Noninvasive Monitoring of Cerebral Perfusion by Transcranial Doppler During Fulminant Hepatic Failure and Liver Transplantation

Abstract

Survival in patients with fulminant hepatic failure (FHF) treated medically, rather than surgically, ranges from 12%-67% (mean 39%), depending on the cause of the disease [1]. The major cause of mortality is increased intracranial pressure (ICP) from brain edema [2]. For patients with a poor prognosis [3], orthotopic liver transplantation may be the definitive treatment [4], even though, only a few years ago, some considered this treatment ineffective by the time patients reached Grade 4 encephalopathy [5]. ICP monitoring allows physicians to use specific therapy to control intracranial hypertension. Continuous measurement of ICP perioperatively in the management of FHF has been associated with a survival rate of 54%-74% in a series of six to 23 patients [6-9], which is generally higher than with medical means [1], and was as high as 92% for the selected group who had undergone liver transplantation [6]. Such invasive monitoring, however, is especially risky in FHF patients with coagulopathy, in whom the incidence of bleeding from ICP monitoring ranges from 5%-22% [6,8] with a mortality rate of 60% [6]. Although the use of ICP monitoring for FHF has become more routine [8], not all centers support the use of this invasive monitoring. We describe a patient with FHF and brain edema who underwent liver transplantation and whose cerebral perfusion was monitored noninvasively by transcranial Doppler (TCD) imaging, as well as invasively by ICP. The noninvasive technique provided adequate information when cerebral perfusion was low, comparable with the invasive technique, and allowed intracranial hypertension to be diagnosed and treated effectively. Case Report A 50-yr-old woman with a 1-mo history of acute hepatitis with negative serologies presented with jaundice only, without ascites or bleeding; however, she became encephalopathic within a week. She was transferred to our tertiary care center for medical management and possible liver transplantation. Within five days, the neurologic status deteriorated from being comatose and reacting to painful stimuli on admission (Stage 2-3) to deep coma without response (Stage 4). Five days postadmission, the trachea was intubated, and the patient was hyperventilated and scheduled for surgery within a week postadmission. Preoperatively, prothrombin time was 25 s and partial thromboplastin time 52 s, which was treated with fresh-frozen plasma. Bilirubin was 39.1 mg% (15.2 mg% direct). Electrolyte and glucose concentrations, blood urea nitrogen, creatinine, and protein were normal. Electrocardiogram showed normal sinus rhythm (79 bpm, low-voltage), echocardiogram showed normal left ventricular function, and ejection fraction was >50%. Chest radiogram was normal. Computed tomography scan of the brain showed normal ventricles and no edema, mass effect, or shift. Electroencephalogram did not reveal focal or lateralizing alterations; TCD imaging revealed normal flow velocity (FV; Figure 1A). The patient had no signs of renal impairment during hospitalization.Figure 1: Transcranial Doppler (TCD) images of cerebral flow velocity from a patient during different stages of liver transplantation and treatment (see Figure 3 and Figure 4 for events). The "cursor" value is diastolic flow velocity (cm/s). This value was determined manually, as well as by the monitor (black shading). A. Baseline: Preoperative flow velocity in the right middle cerebral artery (RMCA) is normal; pulsatility difference is 48 cm/s. B. Before anesthetic induction: Flow velocity is recorded from the RMCA before hemodynamic support and intracranial decompression; pulsatility difference is 60 cm/s (Event 1). During surgery, calculation was made manually and not according to automatic reading of the TCD recording (peak systolic flow velocity is 78 cm/s). C. During revascularization of the transplanted liver: Flow velocity is recorded from RMCA; at this point, the TCD recording mistook systolic "dicrotic notch" for maximal peak, which is actually 75 cm/s; pulsatility difference is 62 cm/s (Event 15). "Depth" refers to the depth to which the probe penetrates. PI, pulsatility index.Because deep coma suppressed neurologic signs, ICP monitoring was initiated. After treatment for coagulopathy (partial thromboplastin time, 40 s) with fresh-frozen plasma, a subdural catheter was inserted and connected to a closed, sterile monitoring system (Model ER, Camino Laboratories, San Diego, CA) for continuous ICP monitoring. This system does not measure pressure directly, but uses fiberoptic technology. Initially, ICP was 10-20 mm Hg; to keep it <15 mm Hg, hyperventilation, optimal head positioning, intravenous (IV) mannitol, and steroids were administered. Treatment was maintained for three days with good response. However, when the patient arrived at the operating room, ICP was 35-40 mm Hg, and cerebral perfusion pressure (CPP; mean blood pressure--ICP) was 45 mm Hg. TCD imaging demonstrated low diastolic flow (Figure 1B; Figure 2; and Figure 3, Event 1). To keep CPP >50 mm Hg, we administered treatment to increase systemic pressure and decrease ICP simultaneously. To increase systemic pressure, the following cardiotonic drugs were administered: dopamine, 5 to 10 micro gram centered dot kg centered dot min-1; epinephrine, bolus and infusion, 0.01 micro gram centered dot kg centered dot min-1; phenylephrine; and calcium chloride when ionized calcium decreased to <2.0 mEq/L. Sodium thiopental, mannitol, and furosemide were given to decrease ICP. Once CPP was >50 mm Hg and diastolic FV increased (Events 2-4 in Figure 2 and Figure 3), we put the patient in the Trendelenburg position temporarily to place central venous and pulmonary artery catheters. Care was taken to place the catheters apart from one another and to use the external jugular, subclavian, and basilic veins, rather than both internal jugular veins, which would increase impedance to venous return from the brain bilaterally.Figure 2: Changes in cerebral perfusion pressure (CPP) and diastolic component of transcranial Doppler flow velocity recording (TCD-D) in a patient during liver transplantation (Events 1-29; periods between events may vary in time). Event 1, before treatment and anesthetic induction; Event 2, after treatment and before anesthetic induction, Event 4, after anesthetic induction and before incision; Event 12, end of Stage 2 of liver transplantation, clamping; Event 15, early Stage 3 (revascularization, unclamping); Event 24, late Stage 3 (manipulations); Event 28, end of surgery before transport of the patient to the intensive care unit.Figure 3: Changes in intracranial pressure (ICP) and diastolic component of transcranial Doppler recording (TCD-D) in a patient during liver transplantation surgery (Events 1-29; periods between events may vary in time; see events in Figure 2).Anesthesia was induced and maintained with fentanyl and isoflurane (0.2%-0.5% range) in 50%-100% O2/air, and IV pancuronium was used for muscle relaxation. In addition to the basic monitoring, the following were monitored continuously: invasive systemic and central pressures, mixed venous O2 saturation, ICP, and TCD. To establish a neurophysiologic status immediately preoperatively, evoked potentials were obtained before any anesthetic or surgical manipulation in the operating room and were normal. After cannulation during anesthesia, TCD images and ICP were within normal limits (Figure 2 and Figure 3, Event 4). The goal during the operation was to maintain CPP >50 mm Hg. Epinephrine and calcium chloride were required throughout the procedure to maintain effective systemic pressure (central venous pressure and wedge pressure at 10-15 mm Hg with fluid administration to maintain filling pressure). Administration of diuretics, furosemide, and mannitol maintained high urine formation and decreased ICP. During Stage 1 surgery (dissection and manipulation of the liver and its vessels), which reduces venous return, crystalloids (saline 0.9% and plasmalyte A) were administered and titrated to central filling pressure and CPP. Also, packed red blood cells, fresh-frozen plasma, and platelets were administered. During Stage 2 (the anhepatic stage that lasted 2.5 h), with the liver vessels clamped, central venous pressure was 5 mm Hg, CPP <25 mm Hg and the diastolic component of the TCD (TCD-D) <5 cm/s (Figure 2, Event 12). All reversed within minutes by volume loading, establishing (as planned) veno-veno bypass, and epinephrine administration,without any use of pharmacologic agent to reduce ICP (i.e., barbiturates). During unclamping and liver reperfusion, CPP was again reduced temporarily (<50 mm Hg), and TCD imaging showed low (<15 cm/s) diastolic FV and high pulsatility difference (PD, 62 cm/s; Figure 1C and Figure 2, Event 15). Intravascular fluid replacement, calcium chloride, and epinephrine were effective. Later, in Stage 3 (neohepatic stage), CPP was within advisable limits (>60 mm Hg), and cerebral FV monitoring was normal (Figure 2 and Figure 3, Events 24-28); all ICP/CPP problems during Stages 2 and 3 were related and treated as systemic pressure/perfusion problems. ICP slowly decreased to 19 mm Hg by the end of surgery, and diuresis was 3500 mL during the 12-h operation. The hemodynamic stability enabled the dose of epinephrine to be decreased to 0.01 micro gram centered dot kg centered dot min-1. A total of 22 L of crystalloid and 16 U of packed red blood cells, 16 U of fresh-frozen plasma, and 10 U of platelets were administered for the entire procedure. The overall relationship between TCD PD and diastolic CPP (diastolic blood pressure--ICP) during anesthesia and surgery is described in Figure 4. There was an inverse relationship with good correlation (r = 0.82, n = 29) for nonlinear regression. Also, TCD FV indices (i.e., PD and pulsatility index [PI]) had a lower nonlinear correlation with mean CPP (mean blood pressure--ICP; Table 1, but that nonlinear correlation was always exponential, second-order polynomial regression type, with a change in relationship <50 mm Hg mean CPP Figure 5, A and B. Evaluation of other intraoperative FV parameters and indices (including PI or mean FV in our study) did not demonstrate as good a relationship with CPP or ICP as did PD Table 1. Only the relationship of PI with diastolic CPP or mean CPP had an r2 value similar or better when compared with the same relationship of PD with cerebral pressures (0.66 vs 0.68 and 0.59 vs 0.47 [exponential nonlinear relationship]), but the PI-CPP relationship had a lower r value for linear relationship when compared with the PD-CPP relationship Table 1. After the operation, the patient was taken to the intensive care unit, where she began to respond appropriately after 48 h. On the third postoperative day, ICP monitoring was discontinued, and on the sixth day her trachea was extubated. She continued to do well with no apparent neurologic deficit and was discharged from the hospital 7 wk after the operation without any additional complications.Figure 4: Correlation of diastolic blood pressure--intracranial pressure (ICP) (diastolic cerebral perfusion pressure [= "diastolic" CPP]), with transcranial Doppler systolic--diastolic flow velocity (pulsatility difference) in a patient during liver transplantation. For nonlinear (exponential polynomial) regression (y = a + bex p[-x/c]): full line; r = 0.82; n = 29 data points; FitstdErr = 5.46; Fstat = 27.52; a = 26.03; b = 56.01; c = 38.79. For linear regression (y = a + bx): dotted line; r = 0.80; n = 29 data points; FitstdErr - 5.66; Fstat = 48.45; a = 71.61; b = -0.61.Table 1: Linear and Nonlinear Correlation (r2 Values) Between Cerebral Perfusion Pressure (Diastolic and Mean) and Intracranial Pressure Values and Flow Velocity (Values and Indices) in a Patient During Orthotopic Liver TransplantationFigure 5: A. Correlation of mean blood pressure--intracranial pressure (ICP) (= mean cerebral perfusion pressure [CPP]), with transcranial Doppler systolic-diastolic flow velocity (pulsatility difference) in a patient during liver transplantation. For nonlinear (exponential polynomial) regression (y = a + bex P[-x/c]): r = 0.68; n = 29 data points; FitstdErr = 7.00; Fstat = 11.71; a = 37.73; b = 99.32; c = 23.77. B. Correlation of mean blood pressure--ICP (= mean CPP), with transcranial Doppler pulsatility index in a patient during liver transplantation. For nonlinear (exponential polynomial) regression (y = a + bex P[-x/c]): r = 0.77; n = 29 data points; FitstdErr = 0.29; Fstat = 18.99; a = 1.11; b = 11.55; c = 12.78.Discussion Management of cerebral edema in patients with FHF depends on objective measurements of CPP or ICP, even though empiric therapy for ICP can be initiated without knowing ICP value. The goal is to keep CPP >40-50 mm Hg and ICP <30-40 mm Hg [4,9] no matter what the cause of the edema. During liver transplantation, however, patients are susceptible to sudden changes in ICP or CPP because of potential systemic hypotension and frequent changes in fluid balance and intravascular volume (blood loss, edema), venous pressures (clamping and unclamping of major vessels), overall hemodynamic stability (low contractility and peripheral vasodilation), and potential intracerebral bleeding (clotting abnormalities) [10-12]. ICP monitoring during orthotopic liver transplantation [6,7,12,13] provided evidence that ICP frequently increases in late Stage 2 and even more frequently in early Stage 3 (new liver perfusion) because of volume overload and postreperfusion cerebral hyperperfusion [13]. Thus, preoperative and intraoperative management is aimed at decreasing intracranial edema and pressure to control and maintain adequate brain perfusion. The options for ICP monitoring are placing a subdural, epidural, or intraventricular catheter; the latter is potentially more useful because of the therapeutic ability to withdraw cerebrospinal fluid. However, it has a higher associated risk for intracerebral bleeding. Placing a subdural catheter may carry less risk; however, it is an invasive procedure as well, and requires expert surgical technique--especially in those patients with coagulopathy who are susceptible to intracranial bleeding [6,8]. An alternative mode of monitoring is TCD, which has been used extensively for monitoring of head injury and cerebral circulatory arrest [14] and has been studied extensively in head-trauma patients [15]. However, in a previous report of three patients with FHF and brain edema, TCD monitoring failed to show any abnormal findings before liver transplantation [16], as we did in our patient. Even though ICP and computed tomography findings did not correlate well in a large series of patients [10,17] ICP monitoring is still considered the best method to evaluate and treat brain perfusion in patients with increased ICP, as well as prevent adverse neurologic outcome. Thus, the value of TCD will, in part, be determined by comparing TCD with ICP monitoring (as demonstrated in our patient). The extensive studies of TCD monitoring in head trauma patients have contributed some information about the relationship between ICP and TCD. Diastolic FV is influenced by cerebral vascular resistance, which is determined mainly by ICP and vessel diameter [15]. TCD images show that diastolic FV becomes hero when ICP equals diastolic blood pressure [18]. This is a conclusive warning sign, at which time TCD images of the diastolic component should be compared with diastolic blood pressure (rather than systolic or mean pressure). PI (difference between systolic and diastolic FV divided by mean FV), which also represents resistance to flow, can be correlated with ICP [15] or CPP (mean blood pressure--ICP) [19] in head-trauma patients. This correlation, however, has been weak [20], individual [15], or confounded when investigated in patients with multiple pathologic subgroups (diffuse, focal, hyperemic lesions) with minimal data derived from each pathology [19]. We think that PD (the difference between systolic and diastolic FV) is an easy variable to calculate and represents the same resistance to flow that PI describes; however, it has never before been correlated with ICP or CPP. Various experimental [21], laboratory [22], and clinical studies [23,24] suggested, however, that the Doppler spectrum can give cerebrovascular resistance information that can be related to ICP or CPP, using FV signal during systole or diastole. Analysis of these pulse-wave shapes was reported about three decades ago with qualitative [25] and quantitative [26] assessment of "damping effect" of the pulse-wave shape. This "damping effect" is produced by partial obstruction of the arterial pathway [26] or by stenosis and vasoconstriction accompanying decreased velocity [27]. The waveform damping can be expressed in terms of a parameter called PI (PI = systolic - diastolic FV/mean FV = total oscillatory energy in FV waveform divided by the energy of the mean forward flow over a cardiac cycle) [26]. During highly increased ICP (>60-70 mm Hg), there is a disproportionate increase in pulsatility (systolic-diastolic difference) of FV peak-to-peak amplitude (with sudden increase in systolic FV and sudden decrease in diastolic FV), probably due to a dramatic reduction intracranial artery transluminal pressure, and to the fact that these arteries become distensible [22]. The PI a common measure to describe the shape of signal waveform [28], but other FV parameters can be used to express peripheral flow resistance: systolic [21], diastolic [29], or mean FV [30]; the area under the curve [29]; and resistance index (RI; RI = systolic - diastolic FV/systolic FV) [31]. However, all the other parameters are considered less sensitive when assessing high ICP and total cerebral tissue resistance [22]. A TCD instrument will read FV directly (including mean) and instantaneously, calculating indices, even though frequent false readings (up to 29%) may present a problem [23,32,33]. That is the reason for searching for a simpler and more dependable variable to express resistance. We think that simple manual calculation of FV pulsatility (systolic-diastolic FV) difference will express resistance in a similar way to PI or RI, with the advantage of accuracy, simplicity, and clinical applicability, which the automated machine-calculated indices are lacking. We also believe that it is more correct physiologically to examine the relationship between TCD readings and CPP, which represents perfusion pressure, rather than ICP, which may not reflect systemic blood pressure or inform us of perfusion pressure. ICP measurement does have value in its own right in terms of prognosis. However, any previous attempt to demonstrate correlation between cerebral FV and cerebral pressure showed that CPP is a better parameter compared with ICP when CBF, cerebral O2 supply/demand ratio, and O2 consumption parameters are evaluated during brain injury [23,24]. A better correlation with PI (previous studies) and PD Table 1 is due to CPP representation of perfusion pressure rather than intracerebral pressure. It is true that when ICP is high enough it is likely to be the limiting factor in brain perfusion, more than the ability to force perfusion up; but that happens only when autoregulation is exhausted and completely lost (ICP >60 mm Hg or mean CPP <35 mm Hg), and then any increase in flow or cerebral perfusion will also increase brain volume and ICP. Because we were interested in the "transient zone" before autoregulation is completely lost, ICP was not the "gold standard" with which to compare. Furthermore, we believe that it may be advantageous to compare "diastolic" CPP (diastolic blood pressure--ICP) with TCD diastolic or TCD PD because of the known relationship between TCD diastolic and diastolic blood pressure [18]. We found that PD was significantly affected when diastolic CPP was <40 mm Hg, which is equivalent to a mean CPP of <60-50 mm Hg Figure 4 and Figure 5. This effect is a consistent characteristic of cerebrovascular resistance and its relationship with CPP or ICP. The correlation of cerebrovascular resistance (expressed by PI) with CPP will change according to ICP levels (<30-40 mm Hg, between 30-40 and 60-70 mm Hg, and >60-70 mm Hg) [22] or CPP levels (>40-55 mm Hg where autoregulation is intact; between 40-55 and 20-35 mm Hg, which is a transient zone; <20-35 mm Hg where autoregulation is exhausted) [21]. The functional increase in cerebrovascular resistance in Phase 3 (no autoregulation) is probably due to an increase in the period of arterial collapse during each cycle, rather than an increase in venous resistance [34], and this change in relationship in Phases 2 and 3 is reflected graphically in the second part of the overall relationship (second-order polynomial relationship), between PD and CPP or between PI and CPP when mean CPP <50 mm Hg Figure 5, A and B. However, at that level (equivalent to diastolic CPP <30), PD may have a tendency to increase if cerebral blood volume and perfusion will be artificially increased while ICP and O2 consumption are still high (i.e., during Stage 3 of liver reperfusion with catecholamine surge), and hyperemia status will be induced (exponential relationship in Figure 4 and Figure 5). On the other hand, PD level is expected to decrease if perfusion pressure with CBF and mean blood pressure (= driving force) all remain low while O2 consumption is still high [35]. Also, the effect of blood pressure (during systemic hypotension) on FV amplitude, measured CBF, and calculated regional cerebrovascular resistance is very similar to the relationship of CPP to these three variables; an exponential-type curve, with two-slope relationship and a "break point" [21]. Intraoperatively, TCD imaging provided adequate information, comparable to that derived from ICP monitoring, when cerebral perfusion was low Figure 2 and Figure 4. Overall, there was an inverse relationship between "diastolic" CPP and TCD PD (r = 0.82 for 29 sets of data; Figure 4); thus, low perfusion pressure or high cerebrovascular resistance (described by PD) correlates well with CPP <50-60 mm Hg or diastolic CPP <30-40 mm Hg Figure 4. That correlation, even though it is derived from one patient, is comparable to that derived from clinical studies of head injury patients: r of 0.71 for a comparison of PI with ICP [16] or r = 0.72 for a comparison with CPP. In our patient, the TCD-D recording decreased to near 0 each time CPP decreased to <45 Figure 2. Even though in some instances of seriously decreased CPP (mean CPP <50 mm Hg) TCD-D decreased to <10 cm/s Figure 2 and PI increased to >1.7 cm/s Figure 3, PD increased in all these instances to >55 cm/s Figure 2 and Figure 3. A decrease in CPP detected by TCD or ICP monitoring was treated to increase systemic pressure and decrease ICP and, when effective, treatment decreased PD to <50 cm/s Figure 3. Low TCD-D or high TCD PD were diagnosed and treated effectively and therefore occurred only briefly, and the patient was neurologically normal postoperatively. A previous report showed that PI was statistically significantly affected when CPP became <70 mm Hg [19]. If this effect is a consistent characteristic of the relationship between CPP and TCD pulsatility or TCD-D, our data from one patient suggest that similar acute changes in the quantitative nature of the relationship can occur at >55 cm/s PD Figure 4 (comparable to CPP <50 mm Hg or diastolic CPP <30 mm Hg), which may represent extremely high flow resistance and low perfusion in the brain. The outcome of patients with FHF and Grade 4 encephalopathy during liver transplantation may be greatly improved by cerebral perfusion monitoring. Evidence such as that from our case shows that the use of TCD monitoring is feasible and that a large-scale study is merited to demonstrate quantitatively the mathematic relationship between brain perfusion (monitored by TCD) and intracranial pressures (ICP, CPP) during diffuse brain edema. Also, future studies should differentiate between cerebral hypoperfusion as opposed to a hyperdynamic state; more important, future studies should answer the question of whether a qualitative change, an acute quantitative change, or a divergence in TCD and CPP relationship can be used to detect dangerously low cerebral perfusion. After establishing such a relationship in a laboratory model, a randomized trial of ICP versus TCD monitoring should be considered, to demonstrate therapeutic efficacy and equivalence with decreased risk.