Association between blood glucose levels in insulin therapy and Glasgow Outcome Score in patients with traumatic brain injury: secondary analysis of a randomized trial

Tao Yuan Department of Neurosurgery, the a liated Lianyungang Oriental Hospital of Xuzhou Medical University Hongyu He Department of Neurosurgery, the a liated Lianyungang Oriental Hospital of Xuzhou Medical University Yuepeng Liu Center for clinical research and translational medicine, the a liated Lianyungang Oriental Hospital of Xuzhou Medical University Jianwei Wang Department of Neurosurgery, the a liated Lianyungang Oriental Hospital of Xuzhou Medical University Xin Kang Department of Neurosurgery, the a liated Lianyungang Oriental Hospital of Xuzhou Medical University Guanghui Fu Department of Neurosurgery, the a liated Lianyungang Oriental Hospital of Xuzhou Medical University Fangfang Xie Department of Neurosurgery, the a liated Lianyungang Oriental Hospital of Xuzhou Medical University Aimin Li Lianyungang No 1 People's Hospital Jun Chen Lianyungang No 1 People's Hospital Wen-xue Wang (  760020210075@xzhmu.edu.cn ) Department of Neurosurgery,the a liated Lianyungang Oriental Hospital of Xuzhou Medical University https://orcid.org/0000-0002-9865-6811


Introduction
Traumatic brain injury (TBI) leads to death and disability in patients with trauma. Secondary brain injury affects the prognosis of patients with TBI signi cantly, while hyperglycemia is one of the important factors inducing secondary brain injury [1]. Recent studies con rmed that hyperglycemia could aggravate the damage of nerve function [2]. Acute hyperglycaemia following TBI, which is de ned as the blood glucose level beyond 200 mg/dL (11 mmol/L) during the early phase of injury, is a common symptom in patients with severe TBI [3,4]. Acute hyperglycaemia following TBI was supposed to be a physiological reaction, which was essential to support the high metabolism in the brain after TBI [3,5]. However, acute hyperglycaemia was also reported to adversely affect the outcomes of patients with TBI [4,6] through exacerbating the secondary injury [7]. Since maximum clinical observations agreed on the existence of a close relationship between acute hyperglycaemia and poor outcomes of patients with TBI [4,[6][7][8][9][10], early care of acute hyperglycaemia after TBI was clinically recommended [11]. However, in clinical practice, hypoglycaemia, which results from excess blood glucose control, was also found to adversely affect the outcomes of patients with TBI [12].
Too high or low blood glucose levels after TBI negatively affect the prognosis of patients with TBI [13,14]. Therefore, the control of the blood glucose level in patients becomes an urgent problem directly related to the prognosis. Therefore, it has been hypothesized that adequate control of acute hyperglycaemia was necessary for patients' bene ts [11,12]. A previous randomized controlled trial (RCT) con rmed the aforementioned hypothesis; 7-13 mmol/L was recommended as the target blood glucose range in insulin therapy. This study used advanced data-analysis tools to con rm our previous ndings and further narrow the target blood glucose range to direct the clinical practice.

Patients
The RCT was registered in ClinicalTrials.gov (NCT02161055) on June [15], hyperglycemia was de ned as rapid blood glucose >7 mmol/L (126 mg/dL). Eligible patients with hyperglycemia after severe TBI were randomly assigned to either the intensive insulin therapy (IIT) group or the non-IIT group in the ratio of 3:1. The IIT group was further subdivided into three subgroups based on the target blood glucose level: 4.4-7.0 mmol/L (strict-control group), 7.1-10.0 mmol/L (moderate-control group), and 10.1-13.0 mmol/L (slight-control group). Computerized randomization was accomplished by investigators who could not contact the participants directly. Patients, outcome assessors, and statisticians were blinded to the information regarding grouping. If the eligible patients consented to the trial, sealed opaque envelopes with a randomly assigned serial number containing the accepted treatment programs were opened. The patients then underwent the corresponding treatment measures. In the case of any error or disclosure about randomization, a new randomization sequence was generated starting from the problematic serial number and applied to subsequent patients.

Inclusion and exclusion criteria
The inclusion criteria were as follows: (1) clinical diagnosis of severe closed TBI [16]; (2) Computed Tomography (CT) con rmation of severe closed TBI; (3) severe TBI following the indications for craniotomy; (4) blood glucose levels of >7.0 mmol/L measured twice by rapid examination within 2 h after admission; (5) Glasgow Coma Score (GCS) of 3-8; (6) age 18-80 years; and (7) no history of diabetes mellitus. The exclusion criteria were as follows: (1) patients with multiple physical injuries; (2) patients with diabetic nephropathy with hemodialysis dependence; (3) patients with neurological disorders before craniocerebral trauma; (4) patients with a history of diabetes before craniocerebral trauma; (5) bilaterally dilated pupils; and (6) refusal of patient's relatives. Moreover, patients could withdraw from the trial under the following conditions: (1) discontinuation of the trial was necessary from a medical point of view, or (2) a request was received from the patient's family to stop the trial.

Blood glucose control in insulin therapy
Within the rst week of hospitalization, rapid blood glucose levels were recorded once every 2 h in each group. Blood glucose measurement: Capillary blood was obtained from the tip of the ring nger to measure the blood glucose level. The blood should be collected from the same nger to ensure an accurate measurement. For measuring the blood glucose level of a patient undergoing transfusion, the blood should be collected from the tip of the nger of the limb without transfusion to ensure the accuracy of measurement.
In the IIT group, the blood glucose levels were monitored and controlled according to the Yale Insulin Infusion Protocol [17]. According to the Yale Insulin Infusion Protocol, the amount of insulin (U) = [fasting blood glucose (mmol/L) × 18-100] × 10 × body weight (kg) × 0.6/(1000 × 2). The insulin for injection (400 U/10 mL) was obtained from Wan-bang Biochemical Pharmaceutical Co., Ltd., China (Lot # 1307230, 1302225, and 1307210). Insulin was infused into the vein at a rate of 0.1 U/(kg × h) using a micropump. During this period, the blood glucose level was monitored once every 2 h, and the insulin dose was adjusted accordingly. If the blood glucose level was higher than the target value, the insulin dose was gradually increased by 1-2 U/h. When the blood glucose level reached the target value, the insulin dose was gradually decreased until its administration was terminated.
On the contrary, in the non-IIT group, the rapid blood glucose level measurement was performed once every 2 h. When the blood glucose level was ≤13.0 mmol/L, no intervention was performed. In the case of blood glucose level >13.0 mmol/L, insulin was subcutaneously injected regularly. Insulin was administered once every 8 h in a fasting state, whereas during venous or enteral nutrition infusion, it was infused 30 min before the nutrition infusion. When the blood glucose level reached ≤13.0 mmol/L, the insulin infusion was terminated.

Measurement of dependent variables
During the original RCT, patients were treated using the following uniform protocol. (1) According to the Guidelines for the management of severe TBI [18], craniotomy for TBI was performed to mainly decompress and remove the hematomas. (2) During the operation, all patients underwent a ventricular puncture. The cerebrospinal uid (CSF) was collected for biochemical analysis and cell culture. (3) All patients were closely monitored in the intensive care unit (ICU) of the Department of Neurosurgery. (4) Therapeutic protocols for severe traumatic brain injury were used. (5) Glucocorticoids, which could cause disorders of glucose metabolism, were not applied regularly in either of the groups. (6) During intravenous injection, glucose and insulin were mixed in a ratio of 5 g (glucose):1 U (insulin) to minimize the effects of exogenous glucose on the blood glucose level. When the patients nished the RCT, routine rehabilitation and daily life training were organized for them. The blood glucose, serum insulin, and blood glycosylated hemoglobin levels, as well as the CSF level of glucose, lactic acid, and chloride, were monitored. CSF was collected during surgery and obtained 1 week after surgery by lumbar puncture. GCS was used to evaluate the severity of the patient's condition, and a recording sheet was used to assess the patient's condition at hospital admission. The Acute Physiology and Chronic Health Evaluation II (APCHE II) score was recorded on each subsequent day in the ICU.

Outcome measurement
The outcome of the study was 5-year Glasgow Outcome Score (GOS). In supplementary analyses, we also analyzed the 6-month GOS. GOS de nitions were as follows: 1, death; 2, persistent vegetative state; 3, severe disability (conscious but disabled), needing daily support; 4, moderate disability (disabled but independent); 5, good recovery, normal active life with minimal de cits.

Statistical analysis
Continuous variables that conformed to normal distribution were described as mean (SD), while those not conforming to normal distribution were described as median (Q1-Q3). Categorical data were presented as numbers and percentages. The univariate analysis was used to scan variables that might contribute to outcomes, while the multivariate regression analysis was used to reveal the independent relationship between insulin therapy and outcomes. A generalized additive model (GAM) was used to investigate dose-response relationships between blood glucose levels and GOS. A linear regression model was employed to estimate the association between blood glucose levels and GOS. The results were presented as β with their 95% con dence intervals (95% CIs). We selected these confounders based on statistical signi cance (P < 0.05) [19]. Adjustments were made for the following potential confounders: pupil changes; GCS before surgery; and APCHE II score.
We further applied a two-piecewise linear regression model to examine the threshold effect of blood glucose levels and GOS. The turning point for blood glucose level was determined using exploratory analyses, which involved moving the trial turning point along with the pre-de ned interval and picking up the point that gave the maximum model likelihood. We also conducted a log-likelihood ratio test comparing the one-line linear regression model with the two-piecewise linear model, as described in previous analyses [20,21].
Dummy variables were used to indicate missing covariate values, which was performed when more than 1% of continuous variables were missing. The two-sided alpha level was set at 0.05. All the statistical analyses were performed using the Empower Stats (www.empowerstats.com, X&Y solutions, Inc., MA, USA) and R software version 3.6.1 (http://www.r-project.org).

Results
Characteristics of the patients A total of 208 participants who participated in the RCT were enrolled in this cohort and followed up to December 2020. Of the total participants, 26 cases were lost to follow-up, and 182 patients were involved in the nal analysis (Fig. 1). For 182 patients, the median of age was 50 (34-59) years, the mean GOS before surgery was 5.46 (1.51) years, the mean APCHE II score before surgery was 28.73 (2.40), the mean blood glucose level before surgery was 19.08 ( (Table 4). Combining slight-control (7.1-10 mmol/L) and moderate-control (10.1-13 mmol/L) IIT groups, we achieved a blood glucose range of 7.1-13 mmol/L, which positively affected the 5-year and 6-month GOS.

Relationships between blood glucose levels and GOS
A threshold, nonlinear association between the average blood glucose level in insulin therapy (mmol/L) and 5-years GOS was found in GAM. The solid red line represented the smooth curve t between variables. Blue bands represented the 95% con dence interval from the t. "Bell"-shaped relationships were found to exist between the average blood glucose level in insulin therapy and 5-year GOS (Fig. 2), in which 6.73 and 8.97 mmol/L (average blood glucose level) were identi ed as in ection points (Table 5).
A similar curve was found between the average blood glucose level in insulin therapy and 6-month GOS (Fig. 2), in which 6.95 and 8.88 mmol/L (blood glucose level) were the in ection points (Table 5).

Multivariate analysis between the average blood glucose level and 5-year GOS
According to the in ection points about 5-year GOS, the average blood glucose level in IIT was assorted into three groups: low-level (<6.73 mol/L), medium-level (³6.73, <8.97 mmol/L), and high-level (³8.97 mmol/L), which then was involved in multivariate regression analysis. The multivariate analysis showed that the 5-year GOS increased by 0.83 (95% CI: 0.22-1.43) in the medium-level group compared with the low-level group; also, the 5-year GOS in the medium-level group also increased by 0.57 compared with that in the high-level group (95% CI: 0.05-1.08) ( Table 6).

Discussion
Insulin therapy was used to clinically control the dramatic increase in the blood glucose level following TBI and bene t the outcomes of patients with TBI; in contrast, this therapy required the target blood glucose range to direct the dosage of insulin [22]. Unlike other brain diseases, such as ischemic stroke and intracerebral hemorrhage, the target blood glucose range for TBI has not been well documented. Although 6.0-10.0 mmol/L was predicted as the appropriate target range of blood glucose [1,22], this range was not fully substantiated by clinical research. In a previous study, based on the effect of IIT on 3and 6-month survival and GOS, we introduced that the blood glucose between 7.1 and 13.0 mmol/L would be the suitable target range for controlling hyperglycemia with insulin therapy following TBI. In the present study, we tried to con rm and further narrow the target blood glucose range in insulin therapy with advanced statistical tools.
During the secondary analysis of RCT with the multivariate regression analysis of the relationship between insulin therapy groups and 5-year and 6-month GOS, a possible ideal blood glucose range, 7.1-13 mmol/L, was supplied, which had a positive effect on the outcomes. As support evidence for the aforementioned analysis, a "bell"-shaped relationship between the average blood glucose level in 7 days in insulin therapy and GOS was found, which was consistent with previous recognition that a too low or too high blood glucose level was harmful to the outcomes of patients [13,14]. With the assistance of curve-tting example and threshold effects analysis, we identi ed two in ection points in the aforementioned curves, following which the blood glucose was segregated into three ranges: low-level, medium-level, and high-level. The multivariate analysis revealed that the 5-year GOS was higher in the medium-level group than in the low-level and high-level groups. The blood glucose range in the mediumlevel group was 6.73-8.97 mmol/L, which was consistent with the blood glucose range reported in a previous RCT and narrower than that previously reported. When using 6-month GOS as the outcome to identify the in ection points, the values of 6.95 and 8.88 mmol/L were achieved. Considering that more positive incidences were involved in 5-year GOS, we adopted 6.73-8.97 mmol/L as the recommended mean blood glucose range for insulin therapy. To quantify the effect of blood glucose level in insulin therapy on GOS, the multivariate analysis was performed between average blood glucose level and 5-year GOS. The result showed that controlling the blood glucose level in the range of 6.73-8.97 mmol/L (medium-level) could increase the 5-year GOS by 0.84 and 0.55 when compared with the low-level group and the high-level group, respectively. This meant that patients could greatly bene t from proper blood glucose control in insulin therapy considering a few GOS changes, indicating a big improvement in prognosis [23].
Due to the in uence of inclusion criteria, our study had some limitations. We selected critical patients with a GCS score of 3-8 and excluded patients with a GCS score of >8. Also, patients with bilateral mydriasis were excluded because of their high mortality. We excluded nonsurgical patients due to some interference differences between surgical and nonsurgical patients. The data used in this study were derived from an RCT that just involved a subset population of patients with TBI. Hence, the conclusion was suitable for a similar subset population. A cohort of patients with TBI of different severity may be a better choice to produce a target blood glucose range in future studies. Besides the aforementioned choice bias, multicenter studies with larger sample size would help achieve a more general conclusion about the target glucose range in insulin therapy.
In conclusion, proper blood glucose range in insulin therapy would improve outcomes of patients with TBI; the range was 6.73-8.97 mmol/L according to our limited analysis.

23.
Zhu C, Chen J, Pan J, Qiu Z, Xu T: Therapeutic effect of intensive glycemic control therapy in patients with traumatic brain injury: A systematic review and meta-analysis of randomized controlled trials. Medicine (Baltimore) 2018, 97(30):e11671.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. CONSORT2010Checklist.doc