Diagnostic accuracy of GFAP and UCH-L1 for intracranial hemorrhage in blunt head trauma: a prospective study

Abstract

Background

Early and accurate identification of intracranial hemorrhage (ICH) following traumatic brain injury (TBI) is crucial, yet conventional imaging techniques like CT have limitations including radiation exposure and overutilization. This study aimed to evaluate the diagnostic utility of two blood-based biomarkers, Glial Fibrillary Acidic Protein (GFAP) and Ubiquitin Carboxy-Terminal Hydrolase L1 (UCH-L1), in detecting ICH in emergency department patients with blunt head trauma.

Methods

This single-center prospective observational study included 442 patients presenting with blunt head trauma at a university emergency department. Plasma GFAP and UCH-L1 levels were measured prior to cranial CT imaging. ROC curve analysis, sensitivity, specificity, and logistic regression were used to assess diagnostic performance and associations between biomarker levels and CT-confirmed ICH.

Results

ICH was detected in 6.1% of patients. GFAP and UCH-L1 both demonstrated high diagnostic performance with AUC values of 0.98 and 0.95, respectively. Sensitivity and specificity were 92.6% and 98.8% for GFAP, and 92.6% and 99.0% for UCH-L1. When both biomarkers were concurrently elevated, sensitivity remained at 92.6% with a negative predictive value of 99.5%. Univariate logistic regression revealed strong associations with ICH: GFAP (OR = 1025.0, 95% CI: 189.4–5548.3) and UCH-L1 (OR = 1284.4, 95% CI: 224.4–7352.3), both p < 0.001.

Conclusion

GFAP and UCH-L1 are highly sensitive and specific biomarkers for early ICH detection in blunt head trauma. Their combined use may enhance triage accuracy, reduce unnecessary imaging, and support rapid clinical decision-making in emergency settings. Larger multicenter studies are warranted to validate these findings and standardize cut-off values.

Background

Traumatic brain injury (TBI) is a significant global health issue that affects millions of individuals each year and leads to substantial morbidity and mortality. Computed tomography (CT) is the gold standard method used to diagnose intracranial hemorrhage (ICH). However, despite advances in neuroimaging techniques, early and accurate detection of ICH in patients with TBI remains a significant clinical challenge [1]. Additionally, concerns related to radiation exposure, accessibility issues, and the risks of unnecessary scans have prompted the search for alternative diagnostic methods [2, 3]. In this context, blood-based biomarkers—particularly Glial Fibrillary Acidic Protein (GFAP) and Ubiquitin C-terminal hydrolase-L1 (UCH-L1)—have emerged as promising tools for identifying patients at risk of ICH and for the early detection of brain injury [4, 5].

GFAP is an astrocyte-specific cytoskeletal protein. It is released into the bloodstream following astroglial injury and is a highly specific marker of structural brain damage. In contrast, UCH-L1 is a neuronal protein involved in protein degradation and homeostasis. Its elevation in blood indicates axonal damage [6, 7]. Several studies have shown that both biomarkers increase in TBI patients, with GFAP generally demonstrating higher specificity than UCH-L1 in detecting ICH [8, 9].

In a large multicenter prospective cohort study conducted by Bazarian et al., it was reported that GFAP and UCH-L1 may provide high sensitivity and negative predictive value for detecting intracranial injuries within the first 12 h following traumatic brain injury (TBI) [10]. In the prospective validation study by Papa et al., GFAP showed sensitivities of ~ 92–95% and specificities of ~ 40–50%, while UCH-L1 demonstrated sensitivities of ~ 90–94% with lower specificity; when combined, the two biomarkers yielded negative predictive values exceeding 98% [11]. Similarly, in the ALERT-TBI multicenter study, GFAP and UCH-L1 together achieved a sensitivity of 97.6% and NPV of 99.6% for ruling out CT-positive lesions. This finding is particularly significant in emergency settings, where the use of blood-based biomarkers could reduce the need for unnecessary head CT scans [10, 11]. However, other studies have shown that GFAP levels typically peak around 20 h after injury and remain elevated for a longer duration compared to UCH-L1. While UCH-L1 rises rapidly and declines within approximately 8 h, GFAP displays a more delayed and sustained elevation, which may make it a more stable biomarker for detecting intracranial hemorrhage (ICH) in later stages of TBI [5, 8, 12]. Comparative studies have suggested that GFAP predict positive CT findings better than other biomarkers such as UCH-L1 and S100B [13].

Given the limitations of CT, these biomarkers offer hope as complementary or alternative tools for identifying patients with ICH. In a multicenter study by Czeiter et al., GFAP and UCH-L1 have been reported to show promising sensitivity and negative predictive value for ruling out ICH, with potential to reduce unnecessary CT scans in selected patients [14]. Previous studies have shown that GFAP and UCH-L1 are associated with intracranial lesions, and that GFAP has been reported to show higher specificity for detecting ICH in some studies. Combined interpretation of these biomarkers has been suggested as a potential approach to improve diagnostic accuracy [1, 15, 16].

However, the variability of these biomarkers across different analytical platforms and patient populations remains a significant issue. For instance, a study comparing two immunoassay platforms—i-STAT^® and Alinity^®—reported a strong correlation between GFAP and UCH-L1 measurements [17]. Additionally, some studies have focused on glial fibrillary acidic protein breakdown products (GFAP-BDP) rather than total GFAP, as these fragments—generated by proteolytic enzymes such as calpain following traumatic brain injury—may appear earlier and more specifically in serum. Nonetheless, the current study did not assess GFAP-BDP levels; thus, the diagnostic potential of these breakdown products remains unexplored [18].

Besides GFAP and UCH-L1, other biomarkers such as microtubule-associated protein 2 (MAP-2) have also been investigated in earlier studies as potential indicators of neuronal injury, although MAP-2 was not included in the present analysis [19]. Similarly, early increases in GFAP and UCH-L1 have been linked to unfavorable clinical outcomes and progression of injury [20]. Furthermore, in patients with polytrauma and hemorrhagic shock, both biomarkers have shown significant correlations with the severity of brain injury and are being considered as prognostic indicators [21].

Although several large studies, including that of Papa et al. [11], have supported the clinical utility of GFAP and UCH-L1 in patients with mild TBI, variability in diagnostic performance has been reported across different study populations and assay platforms. Furthermore, while FDA-approved assays with established thresholds are available, differences in platforms and clinical settings still limit universal implementation. In this context, our study aimed to measure GFAP and UCH-L1 levels in patients presenting to the emergency department with blunt head trauma who underwent brain CT, and to evaluate their relationship with intracranial hemorrhage (ICH).

Methods

Ethical statement

Ethics approval was obtained from the Erzincan Binali Yıldırım University Clinical Research Ethics Committee (Approval No: 2024/112). This study was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants or their legal representatives prior to inclusion in the study. Laboratory and radiologic assessments were performed by personnel blinded to clinical outcomes.

Study design and setting

This prospective observational study was conducted at the Emergency Medicine Department of Erzincan Binali Yıldırım University Mengücek Gazi Training and Research Hospital, a tertiary academic teaching hospital, between June 1, 2024, and January 1, 2025. The emergency department manages approximately 365,000 patient visits annually (about 1,000 per day), including an estimated 36,000–55,000 trauma cases per year, of which approximately 10% represent head trauma. The trauma unit is staffed by board-certified emergency medicine specialists, supported by residents and rotating medical students under faculty supervision. All patients presenting with blunt head trauma to the emergency department during the study period were evaluated for eligibility. Patient recruitment was conducted on a consecutive basis during study period; however, due to practical limitations (e.g., staffing and nighttime hours), this is best described as a consecutive convenience sample.

Sample size estimation was based on an anticipated intracranial hemorrhage (ICH) prevalence of approximately 6% in similar emergency department populations with blunt head trauma, as reported in prior literature. To achieve adequate power for diagnostic performance, we assumed an expected sensitivity of approximately 90% for GFAP/UCH-L1 based on a multicenter validation study by Papa et al. [11]. Considering a 95% confidence level (α = 0.05), 80% statistical power (β = 0.20), and a 5% margin of error, we estimated that at least 25–30 ICH cases would be required. Therefore, enrolling approximately 400–450 patients was calculated to ensure sufficient power, and a total of 442 participants were ultimately included in the final analysis.

Inclusion criteria were:

• Presentation within 24 h post-injury,

• Blunt head trauma without penetration,

• Indication for non-contrast cranial computed tomography (CT) as determined by the treating physcian,

• Blood sampling prior to CT imaging,

• Provision of informed consent by the patient or their legal representative.

Exclusion criteria included:

• Unknown time of trauma,

• Penetrating head trauma,

• Pre-existing significant neurological disorders (defined as prior intracranial surgery intracranial tumors neurodegenerative diseases, epilepsy with ongoing seizures, or prior stroke with residual neurological deficits),

• Pregnancy,

• No CT performed,

• Inability to collect a blood sample,

• Lack of informed consent.

Definitions

Traumatic brain injury (TBI) was defined as blunt head trauma with or without loss of consciousness, presenting within 24 h of injury, and deemed by the treating emergency physician to warrant cranial CT. This definition was selected to capture the full spectrum of head trauma patients undergoing CT in the emergency setting and is consistent with definitions applied in prior biomarker investigations [5, 11]. Intracranial hemorrhage (ICH) was defined radiologically as the presence of epidural, subdural, subarachnoid, intraparenchymal, or intraventricular hemorrhage on CT. Skull fractures without associated hemorrhage were not classified as ICH.

Patients with concomitant extracranial injuries (multi-trauma) were not excluded, provided they met the inclusion criteria for head trauma and underwent cranial CT. However, data collection and analysis focused specifically on head trauma characteristics and intracranial findings. Concomitant injuries were documented but not included in the primary analyses.

Note on GCS

Although GCS is commonly used to define TBI severity, we did not use GCS as an inclusion criterion in order to capture a broader population of patients undergoing CT for head trauma. GCS scores were, however, recorded for descriptive and comparative analyses.

Screening, eligibility, consent, and data collection

All patients presenting with blunt head trauma within 24 h of injury during the study period were screened consecutively using the emergency department triage logs and the CT order queue when study coverage was available (consecutive convenience sampling). Eligibility was determined by the attending emergency medicine specialist according to predefined inclusion and exclusion criteria. Informed consent was obtained face-to-face from the patient or a legal representative after initial stabilization and prior to blood sampling and CT.

Clinical variables—including demographics (age, sex, prior neurological history), trauma mechanism and severity, time from injury, alcohol or drug intoxication, and all Canadian CT Head Rule (CCHR) criteria—were recorded prospectively at the bedside on a structured case report form (CRF) designed for this study. Each CCHR variable was marked as present (+) or absent (−) during the initial evaluation by the attending emergency medicine specialist. Demographic details and time stamps (arrival time, injury-to-blood draw interval) were obtained from patient interview, witness report, and the electronic medical record. Study-specific forms were subsequently transferred into the study database by the investigators.

Patients with concomitant extracranial injuries were eligible provided they met inclusion criteria for blunt head trauma and underwent cranial CT; however, the severity and distribution of extracranial injuries were not systematically recorded.

Laboratory personnel were blinded to clinical and outcome data, and radiologists interpreting cranial CT scans were blinded to biomarker results, which were processed in batches and not available to treating clinicians. A CONSORT-style flow diagram summarizing screened, eligible, consented, excluded, and analyzed patients is provided in the Results.

Sample collection and storage

Venous blood samples were obtained using sterile technique and transferred to EDTA tubes. Samples were centrifuged at 3000 rpm for 10 min, and plasma was separated and stored at − 80 °C in the Biochemistry Laboratory of Erzincan Mengücek Gazi Training and Research Hospital until analysis. To account for biomarker kinetics, the time between trauma and blood collection was recorded, with an average duration of 4.2 ± 1.8 h. This allowed assessment of biomarker levels relative to their known kinetic profiles—UCH-L1 peaks within 8 h and GFAP within 20 h post-injury.

Biomarker measurement via ELISA

Plasma levels of GFAP and UCH-L1 were quantified using commercially available Human GFAP ELISA Kit (FY-EH4526) and Human UCHL1 ELISA Kit (FY-EH3915) (Wuhan Feiyue Biotechnology Co., Ltd.), following the manufacturer’s instructions. According to the manufacturer’s datasheets, assay performance characteristics were as follows: GFAP, sensitivity 9.38 pg/mL, detection range 15.63–1000 pg/mL, inter-assay CV < 10%; UCH-L1, sensitivity 46.88 pg/mL, detection range 78.13–5000 pg/mL, inter-assay CV < 10%. In our cohort, fewer than 5% of samples were below the lower limit of quantification, and none exceeded the upper limit. These assays are research-use-only (RUO) and not FDA-approved for clinical diagnostics; however, similar RUO ELISA platforms have been employed in multiple exploratory TBI studies [22, 23].

Statistical analysis

Statistical analysis was performed using IBM SPSS version 25.0 (SPSS Inc., Chicago, IL, USA). Quantitative variables were presented as mean ± standard deviation, and categorical variables as frequency and percentage. Group comparisons for categorical variables were conducted using Pearson’s Chi-square or Fisher’s Exact test, while numerical data were analyzed using the Mann–Whitney U test.

Receiver Operating Characteristic (ROC) curve analysis was used to evaluate the diagnostic performance of each biomarker, with Area Under the Curve (AUC) values reported. Cut-off values were determined using the Youden Index (J = Sensitivity + Specificity – 1).

Diagnostic parameters including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and their 95% confidence intervals (CIs) were calculated for each cut-off. While manufacturer reference ranges were considered (GFAP: 15.63–1000 pg/mL; UCH-L1: 78.13–5000 pg/mL), the optimal thresholds were determined empirically based on study-specific ROC analysis. Inter-assay coefficients of variation (CV) were within 10%.

Additionally, combined diagnostic performance was assessed by evaluating cases where both biomarkers were concurrently above or below their respective the cut-off. For logistic regression analysis, biomarker concentrations were dichotomized according to the optimal thresholds derived from ROC curve analysis using the Youden Index (GFAP: 136.2 pg/mL; UCH-L1: 311.5 pg/mL). The categorized variables (high vs. low) were entered as binary predictors to calculate odds ratios (ORs) for the presence of ICH. Odds ratios, 95% confidence intervals, and p-values were reported, with p < 0.05 considered statistically significant.

Results

Study population and patient flow

During the study period, a total of 722 patients presenting with blunt head trauma were screened for eligibility. Of these, 210 declined participation after the study protocol was explained. Among the remaining 512 patients, 70 were excluded for the following reasons: penetrating head trauma (n = 4), time since injury > 24 h (n = 19), incomplete data forms (n = 34), pre-existing significant neurological disorders such as prior stroke (n = 4), and loss of sample identification due to detached barcodes (n = 9). The final study population comprised 442 patients who met all inclusion criteria and were included in the analysis.

A detailed flow chart summarizing patient screening, exclusions, and final inclusion is shown in Fig. 1.

Fig. 1.

CONSORT-style flow diagram of patient screening and enrollment

Study population characteristics

Of the 442 included patients, 282 (63.8%) were male and 160 (36.2%) were female, with a mean age of 41.61 ± 25.56 years. Based on cranial CT imaging, intracranial hemorrhage (ICH) was detected in 27 patients (6.1%). There were no statistically significant differences in ICH frequency according to age or gender (p = 0.653 and p = 0.282, respectively).

Among mechanisms of injury, motor vehicle accidents (MVAs) showed the highest rate of ICH (9.5%), followed by falls from height (9.2%). Only MVAs were significantly associated with the presence of ICH (p = 0.019).

Regarding clinical features, loss of consciousness, memory loss, vomiting, high-energy trauma, basilar skull fractures, open/depressed skull fractures, and nonfrontal hematoma were significantly associated with ICH (p < 0.05). Headache, intoxication, and anticoagulant use were not significantly associated.

All demographic, mechanism, and clinical characteristics according to ICH status are summarized in Table 1.

Table 1.

Study population, mechanism, and clinical characteristics

Clinical Characteristics	ICH Present (n = 27)	ICH Absent (n = 415)	Total (n = 442)	p
Age
Median(Min-Max)	53(1–93)	40(0–95)	40(0–95)	0.653ψ
Mean ± Std.	44.15 ± 30.21	41.45 ± 25.26	44.61 ± 25.56
Gender
Male	21 (77.8%)	261 (62.9%)	282 (63.8%)	0.282*
Female	6 (22.2%)	154 (37.1%)	160 (36.2%)
Mechanism of Injury
Fall from height	7 (25.9%)	69 (16.6%)	76 (17.2%)	0.288 *
Motor vehicle accident	16 (59.3%)	152 (36.6%)	168 (38.0%)	0.019 Φ
Assault	0 (0.0%)	24 (5.8%)	24 (5.4%)	0.385 *
Sports injury	0 (0.0%)	7 (1.7%)	7 (1.6%)	1.000 *
Other	4 (14.8%)	163 (39.3%)	167 (37.8%)	0.011 Φ
Clinical Features
Loss of consciousness	8 (29.6%)	13 (3.1%)	21 (4.8%)	< 0.001 Φ
Memory loss	4 (14.8%)	4 (1.0%)	8 (1.8%)	0.001 Φ
Vomiting	4 (14.8%)	10 (2.4%)	14 (3.2%)	0.007 Φ
High-energy trauma	19 (70.4%)	113 (27.2%)	132 (29.9%)	< 0.001 *
Open/depressed skull fracture	6 (22.2%)	20 (4.8%)	26 (5.9%)	0.003 Φ
Basilar skull fracture	4 (14.8%)	4 (1.0%)	8 (1.8%)	0.001 Φ
Anticoagulant use	1 (3.7%)	45 (10.8%)	46 (10.4%)	0.340 Φ
Headache	–	18 (4.3%)	18 (4.1%)	0.616 Φ
Nonfrontal hematoma	2 (7.4%)	4 (1.0%)	6 (1.4%)	0.046 Φ

*: Fisher’s Exact Test, Φ: Pearson Chi Square, ^ψ: Mann–Whitney U Test. Bold p values indicate statistical significance (< 0.05)

Among the 442 patients included, the majority (98.4%) had mild TBI (GCS 13–15), whereas only 0.7% and 0.9% presented with severe or moderate TBI, respectively. This distribution reflects the predominance of mild cases typically observed in emergency department head trauma cohorts. (Table 2)

Table 2.

Association between Glasgow coma scale (GCS) categories and intracranial hemorrhage (ICH)

GCS Category	ICH Absent (n, %)	ICH Present (n, %)	Total (n, %)
Severe (≤ 8)	1 (33.3%)	2 (66.7%)	3 (0.7%)
Moderate (9–12)	1 (25.0%)	3 (75.0%)	4 (0.9%)
Mild (13–15)	413 (94.9%)	22 (5.1%)	435 (98.4%)
Total	415 (93.9%)	27 (6.1%)	442 (100%)

Diagnostic performance of GFAP, UCH-L1, and combined use

Receiver Operating Characteristic (ROC) analysis identified optimal cut-off values for UCH-L1 and GFAP. Cut-off thresholds were determined using the Youden Index (J = Sensitivity + Specificity − 1), which identifies the point that maximizes both sensitivity and specificity. The optimal cut-off for GFAP was 136.2 pg/mL (95% CI: 118.4–153.9), and for UCH-L1 was 311.5 pg/mL (95% CI: 272.1–350.9). Both biomarkers demonstrated high diagnostic performance, with AUC values of 0.95 (95% CI: 0.89–1.00; p < 0.001) for UCH-L1 and 0.98 (95% CI: 0.95–1.00; p < 0.001) for GFAP. When both biomarkers were concurrently elevated, diagnostic accuracy remained high, with sensitivity of 92.6%, specificity of 98.8%, PPV of 83.3%, NPV of 99.5%, and LR + of 76.85 (Table 3; Fig. 2). Confidence intervals (95% CI) were calculated for sensitivity, specificity, PPV (GFAP: 79–87%, UCH-L1: 83–89%), NPV (GFAP and UCH-L1: 98–100%), LR+ (GFAP: 72–82, UCH-L1: 89–103), and LR− (GFAP and UCH-L1: 0.04–0.09).

Table 3.

Diagnostic performance of UCH-L1, GFAP, and combined positivity

	UCH-L1	GFAP	UCH-L1 + GFAP (Both Positive)ϕ
AUC *	0.95 (0.89-1.00)	0.98 (0.95-1.00)		-
Sensitivity (%) *	92.6% (0.77–0.98)	92.6% (0.77–0.98)		92.6% (0.77–0.98)
Specificity (%) *	99.0% (0.98-1.00)	98.8% (0.97-1.00)		98.8% (0.97–1.00)
PPV (%) *	86.2% (0.83–0.89)	83.3% (0.79–0.87)		83.3% (0.79–0.87)
NPV (%) *	99.5% (0.98-1.00)	99.5% (0.98-1.00)		99.5% (0.98–1.00)
Accuracy (%)	98.6%	98.4%		98.4%
LR (+)	96.060	76.850		76.850
LR (-)	0.070	0.070		0.075

AUC: Area Under the Curve; PPV: Positive Predictive Value; NPV: Negative Predictive Value; LR+: Positive Likelihood Ratio; LR-: Negative Likelihood Ratio

*: Diagnostics given with their (95% CI)

ϕ: Combined positivity indicates cases where both UCH-L1 and GFAP were concurrently positive. Sensitivity, specificity, predictive values, and likelihood ratios were calculated from 2 × 2 contingency tables

Fig. 2.

Roc curve for both biomarkers

Logistic regression analysis

Univariate logistic regression showed that elevated plasma concentrations of both biomarkers were strongly associated with the presence of ICH. UCH-L1 was associated with an odds ratio (OR) of 1284.4 (95% CI: 224.4–7352.3; p < 0.001), and GFAP with an OR of 1025.0 (95% CI: 189.4–5548.3; p < 0.001) (Table 4).

Table 4.

Univariate logistic regression analysis of biomarkers for ICH

Protein	Wald	df	Odds Ratio (Exp(β))	95%CI	p-value
UCH-L1	64.66	1	1284.4	224.4–7352.3	< 0.001
GFAP	56.45	1	1025.0	189.4–5548.3	< 0.001

CI: Confidence Interval, df: Degrees of Freedom

Discussion

Traumatic brain injuries (TBIs) remain a major public health issue in both developed and developing countries. Early diagnosis of intracranial hemorrhage (ICH) following TBI is critical for both prognosis and therapeutic success. This study evaluated the diagnostic performance of the glial fibrillary acidic protein (GFAP) and ubiquitin carboxy-terminal hydrolase L1 (UCH-L1) biomarkers in detecting ICH. Our findings show substantial concordance with the international literature.

Among the 442 patients included in our study, 282 (63.85%) were male and 160 (36.15%) were female. No statistically significant association was found between gender and the presence of ICH (p = 0.282). Similarly, age did not show a significant impact on the likelihood of ICH in our population, which had a mean age of 41.33 years (p = 0.653). These results align with the findings of Papa et al. [8] and Welch et al. [5], who reported that GFAP and UCH-L1 levels increase independently of demographic variables such as age and gender.

In terms of injury mechanisms, motor vehicle accidents (MVAs) were significantly associated with ICH (p = 0.019). This observation is consistent with multicenter studies conducted by Bazarian et al. [10], Czeiter et al. [14], and Korley et al. [4]. Falls from height, assaults, and other causes did not reach statistical significance. These findings suggest that in triage processes, biomarker use may serve as a more effective decision-making tool particularly in patients with MVA-related trauma.

Analysis of clinical features revealed that loss of consciousness, memory loss, vomiting, and basal skull fractures were significantly associated with ICH. These results correspond with symptom-based risk classifications reported by Zetterberg et al. [6] and Welch et al. [5]. Notably, incorporating early biomarker levels alongside clinical features such as vomiting and loss of consciousness may enhance clinical decision-making.

In our cohort, most patients presented with mild TBI (GCS 13–15), consistent with the epidemiological pattern reported in emergency department populations, where mild injuries constitute the majority of cases [24, 25]. Only a small proportion had moderate or severe TBI, which is expected in a non-trauma-center population. This predominance of mild cases likely contributed to the relatively low incidence of intracranial hemorrhage (6.1%) and should be considered when interpreting biomarker performance.

Regarding biomarker analysis, the AUC for GFAP was 0.98 with a sensitivity of 92.6% and specificity of 98.8%. UCH-L1 yielded an AUC of 0.95, with the same sensitivity (92.6%) and a specificity of 99.0%. In cases of combined positivity (GFAP + UCH-L1), sensitivity was 92.6%, specificity 98.8%, positive predictive value 83.3%, and negative predictive value 99.5%.

The higher diagnostic performance observed in our study compared with most prior reports may be attributed to several methodological and population-related factors. Blood sampling was performed within a relatively narrow post-injury time window (mean 4.2 ± 1.8 h), coinciding with the known peak release of GFAP and UCH-L1 after traumatic brain injury. Additionally, we used research-use-only ELISA kits rather than FDA-approved point-of-care platforms; inter-assay calibration differences may partly account for higher absolute values and specificity. Furthermore, our single-center cohort was relatively homogeneous in mechanism and demographics, likely reducing biological variability. Lastly, the limited number of ICH-positive cases may have inflated performance estimates. Taken together, these factors may explain the higher AUC and specificity observed and should be considered when comparing our results with multicenter validation studies.

These metrics are in high agreement with those reported by Papa et al. [1, 8, 11, 13], Czeiter et al. [14], Whitehouse et al. [15], and Biberthaler et al. [9]. Notably, in a study by Papa et al. [1] using an FDA-approved biomarker panel, GFAP alone was shown to reliably detect CT-detectable lesions, while UCH-L1 contributed additional specificity by reflecting neuronal injury.

In our univariate logistic regression analysis, elevated levels of both UCH-L1 and GFAP were significantly associated with ICH, with very high odds ratios (OR = 1284.38 and OR = 1025.00, respectively). These results are consistent with studies by Diaz-Arrastia et al. [12], who linked UCH-L1 and GFAP levels with CT findings; by Korley et al. [4], who evaluated functional outcomes in relation to biomarker levels; and by Papa et al. [1], whose 2024 study reported that GFAP levels above 30 pg/mL yielded sensitivity over 98% for lesion detection on CT, while levels above 2300 pg/mL were associated with an ICH rate of 97.1%.

Univariate logistic regression confirmed strong associations between biomarker levels and ICH; however, this analysis provided limited additional information beyond ROC performance. Multivariable logistic regression was not performed due to the relatively small number of ICH cases (n = 27), and thus odds ratios should be interpreted with caution.

Although our findings corroborate prior research, including large multicenter studies [1, 17] and earlier pivotal single-center investigations [14], the single-center nature of the present study limits its generalizability. Future research should include multicenter trials to evaluate biomarker performance across different populations, age groups, and trauma mechanisms. Moreover, while prior work has linked biomarker levels with both radiological findings and functional outcomes [4], potential platform-dependent variability remains a concern, highlighting the need for assay harmonization efforts across laboratories [17].

Additionally, as this study was conducted in a single urban tertiary care center with relatively homogeneous demographic and geographic characteristics, its findings may not be directly applicable to other settings, such as rural or under-resourced healthcare environments. Differences in injury mechanisms, healthcare infrastructure, and patient characteristics could influence biomarker performance. To address these limitations, future research should involve multicenter, geographically diverse cohorts to validate diagnostic thresholds and enhance the external validity of biomarker use across different healthcare contexts. While our findings strongly support the diagnostic accuracy of GFAP and UCH-L1, the generalizability of these results is potentially limited by the demographic and geographic homogeneity of our study population. Conducted in a single urban tertiary care center, the study may not capture the full variability of biomarker expression that could be influenced by differences in race, socioeconomic status, pre-existing health conditions, or environmental exposures prevalent in rural or underserved populations. Additionally, access to emergency care and imaging infrastructure may vary significantly between regions. Future multicenter studies involving diverse populations across multiple healthcare settings—both urban and rural—are essential to validate and refine biomarker thresholds and to enhance external validity.

Furthermore, for these biomarkers to be integrated effectively into clinical workflows, especially in fast-paced emergency departments, point-of-care (POC) testing platforms must be developed and validated. The current reliance on ELISA, while precise, is time-consuming and requires laboratory infrastructure. Portable, automated platforms like i-STAT or Alinity, although promising, involve substantial institutional investment and staff training. Cost-effectiveness analyses should be prioritized in future research to compare biomarker-based triage versus conventional imaging strategies, especially in terms of reduced CT utilization, improved patient throughput, and long-term healthcare savings. National or institutional funding initiatives could facilitate early adoption by supporting device acquisition, training programs, and pilot implementation studies. Ultimately, real-time biomarker integration will depend not only on diagnostic performance but also on pragmatic factors such as speed, cost, and scalability across diverse clinical environments.

The concordance between our findings and those in the literature supports the strong diagnostic utility of GFAP and UCH-L1 for early identification of intracranial injury following head trauma. Particularly, the concurrent positivity of both biomarkers appears to offer robust reliability for clinical decision-making algorithms. The high sensitivity and predictive capabilities of these biomarkers suggest that they may play a pivotal role in guiding patient management, especially in emergency settings where rapid risk stratification is essential.

Although no delayed or missed ICH cases were identified on retrospective chart review, follow-up imaging was not routinely performed for CT-negative patients, as biomarker analyses were conducted post-sampling. Therefore, future studies should prospectively assess delayed hemorrhage and biomarker kinetics through serial imaging to confirm these findings.

Limitations

This study has several limitations. First, its single-center design and relatively homogeneous population may limit the generalizability of the findings. Second, the cut-off values for GFAP and UCH-L1 were derived post hoc using Youden’s index, which may introduce bias and overestimate diagnostic performance. Third, we employed research-use-only ELISA kits rather than FDA-approved point-of-care platforms. This, together with the homogeneous single-center sample and narrow sampling window, may have contributed to the relatively high specificity observed (as discussed above). Fourth, since only patients who underwent cranial CT were included, there is potential for selection bias toward more clinically severe cases. While this design choice was necessary because CT served as the reference standard for ICH, it may have inflated specificity and PPV compared to an unselected head trauma population. Fifth, while Glasgow Coma Scale scores were recorded, outcomes such as hospital admission, neurosurgical intervention, and follow-up imaging were not systematically collected, limiting the ability to correlate biomarker levels with longitudinal prognosis. Sixth, although patients with concomitant extracranial injuries (multi-trauma) were not excluded, the severity and anatomical distribution of these additional injuries were not systematically recorded. This may have introduced unmeasured heterogeneity, as extracranial trauma can potentially influence biomarker levels through systemic inflammatory responses. Finally, as this study was observational, causality cannot be inferred.

Future multicenter prospective studies with diverse populations, standardized assays, and long-term follow-up are warranted to validate and refine these findings.

Conclusion

Our study suggests that the blood-based biomarkers GFAP and UCH-L1 may have diagnostic potential for the early detection of intracranial hemorrhage following blunt head trauma. Both biomarkers demonstrated high apparent sensitivity and specificity within our single-center cohort, with GFAP particularly associated with CT-detectable structural brain damage, and UCH-L1 potentially enhancing diagnostic discrimination by reflecting neuronal injury.

The observed correlation between elevated biomarker levels and CT-confirmed ICH, as well as the high odds ratios identified in logistic regression analysis, indicates a promising association but does not yet establish clinical utility. In particular, the combined positivity of GFAP and UCH-L1 appeared to improve diagnostic reliability; however, these findings should be interpreted cautiously given that the assays used were research-use-only (RUO) ELISA kits, not standardized or FDA-approved clinical tests.

Although our study focused solely on diagnostic accuracy, future research should investigate whether GFAP and UCH-L1 levels can predict clinical outcomes or the need for neurosurgical intervention.

While our findings align with the direction of prior biomarker research, the magnitude of diagnostic performance observed was higher than in most multicenter studies, possibly due to methodological and population-related differences. Therefore, further validation through large-scale, multicenter investigations is essential to confirm optimal thresholds and assess reproducibility across different clinical environments. Despite these limitations, our study contributes additional data supporting the continued investigation of GFAP and UCH-L1 as potential adjuncts to clinical decision-making in traumatic brain injury.

Abbreviations

ICH

Intracranial hemorrhage

TBI

Traumatic brain injury

CT

Computed tomography

GFAP

Glial Fibrillary Acidic Protein

UCH-L1

Ubiquitin Carboxy-Terminal Hydrolase L1

ROC

Receiver Operating Characteristic

AUC

Area Under the Curve

PPV

Positive predictive value

NPV

Negative predictive value

CIs

Confidence intervals

CV

Coefficients of variation

MVAs

Motor vehicle accidents

POC

Point-of-care

Author contributions

Conceptualization: Yasin Bilgin, Fatih Mehmet Sari. Methodology: Yasin Bilgin, Fatih Mehmet Sari, Murat Gunay, Ugur Durmus. Formal Analysis and Statistics: Fatih Mehmet Sari, Serhat Hayme. Investigation and Data Collection: Yasin Bilgin, Fatih Mehmet Sari, Yusuf Kantar. Laboratory Analysis and Sample Processing: Murat Gunay, Ugur Durmus. Writing – Original Draft Preparation: Yasin Bilgin, Fatih Mehmet Sari. Writing – Review & Editing: All authors. Supervision: Yasin Bilgin.

Funding

None.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to [Data collected from institutions under control of general security forces] but are available from the corresponding author on reasonable request.

Declarations

Ethical approval and consent to participate

Access permission and ethical approval have been granted by the Erzincan Binali Yıldırım University Clinical Research Ethics Committee (Approval No: 2024/112). Written informed consent was obtained from all study participants or their legal representatives prior to inclusion in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.