Medicine, P.O. Box 100254, Gainesville, FL 32610-0254, U.S.A.. Sidi A, Messinger G, ...... Pourcelot L. Diagnostic ultrasound for cerebral vascular diseases.
TRANSCRANIAL DOPPLER MONITORING COMPARED WITH INVASIVE MONITORING OF INTRACRANIAL PRESSURE DURING ACUTE INTRACRANIAL HYPERTENSION Avner Sidi, MD,1 Gabriel Messinger, MD,2; 4 and Michael E. Mahla, MD 3
From the 1 Department of Anesthesiology, 2 Division of Critical Care, 3 Department of Neurosurgery, University of Florida College of Medicine, Gainesville, FL, U.S.A. 4 Dr Messinger is currently a¤liated with the Department of Anesthesiology, The Chaim Sheba Medical Center, Tel-Hashomer 52621, Israel.
Sidi A, Messinger G, Mahla ME. Transcranial Doppler monitoring compared with invasive monitoring of intracranial pressure during acute intracranial hypertension. J Clin Monit 1999; 15: 185^195
ABSTRACT. Objective. To determine whether a simple transcranial Doppler waveform variable^pulsatility di¡erence (systolic - diastolic blood £ow velocity) can serve as a measure of critical changes in cerebral perfusion. Methods. Thirteen pigs were anesthetized (anesthesia maintained with halothane) and ventilated to maintain normoxia and normocarbia. To measure mean arterial pressure, hemoglobin, and blood gases, the right carotid artery was cannulated. The right intracranial lateral ventricle was cannulated to measure and increase intracranial pressure; the right internal jugular vein was cannulated in 8 of 13 pigs to measure jugular venous oxygen saturation and to calculate cerebral arteriovenous oxygen content di¡erence. Intracranial pressure was also monitored continuously with a subdural bolt in the contralateral frontal region, and blood £ow velocity in the middle cerebral artery was measured with a transcranial Doppler probe on the right orbital region. Intracranial pressure was increased in increments of 10 to 20 mmHg by infusing saline through the ventriculostomy catheter until the transcranial Doppler indicated that blood £ow velocity had ceased, at which point all variables were allowed to return to baseline. If mean arterial pressure failed to return to baseline, epinephrine, 0.01 to 0.1 mg/kg/min, was infused. Useful data were obtained from 8 pigs and were analyzed separately for pigs that received epinephrine (n = 4) and those that did not (n = 4). Results. Transcranial Doppler measurements correlated more closely with cerebral perfusion pressure = (mean arterial pressure ÿ intracranial pressure) than with intracranial pressure. In the range of 30 to 60 mmHg, cerebral perfusion pressure correlated linearly with the pulsatility di¡erence. The closest nonlinear correlation (third order polynomial relationship) was noted between cerebral perfusion pressure and pulsatility di¡erence (r = 0.8, P < 0.001, n = 217), for the animals that did not receive epinephrine. When a cerebral perfusion pressure 6.5 vol% were used to de¢ne limits of abnormal, pulsatility di¡erence was a sensitive and speci¢c indicator of abnormality in either variable. Pulsatility di¡erence of >70 cm/sec had >77.1% and 86.7% positive accuracy rate, and 60 mmHg, PI increased disproportionately with changes in ICP due to increased sys , decreased dia , and a decreased mean [6]. Pulsatility index was found to have some linear correlation with ICP and total cerebral tissue resistance [6], better than other transcranial Doppler variables used for the same purpose, such as the area under the curve, resistance index (PD/sys ), or sys , dia , or mean [4, 10^ 13]. However, although previous studies with animals or humans documented a relationship between PI and ICP, results have been inconclusive [2^4, 14^17]. Also, the automated calculation of PI by the TCD monitor can be erroneous [3, 16^19]. Even though the PI may appear to be the superior variable, the problems associated with measuring or calculating it using mean measurement directed us and other investigators [4], to evaluate another variable. A simple calculation of the manually derived PD (PD = sys ÿ dia ) was found to have a clear and certain non linear (third order polynomial) relationship with cerebral perfusion pressure (CPP) [4]. Because CPP is a better indicator of perfusion than ICP, the ¢rst step in determining whether PD can serve clinically as a measure of brain perfusion is to determine whether PD correlates with CPP. Besides the PD-CPP relationship, it is also important to establish the relationship between PD and other direct indicators of cerebral perfusion such as cerebral blood £ow (CBF), and oxygen metabolic extraction or cerebral arteriovenous oxygen content di¡erence (C[a-v]O2 ) [20^23]. Venous oxygen saturation in the jugular bulb (SjvO2 ) and C(a-v)O2 derived from SjvO2 re£ect the ratio of global cerebral oxygen supply to demand [20, 24^26]. These indicators have been used to indicate neurologic status [27], and allow detection of hypoperfusion [23, 28], low oxygen supply [29, 30], and increased ICP [31]. Low C(a-v)O2 re£ecting low O2 supply demand ratio can precede tissue damage. However, there is not enough information about the
correlation between SjvO2 or C(a-v)O2 and Doppler cerebral £ow velocity indices (PI or PD) to understand the potential clinical utility of these non-invasive indices. This study was designed to determine in a porcine model of acute di¡use intracranial hypertension the relationship between transcranial Doppler variables (primarily PD) and cerebral pressures, and between transcranial Doppler variables and cerebral metabolic supply and demand. We tested the hypothesis that, up to the upper limit of autoregulation, both cerebral C(a-v)O2 and CPP correlate with middle cerebral artery PD [2, 4, 6, 32]. The data were also used to determine the sensitivity and speci¢city of Doppler measurements for indicating signi¢cant changes in cerebral perfusion using a value of 60 mmHg for the lower threshold of cerebral perfusion and a value of > 6.5% for the lower threshold of C(a-v)O2. MATERIALS AND METHODS This study was approved by the institutional review board, and facilities in which the experimental animals were housed were fully accredited by the American Association for Accreditation for Laboratory Animal Care. Thirteen pigs (weight, 31.8 to 47.7 kg) were anesthetized with ketamine, 10 mg/kg intramuscularly, and placed supine; the trachea was intubated through a tracheostomy with a cu¡ed endotracheal tube. The animals were ventilated to maintain an arterial oxygen tension (PaO2 ) between 80 and 120 mmHg and a carbon dioxide tension (PaCO2 ) between 30 and 40 mmHg, which was con¢rmed by arterial blood gas analysis and continuous monitoring of end-tidal carbon dioxide tension (ETCO2 ) (Novametrix Medical Systems, Inc., Wallingford, CT) and arterial oxygen saturation (SpO2 ) by pulse oximetry (Model 7000, Novametrix). Anesthesia was maintained with halothane, 0.8 to 1.3% and pancuronium, 0.1 to 0.15 mg/kg, intravenously, was used for muscle relaxation. The right carotid artery was cannulated for continuous measurement of mean arterial pressure and the right external jugular vein for infusion of drugs and £uids. In 8 animals, for continuous measurement of SjvO2 , a 4-Fr intravascular optical catheter (Oximetrix Shaw Opticath, Abbott Critical Care Systems, North Chicago, IL) was advanced cephalad into the right internal jugular vein until the tip of the catheter was against the base of the skull or 15 cm of the catheter had been advanced. This technique provided catheter positioning and SjvO2 measurements comparable to those from humans [24]; the position of the catheter for intracerebral vessel location was checked
Sidi et al: Doppler Monitoring of Intracranial Pressure 187
with transient hyperventilation instead of radiologically [26]. The SjvO2 monitor was calibrated in vivo (Oximetrix 3 SO2 /CO Computer, Abbott Critical Care Systems, Chicago, IL). Needle electrodes for continuous recording of lead II electrocardiogram (ECG) were placed subcutaneously. Body temperature was measured with a rectal thermometer and kept constant with a heating mattress. Once monitors were in place, the pigs were placed in the left lateral position. A burr hole was drilled in the right frontoparietal region, and a 16-ga, 13.3-cm Te£on catheter (Angiocath, Desert Co., Sandy, UT) was inserted into the lateral ventricle for infusion of saline to induce intracranial hypertension. For continuous monitoring of ICP, a subdural bolt was inserted in the contralateral frontal region. The transcranial Doppler probe was ¢xed in place against the eye globe to maintain constant insonation depth and angle, by means of a supportive frame. The depth used for the Doppler ultrasonic wave was 40 to 50 mm. The intracranial carotid artery (internal or external, the latter being the main supply for cerebral circulation in swine), its intracranial bifurcation level, and the middle cerebral artery were identi¢ed in the Circle of Willis beyond the bifurcation. The bifurcation served as a marker [33, 34], to identify the middle cerebral artery, and then the middle cerebral artery was insonated. sys , dia , and mean in the middle cerebral artery were measured using a 2-MHz transcranial Doppler system (Transpect, Medasonics, Fremont, CA) with a power intensity of 161 mW/cm2 at a default intensity setting of 100%, a focal distance of 3.8 cm, and pulse repetition that varied from 3.8 to 10.4 kHz, depending on range and gate depth. A range-gate was used to adjust the ultrasonic focus in steps of 2.5 or 5 mm. The axial extension of the sample volume measured 13 mm with a burst width of 12 msec, which could be decreased for greater speci¢city to 7 mm and a burst width of 6 msec. Signals were recorded with a high-pass ¢lter of 150 Hz and a low-pass ¢lter of 16 kHz. After baseline values were determined, ICP was increased in increments of 10 to 20 mmHg. A constant pressure for each increment was maintained for 5 to 10 minutes by manipulating a saline infusion through the ventriculostomy catheter, and Doppler measurements were recorded at each increment until dia was no longer measurable. Data were collected 1^3 times in each minute and at the end of the 5^10 min period. ICP was then allowed to return gradually to normal (< 20 mmHg), as was MAP (>75 mmHg). If ICP failed to return to normal, cerebral spinal £uid was aspirated from the ventriculostomy catheter. If MAP failed to increase to baseline, an epinephrine solution (0.01 to 0.1
mg/kg/min) was infused until the baseline value was restored. The period from the start of the saline infusion to the return of ICP to baseline was referred to as an injury cycle, each animal undergoing 1 to 5 such cycles. Injury cycles were discontinued when ICP failed to decrease below 20 mmHg (even when cerebrospinal £uid could no longer be removed), or CPP to increase above 50 mmHg (even when epinephrine was used). After the experiment, animals were sacri¢ced using potassium chloride. ICP, MAP, and ECG were displayed and recorded continuously on a polygraph (Model 7B, Grass Instrument Co., Quincy, MA). During each injury cycle, transcranial Doppler waveforms were recorded 1 to 3 times every minute and were superimposed on the ICP and ECG waveforms for analysis. Blood hemoglobin and gas analysis were done before testing for calibration of SjvO2 and to con¢rm normoxia and normocarbia (1312 Blood Gas Manager and CO-Oximeter 282, Instrumentation Laboratory Co., Lexington, MA). Hemoglobin and arterial and venous blood gas analyses were performed before and after each test to determine C(a-v)O2. Linear regression was used to evaluate the correlation between £ow velocity variables () and cerebral pressures (ICP, CPP). Polynomial regression was used to determine the correlation between cerebral perfusion/ pressure variables (MAP, CPP) and PD or its logarithmic transformation. A multiple linear regression was used to describe the relationship of C(a-v)O2 to both CPP and PD (Table Curve, SigmaPlot Scienti¢c Graphing System, and SigmaStat Statistical Analysis System, Jandel Scienti¢c, San Rafael, CA). In this fashion, the relationship of the commonly used perfusion pressure (CPP) is compared with metabolic changes that can point to cerebral damage and, the Doppler £ow velocity relationship (PD) is compared with the same metabolic changes. The interrelationship between PD, CPP, C(a-v)O2 was described by a three-dimensional depiction of the relationship among these three variables. This 3-D reconstruction of a Cartesian graph, was generated by plotting curves describing CPP ^ C(a-v)O2 relationship, and PD ^ C(a-v)O2 relationship, each relation for a single animal (n = 4); using an x^y^z coordinate system to plot datapoints, so that each datapoint has an x, y, and z-value [CPP, PD, and C(a-v O2 ) respectively]; (Plot 50, Sigmaplot Scienti¢c Graphic system, Version 5.01, Jandel Scienti¢c, San Rafael, CA). Because the use of epinephrine was a potentially confounding factor a¡ecting cerebral perfusion independently from the direct e¡ect of high ICP on cerebral perfusion, data were analyzed separately for those that
188
Journal of Clinical Monitoring and Computing Vol 15 Nos 3^4 May 1999
did and those that did not need epinephrine (n = 4/ group). For each group, intragroup comparison of slopes and Y-intersects of the logarithmic transformed linear regression analysis was performed by weighted least square regression (Table Curve Software Statistical Analysis System, Jandel Scienti¢c, San Rafael, CA), and intergroup comparison between groups with and without epinephrine was analyzed by a Wilcoxon signed rank test. Using a CPP threshold of 60 mmHg and C(a-v)O2 threshold of 6.5% as indicators of signi¢cant intracranial pathophysiologic changes, T the positive and negative predictive value of PD was determined. A predictive value (positive and negative) of PD for changes in CPP and C(a-v)O2 was expressed as a rate of positive and negative accuracy by a technique previously described [35]. Pearson product moment test was used to evaluate the signi¢cance of correlation between the monitored and calculated variables (BP, ICP, CPP, , and O2 saturation and content). Statistical signi¢cance was set at P < 0.05.
Table 1. Linear regression analysis with correlation r-values between cerebral perfusion and intracranial pressures (mmHg) and hemodynamic and cerebral blood £ow velocity variables (cm/sec) when intracranial pressure (ICP) was increased experimentally in all pigs (n = 338) and only in those in which ICP returned to baseline spontaneously (without epinephrine) (n = 217) Pressure and £ow velocity variables
Cerebral perfusion pressure Intracranial pressure Systolic arterial pressure Diastolic arterial pressure Mean arterial pressure Peak £ow velocity Diastolic £ow velocity Pulsatility di¡erence b Pulsatility index a b
Cerebral perfusion pressure
Intracranial pressure
Without epinephrine
All pigs
Without epinephrine
All pigs
^ 0.289 a 0.908 a 0.906 a 0.948 a ^0.714 a 0.213 ^0.768 a ^0.462 a
^ ^0.163 0.769 a 0.868 a 0.870 a ^0.637 a 0.274 a ^0.681 a ^0.357 a
0.289 a ^ 0.466 a 0.606 a 0.578 a ^0.102 0.532 a ^0.407 a ^0.49 a
^0.163 ^ 0.381 a 0.301 a 0.345 a ^0.089 0.368 a ^0.296 a ^0.229 a
P < 0.001. Systolic ^ diastolic cerebral blood £ow velocity.
RESULTS Five pigs died during the experiment, one from malignant hyperthermia and four from early herniation. The pigs that died from herniation had zero or negative value of systolic (indicating reversed £ow) when ICP was high, low CPP, and did not respond to attempts to decrease ICP or increase perfusion. Data from pigs that died during the study without undergoing a complete injury cycle were not included in the analysis. The surviving pigs included all those in which SjvO2 was measured. Baseline data ( SEM) were within the normal range: MAP 82.7 1.2 mmHg; ICP 15.5 0.5 mmHg; CPP 67.2 1.1 mmHg; sys 69.1 1.1, dia 23.5 0.6; PD 53.0 0.9 cm/sec; PI 0.8 0.1 ; and C(a-v)O2 4.4 0.8 vol%. Both hemoglobin and PaCO2 were stable during the study. Hemoglobin was 9.6 2.4 g/dl (mean SD) at the start of the injury cycle and 9.3 2.4 g/dl at the end, and PaCO2 was 37 2.6 mmHg and 35 7.8 mmHg, respectively. Table 1 demonstrates that the simple linear relationship between the cerebral pressures (CPP or ICP) and di¡erent hemodynamic, perfusion, or doppler £ow velocity parameters (including the more commonly used PI), is far from being perfect. A non-linear correlation was needed here. However, linear or non-linear correlation between PD and CPP was always stronger without epinephrine than with epinephrine (Table 1, Figures 1 and 2). When the logarithmic value of PD was plotted against CPP for each animal with (Figure 3, top) and
Fig. 1. Correlation between pulsatility di¡erence (PD) (systolic diastolic cerebral blood £ow velocity) and cerebral perfusion pressure (CPP) in pigs in which CPP was decreased by increasing intracranial pressure experimentally and returned to baseline spontaneously (4 pigs) without an infusion of epinephrine; n = data points.
without epinephrine (Figure 3, bottom) by linear regression, the individual slopes, intercepts, and their 95% con¢dence limits did not di¡er signi¢cantly, but the correlation (r-value) was lower with epinephrine (Figure 3, top). The slopes and intercepts of the regression analysis between the logarithmic value of PD and CPP of the grouped (rather than the individual) data did not di¡er signi¢cantly between groups by the use of epinephrine. The reason for using a logarithmic expression is to transform the non-linear curve to a linear one, so that linear correlations are possible.
Sidi et al: Doppler Monitoring of Intracranial Pressure 189
Table 2. Predictive value of the pulsatility di¡erance (PD) for changes in cerebral perfusion pressure (CPP) and cerebral C(a-v)O2 expressed as positive and negative accuracy test (% of measurements being ``abnormal'' or ``normal,'' respectively) in cases of measured PD whose con¢rmed condition is ``normal'' or ``abnormal'' according to CPP and cerebral C(a-v)O2 in pigs in which intracranial pressure was increased experimentally
Fig. 2. Correlation between pulsatility di¡erence (PD) (systolic ÿ diastolic cerebral blood £ow velocity) and cerebral perfusion pressure (CPP) in pigs in which CPP was decreased by increasing intracranial pressure experimentally and returned to baseline with an infusion of epinephrine (4 pigs); n = data points.
Transcranial Doppler variables correlated better with CPP than with ICP (Table 1). Both CPP and ICP correlated best with systemic pressure, and the best linear correlation between CPP and a transcranial Doppler variable was with PD (Table 1). The best nonlinear correlation was also between CPP and PD in the non-epinephrine-treated group (r = 0.8, P < 0.0001; Figure 1). When the value of PD was investigated, PD could be predicted to lie within a certain range related to normal or abnormal CPP, assuming CPP >60 mmHg was considered normal. A PD of above 70 cm/sec had a greater than 77% chance of predicting an abnormal CPP and a 0% chance of predicting a normal CPP (Table 2). A PD of below 30 cm/sec predicted an abnormal CPP less than 0.6% of the time and a normal CPP 89.8% of the time. With C(a-v)O2 < 6.5 vol% considered normal (Table 2), a PD 60 cm/sec, 86.7% of C(a-v)O2 values were abnormal (positive accuracy of 85%) and only 14.3% were normal (negative accuracy of 80%). The relationship between PD and C(a-v)O2 followed a certain ``pattern,'' which varied between animals by the extent and duration of intracranial hypertension (Figure 4). The extent of the changes in PD and C(a-v)O2 di¡ered among animals after peak intracranial hypertension, but the pattern of changes was the same (Figure 4). The ``pattern'' of changes was consistent for PD, C(a-v)O2 , and CPP during any given recovery phase in the ``injury cycle'': when CPP increased, PD decreased,
PD range
Positive accuracy
Negative accuracy
CPP < 60 mmHg
C(a-v)O2 > 6.5 vol%
CPP > 60 mmHg
C(a-v)O2 < 6.5 vol%
10 20 30 40 50 60 70 80 90 100 110 120 130
0 0.6 0.6 4.8 19.9 38.0 54.2 77.1 81.5 88.0 91.6 95.8 100
0 0 0 6.7 26.7 66.7 86.7 86.7 86.7 93.3 93.3 100 100
100 100 89.8 57.6 20.3 8.5 3.4 0 0 0 0 0 0
100 100 78.6 67.9 35.7 17.9 14.3 14.3 10.7 7.1 3.6 0 0
The probability that a certain range of PD will be associated with CPP > or or
