Association of vasopressor use with mortality in severe traumatic brain injury: a nationwide retrospective cohort study in Japan

Background: Maintaining of cerebral perfusion pressure (CPP) is frequently incorporated in severe traumatic brain injury management algorithms. However, there is limited evidence on prevalent clinical practices regarding a preferred method for achieving such CPP aims. We conducted a nationwide retrospective cohort study to determine the association between the use of vasopressors and mortality following hospital discharge in patients with severe traumatic brain injury, and to determine whether the use of vasopressors affects emergency department mortality or the occurrence of cognitive dysfunction. Methods: Data were collected between January 2004 and December 2018 from the Japanese Trauma Data Bank, which includes data from 272 emergency hospitals in Japan. Adults aged ≥ 16 years with severe traumatic brain injury, without other major injuries, were examined. A severe traumatic brain injury was dened based on the Abbreviated Injury Scale code and a Glasgow Coma Scale score of 3–8 on admission. Multivariable analysis and propensity score matching were performed. Statistical signicance was assessed using 95% condence intervals (CIs). Results: In total, 10 284 patients were eligible for analysis, with 650 patients (6.32%) included in the vasopressor group and 9634 patients (93.68%) included in the non-vasopressor group. The proportion of deaths on hospital discharge was higher in the vasopressor group than in the non-vasopressor group (81.69% [531/650] vs. 40.21% [3,874/9,634]). This nding was conrmed by multivariable logistic regression analysis (adjusted odds ratio [OR], 5.71; 95% CI: 4.56–7.16). Regarding propensity score-matched patients, the proportion of deaths on hospital discharge remained higher in the vasopressor group than in the non-vasopressor group (81.66% [530/649] vs. 50.69% [329/649]) (OR, 4.33; 95% CI: 3.37–5.57). The vasopressor group had a higher emergency department mortality rate than the non-vasopressor group (8.01% [52/649] vs. 2.77% [18/649]) (OR, 3.05; 95%

Conclusions: In this population, the use of vasopressors for severe traumatic brain injury was associated with higher mortality on hospital discharge. Our results could not reach the conclusion that vasopressor therapy is bene cial in improving clinical outcomes in patients with severe TBI.

Background
Severe traumatic brain injury (TBI) is a leading cause of death and disability in people of all generations worldwide [1][2][3] and is associated with high economic and social costs [4]. Severe TBI is characterized by a primary insult that involves the initial mechanical force of impact and results in loss of brain tissue and neuronal cell death. Secondary injury occurs in the minutes and hours following the primary injury.
Although little can be done to reverse the primary brain damage, secondary brain injury due to dysregulation of cerebral blood ow is potentially preventable. Therefore, acute management of patients after severe TBI aims to minimize secondary brain injury by maintaining cerebral perfusion pressure, de ned as the difference between mean blood pressure and intracranial pressure [5][6][7][8].
Blood pressure control is the mainstay of therapy for several acute cerebral disorders, both in the early emergent phase immediately after brain injury and in the subsequent intensive care unit phase [9]. Hypotension in the acute phase of severe TBI is a key factor associated with poor clinical outcome [7][8][9].
Although vasopressors are considered to be harmful to trauma patients, in cases of a hemorrhagic shock state [10], The most recent international guidelines on the management of severe TBI suggests that maintaining systolic BP at ≥100 mm Hg for patients 50 to 69 years old or at ≥110 mmHg or above for patients 15 to 49 or >70 years old may to decrease mortality and improve outcomes [5]. However, this is based on level evidence, and it cannot specify a preferred method for achieving such systolic BP goals.
Therefore, the aim of this study was to assess the possible association of vasopressor use with mortality in patients with severe TBI, using data from the Japanese Trauma Data Bank (JTDB) registry, which represents the biggest trauma data bank in Japan. Survivors of severe TBI often have cognitive control function de cits [11,12]. Therefore, we also assess association of vasopressor use with the occurrence of cognitive dysfunction.

Study design, population, and setting
This was a nationwide retrospective cohort study conducted using the JTDB database. We included cases registered in the database from January 2004 to December 2018 and patients aged ≥16 years who had a TBI and were transported to a JTDB-participating hospital. We segregated patients with severe TBI based on the Abbreviated Injury Scale (AIS) code (TBI = 140000) and a Glasgow Coma Scale score of 3-8 on admission. We excluded cases with a maximum head AIS score of 6 (unsurvivable injury) or 9 (unspeci ed injury), with other critical injuries (AIS scores of ≥3), in cardiopulmonary arrest on or before hospital arrival, that underwent cardiopulmonary resuscitation (i.e., use of adrenaline during cardiopulmonary resuscitation), or that required interhospital transport. We also excluded cases with missing outcome data or variables required for propensity score (PS) matching [13,14]. This study de ned patients in cardiopulmonary arrest as those whose systolic blood pressure was 0 mmHg and/or heart rate was 0 bpm on or before hospital arrival [15].

Japanese Trauma Data Bank
The JTDB was launched in 2003 by the Japanese Association for the Surgery of Trauma (Trauma Surgery Committee) and the Japanese Association for Acute Medicine (Committee for Clinical Care Evaluation) [15,16], similar to the trauma databases in North America, Europe, and Oceania. By 2018, 272 major emergency medical institutions across Japan had been registered in the JTDB database [16]. The included hospitals have service levels similar to Level I trauma centers in the United States. Data were collected via the Internet from participating institutions. The physicians and medical assistants who attended the AIS coding course were the main inputters of the data [15,16]. The JTDB records trauma patient data including age; sex; mechanism of injury; AIS code (1998 version); Injury Severity Score (ISS); vital signs on hospital arrival; date and time series from hospital arrival to discharge; medical treatments such as interventional radiology, surgery, and computed tomography; complications; and mortality on discharge [15,16]. The ISS was calculated from the top three AIS scores in nine sites classi ed using the AIS codes. The data used in this study were the most recent data available in this registry.

Study endpoints
The primary outcome of this investigation was death on hospital discharge. The secondary outcomes were emergency department (ED) mortality and in-hospital cognitive dysfunction, which diagnosed by medical team in hospital.
Propensity score matching In this study, we selected matched PS analysis as the use of vasopressors was not randomly assigned. A logistic regression analysis was performed to estimate a PS for the prediction of vasopressor use from the available predictors. Confounders were carefully selected from previous reports [9, 10, 17-20], and clinically important confounders were included to estimate PSs. The probability of receiving vasopressors (PS) for each patient was calculated using multivariable logistic regression analysis based on the following 14 variables: age (continuous variable), sex (male/female), year of onset (2004-2006, 2007-2009, 2010-2012, 2013-2015, 2016-2018), Glasgow Coma Scale score (continuous variable), systolic blood pressure (continuous variable) on admission to the ED, surgery indicated for TBI (no/yes), type of injury (blunt/non-blunt), cause of trauma (motor vehicle accident, fall, sports, other), type of TBI (isolated TBI, TBI with skull fracture, TBI with intracranial vessel injury, TBI with other head injuries), prehospital intravenous uid administration (no/yes), use of transfusion in the rst 24 hours (no/yes), past medical history of stroke (no/yes), use of anticoagulant or antiplatelet drugs (no/yes), and ISS (continuous variable). We performed a receiver operating characteristic curve analysis with the area under the curve predicting the use of vasopressors in patients with severe TBI. One-to-one pair matching between the nonvasopressor and vasopressor groups was performed by nearest-neighbor matching without replacement, using a caliper width of 0.02 as the standard deviation of the PS. Covariate balances before and after matching were checked by comparing standardized mean differences (SMDs). An SMD <10% was considered a negligible imbalance between the two groups. In the PS-matched cohort, univariable logistic regression analysis was performed to assess the association between the use of vasopressors and outcomes.

Statistical analysis
We divided patients into two groups (vasopressor and non-vasopressor). Descriptive data are presented as counts and percentages (categorical variables) or medians and interquartile ranges (continuous variables). Outcomes were evaluated using univariable and multivariable logistic regression analyses to assess the independent effect of vasopressor use. Based on these analyses, we calculated the odds ratios (ORs) and 95% con dence intervals (CIs). Based on previous reports [9, 10, 17-20], we selected confounders for multivariable logistic regression analysis on the assumption that none were directly affected by the use of vasopressors. In the multivariable logistic regression model, we adjusted for the aforementioned 14 variables used in the PS calculation. In addition, subgroup analyses were performed to identify the potential bene ts and drawbacks of the use of vasopressors. In each subgroup, a multivariable logistic regression analysis adjusted for the aforementioned variables was performed to assess the independent effect of vasopressor use on mortality on hospital discharge. P for interaction was calculated using the multivariable logistic regression model.
All statistical analyses were conducted using STATA (version 16) (StataCorp LP, 4905 Lakeway Drive, College Station, Texas, USA). Statistical signi cance was de ned as a two-sided p value <0.05 or assessed using 95% CIs. This manuscript was written based on the STROBE statement for the reporting of cohort and cross-sectional studies [21].

Results
A total of 10 284 patients were included in the study. Of those, 650 patients (6.32%) received vasopressors and 9634 patients (93.68%) did not. Fig. 1  GCS, Glasgow Coma Scale; ISS, Injury Severity Score; IV, intravenous injection; PMH, past medical history; PS, propensity score; SBP, systolic blood pressure; SMD, standardized mean difference; TBI, traumatic brain injury. Table 1 shows the baseline characteristics of the PS-matched patients. Following PS matching, 649 patients in each group were included. The area under the receiver operating characteristic curve for PS was 0.73. The characteristics of PS-matched patients were nely balanced in terms of absolute SMD.
Although two groups was slightly imbalanced about type of TBI, the characteristics between the groups were almost similar (sTable1).
The results of the multivariable logistic regression analysis and PS matching for the primary outcome are presented in Table 2. Mortality on hospital discharge was higher in the multivariable logistic regression vasopressor group than in the non-vasopressor group (adjusted OR, 5.71; 95% CI: 4.56-7.16). For PSmatched patients, mortality on hospital discharge was 81.66% (530/649) in the vasopressor group and 50.69% (329/649) in the non-vasopressor group. PS matching analysis illustrated that death on hospital discharge was higher in the vasopressor group than in the non-vasopressor group (OR, 4.33; 95% CI: 3.37-5.57). Table 3 shows the results of the multivariable logistic regression analysis and PS matching for the secondary outcomes. For PS-matched patients, ED mortality in the vasopressor group was higher than that in the non-vasopressor group ( Subgroup analysis suggested that the use of vasopressors for severe TBI was associated with higher mortality on hospital discharge for each factor (Fig. 2)

Discussion
This was a retrospective cohort study conducted to evaluate the effect of vasopressor use on mortality among patients with severe TBI, using a nationwide trauma database in Japan. With robust analyses to adjust for severity, it was found that the use of vasopressors was signi cantly associated with higher mortality, not only on hospital discharge but also in the ED. Subgroup analysis also reiterated the same and suggested that the use of vasopressors for severe TBI was associated with higher mortality on hospital discharge in almost all cases.
Hypotension (de ned as systolic blood pressure <90 mmHg) is a well-known risk factor associated with the occurrence of secondary cerebral damage and poor outcome, especially after TBI [7][8][9]. In contrast, the large IMPACT prospective database found that a systolic blood pressure <120 mmHg is also a strong predictor of unfavorable neurological recovery [22]. Based on these ndings, pharmacological elevation of blood pressure is frequently incorporated in severe TBI management algorithms aimed at preventing or treating cerebral ischemia caused by reduced cerebral perfusion pressure [5]. Further, vasopressors may be used as volume-sparing resuscitating agents to prevent brain edema. However, our study showed that neither mortality nor the occurrence of cognitive dysfunction was improved by the use of vasopressors. Therefore, these results suggest the potential harms of this therapy. A similar recommendation that vasopressors should be avoided in any cases of severe TBI was provided by Lund [23].
Only a few studies have tried to assess whether the use of vasopressors is bene cial for severe TBI, although a number of studies have focused on which type of vasopressor should be used [24,25]. Our ndings could pave the way for future guidelines regarding the use of vasopressors for the treatment of severe TBI.
The penumbra, the brain tissue surrounding the impacted core of the TBI, becomes particularly vulnerable to cell death by hypoxic insults, and preservation of this area is one of the important objectives for using vasopressors. The overall effects of vasopressor use on the brain are probably a consequence of systemic arterial contraction but are also in uenced by other factors such as brain injury, integrity of the blood-brain barrier, and the status of cerebral autoregulation [26]. In the abnormal state of autoregulation after severe TBI, excessive elevation of intracranial pressure favors edema formation by increased capillary hydrostatic pressure across the blood-brain barrier, causing brain herniation [27]. This may also result in unwanted hemodynamic effects, such as intracranial hemorrhage, leading to increased mortality [28,29]. Additionally, vasopressors cause an array of adverse effects among other organs of the body [30, 31]. Catecholamine surge after TBI can lead to peripheral insults induced by the release of proin ammatory substances and result in increased vascular permeability [30]. In this situation, the accentuated proin ammatory response could trigger the development of acute respiratory distress syndrome [31]. Other potential negative systemic side effects of vasopressors include arrhythmia, diuresis, and increased left ventricular afterload.
In patients with severe TBI, low arterial blood pressure, major surgery with bleeding, and blood transfusion can contribute to secondary insults to the brain and aggravate the primary insults and cerebral edema, increasing the risk of developing severe lung injury, or even multiple organ failure. Traditionally, it has been suggested that the clinical effects of brain injuries vary depending on individual characteristics such as sex, age, past medical history, and type of TBI. With multiple confounding factors, planning treatment for TBI is especially challenging. In this study, we used the largest trauma dataset in Japan that allowed us to adjust for confounding factors. To the best of our knowledge, our study included the largest cohort of its kind, to date, that has focused on patients with severe TBI. This study also reveals the utility of vasopressors in some speci c conditions (e.g., in patients with TBI with a history of stroke). In cases of stroke, structural cerebrovascular changes may increase resistance; thus, normal cerebral blood ow can be maintained using vasopressors. However, the use of vasopressors in such a subgroup did not reduce mortality at the time of hospital discharge. These results suggest that vasopressors should be avoided in most cases of severe TBI.
The authors acknowledge a few limitations of this observational study. First, PS matching analysis has the risk of residual selection bias. Some differences in the two groups may still exist, even after PS matching, particularly if data on important confounding factors are not included in the analysis. Second, although the use of vasopressors (mainly catecholamine) for resuscitation was recorded in the JTDB, important information such as the type and dose of vasopressors used and the timing of vasopressor administration was not included, leading to selection bias. Detailed information about the dose, type of vasopressor, mode of administration (bolus and/or continuous), and patients' vital signs when the vasopressor was initiated was not included in the JTDB. However, all TBI patients transported to JTDBparticipating hospitals were treated based on the guidelines for the management of severe TBI, which state that vasopressors are recommended to maintain a systolic blood pressure >110 mmHg [32]. Third, the incidence of cognitive disorders could have been underestimated in our study as the JTDB includes clinical data until hospital discharge, which would largely cater to the acute phase of the injury. Cognitive symptoms after TBI would be accurately evaluated only after recovering from an altered state of consciousness in the acute phase. Additionally, the data included in this study primarily involved cases of blunt trauma, which cannot be extended to penetrating TBI. Lastly, this was an observational study, and there may be other unknown confounding factors. The results of this investigation could not establish causality and remain limited to associations. a randomized controlled trial regarding intracranial pressure ICP management or vasopressors use for maintaining CPP, and regarding blood transfusion or vasopressors use with hypotensive TBI patients.
Blood pressure augmentation to avoid hypotension has been considered to attenuate secondary cerebral damage. Therefore, the management of blood pressure using vasopressors during the emergent phase and intensive care unit phase is based on low-quality evidence and strong recommendations according to recent guidelines for the management of severe TBI [5,9,32,33]. However, our results suggest that vasopressors should be avoided in most cases of severe TBI. Our study has the potential to result in policy and guideline changes for the treatment and management of severe TBI.

Conclusions
In a Japanese population with severe adult TBI associated with blunt trauma, the use of vasopressors was associated with increased mortality, both on hospital discharge and in the ED. Our results could not reach the conclusion that vasopressor therapy is bene cial in improving clinical outcomes in patients with sTBI. Abbreviations AIS, Abbreviated Injury Scale; CI, con dence interval; ED, emergency department; ISS, Injury Severity Score; JTDB, Japanese Trauma Data Bank; OR, odds ratio; PS, propensity score; SMD, standardized mean difference; TBI, traumatic brain injury.

Declarations
Ethics approval and consent to participate: This study was approved by the Ethics Committee of Osaka University Graduate School of Medicine (Osaka, Japan) (approval number: 16260). Personal identi ers were removed from the JTDB database beforehand; thus, the requirement for informed consent was waived.
Consent for publication: Not applicable. Table 2. Primary outcome comparisons between patients with and without vasopressor use before and after PS matching CI, confidence interval; OR, odds ratio; PS, propensity score. Table 3. Secondary outcome comparisons between patients with and without vasopressor use before and after PS matching