Performance of a capnodynamic method estimating cardiac output during respiratory failure - before and after lung recruitment

Hemodynamically unstable patients with severe respiratory failure constitute a medical challenge. In these patients, the interaction between the heart and lungs is frequently affected with increased pulmonary vascular resistance (PVR) and potential right ventricular failure [1]. Lung protective mechanical ventilation commonly improves compliance and oxygenation. However, the accompanying high levels of PEEP can increase the right ventricular afterload with negative effect on the cardiac performance [2, 3]. Hemodynamic monitoring is therefore of great importance to optimize blood perfusion to the injured lungs with least possible strain on the right ventricle.

The capnodynamic method (CO_EPBF) continuously calculates pulmonary blood flow (CO minus the shunted blood fraction) utilizing small variations in partial pressure of alveolar CO₂ (P_ACO₂), automatically controlled by the ventilator [4, 5]. As previously described, the capnodynamic method with inspiratory holds (CO_EPBFinsp) showed acceptable performance in animals with low shunt [4], however when the animals were subjected to lavage-induced lung injury both accuracy and precision were affected [6]. After refinement of the method with the obliged holds inserted in expiration (CO_EPBFexp) instead of inspiration the performance in a porcine model with healthy lungs improved with sustained good trending ability during hemodynamic and ventilatory challenges [5, 7‐9]. In addition, CO_EPBFexp has been shown to provide stable CO monitoring in healthy infants undergoing cleft-lip surgery [8].

Complementary monitoring of pulmonary blood flow during mechanical ventilation in respiratory failure could assist clinicians to perform lung recruitment and adjust PEEP levels, with regards to the right ventricular heart function. [5, 10]. The aim of this study was to evaluate the modified capnodynamic method based on expiratory holds at increased shunt levels during lung injury, before and after lung recruitment and under major hemodynamic changes.

The study was approved by the Uppsala animal research ethical committee (nr. C 47/15) and performed at the Hedenstierna laboratory in Uppsala University, Sweden. The animals used in this study were collected from the same breeding colony (Mångsbo Farm, Uppsala, Sweden). At the farm they had unlimited access to tap water and food on a standardized schedule and kept in a light and temperature-controlled environment. Eight of them were included in another study protocol analysing the effects of ischemia and reperfusion [7]. Results are reported according to the GRRAS guidelines [11].

Data from two animals were excluded in the analysis; one animal was critically unstable after the lung injury and measurements before lung recruitment were not possible. In the second animal a computer failure made CO_EPBF calculation impossible.

Lung injury resulted in respiratory failure including decreased dynamic compliance and ratio of partial pressure of oxygen in arterial blood to inspired fraction of oxygen (P_aO₂/F_iO₂), as well as increased shunt, physiological dead space and partial pressure of CO₂ in mixed venous blood (P_vCO₂). The individual recruitment manoeuvre with PEEP adjustment to 15 (3) cmH₂O (range 11–17 cmH₂O), normalized shunt and compliance to large extent, although physiological dead space and P_vCO₂ were elevated compared to baseline (see Table 1 for changes in respiratory and hemodynamic parameters during the protocol).

The calculated inherent precision of the CO_EPBF and CO_TS during initial baseline conditions were 9% and 6%, respectively. Mean CO during the lung injury protocol measured with CO_EPBFexp and CO_TS was 4.2 and 4.0 L/min and changed in average 30–35% (± 4–10%) during caval balloon inflation and dobutamine infusion (see Fig. 1 for event line). Data for CO measurements was normally distributed. No proportional bias was detected.

Bias (LoA) and PE for the CO_EPBF compared to CO_TS changed from baseline 0.5 (− 0.5 to 1.5) L/min and 30% to − 0.6 (− 2.3 to 1.1) L/min and 39% at LI_P5 and finally 1.1 (− 0.3 to 2.5) L/min and 38% after recruitment manoeuvre and PEEP adjustment (see Fig. 2; Table 2).

Concordance during LI_P5 and LI_PAdj, was 87 and 100% via the four-quadrant plot and 93 and 100% in the polar plot. The mean (95% CI) polar angle during LI_P5 and LI_Padj was − 14.8° (− 40.0° to − 10.5°) and − 2.3° (−22.9° to 18.3°), respectively (see Figs. 3, 4).

We have evaluated the performance of the modified capnodynamic method based on expiratory holds in an experimental lung injury model resulting in high shunt fraction at PEEP 5 cmH₂O and after lung recruitment with PEEP adjustment. In addition, both conditions were evaluated during major CO changes. CO_EPBFexp, without any shunt correction, underestimated CO at high shunt fraction by 14% and overestimated CO by 30% after recruitment and PEEP adjustment. The ability to track changes was only marginally affected at high shunt and improved to 100% after lung recruitment as assessed by the four-quadrant and polar plot analyses.

Respiratory failure is a challenging situation for clinicians working in the operating theatre and intensive care unit. In severe cases, elevated shunt, hypoxemia and hypercapnia can lead to increased pulmonary vascular resistance and right ventricular failure, if not treated adequately. Mechanical ventilation is the mainstay of supportive treatment conveying oxygenation and ventilation at the lowest driving pressure possible. Mechanical ventilation providing continuous CO_EPBF monitoring has the potential to detect almost instantaneous hemodynamic changes, which can help the clinician to optimize oxygen delivery in real time via PEEP titration and adjustment of fluid and vasopressor therapy.

The capnodynamic method indirectly calculates the pulmonary blood flow based on CO₂ elimination kinetics. CO_EPBF is equivalent to the non-shunted fraction of the CO and therefore potentially affected by disease states that increase the shunt fraction. The capnodynamic method based on inspiratory holds performed poorly in a porcine lung-lavage model. Precision was low during high shunt fraction and decreased further after PEEP elevation to 12 cmH₂O, compared to the CO_TS [6]. Interestingly, even before the lavage, a paradoxical rise in CO_EPBFinsp was observed at PEEP 12 cmH₂O [6]. The authors concluded that the observed increase in intrathoracic and airway pressure parallel to the inspiratory pause phases may have caused fluctuation in the pulmonary blood flow per se, possibly disturbing the carbon dioxide signal leading to errors in the obtained CO_EPBFinsp value [6]. The current improved breathing pattern, combining six normal breaths followed by three with added short expiratory hold lowers the average airway pressure during a set of nine breaths [5]. This probably decreases the variation in the pulmonary blood flow, providing more stable conditions for accurate calculations as observed when the CO_EPBFexp was challenged in a porcine model during various ventilatory and hemodynamic changes in healthy lungs [5].

In the current study CO_EPBFexp overestimated CO at higher PEEP levels without markedly affecting precision or trending ability. The expected increase in lung volume, as new regions were re-opened with the RM and kept open with the adjusted PEEP, was captured by the CO₂-based estimation (ELV), an entity included in the capnodynamic equation (see Table 1). However, this entity has not been evaluated with the revised breathing pattern or during lung injury, per se. The reason for the overestimation at higher PEEP levels in the animal model is not entirely clear. At constant metabolism and ventilatory settings, any change in the elimination of CO₂ can be explained by a combination of changes in (1) the effectiveness of the pulmonary blood flow, (2) the area for CO₂ exchange, and the global ventilation–perfusion relationship of the lungs (V/Q ratio). These factors can be different between healthy and injured lungs [23]. The capnodynamic method is based on a homogenous lung model with regards to V/Q ratio. In the injured lung there will be a distribution of varying V/Q conditions over the lung, even after recruitment, especially over distended regions. How these affect the result of the prototype method is not fully understood, but could contribute to the observed over estimation at high PEEP in this study. Previous animal studies and our overall data show conflicting results in terms of higher PEEP (increased lung volume) and accuracy of the CO_EPBF. Interestingly, in a recent clinical study (manuscript) where CO_EPBF was compared to transpulmonary thermodilution during open abdominal surgery, no over estimation was observed even at high PEEP levels (up to 20 cmH₂O) [24].

The current experimental protocol using the improved capnodynamic method departs from our previous study [6] in three ways. Firstly, the animals had experienced caudal ischemia and reperfusion approximately 2 h before [7], possibly adding extra stress and inflammation; secondly, VILI was added to the conventional lavage to induce more severe respiratory failure and thirdly, an individualized recruitment manoeuvre with PEEP adjustment based on compliance, instead of arbitrarily raising it to 12 cmH₂O. Despite the more extreme insult to the animals, percentage error for CO_EPBFexp was roughly unchanged (avg. 38 and 39%) during LI opposite to the high PE (avg. 70 and 75%) observed in the previous study with CO_EPBFinsp. A finding consistent with the overall performance of CO_EPBFexp in healthy lungs.

The accuracy of CO_EPBF is dependent on the ventilation-perfusion status of the lung. When the shunt fraction is increased, CO_EPBF underestimates CO. This can be managed in two ways; either by incorporating shunt correction methods such as the iso-shunt diagram used by the NICO monitor [25] or by combining measurement of oxygen uptake and pulmonary blood flow (via the Fick equation) as applied in the capnotracking method developed by Peyton et al. [26], or alternatively by keeping the lung open with lung recruitment and PEEP adjustment. Recruitment manoeuvres are common in clinical practice both in the operation theatre and ICU and will most likely reduce any significant shunt fraction and improve the overall performance of the capnodynamic method in the commonly used PEEP range. Therefore, CO_EPBF could possibly be used as an independent physiological variable to adjust PEEP for best oxygen delivery as suggested by Gedeon et al., in a study of a small group of patients with acute respiratory failure [10]. For the clinician managing patients with respiratory failure in the operation theatre or ICU, perhaps the most important function of a hemodynamic monitor is to continuously and reliably detect relevant CO changes and response to treatment, as repeatedly shown by the capnodynamic method under experimental conditions.

Despite the severe lung injury, high shunt fraction, recruitment manoeuvre and PEEP adjustment the CO_EPBFexp had a PE of 38 and 39%, when compared to the highly accurate Transonic flow probe. Traditionally, a PE of 30% has been used as a priori to determine interchangeability between the studied and reference method [20]. However, it should be kept in mind that this cut off is primarily based on the inherent precision of the PAC calculating CO via thermodilution during stable hemodynamic situations and a simulation model [27, 28]. Based on the performance of the PAC, the 30% cut-off has recently been challenged by Peyton and Wong, where a PE of 45% was suggested when comparing new non-invasive CO methods in clinical situations when clinical benefit is anticipated [29].

The capnodynamic method calculates CO by changing the time between three out of nine breaths during controlled mechanical ventilation. Any triggered breath will affect the accuracy of CO_EPBF for the next nine breaths. However, the method includes an error function estimating the internal validation of the calculations (not used in this study). This function could theoretically be programmed to filter out CO_EPBF values with an elevated internal error, such as triggered breaths and therefore support the clinician with only stable calculations of pulmonary blood flow.

This study has several limitations. It is a small animal study where the induced respiratory failure is artificial, although the ischemia, reperfusion and VILI might have added to the clinical application. Each animal responded differently to lung injury and lung recruitment resulting in inter-individual differences in treatment and physiological status. As CO_EPBF calculations are based on human data and CO_EPBF and the reference method do not measure the same physiological variable, some difference is to be expected.

In this study, the performance of the revised capnodynamic method was evaluated during lung injury, with high shunt and after lung recruitment with PEEP adjustment. CO_EPBFexp showed good trending ability, especially when CO decreased. Accuracy was affected during high shunt and elevated PEEP although precision might be considered clinically acceptable. Clinical studies validating performance during high risk surgery and intensive care are underway.