This clinical validation study demonstrated, for its primary aim, an excellent agreement between CAPNO-SvO2 and a contemporary blood sample from the pulmonary artery (PAC-SvO2) analysed with a CO-oximeter. The bias was close to zero and the limits of agreement were < 10%, thus falling within the hypothesised clinically acceptable range. The secondary aim of comparing EPBF with thermodilution cardiac output showed a proportional bias of -0.4 l.min− 1 and wider limits of agreement, reflecting the importance of pulmonary shunt flow.
Mixed venous oxygen saturation by the capnodynamic method.
The mixed venous oxygen saturation provides important information about oxygen delivery relative to oxygen needs and aids clinical decisions for critically ill patients. In a large observational cohort of postoperative cardiac patients monitored with a pulmonary artery catheter, low (< 60%) SvO2 on admission to ICU or persisting four hours postoperatively was associated with increased 30-day and one-year mortality [17] as well as early postoperative multiorgan dysfunction [18].
This study used Bland-Altman methodology to compare the performance of the noninvasive CAPNO-SvO2 with invasively obtained PAC-SvO2, adhering to detailed reporting recommendations [19]. The bias for CAPNO-SvO2 compared to PAC-SvO2 was minimal and similar to previous animal studies [5, 6], supporting its accuracy. The limits of agreements were also comparable and within the predefined < 10% limits, indicating that the precision for CAPNO-SvO2 is consistent across both animal and clinical studies. The percentage error was 10–16% throughout the recruitment manoeuvre, demonstrating high precision of the agreement regardless of the level of PEEP. These results suggest that the invasive and intermittent PAC-SvO2 can be interchangeably used with the non-invasive and nearly continuous CAPNO-SvO2.
The PAC-SvO2 used as a reference in this study may have an inherent measurement error related to obtaining and analysing blood samples under real-life clinical conditions. Insufficient sample mixing, air contamination in the syringe and an insufficient sluice volume withdrawn from the pulmonary artery catheter before the sample was taken may have introduced errors in the reference test [20]. Although blood samples were analysed as quickly as practically possible, typically within five minutes, they were not placed in ice or on a rotary shaker which is recommended for optimal results [20]. These limitations to PAC-SvO2 as the clinical reference are important to consider as the observed limits of agreement depend on the inherent precision of both methods. A reference method with smaller inherent measurement error might have narrowed the limits of agreement for CAPNO-SvO2.
The calculation of CAPNO-SvO2 requires, in addition to EPBF, an estimate of oxygen consumption (VO2), which was obtained from the measured VCO2 and a set, defined value for the RQ. Measuring VO2 in respiratory gases is challenging and has limited the use of indirect calorimetry in the perioperative clinical setting [12, 21]. A more practical way to determine RQ in the ICU is to calculate it from the nutritional energy distribution based on the proportion of fat, protein and carbohydrate administered [13]. Previous studies of fasting subjects [22] and postoperative patients in the ICU [12, 13] have shown that RQ is typically about 0.85, whether measured by indirect calorimetry or calculated from nutrient content [13]. An incorrectly set RQ would affect the bias compared to PAC-SvO2. Inserting a higher RQ in the CAPNO-SvO2 formula (Eq. 3) results in higher values. Using a higher RQ of 0.9 (instead of 0.85) in this study would have increased the bias between CAPNO-SvO2 and PAC-SvO2 by 2% points. The negligible bias and acceptable limits of agreement demonstrated in this study with PAC-SvO2 corroborates, however, that the chosen RQ of 0.85 was appropriate for the study cohort of postoperative cardiac patients.
Finally, the agreement between CAPNO-SvO2 and PAC-SvO2 was preserved during changes in PEEP and hence different lung conditions. Since the VCO2 in the CAPNO-SvO2 equation is determined by the blood flow participating in gas exchange (EPBF), its robust performance should be maintained in patients with more severe lung pathology compared to the postoperative patients in this study. Future validation studies of CAPNO-SvO2 are needed to establish this.
Effective pulmonary blood flow by the capnodynamic method
The effective pulmonary blood flow (EPBF) represents the non-shunted fraction of the cardiac output and is therefore a major determinant of gas exchange and systemic oxygen delivery. Maintaining adequate oxygen delivery and timely interventions to mitigate potential reductions in organ oxygenation are important to avoid postoperative organ dysfunction, including acute kidney injury, which remains an important contributor to poor outcomes in cardiac surgery [23, 24]. The nearly continuous monitoring of EPBF by the capnodynamic method is suited to enable prompt recognition and treatment of a low cardiac output state postoperatively.
In this study, the percentage error between EBPF and COTD was slightly higher than that for SvO2 but remained below the 30% limit stated by Critchley and Critchley for interchangeability with COTD [16]. The bias observed was similar to previous reports, even though the presence of non-proportionality was not reported in these earlier studies, and within the usual limits of exclusion (< 15%) in concordance analyses where COTD is used as reference [11, 25–27]. The systematic bias is not surprising since the two variables are not identical physiological entities (EPBF = CO – shunt flow). The observed overall bias of − 0.41 l.min− 1 in this study corresponds to a 10% relative difference, which plausibly represents an average shunt fraction of 10% in the study cohort. In contrast to COTD, the EPBF increased after the PEEP-manoeuvre, likely reflecting a reduction of the shunt with opening of atelectatic lung areas. The divergence between EPBF and CO in the presence of atelectasis has been reported previously [26, 28]. The low percentage error found in this study suggests that EPBF is clinically acceptable for monitoring cardiac output. This is also supported by a recent publication where both EPBF and CAPNO-SvO2 were able to monitor low cardiac output states and the haemodynamic effects of intravascular blood volume changes in a pig model of standardised haemorrhage and subsequent volume resuscitation [10].
Both CAPNO-SvO2 and EPBF are derived from the same measure, VCO2, and the differences in agreement analyses (bias, limits of agreement and percentage error) are likely related to the different clinical reference methods used. The accuracy of blood gas analysis, if the blood sample is properly handled, is superior to thermodilution measurements of CO, which are more variable [11, 29].
Sources of error in the capnodynamic method and study limitations
An obvious limitation of the capnodynamic method is the requirement for controlled mechanical ventilation with satisfactory patient-ventilator synchrony. Data quality deteriorates and might become uninterpretable in the presence of spontaneous breathing efforts. All patients in this study were without spontaneous breathing activity with residual intraoperative neuromuscular blockade present in most. The capnodynamic software features an integrated control function that alerts the user and discards measurements of uncertain data quality during spontaneous breathing activity. Many patients for whom continuous non-invasive assessment of SvO2 is considered important are, however, likely to receive treatment conducive to synchronous, controlled mechanical ventilation.
The calculated EPBF is inversely proportional to the solubility of CO2 in blood, which depends on the haemoglobin concentration used in the content equations [5]. An incorrectly high haemoglobin of 10 g.l− 1 above the true value would result in an underestimation of EPBF by 4%. This would have two counteracting effects on CAPNO-SvO2: while EPBF would be underestimated, the oxygen-carrying capacity of blood would be overestimated (see Methods, Eq. 2), resulting in a true SvO2 of 70% reported as 71.1%. The VCO2 is the primary input variable for the calculation of EPBF. Since VCO2 appears in the numerator and EPBF in the denominator of the CAPNO-SvO2 equation (see Methods, Eq. 3), any error in determining VCO2 is unlikely to significantly affect CAPNO-SvO2, as any inaccuracy in determining VCO2 will proportionally affect both the numerator and denominator. The presence of any postoperative tricuspid regurgitation was not formally assessed and might represent a source of error in the thermodilution measurements used as clinical standard reference.
The study dataset comprised 185 paired observations compared to the 191 estimated in the power calculation. The difference related to patients with missing data after alveolar recruitment. Nonetheless, the power for the actual number of observations (0.88) was very close to the target 0.9 (see Additional File 1, Figure S1). Study enrollment was not consecutive, and patients were sufficiently haemodynamically stable to tolerate the recruitment manoeuvre, meaning that a degree of spectrum bias cannot be excluded.
Advantages of the capnodynamic method and study strengths
The combination of CAPNO-SvO2 and EPBF provides key information on systemic oxygen balance, and both variables are available with minimal change to standard clinical equipment and without requiring any particular user expertise. Furthermore, the capnodynamic variables can be derived within minutes after controlled mechanical ventilation has been initiated, offering nearly continuous signals with built-in signal quality reporting. This study adhered to stringent criteria for the performance and reporting of Bland Altman analyses, with the CAPNO-SvO2 data extracted blinded to the clinical reference standard. Finally, the study was conducted under real-life conditions in the intensive care unit, which supports the external validity of the results.
In conclusion, the results of this study in postoperative cardiac patients admitted to ICU support the use of the capnodynamic algorithm to calculate mixed venous oxygen saturation as interchangeable with blood gas analyses using a pulmonary artery catheter. The effective pulmonary blood flow correlates with thermodilution cardiac output, but the agreement is proportionally influenced by the presence of any pulmonary shunt flow.