Neutrophil-to-Lymphocyte Ratio at Hospital Admission as a Novel Predictor of Early Growth of Intraparenchymal Haemorrhage in Patients With Traumatic Brain Injury

This study aimed to explore the association between the neutrophil-to-lymphocyte ratio (NLR) and early growth of traumatic intraparenchymal haemorrhage (tICH) in patients with traumatic brain injury. A multicentre, observational cohort study was conducted at four hospitals and included patients with cerebral contusion undergoing baseline computed tomography (CT) for haematoma volume analysis within 6 hours after primary injury and who had follow-up visits within 48 hours. Routine blood tests were performed upon admission and analysed with early PIH. Logistic regression and receiver operating characteristic (ROC) analysis was used to explore the predictive value of the NLR for haematoma expansion.


Abstract Background
This study aimed to explore the association between the neutrophil-to-lymphocyte ratio (NLR) and early growth of traumatic intraparenchymal haemorrhage (tICH) in patients with traumatic brain injury.

Methods
A multicentre, observational cohort study was conducted at four hospitals and included patients with cerebral contusion undergoing baseline computed tomography (CT) for haematoma volume analysis within 6 hours after primary injury and who had follow-up visits within 48 hours. Routine blood tests were performed upon admission and analysed with early PIH. Logistic regression and receiver operating characteristic (ROC) analysis was used to explore the predictive value of the NLR for haematoma expansion.

Results
The nal analysis included 1003 patients in the retrospective development and validation cohorts. In the retrospective development cohort, the NLR were higher in the PIH group than in the non-PIH group (P<0.0001).
Multivariate logistic regression analysis revealed that a higher NLR was independently associated with PIH (P<0.0001). ROC curve analysis showed that the NLR had a sensitive ability for predicting PIH (AUC, 0.91 [95% CI, 0.88-0.94]). In the validation study, the NLR had a similar ability to predict PIH.

Conclusion
The NLR can be used to easily assess the growth of tICH and calculated using routine laboratory tests. A high NLR is independently predictive of early growth of tICH and may aid in risk strati cation of patients with tICH on admission.

Background
Despite extensive study and improvements in critical care, outcomes after early intraparenchymal haemorrhage growth in patients with traumatic brain injury (tICH) continue to be poor and di cult to predict.
In developed countries tICH remains the commonest cause of death among individuals younger than 40 years, and in developing countries, the incidence and societal costs of tICH are rising. It is well demonstrated that tICH includes numerous types of insults to the brain, one of the most serious being haemorrhagic cerebral contusion. Occurring in more than 40% of severe tICH cases, intraparenchymal haemorrhage plays an important role in conferring a poor prognosis. In particular, secondary damage resulting from progressive intraparenchymal haemorrhage (PIH), which is de ned as haematoma growth >33% or 5 cm 3 on subsequent computed tomographic (CT) scanning, has been reported in 38%-59% of patients with tICH and is an independent predictor of a poor functional outcome [1,2].
There is a large body of evidence suggesting that neuroin ammation is an important injury mechanism that contributes to ongoing neurodegeneration and neurological impairments associated with tICH. Posttraumatic neuroin ammation is characterized by glial cell activation, leukocyte recruitment, and upregulation of in ammatory mediators. Although many studies have focused on the detrimental effects of neuroin ammation on the injured brain, there are clear bene cial effects that can be achieved if neuroin ammation is identi ed as a novel predictor for haemorrhage growth that can be is crucial for early therapeutic intervention.
Accumulative evidence has shown that many risk factors, such as baseline haematoma volume, early baseline computed tomography (CT) time, Glasgow Coma Scale (GCS), subarachnoid haemorrhage (SAH), subdural haemorrhage (SDH), and coagulopathy, are associated with haematoma expansion. The neutrophil-to-leukocyte ratio (NLR), which is a signi cant indicator for predicting the in ammatory status of patients, has been shown to be a predictor of prognosis among patients with conditions involving the brain including glial tumours [3,4], ischaemic stroke [5], haemorrhagic stroke [6], and convulsive status epilepticus [7]. Moreover, we have previously shown that in patients with severe tICH, the NLR can predict long-term outcomes [8]. However, there has been a lack of research on whether the NLR can predict early haematoma growth in patients with tICH. Therefore, the aim of this study was to assess the value of the NLR for predicting intraparenchymal haematoma growth in patients with tICH.

Patient population
Consecutive patients with primary traumatic cerebral contusion who were admitted to 1 of 4 hospitals (First and Second A liated Hospitals of Shantou University Medical College, Jieyang People's Hospital and Fuzhou General Hospital of Xiamen University) between January 1, 2012, and April 31, 2019, were enrolled in this cohort study.
Patients admitted between 2012 and 2015 were assigned to the retrospective development cohort, and those admitted between 2016 and 2019 were assigned to the prospective validation cohort. The inclusion criteria were as follows: (1) at least 18 years; (2) documentation of a baseline CT scan within 6 hours after brain injury and a follow-up CT within 48 hours after the initial CT; (3) documentation of an initial blood test within 24 hours; and (4) at least one con rmed PIH on the initial CT. The exclusion criteria were as follows: (1) surgery performed before the follow-up CT scan, (2) previous head trauma, (3) previous coagulopathy, or (4) use of antiplatelet or anticoagulant medication.

Data collection
All patients underwent a brain CT scan immediately after admission. The follow-up CT scan was routinely performed within 48 hours of the initial CT or when the patient's condition deteriorated. The haematoma volume was calculated by CT. Inter-reader variability was determined by having the CT image analysed by 2 independent neuroradiologists who were blinded to the treatment (the number of 5-mm slices containing haemorrhage was multiplied by 0.5) [9]. Haematoma expansion was de ned as a 33% or more than 5-mL increase in volume on the follow-up CT scan compared with that on the baseline CT, as previously de ned [10,11] (Figure 1).
Venous blood samples were drawn by venous puncture on admission and stored in tubes containing various anticoagulants for routine blood tests. Routine blood examinations, including examinations of the leukocyte count (reference range, 3.5-9.5 10 9 /L), neutrophil count (reference range, 1.8-6.4 10 9 /L), lymphocyte count (reference range, 1.1-3.2 10 9 /L) and mononuclear cell count (reference range, 0.1-0.6 10 9 /L), were measured for all patients by the routine laboratory assays used at the participating hospitals. The NLR is the number of neutrophils divided by the number of lymphocytes.

Statistical analysis
Data were analysed using SPSS 22 (SPSS Inc., Chicago, Illinois) and MedCalc 18.2.1 (MedCalc Software, Mariakerke, Belgium). Continuous variables are expressed as the means ± standard deviations, and categorical variables are expressed as counts (percentages). Continuous variables were compared by a two-sample t-test, whereas categorical data were analysed using the Pearson χ2 test or Fisher's exact test. A univariate analysis with a non-linear correlation (cubic spline functions) was used to evaluate the shape of the relationship between the continuous variables and outcomes. A multivariate logistic regression model analysis was used to identify the associations between PIH expansion or the indices of in ammation, including the leukocyte count, neutrophil count, lymphocyte count, mononuclear cell count and NLR, and their corresponding risk factors (selection: forward [method = Wald]). The results are presented as odds ratios (ORs) and 95% con dence intervals (CIs). Receiver operating characteristic (ROC) curve analysis was performed to assess the predictive performance for PIH expansion by the NLR values on admission. Using the ROC curve, the cut-off values were estimated, and the corresponding sensitivities and speci cities were calculated based on the area under the curve (AUC) of the ROC curve. Statistical signi cance was set at P<0.05.

General information
Based on the eligibility criteria, 412 of 1498 patients with primary traumatic parenchymal haemorrhage were included in the nal development cohort, and 591 of 1516 patients were included in the nal validation cohort ( Table 1). The baseline clinical characteristics were comparable between the development and validation cohorts.

Prediction of haematoma expansion
In the univariate analysis, SAH, SDH, time to baseline CT, baseline CT haematoma volume, coagulopathy, multiple haematomas (no less than 3 haematomas), lymphocyte count and NLR were associated with haematoma expansion in both cohorts ( Table 2) and were entered into a multivariable logistic regression. Lymphocyte count, SDH, multiple haematomas and the NLR remained signi cant in the multivariable analysis ( Table 3).
The NLR was an independent predictor of haematoma expansion, and for the diagnostic performance for haematoma expansion, the NLR displayed a sensitivity of 0.81 and a speci city of 0.87. The predictive  Figure 3).

Operation group vs non-operation group
Among all 505 patients with haematoma expansion, 166 required surgical intervention. Univariate analysis was performed to identify risk factors for surgical intervention. Statistically signi cant differences in the GCS score, mean arterial pressure (MAP), initial haematoma volume, SAH, SDH, epidural haematoma (EDH), encephalatrophy and NLR were found between those who needed surgical intervention and those who did not (Table 5). Multivariate logistic regression revealed that the GCS score (OR = 0.929, 95% CI, 0.873-0.988), MAP (OR = 1.010, 95% CI, 1.001-1.020), initial haematoma volume (OR = 1.064, 95% CI, 1.035-1.093) and NLR (OR = 1.0256, 95% CI, 1.003-1.047) were risk factors for the requirement for surgical intervention ( Table 6). The NLR values of the operation group and the non-operation group were signi cantly different (P = 0.004).

Discussion
The current study demonstrated that the values of the NLR on admission of the patients with PIH were signi cantly higher than those of the patients without PIH. Furthermore, in the multivariate logistic regression models of predictors of early growth of traumatic intracranial haematoma, the NLR on admission was a signi cant independent predictor of traumatic intracranial haematoma expansion. Importantly, the prognostic performance of the NLR for PIH was higher than that of in the multifactor model, substantiating the potential of the NLR as a new prognostic marker in PIH. Interestingly, in patients with haematoma expansion, the NLR of the surgical group was signi cantly higher than that of the non-surgical group. Taken together, these ndings raise the intriguing possibility that neuroin ammation plays a role in the pathophysiology of PIH.
In a recent study, haematoma expansion occurred in approximately one-third of patients and was strongly associated with poor outcomes [12]. As a potentially modi able determinant of tICH prognosis, haematoma expansion represents an appealing target for acute tICH treatment [13]. In recent years, novel predictors for early haematoma growth in patients with tICH, including leukocyte count [14], coagulopathy [15,16] and the shape of the haematoma on a head CT, have been developed [17]. The presence of SDH has been identi ed as a promising imaging predictor for haematoma growth. Interestingly, our study shows that SDH may be closely related to haematoma expansion when haematoma and SDH appear together on a head CT. In particular, the possibility of haematoma expansion is greatly increased when the haematoma is connected with SDH. The potential mechanisms might be an effect of local pressure around the haemorrhage in cases of PIH. Traumatic haematoma growth often occurs during the rst hours after trauma and has been attributed to the continued bleeding of microvessels that were fractured at the time of the primary injury [18]. Patients with subdural haemorrhage are known to be associated with a more severe presentation and a worse clinical outcome. The pressure around the haematoma decreases as intracranial haematoma in connection with SDH, especially in the early stage, result in the rupture of microvessels around the haematoma and in continuous bleeding.
We provide important novel data showing that compared with SDH, NLR is a more sensitive indicator for predicting for early haematoma growth. A high NLR is re ective of both an elevated innate immune response (more polymorphonuclear leukocytes (PMNs)) and a decreased adaptive immune response (fewer lymphocytes) [19]. Similarly, a growing body of evidence supports the presence of systemic immunosuppression following tICH. The in ammatory response in the hyperacute phase of the disease is not only a nonspeci c stress-related reaction but may also play a key role in the development of haematomas [8]. Zhou at al. described the role of in ammation in intracranial haematoma, from the underlying mechanisms to clinical translation [20]. In particular, the secondary damage caused by in ammation results in neurological deterioration in patients with tICH [21]. Secondary damage is triggered by the presence of intraparenchymal blood, which subsequently activates cytotoxic, excitotoxic, oxidative, and in ammatory pathways [22,23]. The in ammatory response of tICH is characterized by the rapid activation of resident microglial cells and the subsequent in ltration of circulating in ammatory cells, including neutrophils and macrophages [24]. Similarly, in the early stages of tICH, large numbers of in ammatory cells are seen around the haematoma in animal studies [25]. These in ammatory cells release in ammatory cytokines that cause secondary injury around the haematoma, leading to the enlargement of the haematoma. Neutrophils may be the rst leukocyte subtype to enter a haemorrhage [24]. When the leukocyte invasion is greater, the degree of damage increases. Thus, the NLR predicts haematoma expansion due to the role of neutrophils in haematoma. This might explain how the NLR may be used as predictor for haematoma growth.
The NLR derived from the white blood cell differential count, a routine laboratory study, is easy to obtain and calculate, easy to integrate into daily practice, and does not add extra costs. Another advantage of the NLR is that it is more objective than other predictors (subarachnoid haemorrhage, SDH, the shape of the haematoma on a head CT, etc). Furthermore, slight changes in a patient's physical condition may not be re ected on head CT, whereas it may be re ected in the NLR, which could lead to changes in treatment. One drawback of the NLR is the average turnaround time for relevant laboratory results in the emergency department, which is between 30 and 40 minutes. In comparison, head CT images are able to be viewed within 10 minutes.
Some limitations of our study should be acknowledged. First, some bias might have been introduced in the patient selection and data collection. Accurately calculating the haematoma volume, particularly for a few small haematomas, is di cult with normal CT plane scanning. Second, in ammatory cells and other indicators examined were just analysed on admission, and the changes in in ammatory indicators over the development of the patient's condition were not tracked.

Conclusions
Our study reveals the novel and easy-to-use NLR that predicts early intraparenchymal haematoma growth in patients with tICH. The NLR can be easily identi ed by a routine laboratory study and is highly speci c and sensitive for predicting haematoma growth.

Declarations
Ethics approval and consent to participate This study was approved by the Ethics Committee of the First A liated Hospital, Shantou University Medical College.

Consent for publication
All the selected patients have signed informed consents

Availability of data and materials
The datasets used during the current study are available from the corresponding author on reasonable request.         Predictive performance of the single-factor model of the NLR vs. multifactor model in the retrospective cohort and validation cohort. A. retrospective cohort; B. validation cohort.