Cerebral Autoregulation Correlation with Outcomes and Spreading Depolarization in Aneurysmal Subarachnoid Hemorrhage

Abstract

Introduction:

Delayed cerebral ischemia (DCI) remains one of the principal therapeutic targets after aneurysmal subarachnoid hemorrhage (aSAH). While large vessel vasospasm may contribute to ischemia, increasing evidence suggests that physiologic impairment through disrupted impaired cerebral autoregulation (CA) and spreading depolarizations (SD) also contribute to DCI and poor neurologic outcome. This study seeks to explore the inter-measure correlation of different measures of CA, as well as correlation with SD and neurological outcome.

Methods:

Simultaneous measurement of 7 continuous indices of CA were calculated in 19 subjects entered in a prospective study of SD in aSAH undergoing surgical aneurysm clipping. Inter-measure agreement was assessed and the association of each index with modified Rankin score at 90 days and occurrence of spreading depolarization was assessed.

Results:

There were 4102 hours of total monitoring time across the 19 subjects. In time resolved assessment, no CA measures demonstrated significant correlation, however most demonstrate significant correlation averaged over one hour. Pressure reactivity (PRx), oxygen reactivity (ORx), and oxygen saturation reactivity (OSRx) were significantly correlated with mRS at 90 days. PRx and ORx also were correlated with occurrence of SD events. Across multiple CA measure reactivity indices, a threshold between 0.3-0.5 was most associated with intervals containing SD.

Conclusions:

Different continuous CA indices do not correlate well with each other on a highly time resolved basis, so should not be viewed as interchangeable. PRx and ORx are the most reliable indices in identifying risk of worse outcome in aSAH patients undergoing surgical treatment. SD occurrence is correlated with impaired CA across multiple CA measurement techniques, and may represent the pathological mechanism of DCI in patients with impaired CA. Optimization of CA in patients with aSAH may lead to decreased incidence of SD and improved neurologic outcomes. Future studies are needed to evaluate these hypotheses and approaches.

INTRODUCTION

Aneurysmal subarachnoid hemorrhage (aSAH) is a devastating neurologic pathology with high fatality and morbidity. The overall aSAH incidence is 7.9 per 100 000 person-years, or approximately 30,000 cases in the United States every year¹. The primary therapeutic approach focuses on early securing of the aneurysm and treatment to prevent delayed cerebral ischemia (DCI)² and other secondary insults over the first several weeks after ictus³. The pathophysiology of DCI is more complex than previously thought, involving more mechanisms than large vessel vasospasm alone since patients can have severe angiographic vasospasm without DCI and can develop DCI without a large vessel vasospasm^4-7. Recent clinical trials have underscored this seeming paradox, where the L-type calcium antagonist nimodipine demonstrated overall benefit in clinical outcome without significant improvement in angiographic vasospasm⁸, while clazosentan, an endothelin receptor antagonist, improved angiographic vasospasm without improving clinical outcomes⁹. Further, recent trials support a beneficial effect in targeting alternate mechanisms such as spreading depolarization (SD) and micro emboli^10,11. Observational data have further implicated the pathologic role of other mechanisms of DCI including impaired cerebral autoregulation (CA)^4,12-14 and SD^15-17

Cerebral autoregulation (CA) is an important evolutionary mechanism by which the brain is able to maintain a constant cerebral blood flow (CBF) over a dynamic range of systemic blood pressures. Measurement of cerebral autoregulatory status requires some approximation of CBF and a vascular “challenge” in order to determine the response. One method to assess this reserve involves a discrete challenge using blood pressure manipulation or vasodilatory challenges measuring the cerebral blood flow blood flow response⁴. Such approaches do not have the temporal resolution to assess changing conditions over days to weeks encountered with aSAH and traumatic brain injury (TBI). Continuous approximations of CA have therefore been developed by calculating the moving Pearson correlation coefficient between a surrogate CBF measure and the mean arterial pressure (MAP). The most widely used of such measures is pressure reactivity index (PRx), where highly time resolved changes in intracranial pressure (ICP) are used as a proxy measure for CBF^18,19 Other possible surrogates for CBF which provide continuous data have been explored due to the potential confounding with PRx (due to the effect of other factors such as craniectomy and edema on ICP). The oxygen reactivity index (ORx)²⁰ uses brain tissue oxygenation as a surrogate for CBF, and the cerebral blood flow reactivity index (CBFRx)²¹ uses CBF measured by an implanted perfusion probe. In addition, noninvasive measurement of cerebral oxygen saturation using near infrared spectroscopy (NIRS) as a marker for CBF has also been used as a continuous autoregulatory index (OSRx)²². Although the methodology for calculating PRx has become more standardized over the years, there remains significant discrepancy in the literature in the nuanced mathematics of individual reactivity calculations. Variability in which CBF surrogate is used, differences in the hardware used to obtain CA measures, different calculation algorithms, modes of filtration of input/output signals, time windows used for signal processing, artifact exclusion, and patient selection have made it difficult to compare CA indices across studies. Despite the heterogeneity within the literature, many studies have suggested that typical patterns of CA are disrupted after brain injury, including aSAH^23-26.

Spreading depolarizations (SD) are a neurophysiologic event characterized by a slow-moving wave of abrupt shifts in intracellular ion gradients resulting in sustained neuron and astrocyte depolarization²⁷. In response to massive depolarization and resultant release of a variety of vasoactive substrates there is a chaotic spread and variable cerebrovascular response which is characterized by hyperemia under normal circumstances, but may result in spreading ischemia in situations where metabolic substrate is limited^12,15,28. This response may cause a physiologic mismatch between energy supply and demand. After pathologic depolarization of neural tissue, energy demand is extraordinarily high, but impaired availability may prevent appropriate metabolic recovery, resulting in expansion of ischemic injury²⁹. SD occurs frequently in subarachnoid hemorrhage patients and has been closely associated with neurologic decline of DCI and worse outcomes¹⁷. Patterns of SD risk and blood pressure have been explored in traumatic brain injury³⁰ and suggested that impaired autoregulation may be a risk for SD due to the risk of triggering SD in vulnerable tissue due to supply demand mismatch³¹. Such SD then would risk furthering injury to vulnerable regions due to metabolic stress and spreading ischemia²⁷.

In the current study, we sought to better elucidate the relationship between various CA measures and assess how they relate to both clinical outcomes and SD incidence.

METHODS

Study Population

Thirty-two patients at the University of New Mexico Hospital with acute neurological injuries were enrolled in an institutional review board approved prospective observational study on the role of spreading depolarization and other multimodality monitoring in outcomes (UNM HRPO 10-159). Prospective consent from the subject or LAR was obtained for post-operative ECoG (electrocorticographic) recordings with simultaneous collection of data from clinically indicated multimodality monitoring. For the current study, we extracted data only from aSAH subjects who also had intracranial multimodality monitoring.

Physiologic and Neurologic monitoring:

A 1x6 subdural ECoG strip (Auragen: Integra Neurosciences, Plainsboro, NJ) was placed under the research protocol and not used for clinical decision making in these subjects. The electrode strip was placed at the time of clinically indicated surgery prior to replacement of the bone flap, on either the frontal or temporal lobe, depending on the region thought to be most at risk of DCI. The electrode was tunneled subcutaneously, similar to a surgical drain as previously published³². Invasive neuromonitoring was performed for standard clinical care using a multi-lumen bolt (Hummingbird: IRRAS AB, Stockholm, Sweden) containing an integrated external ventricular drain (EVD), PbO2 probe (Licox: Integra Neurosciences, Plainsboro, NJ), and thermal diffusion-based perfusion monitor (Bowman Perfusion Monitor: Hemedex, Waltham, MA). This bolt was placed contralateral to the craniotomy, in the opposite hemisphere from the ECoG strip. Data, including all clinical variables as well as DC ECoG were recorded in a time-locked manner using the Moberg Component Neuromonitoring system (Moberg Research, Ambler PA) and archived in a hospital server. Standard multimodality monitoring included continuous arterial pressure (radial line with transducer at the level of the right atrium), two forms of intracranial pressure (EVD with transducer at the level of the external auditory canal, and parenchymal pressure monitor from the Hummingbird device), PbO2, Perfusion, and regional cerebral oxygen saturation with NIRS (INVOS: Medtronic). The laterality of the cerebral oxygen saturation was recorded in relation to ECoG strip, with monitors on the same side as the ECoG strip labeled as “Strip”, and monitors on the opposite side of the ECoG strip labeled as “NoStrip”. A full spectrum DC coupled amplifier was used for ECoG recordings³³. Multimodal monitoring was continued through the critical portion of the hospitalization at the discretion of the neurosurgeon or neurointensivist. The electrode strip was removed at the bedside with gentle traction when the remainder of the invasive neuromonitoring was removed. SD was scored post hoc using standardized published criteria consisting of a characteristic DC shift, typically with associated transient suppression of high frequency activity and spread to adjacent electrodes³².

Clinical Management:

All patients were treated according to a standardized management protocol for aSAH³. The external ventricular drain was maintained open at 10cm above the external auditory meatus. Systolic blood pressures were allowed to self-regulate with adequate volume support up to 200mmHg. Vasopressors were used only when signs of DCI were present either clinically or radiographically. CA directed therapy using these continuous measures was not directly used in clinical management as all the indices were calculated post hoc, however goal directed protocols were used to treat intracranial hypertension or hypoxia. A standardized, tiered approach to DCI was used beginning with blood pressure augmentation and proceeding to angiography for rescue therapy if there was a lack of clinical or monitoring response to blood pressure augmentation. Glasgow coma score (GCS), Hunt and Hess and World Federation of Neurosurgical Societies (WFNS) scores were recorded at admission in all subjects. The modified Rankin score (mRS) was determined at 90 ±10 days after hospital discharge and was used as the primary outcome measure.

CA Indices and Spreading Depolarization Calculations:

Artifactual values of MAP, ICP, PbO2, perfusion, and regional cerebral oxygen saturation were filtered to ensure that only physiologic values were considered (MAP 45-140mmHg, ICP 0-70mmHg, PbO2 0-50mmHg, perfusion 0-200cc/100g/min, cerebral oxygen saturation 20-100%). Data was exported in one minute mean bins. The PRx value was calculated continuously from the original waveform data using the Moberg Reader using a PRx add-on as the moving 10 second correlation coefficient between the ICP and MAP, using ICP from the parenchymal monitor as this was more continuous than the EVD based ICP. The remaining indices were calculated post hoc using a moving 30-minute moving Pearson correlation coefficient between the CBF surrogate (PbO2, perfusion, cerebral oxygen saturation) and MAP (see figure 1 for additional description). Additional PRx values were also calculated using this approach using the two values for ICP were recorded using the Hummingbird device, one the standard transduced pressure from the EVD (PRx-EVD), and the other from the continuous parenchymal monitor (PRx-Par). All calculated reactivity indices generate a value between +1 and −1, with more positive values indicating impaired CA.

Figure 1: Calculation of Cerebral Autoregulation Measure Reactivity Indexes:

it shows the method of collecting the patient data using multimodal monitoring, followed by filtering them, measuring the reactivity indices in which the pressure reactivity index (PRx)value was calculated using the Moberg Reader and the remaining indices were calculated post hoc using a moving 30-minute time window Pearson correlation coefficient between the cerebral blood flow (CBF) surrogate and mean arterial pressure MAP.

Statistical Analysis:

The inter-measure correlation of CA index was done using simple linear regression to create a matrix of correlation values, with the p-values reported. Comparisons were made between one-minute averages across the entire monitoring interval (where data was available), as well as averages over 1 hour, 12 hours, 24 hours, and whole monitoring interval. Correlations with reactivity indices and clinical outcomes were assessed using the average CA value and outcomes defined as the dichotomized mRS (0-3 vs 4-6). These associations were tested using univariable logistic regression. The association of SD probability across the range of MAP in all subjects was assessed and resultant curve examined for modeling. The correlation of CA measures and frequency of SD events was assessed with binary logistic regression models for the minute-by-minute data, and Poisson regression for the 1 hour, 12 hours, and 24 hours averages. Sequential chi-square modeling was used to assess the optimum threshold for SD prediction. A 2-sided α of less than .05 was considered statistically significant. Statistical analysis was performed using SAS (v9.4: Carey, NC) and GraphPad Prism (v9.1.1: San Diego, CA).

This case series has been reported in line with the PROCESS Guideline³⁴

RESULTS

Of the 32 subjects enrolled in the prospective study, 19 met the inclusion criteria. Table 1 details the demographics and baseline SAH characteristics of the cohort. Figure 2 Part A demonstrates the total number of days of monitoring as well as the percentage of the monitoring period that included valid recording data across all subjects. There was 4102 total hours of monitoring time across subjects. CBFx was the least reliable CA modality with only 40% of the total monitoring time recording valid data. OSRx reliability was also low, with valid data only around 50% of the total monitoring time. All other modalities had greater than 80% valid monitoring time.

Table 1.

Demographics and Baseline aSAH Characteristics

Subject Number	Age	Sex	Aneurysm Location	Admit Glasgow Coma Scale score	Hunt & Hess score	Duration of Monitoring (Hours)	Modified Rankin Score at 90 Days
1	58	F	Ophthalmic ICA	3	3	310	4
2	66	F	ACOM	3	5	113	4
3	58	F	MCA	8	4	190	6
4	55	F	ACOM	14	2	258	5
5	46	F	MCA	13	3	328	0
6	36	F	MCA	7	5	261	2
7	55	M	Ant Cho	7	5	243	1
8	67	F	PCOM	15	2	147	2
9	47	F	ACOM	15	2	141	2
10	62	F	MCA	14	3	17	5
11	69	F	SCA	15	2	372	4
12	54	F	ACOM	15	2	175	2
13	53	F	PCOM	5	5	189	5
14	48	M	ACOM	3	5	235	5
15	43	F	PCOM	15	2	95	1
16	72	F	ACA	3	4	289	5
17	50	F	Carotid Wall	14	3	190	3
18	66	F	PCOM	9	3	287	4
19	44	M	ACOM	15	2	188	2

Figure 2: Comparison of usable, post filtering data time between various autoregulation indices.

Part A: demonstrates the total number of hours of monitoring as well as the percentage of the monitoring period that included valid recording data across all subjects’ index.Pressure reactivity index (PRx), cerebral blood flow reactivity index (CBFRx), oxygen reactivity index (ORx), oxygen scalp reactivity index from the side that has the strip (OSRx Strip), oxygen scalp reactivity index from the opposite side(OSRx No Strip),. PRx calculated from the parenchymal monitor (PRx-Par). PRx calculated from the pressure from the EVD (PRx-EVD)

Part B: Correlation Matrix of Inter-measure Comparability Between Autoregulation Indices: The minute-by-minute analysis demonstrates poor correlation of CA values with each other, indicating that on a highly time resolved scale, the measures do not agree well as to when autoregulation is impaired. The correlation between measures is best using 1-hour averages with 15 out of 21 correlations demonstrating a significant correlation. [* Pressure reactivity index (PRx), † PRx calculated from the pressure from the ventriculostomy tube (PRx-EVD) , ‡ PRx calculated from the parenchymal monitor (PRx-Par), § oxygen reactivity index (ORx), | | cerebral blood flow reactivity index (CBFRx), # oxygen scalp reactivity index from the side that has the strip (OSRx Strip), * *oxygen scalp reactivity index from the opposite side(OSRx No Strip) ]

Figure 2 Part B demonstrates the correlation matrix across each CA index. The minute-by-minute analysis demonstrates poor correlation of CA values with each other, indicating that on a highly time resolved scale, the measures do not agree well as to when autoregulation is impaired. The correlation between measures is best using 1-hour averages with 15 out of 21 correlations demonstrating a significant correlation. The correlations within the 12-hour and 24-hour averages are comparable to the 1-hour averages (supplementary data).

Table 2 displays the univariable logistic regression between the demographic data and overall CA averages with the dichotomized mRS at 90 days. Higher age, lower GCS, higher PRx, ORx, and OSRx measures demonstrated significant association with worse neurologic outcomes. While there was some variability among the various PRx measures, PRx calculated from the parenchymal monitor (PRx-Par) demonstrated perfect separation of good versus poor outcome subjects with a threshold of 0.3. Interestingly, this significant separation was only observed when a 30m rolling average was used (PRx-Par) and not with the high resolution 10s rolling average (Prx) from the same parenchymal monitor. ORx and both literalities of OSRx were also significantly associated with worse outcome.

Table 2.

Univariable Logistic Regression of Whole Admission CA Measure with Outcome (Dichotomized mRS 0–3 vs 4–6)

Variable	Odds Ratio	Lower 95% Confidence Interval	Upper 95% Confidence Interval	p-Value
Demographics
Age	1.1904	1.0178	1.3923	0.0292*
Admit GCS	0.7685	0.6018	0.9815	0.0349*
CA Measure
PRx	1.0298	0.9477	1.1190	0.4886
PRx-EVD	1.1742	0.9828	1.4029	0.0769
PRx-Par	259.907 †	0.0000 †	∞ †	0.0000 *
ORx	1.2080	1.0303	1.4164	0.0199 *
CBFRx	1.0088	0.9360	1.0874	0.8180
OSRx Strip	1.1391	1.0180	1.2746	0.0231 *
OSRx No Strip	1.1632	1.0093	1.3406	0.0368 *

^*Indicates statistical significance

^†Indicates Perfect separation of data

Fig 3 Part B demonstrates a plot of SD probability across 5mmHg MAP bins. We used 30-minute overall bins to assess these associations. A spline curve was fit to assess the overall contour of the relationship due to the non-linear relationship. Similar to results published from TBI subjects³⁰, we noted an increase in SD probability at lower MAP, a long flat section across the mid-range, and decreased probability at higher MAP, suggesting that SD could be a function of autoregulation. We therefore assessed the correlation of individual CA measures and incidence of SD (Table 3). PRx-Par and ORx were significantly associated with SD using logistic or Poisson regression. We then performed sequential chi square analysis (figure 3 Part A) to evaluate the discriminatory power of various thresholds of each autoregulatory index by plotting the chi square value across the range of the autoregulatory index. Focusing on the 0−+1 range (i.e. impaired autoregulation), we identified the peak that identifies the optimal autoregulation threshold to predict SD in 1 hour bins for each index (Figure 3). The most robust peaks were noted with PRx and ORx, with OSRx having essentially no discriminative power to detect SD, even on the same side as the strip electrode (where SD is measured). Interestingly, most thresholds for discriminating SD were in the 0.3-0.5 range, which is similar to thresholds used to discriminate clinical poor outcomes³⁵.

Figure 3.

Part A: Optimum threshold for predicting SD across multiple autoregulation indices where it shows PRx calculated from the parenchymal monitor (PRx-Par) and oxygen reactivity index (ORx) were the variables significantly associated with SD using logistic or Poisson regression SD using a sequential chi-square analysis

Part B: Spline Fitted Curve of MAP versus SD Probability, it demonstrates a plot of Spreading depolarization probability across 5mmHg MAP bins. Using 30-minute overall bins to assess these associations

Table 3.

CA Measure Reactivity Index and Frequency of Spreading Depolarizations

CA Measure	1 Min Avg†	1 Hour Avg ‡	12 Hour Avg‡	24 Hour Avg‡
PRx	0.5464	0.9746	0.8015	0.8293
PRx-EVD	0.6906	0.0972	0.1072	0.0153
PRx-Par	0.0227 *	0.0895	0.0328 *	0.1286
ORx	0.1007	0.0056 *	0.0026 *	0.1417
CBFRx	0.2741	0.6627	0.4992	0.6994
OSRx Strip Side	0.3135	0.4192	0.5686	0.4347
OSRx No Strip Side	0.8881	0.8598	0.3497	0.4157

^*Indicates statistical significance

^†Binary Logistic Regression

^‡Poisson Regression

DISCUSSION

To the authors’ knowledge, this is the first series to measure multimodality CA status simultaneously in patients undergoing SD monitoring after aSAH. Additionally, this cohort is unique in that all CA measures have been simultaneously recorded and the subsequent application of uniform data filtering and analysis has allowed for robust cross measure comparisons. It is therefore relevant and notable that different parameters provide somewhat disparate information, especially when evaluated over a short time window. It is unknown if these data represent complimentary information or varying degrees of error in estimating autoregulatory status, but would potentially have important impact if these data were used to calculate an optimal CPP, which would vary between the method of calculation. The various measures demonstrate the best consensus, however, when averaged over 1-hour interval. We sought to identify multiple methods to validate which measure is the most useful. In terms of raw recording times, we found CBFRx to the least useful due to frequent loss of data due to required recalibrations. This was disappointing since direct CBF measurement would theoretically be the ideal measure to assess autoregulation³⁶. ICP and PbO2 were more reliable measures, with valid data obtained in greater than 85% of the monitoring period.

We then sought to determine which CA measure provided the most relevant outcome information. In terms of clinical outcomes, multiple averaged CA values were associated with outcomes, with odds ratios (ORs) similar to the effect of known predictors of age and GCS at admission^11,37. Of particular note, PRx obtained from the parenchymal monitor demonstrated a perfect separation at a threshold of 0.3, with all values below 0.3 categorized as a good outcome, and all values above 0.3 categorized as poor outcome. Interestingly, the input ICP is the same as used in the standard PRx calculations, but averaged over a longer interval (30 minutes versus 10 seconds), potentially allowing for improved assessment of larger changes in MAP. In other words, the rolling 30-minute average theoretically may have some advantage if there are larger changes in blood pressure during that window. It is important to note that our 30-minute values are probably not the same as the previously published PRX-long³⁸, because it is not a longer average of the 10 second calculation, but rather is derived from the 30-minute raw input data. Using 10 second averages provides different information related to the immediate, compliance related, response of ICP to arterial pulsations, but may not reflect larger trends of blood pressure fluctuations. Using the same 30-minute rolling window, ORx and OSRx also demonstrated significant correlation with neurologic outcome. These predictive results will need to be further validated in larger cohorts to determine if the improved prediction over standard PRx is replicated, especially given that higher resolution PRx data has been suspected to better predict outcome in TBI³⁹.

A limitation of using an overall summary of CA with outcome is that the temporal trends may change and transient severe CA impairment during the critical DCI period may not be adequately assessed by this measure. For that reason, we also sought to determine which parameters were most correlated with SD as a temporally resolved sign of metabolic instability and ischemic risk. This seems to be a reasonable assumption given that recent data have suggested that SD represents the principal mechanism of ischemic progression²⁹. Spreading depolarizations also mediate excitotoxicity⁴⁰ as well as edema progression⁴¹ after brain injury and stroke. We identified PRx and ORx as having the strongest associations with frequency of SD events, providing further support that these measures may have some advantage over others in assessing the risk of SD.

The relationship between impaired autoregulation and SD is most likely complex and cyclical. Our observations link these two physiological disturbances in a proposed mechanism whereby impaired autoregulation increases the risk of SD and ischemic progression. Based on the relationship of SD and MAP noted by Hartings and colleagues in TBI patients³⁰, there appears to be a stable frequency of SD across a wide range of MAP values, with increasing frequency of SD with low MAP and decreasing frequency at high MAP. Their observations support a mechanism whereby SD may be triggered when transient drops in MAP exceed the lower threshold of autoregulatory vasodilation. We replicated a similar trend in this cohort, though the residuals are very wide at the low end of the curve, possibly since aSAH patients are typically maintained at significantly higher MAP than TBI patients. In other conditions where permissive hypertension is less tolerated (stroke, hemorrhage, or TBI), the association of impaired autoregulation could be more pronounced. The minor hypotensive transient drops in our patients are probably not severe enough alone to cause ischemia, however if SD is triggered due to supply-demand mismatch³¹, the added metabolic stress and spreading ischemia may be the final mechanism by which impaired autoregulation leads to DCI as previously shown¹⁷. Impaired CA may therefore predispose patients to transient CBF drops that trigger SD in vulnerable brain, leading to ischemic expansion and potentially further worsening locally impaired CA.

These data have meaningful potential implications for multimodality monitoring. We conclude that there is not one monitor that is dramatically better in terms of clinical outcome or SD prediction. ICP and PbO2 both seem to be both reliable reactivity indices associated with worse clinical outcome and SD. Scalp based measures seemed to have some value in outcome prediction, but not in SD prediction, which may be due to regional variation in where the strip is located and the frontal region evaluated by the INVOS. Further assessments are needed to understand if the differences are related to complimentary physiologic information or error/confounding related to the measurement techniques. We also suggest that monitoring of SD may have a complimentary role in assessing the efficacy of CA directed therapy (such as optimal CPP) on a day-to-day basis. Future studies could be directed at assessing temporal relationship between disrupted CA status and SD frequency as well as SD response to CA directed therapies.

Despite the rich and relatively homogeneous data we present, there are limitations to this approach. All subjects required aneurysm clipping and with advances in endovascular techniques, the generalizability to endovascularly treated patients is unknown. We initially chose to focus on a smaller group with simultaneous monitoring, but these results will need to be expanded to determine the applicability when only subjects with fewer modalities are monitored (such as ICP only). Similarly, in the current study, we did not assess for any potential effect of time latency to SD measurement with impaired autoregulation, so it is possible that the effect of such impairment is more notable if this could be corrected. Finally, though no direct CA directed therapy was used in clinical management (such as optimal CPP calculation) the multimodality parameters were incorporated into the clinical assessment of subjects and so may affect our reported results.

Conclusions

Different continuous CA indices do not correlate well with each other on a highly time resolved basis, so should not be viewed as interchangeable. PRx and ORx are the most reliable indices in identifying risk of worse outcome in aSAH patients undergoing surgical treatment. SD occurrence is correlated with impaired CA across multiple CA measurement techniques, and may represent the pathological mechanism of DCI in patients with impaired CA. Optimization of CA in patients with aSAH may therefore lead to decreased incidence of SD and improved long term neurologic outcomes. Future studies are needed to evaluate these hypotheses and approaches.

Abbreviations:

DCI

delayed cerebral ischemia

CA

cerebral autoregulation

aSAH

aneurysmal subarachnoid hemorrhage

SD

spreading depolarization

ICP

intracranial pressure

CBF

cerebral blood flow

NIRS

near infra-red spectroscopy

PRx

pressure reactivity index

CBFRx

cerebral blood flow reactivity index

ORx

oxygen reactivity index

OSRx

oxygen scalp reactivity index

mRS

modified Rankin score

MAP

mean arterial pressure

PbO2

partial pressure of brain tissue oxygen tension

LAR

legally authorized representative

ECoG

electrocorticographic

GCS

Glasgow coma score

WFNS

World Federation of Neurosurgical Societies