Multimodality Monitoring for Cerebral Perfusion Pressure Optimization in Comatose Patients with Intracerebral Hemorrhage

Abstract

Background and Purpose

Limited data exists to recommend specific cerebral perfusion pressure (CPP) targets in patients with intracerebral hemorrhage (ICH). We sought to determine the feasibility of brain multimodality monitoring (MMM) for optimizing CPP and potentially reducing secondary brain injury after ICH.

Methods

We retrospectively analyzed brain MMM data targeted at perihematomal brain tissue in 18 comatose ICH patients (median monitoring: 164 hours). Physiological measures were averaged over one-hour intervals corresponding to each microdialysis sample. Metabolic crisis (MC) was defined as a lactate/pyruvate ratio (LPR) >40 with a brain glucose concentration <0.7 mmol/L. Brain tissue hypoxia (BTH) was defined as P_btO₂ <15 mm Hg. Pressure reactivity index (PRx) and oxygen reactivity index (ORx) were calculated.

Results

Median age was 59 years, median GCS score 6, and median ICH volume was 37.5 ml. The risk of BTH, and to a lesser extent MC, increased with lower CPP values. Multivariable analyses showed that CPP <80 mm Hg was associated with a greater risk of BTH (OR 1.5, 95% CI 1.1–2.1, P=0.01) compared to CPP >100 mm Hg as a reference range. Six patients died (33%). Survivors had significantly higher CPP and P_btO₂ and lower ICP values starting on post-bleed day 4, whereas LPR and PRx values were lower, indicating preservation of aerobic metabolism and pressure autoregulation.

Conclusions

P_btO₂ monitoring can be used to identify CPP targets for optimal brain tissue oxygenation. In patients who do not undergo MMM, maintaining CPP >80 mm Hg may reduce the risk of BTH.

INTRODUCTION

High blood pressure (BP) after intracerebral hemorrhage (ICH) is associated with hematoma expansion, aggravation of perihematomal edema, and poor clinical outcome.^1–3 The American Stroke Association (ASA) guideline for lowering mean arterial pressure (MAP) below 130 mm Hg and maintaining a cerebral perfusion pressure (CPP) >60 mm Hg in patients with suspicion of increased intracranial pressure (ICP) is based on limited clinical evidence.⁴ Recent studies have shown aggressive BP control is feasible and might be beneficial in reducing early hematoma expansion in patients with acute ICH.^{5, 6} However, these studies were mainly performed in ICH patients with relatively mild deficits. The benefits of aggressive BP control are even more controversial in severe ICH patients.

Brain tissue oxygen tension (P_btO₂) and microdialysis monitoring are used to detect impending ischemia as evidenced by brain tissue hypoxia (BTH) or derangements of oxidative metabolism. BTH and metabolic crisis (MC), defined as elevation of the lactate/pyruvate ratio (LPR) with concurrent brain tissue hypoglycemia, are associated with poor clinical outcome in comatose brain-injured patients.^7–9 Poor neurologic outcome in brain injured patients is associated with autoregulatory failure, a phenomenon in which ICP and P_btO₂ levels correlate positively with arterial BP.^{10, 11} Continuous monitoring of cerebrovascular autoregulation, brain tissue oxygenation, and metabolism may help us better understand the relationship between BP and secondary tissue injury after ICH.^{12, 13} In this study, we sought to establish the feasibility of brain multimodality monitoring (MMM) for guiding CPP management in comatose patients with ICH. Specifically, we hypothesized that a CPP threshold exists below which perihematomal BTH and MC increasingly occurs.

METHODS

Study Population

Nineteen comatose ICH patients underwent MMM between May 2006 and September 2010 in our neuro-ICU according to a standardized protocol. Patients considered eligible for monitoring had a Glasgow Coma Scale score of 3–8, and a supratentorial intraparenchymal or intraventricular hemorrhage (IVH) volume of >30 ml. Exclusion criteria included absence of brain stem reflexes, urgent surgical hematoma evacuation, or Do-Not-Resuscitate status. Probes were placed in the frontal lobe and directed at perihematomal brain tissue within 3 cm of the hemorrhage margin whenever possible. Representative figure for probe location is available online at http://stroke.ahajournals.org. One monitored patient was excluded from the analysis because probes were placed in infarcted tissue. In two patients who underwent hemicraniectomy for ICP control, probes were inserted contralateral to the hemorrhage based on the appearance of bilateral intraventricular hemorrhage (IVH) and global cerebral edema. This observational study was approved by the Columbia University Medical Center Institutional Review Board.

Clinical Management

All patients were treated according to a standardized management protocol. An ICP goal of <20 mm Hg was maintained using a stepwise management strategy.^{4, 14} CPP was targeted at above 60 mm Hg at all times, and was directed at higher target levels on a case-by-case basis as needed to optimize P_btO₂ based on daily review of previous 24-hour data. All patients were ventilated to achieve an arterial oxygen saturation ≥95% and PCO₂ of 30 to 40 mmHg. P_btO₂ measurements were excluded from this analysis when the fraction of inspired oxygen (FiO₂) exceeded 50%.

Data Acquisition

A high resolution data acquisition system (BedmasterEX, Excel Medical Electronics) was used to acquire digital data every 5 seconds. ICP monitoring was performed using a parenchymal ICP probe (Camino System, Integra Neurosciences), P_btO₂ was measured with a Clark type probe (Licox System, Integra Neurosciences), and microdialysis monitoring was performed with a 20K Dalton cutoff catheter with 10 mm membrane length (CMA Microdialysis®). Cerebrovascular pressure reactivity index (PRx)¹⁵ and oxygen reactivity index (ORx)¹⁶ were calculated post-hoc as the running 200-second Pearson correlation coefficient between ICP and MAP (PRx) and P_btO₂ and CPP (ORx). PRx and ORx values range from +1 to −1, with more positive values indicating impaired autoregulation.

Radiological Image Analysis

Admission Brain CT scans were analyzed using MIPAV software (Medical Image Processing, Analysis, and Visualization, version 4.3, National Institutes of Health, Maryland).¹⁷ Regions of hemorrhage on CT scan were outlined slice-by-slice using a semiautomatic threshold approach by a rater blinded to all clinical information.¹⁸ Parenchymal hematoma and IVH volumes were calculated.

Statistical Analysis

All physiological variables were averaged over the time period corresponding to each microdialysis sample (usually every hour). MC was defined as a LPR > 40 and brain glucose < 0.7 mmol/L.¹⁹ BTH was defined as P_btO < 15 mmHg.^{20, 21} Optimal CPP for autoregulation (CPP_prx) for each day was defined as the CPP point with the lowest value of PRx on a PRx-CPP plot.²² Delta CPP was defined as the mean daily CPP – CPP_prx. A positive delta CPP means that the daily observed CPP was higher than the CPP_prx.

Univariate comparisons of pooled data were carried out using a generalized linear model (GLM) using a binomial distribution and logit link function and extended by generalized estimating equations (GEE) using the autoregressive process (AR-1)²³ to handle repeated observations within subject. SPSS 18 software® (SPSS Inc., Chicago, IL, USA) was used for data analysis. A P value < 0·05 was considered statistically significant.

RESULTS

Demographics

Among 18 comatose and ventilated ICH patients, 9 were women and the median age was 59 years (interquartile range (IQR) 42 to 67) (Table 1). Median parenchymal ICH volume was 37.5 ml (IQR 1.8 to 62.3), median IVH volume was 27 ml (IQR 3 to 49), median GCS score was 6 (IQR 4 to 8), and the median MMM time was 164 hours (IQR 88 to 204). Thirteen patients had deep hemorrhages, all but one with co-existing IVH. No complications occurred as a result of MMM probe insertion. Six patients (33%) died in the hospital; all 12 survivors were discharged to either acute or sub-acute rehabilitation facilities. Life support was actively withdrawn in five; one patient (No. 16) died of brain death related to ICP crisis.

Table 1.

Demographic and Clinical Features

No.	Age	Sex	Location	Etiology	GCS	Percent of MMM with MC	Percent of MMM with BTH	In-hospital death	ICH volume (ml)	IVH volume (ml)	Onset to MMM (hrs)	Duration of MMM (hrs)
1	55	F	Parietal	HT	7	2	3	No	41	0.3	27	66
2	40	F	Thalamic	HT	8	66	67	Yes	38	7	29	210
3	41	M	Caudate	HT	8	28	34	No	1	73	41	76
4	78	M	Putaminal	HT	4	0	3	No	2	91	47	288
5	57	M	Putaminal	Cocaine	7	51	36	Yes	63	0	17	403
6	61	F	Thalamic	HT	8	0	18	No	22	32	28	202
7	77	F	Caudate	HT	7	1	0	No	0.2	82	12	102
8	48	F	Putaminal	MMD	4	2	16	No	62	2	27	152
9	53	F	Putaminal	HT	8	68	70	Yes	37	27	26	81
10	64	F	Thalamic	HT	5	0	11	No	28	2	21	91
11	28	M	Temporal	HT	5	1	12	No	88	0	48	198
12	80	M	Frontal	AA	8	12	2	No	67	3	47	42
13	32	M	Primary IVH	HT	3	34	78	Yes	0	30	20	173
14	65	F	Putaminal	HT	5	23	69	No	54	26	24	142
15	65	F	Primary IVH	HT	4	0	12	No	0	84	20	166
16	43	M	Putaminal	HT	4	56	79	Yes	56	33	6	163
17	73	M	Putaminal	HT	4	25	6	Yes	19	49	41	173
18	64	M	Putaminal	HT	7	14	16	No	77	22	34	281

GCS, Glasgow coma scale; MMM, multimodality monitoring; MC, metabolic crisis; BTH, brain tissue hypoxia; M, male; F, female; HT: Hypertension; MMD: Moyamoya disease; AA: amyloid angiopathy

Relationship of CPP to Brain Tissue Oxygenation and Metabolic Crisis

Analysis of 24-hour data frequently revealed linear relationships between PbtO2 and CPP which tended to resolve over time (A supplement figure is available online at http://stroke.ahajournals.org). The probability of BTH increased significantly from 21% to 58% as CPP declined from >90 to <50 mm Hg (Figure 1). A less pronounced relationship existed between CPP and MC: the probability was 17% when CPP was below 70 mm Hg, and steadily fell to 3% when CPP was above 110 mm Hg.

Figure 1. Probability of BTH and MC as a function of CPP in the entire study population.

The risk of cerebral metabolic crisis and brain tissue hypoxia increased progressively with decreasing cerebral perfusion pressure (CPP) values. Histogram shows the distribution of mean hourly CPP values.

Predictors of Brain Tissue Hypoxia

Univariate analysis showed that patients were 18% less likely to experience BTH for every 10 mm Hg increase in CPP (odds ratio (OR) 0.82, 95% confidence interval (CI) 0.69 – 0.96, P = 0.01) (Table 2). ETCO₂ and impaired pressure autoregulation (PRx > 0.2) were also significantly associated with BTH. In a multivariable GEE model, low CPP and ETCO₂ levels < 34 mm Hg (dichotomized on median value) were significantly associated with BTH after adjusting for age, admission GCS, APACHE-II subscore, and PRx > 0.2 (Table 3). The odds for BHT reached statistical significance when CPP was <80 mm Hg compared to CPP ≥100 mm Hg as a reference (OR 1.5, 95% CI 1.1–2.1).

Table 2.

Univariate Analysis of Risk Factors for Brain Tissue Hypoxia and Metabolic Crisis

	Brain tissue hypoxia		Metabolic crisis
	OR (95% CI)	P value	OR (95% CI)	P value
Demographics
Female	1.63 (0.46–5.71)	0.45	1.88 (0.25–14.35)	0.54
Age, per year	1.00 (0.97–1.04)	0.91	0.98 (0.92–1.03)	0.39
Past Medical History
Hypertension	0.83 (0.29–2.36)	0.73	0.57 (0.09–3.78)	0.56
Diabetes mellitus	2.13 (0.54–8.39)	0.28	2.66 (0.34–21.02)	0.35
Smoking	1.58 (0.45–5.52)	0.47	4.25 (0.71–25.52)	0.11
Baseline Clinical
ICH volume, per ml	0.99 (0.97–1.02)	0.79	1.01 (0.98–1.03)	0.57
IVH volume, per ml	1.00 (0.97–1.04)	0.76	0.93 (0.87–1.00)	0.06
Glasgow coma scale	1.12 (0.80–1.56)	0.50	0.98 (0.85–1.11)	0.34
APACHE-II subscore*	0.95 (0.86–1.04)	0.23	0.95 (0.82–1.09)	0.45
Daily Variables
ICH day	0.90 (0.73–1.11)	0.32	1.06 (0.93–1.20)	0.40
Hourly Variables
CPP, per 10 mm Hg	0.82 (0.69–0.96)	0.014	0.98 (0.90–1.07)	0.71
ICP, per mm Hg	1.02 (0.98–1.07)	0.31	0.99 (0.97–1.01)	0.22
ETCO2, per mm Hg	0.85 (0.78–0.93)	< 0.001	0.95 (0.90–1.01)	0.11
Serum glucose, per mmol/L	1.01 (1.00–1.02)	0.06	1.00 (0.99–1.02)	0.30
PRx > 0.2	3.05 (1.47–6.31)	0.003	0.98 (0.76–1.27)	0.88
ORx > 0.2	0.89 (0.45–1.75)	0.75	1.12 (1.02–1.23)	0.02

OR, odds ratio; CI, confidence interval; CPP, cerebral perfusion pressure; ETCO₂, end tidal carbon dioxide; PRx, pressure reactivity index; ORx, oxygen reactivity index (refer to text for definitions). P values in bold are significant.

Table 3.

Multivariable Model for Predicting Brain Tissue Hypoxia

	Brain Tissue Hypoxia
	Adjusted OR (95%-CI)	P Value
ETCO2 <34 mmHg	1.4 (1.2–1.8)	< 0.01
Ranges of CPP
>100 mmHg	Reference group
90–100 mmHg	1.0 (0.8–1.3)	0.97
80–90 mmHg	1.3 (0.9–1.7)	0.11
70–80 mmHg	1.5 (1.1–2.1)	0.01
60–70 mmHg	1.7 (1.2–2.5)	<0.01
<60 mmHg	1.8 (1.3–2.4)	<0.01

Multivariable model using GEE adjusted for the variables listed. P values in bold are significant. ETCO₂, end tidal carbon dioxide; PRx, pressure reactivity index; CPP, cerebral perfusion pressure.

Predictors of Metabolic Crisis

Univariate analysis showed that patients were 12% more likely to have MC when their ORx was above 0.2 (OR: 1.12; 95%-CI: 1.0–1.2, P=0.02). However, in a multivariable model no variables were statistically significant for predicting MC (Table 2).

Factors associated with In-Hospital Mortality

No baseline characteristics, including age, GCS score, and ICH or IVH volume, were associated with in-hospital mortality. By contrast, all but one of the continuously-recorded variables were significantly different in in survivors and non-survivors. Patients who survived had lower ICP and higher CPP values with similar MAP levels; slightly higher P_btO₂ and microdialysis glucose values; and lower LPR values (Table 4). Delta CPP was significantly greater in survivors compared to those who died. Both PRx and ORx were significantly lower in survivors, indicating greater preservation of autoregulation. Analysis of time-series data comparing survivors and non-survivors showed that survivors had significantly higher CPP and P_btO₂ levels starting on post-bleed day 4, and lower ICP values starting on day 5 (Figure 2). By contrast, survivors had significantly lower PRx and LPR values, indicating preserved pressure autoregulation and aerobic metabolism respectively, throughout the entire monitoring period.

Table 4.

Comparison of Clinical and Multimodality Variables in Survivors and Non-Survivor

	Survivors (N=12)	Non-Survivors (N=6)	P value
Clinical Variables*
Age, median (IQR)	64 (50–74)	48 (38–61)	0.13
ICH volume, ml, median (IQR)	35 (1–66)	32 (14–58)	0.54
IVH volume, ml, median (IQR)	24 (1–79)	14 (0–37)	0.71
GCS score, median (IQR)	6.0 (4.3–7.8)	5.5 (3.8–8.0)	0.60
APACHE-II subscore, median (IQR)	20 (18–25)	19 (14–23)	0.30
Multimodality Data Variables †
ICP, mean (SD)	7.8 (8.3)	16.4 (7.4)	< 0.001
CPP, mean (SD)	93 (20)	85 (196)	< 0.001
Delta CPP (CPP – CPPPRx)	5.0 (11.9)	−2.2(12.2)	0.003
MAP, mean (SD)	100 (18)	100 (19)	0.86
PbtO2, mean (SD)	20 (11)	18 (12)	<0.01
LPR, mean (SD)	27 (9)	39 (11)	< 0.001
MD glucose, mean (SD)	1.2 (0.8)	1.0 (1.1)	0.002
PRx, mean (SD)	0.05 (0.24)	0.32 (0.23)	< 0.001
ORx, mean (SD)	0.02 (0.15)	0.12 (0.16)	< 0.001

^*Statistical comparisons were performed using the Mann-Whitney test

^†Generalized estimating equations.

Figure 2. Serial changes of physiological variables during the first nine days after ICH.

Non-survivors had persistently elevated intracranial pressure with decreased cerebral perfusion pressure from postbleed days 4 or 5, respectively (A) (P < 0.05). Brain tissue oxygen tension (P_btO₂) and degree of aerobic metabolism (lactate/pyruvate ratio) were also disturbed in non-survivors (B and C). Pressure reactivity index were higher in non-survivors for whole monitored time, suggesting more disturbance of autoregulation (D). All statistical analyses were performed using generalized estimating equation. Error bars represent 95 percent confidence interval. P_btO₂: partial brain oxygen tension*: P value < 0.05, A.U.: arbitrary unit.

DISCUSSION

We retrospectively analyzed data to determine whether P_btO₂ and microdialysis monitoring can identify CPP thresholds that might minimize the risk of secondary brain injury. We found that the risk of BTH increased significantly when CPP fell below 80 mm Hg compared to >100 mm Hg as a reference range. Non-survivors had early and sustained impairment of pressure autoregulation (high PRx) and aerobic metabolism (high LPR), and with delayed reductions in CPP and P_btO₂ in the setting of increased ICP.

Established clinical predictors of mortality, including age, GCS score, and volume of hemorrhage were not significantly different in those who died or survived, owing to the similar severity of illness across patients that we studied. By contrast, our findings indicate that mortality was associated with impaired cerebral autoregulation, lower CPP and higher ICP values, and a greater burden of BTH and anaerobic tissue metabolism. Non-survivors had persistently positive PRx values (Figure 2), indicating impaired autoregulation. The prognostic meaning of PRx has been addressed by other groups, showing that low CPP contributed to poor outcome among ICH patients with impaired cerebrovascular reactivity (PRx>0.2).²⁴ We also found persistent extracellular LPR elevations and lower brain tissue glucose levels among those who died. Cerebral lactate and LPR elevation been associated with mortality after severe traumatic brain injury (TBI) and poor-grade SAH,^{9, 25} but to our knowledge has not yet been linked to poor outcome after ICH. The lack of association between MC with CPP in our study supports the concept that persistent mitochondrial dysfunction may play a more important role than hypotension as a cause of impaired energy metabolism.²⁶

Survivors had a progressive reduction in ICP and increase in CPP after day 3 (Figure 2). Even among the non-survivors, mean ICP was maintained less than 20 mm Hg throughout the entire monitoring period, which likely reflects the aggressive management protocol that we adhered to. Survivors in our study not only had higher CPP values overall, but also higher delta CPP (CPP – CPP_PRx) values, indicating less likelihood for CPP to fall below the threshold for optimal pressure reactivity. Further studies are needed to determine whether goal-directed CPP optimization aimed at minimizing BTH can improve outcome after ICH.

Our study has several important limitations. The subject cohort was small, and generalizability is hampered by the fact that all patients were treated in a single facility. Hospital mortality is widely accepted as a reliable and valid endpoint for clinical research, but there is increasing awareness that decisions to withdraw care can influence survival. Since caregivers were not blinded to the MMM results, it is conceivable that this data could have affected decision-making, but we feel that this is unlikely. Life support was actively withdrawn in all but one of the six patients who died. Decisions to continue or withhold life support are an important potential source of bias that may have influenced our mortality analysis. Unfortunately we did not measure long-term survival and functional recovery in this retrospective, hospital-based study. Given the fact that almost every MMM variable that we recorded was associated with mortality, larger studies are needed to better understand their relative importance for predicting outcome.

In conclusion, our findings suggest that MMM is a feasible method for optimizing perfusion and possibly minimizing secondary injury in comatose ICH patients. Since this analysis was viewed as exploratory and hypothesis generating, our findings clearly require independent confirmation in a larger multi-center patient population.