Noninvasive Cardiac Output Estimation by Inert Gas Rebreathing in Mechanically Ventilated Pediatric Patients

www.jpeds.com • THE JOURNAL OF PEDIATRICS

ORIGINAL ARTICLES

Noninvasive Cardiac Output Estimation by Inert Gas Rebreathing in Mechanically Ventilated Pediatric Patients Amanda M. Perak, MD1,*, Alexander R. Opotowsky, MD, MPH1,2, Brian K. Walsh, RRT-NPS3, Jesse J. Esch, MD, MSc1, James A. DiNardo, MD4, Barry D. Kussman, MB, BCh4, Diego Porras, MD1, and Jonathan Rhodes, MD1 Objective To assess the feasibility and accuracy of inert gas rebreathing (IGR) pulmonary blood flow (Qp) estimation in mechanically ventilated pediatric patients, potentially providing real-time noninvasive estimates of cardiac output. Study design In mechanically ventilated patients in the pediatric catheterization laboratory, we compared IGR Qp with Qp estimates based upon the Fick equation using measured oxygen consumption (VO2) (FickTrue); for context, we compared FickTrue with a standard clinical short-cut, replacing measured with assumed VO2 in the Fick equation (FickLaFarge, FickLundell, FickSeckeler). IGR Qp and breath-by-breath VO2 were measured using the Innocor device. Sampled pulmonary arterial and venous saturations and hemoglobin concentration were used for Fick calculations. Qp estimates were compared using Bland-Altman agreement and Spearman correlation. Results The final analysis included 18 patients aged 4-23 years with weight >15 kg. Compared with the reference FickTrue, IGR Qp estimates correlated best and had the least systematic bias and narrowest 95% limits of agreement (results presented as mean bias ±95% limits of agreement): IGR −0.2 ± 1.1 L/min, r = 0.90; FickLaFarge +0.7 ± 2.2 L/min, r = 0.80; FickLundell +1.6 ± 2.9 L/min, r = 0.83; FickSeckeler +0.8 ± 2.5 L/min, r = 0.83. Conclusions IGR estimation of Qp is feasible in mechanically ventilated patients weighing >15 kg, and agreement with FickTrue Qp estimates is better for IGR than for other Fick Qp estimates commonly used in pediatric catheterization. IGR is an attractive option for bedside monitoring of Qp in mechanically ventilated children. (J Pediatr 2016;177:184-90).

S

erial measurements of cardiac output or pulmonary blood flow (Qp) facilitate optimal management of critically ill children. Unfortunately, reliable bedside measurements of these variables are not readily available, and clinicians frequently rely upon subjective assessments despite poor agreement with objective measurements.1 Inert gas rebreathing (IGR) is an established noninvasive technique to estimate Qp and, in the absence of significant intrapulmonary or intracardiac shunting, systemic cardiac output as well. The latest iterations of IGR technology use 0.5% nitrous oxide (N2O) as the soluble indicator gas and photoacoustic spectroscopy (PAS) for gas analysis, and are able to nontoxically estimate Qp at the bedside in under a minute. IGR reliably estimates Qp in adults,2-8 and recently, we demonstrated that IGR reliably estimates Qp in children with no intracardiac shunt or pure right-to-left shunt.9 However, the studied PAS-based device is designed for spontaneously breathing patients and cannot be directly employed in critically ill, mechanically ventilated patients. A ventilator adaptor assembled from standard medical equipment (anesthesia bag, connectors, and pneumatic valves) is available for insertion between the ventilator circuit and endotracheal tube (ETT) to address this issue, but to date, no study has validated the use of the PAS-based device in this population. From the 1Department of Cardiology, Boston Children’s Hospital, Boston, MA; 2Division of Cardiovascular The primary aim of this study was to assess the feasibility and accuracy of IGR Medicine, Department of Medicine, Brigham and in mechanically ventilated pediatric patients using the PAS-based device. IGR Qp Women’s Hospital, Boston, MA; 3Division of Critical Care, Department of Anesthesiology, Perioperative and Pain estimates were compared with reference Qp estimates obtained on the basis of the Medicine, Boston Children’s Hospital, Boston, MA; and 4Division of Cardiac Anesthesia, Department of Fick principle in patients undergoing clinically indicated cardiac catheterization. Anesthesiology, Perioperative and Pain Medicine, Boston Given the critical importance of oxygen consumption (VO2) in the Fick equaChildren’s Hospital, Boston, MA *Current address: Division of Cardiology, Ann and Robert tion and the well-described inaccuracies of VO2 estimated by published equations,10-12 H. Lurie Children’s Hospital of Chicago, Chicago, IL. we determined Fick Qp using measured VO2 (FickTrue) as the reference test. To Supported by the Tommy Kaplan Fund for Cardiovascular Sciences (to A.P.) and the Dunlevie Family Fund (to A.O.). provide context for the comparison of IGR and FickTrue Qp estimates, we also

ETT FickLaFarge FickLundell FickSeckeler FickTrue IGR

Endotracheal tube Fick using LaFarge-based VO2 Fick using Lundell-based VO2 Fick using Seckeler-based VO2 Fick using measured VO2 Inert gas rebreathing

N2O PAS Qp Qs SF6 VO2

Nitrous oxide Photoacoustic spectroscopy Pulmonary blood flow Systemic blood flow Sulfur hexafluoride Oxygen consumption

Gas canisters, MiniValve adaptor, and disposables for inert gas rebreathing measurements, as well as technical support funds were also provided by InnoCC (Odense, Denmark). Although InnoCC provided technical assistance when requested by the authors during the initial study design and testing, InnoCC had no role in study design, collection, analysis or interpretation of data, drafting or editing of the manuscript, or the decision to submit the paper for publication. The authors declare no conflicts of interest. 0022-3476/$ - see front matter. © 2016 Elsevier Inc. All rights reserved. http://dx.doi.org10.1016/j.jpeds.2016.07.007

184

Volume 177 compared Fick Qp using equation-based VO2 values with FickTrue Qp because these alternative Fick Qp estimates are commonly employed in pediatric catheterization. Finally, we performed comparisons between VO2 values that were directly measured, equation-based, and reverse-calculated from the Fick equation (using measured oxygen saturations, hemoglobin concentration, and IGR Qp).

Methods Patients weighing >9 kg who were scheduled for clinically indicated cardiac catheterization with endotracheal intubation and mechanical ventilation were recruited for the study between September 2014 and March 2015. The weight threshold was chosen based upon theoretical concerns about dead space proportions and gas sampling rate. Patient-related exclusion criteria included left-to-right shunt (based on our previous work demonstrating the unreliability of IGR in this group9) or moment-to-moment instability during IGR or saturation measurements. Catheterization-related exclusion criteria included missing data or implausible saturation data, defined as mixed venous or pulmonary artery saturation >85%. IGR-related exclusion criteria included use of N2O (the soluble test gas) for induction of anesthesia or presence of air leak around the cuffed ETT. Written informed consent to participate in the study was provided by all patients or legal guardians, and the study protocol was approved by the Boston Children’s Hospital Institutional Review Board.

•

October 2016

Hemoglobin was measured from a peripheral blood sample prior to catheterization. Anesthetic induction and endotracheal intubation were performed and mechanical ventilation initiated by a pediatric cardiac anesthesiologist per routine practice. After intubation with a cuffed ETT and stabilization in 21% oxygen (air), the IGR ventilator adaptor was inserted between the ETT and ventilator, and breath-by-breath VO2 and IGR Qp were measured using the PAS device as described below. After completion of the PAS measurements, the adaptor was removed from the ventilator circuit. The process was completed either during sterile preparation of the patient or while vascular access was obtained, with the complete protocol typically requiring less than 10 minutes (3-5 minutes for VO2, 1 minute for IGR), and the catheterization continued without interruption. Clinical and physiologic data were collected on each participant. PAS Device Configuration The PAS-based Innocor device and MiniValve ventilator adaptor (both Innovision, Odense, Denmark) were used to measure VO2 and Qp as shown in Figure 1. The MiniValve adaptor inserts between the ventilator and ETT and is composed of a filter (Hygroboy; Nellcor-Covidien, Boulder, Colorado), a pneumotach (Hans Rudolph 4700B, Shawnee, Kansas), 3 pneumatic valves, and an anesthesia bag. The baseline valve configuration of the adaptor allows the patient to be ventilated by the mechanical ventilator (Figure 1, A); 51.6 mL of dead space is added to the ventilator circuit. When the rebreathing program is activated, the anesthesia bag fills with the predefined test gas mixture and the valve configuration adjusts

Figure 1. A, MiniValve adaptor baseline configuration. The patient’s ETT communicates with the ventilator; the filter, pneumotach, and gas sample line are interposed. B, Rebreathing configuration. The patient’s ETT communicates with the rebreathing bag with the filter, pneumotach, and gas sample line interposed; the ventilator communicates with the room’s air. Modified and reprinted with permission from Innovision. 185

THE JOURNAL OF PEDIATRICS • www.jpeds.com such that the ventilator is open to the room and the ETT communicates with the anesthesia bag (Figure 1, B); the adaptor dead space becomes 61.8 mL. In both configurations, the gas sample line routes 120 mL/min of respiratory gas to the PAS gas analyzer, and both the gas analyzer and pneumotach provide raw data to the Innocor device, which displays a variety of outputs. All PAS measurements were undertaken by one of 3 trained investigators without knowledge of the catheterization data or equation-based VO2 estimates. The PAS device was calibrated to room air, volumes, and time delay once daily using a 1-Lr calibration syringe (Hans Rudolph, Shawnee, Kansas). IGR Qp Measurement Prior to initiating a rebreathing test, the alveolar N2O concentration was measured by the PAS device via the gas sampling line in the baseline MiniValve configuration. To proceed, this concentration was required to be <0.0025%. The basic technique of IGR with the Innocor system in spontaneously breathing patients has been reported in detail2-4 and is the basis for the protocol in ventilated patients. Briefly, an anesthesia bag is filled with test gas, which is a mixture of a blood-soluble gas (0.5% N2O), an inert insoluble gas (0.1% sulfur hexafluoride [SF6]), oxygen (28.3%), and nitrogen (the balance). The system allows for further enrichment of oxygen if needed, but for this study the test gas described above was used. During rebreathing of the test gas in a closed system (Figure 1, B) over a 5- to 6-breath, 10- to 20-second interval, the PAS analyzers continuously measure gas concentrations from the sample line. Equilibration of the insoluble SF6 concentration occurs after 2-3 breaths, and this both indicates adequate mixing of test gas within the total system (lungs, ETT, and MiniValve adaptor including rebreathing bag), and allows calculation of the total system volume. After this mixing has occurred, the subsequent rate of N2O disappearance reflects its uptake by pulmonary blood flow. These data (total system volume and rate of N 2 O disappearance) permit calculation of the effective Qp that participates in gas exchange. For this study of mechanically ventilated patients, the test gas volume was set somewhat higher (generally by 100200 mL) than the patient’s tidal volume prior to starting the test (which had been selected by the anesthesiologist to provide adequate ventilation as per routine practice); this increase was to account for the continuous gas sampling (120 mL/min) and MiniValve adaptor dead space (61.8 mL). Manual bag ventilation during the rebreathing test was titrated to allow complete emptying of the bag with each inspiration, ~ 12-15 breaths/min in most cases. After completion of the rebreathing test, the Innocor displays of measured Qp and gas time-concentration curves were reviewed by the investigators and checked for acceptability (adequate equilibration of SF6 prior to analysis of N2O disappearance slope, absence of spikes or other artifacts in the curves). 186

Volume 177 Generation of VO2 and Fick Qp Estimates The Appendix (available at www.jpeds.com) provides detailed descriptions of the PAS VO 2 measurement, VO2−estimating equations, cardiac catheterization procedure, Fick calculations, and reverse-Fick VO2 equation derivation. Briefly, VO2 was measured with the PAS device via the breath-by-breath method of expired gas analysis,13,14 and the value was averaged over >1 minute after stability. VO2 was also estimated using the equations of LaFarge,15 Lundell,16 and Seckeler.17 Cardiac catheterization was performed as per routine (without knowledge of the IGR data), and oxygen saturation data for the pulmonary vein (or where appropriate, systemic artery) and pulmonary artery were recorded from the catheterization report. The Fick equation18 was then applied to calculate Qp estimates from the measured (FickTrue) or equationbased (Fick using LaFarge-based VO2 [FickLaFarge], Fick using Lundell-based VO2 [Fick Lundell ], and Fick using Seckelerbased VO2 [FickSeckeler]) VO2 values, measured hemoglobin concentration, and sampled oxygen saturations. Finally, the Fick equation was rearranged and applied to back-calculate VO2 from the IGR Qp and measured hemoglobin concentration and oxygen saturations. Statistical Analyses Feasibility is described qualitatively. Categorical data are presented as numbers (percentages) and continuous data are presented as median (IQR). Agreement between tests was primarily assessed with Bland-Altman analysis, but Spearman correlation coefficients are also reported. These comparisons were made for IGR Qp with FickTrue Qp as well as for each of the 3 other types of Fick Qp estimates (FickLaFarge, FickLundell, and FickSeckeler) with FickTrue Qp. Secondarily, agreement was analyzed between measured VO2, equation-based VO2 (LaFarge, Lundell, and Seckeler), and VO2 reverse-calculated from the Fick equation. Statistical analyses were performed using version 9.3 of the SAS System for Windows (SAS Institute Inc, Cary, North Carolina) and version 6.0 of GraphPad Prism (GraphPad Software, La Jolla, California).

Results PAS and catheterization measurements were attempted in 29 patients, and 18 patients were included in the final analysis. Of the 11 excluded patients, 7 were for reasons unrelated to PAS measurement (Figure 2; available at www.jpeds.com). Characteristics of the 18 patients whose data were analyzed are presented in Table I. This final sample was 50% female, with an age range of 4-23 years and weight range of 16.6-72.4 kg. Cardiac diagnoses are listed in Table II. Three patients (17%) had right-to-left intracardiac shunting; the remainder had no shunt lesion. Qp as measured by the reference test, FickTrue, ranged from 0.8 to 6.0 L/min. All 18 included patients demonstrated stable, physiologic-range vital signs including capnography and pulse oximetry at the time of IGR measurement and catheterization saturation acquisition. Perak et al

ORIGINAL ARTICLES

October 2016

Table I. Demographic and anthropometric characteristics of the study patients Patient characteristics Female Age (y) Height (cm) Weight (kg) BSA (kg/m2)

9 (50%) 11 [7-17] 148 [118-161] 41.2 [23.1-54.4] 1.30 [0.86-1.84]

Continuous variables are presented as median [IQR]; categorical variables are presented as N (%).

Feasibility of PAS-Based Measurements There were no adverse events. Of the 29 patients evaluated, 4 had to be excluded from the analysis specifically for issues that precluded accurate PAS measurements: 3 with N2O used for anesthetic induction and 1 with ETT leak. Notably, 1 patient who had received N 2 O for anesthetic induction was reevaluated at the conclusion of the catheterization procedure (>1 hour after induction) and still had levels of alveolar N2O that were too high (>0.0025%) to permit accurate IGR measurement. All of the 18 patients in the final sample after the prespecified exclusions had stability of the VO2 curves and proper equilibration and smoothness of the IGR gas time-concentration curves on visual inspection. However, review of IGR gastime concentration curves for the 11 excluded patients identified 3 patients with poor quality data (spikes, poor mixing); these 3 patients had been excluded for reasons unrelated to PAS (left-to-right shunt in 2, instability in 1), but notably each had body weight <15 kg. In contrast, 3 of the 18 patients in the final study sample had body weights 15-20 kg (16.6, 17.0, and 17.3 kg, respectively); the IGR gas time-concentration curves for these 3 patients were adequate for analysis. Agreement Between Qp Measurements Figure 3 demonstrates better agreement with FickTrue Qp estimates for IGR compared with the other Fick Qp estimates. Compared with FickTrue, IGR Qp had mean bias −0.2 L/min,

95% limits of agreement −1.3 to +1.0 L/min, and r = 0.90 (P < .001). On the other hand, compared with FickTrue, FickLaFarge Qp had larger mean bias (+0.7 L/min), wider 95% limits (−1.4 to +2.9 L/min), and r = 0.80 (P < .001). Similarly, FickSeckeler vs FickTrue had mean bias +0.8 L/min, 95% limits −1.6 to +3.3 L/ min, r = 0.83 (P < .001). Agreement with FickTrue was poorest for FickLundell: mean bias +1.6 L/min, 95% limits −1.3 to +4.5 L/ min, and r = 0.83 (P < .001). Considered from another perspective, IGR Qp was within 20% of FickTrue Qp in 72% of patients. In contrast, Qp estimates based on FickLaFarge, FickSeckeler, and FickLundell were within 20% of FickTrue Qp in only 39%, 33%, and 6%, respectively. Agreement between VO2 Values Bland-Altman analysis of agreement between VO2 values is shown in Figures 4 and 5 (available at www.jpeds.com). Mean measured VO2 for the 18 included patients was 138 mL/min. Agreement with measured VO2 was similar for LaFarge and Seckeler equations (LaFarge bias +28 mL/min, 95% limits ± 93 mL/min, r = 0.71; Seckeler + 32 mL/min, ± 103 mL/min, r = 0.65), and it was poorer for the Lundell equations (+66 mL/ min, ±105 mL/min, r = 0.62). When compared with the VO2 reverse-calculated from the Fick equation (Figure 5), the measured VO2 had minimal bias of +8 mL/min, 95% limits of agreement ±65 mL/min, and r = 0.73 (P < .001); the equationbased VO2 values showed poorer agreement with reversecalculated VO2 (LaFarge: +36 mL/min, ±67 mL/min, r = 0.66; Seckeler: +40 mL/min, ±78 mL/min, r = 0.64; Lundell: +74 mL/ min, ±86 mL/min, r = 0.60).

Discussion This report builds on previous investigations by demonstrating the feasibility and accuracy of Qp measurement using current-generation IGR and PAS technology in mechanically ventilated pediatric patients with and without structural heart disease. This experience demonstrates that IGR is consistently feasible in patients weighing >15 kg with no ETT leak

Figure 3. Bland-Altman plots comparing Qp measurements. A, For IGR vs FickTrue Qp, mean bias of IGR was −0.2 L/min and 95% limits of agreement were −1.3 to +1.0 L/min. B, For FickLaFarge vs FickTrue Qp, mean bias of FickLaFarge Qp was +0.7 L/min and 95% limits were −1.4 to +2.9 L/min. C, For FickSeckeler vs FickTrue Qp, mean bias of FickSeckeler was +0.8 L/min and 95% limits were −1.6 to +3.3 L/min. D, For FickLundell vs FickTrue Qp, mean bias of FickLundell was +1.6 L/min and 95% limits were −1.3 to +4.5 L/min. Blue lines show mean bias and dotted lines show 95% limits of agreement. Mean Qp by FickTrue was 2.9 L/min for all comparisons. Noninvasive Cardiac Output Estimation by Inert Gas Rebreathing in Mechanically Ventilated Pediatric Patients

187

THE JOURNAL OF PEDIATRICS • www.jpeds.com Table II. Primary cardiac diagnoses of the study patients, in order of decreasing frequency Diagnoses

N

Structurally normal heart; history of arrhythmia* Tetralogy of Fallot (pulmonary stenosis/atresia), status post repair Functionally single ventricle, status post Fontan Hypertrophic cardiomyopathy Truncus arteriosus, status post repair Double-outlet, hypoplastic right ventricle, pulmonary atresia, status post 1.5-ventricle repair Isolated right pulmonary artery from the ascending aorta, status post repair Multiple cardiac rhabdomyomas; tuberous sclerosis Midaortic syndrome; neurofibromatosis Total

4 4 4 1 1 1 1 1 1 18

*Three with history of supraventricular tachycardia because of an accessory pathway, 1 status postventricular fibrillation cardiac arrest; all were in sinus rhythm at the time of the study data acquisition.

and no recent exposure to the blood-soluble test gas, N2O. IGR estimates of Qp (which, in the absence of shunts, equals cardiac output) agreed well with Fick estimates using measured VO2 (FickTrue). Moreover, given that Fick Qp estimates using measured VO2 (FickTrue) are considered the reference standard, it is notable that FickTrue’s agreement with IGR was considerably stronger than was FickTrue’s agreement with any of the Fick Qp estimates using equation-based VO2 values (FickLaFarge, FickLundell, or FickSeckeler). Correspondingly, each of the equations for VO2 showed significant bias compared with both measured VO2 and VO2 calculated from the Fick equation using IGR Qp. Although IGR was feasible in most patients, 3 distinct exceptions should be noted. First, patients with ETT leaks do not satisfy the criteria for a closed rebreathing system and IGR should not be performed in this context. Second, patients with recent exposure to anesthetic-dose N2O seem to have prolonged presence of alveolar N2O at levels high enough to preclude IGR testing, which relates to the fact that the N2O concentrations used for the IGR test are extremely low (initially 0.5%, with abrupt decreases upon respiratory dilution and bloodstream uptake). In contrast, alveolar washout of the small amount of N2O used for an IGR test generally occurs in about 5 minutes, allowing for a serial monitoring of Qp by IGR.9,19 Third, patients <15 kg seem to have difficulty achieving the rapid and complete gas mixing required for IGR with the current MiniValve adaptor, which leads to unreliable results. This failure of mixing is likely due to disproportionately large dead space to tidal volume ratios with the adaptor in place, as well as excessive reduction of tidal volume by continuous gas sampling from the closed IGR circuit. These factors could probably be mitigated if the adaptor were further miniaturized. In addition, it should be noted that cardiac patients with significant left-to-right shunt lesions were excluded from this study, as IGR is unreliable in this population. Our previous work demonstrated a large negative bias of IGR Qp and poor agreement with the reference tests, probably related to early recirculation of the soluble test gas, N2O.9 In this small study, IGR measured Qp with less mean bias and greater precision than did other methods commonly em188

Volume 177 ployed in the pediatric catheterization laboratory. This suggests that the level of accuracy and precision of the IGR technology will be acceptable to clinicians for many applications. In particular, availability of the MiniValve adaptor now permits the use of IGR in the critically ill, mechanically ventilated patients for whom cardiac output measurements might have the greatest clinical value. Although the inability to apply current IGR technology to patients with weight <15 kg or leftto-right shunts may restrict its use in the pediatric cardiac intensive care unit, these limitations may be less restrictive in the pediatric general medical and surgical intensive care units (and use of N2O anesthetic for surgery can be avoided, or if it used, the PAS device can determine when washout is sufficient for IGR measurement). In fact, for patients with weight >15 kg and no shunt, the best use of IGR may well be as a noninvasive cardiac output monitor, because, in the absence of shunt, Qp is equal to cardiac output (systemic blood flow [Qs]) and no blood sampling would be required. The noninvasive nature and ease of use of the point-of-care PAS-based IGR device are significant advantages in the pediatric intensive care unit; we believe that with proper training, IGR measurements can be reliably performed by intensivists, respiratory therapists, and other clinicians providing care to intubated patients. This study builds upon a growing body of literature documenting the accuracy of current-generation PAS-based IGR technology for estimating cardiac output. Several studies of IGR in spontaneously breathing adults with heart failure or pulmonary hypertension have shown that it is accurate and precise,2-5,7,8 even in the presence of significant parenchymal lung disease,6 and has incremental value to predict outcomes.20,21 Recently, we demonstrated that IGR is also accurate in spontaneously breathing children and adults with congenital heart disease, as long as no significant left-to-right shunt lesion is present.9 Although the present study is unique as the first evaluation of current IGR technology in intubated patients, there are several reports of cardiac output estimation in intubated patients using other IGR devices. For example, Reutershan et al22-24 reported 3 studies of intubated, critically ill adults using a previous generation IGR device which used a similar PAS-based gas analyzer, but 10-fold higher concentrations of the test gases and a more cumbersome ventilation system. These studies demonstrated that when IGR Qp estimates were combined with an arterial saturation measurement to correct for intrapulmonary shunt, the resulting IGR-based Qs agreed well with Qs measured by thermodilution. Furthermore, IGR Qp, as an estimate of effective Qp, was useful for predicting which patients with acute respiratory distress syndrome would respond to prone positioning.24 Future study is needed in larger cohorts of pediatric medical, surgical, and cardiac critical care patients to evaluate the value of current IGR technology for monitoring systemic cardiac output (Qs) and effective Qp in these settings, as well as to evaluate the use of IGR in intubated patients <15 kg when smaller ventilator adaptors with less dead space become available. Perak et al

October 2016 Of particular importance to pediatric cardiologists are our observations that equation-based (LaFarge, Lundell, or Seckeler) VO2 values agreed poorly with measured VO2, and the agreement between measured VO2 and Fick-calculated VO2 was stronger than the agreement between equation-based VO2 and either measured or Fick-calculated VO2. These findings are consistent with a number of studies that have critically assessed the accuracy of equations used to estimate VO2 in the pediatric cardiac catheterization laboratory and have concluded that relying on an assumed VO 2 can be a major source of error.11,12,17,25,26 The potential clinical significance of such error in estimation of VO2, and, therefore, an equivalent percentage of error in the Fick Qs and Qp, as well as subsequent pulmonary vascular resistance calculations, should not be underestimated. Several authors have pointed out the implications of unreliable pulmonary vascular resistance measurements in particular, given the importance of this value in driving diagnosis, prognosis, and therapy for pulmonary hypertension,27 as well as potential impact on risk stratification or even candidacy for cardiac surgeries including heart transplant, closure of intracardiac shunts, and single ventricle palliation.28,29 With these limitations and potential implications in mind, alternative methods of Qp estimation, such as IGR, may become an important addition to the invasive assessments used to aid in clinical decision-making in children with cardiovascular disease. This was a small study of IGR feasibility and accuracy, with conclusions and generalizability limited by the sample size. Our sample was drawn from a population of pediatric cardiac patients undergoing nonurgent cardiac catheterization; thus study in other target populations including critically ill children in the medical, surgical, and cardiac intensive care units is needed. For example, we did not evaluate IGR in children with requirements for high inspired oxygen concentration or positive end-expiratory pressure levels, which may be important because the IGR test gas for the studied PAS device has a maximal oxygen concentration of 57.5% and positive endexpiratory pressure is not provided during the (~20-second) IGR test. Furthermore, because we did not perform repeated IGR trials to systematically assess IGR repeatability or precision, we cannot determine what proportion of the 95% limits of agreement between IGR and FickTrue Qp is due to imprecision in the IGR vs the Fick measurement.30 Finally, our IGR and catheterization saturation data were not obtained at precisely the same moment in all cases; therefore we cannot exclude the possibility that the true, underlying Qp was different at the time of IGR vs saturation measurement. However, the time interval between measurements was minimal, and vital signs (including capnography and pulse oximetry) were stable and physiologic during IGR and saturation data acquisition for all analyzed patients (making serious acid-base derangements unlikely). Moreover, if random variation in Qp were present, our results would tend to underestimate the true agreement between IGR and FickTrue Qp. Similarly, hemoglobin concentration was often measured on the day prior to catheterization (as is our institution’s clinical practice), but if random variation were present, agreement between IGR and FickTrue Qp would be underestimated.

ORIGINAL ARTICLES In pediatric patients >15 kg who are intubated and mechanically ventilated, use of the ventilator-adapted PAS device is feasible and produces IGR Qp estimates that agree well with Fick Qp estimates when VO2 is measured. Equation-based VO2 agrees poorly with measured VO2 and, therefore, Qp calculated by Fick using equation-based VO2 agrees poorly with that using measured VO2. Further study of IGR is warranted, as it may be an attractive option for bedside monitoring of Qp in intubated pediatric patients, as well as a useful adjunctive measure to invasive testing in patients for whom the Qp estimate and related variables are critical. ■ We are grateful to Jørgen Grønlund Nielsen and Peter Clemensen (InnoCC) for technical support. Submitted for publication Mar 2, 2016; last revision received May 13, 2016; accepted Jul 6, 2016 Reprint requests: Jonathan Rhodes, MD, Department of Cardiology, Boston Children’s Hospital, 300 Longwood Ave, Boston, MA 02115. E-mail: [email protected]

References 1. Tibby SM, Hatherill M, Marsh MJ, Murdoch IA. Clinicians’ abilities to estimate cardiac index in ventilated children and infants. Arch Dis Child 1997;77:516-8. 2. Gabrielsen A, Videbaek R, Schou M, Damgaard M, Kastrup J, Norsk P. Noninvasive measurement of cardiac output in heart failure patients using a new foreign gas rebreathing technique. Clin Sci 2002;102:24752. 3. Peyton PJ, Thompson B. Agreement of an inert gas rebreathing device with thermodilution and the direct oxygen Fick method in measurement of pulmonary blood flow. J Clin Monit Comput 2004;18:373-8. 4. Agostoni P, Cattadori G, Apostolo A, Contini M, Palermo P, Marenzi G, et al. Noninvasive measurement of cardiac output during exercise by inert gas rebreathing technique: a new tool for heart failure evaluation. J Am Coll Cardiol 2005;46:1779-81. 5. Farina S, Teruzzi G, Cattadori G, Ferrari C, De Martini S, Bussotti M, et al. Noninvasive cardiac output measurement by inert gas rebreathing in suspected pulmonary hypertension. Am J Cardiol 2014;113:546-51. 6. Saur J, Trinkmann F, Doesch C, Scherhag A, Brade J, Schoenberg SO, et al. The impact of pulmonary disease on noninvasive measurement of cardiac output by the inert gas rebreathing method. Lung 2010;188:43340. 7. Saur J, Fluechter S, Trinkmann F, Papavassiliu T, Schoenberg S, Weissmann J, et al. Noninvasive determination of cardiac output by the inert-gasrebreathing method – comparison with cardiovascular magnetic resonance imaging. Cardiology 2009;114:247-54. 8. McLure LER, Brown A, Lee WN, Church AC, Peacock AJ, Johnson MK. Non-invasive stroke volume measurement by cardiac magnetic resonance imaging and inert gas rebreathing in pulmonary hypertension. Clin Physiol Funct Imaging 2011;31:221-6. 9. Marma AK, Opotowsky AR, Fromm BS, Ubeda-Tikkanen A, Porras D, Rhodes J. Noninvasive cardiac output estimation by inert gas rebreathing in pediatric and congenital heart disease. Am Heart J 2016;174:80-8. 10. Narang N, Thibodeau JT, Levine BD, Gore MO, Ayers CR, Lange RA, et al. Inaccuracy of estimated resting oxygen uptake in the clinical setting. Circulation 2014;129:203-10. 11. Fakler U, Pauli C, Hennig M, Sebening W, Hess J. Assumed oxygen consumption frequently results in large errors in the determination of cardiac output. J Thorac Cardiovasc Surg 2005;130:272-6. 12. Li J, Bush A, Schulze-Neick I, Penny DJ, Redington AN, Shekerdemian LS. Measured versus estimated oxygen consumption in ventilated patients with congenital heart disease: the validity of predictive equations. Crit Care Med 2003;31:1235-40.

Noninvasive Cardiac Output Estimation by Inert Gas Rebreathing in Mechanically Ventilated Pediatric Patients

189

THE JOURNAL OF PEDIATRICS • www.jpeds.com 13. Sheth SS, Maxey DM, Drain AE, Feinstein JA. Validation of the Innocor device for noninvasive measurement of oxygen consumption in children and adults. Pediatr Cardiol 2013;34:847-52. 14. Guo L, Cui Y, Pharis S, Walsh M, Atallah J, Tan M-W, et al. Measurement of oxygen consumption in children undergoing cardiac catheterization: comparison between mass spectrometry and the breath-bybreath method. Pediatr Cardiol 2013;35:798-802. 15. LaFarge CG, Miettinen OS. The estimation of oxygen consumption. Cardiovasc Res 1970;4:23-30. 16. Lundell BP, Casas ML, Wallgren CG. Oxygen consumption in infants and children during heart catheterization. Pediatr Cardiol 1996;17:20713. 17. Seckeler MD, Hirsch R, Beekman RH, Goldstein BH. A new predictive equation for oxygen consumption in children and adults with congenital and acquired heart disease. Heart 2015;101:517-24. 18. Fick A. Uber die messung des Blutquantums in den Hertzvent rikeln. Sitzber Physik Med Ges Wurzburg 1870;36. 19. Damgaard M, Norsk P. Effects of ventilation on cardiac output determined by inert gas rebreathing. Clin Physiol Funct Imaging 2005;25: 142-7. 20. Goda A, Lang CC, Williams P, Jones M, Farr MJ, Mancini DM. Usefulness of non-invasive measurement of cardiac output during submaximal exercise to predict outcome in patients with chronic heart failure. Am J Cardiol 2009;104:1556-60. 21. Lang CC, Karlin P, Haythe J, Lim TK, Mancini DM. Peak cardiac power output, measured noninvasively, is a powerful predictor of outcome in chronic heart failure. Circ Heart Fail 2009;2:33-8. 22. Reutershan J, Ressel M, Wagner T, Schröder T, Dietz K. Fretschner R. Non-invasive determination of effective pulmonary blood flow: evalua-

190

Volume 177

23.

24.

25.

26.

27.

28.

29.

30.

tion of a simplified rebreathing method. J Med Eng Technol 2003;27:1949. Reutershan J, Kapp T, Unertl K. Fretschner R. Noninvasive determination of cardiac output in ventilated patients. Clinical evaluation of a simplified quick method. Anaesthesist 2003;52:778-86. Reutershan J, Schmitt A, Dietz K, Fretschner R. Non-invasive measurement of pulmonary blood flow during prone positioning in patients with early acute respiratory distress syndrome. Clin Sci 2004;106:3-10. Schmitz A, Kretschmar O, Knirsch W, Woitzek K, Balmer C, Tomaske M, et al. Comparison of calculated with measured oxygen consumption in children undergoing cardiac catheterization. Pediatr Cardiol 2008;29:10548. Rutledge J, Bush A, Shekerdemian L, Schulze-Neick I, Penny D, Cai S, et al. Validity of the LaFarge equation for estimation of oxygen consumption in ventilated children with congenital heart disease younger than 3 years – a revisit. Am Heart J 2010;160:109-14. Fares WH, Blanchard SK, Stouffer GA, Chang PP, Rosamond WD, Ford HJ, et al. Thermodilution and Fick cardiac outputs differ: impact on pulmonary hypertension evaluation. Can Respir J 2012;19:261-6. Giglia TM, Humpl T. Preoperative pulmonary hemodynamics and assessment of operability: is there a pulmonary vascular resistance that precludes cardiac operation? Pediatr Crit Care Med 2010;11:S57-69. Shanahan CL, Wilson NJ, Gentles TL, Skinner JR. The influence of measured versus assumed uptake of oxygen in assessing pulmonary vascular resistance in patients with a bidirectional Glenn anastomosis. Cardiol Young 2003;13:137-42. Critchley LA, Critchley JA. A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. J Clin Monit Comput 1999;15:85-91.

Perak et al

ORIGINAL ARTICLES

October 2016

Appendix Methods Breath-by-Breath VO2 Measurement. VO2 was measured with the breath-by-breath method of expired gas analysis, as previously reported13 and validated against mass spectroscopy in intubated pediatric patients.14 With the MiniValve apparatus in the baseline configuration (Figure 1, A), the PAS device calculates the difference between oxygen volume inspired and expired in each breath by integrating the product of oxygen concentration and flow in the respiratory gas over the entire respiratory cycle. Real-time VO2 curves were displayed, and the VO2 value was averaged over >1 minute after stability of the curve. Estimation of VO2 with Predictive Equations. For each patient, indexed VO2 was estimated using the equations of LaFarge,15 Lundell,16 and Seckeler.17 The absolute (unindexed) VO2 was obtained by multiplying the calculated indexed value by body surface area (BSA). The LaFarge equations, derived in 1970 in a mixed cohort of 879 sedated patients ages 3-40 years, were chosen for this analysis because they are used to generate VO 2 values in most pediatric catheterization laboratories.18 The Lundell equations are one of the other more prevalent methods used for VO2 estimation in pediatrics and were derived in a cohort of 504 patients with an age range (0-27 years) similar to that in the current analysis; however, as with the LaFarge cohort, patients in this cohort were also sedated (rather than under general anesthesia). The Seckeler equation was recently developed in a cohort of 502 patients aged 0-59 years, and as far as we are aware, it is the only formula for VO2 estimation that was derived in patients under general anesthesia and with an age range inclusive of that in our analysis. The LaFarge formulas were used as follows:

for males: VO2 BSA [(mL/min ) /m 2 ]

= 138.1 − 11.49 × ln (age ) + 0.378 × HR , and

for females: VO2 BSA [(mL/min ) /m 2 ] = 138.1 − 17.04 × ln (age ) + 0.378 × HR,

where age is in years and HR is the heart rate recorded at the time of IGR and saturation sampling, in beats per minute. The Lundell formulas were used as follows:

for patients < 3 years of age: VO2 BSA [(mL/min ) /m 2 ] = 3.42 × height − 7.83 × weight + 0.38 × HR − 54.1 for males ≥ 3 years of age: VO2 BSA [(mL/min ) /m 2 ] = 0.79 × HR − 7.4 × BSA A + 108.1 for females ≥ 3 years of age: VO2 BSA [(mL/min ) /m 2 ] = 0.77 × HR − 5.2 × BSA + 106.8

where height is in cm, weight is in kilograms, and HR is heart rate in beats per minute. The Seckeler formula was used as follows:

VO2 BSA [(mL/min ) /m 2 ]

= 242.1 + 9.7 × [ ln (age ) − 34 × ln ( weight ) − 9.6 × (single ventricle ) − 11.2 (critical illness ) ,

where age is in years, weight is in kilograms, and the dichotomous variables “single ventricle” and “critical illness” are each given a value of 1 for yes or 0 for no. Cardiac Catheterization. Clinically indicated catheterization was performed as per routine: end-hole catheters were used to obtain blood samples from the pulmonary artery, systemic artery, and when clinically indicated, pulmonary vein. Oxygen saturations were measured from these samples with an oximeter (Avoximeter 1000E; Accriva Diagnostics, San Diego, California). The resulting saturation data were recorded from the final catheterization report. Fick Calculations of Qp and VO2. The Fick equation,19 based upon the body’s oxygen consumption divided by the arteriovenous oxygen content differential, was used to calculate 4 Qp estimates: (1) FickTrue, using PAS-based measurements of VO2; (2) FickLaFarge, using VO2 from the LaFarge equations; (3) FickLundell, using VO2 from the Lundell equations; and (4) FickSeckeler, using VO2 from the Seckeler equation. The arteriovenous oxygen content differential was calculated using the measured hemoglobin and oxygen saturations from the pulmonary vein (or where appropriate, systemic artery) and pulmonary artery. To back-calculate the VO2 using the IGR measurement of Qp in combination with the measured hemoglobin and oxygen saturations, the Fick equation was rearranged as follows:

Qp = VO2

{[Sa O2 pv

100 × 13.6 × Hgb + 0.03( Ppv O2 )]

− [ Sa O2 pa 100 × 13.6 × Hgb + 0.03( Ppa O2 )]}

But because the study was done without significant supplemental oxygen, PpvO2 (as well as PpaO2) is negligible. Thus,

VO2 = Qp × {[Sa O2 pv 100 × 13.6 × Hgb] − [Sa O2 pa 100 × 13.6 × Hgb]}

Rearranging and simplifying:

VO2 = Qp × 13.6 × Hgb × ( Sa O2 pv − Sa O2 pa ) 100 , and VO2 = Qp × 0.136 × Hgb × ( Sa O2 pv − Sa O2 pa ) ;

where Qp is in L/min, VO2 is in mL/min, Hgb is hemoglobin (g/dL), SaO2pv and SaO2pa are oxygen saturations in the pulmonary vein (or systemic artery) and pulmonary artery, respectively, and PpvO2 and PpaO2 are the partial pressures of oxygen in the pulmonary vein (or systemic artery) and pulmonary artery, respectively.

Noninvasive Cardiac Output Estimation by Inert Gas Rebreathing in Mechanically Ventilated Pediatric Patients

190.e1

THE JOURNAL OF PEDIATRICS • www.jpeds.com

Volume 177

Figure 2. Flow diagram of study participants.

Figure 4. Bland-Altman plots comparing assumed with measured VO2. A, For LaFarge vs measured VO2, mean bias of LaFarge VO2 was +28 mL/min and 95% limits were −66 to +121 mL/min. B, For Seckeler vs measured VO2, mean bias of Seckeler VO2 was +32 mL/min and 95% limits were −70 to +135 mL/min. C, For Lundell vs measured VO2, mean bias of Lundell VO2 was +66 mL/min and 95% limits were −39 to +171 mL/min. Blue lines show mean bias and dotted lines show 95% limits of agreement. Mean measured VO2 was 138 mL/min for all comparisons.

190.e2

Perak et al

October 2016

ORIGINAL ARTICLES

Figure 5. Bland-Altman plots comparing measured and assumed VO2 to Fick-calculated VO2. A, For measured vs Fickcalculated VO2, mean bias of measured VO2 was +8 mL/min and 95% limits were −57 to +73 mL/min. B, For LaFarge vs Fickcalculated VO2, mean bias of LaFarge VO2 was +36 mL/min and 95% limits were −31 to +103 mL/min. C, For Seckeler vs Fick-calculated VO2, mean bias of Seckeler VO2 was +40 mL/min and 95% limits were −37 to +118 mL/min. D, For Lundell vs Fick-calculated VO2, mean bias of Lundell VO2 was +74 mL/min and 95% limits were −12 to +160 mL/min. Blue lines show mean bias and dotted lines show 95% limits of agreement. Mean Fick-calculated VO2 was 130 mL/min for all comparisons.

Noninvasive Cardiac Output Estimation by Inert Gas Rebreathing in Mechanically Ventilated Pediatric Patients

190.e3