Cerebral blood flow during cardiopulmonary bypass in pediatric cardiac surgery: the role of transcranial Doppler – a systematic review of the literature

The available instruments emit pulsed-wave ultrasounds at 2–4 MHz who penetrate and scatter the tissue. The ultrasonic waves are then backscattered from moving red blood cells with a shifted frequency toward the receiver, placed in the same doppler probe of the transmitter. The frequency shift lies in the audible range. The frequency (Doppler) shift can be expressed as follows:

F = 2 * Fo * v * cos∝/c

where F is Doppler shift (Hz), Fo = mean frequency of the emitted ultrasound, v = blood flow velocity (cm/s), ∝ = angle between the direction of the transmitted sound beam and the axis of the blood flow and c = velocity of sound in tissue. Thus, F is proportional to blood flow velocity (v) if the angle and the frequency of emitted ultrasound remains constant. The formula also shows that the angle between the beam's direction and blood should be kept close to 0 in order to minimize the possibility of measuring errors (the maximum value of F can be found at ∝ = 0).

Several positions can be used to examine the basal cerebral arteries (Figure 1), but the most reproducible and comfortable tecnique for clinical use in pediatric patients is to insonate the middle cerebral artery (MCA) through the temporal window which can be found about 1 cm in front of the external auditory meatus and roughly 1–2 cm above the zygomatic arch. The ultrasonic beam is directed horizontally. The depth of the sample volume and angle of insonation is adjusted until the bifurcation of the MCA and the anterior cerebral artery (ACA) is found (positive deflection toward the transducer from the MCA and a retrograde signal with a negative deflection from the ACA). Normally the clinician should find a stable baseline at the beginning of evaluation and compare it with further values (Table 1).

According to the Hagen-Poiseuille law, the flow in a rigid tube is

F = P * π * r⁴/8 * η * l

Where r is the radius of the tube, l is the length of the tube, P is the difference between the pressure at the beginning and at the end of the tube, η is the viscosity of the fluid.

The relation between the flow and the velocity (v) is

F = v * π * r²

From the previous formula and from the Hagen-Poiseuille law we obtain

v = P * r²/8 * η * l

The previous formulas allow us to conclude that the flow (ml/min) in a vessel is dependent on the radius of the vessel and on the flow velocity within the vessel. The velocity is dependent on viscosity (hematocrit) of the blood, on the radius of the vessel and on the perfusion pressure in the vessel. Assuming that the MCA diameter remains unchanged, any changes in velocity would reflect changes in flow. Several experimental works have provided evidences that the diameter of the MCA does not change significantly during cardiac operations, the vasomotor action being confined to resistance arteries and arterioles, as demonstrated by previous works [2, 3].

Many studies have used TCD in order to evaluate changes of cerebral perfusion measuring blood flow velocities [4, 5]. As the diameter of the insonated vessel is not know in any individual, it is not possible to evaluate the absolute value of cerebral blood flow. However, the variation in cerebral blood flow (CBF) is determinant, rather than the absolute value. The reliability of correlation between changes in MCA velocity and CBF (measured using intravenous Xenon-133) has been first validated by Bishop and coworkers [6] in an experimental work that compared the MCA flow velocities and CBF in symptomatic patients with cerebrovascular disease in normo and hypercapnic conditions. These findings have been confirmed in an experimental work by Rosemberg [7], who demonstrated that changes in cerebral blood flow velocity are useful qualitative measures of changes in cerebral blood flow in paralyzed newborn lambs. Trivedi et al. [8] showed marked similarities between changes in the velocity in the MCA and changes in the estimated hemispheric CBF measured with Xenon-133 clearance technique (Figure 2).

The introduction of deep hypotermic cardiopulmonary bypass (DHCPB) with or without deep hypothermic circulatory arrest (DHCA) in children who need complex aortic arch reconstruction has substantially improved operating conditions and therefore reduced cardiac morbidity. The aim of hypothermia during CPB is to reduce metabolic activity, CBF and cerebral metabolic rate of oxygen (CMRO₂), so that it is possible to maintain energy stores and provide organ protection during low flow state [9, 10]. Profound hypothermia with continuous low-flow cardiopulmonary bypass (low-flow CPB) has been suggested as being superior to DHCA in preventing neurological damage [11, 12], hypothetically providing an indefinite period of cerebral perfusion. The most important factor ruling cerebral hemodynamics is cerebral autoregulation, in order to maintain CBF constant throughout a wide range of arterial pressure. This autoregolatory mechanism is deeply affected by temperature. Taylor and coworkers [13] found that autoregulation is preserved during normothermic CPB, it begins to be altered at temperature less than 25°C, and it is lost at temperature less than 20°C, while previous studies had shown that autoregulation is intact during moderately hypothermic CPB (25° to 32°C) [10], and it is lost during deep hypothermic CPB (18° to 22°C) (Figure 3). However, this loss of autoregulation is most likely caused by a state of "cold-induced vasoparesis", in which cerebral vascular resistance increases with temperature reduction. Jonassen and coworkers [14] showed that a significant proportion of patients treated with profound hypothermia and either low-flow CPB or circulatory arrest exhibited a TCD pattern consistent with increased cerebral vascular resistance in the early postoperative period, whereas this pattern was not present in patients treated with moderately hypothermic CPB. There was a tendency for this pattern to occur with greater frequency in patients who had a period of circulatory arrest (Figure 4). While CBF decreases in a linear manner, CMRO₂ decreases exponentially with temperature reduction. Therefore, CBF/CMRO₂ during DHCPB increases favouring luxury perfusion of the brain. Normal coupling of CBF/CMRO₂ is present before and after CPB, as well as during normothermic CPB and α-stat management [10]. During CPB rewarming, CBF returns to baseline values, except in patients exposed to periods of DHCA where CBF remains decreased (Figure 5) [15]. Astudillo et al. [16] demonstrated that low cerebral perfusion immediately following DHCA is characterized by a prolonged period of absent diastolic CBFV in MCA: this finding was explained by an increased intracranial pressure after total circulatory arrest procedure, while patients subjected to continuous low-flow perfusion technique showed a CBFV close to baseline values at skin closure. It must be remarked however, that a significant age difference between the two groups was present, being the patients subjected to circulatory arrest younger than the patients in the non arrest group. An interesting finding from the group of Rodriguez [17] is that a delay in rewarming on reperfusion after DHCA improved recovery of a diastolic doppler signal compared with patients who underwent immediate rewarming. In the group undergoing cold reperfusion, postbypass CBF velocity was not different from baseline (Figure 6). Taylor and coworkers [13] used the TCD as an indicator of perfusion during repair of congenital heart defects requiring moderate or profound hypothermia and low-flow CPB. The principal finding of this study was the immediate loss of detectable CBFV in the middle cerebral artery when cerebral perfusion pressure (CPP) decreased below 9 mmHg: they clearly conclude that CPP is a crucial parameter, rather then pump flow rate, in impacting brain perfusion. They also confirmed the loss of cerebral autoregulation between 23° and 25°. However, Jonassen et al [14], in agreement with data from van der Linden [2], showed detectable cerebral blood flow at pump flow rate and mean arterial pressure (MAP) values lower than those reported by Taylor. Possible explanations may include the use of vasodilators, the increased sensitivity of the TCD apparatus by removal of the low-pass filter and the avoidance of jugular central venous lines that can theoretically impede regional cerebral venous drainage in small infants and thus decreasing CPP. A more recent study from Zimmermann et al. [18] performed in 28 neonates undergoing the arterial switch operation with α-stat acid-base management, showed that cerebral perfusion can be detected by TCD in the MCA in some neonates at bypass flow as low as 10 ml/kg per minute. However a minimum bypass flow rate of 30 ml/kg per minute was needed to detect cerebral perfusion in all neonates. All patients with a MAP of 19 mmHg or greater, regardless of pump flow rate, had detectable cerebral perfusion by TCD but correlation between MAP and CPB pump flow rates was minimal, confirming the conclusions of Taylor that mean arterial blood pressure alone is a poor indicator of CPP.

The aim of regional low flow perfusion (RLFP) technique is to minimize the effect of DHCA on the brain maintaining cerebral blood volume and oxygen saturation while providing the same surgical exposure obtained with DHCA alone. Using near infrared spectroscopy (NIRS) technology Pigula et al. [19] documented significant decreases in both cerebral blood volume and oxygen saturation in children who underwent repair with DHCA as compared with children with RLFP. When the use of RLFP rate is guided by NIRS, cerebral oxygen saturation was the same of that provided by standard CPB support. Andropoulos and coworkers [20] found that the use of TCD ultrasonography might add further informations during bypass and RLFP. In particular, since NIRS monitor is not able to display oxygen saturation higher than 95%, it only detects inadequate blood flow (desaturation) but not excessive CBF (hypersaturation). The potential dangers of excessive CBF include cerebral edema and intracranial hemorrage. The use of TCD is thus of great importance in order to maintain cerebral oxygen saturations and blood flow velocities within 10% of baseline, in order to prevent cerebral hyperperfusion during periods with high NIRS saturation values. Noteworthy, a significant difference in the mean bypass flow rate necessary to maintain the baseline oxygen saturation and CBFV was present between the two studies. Reacquisition of baseline cerebral blood volume and cerebral oxygen saturations were accomplished with a RLFP of 20 ml/Kg/min and 63 ml/Kg/min by Pigula and Andropoulos, respectively. A possible explanation of this significantly different perfusion management could be attributed to the fact that Andropoulos and coworkers obtained a higher cerebral vasodilation due to a different acid-base management technique (Ph-stat versus α-stat used by Pigula) and to the use of phenoxybenzamine or phentolamine hydrocloride during CPB.

It is well known that CO₂ has a profound influence on CBF [21] with increasing CO₂ resulting in an increase in CBF. During CPB, CBF increases with increasing arterial carbon dioxide tension, but this response is diminished by deep hypothermia and age less than 1 year. In the pH-stat management PaCO₂ is maintained at 40 mmHg regardless of temperature changes (temperature-corrected), while α-stat management PaCO₂ is not adjusted (temperature-uncorrected). Trivedi et al. [8] showed a decrease of CBF (evaluated by Xenon-133 and CBFV) during 28° bypass in patients subjected to α-stat management. This was associated with a significant reduction in PaCO₂, while there was no reduction in patients subjected to pH-stat management (Figure 2). Other studies, however, performed by another group demonstrated that during α-stat management cerebral blood flow velocity measured by TCD was less pressure passive than during pH-stat management and that there was a better matching of cerebral metabolism to CBFV with α-stat management [22].

There is an inverse relation between hematocrit and and CBFV during DHCPB in neonates and infants, although the optimal hematocrit to perform CPB has not yet been determined [23].

Rodriguez et al. [24] showed that aortic and venous cannulations for CPB in children can hesitate in transient reductions in CBFV and MAP and that usually these alterations are compensated within the next 40 seconds. Reported undesirable cerebral effects were more frequent in young infants with cannulation of right atrium. Superior Vena Cava obstruction during venous cannulation resulted in an increased pressure in the internal jugular vein and in the absence of diastolic Doppler flow (Figure 7).

Recent studies showed the beneficial effects of simultaneous neurological monitoring (NIRS, TCD, processed electro encephalography). Austin et al. [25] showed the potential benefit of intraoperative interventions based on an intervention algorithm previously used in adult cardiac surgery and modified for pediatric use (Table 7). The algorithm design triggers specific interventions when specific clinical parameters change. Two groups of patients were randomized to the intervention or no intervention group in response to any detected perfusion abnormality. 70% of patients in this study experienced significant changes in one or more monitored variables. The incidence of neurological sequelae (i.e. seizure, movement, vision or speech disorders) between patients with no change of neurophysiologic values and patients who received interventions in response to a change in any of the monitored variables was the same (6–7%), while neurological damage was 26% among patients who did not receive any treatment. The proportion of neurological sequelae in no-interventions patients discharged from the hospital within 1 week (32%) was significantly lower than that observed in either the intervention (51%) or no change group (58%). The TCD was responsible for 37% of detected problems, while in another study from Andropoulos et al. [26] the TCD alone was responsible for only the 10% of interventions.

TCD can also easily detect and count cerebral emboli; they are described as high intensity transient signals (HITS). A study from O'Brien [27] showed that microemboli can be detected in the carotid arteries of children during repair of congenital heart disease and are especially prevalent immediately after release of the aortic cross clamp. However the TCD cannot distinguish between true emboli and false positive artefacts. The number of emboli detected during pediatric congenital heart surgery does not seem to correlate with postoperative neurological injuries.