Cardiac output (CO) is an important hemodynamic measure that helps to guide patient therapy both in the operating room and in the intensive care unit. The “gold standards” of CO measurement are techniques that use dye dilution or thermodilution or that apply the Fick principle to metabolism of oxygen (^1–3)). The most common standardized method of CO assessment in children and adults is the thermodilution technique via pulmonary artery catheter (^2,4). Assessment of CO in pediatric patients by pulmonary artery catheterization is often technically difficult and the benefit of its use is often outweighed by the inherent risks of invasive catheterization (carotid artery puncture, cardiac arrhythmia, bleeding, embolism, clotting, and infection) (⁵). Alternative techniques for noninvasive assessment of CO (e.g., thoracic bioimpedance and transesophageal doppler ultrasound) have not been able to accurately and consistently estimate CO over a wide range of CO and disease processes (^6–9).

Recently, a noninvasive CO monitor (NICO) has been developed that uses partial rebreathing of carbon dioxide (CO₂) and determines CO via the Fick principle for CO₂ (Novametrix Medical Systems Inc. [Respironics], Wallingford, CT) (¹⁰). In animal studies, Bland-Altman analysis and regression analysis of noninvasive CO measurements, determined by the NICO monitor, were accurate and showed tight correlation when compared with thermodilution measurements of CO, even in the presence of lung disease.¹ The NICO monitor has also been clinically validated in adults and is approved by the Food and Drug Administration for use with tidal volumes ≥200 mL.^2–4 No clinical studies have confirmed the accuracy of CO determination using partial rebreathing in pediatric patients.

Cardiopulmonary relationships and lung anatomy and physiology differ in infants and children from those of adults (^11,12). Because the measurement of CO using partial rebreathing may be influenced by these differences, evaluation of its accuracy is important in pediatric patients. Thus, the purpose of this study was to assess the accuracy of noninvasive CO measurement using partial rebreathing compared with measurement obtained via thermodilution in children.

Methods

This was a prospective observational study and was approved by the IRB of The Children’s Hospital of Philadelphia. Written informed consent was obtained from each patient’s parent or legal guardian.

Thirty-seven pediatric patients aged 16 mo to 12 yr, weighing between 9.4 and 42 kg undergoing cardiac catheterization to evaluate hemodynamic function were studied (Table 1). Patients were excluded if they weighed <8 kg, had intracardiac mixing lesions, or had documented pulmonary hypertension. An 8-kg minimum weight requirement was chosen based on the NICO monitor’s Food and Drug Administration approved minimum tidal volume of 200 mL. Patients with intracardiac mixing lesions were excluded because of inaccurate estimates of CO by thermodilution. Children with pulmonary hypertension, defined as having systolic pulmonary artery or right ventricular blood pressure more than half of the systemic systolic blood pressure, were not studied because the potential increase in serum CO₂ content during partial rebreathing could further increase pulmonary vascular resistance.

Each patient received a general anesthetic consisting of an IV induction with a thiopental bolus (2–5 mg/kg) or a propofol bolus (2–3 mg/kg) followed by muscle relaxation with a pancuronium bolus (0.1 mg/kg). Anesthesia was maintained with a propofol infusion (25–100 μg · kg⁻¹ · min⁻¹), inhaled isoflurane (0.2%–0.4%), and inspired oxygen (21%). Patients’ tracheas were intubated with either a cuffed or uncuffed tracheal tube such that there was no audible airway leak at <30 cm H₂O. The patient’s mechanical ventilation was titrated to prospectively designated standard target ranges for tidal volume, arterial pH, partial pressure of CO₂, and partial pressure of oxygen using peak inspiratory pressures ≤30 cm H₂O without the use of positive end-expiratory pressure. Ventilation was achieved such that there was no audible air leak, and partial pressure of CO₂ was targeted to 40 mm Hg. Each patient was continuously monitored with pulse oximetry, end-tidal CO₂, noninvasive arterial blood pressure, electrocardiogram, and temperature measurement.

The cardiologist placed a pulmonary artery catheter (6F; Arrow International Inc., Reading, PA) under fluoroscopic guidance. CO values derived from thermodilution (CO_TD) were obtained after injection of 5 or 10 mL of cold saline (0°C) with appropriate calibration constant based on the length of the catheter and volume of iced saline injectate. Three to 4 CO_TD measurements were made and the mean was recorded. Measurements that were more than a 15% deviation from the mean were discarded by the cardiologist.

The NICO CO₂ sensor was calibrated to zero before initiating measurements by opening the system to atmosphere. Twenty-one percent inspired oxygen was entered into the data of the NICO monitor. NICO measures were only taken during the period of thermodilution measurement. The length of the rebreathing tubing was maintained in the smallest position. The NICO sensor (Novametrix Medical Systems Inc.) allowed 50 s of partial rebreathing of exhaled CO₂. Three to 4 noninvasive CO values were determined with the NICO monitor (CO_NI) (software version 3.1, fast mode) simultaneously with CO_TD assessment. The same number of CO_TD measurements accepted by the cardiologist was taken noninvasively.

The NICO monitor calculated CO using the Fick equation for CO₂. Assuming that dead-space volume and the mixed venous content of CO₂ remained constant during rebreathing, CO was calculated using the change in CO₂ excretion and change in end-tidal CO₂ based on the slope of the CO₂ dissociation curve from hemoglobin (a detailed review is presented elsewhere) (¹³). All values reported on the NICO monitor were included and the calculated mean was recorded.

A sample size of 37 patients was chosen based on the width of the 95% confidence interval. There is minimal change of this confidence interval in larger sample sizes.

A linear regression plot of the means of CO_NI versus CO_TD was graphed and Pearson’s correlation coefficient (r) was calculated. Evaluation of cardiac performance in children is usually described in terms of cardiac index (CI = CO divided by body surface area [BSA]) because BSA increases from infancy to adolescence. Thus, we calculated the r value from a second linear regression plot of mean indexed CO derived from noninvasive measurement (CI_NI) versus the mean indexed CO derived from thermodilution (CI_TD). Statistical analysis was performed with the Student’s t-test. Significance was set at P < 0.05.

Bland-Altman analysis was performed by plotting the differences between the means of CO_TD and CO_NI versus the average of the means (¹⁴). Bias, the deviation of a measurement from CO_TD, was represented by the mean of the differences between CO_TD and CO_NI. Precision, the ability to reproduce the same measurement given the same conditions, was represented by the 95% confidence interval of the mean (¹⁴). Accuracy was defined as the degree to which a noninvasive measurement represented the thermodilution measurement. A second Bland-Altman analysis was plotted for indexed measurements.

The percentage difference between measures of CO_NI and CO_TD and indexed COs was also performed.

To determine the effect of the BSA and tidal volume on the differences between measurements obtained by partial rebreathing and thermodilution, we performed post hoc sensitivity analyses (¹⁵). We plotted BSA and tidal volume against the differences between measurements of CO_NI and CO_TD and against the differences between CI_NI and CI_TD.

Results

Linear regression analysis between CO_NI and CO_TD yielded an r value of 0.83 (P < 0.03) (Fig. 1A). Bland-Altman analysis demonstrated a bias of −0.27 L/min and precision of ±1.49 L/min (Fig. 1B). Interestingly, as seen in Figure 1B, CO_NI underestimated CO_TD at an average CO <3 L/min. As the average CO increased to >3 L/min, CO_NI consistently overestimated CO_TD. The mean difference between all CO_NI and CO_TD measurements was 3.2% ± 17.2%.

Linear regression analysis of CI_NI and CI_TD yielded a decrease in the correlation coefficient to an r value of 0.67 (P = 0.15) (Fig. 1C). Bland-Altman analysis of indexed values demonstrated a bias of −0.18 L · min⁻¹ · m⁻² and a precision of ±2.13 L · min⁻¹ · m⁻² (Fig. 1D). Here too, CI_NI underestimated CI_TD at lower average CIs and overestimated CI_TD at higher average CIs (Fig. 1D). The mean difference between all CI_NI and CI_TD measurements was 5.0% ± 17.0%.

With both CO_NI and CI_NI, each measurement was highly repeatable with only 4% error ±6 from the mean measurement in each patient (vertical error bars, Fig. 1, A and C).

Post hoc sensitivity analysis of BSA and tidal volume as a function of the differences between CO_NI and CO_TD (Fig. 2A) demonstrated that across all BSAs and tidal volumes evaluated, the difference between the 2 measures of CO was <2 L/min. However, post hoc analysis of indexed measurements (Fig. 2B) showed that the differences between the two measures of CI reached values as high as 2.3 L · min⁻¹ · m⁻² especially in patients with BSAs of ≤0.6 m² (closed circles). Ten of the 11 patients with BSAs of ≤0.6 m² were ventilated with tidal volumes of <300 mL (gray circles, Fig. 2B). The five patients causing the most variation with respect to noninvasive CI estimation were in this group.

Discussion

Historically, measurements of CO are clinically useful if they are within 10%–20% of the accepted standard (¹⁶). When comparing measurements obtained by two different techniques, a correlation coefficient of ≥0.8 indicates a strongly positive relationship (¹⁷). An r value between 0.5 and 0.8 indicates moderate correlation (¹⁷). Prior studies in adults comparing CO_NI and CO_TD reported r values of 0.78 or better (¹⁸). In this pediatric cardiac catheterization population, comparison of NICO-derived CO_NI with CO_TD yielded an r value of 0.83. This represents a strongly positive correlation and is comparable with results obtained in adult studies. A bias of −0.27 L/min is quite acceptable in clinical practice and the precision (±1.49 L) was relatively narrow. Although 1.49 L precision compares well to results from prior studies (Table 2), this degree of error could be quite significant in patients with smaller COs (¹⁸). However, precision represents the second standard deviation around the mean (bias) and most of the error was <1.49 L. In fact, the average percentage difference between CO_NI and CO_TD measurements was 3.2% ± 17.2%. This is within clinically accepted error.

The relationship between CO_NI and CO_TD weakened when measurements were indexed to BSA. The correlation decreased to 0.67 and the precision widened to ±2.13 L · min⁻¹ · m⁻². The average percentage difference between CI_NI and CI_TD measurements also increased to 5.0% ± 17.0%. This means that a number of the measures had an error more than the 10%–20% clinically accepted.

Post hoc sensitivity analysis demonstrated that the differences between CO_NI and CO_TD were fairly uniform across all BSAs and tidal volumes. However, the differences between CI_NI and CI_TD increased dramatically at BSAs <0.6 m². The majority of these patients had tidal volumes of <300 mL. Thus, in patients with BSA <0.6 m² and in those ventilated with tidal volumes of <300 mL, noninvasive determination of CI with partial rebreathing was less accurate.

A potential explanation for the diminishing correlation between the measures of CI_NI and CI_TD when the data were indexed is that differences between the two types of measures were magnified when BSAs were <1 m². For instance, in a patient with a BSA of 0.5 m², if the difference between CO_NI and CO_TD is 0.5 L/min, the difference increases to 1 L · min⁻¹ · m⁻² when the data are indexed. By the same token, differences between CI_NI and CI_TD would decrease in patients with a BSA of >1 m².

Furthermore, the 200-mL tidal volume as a lower limit for sensor function probably accounts for the major differences between noninvasively measured CI and indexed measurements obtained by thermodilution in the smaller children. Potentially, the smaller volumes of gas moving through the sensor could lead to error in the calculation of CO. Based on the post hoc data, it seems that 300 mL is the smallest tidal volume that yielded acceptable noninvasive COs and CIs.

Most clinical studies evaluating the partial rebreathing technique have reported overestimation of lower COs noninvasively and underestimation of higher COs (¹⁹). In contrast, our study in children shows that CO_NI underestimated CO_TD at lower average COs and overestimated CO_TD at higher COs. Although opposite to results reported in adults, our findings may be explained by the sources of error proposed by Yem et al. (¹⁹). Recirculation of mixed venous blood with increased CO₂ content may be increased in children with higher COs leading to overestimation of CO. In lower CO states, recirculation may be decreased and CO may be underestimated.

Another source of error is rebreathing time. At lower COs, the rebreathing time needed to accurately determine pulmonary blood flow is longer than the 50 seconds of rebreathing performed noninvasively (¹⁹). Furthermore, the rebreathing time required for accurate measurement in higher CO states is <24 seconds (¹⁹). Therefore, partial rebreathing may underestimate lower COs and overestimate higher COs in children solely on the basis of rebreathing time.

Despite these potential sources of error, the partial rebreathing technique measured CO reproducibly with little error between measures in each patient. Underestimation of lower COs and overestimation of higher COs coupled with highly reproducible values, make the NICO a potential monitor to trend CO during dynamic states. This is because a low CO value reported on the NICO is likely to be truly low and a high CO value is likely to be truly high.

Comparing a technique to thermodilution, which has a margin of error associated with it, has its limitations. Compared with direct Fick method, thermodilution has a bias and precision of 2.3 ± 1.6 L/min (²⁰). In critically ill patients, the bias and precision has been reported to be −0.17 ± 3.13 L/min (²¹). In infants and children, thermodilution measurements vary from Fick-determined measurements by as much as 21% (²²).

In this study, the majority of patients evaluated had relatively healthy lungs and near normal cardiovascular function during a period of static hemodynamics. Demonstration of acceptable correlation of NICO with thermodilution measurement of CO in these patients does not indicate that noninvasive assessment will be accurate in less healthy subjects and during periods of hemodynamic instability. Recognizing the limitations of examining CO measurements in a relatively healthy population, we intended this project to be a pilot study to determine the accuracy of NICO given the nuances of pediatric cardiopulmonary anatomy and physiology. Further investigations will focus on critically ill children with more complex disease processes and a wider range of CI and will build on the results of this investigation. Although ours was not designed as a safety study, it is worth noting that there were no adverse hemodynamic events or respiratory compromise during partial rebreathing.

A further limitation of this study is the exclusion of patients with pulmonary hypertension and intracardiac mixing lesions. Anesthesiologists and intensivists often treat these subgroups of patients empirically and would benefit from accurate CO measurement to guide therapy.

Finally, the 200-mL tidal volume is a major limitation. The current sensor and technology are designed for adult patients with larger tidal volumes. Herein, we have demonstrated that, in terms of accuracy, the lower limit of tidal volume is approximately 300 mL for this sensor. Furthermore, we have shown that with the current technology, the smaller body size limit is a BSA of 0.6 m².

Our exclusion of patients weighing <8 kg was an important but necessary limitation based on the 200-mL tidal volume limitation. With 8 kg as the lower limit, we could safely deliver up to 200 mL of tidal volume (25 mL/kg) without causing excessive barotrauma or excessive inflating pressure (peak pressure <30 cm H₂O). However, this subgroup, including those with a BSA <0.6 m², may be the most important to evaluate with regard to CO measurement. At the time of this study, technology to use partial rebreathing with tidal volumes <200 mL had not been developed. Until this technology becomes available, application of NICO to smaller patients with a tidal volume limitation may be problematic.

Given these limitations, we conclude that noninvasive measurement of CO using partial rebreathing technology may provide clinically acceptable assessment of CO in hemodynamically stable children. Further study is required to establish the accuracy of this technology in patients with a BSA of <0.6 m² and in patients ventilated with <300 mL tidal volume. In addition, determination of CO via partial rebreathing needs to be assessed in children who are hemodynamically unstable.