Performance of Noninvasive Partial CO²Rebreathing Cardiac Output and Continuous Thermodilution Cardiac Output in Patients Undergoing Aortic Reconstruction Surgery

After institutional review board approval, 28 patients undergoing elective aortic reconstruction for infrarenal abdominal aortic aneurysm were enrolled in this study when their informed consent was obtained. Anesthetic management was standardized in these patients as follows. After epidural catheterization at the Th10/11 or 11/12 interspace, anesthesia was induced with intravenous fentanyl and propofol and maintained with sevoflurane inhalation with or without nitrous oxide. Patients were paralyzed with vecuronium and mechanically ventilated with either the AS/3 ADU (Datex-Ohmeda, Helsinki, Finland) or KION (Siemens, Solna, Sweden) anesthesia machine. Tidal volume and respiratory rate were initially set at 10 ml/kg and 10 breaths per minute, respectively, and were adjusted to maintain Paco²at 35–45 mmHg. After tracheal intubation, cardiac output was continuously monitored using the NICO monitor (software version 3.1). Cardiac output data obtained with the NICO in average mode were downloaded to a computer for analysis. The NICO sensor was placed distal to the heat and moisture exchanger (Hygrobac S, DAR-Mallinckrodt, Mirandola, Italy). Meticulous attention was paid to maintenance of adequate rebreathing circuit volume during monitoring. An 8-French pulmonary artery catheter (746HF8, Baxter Healthcare, Irvine, CA) was inserted via  the right internal jugular vein. Continuous thermodilution cardiac output (CCO) and mixed venous oxygen saturation were continuously monitored with a Vigilance monitor (Baxter Healthcare, Irvine, CA). The calibration of mixed venous oxygen saturation and subsequent blood gas analysis were performed using a standard blood gas analyzer with Cooximeter (Stat Profile M, Nova Biochemicals, Waltham, MA). These data were downloaded to another computer every 30 s, and data averaged for 3 min before bolus thermodilution cardiac output (BCO) measurement were used for analysis. Fluid administration, blood transfusion, epidural injection of local anesthetics, administration of inotropic and vasodilatory drugs, and other anesthetic management procedures were performed at the discretion of the attending anesthesiologist.

BCO measurements were made at the following predetermined time points: (1) after anesthetic induction (Postinduction); (2) during aortic cross-clamping (XC); (3) at reperfusion of unilateral iliac artery (Declamp); and (4) during peritoneal closure (Endop). The injection of ice-cold saline was repeated four times within 3 min and averaged data were used for analysis. These BCO measurements were performed following the stabilization of intravascular fluid status and cardiac performance. This period was arbitrarily defined as being at least 5 min, during which there was no fluid challenge or change in pharmacologic intervention. Blood gas analysis data were input to the NICO monitor before each BCO measurement to allow precise estimations of shunt fraction and NICO values. At each BCO measurement, the shunt fraction was calculated using the standard formula (see  Appendix).

NICO monitoring was terminated postoperatively, but CCO monitoring was continued during the postoperative period in some patients when clinically indicated. After patients were fully awake and showed stable hemodynamics and acceptable ventilation and oxygenation, they were extubated in the operating room before being transferred to the intensive care unit.

Data are expressed as mean ± SD and were statistically analyzed using the Statview 5.0 software (SAS Institute, Chicago, IL). Correlations between NICO and BCO data and between CCO and BCO data sets were determined by linear regression. A Bland-Altman plot was used to compare the bias (the mean of the differences) and precision (SD of bias) of NICO and CCO against BCO. 11In addition, the relative error (expressed in %) was defined as 100*([either NICO or CCO-BCO]/BCO). 12Intraoperative changes in other parameters were analyzed using a one-way ANOVA with repeated measurements. A value of P < 0.05 was considered significant. Multiple comparisons with Scheffé correction were performed when significant changes were detected.

Demographic and operative data are summarized in table 1. Intraoperative cardiac output as measured by NICO, CCO, and BCO; other variables are summarized in table 2. The interval between the aortic cross-clamp and the second measurement (XC), and the aortic declamp and the third measurement (Declamp) were 17 ± 5 and 20 ± 6 min, respectively. Overall correlations between NICO and BCO and CCO and BCO are expressed in figure 1. These data show that both NICO and CCO measurements correlated significantly with BCO, and that the correlation coefficients were remarkably similar (r = 0.80 and 0.81, respectively). Figure 2shows the Bland-Altman plots of NICO and CCO against BCO. Overall, the bias and precision between NICO and BCO was −0.58 ± 0.90 l/min. The bias and precision between CCO and BCO for all measurements was 0.38 ± 1.17 l/min. These data show that NICO underestimated cardiac output, whereas CCO overestimated cardiac output, when compared with BCO measurements.

A comparison of bias and precision of both NICO and CCO against BCO at four intraoperative stages during aortic reconstruction is shown in table 3. The bias between NICO and BCO was small before cross-clamping, but increased following aortic cross-clamp and peaked after reperfusion of the unilateral iliac artery. However, the precision of the data remained within the same range. The bias and precision of CCO against BCO was comparable with that of NICO after anesthetic induction, and remained low during the aortic cross-clamp and early reperfusion periods. At the end of surgery, however, CCO overestimated cardiac output by 0.72 l/min and the deviation was larger than that observed in the NICO system. The relative error for all measurements as compared with BCO was −11.5 ± 19.4% for NICO and 8.2 ± 23.0% for CCO. Among the total number of measurements (n = 112), 20 NICO measurements and 18 CCO readings exceeded the relative error by 30%. The time-dependent change of the relative error is summarized in table 3. Figure 3shows the relationship between the bias of NICO against BCO and four other parameters with the potential to affect the precision of NICO measurements: Paco²-Pet CO²difference, P/F ratio, Vco², and shunt fraction (Qs/Qt). The data indicate that deviations in these parameters did not directly account for the differences between NICO and BCO because no specific trend was noted in the relationship between the bias and the value of these parameters.

This study had four main findings. The first was that the NICO monitor provides a reasonably accurate overall estimate of cardiac output. The second finding, however, was that the NICO system experienced decreases in accuracy following aortic cross-clamping and declamping and that it consistently underestimated cardiac output as compared with BCO. The third finding was that, overall, in terms of bias and precision, NICO and CCO methods are comparable relative to BCO. The fourth finding was that analyses of the Paco²-Petco²difference, P/F ratio, Vco²and shunt fraction showed that none of these parameters could account for the increased bias in NICO estimates of cardiac output following reperfusion.

The NICO monitor uses the differential CO²Fick partial rebreathing method 9,10to measure pulmonary capillary blood flow noninvasively and continuously in mechanically ventilated patients. The NICO algorithm is predicated on the following assumptions: (1) a stable CO²dissociation curve to convert Petco²to Caco²; (2) constant mixed venous CO²content; (3) constant dead space; and (4) stable pulmonary capillary blood flow during the measurement cycle. In addition to calculating cardiac output, it is necessary to estimate the shunt fraction correctly using Nunn's isoshunt plots by entering hemoglobin, Pao², and Paco²values. This device requires minimal operator experience and is not subject to electromagnetic interference during surgery. These characteristics suggest clinical advantages to the use of this device, but only a limited number of studies regarding the accuracy of the NICO monitor have been conducted. 13–17 

In our study, the agreement between NICO and CCO after anesthetic induction (−0.1 ± 0.61 l/min) was comparable with these previous findings, and we suggest that NICO provides an accurate estimate of cardiac output in stable conditions. However, the reports cited above did not specifically address the accuracy of the NICO monitor in the context of intraoperative events. This prompted us to investigate how aortic cross-clamping and declamping affect the accuracy of NICO monitoring and we found that bias increased significantly after aortic cross-clamp release.

Several factors are implicated in the difference between NICO and thermodilution cardiac output monitoring. Tachibana et al.  reported a bias and precision of 0.28 ± 2.04 l/min during large tidal ventilation, but more discrepant results (−1.6 ± 2.24 l/min) during low tidal ventilation. 16They speculated that slower changes in CO²stored in the body and decreased mixed venous CO²content following the decrease in tidal volume result in increased bias. 16It is conceivable that the increased bias immediately following reperfusion may be caused by an abrupt increase in Vco²that cause similar effects to decreasing tidal volume. However, this mechanism may not fully account for our finding because the measurement was performed after Vco²had stabilized. Alternatively, subclinical pulmonary dysfunction might negatively affect estimates of the shunt fraction and, consequently, of cardiac output. Maxwell et al.  used a porcine model of severe chest trauma and hemorrhagic shock in an experimental setting to compare the performance of the NICO monitor with CCO. 15In that study, bias and precision was reported as 0.01 ± 0.69 l/min. In addition, Odenstedt et al.  reported good agreement between NICO and both CCO and BCO (bias ± 1 SD, −0.04 ± 1.72 l/min) in critically ill patients, although they noted significant underestimation of shunt fraction and Vco². 17Although pulmonary oxygenation remains relatively stable during surgery, subclinical pulmonary dysfunction, especially increased dead space caused by ischemia-reperfusion, may contribute to the increase in bias. However, in our study no factor could be singled out to account for the increase in bias in NICO measurements after reperfusion. We speculate that relatively acute changes in Vco²and mixed venous CO²content persisting more than 15 min after reperfusion may be responsible for the underestimation of cardiac output as measured by the NICO monitor after aortic declamping. Alterations in pulmonary dead space and intrapulmonary shunt not reflected in the blood gas analysis may also contribute to this change. In addition, intraoperative fluctuations in such parameters as pH, temperature, hemoglobin concentration Vco², dead space, and shunt can also be potential sources of error. Further study is obviously needed to clarify which factors influence the accuracy of NICO after reperfusion.

Comparisons between NICO and other continuous cardiac output monitoring systems provide clinically relevant information about the usefulness of NICO, which, all other considerations aside, is important because continuous and automatic monitoring is preferable in a clinical setting. Previous investigations report the bias and precision between CCO and BCO as 0.49 ± 1.01, 0.12 ± 0.84, and 0.40 ± 1.26 l/min, respectively, in either the Baxter Vigilance or Abbott Opti-Q system (Abbott, Chicago, IL). 18–20Our data show a comparable degree of agreement between CCO and BCO. Although we are not able to specifically address the reason of increased bias of CCO at the end of surgery, increased cardiac output during this period may account for the increased bias of CCO. 21 

In regard to patient outcome, the routine use of pulmonary artery catheters in aortic surgery is associated with conflicting results. Berlauk et al.  report that preoperative hemodynamic optimization through the use of pulmonary artery catheters improved outcome in atherosclerotic aortic occlusive disease patients. 22On the contrary, Valentine et al.  concluded that the routine use of pulmonary artery catheters is not beneficial and may be associated with a higher rate of intraoperative complications. 6 

These conclusions prompted us to investigate whether monitoring cardiac output using less invasive methods is adequate for intraoperative monitoring during elective aortic surgery. Several studies demonstrate that the monitoring of cardiac output and related parameters during aortic cross-clamping provides clinically useful information. For example, it has been shown that changes in cardiac output and CO²production during aortic cross-clamping are a predictor of hypotension at declamping. 23Whalley et al.  reported that the increase in systemic vascular resistance during aortic cross-clamping positively correlated with a base deficit after reperfusion. 24These findings indicate that cardiac output and systemic vascular resistance, which can be monitored by NICO and the central venous catheter, may make it possible to predict declamping shock and take preventive measures. However, NICO cannot provide information regarding intravascular fluid status and may not be useful in the guidance of intraoperative fluid administration for patients with significant cardiac or pulmonary dysfunction. In this regard, it cannot be recommended as a replacement for pulmonary artery catheter monitoring.

In conclusion, it was determined that NICO provides an accurate estimate of cardiac output after anesthetic induction but underestimates cardiac output after aortic cross-clamping and declamping. Changes in Vco²and subsequent total body CO²content change may be responsible for this bias. CCO shows a relatively smaller degree of bias than NICO but overestimates cardiac output at the end of the surgery. The performance of NICO is comparable with CCO in this clinical setting, and we believe that with cautious evaluation of data NICO may prove to be useful during elective aortic reconstruction surgery.

Shunt fraction = (Cco²− Cao²) / (Cco²− Cvo²).

Cco²= ([Hb]* 1.39) + PAo²* 0.031.

Cao²= 1.39 * Sao²[Hb]+ 0.0031 * Pao².

Cvo²= 1.39 * Svo²*[Hb]+ 0.0031 * Pvo².

PAo²= Fio²(P^B− P^H2O) − Paco²/ R.

Where P^B= 760, P^H2O= 47, R = 0.8.

Cao²= oxygen content of arterial blood; Cco²= oxygen content of pulmonary capillary blood; Cvo²= oxygen content of mixed venous blood; Fio²= fraction of inhaled oxygen concentration; Hb = blood hemoglobin concentration; Paco²= carbon dioxide partial pressure of arterial blood; PAo²= oxygen partial pressure of alveolar gas; Pao²= oxygen partial pressure of arterial blood; P^B= barometric pressure; P^H2O= partial pressure of water vapor; Pvo²= oxygen partial pressure of mixed venous blood; R = respiratory quotient.