Relationship between Measures of Cerebrovascular Reactivity and Intracranial Lesion Progression in Acute Traumatic Brain Injury Patients: A CENTER-TBI Study

Abstract

Failure of cerebral autoregulation has been linked to unfavorable outcome after traumatic brain injury (TBI). Preliminary evidence from a small, retrospective, single-center analysis suggests that autoregulatory dysfunction may be associated with traumatic lesion expansion, particularly for pericontusional edema. The goal of this study was to further explore these associations using prospective, multi-center data from the Collaborative European Neurotrauma Effectiveness Research in TBI (CENTER-TBI) and to further explore the relationship between autoregulatory failure, lesion progression, and patient outcome. A total of 88 subjects from the CENTER-TBI High Resolution ICU Sub-Study cohort were included. All patients had an admission computed tomography (CT) scan and early repeat scan available, as well as high-frequency neurophysiological recordings covering the between-scan interval. Using a novel, semiautomated approach at lesion segmentation, we calculated absolute changes in volume of contusion core, pericontusional edema, and extra-axial hemorrhage between the imaging studies. We then evaluated associations between cerebrovascular reactivity metrics and radiological lesion progression using mixed-model regression. Analyses were adjusted for baseline covariates and non-neurophysiological factors associated with lesion growth using multi-variate methods. Impairment in cerebrovascular reactivity was significantly associated with progression of pericontusional edema and, to a lesser degree, intraparenchymal hemorrhage. In contrast, there were no significant associations with extra-axial hemorrhage. The strongest relationships were observed between RAC-based metrics and edema formation. Pulse amplitude index showed weaker, but consistent, associations with contusion growth. Cerebrovascular reactivity metrics remained strongly associated with lesion progression after taking into account contributions from non-neurophysiological factors and mean cerebral perfusion pressure. Total hemorrhagic core and edema volumes on repeat CT were significantly larger in patients who were deceased at 6 months, and the amount of edema was greater in patients with an unfavourable outcome (Glasgow Outcome Scale-Extended 1–4). Our study suggests associations between autoregulatory failure, traumatic edema progression, and poor outcome. This is in keeping with findings from a single-center retrospective analysis, providing multi-center prospective data to support those results.

Introduction

Failure of cerebral autoregulation has been linked to unfavorable patient outcomes in acute traumatic brain injury (TBI).^1,2 Various continuous indices of cerebrovascular reactivity have been used as surrogate measures of autoregulation in patients requiring neurocritical care. The most widely described indices are derived from correlating slow-wave vasogenic fluctuations in simultaneous measurements of mean arterial pressure (MAP) or cerebral perfusion pressure (CPP) with intracranial pressure (ICP) waveforms. A recent study demonstrated the prognostic significance of three of these ICP-derived metrics—PRx (correlation between ICP and MAP), PAx (correlation between MAP and pulse amplitude of ICP [AMP]), and RAC (correlation between AMP and CPP)—with respect to 6- to 12-month global patient outcome in patients enrolled in the CENTER-TBI high-resolution monitoring cohort.³ However, the mechanisms through which such abnormal vascular physiology contribute to poorer outcomes remain unclear.

Possible pathophysiological phenomena linking autoregulatory failure to outcome include ischemia, hyperemia, and intracranial hypertension.⁴ One candidate pathway which has received little intention in this context is delayed progression of intracranial lesions. A retrospective exploratory analysis performed by our group suggested associations between impaired cerebrovascular reactivity and expansion of pericontusional edema in a cohort of 50 acute TBI patients.⁵ A potential explanation for these findings may be that large fluctuations in blood flow resulting from perturbations in autoregulatory capacity overwhelm fragile microvessels located in the periphery of the hemorrhagic contusion core. This is in line with experimental animal data showing that the energy transferred during TBI can trigger progressive microvascular dysfunction in brain tissue adjacent to the primary injury site in the absence of overt bleeding or vessel fracture.^6,7 Better characterizing the relationship between cerebrovascular derangements and radiological outcomes in acute TBI could allow for more individualized management strategies by allowing for identification of patients at high risk of lesion expansion, thus informing decisions regarding follow-up imaging and the need for more aggressive medical or surgical therapies.

The primary aim of this study was to confirm the univariate associations between autoregulatory failure and traumatic edema progression our group has previously reported, this time using data obtained from the larger, prospective, multi-center, high-resolution cohort of CENTER-TBI. Our secondary aims were to 1) determine which cerebrovascular reactivity metrics show the strongest associations with lesion expansion after adjusting for non-neurophysiological predictors of radiological progression in a multi-variate analysis and 2) explore relationships with long-term patient outcome.

Methods

Patient population

All patients enrolled in the CENTER-TBI high-resolution ICU cohort were screened for inclusion in this study. These patients were recruited between January 2015 and December 2017 in a total of 21 centers across Europe. All patients included in this cohort were admitted to an intensive care unit (ICU) with a diagnosis of acute TBI and had high-frequency physiological and intracranial recordings performed during their ICU stay. Decisions regarding the need for invasive ICP monitoring and ICP-directed therapies were made in accordance with the Brain Trauma Foundation guidelines.⁸ An additional inclusion criterion specific to our analysis was the availability of an admission computed tomography (CT) scan and repeat scan performed during the acute period, defined as within 7 days of injury. Patients who underwent craniotomy for hemorrhage evacuation or a decompressive craniectomy were excluded unless two CT scans had been performed before surgical decompression. This decision was based on our inability to perform reliable serial lesion measurements in this subgroup and considering evidence that decompressive craniectomy contributes to iatrogenic expansion of intraparenchymal lesions, introducing a source of confounding bias.^9,10

Ethics approval

Data collected as part of CENTER-TBI were subject to national and local ethics review processes. UK ethics approval is outlined in IRAS No: 150943; REC 14/SC/1370. In addition, the CENTER-TBI study (EC grant 602150) was conducted in accordance with all relevant laws of the European Union, if directly applicable, and all relevant laws of the country where the recruiting sites were located, including, but not limited to, the relevant privacy and data protection laws and regulations, the relevant laws and regulations on the use of human materials, and all relevant guidance relating to clinical studies, including, but not limited to, the ICH Harmonised Tripartite Guideline for Good Clinical Practice (CPMP/ICH/135/95) and the World Medical Association Declaration of Helsinki entitled “Ethical Principles for Medical Research Involving Human Subjects.” Informed consent by the patients and/or the legal representative/next of kin was obtained, accordingly to the local legislations, for all patients recruited in the Core Dataset of CENTER-TBI and documented in the electronic case report form.

Data collection

Patient demographics and baseline clinical variables, including age, sex, injury mechanism, and admission laboratory values, Glasgow Coma Scale (GCS), and pupillary responses, were prospectively recorded at the treatment site. A first CT scan was performed on the day of admission, and at least one repeat imaging study was obtained in the first week after injury, with the specific timing and frequency of follow-ups dictated by clinical indication. Defaced imaging data files were made available to investigators for further analysis. High-frequency neurophysiological monitoring was started after admission to the ICU and maintained throughout the ICU stay, with the goal of initiating recordings within 24 h of injury. Global patient functional outcome, including Extended Glasgow Outcome Scale (GOSE) scores, was obtained 6 months post-injury by in-person interview or postal questionnaires. Data were extracted from version 2.0 of the centralized core CENTER-TBI registry by the Neurobot platform.

Neurophysiological signal acquisition and processing

Invasive arterial blood pressure readings were obtained in all patients. ICP was measured by an intraparenchymal probe (Codman ICP MicroSensor; Codman & Shurtleff Inc., Raynham, MA or Camino ICP Monitor; Integra Life Sciences, Plainsboro, NJ). Signals were recorded using digital data transfer or digitized by an A/D converter (DT9800 series; Data Translation, Marlboro, MA) and sampled at a frequency of 100 Hz or higher using ICM+ software (Cambridge Enterprise Ltd, Cambridge, UK; https://icmplus.neurosurg.cam.ac.uk) or Moberg CNS Monitor (Moberg Research Inc., Ambler, PA). Signal artefacts were removed through a combination of manual curation by experts and automated methods. Flat lines, data related to monitor disconnection, mislabeled data (e.g., period of electrocardiography data found in ICP signal), and non-sensical data were manually removed by experts. Automated filters were subsequently applied to exclude values clearly outside the modality-specific physiological ranges not removed during the manual curation step. Processing of the acquired signal was then performed using ICM+ software as previously described.^3,11 Continuous cerebrovascular reactivity indices were derived by calculating the moving correlation coefficient between slow-wave fluctuations in ICP and MAP (PRx), pulse amplitude of ICP and MAP (PAx), and pulse amplitude of ICP and CPP (RAC). These three indices range from −1.0, which indicates intact autoregulation, and +1.0, which typically corresponds to severely impaired autoregulation. Minute-by-minute data were obtained for each patient for the entire recording period, with a median total duration of 135 h (interquartile range [IQR], 92–184).

Semiautomated volumetric lesion analyses

All CT scans performed within 7 days of injury for each participant were downloaded from the CENTER-TBI core database and automatically segmented using an adaptation of DeepMedic,¹² a three-dimensional convolutional neural network (CNN). This CNN was trained and validated on a set of 98 CT scans derived from a separate cohort of 27 TBI patients and consists of three parallel pathways processing image patches at different resolution. This step yielded automated lesion segmentations divided into three lesion types: extra-axial hemorrhage (EAH), intraparenchymal hemorrhage (IPH), and perilesional edema (Fig. 1). Each of these automated predictions was then visually inspected by a clinician expert, and any false positive or false negative was manually corrected using an open-source segmentation software (ITK-snap^©, version 3.8.0-beta).¹³

FIG. 1.

Semiautomated lesion segmentations. (A) Original image. (B) Automatic lesion prediction. (C) Semiautomated lesion map after manual corrections by expert. Green label: extra-axial hemorrhage. Red label: intraparenchymal hemorrhage (contusion core). Blue label: pericontusional edema. Color image is available online.

Final lesion volumes (milliliters [mL] of blood or edema) were extracted from these corrected lesion maps. We calculated intra- and inter-rater reliability (IRR) using the intraclass coefficient calculated on a random subset of 25 scans segmented by a second (inter-RR) or the same (intra-RR) expert. We applied corrections for rater bias and rater-scan interactions. We achieved satisfactory agreement with inter-rater intraclass correlation coefficients (ICCs) of 0.98, 0.90, 0.98, and 0.92 for IPH, edema, EAH, and intraventricular hemorrhage, respectively, and intra-rater ICCs of 0.98, 0.97, 0.98, and 0.96, respectively.

The corrected volumes were also projected to an atlas template using affine registration methods to obtain their specific neuroanatomical correlates. We used a customized version of the multi-atlas label propagation with expectation-maximization–based refinement method with alignment to the patients' images in native space.¹⁴ This parcellated the brain into 134 regions for gray matter regions with propagation to white matter structures using the technique described by Schiffer and colleagues.^15,16 For the purpose of this study, these localized segmentation maps were collapsed into larger groupings to classify intraparenchymal lesions as either deep (brainstem, thalami, basal ganglia, or cerebellum) or superficial (frontal, temporal, insular, parietal, or occipital lobes).

Outcome measures

The primary outcome of interest was the relationship between autoregulatory status and radiological lesion progression in the acute TBI phase. For each lesion subtype, we calculated a continuous measure of lesion progression by subtracting the absolute lesion volume on the initial scan from the repeat scan (represented as “delta” volumes [Δ], measured in mL). For each patient, the early repeat CT scan with the highest total lesion load was included in the final analyses in an attempt to capture peak hemorrhage and edema formation in the acute period. We adjusted for timing to first and repeat CT scan in the multi-variate models described below to control for the different time points between participants.

We also calculated mean PRx, PAx, and RAC values for each patient for the period between the initial and repeat scan, as well as the percentage of time spent above previously identified clinical threshold for each index (PRx >0, PAx >0, and RAC >−0.10) and mean hourly dose above these same thresholds. These specific thresholds were chosen based on the fact that they have shown similar associations with functional outcome in this patient population compared to the more conservative values of 0.25 and 0.35 for PRx and PAx and −0.05 for RAC.³

Functional outcomes were mortality (dichotomized into alive, GOSE 2–8; dead, GOSE 1) and global patient outcome at 6 months (dichotomized into favorable: GOSE 5–8 and unfavourable outcome: GOSE 1–4).

Statistical analysis

Data were analyzed using R (version 3.5.3; https://www.R-project.org/) and JMP Pro software (1989–2019, version 14; SAS Institute Inc., Cary, NC). Normality of continuous variables was assessed using normal quantile plots and the Shapiro-Wilk test. The alpha was set at 0.05 for significance, with adjustment for multiple comparisons within individual analyses using the Holm-Bonferroni method.

Descriptive statistics are presented as mean ± standard deviation or median (IQR) for continuous variables and proportion of the total sample for categorical variables. Univariate associations between physiological variables, and lesion progression data were assessed using a mixed-effect regression model to account for repeated within-subject measurements at different time points. We opted to analyze absolute, rather than relative, changes in lesion volumes because the latter tend to accentuate expansion rates for small lesions, which may not be as clinically significant.³⁵ In a separate univariate analysis, we looked at the non-neurophysiological parameters which have previously shown the strongest associations with traumatic lesion progression to evaluate their predicted value in our sample: initial lesion volume, timing of initial and repeat CT relative to injury, and presence of a coagulopathy on admission or between the two imaging studies.^17–25 We also included mean CPP as a potential confounding factor, based on the concern that clinicians may tend to target a higher CPP in patients with more severe intracranial injuries—which could potentially contribute to hemorrhage or edema expansion.^26,27

Independent variables showing significant associations in the univariate analysis after correcting for multiple comparisons were entered in a multi-variate mixed-effect model. The goal was to determine whether cerebrovascular reactivity metrics added predictive value over previously identified risk factors. Multi-variate regressions were adjusted for age, sex, and baseline GCS score, because these covariates have shown strong associations with a range of outcomes in TBI (see Iaccarino, prognostic paper model). The models were successively reduced and compared using the corrected Akaike information criterion (AICc) score. To account for the skewed nature of some of the lesion progression distributions (small proportion of patients with large delta volumes), we also performed a sensitivity analysis in which the regression coefficients were bootstrapped using non-parametric techniques (n = 1000 repetitions). Difference in lesion volumes between dichotomized outcome groups were compared using the Mann-Whitney U test. Impact of lesion burden on mortality was assessed using univariate logistic regression.

Results

Patient characteristics

One hundred twenty-two patients from the high-resolution ICU cohort had at least two CT scans available within the first week of injury and met all other inclusion criteria. During curation of the intracranial recording data, an additional 34 patients had to be excluded because of the presence of external ventricular drain (EVD) creating excessive artefact in the high-frequency signals (a reliable ICP can only be measured when the drainage system is closed causing frequent interruptions in the recordings on top of artefact introduced by drain flushes), leaving a final sample size of 88 participants. Patient demographics and baseline clinical characteristics for this sample are reported in Table 1.

Table 1.

Patient Characteristics

Characteristic	No. of patients (%)
Total patients	88
Mean age, years	43 ± 18
Female sex	14 (16)
Mechanism of injury
High velocity	38 (44)
Falls	27 (31)
Other blunt mechanism	23 (25)
Mean total injury severity (ISS)	40 ± 15
Clinical severity (GCS)
Mild (13–15)	10 (11)
Moderate (9–12)	16 (18)
Severe (3–8)	58 (66)
N/A	4 (5)
Pupils (uni- or bilateral unreactive)	21 (7)
Secondary insult
Hypotension	13 (15)
Hypoxia	14 (16)
Coagulopathy
INR >1.2	8 (9)
aPTT >35 sec	14 (16)
Platelets <100	4 (5)
Fibrinogen <150	8 (9)
Any	21 (23)
Pre-injury antithrombotic use	3 (3)
Global functional outcome
6-month, no. favorable outcome	37 (42)
6-month, alive	66 (75)

aPTT, activated partial thromboplastin time; GCS, Glasgow Coma Scale; ISS, Injury Severity Score; INR, international normalized ratio; N/A, not applicable.

Imaging findings

Initial CT scans were obtained within 3.3 ± 4.2 h of injury. Median Marshall and Rotterdam scores were 2 (IQR, 2–6) and 3 (IQR, 2–4), respectively. Midline shift was present on 10 of the initial imaging studies (11%), and basal cistern compression was present on 28 (32%). Distribution of contusion sizes (intraparenchymal core plus edema) and degree of progression for each of the lesion subtypes are shown in Figure 2. Average timing for the repeat scan was 40.0 ± 35.1 h from injury (Fig. 3).

FIG. 2.

Imaging findings. (A) Distribution of contusion volumes (hemorrhagic core + surrounding edema) on initial CT. (B) Median absolute lesion volumes on initial and repeat scan. CT, computed tomography.

FIG. 3.

Timing of initial and repeat. Boxplot showing median and mean (diamond) timing in terms of hours from injury for initial and repeat CT scan. CT, computed tomography.

Univariate associations with traumatic lesion growth

Neurophysiological characteristics for the entire recording period are summarized in Table 2. Impairment in cerebrovascular reactivity was significantly associated with progression of pericontusional edema and, to a lesser degree, IPH. The strongest relationships were observed between RAC-based metrics and edema formation. PAx showed weaker, but consistent, associations with hemorrhagic contusion core expansion and edema progression. In contrast, progression of EAH was only weakly and negatively associated with percentage of time spent with RAC >−0.10, but this did not remain significant after correcting for multiple comparisons. There were also no significant relationships with PRx after adjusting for multiple testing. p values for each of these univariate associations are displayed in Table 3, and corresponding scatterplot diagrams are available as Supplementary Figure S1 to better illustrate dose-response relationships.

Table 2.

Neurophysiological Characteristics

Characteristic	Median (IQR)
Duration of intracranial recordings (h)	135 (92–184)
ICP (mm Hg)	12.5 (9.5–15.0)
CPP (mm Hg)	71.5 (64.4–76.8)
% time with ICP >22 mm Hg	3.3 (0.8–7.9)
Mean PRx	–0.01 (−0.11 to 0.10)
% time PRx positive	47.0 (35.3–62.1)
Mean hourly dose of PRx >0	7.7 (5.3–10.9)
Mean PAx	–0.09 (−0.22 to 0.09)
% time PAx positive	37.8 (25.1–61.3)
Mean hourly dose of PRx >0	5.3 (3.6–10.2)
Mean RAC	–0.41 (−0.57 to −0.19)
% time RAC >−0.10	20.2 (7.4–35.7)
Mean hourly dose of RAC >−0.10	2.6 (1.2–5.6)

CPP, cerebral perfusion pressure; ICP, intracranial pressure; PAx, pulse amplitude index (correlation between pulse amplitude of intracranial pressure and mean arterial pressure); PRx, pressure reactivity index (correlation between intracranial pressure and mean arterial pressure); RAC, correlation between pulse amplitude of intracranial pressure and cerebral perfusion pressure.

Table 3.

Univariate Associations between Cerebrovascular Reactivity Measures and Radiological Progression

	Δ Core	Δ Edema	Δ Extra-axial
Measures and progression	p value	p value	p value
Mean PRx	0.317	0.290	0.356
Mean PAx	0.009(+)	0.0076(+)	0.207
Mean RAC	0.064	<0.001(+)	0.156
% time spent PRx positive	0.041	0.146	0.344
% time spent PAx positive	0.002(+)	0.004(+)	0.283
% time spent RAC >−0.10	0.024(+)	<0.001(+)	0.041(–)
Mean hourly dose PRx positive	0.596	0.476	0.565
Mean hourly dose PAx positive	0.019(+)	0.030(+)	0.304
Mean hourly dose RAC >−0.10	0.139	0.006(+)	0.172

Significant p values are italicized (α = 0.05), and the direction of the relationship between the dependent and independent variable is indicated in parentheses.

Δ Core, absolute difference in total volume (mL) of hemorrhagic contusion core between initial and repeat scan; Δ Edema, absolute difference in volume (mL) for pericontusional edema between initial and repeat scan; Δ Extra-axial, absolute difference in total volume (mL) of extra-axial hemorrhage between initial and repeat scan; PAx, pulse amplitude index (correlation between pulse amplitude of intracranial pressure and mean arterial pressure); PRx, pressure reactivity index (correlation between intracranial pressure and mean arterial pressure); RAC, correlation between pulse amplitude of intracranial pressure and cerebral perfusion pressure.

Given the lack of clear associations with EAH, the rest of the analysis was focused on distinguishing factors preferentially contributing to progression of contusional edema versus the hemorrhagic contusion core. With respect to non-neurophysiological factors, lesion volume on the initial CT was the strongest predictor of hemorrhagic core expansion, whereas time to repeat CT was most significantly associated of edema development (Table 4). Mean CPP during the recording period was weakly associated with core, but not edema, progression. Time to repeat CT was not significantly associated with progression of contusion core, but was included in the multi-variate analyses given that it met our pre-specified threshold of p < 0.20 (p = 0.14). Time to initial CT and presence of a documented coagulopathy on admission or during the between-scan interval (broadly defined as an international normalized ratio [INR] >1.2 or activated partial thromboplastin time [aPTT] >35 sec or platelets <100/nL or fibrinogen <150 mg/dL) did not relate to intraparenchymal lesion growth in our sample. No new patterns emerged after categorizing the lesion volumes into “deep” versus “superficial” (data not shown but available on request).

Table 4.

Univariate Associations for Factors Previously Associated with Lesion Growth

	Δ Core	Δ Edema
Factor	p value	p value
Initial lesion volume	0.001(+)	0.256
Time to initial CT	0.971	0.297
Time to repeat CT	0.14	<0.001(+)
Coagulopathy on admission	0.744	0.517
Coagulopathy between CTs	0.578	0.677
Mean CPP	0.035(+)	0.097

Significant p values are italicized (α = 0.05), and the direction of relationship between the dependent and independent variable is indicated in parentheses.

Δ Core, absolute difference in total volume (mL) of hemorrhagic contusion core between initial and repeat scan; Δ Edema, absolute difference in volume (mL) for pericontusional edema between initial and repeat scan; CPP, cerebral perfusion pressure; CT, computed tomography.

Multi-variate analyses

Tables 5 and 6 present the results of our multi-variable models for core and edema growth. Cerebrovascular reactivity metrics remained strongly predictive of lesion progression after taking into account contributions from non-neurophysiological factors, mean CPP, and after adjusting for age, sex, and baseline GCS. The model that best explained core expansion consisted of initial lesion volume, percentage of the recording time spent with a positive PAx index and mean CPP. Amount of edema formation was best predicted by time from injury to the repeat CT scan and percentage of time spent with an RAC index value of >−0.10. Although the regression parameter estimate for mean CPP was not significant (p = 0.06), including it in the final model slightly improved the fit for our data (delta AICc −1.3; models unadjusted for CPP and corresponding AICcs available as Supplementary Tables S2 and S3). Characteristics of the excluded patients are reported in Supplementary Table S1. After bootstrapping the confidence intervals for each regression coefficient in a sensitivity analysis, all the predictors remained significant except for initial lesion volume in the contusion core model.

Table 5.

Multi-Variate Model of Contusion Core Progression^*

	Δ Core
Progression	p value	Parameter mean [95% CI]
Initial core lesion volume	0.004	0.253 [0.084–0.422]
Mean CPP	0.008	0.150 [0.040–0.259]
% time PAx positive	0.046	0.067 [0.001–0.130]

Significant p values and confidence intervals (CI) are italicized (α = 0.05).

^*Adjusted for age and baseline Glasgow Coma Scale.

CPP, cerebral perfusion pressure; PAx, pulse amplitude index.

Table 6.

Multi-Variate Model of Pericontusional Edema Progression^*

	Δ Edema
Progression	p value	Parameter mean [95% CI]	[Bootstrap 95% CI]
Time to repeat CT	<0.001	0.130 [0.064–0.195]	0.128 [0.049–0.225]
% time RAC >−0.10	<0.001	0.221 [0.104–0.338]	0.217 [0.066–0.364]
Mean CPP	0.061	0.189 [−0.009 to 0.389]	0.175 [−0.037 to 0.395]

Significant p values and confidence intervals (CI) are italicized (α = 0.05).

^*Adjusted for age and baseline Glasgow Coma Scale.

CT, computed tomography; CPP, cerebral perfusion pressure.

Relationship between acute lesion burden and functional outcome

Total hemorrhagic core and edema volumes on repeat CT were significantly larger in patients who were deceased at 6 months (see Table 7). Amount of edema was also greater in patients with an unfavorable outcome (GOSE 1–4), but the amount of core did not significantly differ between the dichotomized outcome groups. Degree of progression between the two CTs (Δ) was consistently higher in patients with a poor outcome, but this did not reach statistical significance. Each milliliter increase in volume of core and edema measured on the repeat scan increased the odds of mortality at 6 months by 6.6% and 4.0%, respectively (see Table 8). Six patients had missing GOSE information and were excluded from these comparisons. Characteristics of the excluded patients are reported in the Supplementary Materials. Relationships between the different cerebrovascular reactivity indices and functional outcome in the CENTER-TBI high-resolution ICU cohort have been previously described and were not re-evaluated in this study.³

Table 7.

Lesion Progression Volumes for Dichotomized Outcome Groups

	Alive (n = 66)	Dead (n = 16)	p value
Δ Core (mL)	0.3 (0.0– to 3.2)	2.7 (0.2–7.1)	0.104
Δ Edema (mL)	1.0 (0.0– to 9.2)	8.3 (−0.3 to 17.8)	0.242
Total core on repeat CT (mL)	1.9 (0.1– to 6.8)	7.4 (1.1–25.9)	0.014
Total edema on repeat CT (mL)	3.8 (0.2– to 16.8)	19.8 (7.5–51.3)	0.006

	Favorable (n = 37)	Unfavorable (n = 45)	p value
Δ Core (mL)	0.3 (0.0– to 4.6)	1.2 (0.0–4.6)	0.605
Δ Edema (mL)	1.1 (0.0– to 9.0)	3.5 (−0.1 to 13.9)	0.827
Total core on repeat CT (mL)	1.5 (0.1– to 6.0)	3.7 (0.2–10.9)	0.16
Total edema on repeat CT (mL)	2.5 (0.2– to 13.7)	9.3 (0.9–30.1)	0.04

Volumes presented as median (interquartile range). Significant p values are italicized.

Δ Core, absolute difference in total volume (mL) of hemorrhagic contusion core between initial and repeat scan; Δ Edema, absolute difference in volume (mL) for pericontusional edema between initial and repeat scan; CPP, cerebral perfusion pressure; CT, computed tomography.

Table 8.

Univariate Logistic Regression of Lesion Volume versus Mortality at 6 Months

	Unit odds ratio for mortality	[95% CI]	p value
Total core on repeat CT (mL)	1.066	[1.02–1.12]	0.002
Total edema on repeat CT (mL)	1.04	[1.01–1.07]	0.002

Significant p values are italicized.

CT, computed tomography; CI, confidence interval.

Discussion

Using prospectively collected data from the CENTER-TBI high-resolution cohort, we evaluated the relationship between autoregulatory failure and traumatic lesion progression. We observed strong associations between impaired cerebrovascular reactivity and edema formation, weaker associations with expansion of the hemorrhagic contusion core, and little or no relationship with changes in EAH. These findings are in keeping with our results from an exploratory analysis on a local retrospective cohort of acute TBI patients.⁵ In the present study, autoregulatory dysfunction remained predictive of lesion growth after adjusting for previously identified risk factors and potential confounders in a multi-variate model.

There has been significant interest in identifying the factors which best predict the evolution of intracranial hemorrhage in acute TBI given that it represents a major cause of neurological deterioration and morbidity.^{17,18,23–25} Substantially less effort has been invested trying to characterize differences between the hemorrhagic lesion core versus the edematous changes that permeate the primary injury penumbra, where more salvageable brain tissue may reside. This probably relates to the fact that the traditional ellipsoid approximations, which have been most commonly used in this setting, cannot adequately delineate core-edema boundaries.^25,28,29 Using a novel, semiautomated approach at segmentation, we were able to obtain precise volumetric measurement and parcel out different aspects of lesion morphology.

This allowed us to demonstrate that autoregulatory dysfunction is preferentially associated with edema formation. According to experimental data, the transfer of kinetic forces at the moment of the traumatic impact triggers metabolic changes in surrounding neural and vascular tissues, which may compromise their functionality in the absence of gross structural damage.⁷ This may render the parenchyma adjacent to the primary injury site particularly vulnerable to the large fluctuations in blood flow allowed by compromised autoregulatory mechanisms and promote both cytotoxic and vasogenic edema development pathways.^30,31 An alternative hypothesis is that the worsening mass effect attributed to edema expansion could interfere with normal cerebrovascular physiology. This could explain the strong associations with RAC-based metrics, which have been postulated to contain information pertaining to cerebral compensatory reserve.¹¹ This hypothesis is, however, not supported by the fact that patients with comparatively larger extra-axial lesion volumes did not exhibit a similar degree of cerebrovascular reactivity impairment. It also would not explain why these relationships held true at the lower ranges of edema formation, where we would not expect significant depletion of compensatory reserve.

Only PAx-based measures—in particular percent time spent with a positive PAx value—showed associations with core expansion. There exists preliminary evidence that this index is more robust than PRx at higher levels of intracranial compliance.³² It may therefore have been more sensitive to reduction in autoregulatory capacity in our sample given that ICPs were not significantly elevated for most patients (see Table 2). The high predictive value of initial lesion volume that we observed with respect to core expansion is in agreement with the rest of the literature.^{17–19,25,33} The literature has been less consistent about the impact of coagulopathy, which may be attributed to differences in the definition of coagulopathy used.^17,19,23,25 In this study, we relied on fairly conservative thresholds based on conventional laboratory assays, which may not capture the full spectrum of trauma-associated deficiencies in the coagulation cascade.

The improvement in multi-variate fit after including mean CPP in the models, despite the borderline (Δ core) or lack of (Δ edema) statistical significance in the univariate analyses, suggests a possible interaction between deficient autoregulation, increased perfusion pressures, and lesion progression. Physiological principles dictate that high perfusion pressures could translate in supraphysiological cerebral blood flow patterns in a pressure-passive neurovascular environment, potentially making the combination of high CPP and impaired autoregulation especially deleterious from a lesion progression standpoint.

It is interesting that the total amount of edema on repeat scan best predicted unfavorable outcome at 6 months, whereas incremental increases in the amount of intraparenchymal bleeding conferred a greater risk of mortality compared to edema. This once again highlights the value of using approaches that can provide morphology-specific measurements. Establishing rational therapeutic targets to mitigate secondary structural and functional tissue damage in TBI will require an understanding of the respective clinical course and risk factors associated with each lesion subtype.

Expansion of EAH is dictated by different pathophysiological mechanisms, and it did not show significant associations with autoregulatory capacity in our study. It must, however, be emphasized that patients with large epidural or subdural hematomas requiring immediate evacuation would have been excluded from our sample, introducing an inherent selection bias. Further, degree of progression for the remaining EAHs was relatively low, which may have reduced our power to detect significant relationships. The lower rate and severity of progression we observed may also reflect the redistribution of conservatively managed EAHs over time, which makes serial volume comparisons less reliable.³⁴

Limitations

A number of limitations should be mentioned. First, despite relying on multi-center data, the overall sample size was relatively low owing to the exclusion of surgical and EVD patients. This introduced some selection bias and prevented us from performing an internal validation of our multi-variate models. More important, the intervals for the repeat imaging studies and the exact time frame for the intracranial recordings were not pre-specified in this study, but left to the discretion of the clinical team based on clinical indication for pragmatic reasons. Despite our efforts to control for the discrepancies in time points by incorporating random effects into our models and by adjusting for the timing of the radiological assessments, there may be residual bias from temporally dependent variables. In addition, our results do not account for heterogeneity in the treatments received by patients enrolled in the high-resolution cohort including ICP-lowering therapies and the use of different vasoactive agents. Given that this was an observational study with a comparative effectiveness research mandate, participating ICUs were not required to adhere to a specific treatment protocol.

Conclusion

Despite these limitations, our study was able to demonstrate robust associations between autoregulatory failure, traumatic edema progression, and poor outcome. Future studies should aim to externally validate these findings, ideally using prospective, multi-center data. A separate analysis of surgical patients should also be considered to determine whether impairment in cerebrovascular reactivity contribute to post-surgical edema and rebleeding and explore relationships with outcome in this subset of more severely injured patients.

CENTER-TBI High Resolution Sub-Study Participants and Investigators

Audny Anke, University Hospital Northern Norway, Tromso Norway; Ronny Beer, Medical University of Innsbruck, Innsbruck, Austria; Bo-Michael Bellander, Karolinska University Hospital, Stockholm, Sweden; Erta Beqiri, Niguarda Hospital, Milan, Italy; Andras Buki, University of Pécs, Hungary; Manuel Cabeleira University of Cambridge, Addenbrooke's Hospital, Cambridge, United Kingdom; Marco Carbonara, Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico, Milan, Italy; Arturo Chieregato, Niguarda Hospital, Milan, Italy; Giuseppe Citerio, ASST di Monza, Monza, Italy and Università Milano Bicocca, Milano, Italy; Hans Clusmann, Medical Faculty RWTH Aachen University, Aachen, Germany; Endre Czeiter, University of Pecs, Hungary; Marek Czosnyka, University of Cambridge, Addenbrooke's Hospital, Cambridge, United Kingdom; Bart Depreitere, University Hospitals Leuven, Leuven, Belgium; Ari Ercole, University of Cambridge, Addenbrooke's Hospital, Cambridge, United Kingdom; Shirin Frisvold, University Hospital Northern Norway, Tromso, Norway; Raimund Helbok, Medical University of Innsbruck, Innsbruck, Austria; Stefan Jankowski, Sheffield Teaching Hospitals NHS Foundation Trust, Sheffield, United Kingdom; Danile Kondziella, Region Hovedstaden Rigshospitalet, Copenhagen, Denmark; Lars-Owe Koskinen, Umeå University, Umeå, Sweden; Ana Kowark, University Hospital of Aachen, Aachen, Germany; David K. Menon, University of Cambridge, Addenbrooke's Hospital, Cambridge, United Kingdom; Geert Meyfroidt, University Hospitals Leuven, Leuven, Belgium; Kirsten Moeller, Region Hovedstaden Rigshospitalet, Copenhagen, Denmark; David Nelson, Karolinska University Hospital, Stockholm, Sweden; Anna Piippo-Karjalainen, Helsinki University Central Hospital, Helsinki, Finland; Andreea Radoi, Vall d'Hebron University Hospital, Barcelona, Spain; Arminas Ragauskas, Kaunas University of Technology and Vilnius University, Vilnius, Lithuania; Rahul Raj, Helsinki University Central Hospital, Helsinki, Finland; Jonathan Rhodes, NHS Lothian & University of Edinburg, Edinburgh, United Kingdom; Saulius Rocka, Kaunas University of Technology and Vilnius University, Vilnius, Lithuania; Rolf Rossaint, University Hospital of Aachen, Aachen, Germany; Juan Sahuquillo, Vall d'Hebron University Hospital, Barcelona, Spain; Oliver Sakowitz, Klinikum Ludwigsburg, Ludwigsburg, Germany and University Hospital Heidelberg, Heidelberg, Germany; Peter Smielewski, University of Cambridge, Addenbrooke's Hospital, Cambridge, United Kingdom; Nino Stocchetti, Milan University and Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico, Milano, Italy; Nina Sundström, Umea University, Umea, Sweden; Riikka Takala, Turku University Central Hospital and University of Turku, Turku, Finland; Tomas Tamosuitis, Kaunas University of Health Sciences, Kaunas, Lithuania; Olli Tenovuo, Turku University Central Hospital and University of Turku, Turku, Finland; Peter Vajkoczy, Universitätsmedizin Berlin, Berlin, Germany; Alessia Vargiolu, ASST di Monza, Monza, Italy; Rimantas Vilcinis, Kaunas University of Health Sciences, Kaunas, Lithuania; Stefan Wolf, Universitätsmedizin Berlin, Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany; Alexander Younsi, University Hospital Heidelberg, Heidelberg, Germany; Frederick A. Zeiler, University of Cambridge, Addenbrooke's Hospital, Cambridge, United Kingdom and Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, Manitoba, Canada.

Funding Information

Additional funding was obtained from the Hannelore Kohl Stiftung Germany), from OneMind (USA) and from Integra LifeSciences Corporation (USA). F.M. has received salary support for dedicated research time from the Canada Cambridge Scholarship funded by the Cambridge Commonwealth Trust. F.A.Z. receives support from the National Institutes of Health (NIH) through the National Institute for Neurological Diseases and Stroke (NINDS) and the National Institute for Biomedical Imaging and Bioengineering (NIBIB), University of Manitoba Thorlakson Chair in Surgical Research Establishment Fund, University of Manitoba VPRI Research Investment Fund (RIF), Winnipeg Health Sciences Centre Foundation, and the University of Manitoba Rudy Falk Clinician-Scientist Professorship. V.N. is supported by an Academy of Medical Sciences/The Health Foundation Clinician Scientist Fellowship. These studies were also supported by infrastructure provided by the NIHR Cambridge Biomedical Research Centre (BRC), which is a partnership between Cambridge University Hospitals NHS Foundation Trust and the University of Cambridge, funded by the National Institute for Health Research (NIHR). In addition, P.J.A.H. was supported by an NIHR Research Professorship and DKM by an NIHR Senior Investigator Award. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.

Author Disclosure Statement

P.S. and M.C. receive part of the licensing fees for the software ICM+ used for data collection and analysis in this study.