Neurology

Traumatic brain injury and intraparenchymal hemorrhage progression: Blood pressure variability matters

a b s t r a c t

Introduction: Blood pressure variability (BPV) has been shown to correlate with intraparenchymal hematoma progression (HP) and worse outcomes in patients with spontaneous intracerebral hemorrhage . However, this association has not been elucidated in patients with traumatic Intraparenchymal hemorrhage or contusion (tIPH). We hypothesized that 24 h-BPV from time of admission is associated with hemorrhagic progression of contusion or intraparenchymal hemorrhage (HPC), and worse outcomes in patients with tIPH.

Method: We performed a retrospective observational analysis of adult patients treated at an academic regional Level 1 trauma center between 01/2018-12/2019. We included patients who had tIPH and >= 2 computer tomog- raphy (CT) scans within 24 h of admission. HP, defined as >=30% of admission Hematoma volume, was calculated by the ABC/2 method. We performed stepwise multivariable logistic regressions for the association between clin-

ical factors and outcomes.

Results: We analyzed 354 patients’ charts. Mean age (Standard Deviation [SD]) was 56 (SD = 21) years, 260 (73%) were male. Mean admission hematoma volume was 7 (SD =19) cubic centimeters (cm3), 160 (45%) had HP. Coefficient of variation in systolic blood pressure (SBPCV) (OR 1.03, 95%CI 1.02-1.3, p = 0.026) was significantly associated with HPC among patients requiring External ventricular drain (EVD). Difference between highest and lowest systolic blood pressure (SBPmax-min) (OR 1.02, 95%CI 1.004-1.03, p = 0.007) was associated with hospital mortality.

Conclusion: SBPCV was significantly associated with HP among patients who required EVD. Additionally, increased SBPmax-min was associated with an increase in mortality. Clinicians should be cautious with patients’ blood pressure until further studies confirm these observations.

(C) 2021

  1. Background

Traumatic brain injury is the leading cause of death in Trauma victims and frequently causes disability and neurological se- quelae in survivors [1]. TBI is of special concern because we have a growing population of TBI survivors with significant TBI-related

* Corresponding author at: 22 South Greene Street, Suite T3N45, Baltimore, MD 21201, USA.

E-mail addresses: [email protected] (C. Tran), [email protected] (A. Aligabi), [email protected] (J. Olexa), [email protected]

(U. Bodanapally), [email protected] (G. Schwartzbauer), [email protected] (Q.K. Tran).

disabilities long after acute medical treatment and rehabilitation [2]. The most common mass lesions associated with traumatic brain in- jury include intraparenchymal hemorrhage (IPH) and contusions [2,10]. Hemorrhagic progression of contusion is detrimental because it results in irrevocable loss of brain tissue, but may be preventable with appropriate medical treatment [3]. Currently, the treatment for traumatic IPH is supportive medical care, which includes the moni- toring of intracranial pressure as necessary, and avoidance of hypotension [4].

Intraparenchymal hemorrhage is a disease that may either develop spontaneously (sICH) or due to a Traumatic event (tIPH). It is important to differentiate between traumatic and non-traumatic ICH types be- cause they have different etiologies and affect different patient

https://doi.org/10.1016/j.ajem.2021.12.005

0735-6757/(C) 2021

populations [5]. Non-traumatic spontaneous ICH is often associated with systemic hypertension resulting in hypertensive hemorrhages, and primarily affects older patients [5]. In contrast, traumatic IPH is caused by physical or traumatic events that result in blood vessel dis- ruption secondary to the impact [5]. The resultant changes in molecular signaling cascades within the cerebral Endothelial cells makes them particularly sensitive to systemic blood pressure variation [6].

blood pressure variability has been shown by Duan and col- leagues to influence spontaneous ICH progression, resulting in poorer outcomes [7]. BPV is defined as the average of absolute differences be- tween consecutive blood Pressure measurements (successive variation [SBPSV]), variation of blood pressure during a period of time (standard deviation [SBPSD]), or coefficient of variation of SBP (SBPCV) [9]. In other words, SBPSV is how “quickly” patients’ systolic blood pressure changes from one measurement to the next. Additionally, SBPSD and SBPCV refer to how “tightly” patients’ blood pressures are controlled over time. It was postulated that in those with spontaneous ICH, cere-

Table 1

Demographic characteristics of 354 patients with traumatic brain injury.

All patients

No Hematoma expansion

Hematoma expansion

>=30%

P

N = 354

N = 194

N = 160

Age, mean (SD)

56 (21)

54 (20)

0.29

Gender, male, N (%)

260 (73)

143 (74)

117 (73)

0.99

BMI, mean (SD)

26 (5)

25 (5)

0.93

direct admission, N (%)

262 (74)

128 (66)

134 (84)

0.003

Transfer from other hospital, N (%)

91 (26)

66 (34)

25 (16)

Past medical history, N (%) Any anticoagulation

22 (6)

11 (6)

11 (7)

0.77

Any Antiplatelet therapy

54 (15)

29 (15)

25 (16)

0.84

Hypertension

135 (38)

77 (40)

58 (36)

0.56

Diabetes

45 (13)

32 (16)

13 (8)

0.08

Mechanism of injury, N (%)

Fall

200 (56)

107 (55)

93 (58)

0.67

bral perfusion pressure (CPP) is dependent on mean arterial pressure

MVC

91 (26)

53 (27)

38 (24)

0.62

(MAP) [8], where hypertension is associated with increased risk of

Other blunt trauma

43 (12)

21 (11)

22 (14)

0.52

bleeding while hypotension is associated with poor CPP. However,

Any penetrating trauma

7 (2)

5 (3)

2 (1)

0.62

there is a paucity of literature pertaining to the relationship between BPV and traumatic IPH.

Unknown 13 (4) 8 (4) 5 (3) 0.99

Admission laboratory values, mean (SD)

While the current practice in treating patients with tIPH is to prevent significant hypertensive and hypotensive episodes, knowing whether BPV is associated with hematoma progression will better highlight the

Serum alcohol concentration (mg/dL)

admission lactate level (mmol/L)

44 (92) 37 (85) 52 (100 0.14

3.0 (2) 2.7 (1.7) 3.3 (2.5) 0.01

importance of blood pressure management during hospital admission.

The purpose of our study was to investigate the effects of blood pressure variability on hematoma progression during the first 24 h after patient admission to a regional trauma center for traumatic IPH or contusion, as well as the effects of the first 24-h BPV on these patients’ hospital out- comes. We hypothesized that 24 h-BPV from time of admission is asso- ciated with hematoma progression and worse outcomes in patients with traumatic IPH.

  1. Methods
    1. Study design & patient selection

After obtaining institutional review board approval, all adult (>=18 years old) patients with isolated traumatic intracranial hemor- rhage and >= 2 computed tomography (CT) scans who were admitted to an academic regional Level I trauma center between 01/01/2018

and 12/31/2019 were included in this retrospective study. Based on clinical protocol, all patients who arrived at our trauma center were evaluated by the trauma and neurosurgery teams and received appro- priate resuscitation and interventions, including repeat head CT scans 6-24 h after admission.

We excluded patients who presented with acute TBI with isolated subdural, epidural, and/or Subarachnoid hemorrhages in the absence of IPH or contusion, as well as all patients with concomitant injuries. We also excluded all patients from our analysis who did not undergo follow-up Head CT imaging within 24 h after injury, an indicator that any hematoma identified on initial CT scan was not significant enough for further workup. Lastly, we excluded patients whose length of stay in the hospital was <24 h because we were interested in patients’ 24-h BPV.

    1. Independent variables

We obtained Demographic and clinical data from multiple sources including Emergency Department records, transportation team docu- ments, and our institution’s electronic medical records (EMR). All de- mographic and clinical data, as well as any medical interventions received by the patient within the first 24 h of hospital admission, were recorded and can be referenced in Table 1.

Admission INR 1.1 (0.4) 1.1 (0.5) 1.1 (0.3) 0.72

BMI, body mass index; INR, international normalized ratio; MVC, motor vehicle collision; SD, standard deviation.

CT imaging studies obtained as part of the patients’ clinical care were used to calculate hematoma volumes. We calculated hematoma vol- umes in a quantitative manner using the (A x B x C)/2 method [10-13].

    1. Blood pressure variability

All systolic blood pressure measurements during the initial 24 h of hospital admission were obtained from Arterial line readings docu- mented in patient charts. From these values, we identified the highest systolic blood pressure (SBPmax) and the lowest systolic blood pressure (SBPmin), such that SBPmax-min = SBPmax– SBPmin. In addition, we also analyzed systolic blood pressure variation with three components that have been previously studied in patients with spontaneous intraparenchymal hemorrhage [9,14]: standard deviation in SBP (SBPSD), successive variations in SBP (SBPSV), and coefficient variation in SBP (SBPCV).

The SBPSD, SBPSV, and SBPCV were calculated as previously described [14].

    1. Outcome measures

Our primary outcome was the number of patients (percentage) who demonstrated hematoma progression. Cepeda and associates defined a significant hematoma progression as any volume increase >=30% be-

tween the initial and follow-up CT scans [12], using the following equa-

tion:

Hematoma progression 2nd volume-1st volume *100

1/4

1st volume

We also investigated the secondary outcome of hospital mortality.

    1. Data collection and management

Research team members were first trained by the PI and senior re- search team members, including a neurosurgeon, for data collection

and hematoma volume calculations. For training, we used sets of 10 pa- tients’ charts until each research team members’ results of all indepen- dent variables reached 90% agreement with a senior research team member. An investigator also randomly checked 10% of data collection, including the hematoma volumes, for good Interrater agreement of

>90%. To reduce the risk of bias, investigators extracted data in separate sections. Data was extracted to a standardized Microsoft Excel spread- sheet (Microsoft Corp, Washington, USA).

    1. Sample size calculation

We performed a multiple logistic regression to measure the associa- tions between demographic, clinical independent variables, and hema- toma progression. The sample size needed for multiple logistic regression was calculated according to the equation [15]:

Sample size 1/4 number of counts per independent variable

* number of independent variables incidenceofoutcome

Prior studies suggested the need for 5-10 counts per independent variable [16,17], a multivariable logistic regression that supports 10 in- dependent variables, and the incidence of hematoma progression as 30% [2]. We concluded that our multivariable logistic regression would need approximately 300 patients to support up to 10 indepen- dent variables.

    1. Data analysis

We used descriptive analysis with mean [standard deviation (SD)], median [Interquartile Range (IQR)] or percentages to present our demo- graphic and clinical data. Continuous data was analyzed using the stu- dent t-test or the Mann-Whitney U test. Categorical data was compared using a chi-squared test or Fisher-exact test.

We performed a forward stepwise multivariable logistic regression to assess patients’ demographic, clinical independent variables, and di- chotomous outcomes (No Hematoma progression vs. Yes Hematoma progression, dead vs. alive). To optimize the fitness of our multivariable logistic regression, we used the Classification and Regression Tree (CART) model to identify the 10 most important independent variables. This process was shown to significantly improve the predictive perfor- mance of a multivariable regression [18]. The CART algorithm examines each independent variable and its interactions with other variables, then performs a series of splits to maximize the sensitivity and specific- ity of the classification. Through many interactions, the CART model cre- ates a decision tree and assigns the strongest predictor for a given outcome a “relative variable importance” of 100%. Other significant in- dependent variables are assigned values as a percentage of the strongest predictor. We used a 10-fold cross-validation technique for our CART model with 37 predictors (Appendix 1). We defined the minimum number of cases to allow a splitting of 3 (N = 3) and we did not limit the number of maximum tree depth. Once we identified the important variables, we selected the top 10 variables for our multivariable logistic regressions (Appendix 2).

We performed 2 models of multivariable logistic regressions for the primary outcome of Yes Hematoma Progression vs. No Hematoma Pro- gression. The first model contained the values of intracranial opening pressure (IOP) as an independent variable, one of the important vari- ables determined by CART analysis. However, the first model only contained 69 patients with IOP values, therefore we performed a second model that did not contain IOP but included the other 9 important var- iables suggested by CART analysis. This second model included all 354 patients.

We presented the results from our stepwise multivariable logistic regressions as odd ratios (OR), 95% Confidence Interval (95% CI), p-value. We assessed our multivariable logistic regressions with the Hosmer-Lemeshow goodness-of-Fit test, and the Area Under the Re- ceiver Operating Characteristic curve (AUROC). A Hosmer-Lemeshow test’s p-value <0.05 indicates that the logistic regression has a good fit of the independent variables. The AUROC assessed the discriminatory capability of our multivariable logistic regressions. A regression model with an AUROC of 1 showed perfect discriminatory capability, while an AUROC of 0.5 showed poor discrimination between the dichotomous outcomes. Once our multivariable logistic regressions identified the sig- nificant continuous variables, we planned to perform probit logic anal- yses to identify absolute values of these continuous variables and their predictions for patients’ outcomes.

All statistical analyses were performed with Minitab Version 19 (www.minitab.com; Minitab LLC, State College, Pennsylvania, USA). Analyses with two-tailed P-value <0.05 were considered statistically significant.

  1. Results
    1. Patient characteristics

A total of 473 patients were retrospectively reviewed in detail from our database of 1632 TBI patients who were admitted to our quaternary medical center in the 2-year period. From this cohort, we included 354 patients that matched our inclusion criteria after manual chart review (Fig. 1).

The mean age was 56 (SD = 21) years, and 260 (73%) were male and 94 (27%) were female. The two groups’ characteristics were similar ex- cept for one variable: the admission lactate level was 3.3 (SD 2.5) in those with hematoma progression compared to 2.7 (SD 1.7) in those without progression (p = 0.01).

    1. Patient management

The intervention data for these 2 groups were compared in Table 2. There was no significant difference between the management of pa- tients with and without hematoma progression.

Image of Fig. 1

Fig. 1. Patient selection diagram.

CT, computed tomography; IPH, intraparenchymal hemorrhage.

Table 2

Managements of patients with traumatic brain injury.

All patients

No Hematoma Expansion

Hematoma Expansion >=30%

P

N = 354

N = 194

N = 160

Any mechanical ventilation, N (%)

202 (57)

104 (54)

98 (61)

0.32

Any intravenous crystalloids, N (%)

309 (87)

164 (85)

145 (91)

0.19

24 h IVF input (ml), mean (SD)

3600 (300)

1800 (1800)

2000 (1700)

24 h fluid balance

1200 (2100)

1200 (2100)

1200 (2200)

Any anti-seizure medication, N (%)

87 (25)

49 (25)

38 (24)

0.86

Any phenytoin, N (%)

73 (21)

39 (20)

34 (21)

0.86

Any levetiracetam, N (%)

14 (4)

10 (5)

4 (3)

0.72

Any Hyperosmolar therapy, N (%)1

108 (31)

49 (25)

59 (37)

0.07

Type of hyperosmolar therapy – 3% saline, N (%)1

108 (31)

49 (25)

59 (37)

Type of hyperosmolar therapy – mannitol, N (%)1

10 (3)

3 (2)

7 (4)

Any blood products, N (%) [ 1]

107 (30)

50 (26)

57 (36)

0.12

Type of blood products – pRBC, N (%)

64 (18)

29 (15)

35 (22)

0.2

Type of blood products – FFP, N (%)

41 (12)

17 (9)

24 (15)

0.19

Type of blood products – platelet, N (%)

61 (17)

29 (15)

32 (20)

0.35

Any EVD, N (%)

80 (23)

38 (20)

42 (26)

0.31

Opening pressure (cm H20), mean (SD)

15 (8)

14 (7)

16 (8)

0.43

Any craniectomy, N (%)

53 (15)

22 (11)

31 (19)

0.11

IVF, intravenous fluids; pRBC, Packed red blood cells; FFP, fresh frozen plasma; EVD, external ventricular drain.

1 Patients may receive more than one item.

    1. Patients’ blood pressure variability and hospital outcome

All indices of BPV were similar between patients with and without hematoma progression (Table 3). Bivariate analysis revealed a

Table 3

Table 4

Multivariable logistic regressions for association of clinical factors and TBI patients’ pri- mary outcome (hematoma progression >=30%). Top 10 variables that were identified by the Classifications And Regression Tree (CART) model was included in the regression. Only factors with statistically significant association were reported. First multivariable logistic regression contained the independent variable intracranial opening pressure and 9 other important variables. The second model contained 9 other important variables but did not contain intracranial opening pressure so the model could include more patients.

Blood pressure variability and outcomes among patients with traumatic brain injury.

Independent

Model with opening

Model without opening

All

No Yes P

variables

pressure 1

pressure 2

patients

hematoma expansion

hematoma expansion

(N = 69) (N = 354)

OR 95% CI P VIF OR 95% CI P VIF

Total number of BP measurements, median [IQR]

CV

mean (SD)

Repeat hematoma volume (ml), mean (SD)

N = 354 N = 194 N = 160

31 [7-32] 31 [7-40] 34 [23-42] 0.32

SBPmax, mean (SD)

178 (31)

178 (30)

177 (33)

0.72

SBPmin, mean (SD)

91 (28)

93 (30)

89 (26)

0.15

SBPSV, mean (SD)

14 (6)

14 (6)

14 (5)

0.91

SBPSD, mean (SD)

17 (7)

16 (8)

17 (7)

0.78

SBP , mean (SD)

16 (6)

15 (6)

16 (7)

0.17

First hematoma volume (ml),

7 (19)

9 (25)

4 (8)

0.023

8 (18) 7 (15) 10 (20) 0.06

Hematoma progression

SBPCV 1.03 1.02-1.3 0.026 1.6 1.03 0.99-1.07 0.098 1.03

Direct admission 4.8 4-100+ 0.002 1.25 2.3 1.3-3.9 0.002 1.01

GCS, Glasgow Coma Scale score; OR, Odds Ratio; SBPCV, Coefficient variation in systolic blood pressure; SBPSV, successive variation in systolic blood pressure; VIF, Variance Inflation Factor.

1 The Hosmer-Lemeshow test’s p-value = 0.47, AUROC = 0.82.

2 The Hosmer-Lemeshow test’s p-value = 0.15, AUROC = 0.65.

Time to follow-up CT scan (hours), median [IQR]

6 [4-11] 6 [4-16] 5.7

[3.75-8.3]

0.03

significant correlation between Glasgow Coma Scale score and

Hematoma progression 0-10%,

N (%)

Hematoma progression, 11-20%,

N (%)

Hematoma progression, 21-30%,

N (%)

Hematoma progression, >30%,

N (%)

GCS score, median [IQR]

158 (45) 158 (81) NA NA

22 (6) 22 (11) NA NA

14 (4) 14 (7) NA NA

160 (45) 0 160 NA

hematoma volume progression. While both groups shared similar GCS

scores at admission, those who experienced a hematoma progression had a significantly lower GCS score at 24 h post-admission (10 [IQR 7-14] vs 12 [IQR 9-15] p = 0.044) and a significantly lower GCS score

Table 5

Multivariable logistic regressions for association of clinical factors and TBI patients’ mortal-

At admission 7 [3-13] 13 [7-15] 12 [6-14] 0.43

At 24 h 11 [7-14] 12 [9-15] 10 [7-14] 0.044

On hospital day 5 14 [10-15] 14 [10-15] 13 [8-15] 0.003

Disposition, N (%)

Discharge home 91 (26) 51 (26) 40 (25) 0.86

Acute rehabilitation

167 (47)

91 (47)

76 (48)

0.88

Hospital death1

Skilled nursing home

34 (10)

21 (11)

13 (8)

0.47

Best hospital day 5 GCS

0.61

0.53-0.69

0.001

1.0

Hospice/death

52 (10)

26 (13)

26 (16)

0.55

Difference between SBPmax-SBPmin

1.02

1.004-1.03

0.007

1.0

ity. Top 10 variables that were identified by the Classifications And Regression Tree (CART) model was included in the regression. Only factors with statistically significant association are reported.

Independent variables OR 95% CI P VIF

Other 9 (3) 4 (2) 5 (3) 0.99

SBPmax, maximum systolic blood pressure; SBPmin, minimum systolic blood pressure; SBPSV, successive variation in systolic blood pressure; SBPSD, standard deviation of systolic blood pressure; SBPCV, coefficient of variation of systolic blood pressure; CT, computerized tomography; GCS, Glasgow coma scale.

GCS, Glasgow Coma Scale score; OR, Odds Ratio; SBPCV, coefficient variation in systolic blood pressure; SBPmax, highest systolic blood pressure; SBPmin, lowest systolic blood pressure; SBPSV, successive variation in systolic blood pressure; VIF, Variance Inflation Factor.

1 The Hosmer-Lemeshow test’s p-value = 0.13, AUROC = 0.93.

Image of Fig. 2

on hospital day 5 (13 [IQR 8-15] vs 14 [IQR 10-15] p = 0.003), com- pared to patients without hematoma progression (Table 3).

    1. Primary outcomes

We performed 2 multivariable logistic regression analyses for the as- sociation between clinical factors and hematoma progression: model 1 with intracranial opening pressure, and model 2 which did not include intracranial opening pressure (Tables 4 and 5), as explained in our sub- section of statistical analyses. Results from model 1 showed that each 1 mmHg change in SBPCV (OR 1.03, 95% confidence interval (CI) 1.02-1.3, p = 0.026) was associated with a 3% higher likelihood of hematoma progression among patients who had values for intracranial opening pressure. In other words, a 10-mmHg increase of SBPCV in these patients was associated with 34% increased odds of hematoma progression. Model 1 also reported very good discriminatory capability between Yes Hematoma progression and No Hematoma progression with an AUROC of 0.82. However, the associa- tion between SBPCV and hematoma progression was no longer significant in model 2 (OR 1.03, 95% CI 0.99-1.07, p = 0.098). Model 2 also had lower discriminatory capability with an AUROC of 0.65. Both models showed good fit of the data (Hosmer-Lemeshow’s p-values

<0.05) and there was no multicollinearity (VIF was approximately 1

for all included independent variables).

    1. Secondary outcomes

Multivariable logistic regression analysis was also performed for hospital death (Table 5). Two variables were significantly associated with hospital death: patients’ best GCS on hospital day 5 (OR 0.61, 95% CI 0.53-0.69, p = 0.001) and greater SBPmax-min (OR 1.02, 95% CI 1.004-1.03, p = 0.007). Probit logic analyses also identified that an average of 20-mmHg difference between SBPmax and SBPmin over 24 h to be associated with 2% of mortality (Fig. 2A) while a 50- mmHg difference was associated with 5% of mortality. Similarly, 10% of patients who had average 24-h SBPmax at 160 mmHg were pre- dicted to die during hospital stay (Fig. 2B). Additionally, 13% of pa- tients with 24-h SBPmin of 90 mmHg were predicted to die during hospital stay (Fig. 2C).

  1. Discussion

In this study of patients with traumatic intraparenchymal hemor- rhage, we identified several indices of blood pressure variability signifi- cantly associated with patient outcomes. A greater systolic blood pressure coefficient of variation (SBPCV) was found to be associated with a higher likelihood of hematoma progression >=30% among patients requiring neurosurgical intervention. Furthermore, larger

differences between highest and lowest SBP (SBPmax-min) were associated with a greater odds of hospital death.

We found that SBPCV, an indicator of how tightly the blood pressure was controlled, was associated with significant hematoma progression

>30% among patients who had reported intracranial opening

pressures. This suggests that, regardless of absolute blood pressure

values, fluctuations in blood pressure during the first 24 h of hospital admission was associated with hematoma progression in this patient population. The findings from our study indicate that a common mechanism is shared between blood pressure variability and hematoma progression in those with spontaneous ICH vs traumatic ICH. Prior literature proposed that large fluctuations in systolic blood pressure impairs Cerebral autoregulation and that BPV increases oncotic and hydrostatic pressure gradients in the peri-hematomal region that may exacerbate cerebral edema [9,19].

While SBPCV was a significant factor in hematoma progression in our multivariable logistic regression model including intracranial opening pressure, it was not found to be significantly associated with hematoma progression in the model without intracranial opening pressure, nor in our probit logic analysis. However, this model had lower discriminatory capability (Table 4). This strongly suggests that SBPCV is particularly important among patients who require external ventricular drain (EVD) placement for intracranial pressure monitoring. While further studies are needed to confirm our observation, our results show that “tight” control of blood pressure in critically ill patients, especially those requiring surgical intervention, may be beneficial to prevent hematoma progression.

Our finding that the GCS score is significantly associated with hospi- tal mortality is consistent with findings from previous literature. Tuteja et al. showed that there was a significant correlation between triage GCS and in-hospital death in patients with spontaneous ICH, where higher triage GCS scores were correlated with lower risk of in-hospital death [14]. Previous studies among patients with sICH [20,21] or TBI [22] both suggested an association between patients’ GCS on hospital day 5 and their outcome. However, previous studies involving TBI patients only measured the association of TBI patients and the risk of receiving tracheostomy [22]. Our study further elucidated that GCS scores were significantly associated with patients’ mortality in those with traumatic IPH. However, we could not exclude reverse causality as family may opt to withdraw life support for patients with poor GCS on hospital day 5. Nonetheless, this observation suggests that GCS score could be a potential predictor for TBI patients’ outcome, which could be applied clinically.

SBPmax-min, an index of blood pressure variability, was also shown to correlate with hospital mortality. This finding brings new insight regarding the association between hypotension and poor outcomes in patients with TBI. Having low systolic blood pressures not only reduces cerebral perfusion pressure, but also exacerbates blood pressure variability. This suggests that not only should hypotension be avoided in patients with traumatic IPH, but reducing fluctuations be- tween high and low systolic blood pressure may also improve patient outcomes.

Our findings of association between blood pressure variability and TBI patients’ outcomes could not exclude reverse causality between pa- tients’ underlying disease and their blood pressure. Due to severe dis- ease, some patients would develop hypotensive episodes from brain herniation, causing high blood pressure variability. We performed a post hoc analysis involving 69 patients who required EVD (Appendix 3). We observed that the patients who had high opening pressure (>=

20 cm H20) were associated with higher percentages of hypotensive

Fig. 2. Probit logic analysis for absolute values of SBPmax, SBPmin, SBPmax-min and their predictions for mortality.

  1. Absolute values of difference between highest and lowest systolic blood pressure (SBPmax-min) and their predictions for mortality. The x-axis represents the absolute values of SBPmax-min. Y-axis represents the prediction of mortality associated with such values.
  2. Absolute values highest systolic blood pressure (SBPmax) and their predictions for mortality. The x-axis represents the absolute values of SBPmax. Y-axis represents the prediction of mortality associated with such values.
  3. Absolute values of lowest systolic blood pressure (SBPmin) and their predictions for mortality. The x-axis represents the absolute values of SBPmin. Y-axis represents the prediction of mortality associated with such values.

episodes (systolic blood pressure <= 90 mmHg), compared to those who had opening pressure <= 19 cm H2O. However, the difference was not statistically significant, most likely because we did not have enough power to detect the difference. Further investigations with adequate

sample size are needed to provide further information about the association of blood pressure variability and patients’ disease severity.

  1. Limitations

In addition to its retrospective nature, our study has several limita- tions. Most importantly, we were only able to obtain blood pressure values that were recorded in the patients’ charts. We acknowledge that this does not account for the dynamic nature of hemodynamic monitoring and attempted to alleviate this concern by including all SBP values recorded within the first 24 h of each patient’s admission. However, we did not collect the time when each blood pressure was collected so we could not correlate whether the highest or lowest blood pressure measurements corresponded to patients’ clinical course such as signs of brain herniation. Furthermore, smaller hemato- mas were more difficult to calculate because we were limited by the cross sectional views on CT scan, although we utilized quality assur- ance processes to maximize our accuracy. This may account for vari- ability in hematoma volumes, especially in the No Hematoma Progression group which demonstrated smaller volumes on average. This could explain why our CART model indicated that initial hema- toma volume was an important variable associated with hematoma progression, while our multivariable logistic regressions did not find any statistical significance. Lastly, intracranial opening pressure was identified as a relatively important factor by the CART model but was not a significant factor in our multivariable logistic regression, likely due to its small sample size. Additionally, we were not able to compare the intracranial pressure among patients who required EVD, as it was beyond the scope of our study. These values may have provided further information about blood pressure variability and patients’ outcomes.

  1. Conclusions

In patients with traumatic intracranial hemorrhage or contusion, those who required intracranial pressure monitoring and experienced a greater coefficient of variation in systolic blood pressure (SBPCV), were associated with higher odds of developing hematoma progression. Furthermore, the difference between highest and lowest blood pressure (SBPmax-min) was another index of blood pressure variability which was associated with an increased likelihood of hospital death. Our study suggests that, until further investigations are available, clinicians should minimize blood pressure variability within the first 24 h after traumatic IPH to decrease the odds of hematoma progression and improve patient outcomes.

Meetings presented

50th Annual Congress of the 2021 Society of Critical Care Medicine.

February 2020. Virtual Congress.

Disclosures of funding

No funding was provided for this manuscript.

Declaration of Competing Interest

The authors declared no conflict of interest for this manuscript.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2021.12.005.

Appendix 1. All (37) independent variables collected during initial patient chart review and analyzed using the Classification and Regression Tree (CART) predictive model

Unlabelled image

Appendix 2. Classification And Regression Tree (CART) model to show relatively important factors associated with hematoma progression, and hospital mortality

Appendix 2A. Factors and their relative variable importance values associated with hematoma progression. The top 10 variables were included in the multivariable logistic regression for hematoma progression

Unlabelled image

Appendix 2B. Factors and their relative variable importance values associated with hospital mortality. The top 10 variables were included in the multivar- iable logistic regressions for patients’ outcomes

Unlabelled image

BAL, blood alcohol level; BMI, body mass index; BP, blood pressure; GCS, Glasgow Coma Scale; IVF, intravenous fluids; INR, international normalized ratio; MVC, motor vehicle collision; PE, pulmonary embo- lism; pRBC, packed red blood cells; SBPmax, maximum systolic blood pressure; SBPmin, minimum systolic blood pressure; SBPSV, successive variation in systolic blood pressure; SBPSD, standard deviation of systolic blood pressure; SBPCV, coefficient of variation of systolic blood pressure.

Appendix 3. Subgroup analysis of blood pressure variability and out- comes among patients with TBI requiring EVD

All Patients

High opening pressure (>=20

Normal opening

P

(n=69)

mmHg) (n=22)

pressure (<=19

mmHg)

(n=47)

186

SBPmax, mean (SD)

(30)

173 (20)

191 (32)

0.005

SBPmin, mean (SD)

79 (27)

81 (35)

78 (24)

0.77

SBPSV, mean (SD)

16 (6)

14 (4)

17 (6)

0.007

SBPSD, mean (SD)

19 (7)

17 (6)

19 (7)

0.17

SBPCV, mean (SD)

19 (5)

17 (5)

19 (5)

0.18

Number of patients with any

episode of hypertension

(SBP >=160 mmHg), N (%) 60 (87)

Number of patients with any

14 (64)

46 (98)

0.19

episode of hypotension (SBP 69

<=90 mmHg), N (%) (100)

22 (100)

47 (100)

1

readings/total BP readings in 31

24 hours (%) [IQR] [16-43]

27 [13-36]

31 [20-45]

0.36

Number of hypotensive BP

readings/total BP readings in 3.2

24 hours (%) [IQR] [0-8.3]

3.5 [0-8.1]

3.1 [0-8]

0.97

Initial hematoma volume (ml),

mean (SD) 17 (28)

26 (34)

14 (25)

0.13

Hematoma expansion, >30%, N

(%) 50 (67)

15 (75)

36 (65)

0.43

GCS score, median [IQR]

At admission 6 [6]

7 [7]

6 [5]

0.84

At 24 hour 7 [4]

7 [3]

7 [4]

0.37

On hospital day 5 8 [5]

8 [7]

8 [5]

0.26

Disposition, N (%)

Discharge home 2 (3)

0 (0)

2 (4)

1

Acute rehabilitation 42 (56)

9 (45)

33 (60)

0.25

Skilled Nursing home 4 (5)

0 (0)

4 (7)

0.22

Hospice/Death 22 (29)

10 (50)

12 (22)

0.018

Other 2 (3)

0

2 (4)

0.023

Number of hypertensive BP

CT, computerized tomography; EVD, external ventricular drain; GCS, Glasgow coma scale; IQR, interquartile range; ml, millimeter; mm Hg, millimeter of mercury; SBPmax, maximum systolic blood pressure; SBPmin, minimum systolic blood pressure; SBPSV, successive variation in systolic blood pressure; SBPSD, standard deviation of systolic blood pressure; SBPCV, coefficient of variation of systolic blood pressure; SD, standard deviation; TBI, traumatic brain injury.

References

  1. Orito K, Hirohata M, Nakamura Y, Yamamoto M, Takeshige N, Aoki T, et al. Predictive value of leakAge SIgns for pure brain contusional hematoma expansion. J Neurotrauma. 2018;35(5):760-6.
  2. Carnevale J, Segar D, Powers A, Shah M, Doberstein C, Drapcho B, et al. Blossoming contusions: identifying factors contributing to the expansion of traumatic intracere- bral hemorrhage. J Neurosurg. 2018;129(5):1305-16.
  3. Kurland D, Hong C, Aarabi B, Gerzanich V, Simard M. Hemorrhagic progression of a contusion after traumatic brain injury: a review. J Neurotrauma. 2012;29(1):19-31.
  4. Galgano M, Toshkezi G, Qui X, Russell T, Chin L, Zhao L-R. Traumatic brain injury: current treatment strategies and future endeavors. Cell Transplant. 2017;26(7): 1118-30.
  5. Quinones-Ossa G, Durango-Espinosa Y, Padilla-Zambrano H, Moscote-Salazar L, Keni R, Deora H, et al. The puzzle of spontaneous versus traumatic intracranial hemor- rhages. Egypt J Neurosurg. 2020.;35(13).
  6. Jha R, Bell J, Citerio G, Hemphill J, Kimberly W, Narayan R, et al. Role of sulfonylurea receptor 1 and glibenclamide in traumatic brain injury: a review of the evidence. Int J Mol Sci. 2020;21(2):409.
  7. Duan Z, Tao L, Tang T, Fu C, Liu J, Chen B, et al. Effect of blood pressure variability on early Neurological deterioration in single small subcortical infarction with parental arterial disease. eNeurologicalSci. 2017;8:22-7.
  8. Jantzen JP. Prevention and treatment of Intracranial hypertension. Best Pract Res Clin

Anaesthesiol. 2007;21(4):517-38.

  1. Manning L, Hirakawa Y, Arima H, Wang X, Chalmers J, Wang J, et al. Blood pressure variability and outcome after acute intracerebral haemorrhage: a post-hoc analysis of INTERACT2, a randomised controlled trial. Lancet Neurol. 2014;13(4):364-73.
  2. White C, Griffith S, Caron JL. Early progression of traumatic cerebral contusions: characterization and risk factors. J Trauma Inj Infect Crit Care. 2009;67(3):508-15.
  3. Chang E, Meeker M, Holland M. Acute traumatic intraparenchymal hemorrhage: risk factors for progression in the early post-injury period. Neurosurgery. 2006;58(4): 647-56.
  4. Cepeda S, Gomez P, Castano-Leon AM, Martinez-Perez R, Munarriz P, Lagares A. Traumatic intracerebral hemorrhage: risk factors associated with progression. J Neurotrauma. 2015;32(16):1246-53.
  5. Alahmadi H, Vachhrajani S, Cusimano M. The natural history of brain contusion: an analysis of radiological and clinical progression. J Neurosurg. 2010;112(5):1139-45.
  6. Tuteja G, Uppal A, Strong J, Nguyen T, Pope K, Jenkins R, et al. Interventions affecting blood pressure variability and outcomes after intubating patients with spontaneous intracranial hemorrhage author links open overlay panel. Am J Emerg. 2019;37(9): 1665-71.
  7. Peduzzi P, Concato J, Kemper E, Holford T, Feinstein A. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol. 1996;49(12):1373-9.
  8. Vittinghoff E, McCulloch C. Relaxing the rule of ten events per variable in logistic and Cox regression. Am J Epidemiol. 2007;165(6):710-800.
  9. Smeden M, Groot J, Moons K, Collins G, Altman D, Eijkemans M, et al. No rationale for 1 variable per 10 events criterion for binary logistic regression analysis. BMS Med Res Methodol. 2006;16(1):163.
  10. Ma X. Introduction, advantages of CART. Using Classification and Regression Trees: A Practical Primer. Information Age Publishing, INC; 2018 www.infoagepub.com.
  11. Sykora M, Diedler J, Rupp A, Turcani P, Rocco A, Steiner T. Impaired baroreflex sen- sitivity predicts outcome of acute intracerebral hemorrhage. Crit Care Med. 2008;36 (11):3074-9.
  12. Cooper D, Rosenfeld J, Murray L, Arabi Y, Davies A, D’Urso P, et al. Decompressive craniectomy in diffuse traumatic brain injury. N Engl J Med. 2011;364:1493-502.
  13. Maas M, Francis B, Sangha R, Lizza B, Liotta E, Naidech A. Refining prognosis for in- tracerebral hemorrhage by early reassessment. Cerebrovasc Dis. 2017;53(3-4): 110-6.
  14. Jenkins R, Morris N, Haac B, Besien R, Stein D, Badjatia N, et al. outcome predictors for severely brain-injured patients directly admitted or transferred from emergency departments to a trauma center. World J Emerg Med. 2020;11(2):120-1.