Moderate-to-severe traumatic brain injury (TBI) is one of the leading causes of death and neurological dysfunction (1, 2). About 62.9–72% of these patients were diagnosed with diffuse axonal injury (DAI) by magnetic resonance imaging (MRI) (3, 4). DAI is caused by acceleration-deceleration or rotational forces on brain tissues, resulting in axonal shear injuries and delayed axonal disconnection (5, 6). Patients suffering from DAI are often accompanied by a 6-h loss of consciousness, and the diagnosis and severity grading of DAI within the first 6 h may reduce the time to investigate the cause of the coma and provide opportunities for optimized treatment strategies (6, 7). The gold standard for diagnosing and grading DAI is histopathology, but for surviving patients, MRI is considered the primary basis for diagnosing DAI, because it shows subtle axonal damage and hemorrhages that cannot be detected by computed tomography (CT) (8). However, for patients with severe TBI who require multiple medical devices and drug treatments, lengthy queuing and examination times for MRI, coupled with the limitations of monitoring equipment due to high magnetic fields, hinder the application of MRI in the acute phase of injury (9). Therefore, the identification of DAI in the early stage of TBI is an urgent problem to be solved at present.

The role of biomarkers in the diagnosis and prognostic assessment of DAI has received increasing attention (9–11). Neuron-specific enolase (NSE) is an isoenzyme of the glycolytic enzyme enolase which exists almost in neuronal cells, with the highest concentrations in the central nervous system. The molecular weight of 78 KD facilitates the leakage of NSE into the extracellular compartment and blood after the structural damage of nerve cells (12). NSE has been demonstrated to be increased in serum following various types of TBI, making it potentially a sensitive and specific biomarker of neuronal cell damage (12, 13). However, small cell lung cancers, red blood cells, platelets, adipose tissue, and smooth muscle contain NSE as well (14).

The Glasgow Coma Scale (GCS) is recognized as one of the most important neurological tools for assessing neurological function in patients with TBI (15). The admission GCS score has been confirmed to be strongly associated with the severity and prognosis of patients with TBI, but it is susceptible to sedation or endotracheal intubation, and patients suffering from DAI may present with lower GCS scores in the acute phase, leading to an inability to accurately distinguish between the grade of DAI and long-term prognosis (15, 16).

In summary, the early diagnosis of DAI cannot be accomplished with a single index so far, although part of it can be reflected by serum NSE or admission GCS score. We conjecture that better evaluation performance may come from the combination of NSE and GCS. This study aimed to test the hypothesis that the serum NSE level to admission GCS score ratio (NGR) is an effective indicator for identifying DAI in the acute phase of injury.

This retrospective study indicated that an increased value of NGR on admission could be employed as an independent predictor of DAI in patients with moderate-to-severe TBI. NGR demonstrated high diagnostic accuracy for DAI (AUC = 0.9493; sensitivity, 90.91%; and specificity, 85.71%). Patients with favorable neurological outcomes had significantly lower NGR as compared with those with unfavorable neurological outcomes.

Diffuse axonal injury is usually caused by external injury involving shearing force on brain tissues, and it manifests in the form of focal axonal shear injuries and axonal breakage (5, 6, 19). More than 50 million people suffer from TBI each year worldwide, among them, patients with moderate-to-severe TBI with DAI are more likely to have long-term sequelae or serious neurological deficits, even death (20–22). During initial resuscitation for TBI, effective and accurate prediction factors for DAI can contribute to support medical decisions for physicians and caregivers, from diagnosis to interventions (6, 9).

In recent years, many DAI biomarkers have been discovered by previous studies, such as tau protein, β-APP, NSE, S-100 calcium-binding protein B(S-100B), Ubiquitin C-terminal hydrolase-L1 (UCH-L1), spectrin, glial fibrillar acidic protein (GFAP), and neurofilament light (NfL) (5, 9, 18, 19, 23–26). GFAP and UCH-L1 have been shown to discriminate patients regarding the presence or absence of brain lesions on initial CT scan (26) and authorized for clinical use in the evaluation of mild TBI. Furthermore, Abbott has developed a rapid version on the i-STAT™ Alinity™ platform which received FDA approval in 2021 (27). We expect this device to be widely used in China and other countries to improve the diagnostic efficiency of DAI. Positive immunohistochemistry for β-APP is undetectable in normal brain tissue samples, whereas the immune response is enhanced starting 2 h after TBI. Following DAI, the disruption of axoplasmic transport leads to the accumulation of β-APP in axons, bringing its concentration to detectable levels (24). As the main cytoskeletal components of nerve cells, NfL and spectrin play important roles in maintaining axonal caliber and neuron morphology. Increased levels of these biomarkers result from axonal cell damage (28). Unfortunately, these biomarkers above are currently considered as the diagnostic indicators for DAI in laboratory settings and forensics and are not widely used in the management of patients with TBI (24). Tau protein, which accounts for a large proportion of microtubule-associated proteins, is released in great quantities into the cerebrospinal fluid after patients suffered DAI. Therefore, the content of tau protein in cerebrospinal fluid can be regarded as an indicator for quantitative evaluation of the severity of axonal injury. However, the tau protein detected in serum has not been found to be helpful for effectively evaluating prognosis. As a result, it is not enough to be used as an early diagnostic indicator of DAI in emergency treatment (24). S-100B and NSE have been investigated as potential biomarkers for predicting early neurological outcome or mortality. However, as independent predictors, the results are unsatisfactory. In fact, these molecules in tissues other than brain tissue severely limit their application, since serum levels of these biomarkers may also be increased in the serum of trauma patients without TBI (29, 30).

All hospitals above second-level in China have set up trauma treatment centers in EDs, all of which are able to treat patients with TBI. However, the detection capabilities of each hospital are inconsistent, and many new inspection and testing methods are only routinely implemented in third-level hospitals. Therefore, the diagnostic capability of DAI after injury is currently very limited. This study was dedicated in exploring the use of conventional testing methods combined with clinical characteristics to improve the diagnostic ability of DAI and avoiding the impact of missed diagnosis or misdiagnosis on the formulation of treatment strategies. By analyzing the demographic and clinical characteristics of eligible patients with TBI, no significant differences were found between patients with and without DAI with respect to sex, age, and cause of trauma. The non-DAI group had significantly lower serum NSE levels, Marshall CT score, and NGR compared with the DAI group (p < 0.0001). The admission GCS scores were significantly lower in patients with DAI (p < 0.0001). These conclusions are consistent with previous research findings (6, 19, 31). Considering the low sensitivity of CT examination for microbleeds and axonal damage caused by DAI in the acute phase (32, 33). Additionally, serum NSE is widely used by medical institutions at all trauma treatment centers in China because of its advantages of being easily accessible, low-cost (30–40 RMB), objective, and repeatable (12, 34, 35). We excluded Marshall CT score and combined serum NSE levels within 6 h of injury and admission GCS scores to form a simple, objective, and stable predictor.

In a study enrolling 176 patients, 106 patients were assigned to the non-DAI group and 70 were assigned to the DAI group. Logistic regression analysis identified that the GCS and Rotterdam CT scores were strongly related to the presence of DAI on subsequent MRI, with adjusted odd ratios (ORs) of 0.877 and 1.047, respectively (18). In our study, the correlation of admission GCS scores and DAI was also proved by the Pearson correlation analysis (r = −0.625, p < 0.0001), as did Marshall CT (r = 0.402, p < 0.0001). The strong correlation of the variable with DAI was shown by the large absolute value of Pearson r, with NSE (r = 0.596, p < 0.0001) being more significant. However, the largest Pearson r value was exhibited by NGR (r = 0.755, 95% CI 0.664–0.824, p < 0.0001), indicating that NGR was most correlated with the presence of DAI on clinical MRI.

Serum levels of NSE at admission were related to initial GCS scores and 6-month Glasgow Outcome Scale scores. Thus, serial measurements of serum NSE may help assess brain damage. Park et al. analyzed data concerning pediatric patients with TBI (36). According to Skandsen et al., DAI was found in nearly three-quarters in patients with moderate and severe TBI in the acute phase and GCS was correlated to the outcome only if patients with TBI had accompanying DAI (4). To further confirm the predictive power of significantly correlated factors for an early diagnosis of DAI, ROC curves were employed to assess the diagnostic accuracy of NGR, NSE, and GCS on admission for identifying DAI after injury. The final results showed that NGR had the highest diagnostic accuracy (AUC = 0.9493; sensitivity, 90.91%; and specificity, 85.71%) compared with GCS and NSE.

Based on a feature selection method using a tree-based ensemble algorithm, GCS, age, glucose, and fibrin/fibrinogen degradation products were identified as the effective prognostic factors for poor in-hospital outcome. Matsuo et al. analyzed data from 232 patients with TBI (37). Our study found that NGR was a more convenient and practical indicator for predicting 6-month GOSE because of its advantages of being easily accessible and objective.

There are several limitations to our present study. First, the sample size is not large enough to evaluate the predictive value of NGR for the occurrence of DAI after injury. Second, the pathological process after TBI is relatively complex, and there are many differential metabolites in the blood (38, 39), but only the serum NSE levels combined with GCS on admission were employed to predict the appearance of DAI and 6-month GOSE. Plasma neurofilament light (NfL), glial fibrillary acidic protein (GFAP), and ubiquitin C-terminal hydrolase-L1 (UCH-L1) will be considered for inclusion in predictive models in expanded studies after they are widely used in routine clinical practice in China. Third, NSE levels are time-dependent factor (12, 35), but in our study, the effect of dynamic changes of NSE levels on the 6-month prognosis of moderate to severe TBI was not further evaluated. Fourth, serum of trauma patients with TBI cannot fully reflect the pathological process of brain tissue injury in the early stage, which may bias the hyper-acute assessment.

Our study demonstrated that NGR on admission could serve as an independent predictor of DAI with moderate-to-severe TBI and the best cut-off value of NGR on admission was 4.25 with a sensitivity of 90.91% and a specificity of 85.71%. This indicates that NGR is an easily accessible, objective, and reliable indicator to predict DAI in the acute phase of injury.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by Ethics Review Committee of Haining People's Hospital. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

RJ and WC contributed to the initial idea for this study. WC, RJ, and GW completed the study design. ZZ, RC, and CY developed and revised the search strategy. WS and GW contributed to consults about clinical issues. WS and RC extracted the data. WC analyzed the extraction data. RJ, CY, GW, and WC contributed to the revision of the draft. All authors approved the final manuscript prior to submission.

This study was supported by the Tianjin Research Program of Application Foundation and Advanced Technology (grant 19YFZCSY00650).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.