Introduction

Severe traumatic brain injury (sTBI) is one of the most lethal forms of trauma and is characterized by high mortality and disability rate.¹ Mechanistically, brain tissues are attacked by external forces, leading to irreversible primary brain injury and subsequent secondary brain injury.² The latter has intricate pathophysiological mechanisms, including an excessive inflammatory response, mitochondrial dysfunction, and oxidative imbalance, subsequently leading to blood-brain barrier destruction and exacerbation of brain edema, finally causing neuronal death, neurological impairments, and even death of patients.^3,4 Acute lung injury (ALI) is a complication of high-frequency sTBI management that can easily lead to poor prognosis in patients.⁵ The Glasgow Coma Scale (GCS) and Rotterdam computed tomography (CT) classification are the two favorite approaches for severity estimation of sTBI.^6,7 Concerning assessment of neurological status after sTBI, the Glasgow Outcome Scale (GOS) is accepted as a usable avenue in clinical practice.⁸ Given that blood biomarkers are easy to obtain, they have been extensively studied in recent decades with regard to the prognostic significance of sTBI during recent decades.^9–11

The blood-brain barrier (BBB) is a paramount structure of the central nervous system, and compromised BBB structure or function is a pathological hallmark of various neurological diseases.¹² Tight junction (TJ)-associated proteins maintain integrity and function of blood-brain barrier (BBB).¹³ Occludin was the first identified transmembrane TJ protein and massively contributes to BBB integrity.¹⁴ Blood occludin may be a biochemical marker of BBB destruction and brain edema.^15–18 Moreover, higher serum occludin levels after acute ischemic stroke are independently associated with early neurological deterioration, hemorrhage transformation, and poor prognosis.^19,20 These findings suggest that serum occludin is a biomarker of brain injury. In addition, occludin is abundant in alveolar cells and essential for maintaining normal alveolar barrier integrity in ALI as a result of ventilation or acute respiratory distress syndrome,^21,22 meaning that occludin may be an indicator of ALI. Currently, several neuron- or glia- specific biomarkers, such as neuron specific enolase, S100-beta protein, glial fibrillary acidic protein, neurofilament light, tau and ubiquitin carboxyl hydrolase L1, have been fully studied regarding their prognostic values in sTBI.^23–25 However, occludin is different from the preceding biomarkers, because it functions as a TJ protein with apparent participation in both brain injury and lung injury,^19–22 therefore proffering a notable insight into clinical study in relation to ALI and neurological prognosis following sTBI. Here, serum occludin levels were measured in humans with sTBI in an effort to discern its relevance to trauma severity, ALI, and neurological outcome, while assessing the mediating role of ALI on its association with poor prognosis.

Materials and Methods

Study Design. Subject Enrollments and Ethical Statements

This observational analytic study was conducted between September 2021 and October 2023 at two comprehensive hospitals: Shengzhou People’s Hospital (Shengzhou, China) and Longquan People’s Hospital (Longquan, China). Intrinsically, two sub-studies, namely the cross-sectional analysis and prospective cohort assessment, composed this clinical investigation. First, controls and patients with sTBI were employed to unravel the evolutionary course of serum occludin levels subsequent to sTBI. In the latter part, patients were used to determine the role of serum occludin as a prognostic metric for sTBI. Patients were required to be at least 18 years old, stay in a state of pretraumatic GOS score of 1, be admitted to the hospital within post-trauma twelve hours on account of blunt, isolated head trauma, and experience post-resuscitation GCS scores from 3 to 8. Subsequently, we excluded patients with certain neurological illnesses, such as cerebrovascular accidents, brain neoplasms, and head trauma, as well as those with specific conditions or severe diseases, such as the use of steroid hormones, infections within the past month, pregnancies, malignancies, heart failure, and ascites. Controls comprised volunteers who were free from some chronic illnesses, such as hypertension, diabetes, and dyslipidemia, and had normal results in some routine tests, such as blood erythrocyte counts, blood hemoglobin levels, blood leukocyte counts, blood neutrophil counts, and blood potassium levels. This study was performed in accordance with the Declaration of Helsinki and follow-up updates. The study protocol was approved by the Institutional Review Committees of the Shengzhou People’s Hospital (No. 2021–039-01) and Longquan People’s Hospital (No. 2018–023). The lawful representatives of patients and controls were familiarized with the study details and then independently provided written informed consent forms.

Data Collection and Neurological Functional Assessment

In two centers, operational modalities and other therapeutic methods were adopted to treat patients based on the protocol specified in the guidelines for the management of sTBI.²⁶ All patients were consecutively enrolled and their admission time since trauma was documented. At the emergency center, we inquired about their age, sex, tobacco smoking, alcohol consumption, hypertension, and diabetes mellitus. Traumatic causes were dichotomized into traffic accidents and other types. Readings were obtained following non-invasive measurements of arterial blood pressure at the emergency center. So as to mirror the severity of the sTBI, post-resuscitation GCS scores at admission were recorded by trained assessors, who were blinded to the other clinical materials. Based on head CT imaging, positive radiological hemorrhagic manifestations were categorized into two types: mass and non-mass lesion bleeding. The former encompasses hematomas localized in the intracerebral, epidural, and subdural cavities, while the latter consists of bleeding located in the intraventricular and subarachnoid spaces. The non-bleeding appearance included brain contusions and pneumocephalus. Rotterdam CT scale scores were computed in accordance with radiological data. Operation within 24 hours and seizure onset subsequent to sTBI were registered. In-hospital ALI was diagnosed according to previous criteria as follows: acute emergences of clinical symptoms with the ratio of partial pressure of arterial oxygen to fractional inspired oxygen equal to or below 300, findings of bilateral infiltrates via chest radiograph and exclusion of left arterial hypertension.²⁷ Via the structured interviews, GOS scores at six-month mark post-sTBI were documented by the experienced clinician through a telephone visit, in order to reflect neurological function, and scores of 1–3 is indicative of a poor prognosis.²⁸

Immune Test

All standardized procedures were strictly performed in two centers according to pre-established specifications. Blood samples were collected via venous access from the patients and controls. A rapid action was done to put blood samples in 5 mL gel-prefilled biochemistry tubes, and after centrifugation 2000 × g for 10 minutes, the supernatant was extracted with transferal into Eppendorf tubes for storage at −80°C conditions awaiting measurements. Within three months of preservation, serum specimens must be thawed for quantification of occludin levels. Serum occludin levels were detected in duplicate by the same professional personnel using a sandwich enzyme-linked immunosorbent assay kit (item number, CSB-EL016263HU; CUSABIO, China). As for this kit, the detection range was from 23.5 to 1500 pg/mL, the minimum detectable limit was 5.8 pg/mL, and the inter- and intra-assay precision coefficients of variation were below 10%.

Statistical Analysis

The SPSS statistical package (version 20.0; SPSS Inc., Chicago, Illinois, USA) was used for statistical analyses. The data were divided into categorical and continuous variables. The former was expressed in the form of frequencies (proportions). For the latter, the Shapiro–Wilk test or Kolmogorov–Smirnov test was performed to identify normality. Normally distributed variables were reported as means (standard deviations, SDs) if non-normally distributed, and as medians (percentiles 25th-75th). To compare the data between two groups or among multiple groups, the independent t-test, Mann–Whitney U-test, Pearson’s Chi-square test, Fisher’s exact test, and Kruskal–Wallis test were employed as needed. Bivariate correlations were investigated using the Spearman’s test. Here, five dependent variables were selected, including continuous GOS scores, serum occludin levels, ALI, poor prognosis and ordinal GOS scores, and many kinds of multivariate statistical methods were employed, covering the multivariate linear regression analysis, binary logistic regression analysis and ordinal regression analysis. Significantly distinct factors in the univariate analyses were included in the multivariate models to identify the independent variables. Discrimination efficiencies were estimated from the receiver operating characteristic (ROC) curve using MedCalc statistical software version 17.4 (MedCalc Software, Mariakerke, Belgium). The R software (version 3.5.1; https://www.r-project.org), variance inflation factor (VIF) was calculated, and the nomogram, restricted cubic spline, decision curve, forest plot, and graphs of mediation and subgroup analyses were obtained. Scatter plots, bar plots, and ROC curves were plotted using the GraphPad Prism 7.01 (GraphPad Software, Inc., San Diego, California, USA). Two-sided statistical significance was defined as P < 0.05.

Results

Participant Features

After consecutive recruitment, 295 patients with sTBI met the inclusion criteria. Subsequently, we excluded 49 patients, including 18 with previous neurological diseases, 21 with some specific conditions or severe diseases, 3 rejected study entry, 2 lost to follow-up, 2 provided inadequate information, and 3 had unqualified blood samples. Ultimately, 246 patients were included in this analysis. The baseline patient characteristics are shown in Table 1. One hundred controls were aged from 20 to 71 years (mean, 44.8 years; SD, 13.3 years), the male to female was 1.33 (57/43), and included 34 tobacco smokers and 33 alcohol drinkers. From a statistical perspective, the above four variables were not significantly different between controls and patients (all P>0.05).

Table 1 Baseline Characteristics and Correlativity Analyses in Severe Traumatic Brain Injury

Serum Occludin Levels and sTBI-Related Severity

Serum occludin levels were markedly higher in the patients than in the controls (P<0.001; Figure 1). In subgroups with GCS scores spanning from 3 to 8, significant differences were existent in terms of serum occludin levels among six subgroups, with the highest levels in subgroup of GCS score 3, followed by those with the scores from 4 to 6 and the lowest levels in that of GCS score 8 (P<0.001; Figure 2); consistently, serum occludin levels were significantly in opposite proportion to GCS scores (P<0.001; Figure 3). Alternatively, Serum occludin levels substantially increased in the order of Rotterdam CT scores, varying from 3 to 6 (P<0.001; Figure 4), and serum occludin levels were markedly enhanced with increasing Rotterdam CT scores (P<0.001; Figure 5). To be continued, aside from GCS scores and Rotterdam CT scores, other factors of notable correlation with serum occludin levels were abnormal cisterns, midline shift above 5 mm, subarachnoid hemorrhage, brain contusion, ALI and blood glucose levels (all P<0.05; Table 1). Excluding ALI as a subsequent event and three components of the Rotterdam CT scoring system, namely abnormal cisterns, midline shift > 5 mm, and subarachnoid hemorrhage, the remaining four variables were forced into the multiple-factorial model. As a result, Serum occludin levels were independently correlated with GCS score [beta (β), −2.530; 95% confidence interval (CI], −3.241--1.819; VIF, 1.119; P=0.001) and Rotterdam CT score (β, 2.427; 95% CI, 1.232–3.622; VIF, 1.538; P=0.001).

Figure 1 Variation of serum occludin levels subsequent to severe traumatic brain injury. There was a substantial increase in serum occludin levels in patients with severe traumatic brain injury compared to controls (P<0.001).

Figure 2 Disparities of serum occludin levels at a basis of Glasgow coma scale scores following severe traumatic brain injury. Serum occludin levels of patients were notably distinct among the six subgroups subdivided based on Glasgow coma scale scores post-severe traumatic brain injury, manifested with the highest levels in the subgroup with a score of 3, the medium levels in those with scores from to 4–7 and the lowest levels in those with a score of 8 (P<0.001). Abbreviation: GCS, Glasgow coma scale.

Figure 3 Relationship between serum occludin levels and Glasgow coma scale scores of humans diseased of severe traumatic brain injury. Serum occludin levels were inversely proportional to the Glasgow Coma Scale scores among individuals with severe traumatic brain injury (P<0.001). Abbreviation: GCS, Glasgow coma scale.

Figure 4 Distinction of serum occludin levels according to Rotterdam computed tomography scores in patients experiencing severe traumatic brain injury. Serum occludin levels of cases showed significant differences among the four subgroups categorized in accordance with Rotterdam computed tomography scores after severe traumatic brain injury, marked by the highest levels in the subgroup with score 6, the medium levels in those with scores 4–5 and the lowest levels in those with score 3 (P<0.001). Abbreviation: CT, computed tomography.

Figure 5 Relation of serum occludin levels to Rotterdam computed tomography scores among individuals suffering from severe traumatic brain injury. Serum occludin levels were positively correlated with Rotterdam computed tomography scores after severe traumatic brain injury (P<0.001). Abbreviation: CT, computed tomography.

Serum Occludin Levels and 6-Month Neurological Functional Status Following sTBI

As outlined in Figure 6, six-month GOS scores were intimately correlated with the serum occludin levels (P<0.001; Figure 6). In addition, the GOS scores were closely related to age, diabetes mellitus, GCS scores, Rotterdam CT scores, abnormal cisterns, midline > 5 mm, subarachnoid hemorrhage, intracerebral hemorrhage, ALI, blood glucose levels, and blood leukocyte counts (all P<0.05; Table 1). Similar to the preceding interpretation, abnormal cisterns, midline > 5 mm, and subarachnoid hemorrhage were excluded from the multivariate model. Therefore, the model was inclusive of the rest, leading to the revelation that GOS scores were independently associated with GCS scores (β, 0.224; 95% CI; 0.120–0.327; VIF, 1.991; P=0.001), Rotterdam CT scores (β, −0.491; 95% CI, −0.659--0.322; VIF, 1.793; P=0.002), ALI (β, −0.351; 95% CI, −0.682--0.020; VIF, 1.267; P=0.038), and serum occludin levels (β, −0.024; 95% CI, −0.041--0.007; VIF, 1.843; P=0.006).

Figure 6 Correlation of serum-based occludin levels with Glasgow outcome scale scores in subjects diseased of severe traumatic brain injury. Serum-based occludin levels were inversely correlated with the Glasgow outcome scale scores of patients with severe traumatic brain injury (P<0.001). Abbreviation: GOS, Glasgow outcome scale.

Five subgroups based on GOS scores had an exhibition of pronouncedly distinct serum occludin levels, marked by the highest levels in the subgroup with a score of 1, a subsequent gradual decline in the levels with increasing scores from 2 to 4, and the lowest levels in the subgroup with a score of 5 (P<0.001; Figure 7). Among the five subgroups, there were ten variables of substantial disparity: age, GCS score, Rotterdam CT score, abnormal cisterns, midline > 5 mm, subarachnoid hemorrhage, serum occludin levels, ALI, blood glucose levels, and blood leukocyte counts (all P<0.05; Table 2). Considering that the Rotterdam CT classification is composed of abnormal cisterns, midline > 5 mm, and subarachnoid hemorrhage, these three factors were not included in the multivariate model. The other seven variables were entered into the ordinal multivariate model, and then it was confirmed that GOS scores were independently associated with GCS scores (β, 0.539; 95% CI; 0.307–0.770; VIF, 1.959; P=0.003), Rotterdam CT scores (β, −0.960; 95% CI, −1.325--0.596; VIF, 1.738; P=0.005), ALI (β, −0.763; 95% CI; −1.340--0.186; VIF, 1.352; P=0.010), and serum occludin levels (β, −0.047; 95% CI, −0.084--0.010; VIF, 1.813; P=0.013).

Table 2 Relevant Factors to Ordinal Glasgow Outcome Scale Scores After Severe Traumatic Brain Injury

Figure 7 Differences of serum-based occludin levels among subgroups with distinct Glasgow outcome scale scores post-severe traumatic brain injury. Serum-based occludin levels were substantially higher in patients with a Glasgow outcome scale score of 1, followed by those with scores of 2–4, and were markedly lower in those with a score of 5, following severe traumatic brain injury (P<0.001). Abbreviation: GOS, Glasgow outcome scale.

As depicted in Figure 8, serum occludin levels were significantly higher in patients with a poor prognosis than in those with a good prognosis (P<0.001). Meanwhile, as shown in Figure 9, serum occludin levels displayed strong discrimination efficiency for likelihood of poor prognosis and the optimal value was selected as 6.6 ng/mL, which anticipated poor prognosis with 86.9% sensitivity and 58.9% specificity (the maximum Youden index at 0.458). Figure 10 shows the presence of a linear correlation between serum occludin levels and the possibility of a poor prognosis (P for nonlinear >0.05). The results of the univariate analyses are outlined in Table 3, showing that several factors, such as age, diabetes mellitus, GCS scores, Rotterdam CT scores, abnormal cisterns, midline > 5 mm, subarachnoid hemorrhage, subdural hematoma, ALI, serum occludin levels, blood glucose levels, and blood leukocyte counts, were substantially different between patients with and without poor prognosis (all P<0.05). Thereafter, other significantly different parameters, but not the three components of the Rotterdam CT assessment system, such as abnormal cisterns, midline > 5 mm, and subarachnoid hemorrhage, were integrated into the binary multivariate model. A further finding was that serum occludin levels [odds ratio (OR), 1.051; 95% CI, 1.006–1.098; VIF, 1.611; P=0.026], GCS scores (OR, 0.621; 95% CI, 0.467–0.827; VIF, 1.431; P=0.001), ALI (OR, 2.926; 95% CI, 1.564–5.474; VIF, 2.128; P=0.015) and Rotterdam CT scores (OR, 2.665; 95% CI, 1.639–4.334; VIF, 2.116; P=0.002) remained as the four independent predictors of poor prognosis after sTBI. Moreover, several conventional variables such as age, sex, hypertension, diabetes, alcohol consumption, and cigarette smoking negligibly moderated the relationship between serum occludin levels and the possibility of poor prognosis in patients with sTBI (all P for interaction >0.05; Figure 11). The preceding four factors were consolidated to create a combined model to forecast poor prognosis. The model was described in the form of a nomogram (Figure 12), had a good fit (Figure 13), possessed satisfactory clinical benefits (Figure 14), and had a substantially higher prognostic capability than its four components under the ROC curve (all P<0.05; Figure 15).

Table 3 Pertinent Variables to Six-month Functional Outcome and Acute Lung Injury After Severe Traumatic Brain Injury

Figure 8 Serum occludin levels between patients with the development of poor prognosis and those without such a condition after severe traumatic brain injury. Occludin levels in the serum of patients with severe traumatic brain injury were significantly higher in patients with poor prognosis than in those without such adverse events (P<0.001).

Figure 9 Receiver operating characteristic curve as to serum occludin levels for predicting poor prognosis subsequent to severe traumatic brain injury. Serum occludin levels were at a state of effective discrimination ability for poor prognosis, and the optimal value of the levels was designated as 6.6 ng/mL, yielding the maximum Youden index for predicting prognosis in severe traumatic brain injury. Arrow indicates threshold value of serum occludin levels. Abbreviations: AUC, area under curve; 95% CI, 95% confidence interval.

Figure 10 Restricted cubic spline for assessment of linear relationship between serum occludin levels and probability of poor prognosis following severe traumatic brain injury. Serum occludin levels exhibited a linear relationship with the risk of poor prognosis after severe brain injury (P for nonlinear > 0.05).

Figure 11 Forest plot assessing subgroup analysis with respect to serum occludin levels’ association with poor prognosis following severe traumatic brain injury. The relationship between serum occludin levels and poor prognosis after severe brain injury was not moderated by the conventional variables in the graph (all P interaction > 0.05). Abbreviations: OR, odds ratio; 95% CI, 95% confidence interval.

Figure 12 Nomogram evaluating combined model of prognosis prediction in severe traumatic brain injury. The model was composed of four independent predictors of poor prognosis after severe traumatic brain injury. Each variable corresponded to a separate score, and the aggregative scores of the four factors pointed to a risk of poor prognosis. Abbreviations: ALI, acute lung injury; GCS, Glasgow Coma Scale; CT, Rotterdam computed tomography.

Figure 13 Calibration curve showing model fit of prognosis anticipation following severe traumatic brain injury. The model was constructed based on four independent prognostic factors after severe brain injury. The model showed good fit.

Figure 14 Decision curve displaying clinical validities of prognosis prediction of prognosis model and its components after severe traumatic brain injury. The model was configured based on four predictive factors of poor prognosis following severe traumatic brain injury. This model is clinically valid for severe traumatic brain injury. Abbreviations: ALI, acute lung injury; GCS, Glasgow Coma Scale; CT, Rotterdam computed tomography.

Figure 15 Receiver operating characteristic curve demonstrating prognostic efficiency of merged model and its members following severe traumatic brain injury. The model was established by integrating the four predictors of poor prognosis after severe traumatic brain injury. The model possessed substantially higher predictive ability than the other four independent variables (**P<0.01). Abbreviations: GCS, Glasgow Coma Scale; CT, Rotterdam computed tomography classification; AUC, area under curve; 95% CI, 95% confidence interval.

Serum Occludin Levels in Relation to ALI and Mediation Role of ALI

As shown in Figure 16, serum occludin levels were markedly enhanced in ALI patients (P<0.001). Simultaneously, as displayed in Figure 17, serum occludin levels exhibited effective predictive value for ALI probability, and a value more than 7.7 ng/mL distinguished ALI risk with 87.3% sensitivity and 51.4% specificity (the maximum Youden index of 0.387). Serum occludin levels linearly correlated with the likelihood of ALI (P for nonlinear >0.05; Figure 18). As shown in Table 3, age, GCS score, Rotterdam CT score, abnormal cisterns, midline > 5 mm, subarachnoid hemorrhage, subdural hematoma, serum occludin levels, blood glucose levels, and blood leukocyte counts were significantly different between the ALI groups (all P<0.05). Abnormal cisterns, midline > 5 mm, and subarachnoid hemorrhage were not included in the multivariate model because of their contribution to the Rotterdam CT classification. The other factors were added to the binary multivariate model, and subsequently, it was proved that serum occludin levels (OR], 1.058; 95% CI, 1.015–1.102; VIF, 1.778; P=0.008), GCS scores (OR, 0.673; 95% CI, 0.495–0.916; VIF, 1.911; P=0.004), and Rotterdam CT scores (OR, 2.048; 95% CI, 1.334–3.142; VIF, 1.639; P=0.001) kept independently associated with post-sTBI ALI. Age, sex, hypertension, diabetes, alcohol consumption, and cigarette smoking did not significantly interact with serum occludin levels in association with ALI (all P for interaction >0.05; Figure 19). The independent predictive variables for ALI were merged to form a combined model for ALI prediction. A nomogram was used to visualize the model (Figure 20). The model showed good stability in ALI prediction (Figure 21), exhibited acceptable clinical validity (Figure 22), and outperformed any of its three members in ALI prognostication under the ROC curve (all P<0.05; Figure 23). ALI in part mediated association between serum occludin levels and poor prognosis, and its indirect effects constituted 26.0% of the total (Figure 24) with effectiveness of comparative stability (Figure 25).

Figure 16 Serum occludin levels between patients experiencing acute lung injury and those not suffering from this condition following severe traumatic brain injury. Occludin levels in the serum of patients diagnosed with severe traumatic brain injury were significantly higher in those with acute lung injury than in those without acute lung injury (P<0.001). Abbreviation: ALI, acute lung injury.

Figure 17 Receiver operating characteristic curve assessing predictive ability of serum occludin levels for acute lung injury secondary to severe traumatic brain injury. Serum occludin levels had a strong discrimination capability for the development of acute lung injury following severe traumatic brain injury, and its optimal value was chosen to be 7.7 ng/mL for forecasting acute lung injury in severe traumatic brain injury with the maximum Youden index. Arrow indicates threshold value of serum occludin levels. Abbreviations: AUC, area under curve; 95% CI, 95% confidence interval.

Figure 18 Restricted cubic spline verifying linear correlation of serum occludin levels with likelihood of acute lung injury post-severe traumatic brain injury. Serum occludin levels were linearly correlated with the risk of acute lung injury after severe traumatic brain injury (P for nonlinear > 0.05).

Figure 19 Forest plot describing subgroup assessment of association of serum occludin levels with acute lung injury secondary to severe traumatic brain injury. Serum occludin levels had no significant interactions with the common variables in prognosis prediction after severe brain injury (all P > 0.05). Abbreviations: OR indicates odds ratio; 95% CI, 95% confidence interval.

Figure 20 Nomogram outlining combined model for predicting acute lung injury subsequent to severe traumatic brain injury. The model was composed of three independent predictors of acute lung injury following severe traumatic brain injury. Each variable corresponded to its respective score, and the cumulative scores of these factors were related to the possibility of acute lung injury. Abbreviations: GCS, Glasgow coma scale; CT, Rotterdam computed tomography classification.

Figure 21 Calibration curve portraying model fit of acute lung injury secondary to severe traumatic brain injury. The model was created using three independent factors of acute lung injury severe brain injury. The model was stably run in the context of severe traumatic brain injury.

Figure 22 Decision curve depicting clinical benefits of combined model and its components in predicting acute lung injury after severe traumatic brain injury. The model comprises three predictive factors for acute following a severe traumatic brain injury. This model is clinically beneficial for the treatment of severe traumatic brain injury. Abbreviations: ALI, acute lung injury; GCS, Glasgow Coma Scale; CT, Rotterdam computed tomography.

Figure 23 Receiver operating characteristic curve delineating effectiveness of merged model and its members in prognosticating acute lung injury following severe traumatic brain injury. The model was developed by incorporating three predictors of acute lung injury following severe traumatic brain injury. The model had a markedly higher predictive capability than the other three independent predictors (*P<005; **P<0.01). Abbreviations: GCS, Glasgow Coma Scale; CT, Rotterdam computed tomography classification; AUC, area under curve; 95% CI, 95% confidence interval.

Figure 24 Mediation role of acute lung injury on association of serum occludin levels with poor prognosis after severe traumatic brain injury. Prognosis association of serum occludin levels was partially mediated via acute lung, and its mediation effect accounted for 26.0% of all.

Figure 25 Sensitivity analysis of mediation effect of acute lung injury on prognosis association of serum occludin levels following severe traumatic brain injury. Under both control and treatment conditions, there was statistically satisfactory robustness in the mediation effect of acute lung injury on the prognostic association of serum occludin levels in severe traumatic brain injury, because sensitivity analysis showed that both ρ values amounted to 0.3.

Discussion

To the best of our knowledge, this is the first study to investigate the prognostic implications of serum occludin level in patients with sTBI. First, serum occludin levels were significantly increased after sTBI. Second, serum occludin levels were independently associated with severity indicators such as GCS scores and Rotterdam CT classification. Third, serum occludin levels were independently associated with poor prognosis and ALI. Fourth, serum occludin levels were combined with other independent predictors of poor prognosis and ALI to form two models of satisfactory clinical efficacy. Finally, ALI may mediate the association between serum occludin levels and poor prognosis in this cohort of patients with sTBI. Overall, the above evidence supports the hypothesis that serum occludin may act as a prospective biochemical marker for the assessment of severity and anticipation of prognosis during sTBI management.

Acute brain injury, including sTBI, is characterized by BBB disruption.²⁹ BBB destruction is attributable to complex processes involving inflammation, oxidation, and amino acid toxicity reactions, and its damage further results in pathophysiological injury to the brain, encompassing cerebral edema, neuronal death, and even neurological impairments.³⁰ Therefore, maintaining BBB integrity is critical. Occludin, a member of the TJ protein family, is an essential protein that structurally composing the BBB.¹⁴ Under certain pathological circumstances, occludin dissolution or anomalous occludin enhances BBB permeability.^29–32 To date, exploring occludin-targeting medications has become a method for improving BBB permeability, thereby restoring neuronal function and the subsequent recovery of neurological function in acute brain injury diseases.^33–35

In rats with cerebral ischemia from middle cerebral artery occlusion, blood occludin levels were elevated in a time-dependent fashion, with a prompt enhancement at 4.5 hours following ischemia, which concorded with the loss of occludin from ischemic cerebral microvessels and obvious BBB leakage at 4.5 hours following ischemia.¹⁶ In another similar experiment in rats subjected to middle cerebral artery occlusion, the temporal trends of blood occludin levels were consistent with their loss from ischemic cerebral microvessels.¹⁷ BBB breakdown is strongly linked to the development.³⁶ In a recent study, serum occludin was verified as an identifiable biomarker of perihematomal edema after human intracerebral hemorrhage in the framework of a multivariate analysis.³⁷ These data suggest that blood occludin is a potential metric for early BBB damage.

Serum occludin levels were significantly higher in patients diagnosed of intracerebral hemorrhage than in controls,³⁷ and patients with acute ischemic stroke versus controls with pseudo-strokes had profoundly increased serum occludin levels.²⁰ Additionally, blood occludin levels were notably higher in mice with TBI than in those under normal conditions, as well as in 20 patients with TBI than in 20 healthy controls.²⁹ Consistently, serum occludin levels were substantially elevated after sTBI in this cohort of 246 subjects compared with 100 controls. In response to acute brain injury, occludin from cerebral microvessels can be released into the blood circulation;^16,17 therefore, serum occludin levels may increase, shedding light on the possibility of serum occludin as a prognostic biomarker of acute brain injury.

Regarding the relationship between serum occludin levels and sickness severity in acute brain injury diseases, a study of acute ischemic stroke using univariate analysis showed that serum occludin levels were highly correlated with cerebral infarct volume and National Institutes of Health Stroke Scale (NIHSS) scores.¹⁹ Nonetheless, another study reported that plasma occludin levels were not related to NIHSS scores following acute ischemic stroke.³⁸ Considering that the former cohort had median NIHSS scores of 14,¹⁹ while the latter cohort had scores of 3,³⁸ a large severity disparity may be an interpretation of the result inconsistency. In the current study of 246 patients with sTBI, serum occludin levels were independently associated with the GCS and Rotterdam CT scores. Thus, this study may provide a new clue to solidify the concept that serum occludin levels mirror the magnitude of sTBI trauma.

Among two cohorts of patients with acute ischemic stroke, serum occludin levels were reported to be independently associated with several outcome variables such as early neurological deterioration, hemorrhagic transformation, and poor ninety-day prognosis.^19,20 In our study of sTBI humans, post-trauma six-month neurological prognosis indicated by continuous GOS, binary GOS, and ordinal GOS scores, together with ALI, were designated as the outcome variables of interest. Intriguingly, serum occludin levels were independently associated with all the outcome variables. Furthermore, restricted cubic splines were plotted, and subgroup analyses were performed to enhance the robustness of the results that serum occludin may have the opportunity to be a prognostic indicator of poor prognosis and ALI subsequent to sTBI.

Serum occludin levels, along with NIHSS scores, were the two independent predictors of hemorrhage transformation following reperfusion in patients with acute ischemic stroke, and the two indicators were combined to create a model with a markedly higher area under the ROC curve over both.³⁹ In another study of acute ischemic stroke, serum occludin levels combined with other clinical risk factors, such as the Alberta Stroke Program Early CT scores and endovascular treatment, which were all independent predictors of hemorrhagic transformation, substantially enhanced the accuracy of the predictive effect.⁴⁰ In our study, the four independent predictors of poor prognosis, namely, serum occludin levels, GCS scores, Rotterdam CT scores and ALI, were merged to build a prognosis model; analogously, the three independent predictive factors of ALI, ie, serum occludin levels, GCS scores and Rotterdam CT scores, to establish an ALI model. The two models were validated with a good fit, satisfactory clinical validity, and high predictive efficiency using several statistical tools. Taken together, serum occludin may be an auxiliary prognosticator of sTBI with an increasing ability to predict ALI and neurological function.

Occludin strongly preserves alveolar barrier integrity, with extensive distribution in alveolar cells,^21,22 reinforcing serum occludin as a possible indicator of ALI. ALI has been shown to be a determinant of poor prognosis following sTBI, and concurrently, serum occludin levels have been found to be an independent predictor of ALI. ALI may be a link between serum occludin levels and poor prognosis following sTBI. ALI partially mediated the association between serum occludin levels and poor prognosis in patients with sTBI in the present statistics. In other words, ALI may elucidate the link between serum occludin levels and a poor prognosis after sTBI. However, the intricate interplay between these factors warrants further investigation.

Several limitations or prospects must be mentioned here. First, although the potential clinical utility of serum occludin as a prognostic biomarker of sTBI may be recognized, clinical implementation challenges, such as assay standardization and cost- effectiveness, must be overcome in future. Second, partial mediation of ALI on prognostic association of serum occludin levels were found in this host of patients with sTBI, while the limitations of mediation analysis should be acknowledged given possible existence of inadequately large sample size in observational data could affect robustness of results by mediation analysis. Third, serum occludin levels were gauged merely at admission and therefore their temporal trends were not reported in the present study. To finish this work in future may be informative in clinical practice. Finally, the current study had an adequate sample size based on statistical analysis, but it is a single-cohort observational analysis after all; so, before routine clinical application of serum occludin as a prognostic biochemical indicator of sTBI, a larger and even multi-center cohort study with diverse populations and settings is in future necessitated for the sake of better enhancing reliability and scientificity of the conclusions.

Conclusions

There was a notable increase in serum occludin levels following sTBI, which was independently associated with severity, ALI, and six-month poor prognosis. Moreover, the association between serum occludin levels and poor prognosis is partly mediated by ALI. In summary, serum occludin level may be a prognostic predictor of clinical prospects, and ALI may partially explain the association between serum occludin level and poor prognosis after sTBI.