RESEARCH HUMAN CLINICAL STUDIES

Transcription

1 TOPIC RESEARCH HUMAN CLINICAL STUDIES RESEARCH HUMAN CLINICAL STUDIES The Relationship Between Intracranial Pressure and Brain Oxygenation in Children With Severe Traumatic Brain Injury Ursula K. Rohlwink, MSc* Eugene Zwane, PhD A. Graham Fieggen, MD* Andrew C. Argent, MD Peter D. le Roux, MDk Anthony A. Figaji, MD, PhD* *School of Child and Adolescent Health, Division of Neurosurgery, Red Cross War Memorial Children s Hospital, University of Cape Town, Cape Town, Western Cape, South Africa; Department of Biostatistics, University of Swaziland, Kwaluseni, Swaziland; Division of Pediatric Critical Care, Red Cross War Memorial Children s Hospital, University of Cape Town, Cape Town, Western Cape, South Africa; kdepartment of Neurosurgery, University of Pennsylvania, Philadelphia, Pennsylvania Correspondence: Anthony Figaji, MD, PhD, ICH Building, Room 617, Red Cross War Memorial Children s Hospital, Klipfontein Road, Rondebosch, Cape Town, South Africa. Anthony.Figaji@uct.ac.za Received, April 12, Accepted, November 1, Published Online, November 30, BACKGROUND: Intracranial pressure (ICP) monitoring is a cornerstone of care for severe traumatic brain injury (TBI). Management of ICP can help ensure adequate cerebral blood flow and oxygenation. However, studies indicate that brain hypoxia may occur despite normal ICP and the relationship between ICP and brain oxygenation is poorly defined. This is particularly important for children in whom less is known about intracranial dynamics. OBJECTIVE: To examine the relationship between ICP and partial pressure of brain tissue oxygen (PbtO2) in children with severe TBI (Glasgow Coma Scale score #8) admitted to Red Cross War Memorial Children s Hospital, Cape Town. METHODS: The relationship between time-linked hourly and high-frequency ICP and PbtO2 data was examined using correlation, regression, and generalized estimating equations. Thresholds for ICP were examined against reduced PbtO2 using age bands and receiver-operating characteristic curves. RESULTS: Analysis using more than 8300 hourly (n = 75) and 1 million high-frequency data points (n = 30) demonstrated a weak relationship between ICP and PbtO2 (r =0.05 and r = 0.04, respectively). No critical ICP threshold for low PbtO2 was identified. Individual patients revealed a strong relationship between ICP and PbtO2 at specific times, but different relationships were evident over longer periods. CONCLUSION: The relationship between ICP and PbtO2 appears complex, and several factors likely influence both variables separately and in combination. Although very high ICP is associated with reduced PbtO2, in general, absolute ICP has a poor relationship with PbtO2. Because reduced PbtO2 is independently associated with poor outcome, a better understanding of ICP and PbtO2 management in pediatric TBI seems to be needed. KEY WORDS: Brain oxygenation, Children, Head injury, Intracranial monitoring, Intracranial pressure, Traumatic brain injury Neurosurgery 70: , 2012 DOI: /NEU.0b013e318243fc59 Copyright ª 2011 by the Congress of Neurological Surgeons Traumatic brain injury (TBI) is a major contributor to the burden of disease in children and adults worldwide. 1,2 The poor outcome associated with severe TBI may be improved by effective medical care, which in large part aims to prevent or ameliorate secondary insults. 3-7 Currently, intracranial pressure ABBREVIATIONS: CPP, cerebral perfusion pressure; GEE, generalized estimating equation; ICP, intracranial pressure; PbtO2, partial pressure of brain tissue oxygen; ROC, receiver-operating characteristic; TBI, traumatic brain injury (ICP) monitoring and management are a cornerstone of treatment for severe TBI. Increased ICP is common after TBI and may exacerbate the primary injury through local tissue pressure effects, reduction in cerebral perfusion pressure (CPP) and compromised brain metabolism, including brain oxygenation. One of the major reasons to treat ICP, therefore, is to ensure adequate brain perfusion and avoid ischemia. Brain ischemia is an important secondary injury after TBI and may increase the risk of death or disability in these patients. 8 Therefore, prompt detection and treatment of ischemia may minimize the potential for irreversible brain damage VOLUME 70 NUMBER 5 MAY

2 INTRACRANIAL PRESSURE AND BRAIN OXYGENATION Observational clinical studies using direct measures of brain oxygen, such as partial pressure of brain tissue oxygen (PbtO2), have found that reduced PbtO2 is associated with poor outcome in both adult and pediatric TBI However, direct measures of brain perfusion or oxygenation are not routinely used in the management of patients with severe TBI, especially not in children. Recent evidence has suggested that reduced PbtO2 may occur despite apparently adequate ICP and CPP control in both adults and children with TBI In particular, very low brain tissue oxygen (PbtO2,10 mm Hg/1.33 kpa) may occur in as many as one third of children with severe TBI despite adherence to current international management guidelines. 18 These episodes appear to be important because reduced PbtO2 is associated with poor outcome and the association between ICP and poor outcome may, in part, be explained by reduced PbtO2. 12 In some patients it is apparent that ICP control is directly associated with improved PbtO2, ie, there is an inverse relationship between ICP and PbtO2. For example, refractory increased ICP tends to be associated with low PbtO2 that responds to decompressive craniectomy in both adult 19 and pediatric 20 TBI. On the other hand, other factors may cause parallel changes in ICP and PbtO2. For example, hyperemia could theoretically increase both ICP and PbtO2 simultaneously, and so the relationship between these 2 variables may notalwaysbeasexpected. The relationship between ICP and PbtO2 has important implications for treatment protocols, particularly because monitors for brain perfusion or oxygenation are not yet recommended for children in international guidelines. A better understanding of the relationship between ICP and PbtO2 therefore may be of value. This is of particular importance for children in whom the interpretation of intracranial relationships may be more complex than in adults. Therefore, we sought to examine the relationship between ICP and PbtO2 in a large cohort of children with severe TBI. Specifically, we aimed to (1) examine the relationship between ICP and PbtO2 from time-linked paired observations of these variables, (2) explore various critical thresholds for ICP and PbtO2, and (3) examine digital data trends depicting the relationship between ICP and PbtO2. METHODS Approval for this study was obtained from the institutional review board of the University of Cape Town. Patient consent for inclusion in the study was obtained from the next of kin. Patient Selection This was a retrospective study of prospectively collected data for all children who underwent ICP and PbtO2 monitoring for severe TBI (postresuscitation Glasgow Coma Scale score of #8) from June 2006 to June All patients were younger than 15 years old. ICP and PbtO2 Monitors ICP was measured with an intraparenchymal monitor (Codman ICP Express; Codman, Raynham, Massachusetts) or Camino (Integra Neurosciences, Plainsboro, New Jersey). ICP catheters were placed on the same side as PbtO2 monitors (Licox; Integra Neurosciences) in normalappearing white matter on head computed tomography scan in the right frontal lobe or in the more injured hemisphere. Correct position of the catheters was confirmed on follow-up computed tomography. Intracranial monitoring was started as soon as possible after admission and continued until ICP and PbtO2 were stable for more than 24 hours without treatment other than sedation for ventilator care or until the patient died. Patient Care Patients were cared for using a local algorithm based on current recommendations for the management of TBI in children. 21 All patients with severe TBI are resuscitated on admission, intubated, and mechanically ventilated in the pediatric intensive care unit. Significant intracranial hematomas are immediately surgically evacuated. Increased ICP (.20 mm Hg/2.67 kpa) is treated in a stepwise fashion. 22,23 In general, we aim to keep CPP greater than 50 mm Hg/6.6 kpa in children older than 2 years of age and more than 45 mm Hg/5.9 kpa in children younger than 2 years of age. Reduced PbtO2 is defined as less than 20 mm Hg (,2.67 kpa) 11 and is treated using an algorithmic approach, beginning with identification of the cause for reduced PbtO2, which was reported elsewhere. 12 Data Collection Time-linked data of physiological parameters were extracted from the hourly recordings taken in the intensive care unit (75 patients). Since April 2008, ICP and PbtO2 data were also captured using a computerized software recording system (ICMPlus; Cambridge University, Cambridge, United Kingdom) that sampled variables at 50 Hz. From this raw data file, variables were averaged over 10 seconds, and these 10-second averages were used for subsequent data analysis. Among this cohort of 75 patients, we had 30 patients who had complete data sets from continuous recording that had been manually cleaned of all artifacts. The first 2 hours of PbtO2 monitoring data were discarded to allow for probe stabilization. Data were excluded from patients who were terminal on admission (brain dead or Glasgow Coma Scale score of 3 with fixed dilated pupils who died within the first 12 hours), from beyond the first hour of PbtO2 = 0, from catheters placed incorrectly or that were in a pericontusional location, and when recorded during tests, such as autoregulation and ventilation tests. Data Analysis Data were analyzed using several different analyses, including the timelinked relationship between ICP and PbtO2, controlling for other variables, overall critical thresholds of ICP, age-related thresholds of ICP, and thresholds for individual patients. Statistical significance was set at 0.05 for all analyses, and data were analyzed using SPSS version 19.0 (SPSS Inc, Chicago, Illinois). We examined 2 groups of data. In group A, we examined hourly variables recorded from the nursing observations of all patients (N = 75). Although these were only hourly data samples, they allowed a larger cohort of patients to be examined. A previous analysis done by our group (unpublished data) using a Bland-Altman plot showed that for hourly and continuously recorded data, 95% of values were in agreement. However, within this group, we also had 30 patients for whom we had high-frequency data (group B). After manual cleaning, we had more than 1 million time points for all variables to examine separately in this group for finer analysis. NEUROSURGERY VOLUME 70 NUMBER 5 MAY

3 ROHLWINK ET AL Time-Linked Relationship Between ICP and PbtO2 Analyses were conducted in a sequential manner, starting with raw data of paired ICP and PbtO2 observations, followed by increasingly defined analyses to explore different methods of analyzing the relationship between these 2 variables. The General Relationship Between ICP and PbtO2 Paired observations for ICP and PbtO2 were examined with Spearman s correlation and a generalized estimating equation (GEE) model to control for interindividual variation. Simple logistic regression was conducted using PbtO2 dichotomized at 10 mm Hg as the dependent variable because this threshold has a well-described independent association with poor outcome ,14,24,25 We examined the impact of several other factors that may influence this relationship, including whether patients had a monitor placed on the side of a significant contusion, whether a craniotomy/craniectomy had been performed, and the time after injury. Where we used continuous data, we also checked for the possibility that a lag-phase difference in the PbtO2 response may influence our results by reanalyzing the data with the PbtO2 values offset by 10 and 20 seconds. Relationship Between ICP and PbtO2 Controlled for Systemic Variables To examine the possible influence of confounding variables, we used multiple linear regression and a GEE model to control for mean arterial pressure, PaO 2,PaCO 2, and hemoglobin. For this analysis, we only used ICP and PbtO2 data from time points when observations for these other variables were available, ie, when arterial blood gases were taken. Critical Threshold Values of ICP and PbtO2 Values for ICP were examined in 5 mm Hg increments to determine whether critical ICP thresholds for changes in PbtO2 could be defined. Given that various studies have reported different thresholds for PbtO2 in association with outcome and the fact that although our treatment threshold for PbtO2 was 20 mm Hg, episodes of lower PbtO2 were common and PbtO2 less than 10 mm Hg had a stronger association with poor outcome in earlier work at our institution, 12 we examined the sensitivity and specificity of these ICP thresholds against several PbtO2 thresholds in 5 mm Hg increments from 5 mm Hg to 20 mm Hg. The relative risks for each threshold of ICP for the different thresholds of PbtO2 were calculated. To determine which ICP threshold level would provide the best balance between sensitivity and specificity to detect reduced PbtO2, receiveroperating characteristic (ROC) curves were constructed. The ROC curve displays the efficacy of a range of tests (in this case, ICP thresholds) by plotting sensitivity against 1-specificity with both axes ranging from 0 to 1. The diagonal of such a graph denotes that the test in question would accurately predict disease (in this case, PbtO2 below threshold) with no greater power than chance. The area under the curve equals the probability that a random sample with a positive result has a higher value than a random sample with a negative result and equals 0.5 under the diagonal (null hypothesis). The effectiveness of a test to detect disease would be found by examining how far the curve lies away from the diagonal in the direction of the y axis and is indicated by the area under the curve, where the perfect curve would have an area of ,27 ROC curves were constructed for ICP against PbtO2 less than 5 mm Hg, less than 10 mm Hg, and less than 20 mm Hg. Age-Related Thresholds Because children are developing physiologically and may have different normative values for ICP at different ages, we also considered that age may influence our results. To date, there are few robust data on age-related ICP treatment thresholds, and no recommendations of specific age-related thresholds for ICP treatment are outlined in the pediatric TBI guidelines. We used data from Chambers et al, 28 who assessed ICP thresholds in association with outcome. Accordingly, our cohort was divided into the following age categories, for which the respective ICP thresholds were adopted: 2 to 6 years (6 mm Hg), 7 to 10 years (9 mm Hg), and 11 to 14 years (13 mm Hg). ROC curves were plotted for each age group. To calculate the sensitivity and specificity of these age-related thresholds for reduced PbtO2, threshold ICP values for each group were designated by the symbol X (therefore X = 6, 9, or 13 mm Hg depending on the age category). PbtO2 was dichotomized at 10 mm Hg. These dichotomized data for each patient were then sorted for (1) ICP, X, and ICP $ X, (2) ICP, (X 1 5) and ICP $ (X 1 5), and (3) ICP, (X 1 10) and ICP $ (X 1 10). Therefore, data were examined using 3 ICP levels: age threshold, age threshold 1 5 mm Hg, and age threshold 1 10 mm Hg. Relationship Between ICP and PbtO2 in Individual Patients Because TBI may have significant interindividual heterogeneity and because examination of data in a large cohort may mask significant relationships seen in individual patients, additional ROC curves for individual patients were constructed using hourly and electronic data (where available). Relationship Between ICP and PbtO2 in Individual Patients at Selected Time Points Because of temporal heterogeneity in TBI, it is possible that the relationship between ICP and PbtO2 when averaged over a long time period, even in a single patient, may appear weak, but may demonstrate a stronger relationship when examined over a more limited time frame. Therefore, electronic data trends for individual patients were reviewed to select examples of the dynamic relationship between ICP and PbtO2. Data were manually reviewed to exclude artifacts and any observations related to external manipulation. RESULTS Data for 81 patients who underwent ICP and PbtO2 monitoring for severe TBI from June 2006 to June 2009 were reviewed. Sixteen percent of the data (1601 hours) were excluded for the following reasons: poor placement or penumbral probe location (n = 3 patients), the patient was terminal on admission (n = 3 patients), data were recorded after brain death was diagnosed or PbtO2 was 0, data were recorded during dynamic tests such as autoregulation and hyperoxia tests, and data were missing, in part because of unintentional dislodging of one of the probes. Exclusion of data was distributed across the cohort. This left 75 patients in the study group; the general demographic, clinical, and physiological data for these patients are displayed in Tables 1 and 2. Of these patients, 15 had monitors placed in the more-injured hemisphere, and the rest had monitoring either of the less injured hemisphere or right frontal as per standard 1222 VOLUME 70 NUMBER 5 MAY

4 INTRACRANIAL PRESSURE AND BRAIN OXYGENATION protocol in patients who had diffuse brain injury, which made up the majority of patients in this cohort. Time-Linked Relationship Between ICP and PbtO2 General Relationship Between ICP and PbtO2 Spearman s correlation using 8389 paired observations of hourly data in 75 patients (group A) for ICP and PbtO2 revealed a weak correlation between ICP and PbtO2 (r = 0.05, P,.01). The scatterplot for this relationship is shown in Figure 1. Logistic regression analysis showed that higher values of ICP were associated with a lower probability of PbtO2 of 10 mm Hg or greater; however, this relationship also was weak (OR, 0.93; 95% CI, ). In addition, when the GEE model was used, no significant relationship was observed between ICP and PbtO2 (P =.86). Results from Spearman s correlation for data analyzed using the continuous data set in 30 patients (group B,.1 million time points) were similar (r = 0.04, P,.05). When we examined only periods of hypotension (defined as mean arterial pressure less than the 5th percentile for age and sex; n = data points from the continuous data, group B), the r value for this relationship was slightly higher (r = 0.19, P,.05). Whether patients had a craniotomy or craniectomy performed made no significant difference in the results. Ten patients underwent a craniotomy or craniectomy before monitoring was started, and TABLE 1. Admission Clinical and Demographic Characteristics (N = 75) a Characteristic Value Age, y (range) (4 mo-14 y) Sex Male 49 Female 26 Postresuscitation GCS score, median (range) 6 (3-8) Motor component score on GCS, 4 (1-5) median (range) Pupil reaction on admission, no. (%) Bilaterally reactive 61 (81) Unilaterally nonreactive 4 (5.3) Bilaterally nonreactive 10 (13.3) Mechanism of injury, no. (%) MVA passenger 17 (22.6) MVA pedestrian 43 (57.3) Crush injury 3 (4) Gunshot wound 4 (5.3) Fall from height 1 (1.3) Stab to head 2 (2.6) NAI 2 (2.6) Blunt head injury/assault 2 (2.6) Other 1 (1.3) Polytrauma 46 (61.3) a GCS, Glasgow Coma Scale; MVA, motor vehicle accident; NAI, nonaccidental injury. Values are expressed as mean 6 standard deviation (or median and range where specified) or as numbers and percentages. Polytrauma was defined as all injuries to organ systems other than the head, excluding superficial wounds. TABLE 2. Summary of Physiologic Monitored Variables for Total Monitored Time a Variable Value PbtO2, mm Hg (0-98.9) b MAP, mm Hg ICP, mm Hg (0-100) CPP, mmhg PaO 2, mm Hg 128 (92-188) (32-586) PaCO 2,mmHg SaO 2, % 99 (98-100) range FiO 2, % 40 (40-60) range Hemoglobin, g/dl Serum sodium, mmol/l a CPP, cerebral perfusion pressure; FiO 2, inspired fraction of oxygen; ICP, intracranial pressure; MAP, mean arterial pressure; PbtO2, partial pressure of brain tissue oxygen; SaO2, pulse oximetry. Values reported as mean 6 standard deviation or median (interquartile range) and range depending on distribution characteristics of observations. Descriptive data calculated for 9452 observations, terminal data beyond first hour PbtO2 = 0 excluded. b The extreme value of 98.9 mm Hg for PbtO2 occurred in 1 patient after cardiac arrest and resuscitation. a further 8 had surgery during the course of their monitoring. When we excluded these patients (excluded the former completely and excluded all data collected after surgery for the latter), the r value for the remaining patients was 0.02 (P..05). There was no significant relationship observed between the ICP/PbtO2 correlation and outcome according to the Glasgow Outcome Score (P =.31). When we separately examined patients based on where the monitor had been placed, the r value for the ICP/PbtO2 relationship when the monitor was in the more injured hemisphere (n = 15) was 0.27 (P,.01), and in typical diffuse injury or the less injured hemisphere (n = 60) the r value was 0.01 (P =.67).To FIGURE 1. Scatterplot of the correlation between intracranial pressure and brain oxygenation. Scatterplot of the 8389 paired observations of intracranial pressure (ICP) and partial pressure of brain tissue oxygen (PbtO2). r value (P value) are recorded at the top of the figure. NEUROSURGERY VOLUME 70 NUMBER 5 MAY

5 ROHLWINK ET AL examine the influence of time after injury, we chose 12-hour periods (from continuous data set, group B) on day 1 (first 12 hours of monitoring) and day 4 (72-84 hours). The r values were 0.07 and 20.06, respectively. We examined the issue of whether a lag phase in the PbtO2 response possibly influenced our results by analyzing the continuous data for 2 patients and offset the PbtO2 values by 10 and 20 seconds. There was no difference in the result: r = 0.19/0.18/0.18 (normal/10-second/20-second offset) and r = 20.02/20.03/20.03, respectively. Relationship Between ICP and PbtO2 Controlled for Other Physiological Variables Linear regression analysis using ICP and PbtO2 data recorded concurrently with arterial blood gas observations (n = 400) did not demonstrate a relationship between PbtO2 and ICP (r = 20.11, P =.27). This analysis controlled for blood pressure, PaO 2,PaCO 2, and hemoglobin. Results from the GEE models were similar (Table 3). Figure 2 shows scatterplots for PbtO2 and CPP (using the continuous data set of group B) and PbtO2 against PaO 2 and PaCO 2 (using the arterial blood gas data set). Note that the r values of the scatterplots are different from the GEE values in Table 3 because the latter accounts for interindividual differences. Although PbtO2 is known to be strongly influenced by PaO 2,ther value for this relationship was still relatively small, in part presumably because higher inspired fractions of oxygen were used as a treatment for low PbtO2 and was also more likely in patients who may have had increased ICP. The negative r value (nonsignificant) with PaCO 2 is interesting; however, this does not take into account interindividual differences. Also, even though increased CO 2 is associated with cerebral vasodilation, the effect on ICP caused by increased cerebral blood volume may compromise PbtO2. Critical Threshold Values of ICP and PbtO2 Table 4 shows the frequencies of PbtO2 data in each ICP category. The distribution of values for PbtO2 less than 5 mm Hg appears skewed to ICP values greater than 40 mm Hg, but these represent data from 3 patients and the total number of values for PbtO2 less than 5 mm Hg account only for 0.7% of the total data. Table 5 shows the sensitivities, specificities, and relative risks of ICP thresholds for reduced PbtO2. In general, the TABLE 3. Generalized Estimating Equation Results for Brain Tissue Oxygen and Several Variables a Variable Coefficient (P) ICP (P =.6) MAP 0.19 (P,.01) PaO (P,.01) PaCO (P,.01) Hb 0.45 (P =.4) a Hb, hemoglobin; ICP, intracranial pressure; MAP, mean arterial pressure; PbtO2, partial pressure of brain tissue oxygen. relative risk for reduced PbtO2 increased as the ICP threshold increased. The ROC curves for ICP in general demonstrated that the curves lay mostly close to the diagonal line, depicting a limited overall relationship between ICP and PbtO2. The 3 ROC curves are shown in Figure 3. Using 1-mm Hg increments for the PbtO2 thresholds between 10 and 20 mm Hg in the ROC curve analysis did not change the result; no threshold yielded an area under the curve greater than Age-Related Thresholds An age-related ICP threshold of X (6, 9, or 13 mm Hg, depending on the age category) detected 90.4% of PbtO2 episodes less than 10 mm Hg, ie 90% of observations of PbtO2, 10 mm Hg occurred when ICP was above the age threshold, but this only had a specificity of 18.7%. This is possibly because the chosen ICP thresholds are very low. However, these are data that were reported for suggested age-corrected ICP thresholds in children. 28 Consistent with this, 81% of all ICP observations occurred at ICP values greater than the age threshold. ICP = (X 1 5) (11, 14, 18 mm Hg) had a sensitivity of 63.3% and a specificity of 46.3% for detecting episodes of PbtO2 less than 10 mm Hg. ICP = (X 1 10) (16, 19, or 23 mm Hg) detected 45.6% of cases with a specificity of 74.9%. The age-related ROCs, similar to those derived for the entire cohort, demonstrated a limited relationship between ICP and PbtO2. ICP and PbtO2 in Individual Patients ROC curves for individual patients showed significant variation but in general were not better than ROC curves for pooled data. For example, ROC curves for 2 individual patients revealed that the one patient had an ROC curve that lay above the diagonal (area = 0.64) and the other had an ROC curve below the diagonal (area = 0.27). These differences were typical of many patients. These data suggest that even for individual patients, the relationship between ICP and PbtO2 is complex. Case Illustrations Examination of electronic data revealed several different patterns in the relationship between ICP and PbtO2. A selected case (Figure 4) shows data for ICP, PbtO2, and mean arterial pressure recorded over 3 hours in a child with diffuse TBI and illustrates some of the variability in the relationship between ICP and PbtO2. Figure 4A shows a linear positive relationship between blood pressure and both ICP and PbtO2, as seen with impaired pressure autoregulation. Figure 4B demonstrates a linear positive relationship between ICP and PbtO2 because they both increase despite stable blood pressure. When the ICP subsequently increases to higher values, the relationship appears to change to a negative one (Figure 4C). This is followed by a much greater increase in ICP, which is associated with a significant reduction in PbtO2 (Figure 4D). This single case illustrates the temporal heterogeneity in TBI pathophysiology and the changing relationship between various physiological variables. Similar 1224 VOLUME 70 NUMBER 5 MAY

6 INTRACRANIAL PRESSURE AND BRAIN OXYGENATION FIGURE 2. Scatterplots for several variables plotted against PbtO2. A, scatterplot of time-linked cerebral perfusion pressure (CPP) and PbtO2 using continuous data set (group B); B, C, scatterplots from arterial blood gas data set for partial pressure of brain tissue oxygen (PbtO2) plotted against PaO 2 and PaCO 2. observations were identified in many other patients, emphasizing the importance of continuous monitoring of various parameters to best understand the pathophysiology in individual patients over time. This patient made a good clinical recovery (Glasgow Outcome Score 5). Several other episodes in different patients are shown in Figure 5 as further examples of physiological variability. Figure 5A shows a typical negative response of PbtO2 to several large and sudden increases in ICP that were poorly controlled. The PbtO2 decreases immediately and to very low levels, highlighting the necessity of aggressive control of ICP needed in this patient. Figure 5B shows the trace of an 11-year-old boy and how even a moderate increase in ICP, well below the recommended treatment threshold, can affect PbtO2. Arguably, control of ICP at a lower threshold may benefit the patient, depending on the range of the PbtO2 decrease; however, if the PbtO2 change is within normal limits, perhaps this is an acceptable change that needs no intervention. In contrast to these inverse relationships TABLE 4. Frequency Table for Dichotomized Intracranial Pressure and Brain Oxygenation Data a PbtO2 ICP, mm Hg, $20 Totals, X X X $ Totals a ICP, intracranial pressure; PbtO2, partial pressure of brain tissue oxygen. Tabulation of the number of observations in ICP and PbtO2 categories. X denotes no observations in that category. between ICP and PbtO2, Figure 5C shows a parallel increase in both ICP and PbtO2 that was not related to a change in blood pressure and for which the cause was not identifiable. Figure 5D shows a steady increase in ICP that is associated initially with a steady PbtO2 increase. The PbtO2 direction then changes and becomes negative, presumably because the ICP has crossed a particular threshold (the line demonstrated in Figure 5D), after which PbtO2 is influenced by the increasing tissue pressure effect. This may be helpful in recognizing a threshold at which TABLE 5. Sensitivities and Specificities Calculated for Dichotomized Data of Intracranial Pressure Thresholds Against Various Categories of Reduced Brain Tissue Oxygen a ICP (mm Hg) Sensitivity Specificity Relative Risk PbtO2, (82-98) (22-24) 4.05 ( ) (65-88) (54-56) 4.53 ( ) (50-76) (80-82) 7.61 ( ) (34-61) 92.7 (92-93) ( ) (34-61) 97.1 (96-97) ( ) (34-61) 99.3 (99-99) ( ) PbtO2, (75-88) (22-24) 1.46 ( ) (52-68) 55.6 (54-56) 1.94 ( ) (34-50) (80-82) 3.09 ( ) (18-32) (92-93) 4.0 ( ) (15-29) (96-97) 8.65 ( ) (13-26) (99-99) ( ) PbtO2, (75-80) (23-25) 1.13 ( ) (48-54) (55-57) 1.29 ( ) (26-32) (81-83) 1.76 ( ) (12-16) (92-93) 2.0 ( ) (4-7) (96-97) 2.02 ( ) (2-4) (99-99) 3.36 ( ) a ICP, intracranial pressure; PbtO2, partial pressure of brain tissue oxygen. Values are reported with 95% confidence intervals. NEUROSURGERY VOLUME 70 NUMBER 5 MAY

7 ROHLWINK ET AL FIGURE 3. A-C, Receiver-operating characteristic curves (ROC) for PbtO2 categories less than 20 mm Hg, less than 10 mm Hg, and less than5 mm Hg. AUC, area under the curve. the ICP becomes an influencing factor on PbtO2. However, in our experience, we see this commonly, but the threshold varies between patients and even in an individual at different times. Figure 5E shows an episode of increased ICP in which the PaCO 2 was decreased as a test of CO 2 responsiveness (decreased by 0.7 kpa). At the time of the CO 2 decrease (first line), the ICP promptly decreases and then increases again when the PaCO 2 was restored to baseline (second line). Despite the fact that decreasing PaCO 2 leads to vasoconstriction, here the PbtO2 and cerebral blood flow increase in response, presumably because of the effect of the reduced ICP. In this patient, a lower CO 2 may be of benefit at this point, keeping in mind that the circumstances may change over time. In our experience, a decrease in PaCO 2 can have different effects on PbtO2 and cerebral blood flow, depending on the prevailing physiology. Finally, Figure 5F illustrates a patient with impaired pressure auotregulation in whom blood pressure changes are followed passively by parallel changes in ICP and PbtO2. In this patient, optimal control of blood pressure in a narrow range is of critical importance, based on the ICP and PbtO2 response. Patients with intact autoregulation typically have a different response. Of interest, the correlation (r) between ICP and PbtO2 in Figure 5A was (P,.01), and the correlation in Figure 5F was 0.9 (P,.01). DISCUSSION This study examined time-linked ICP and PbtO2 observations using several exploratory methods in 75 children with severe TBI. The main findings can be summarized as follows: (1) The overall relationship between ICP and PbtO2 is weak. Controlling for other factors that may affect PbtO2 did not improve this relationship. (2) The ICP-PbtO2 relationship varies among patients and within an individual patient over time. (3) Although a relationship between high ICP and low PbtO2 is observed at times in individual patients, specific threshold values for ICP are not predictive of low PbtO2. (4) There can be a strong negative relationship between ICP and PbtO2, but at times, there also may be a positive relationship depending on the underlying cerebral pathophysiology, such as cerebral hyperemia or impaired pressure autoregulation. Together these data suggest that the ICP-PbtO2 relationship is likely a complex relationship influenced by several factors. Although the main finding is negative, ie, a poor relationship between ICP and PbtO2, we believe this is important to further our understanding of the interaction between physiological variables in acute brain injury. From a practical standpoint, these data suggest that a simple linear relationship between ICP and PbtO2 cannot be assumed; when ICP is normal, PbtO2 may still be low, and when ICP is increased, the response of PbtO2 may be variable depending on the underlying pathophysiology and may, in some cases, be increased. Time-Linked Relationship Between ICP and PbtO2 The overall correlation between ICP and PbtO2 was weak (r = 0.05). Although the nonlinear regression line showed a negative trend only at very high values of ICP, from our clinical experience (and see Figures 4 and 5D), it is evident that increased ICP often may compromise PbtO2 and that this may occur at different ICP thresholds. During specific episodes, however, as demonstrated in our illustrations, the correlation 1226 VOLUME 70 NUMBER 5 MAY

8 INTRACRANIAL PRESSURE AND BRAIN OXYGENATION FIGURE 4. Illustration of the variable relationship between intracranial pressure (ICP) and partial pressure of brain tissue oxygen (PbtO2). High-frequency data recording (each data point represents an average of 10 seconds of 50-Hz recording) in a single patient over approximately 3.5 hours for ICP, PbtO2, and mean arterial pressure (MAP). This figure illustrates the different relationships that may be seen between ICP and PbtO2 and includes an example of the potential influence of MAP. No specific treatment for PbtO2 was administered during this period. See text for full description. between ICP can be strongly positive or negative, depending on the prevailing dynamics. Overall, however, the averaged relationship is weak, and controlling for several other systemic physiological variables showed no improvement of this result. Furthermore, ROC curves did not show a high diagnostic accuracy for any ICP level in detecting reduced PbtO2, even when age bands were taken into consideration or individual patients were examined. Although the relative risk of reduced PbtO2 did increase as the ICP threshold increased, very little of the variability in PbtO2 was attributable to ICP (R² = 0.01, in regression analysis). In general, as the ICP threshold increased, sensitivity for episodes of reduced PbtO2 decreased and specificity increased. An ICP threshold of 20 mm Hg detected only 42% of episodes of PbtO2 less than 10 mm Hg and 29% of PbtO2 less than 20 mm Hg. This suggests that in the context of normal ICP, other factors are commonly responsible for reduced PbtO2. Conversely, even when ICP was high, PbtO2 was not always reduced. In fact, when ICP was greater than 40 mm Hg (Table 4), more than half of the PbtO2 observations were still greater than 20 mm Hg. Although in part this may reflect the influence of interventions to treat low PbtO2, it appears that other factors, such as hyperemia and impaired pressure autoregulation, may cause simultaneous increases in ICP and PbtO2. Why Is the Relationship Between ICP and PbtO2 Variable and What Are the Implications? There are several possible mechanisms in TBI that potentially influence the relationship between ICP and PbtO2. First, some factors may cause the expected opposite changes in ICP and PbtO2. For example, cerebral edema may increase ICP and decrease PbtO2 by the reduction in CPP or by increasing the diffusion distance of oxygen from the capillary to the cell. However, this relationship of increased ICP and reduced PbtO2 is not inevitable: our previous work suggests that almost one third of episodes of PbtO2 less than 10 mm Hg occur when patients appear to be optimally managed according to current guidelines. 18 Second, several physiological variables may influence PbtO2 independently of ICP, such as systemic oxygenation, cerebral vasospasm, and increased demand. Third, other factors may cause parallel changes in ICP and PbtO2, such as hyperemia, hypercarbia, and increased blood pressure (when pressure autoregulation is impaired). When the brain is more dysregulated, it is likely that the relationship between ICP and NEUROSURGERY VOLUME 70 NUMBER 5 MAY

9 ROHLWINK ET AL FIGURE 5. A, negative relationship between intracranial pressure (ICP) and partial pressure of brain tissue oxygen (PbtO2). B, negative relationship at a low value of ICP. C, positive relationship between ICP and PbtO2. D, positive, then negative relationship between ICP and PbtO2. E, CO 2 decrease test. F, impaired autoregulation. MAP, mean arterial pressure; CBF, local cerebral blood flow (as measured with Hemedex and Bowman perfusion monitor). See text for a full discussion. PbtO2 (and blood pressure) may be more strongly positive, certainly at lower levels of ICP. The results from the more injured hemisphere, in which the r value was higher (and significant despite a smaller sample), is in keeping with this. However, even in these patients the r value remains small, presumably because of the many other factors affecting this relationship. An ICP monitor is recommended in patients with severe TBI and is considered a standard in most intensive care units. PbtO2 monitors were included in the adult Guidelines for Severe TBI in but are not yet standard of care. Emerging data from several centers suggest, however, that reduced PbtO2 is strongly associated with poor outcome in adults and children 11-13,15,29,30 and that PbtO2 monitoring 1228 VOLUME 70 NUMBER 5 MAY

10 INTRACRANIAL PRESSURE AND BRAIN OXYGENATION and treatment may benefit adult severe TBI patients when combined with ICP control. 17,31,32 The use of monitors of brain oxygenation, brain perfusion, and metabolism therefore has increased. 10,24,33-37 Appropriate use of these monitors may help interpret the pathophysiology underlying increased ICP in individual patients and help direct both the method of treating ICP and the threshold at which ICP should be treated. This may be of particular value in children for whom there is little evidence of appropriate thresholds for ICP treatment given their changing physiological profiles with age. 4 Study Limitations There are several potential limitations of this study. First, this was a retrospective study in a population treated for high ICP and reduced PbtO2; untreated ICP and PbtO2 may assume different relationships; however, this would be ethically unjustifiable and the results of this study therefore are relevant to patients being actively treated. This is a limitation of studying almost all variables in TBI and their relationships both with outcome and with other variables because there are very few monitors or results that we do not respond to. Although normal or high PbtO2 in the face of high ICP may be a treatment effect, in our experience, this is often not the case, and there may be several physiological reasons that explain a variable relationship between ICP and PbtO2. We have observed many patients in which ICP and PbtO2 simultaneously increase, which may reflect a hyperemic process; at other times, an increase in blood pressure has increased both ICP and PbtO2 in parallel in patients with impaired pressure autoregulation. Based on our clinical experience, we expect that even in the absence of any treatment for PbtO2, a similar result between ICP and PbtO2 would be obtained. Second, although age was considered in 3 bands, there are no robust data on what normal ICP is for all age categories or what constitutes significantly increased ICP for age. Despite the general belief that younger children should have ICP treated at lower thresholds, there is no evidence currently to develop firm recommendations for this practice. 23 Third, although all values of PbtO2 and ICP were time linked and therefore comparable, the use of hourly data may be limiting. However, analysis of data from individual patients with high-frequency data collection using more than 1 million time points were consistent with hourly analysis and revealed varying relationships between ICP and PbtO2, and ROC curves constructed with electronic data were not better than those using hourly data (data not shown). Fourth, the PbtO2 monitor used in this study examines a focal area of brain tissue. It is possible that data from a global measure of brain oxygenation may reveal different results. However, no reliable and continuous measure of global brain oxygenation currently exists, and there is some evidence to suggest that PbtO2 measured in relatively uninjured white matter reflects global changes. 38,39 Fifth, PbtO2 may be affected by several variables. PbtO2 measures brain tissue oxygen tension and is therefore strongly influenced not only by cerebral blood flow (and other factors that influence cerebral blood flow), but also by arterial oxygen tension and the diffusion of oxygen through brain tissue. 40,41 Although PbtO2 can decrease as CBF decreases, a PbtO2 monitor is not an ischemia monitor. 39,42-45 This may explain in part the dynamic variability in TBI, as there are many factors that influence ICP, brain perfusion, brain oxygenation, and brain metabolism. Our results are consistent with this. Still, even in the presence of many such factors, if ICP had a standard relationship with PbtO2, we believe that this would be evident in such a large analysis. Also, controlling for several other variables that we would expect to influence PbtO2 did not improve the results, nor did controlling for interindividual variability. Many other factors should be examined in detail with respect to their relationship to PbtO2, for example CPP. CPP is often considered to be as important as, or even more important than, ICP in children with TBI. Correlation between CPP and PbtO2 in this cohort was also weak, and we intend to examine this relationship in greater detail in a future study. Sixth, these results are relevant only to a pediatric population and may not apply to adult patients because there are differences in cerebrovascular pathophysiology in TBI between adults and children. 35,46 However, a plot of the simple relationship between ICP and PbtO2 in 27 adult patients (n = 2821 observations) 29 showed a distribution similar to that of our results. In that study, even though high ICP increased the relative risk of low PbtO2, the number of observations of reduced PbtO2 was still 3 times greater when ICP was reduced than when it was high. Finally, several factors discussed here, such as autoregulation, time after injury and age, and perhaps other factors, such as cardiovascular and respiratory instability, may affect this relationship, and these are all worth detailed separate examination that is beyond the scope of this article, but highlights the heterogeneity of the population and informs further studies. CONCLUSION There is no Class I evidence to suggest that control of ICP or PbtO2 or the use of these monitors should be a standard of care for all severe TBI. Nevertheless, an ICP monitor often is considered standard in centers that treat TBI patients, and the frequency of PbtO2 monitoring is increasing. Our results suggest that the relationship between ICP and PbtO2 is variable and is likely dependent on several complex interactions. What determines PbtO2 and how best to treat compromised PbtO2 are only beginning to be elucidated. 32,40,41 Randomized, controlled trials are now planned for both ICP- and PbtO2-based care. However, until these studies are performed, we suggest that more data are required to understand the causes and nature of increased ICP and reduced PbtO2 and that current strategies to treat both need ongoing review. Perhaps the data also suggest how challenging these randomized, controlled trials may be, given the heterogeneity and complexity demonstrated in these patients. Disclosures Drs Figaji and le Roux have received funding from the Integra Foundation for the study of cerebral perfusion pressure in children in Dr le Roux also received NEUROSURGERY VOLUME 70 NUMBER 5 MAY

11 ROHLWINK ET AL research support from Integra and is a consultant for Codman and a member of Edge Therapeutics Scientific Advisory Board. Dr Figaji received honoraria for speaking engagements from Codman and Integra Neurosciences and funding from the National Research Foundation. Ms. Rohlwink receives funding from the Wellcome Trust. The other authors have no personal financial or institutional interest in any of the drugs, materials, or devices described in this article. REFERENCES 1. Ghajar J. Traumatic brain injury. Lancet. 2000;356(9233): Maas AI, Stocchetti N, Bullock R. Moderate and severe traumatic brain injury in adults. Lancet Neurol. 2008;7(8): Elf K, Nilsson P, Enblad P. Outcome after traumatic brain injury improved by an organized secondary insult program and standardized neurointensive care. Crit Care Med. 2002;30(9): Adelson PD, Bratton SL, Carney NA, et al. Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. Chapter 17. Critical pathway for the treatment of established intracranial hypertension in pediatric traumatic brain injury. Pediatr Crit Care Med. 2003;4 (3 suppl):s65-s Chesnut RM. Secondary brain insults after head injury: clinical perspectives. New Horiz. 1995;3(3): Chesnut RM. Care of central nervous system injuries. Surg Clin North Am. 2007; 87(1): ; vii. 7. Tang ME, Lobel DA. Severe traumatic brain injury: maximizing outcomes. Mt Sinai J Med. 2009;76(2): Graham DI, Ford I, Adams JH, et al. Ischaemic brain damage is still common in fatal non-missile head injury. J Neurol Neurosurg Psychiatry. 1989;52(3): Hlatky R, Valadka AB, Robertson CS. Intracranial hypertension and cerebral ischemia after severe traumatic brain injury. Neurosurg Focus. 2003;14(4):e Lang EW, Mulvey JM, Mudaliar Y, Dorsch NW. Direct cerebral oxygenation monitoring a systematic review of recent publications. Neurosurg Rev. 2007;30 (2):99-106; discussion Maloney-Wilensky E, Gracias V, Itkin A, et al. Brain tissue oxygen and outcome after severe traumatic brain injury: a systematic review. Crit Care Med. 2009;37(6): Figaji AA, Zwane E, Thompson C, et al. Brain tissue oxygen tension monitoring in pediatric severe traumatic brain injury. Part 1: relationship with outcome. Childs Nerv Syst. 2009;25(10): Valadka AB, Gopinath SP, Contant CF, Uzura M, Robertson CS. Relationship of brain tissue PO2 to outcome after severe head injury. Crit Care Med. 1998;26(9): Brain Trauma Foundation. American Association of Neurological Surgeons. Congress of Neurological Surgeons, et al. Guidelines for the management of severe traumatic brain injury. X. Brain oxygen monitoring and thresholds. J Neurotrauma. 2007;24(suppl 1):S65-S van den Brink WA, van Santbrink H, Steyerberg EW, et al. Brain oxygen tension in severe head injury. Neurosurgery. 2000;46(4): ; discussion Chang TP, Nager AL. Pediatric traumatic brain injury: the utility of betanatriuretic peptide. J Trauma. 2010;68(6): Stiefel MF, Udoetuk JD, Spiotta AM, et al. Conventional neurocritical care and cerebral oxygenation after traumatic brain injury. J Neurosurg. 2006;105(4): Figaji AA, Fieggen AG, Argent AC, Leroux PD, Peter JC. Does adherence to treatment targets in children with severe traumatic brain injury avoid brain hypoxia? A brain tissue oxygenation study. Neurosurgery. 2008;63(1):83-91; discussion Ho CL, Wang CM, Lee KK, Ng I, Ang BT. Cerebral oxygenation, vascular reactivity, and neurochemistry following decompressive craniectomy for severe traumatic brain injury. J Neurosurg. 2008;108(5): Figaji AA, Fieggen AG, Argent AC, Le Roux PD, Peter JC. Intracranial pressure and cerebral oxygenation changes after decompressive craniectomy in children with severe traumatic brain injury. Acta Neurochir Suppl. 2008;102: Adelson PD, Bratton SL, Carney NA, et al. Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. Chapter 1: introduction. Pediatr Crit Care Med. 2003;4(3 suppl):s2-s Adelson PD, Bratton SL, Carney NA, et al. Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. Chapter 5. Indications for intracranial pressure monitoring in pediatric patients with severe traumatic brain injury. Pediatr Crit Care Med. 2003;4(3 suppl):s19-s Adelson PD, Bratton SL, Carney NA, et al. Guidelines for the acute medical management of severe traumatic brain injury in infants, children, and adolescents. Chapter 6. Threshold for treatment of intracranial hypertension. Pediatr Crit Care Med. 2003;4(3 suppl):s25-s Wartenberg KE, Schmidt JM, Mayer SA. Multimodality monitoring in neurocritical care. Crit Care Clin. 2007;23(3): Maloney-Wilensky E, Le Roux P. The physiology behind direct brain oxygen monitors and practical aspects of their use. Childs Nerv Syst. 2010;26(4): Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem. 1993;39(4): Chambers IR, Treadwell L, Mendelow AD. Determination of threshold levels of cerebral perfusion pressure and intracranial pressure in severe head injury by using receiver-operating characteristic curves: an observational study in 291 patients. J Neurosurg. 2001;94(3): Chambers IR, Jones PA, Lo TY, et al. Critical thresholds of intracranial pressure and cerebral perfusion pressure related to age in paediatric head injury. J Neurol Neurosurg Psychiatry. 2006;77(2): Chang JJ, Youn TS, Benson D, et al. Physiologic and functional outcome correlates of brain tissue hypoxia in traumatic brain injury. Crit Care Med. 2009;37 (1): van Santbrink H, Maas AI, Avezaat CJ. Continuous monitoring of partial pressure of brain tissue oxygen in patients with severe head injury. Neurosurgery. 1996;38 (1): Narotam PK, Morrison JF, Nathoo N. Brain tissue oxygen monitoring in traumatic brain injury and major trauma: outcome analysis of a brain tissue oxygen-directed therapy. J Neurosurg. 2009;111(4): Spiotta AM, Stiefel MF, Gracias VH, et al. Brain tissue oxygen-directed management and outcome in patients with severe traumatic brain injury. J Neurosurg. 2010;113(3): Narotam PK, Burjonrappa SC, Raynor SC, Rao M, Taylon C. Cerebral oxygenation in major pediatric trauma: its relevance to trauma severity and outcome. J Pediatr Surg. 2006;41(3): Goodman JC, Robertson CS. Microdialysis: is it ready for prime time? Curr Opin Crit Care. 2009;15(2): Charalambides C, Sgouros S, Sakas D. Intracerebral microdialysis in children. Childs Nerv Syst. 2010;26(2): Vajkoczy P, Roth H, Horn P, et al. Continuous monitoring of regional cerebral blood flow: experimental and clinical validation of a novel thermal diffusion microprobe. J Neurosurg. 2000;93(2): Jaeger M, Soehle M, Schuhmann MU, Winkler D, Meixensberger J. Correlation of continuously monitored regional cerebral blood flow and brain tissue oxygen. Acta Neurochir (Wien). 2005;147(1):51-56; discussion Carmona Suazo JA, Maas AI, van den Brink WA, van Santbrink H, Steyerberg EW, Avezaat CJ. CO2 reactivity and brain oxygen pressure monitoring in severe head injury. Crit Care Med. 2000;28(9): Scheufler KM, Lehnert A, Rohrborn HJ, Nadstawek J, Thees C. Individual value of brain tissue oxygen pressure, microvascular oxygen saturation, cytochrome redox level, and energy metabolites in detecting critically reduced cerebral energy state during acute changes in global cerebral perfusion. J Neurosurg Anesthesiol. 2004;16(3): Menon DK, Coles JP, Gupta AK, et al. Diffusion limited oxygen delivery following head injury. Crit Care Med. 2004;32(6): Rosenthal G, Hemphill JC III, Sorani M, et al. Brain tissue oxygen tension is more indicative of oxygen diffusion than oxygen delivery and metabolism in patients with traumatic brain injury. Crit Care Med. 2008;36(6): Hlatky R, Valadka AB, Goodman JC, Contant CF, Robertson CS. Patterns of energy substrates during ischemia measured in the brain by microdialysis. J Neurotrauma. 2004;21(7): Huchzermeyer C, Albus K, Gabriel HJ, et al. Gamma oscillations and spontaneous network activity in the hippocampus are highly sensitive to decreases in po2 and concomitant changes in mitochondrial redox state. J Neurosci. 2008;28(5): Valadka AB, Hlatky R, Furuya Y, Robertson CS. Brain tissue PO2: correlation with cerebral blood flow. Acta Neurochir Suppl. 2002;81: Gupta AK, Hutchinson PJ, Al-Rawi P, et al. Measuring brain tissue oxygenation compared with jugular venous oxygen saturation for monitoring cerebral oxygenation after traumatic brain injury. Anesth Analg. 1999;88(3): Giza CC, Mink RB, Madikians A. Pediatric traumatic brain injury: not just little adults. Curr Opin Crit Care. 2007;13(2): VOLUME 70 NUMBER 5 MAY