; TBI, traumatic brain injury; CT+, CT-positive (patients with acute traumatic intracranial lesions on head CT); CT−, CT-negative (patients with no acute traumatic intracranial lesions on head CT); cutoff, GFAP cutoff level for a trauma-positive head CT.

aTRACK-TBI; studies contain some overlapping cases, also GFAP-BDP levels were measured.

bMostly the same sample in both studies.

cPartly the same sample in both studies.

dGFAP-BDP levels were also measured.

eMostly the same sample in both studies.

fMostly the same BANDITS sample in both studies.

GFAP platforms:

1. Quanterix Corporation, Lexington, MA, USA (method:single molecule array).

2. Banyan Biomarkers, Alachua, FL, USA [method: enzyme-linked immunosorbent assay (ELISA)].

3. BioVendor, Brno, Czech Republic; Candler, NC, USA; and Heidelberg, Germany (method: ELISA).

4. The Evidence Investigator Cerebral Custom Array IV Randox Laboratories Ltd., Crumlin, County Antrim, United Kingdom (method: digital immonoassay technology).

5. Meso Scale Discovery, Gaithersburg, MD, USA (method: electro-chemiluminescent immunoassay).

6. LIAISON® GFAP and S100B assay, AB Sangtec Medical, Bromma, Sweden (method: monoclonal immunoluminometric assay).

7. Future diagnostics, Wijchen, the Netherlands (method: 2-site luminometric immunoassay).

8. Ridascreen Risk Material 10/5, R-Biopharm AG, Darmstadt, Germany (method: ELISA).