Extracorporeal Arteriovenous Ultrasound Measurement of Cardiac Output in Small Children

The clinical examination of a severely hemodynamically unstable child has obvious limitations and may give unreliable estimates of the overall hemodynamic status. The cause of arterial desaturation, degree of hypovolemia, and systemic hypoperfusion are parameters that may be particular difficult to evaluate because of compensatory mechanisms.^1–5  Easily adapted and safe monitoring devices with good reliability and reproducibility are needed to address these parameters, especially in children.^6–8 

An available invasive cardiac output (CO) measuring method, such as thermodilution technique, although applicable for adults, has size restrictions for use in young children and may cause a substantial risk of complications (bleeding, thrombosis, embolism, and arrythmia). Noninvasive methods are less accurate and precise in measuring CO in small children (electrical bioimpedance, carbon dioxide rebreathing, and Doppler methods), require high level of training (echocardiography), or are clinically impractical and time consuming (magnetic resonance imaging).^9,10 

A technology that uses ultrasound detection of blood dilution by means of a saline bolus injection and an extracorporeal arteriovenous loop between existing central venous and peripheral arterial lines has been developed for hemodynamic assessment in young children and neonates.¹¹  It is minimally invasive, because it uses existing catheters and does not require additional invasive procedures. It measures total CO, including coronary blood flow.

The hypothesis of the study is that blood dilution detected by ultrasound is comparable with the reference method, perivascular flow probe around aorta to determine cardiac output. The aim of the study was to estimate the precision of this technology in young children compared with CO measured by a perivascular flow probe around the aorta (reference method) through five simultaneously replicated CO measurements and to assess the agreement by quantifying the difference (bias) between the two methods.

Forty-seven children undergoing elective cardiac surgery for correction of atrial septal defect and/or ventricular septal defect were enrolled in the study. Inclusion criteria were informed written parental consent, weight of less than 15 kg, and atrial septal and/or ventricular septal defects. Because lack of repeatability of the CO measurement can interfere with the comparison of the two methods, it was important that the CO measurement and the stroke volume were fairly constant during the measurement. Exclusion criteria were, therefore, shunts (undiagnosed extracardiac or significant residual shunt after the surgical correction), perioperative arrythmias (supraventricular, nodal tachycardia, and atrioventricular heart block), and/or significant valvular regurgitations (aortic, mitral, tricuspid, and pulmonary valvular leaks). This study was approved and registered by the Ethics Committee of Lund University, Lund, Sweden (Dnr 2013636).

The technology is based on the premise that the ultrasound velocity of blood changes linearly with the level of blood dilution caused by injection of a specified bolus volume of body-temperature isotonic saline. The ultrasound velocity in blood is 1,560 to 1,585 m/s and decreases toward the ultrasound velocity of saline (1,530 m/s) after a bolus injection of saline. The device, developed by Transonic Systems Inc. (USA), uses an extracorporeal arteriovenous loop connected between existing arterial and central venous lines. The loop is connected to an external roller pump, which maintains a stable blood flow in the loop of 10 to 12 ml/min. The loop contains specific venous and arterial segments to which external ultrasound sensors fit. The sensors measure the ultrasound velocity and the blood flow at the out-flow and in-flow parts of the loop circuit.

A measurement session begins by entering the patient’s weight, length, arterial blood pressure, central venous pressure, and heart rate into the device. The connection stopcock to the out- and in-flow segments of the arteriovenous loop, primed with 5 ml of body-temperature, heparinized isotonic saline, is opened. The roller pump starts, and a small amount of body-temperature physiologic isotonic saline (0.5 to 1.0 ml/kg) is injected at the out-flow segment of the loop on the venous side before the venous ultrasound sensor. The saline is warmed to 37°C body temperature by a bag warmer that is connected to the device. The volume and time of the injected saline is determined by the venous ultrasound sensor. The saline is completely mixed as the blood passes through the cardiopulmonary circulation and gives rise to a homogenous blood dilution on the arterial side. The final blood dilution that occurs in the systemic arterial circulation is detected by the arterial ultrasound sensor at the arterial in-flow segment of the loop, and an ultrasound velocity curve is generated (fig. 1).

Because the technology records the ultrasound velocity simultaneously at both the out-flow and in-flow segments of the loop at a constant blood flow rate, not only can the area under the curve be analyzed, but the time of occurrence and form of the dilution curve after it passes through the lungs and heart can be used to calculate total end-diastolic cardiac volume, central blood volume, and active blood volume and to determine and detect cardiac shunts.^12–17  CO is calculated by analyzing the area under curve based on the Stewart–Hamilton indicator dilution principle.^18–20 

AU-series confidence perivascular flow probes (Transonic Systems Inc.) were used in this study. The flow probe is custom-designed to fit around vessels to measure blood flow in real time by ultrasound transit-time technology. Transit-time technology uses four crystals and wide-beam illumination to send ultrasonic signals back and forth across the vessel, alternately intersecting the blood in upstream and downstream directions. The flowmeter derives an accurate measure of the “transit time” it takes for the wave of ultrasound to travel from one transducer to the other. The difference between the upstream and downstream integrated transit times is a measure of true volume flow, not velocity.

The flow probes are available in different diameter sizes (8 to 24 mm) and can be used multiple times because they can undergo standard sterilization. The probes come ready to use and calibrated from the manufacturer with a certified length of use for more than 1 yr.

In our study, the flow probe was applied to the aorta approximately 1 cm distal to the origin of the coronary arteries (fig. 2). The flow probe was then connected to an Optima dual-channel HT363 Flow-QC meter (Transonic Systems Inc.). AureFlo diagnostic software (Transonic Systems Inc.) was used to visualize a good signal of pulsating aortic blood flow waveform and record CO. Transit-time ultrasound perivascular flow probes are considered the standard reference method for cardiac output estimation and have been verified in number of studies.^21,22 

Anesthesia was induced using fentanyl (5 μg/kg) and penthothal (5 mg/kg) and maintained with isoflurane (0.5 to 1.0%). Pancuronium (0.2 mg/kg) was given to facilitate intubation with a cuffed endotracheal tube. As is routine in children undergoing cardiac surgery, all subjects had a peripheral arterial catheter placed in the radial artery and a central venous catheter placed in the right internal jugular vein. The catheters were connected to the arteriovenous loop of the CO device and were primed with heparinized (2 units/ml) 37°C physiologic saline. Ultrasound sensors were placed on the venous and arterial segments of the arteriovenous loop before surgery. After surgical correction and weaning from cardiopulmonary bypass, transesophageal echocardiography was performed to exclude residual intracardiac shunts or valve regurgitations. When a stable sinus rhythm and normal body temperature were achieved, the surgeons applied the perivascular flow probe around the aorta, and measurements were initiated. Each measurement session consisted of five consecutively repeated CO measurements with injections of body temperature physiologic saline boluses and, simultaneously, five readings of aortic blood flow measured with the transit-time ultrasound periaortic flow probe.

Statistical analysis was performed using Statistica version 12 (Statsoft, USA). No statistical power calculation was conducted before the study, because the bias and SD of the bias between the two methods were unknown, and the CI for the 95% limits of agreement was not possible to estimate. The sample size was based on previous experience with this design. All data are expressed as mean ± SD unless indicated otherwise.

The degree of variation of each technique was presented as the coefficient of error (CE) of average repeated measurements, calculated as ratio of the coefficient of variation (CV) divided with the square root of the number (n) of repeated measurements (CE = CV/√n). Precision of a technique is considered to be two times the coefficient of error as suggested by Cecconi et al.²³ 

Bland–Altman analysis was used to estimate bias between the different techniques while accounting for the repeated measurements within each individual.²⁴  The mean difference (bias) between cardiac output with saline dilution and ultrasound detection (CO_UD) minus cardiac output with perivascular flow probe around aorta (CO_PVFP) was calculated and plotted against the average of the comparison (CO_UD + CO_PVFP)/2. The 95% limits of agreement were calculated as mean bias ± 1.96 × SD (SD of the bias). Limit of agreement analysis was performed to determine whether the two methods agreed sufficiently with each other so that one could replace the other. The 95% CI of the bias and the limits of agreement were determined after testing for normal distribution using Levene’s test.²⁵ 

According to Critchley and Critchley,²⁶  the percentage error (PE) was calculated as 1.96 × SD of the bias/mean cardiac output of the reference method × 100%.

A total of 47 children were enrolled in the study. Four children were excluded before surgery: one because of hemodynamic instability; another because an ultrasound sensor came loose underneath the surgical drape; and two because the internal calibration date for the tested device had expired, and the monitor refused to accept data. This resulted in 43 children being included in the study, in which a total of 215 paired CO_UD and CO_PVFP measurements were performed (table 1). Mean age was 356 days (range, 30 to 1,303 days); mean weight was 7.1 kg (range, 2.7 to 13.6 kg); and mean body surface area was 0.36 m² (range, 0.18 to 0.59 m²). There were no missing data from the included children.

The mean CO_UD was 1.28 l/min (range, 0.46 to 2.98 l/min) and CO_PVFP 1.20 l/min (range, 0.42 to 2.70 l/min). Means and SDs were normally distributed. The coefficient of error for repeated CO_UD measurements was 1.8%, resulting in precision of 3.6% for CO_UD. Coefficient of error for CO_PVFP was 2.5%, resulting in precision of 5.0% for CO_PVFP.

Bland–Altman analysis (fig. 3) showed that the bias between CO_UD and CO_PVFP was 0.08 l/min (95% CI, 0.05 to 0.10), and the limits of agreement were −0.24 l/min (95% CI, −0.17 to −0.32) and 0.40 l/min (95% CI, 0.33 to 0.47). The percentage error between CO_UD and CO_PVFP was 26.6%.

Three patients had insignificant residual shunts after the surgical correction. No valvular regurgitations were observed that affected the results. There were no adverse effects in any patient related to measurements.

In the present study, we found that repeated CO measurements by the tested method, which uses saline bolus injections and ultrasound detection of the expected blood dilution (CO_UD), showed a similar precision as our reference method, a perivascular flow probe around the aorta (CO_PVFP). We believe that CO measurement with a periaortic flow probe is the most accurate technique to be used for comparison studies using repeated CO measurements in vivo. There was a small but significant bias of the measured CO between CO_UD and CO_PVFP according to the Bland–Altman analysis of 0.08 l/min. This finding of a lower CO_PVFP compared with CO_UD is consistent with a coronary blood flow of approximately 4 to 10% of the measured CO, which agrees with suggestions from earlier publications.²⁷ 

Earlier studies using perivascular flow probes in animals have applied the flow probe around the pulmonary artery as a measurement of total CO, which is also measured by CO_UD. This had been our intention, but we discovered that the peripulmonary flow probe tended to compress the right coronary vessels, resulting in a reduction of CO. This led us to abandon its use for the simultaneously repeated CO measurements and rely solely on the blood flow probe around the aorta, although the total CO missed the drain by the coronary blood flow.

There will always be a physiologic variability in CO caused by ventilation and cardiac filling, which result in various degree of imprecision during the CO measurements. The CO_PVFP analysis recorded beat-to-beat fluctuations in CO caused by the ventilation, changes in cardiac filling, and changes in coronary blood flow. This can be the reason for a slightly but not significantly higher coefficient of error (2.5%) compared with CO measured with CO_UD (1.8%). In addition, the dilution curve, which is used by the CO_UD to calculate CO, reflects a mean of several heart beats, which may give a more stable value. We suspected that five consecutive repeated injections of 0.5 to 1.0 ml/kg of saline could decrease the precision of the CO_UD analysis, because of a potential dilutional effect on the blood by multiple saline injections, but this did not affect the precision of our analysis. Because we conducted a trial with a high number (n = 5) of repeated measurements and did not remove any measurements, we believe that our analysis is accurate and indicates that the dilution of the blood with bolus doses of saline and detection by ultrasound sensors in a constant extracorporeal arteriovenous loop flow gives stable CO measurements compared with, for example, thermodilution, which still is regarded to be the gold standard for CO measurement in children and a reference method in validation studies.^28,29 

Although our analysis was negatively influenced by the bias caused by the coronary blood flow between CO_UD and CO_PVFP, our results show a percentage error of 26.6%, which is less than the 30% limit, that has been concluded by Critchley and Crithchley²⁶  as an acceptable limit of a new technique to be equivalent to the reference method.

One limitation to the tested technology is that the arterial blood pressure monitoring has to be closed during the CO measurement. The device, in its present form, only allows for intermittent but not continuous CO measurements as would be preferable for a monitoring device. In addition, one central venous line is occupied for the injection and the loop circulation during the measurement. However, there is no patient blood loss because all blood in the arteriovenous loop is flushed back into the circulation after measurements. Only small-volume (0.5 to 1.0 ml/kg) body-temperature saline boluses are needed to do the measurements, and there is no drop in heart rate as often is seen during measurements with a thermodilution catheter because of cold saline boluses.

In summary, the tested technology, which uses a saline bolus into the venous side and ultrasound sensors on an extracorporeal loop to detect the blood dilution both on the venous and arterial side for calculation of CO, has a precision comparable with the precision obtained by continuous CO measurement by a periaortic flow probe in young children. A Bland–Altman plot shows that the tested method detects a small expected bias between CO_UD and CO_PVFP caused by the coronary blood flow. It is easy to apply clinically in children with arterial and central venous catheters in place. The technology could potentially be a promising alternative as a reference method for comparison studies of CO in young children.

The authors thank their pediatric cardiac surgical team at the Children’s Hospital, University Hospital of Lund, Lund, Sweden, for help, patience, and understanding; and Nikolai Krivitski, Ph.D., DSc, Transonic, Ithaca, New York, for technical assistance.

Supported by project grants from the Swedish Children Heart Association, Stockholm, Sweden, and the Anna and Edwin Bergen Foundation, Lidingö, Sweden.

The authors declare no competing interests.