A modified breathing pattern improves the ...

J Clin Monit Comput DOI 10.1007/s10877-016-9891-z

ORIGINAL RESEARCH

A modified breathing pattern improves the performance of a continuous capnodynamic method for estimation of effective pulmonary blood flow Caroline Ha¨llsjo¨ Sander1,2 • Thorir Sigmundsson1,2 • Magnus Hallba¨ck3 • Fernando Suarez Sipmann4,5 • Mats Wallin2,3 • Anders Oldner1,2 • Ha˚kan Bjo¨rne1,2

Received: 22 March 2016 / Accepted: 25 May 2016 Ó Springer Science+Business Media Dordrecht 2016

Abstract In a previous study a new capnodynamic method for estimation of effective pulmonary blood flow (COEPBF) presented a good trending ability but a poor agreement with a reference cardiac output (CO) measurement at high levels of PEEP. In this study we aimed at evaluating the agreement and trending ability of a modified COEPBF algorithm that uses expiratory instead of inspiratory holds during CO and ventilatory manipulations. COEPBF was evaluated in a porcine model at different PEEP levels, tidal volumes and CO manipulations (N = 8). An ultrasonic flow probe placed around the pulmonary trunk was used for CO measurement. We tested the COEPBF algorithm using a modified breathing pattern that introduces cyclic end-expiratory time pauses. The subsequent changes in mean alveolar fraction of carbon dioxide were integrated into a capnodynamic equation and effective pulmonary blood flow, i.e. non-shunted CO, was calculated continuously breath by breath. The overall agreement between COEPBF and the reference method during all interventions was good with bias (limits of agreement) 0.05 (-1.1 to 1.2) L/min

& Caroline Ha¨llsjo¨ Sander [email protected] 1

Department of Anaesthesiology, Surgical Services and Intensive Care Medicine, Karolinska University Hospital, 171 76 Solna, Stockholm, Sweden

2

Department of Physiology and Pharmacology, Karolinska Institutet, Stockholm, Sweden

3

Maquet Critical Care AB, Solna, Sweden

4

Hedenstierna’s Laboratory, Department of Surgical Sciences, Section of Anaesthesiology and Critical Care, Uppsala University, Uppsala, Sweden

5

CIBER de enfermedades respiratorias (CIBERES), Instituto Carlos III, Madrid, Spain

and percentage error of 36 %. The overall trending ability as assessed by the four-quadrant and the polar plot methodology was high with a concordance rate of 93 and 94 % respectively. The mean polar angle was 0.4 (95 % CI -3.7 to 4.5)°. A ventilatory pattern recurrently introducing end-expiratory pauses maintains a good agreement between COEPBF and the reference CO method while preserving its trending ability during CO and ventilatory alterations. Keywords Monitoring  Carbon dioxide  Cardiac output  Perioperative

1 Introduction Goal directed protocols for haemodynamic optimisation may improve outcome in high-risk surgical patients [1–3]. In this setting a flow based parameter like cardiac output (CO) could be used to guide the haemodynamic interventions [4–7]. In mechanically ventilated subjects alterations of mean alveolar content of carbon dioxide can be easily induced by cyclic variations in the ventilatory pattern and used for calculation of effective pulmonary blood flow (EPBF), i.e. the non-shunted fraction of CO. An algorithm based on a capnodynamic equation can be used for this continuous calculation of EPBF [8, 9]. Our group has previously evaluated a novel mathematical approach to a capnodynamic equation during significant CO alterations in a porcine model [8, 9]. The ventilatory pattern used in those experiments followed the pattern of ventilatory change used by Gedeon et al. [10] in dogs. The capnodynamic method (COEPBF) showed low bias and good trending abilities during the haemodynamic

123

J Clin Monit Comput

interventions and displayed prerequisites for a short response time. However, in a subsequent study where ventilatory settings were altered, COEPBF had a paradoxical rise in contrast to CO when positive end-expiratory pressure (PEEP) was increased [9]. Data suggested that this paradoxical behaviour of COEPBF could be related to the chosen breathing pattern that consisted of a recurrent pattern of ten breaths: five normal breaths and five breaths with an inspiratory pause prolonging the inspiratory time [8] (Fig. 1). The resulting phases of increased intrathoracic and airway pressure during the inspiratory pause phases could affect EPBF per se as well as potentially cause an overestimation of COEPBF at higher PEEP levels. The capnodynamic equation assumes that pulmonary perfusion remains constant for the sequences of breathing cycles that are analysed. Any variation of the pulmonary blood flow during the breaths in the breathing pattern cycle may introduce disturbances in the measured carbon dioxide signal that will lead to errors in the obtained COEPBF value. This effect may be enhanced at high PEEP levels, as endinspiratory pressures will be increased further. It is therefore likely that end-expiratory pauses exert a smaller influence on pulmonary blood flow since the impact on intra thoracic pressure and CO is less pronounced compared to inspiratory pauses [10, 11]. A pattern with endexpiratory pauses was piloted by Peyton et al. [12] comparing a differential carbon dioxide method built on the principle of Fick for calculation of pulmonary blood flow measurement to bolus thermodilution in cardiac surgery patients. However, in that study the effect of differencing PEEP was not investigated [12]. Furthermore, our previous experiments also suggested that the number of breathing cycles included in the ventilatory pattern could be reduced in order to shorten the response time. The aim of this study was to test the hypothesis that COEPBF based on a modified breathing pattern implementing expiratory holds would improve its performance. For this purpose we assessed the agreement and trending ability of COEPBF with this modified breathing pattern

during significant alterations in CO and ventilator settings in a porcine model. As a secondary aim we compared the performance of COEPBF with two clinically used methods, often considered to be reference methods for CO monitoring in humans, the pulmonary artery catheter (COPAC) and the transpulmonary thermodilution (COTPT) techniques [13, 14].

2 Materials and methods 2.1 Anaesthesia and surgical preparation The study was approved by the Animal Research Ethical Committee and performed at the Hedenstierna Laboratory in Uppsala University, Sweden. The anaesthetic and surgical procedures have been described in detail previously [8]. In brief, eight pigs with a mean weight of 34 kg (32–36 kg) were anaesthetised, catheterised and mechanically ventilated in a volume controlled mode with a tidal volume (TV) of 8 mL/kg, FiO2 0.40 and PEEP 5 cm H2O (Servo-i, Maquet, Solna, Sweden). A 13.5 Fr catheter was inserted into the femoral artery for controlled bleeding and a pulmonary artery catheter (Edwards, Irving, CA, USA) was inserted via v. jug interna for blood sampling, pressure and CO recordings. An 8 Fr Fogarty catheter was inserted into the inferior cava vein for controlled preload reduction. Via ultrasound guidance a 4 Fr 8 cm artery catheter (PiCCO Pulsion, Munich, Germany) was placed in the contralateral femoral artery and connected to the PICCO2 monitor for transpulmonary thermodilution. An ultrasonic flow probe was placed around the pulmonary trunk through a left-sided mini-thoracotomy for reference CO measurements (COTS; T 401; Transonic system Inc., Ithaca NY, USA). The size of the probe was adjusted to the size of the pulmonary trunk (18 or 20 mm). After the chest was closed the animal was repositioned into the supine position and a recruitment manoeuvre was performed to re-expand the lungs. The COTS was considered to be the primary comparator since it is an invasive gold standard method for CO measurements. Pressure readings and haemodynamic signals were recorded and sampled (at 100 Hz) into a data acquisition system (Acknowledge, version 3.2.7, Bio Pac Systems, Santa Barbara, CA, US) to allow post processing of data. 2.2 The ventilatory pattern

Fig. 1 The continuously on-going ventilatory pattern consisting of ten breathing cycles; five normal cycles followed by five respiratory cycles with an inspiratory hold

123

The tested modified ventilatory pattern consisted of nine breathing cycles including six normal cycles followed by three cycles with an added end-expiratory pause (Fig. 2).

J Clin Monit Comput Fig. 2 The upper panel shows the continuous on-going ventilatory pattern consisting of nine breathing cycles, six normal cycles are followed by three respiratory cycles with an end-expiratory hold displayed by the airway pressure (cmH2O) The lower panel shows the corresponding variations in the end-tidal carbon dioxide levels (mmHg)

Since the ventilatory pattern is constantly on-going adequate minute ventilation was maintained by increasing the rate of the breaths without imposed pauses to compensate for the lower rate of the breathing cycles with imposed pauses. This particular breathing pattern, consisting of three respiratory cycles with an imposed expiratory pause followed by six normal respiratory cycles creates cyclic variations in the mean alveolar carbon dioxide fraction of approximately 4–8 mmHg. 2.3 Calculation of effective pulmonary blood flow The capnodynamic equation (see below) describes a mole balance of carbon dioxide between the transport of carbon dioxide to and the elimination from the lung and the rate of change of the carbon dioxide content in the lung induced by the alternating ventilatory pattern. The left-hand side of the equation expresses the difference in mean alveolar carbon dioxide content in the lung between two breaths. Effective lung volume (ELV) represents the gas volume in the lung containing carbon dioxide. The first term on the right-hand side of the equation, EPBF, represents the circulatory supply of carbon dioxide to the lung. Dtn expresses the duration of each breathing cycle, which varies due to the imposed expiratory holds. EPBF and CvCO2 are assumed to be constant during a measurement cycle and are, together with ELV, determined from the mathematical model. The carbon dioxide content of pulmonary end-capillary blood, CcCO2, is calculated from the measured alveolar partial pressure, PACO2, which is assumed to correlate closely to the partial pressure of carbon dioxide of the pulmonary end-capillary blood. To convert the partial pressure of carbon dioxide to concentration, the

dissociation curve for carbon dioxide in blood described by Capek et al. [11] was used. The last term VTCOn2 is measured by the ventilator and represents the quantity of carbon dioxide exhaled by the nth tidal volume. FACO2 is estimated from the midpoint value of the alveolar plateau (phase III) of the volumetric capnogram [15]. During the imposed expiratory pause all valves of the ventilator are closed and active expiration is ended immediately after the pause is activated. Thus all breaths had the same active expiratory time and consequently the duration of the expiratory portion of the capnogram did not vary, only the carbon dioxide level varied between breaths. The induced variations in expired carbon dioxide were created by automatic alteration of effective ventilation through periodically imposed end-expiratory pauses generated by a Servo-i ventilator (Maquet Critical Care, Solna, Sweden) with slightly modified software. Carbon dioxide and gas flow were measured by the ordinary mainstream carbon dioxide sensor and flow sensors of the Servo-i ventilator, which were connected to a lap top computer where all the mathematical analysis were carried out with software written in MatlabTM (Mathworks). ELV  ðFA COn2  FA CO2n1 Þ
¼ EPBF  Dtn  Cv CO2  Cv COn2  VTCOn2 ELV, Effective lung volume (L); EPBF, effective pulmonary blood flow (L/min); n, current breath; n - 1, previous breath; FACO2, mean alveolar carbon dioxide fraction; CvCO2, mixed venous carbon dioxide concentration (Lgas/Lblood); CcCOn2, pulmonary end-capillary carbon dioxide concentration (Lgas/Lblood); VTCOn2, volume (L) of carbon dioxide eliminated by the current, nth, breath; Dtn, current breath cycle time (min).

123

J Clin Monit Comput

Each breathing cycle results in a new equation. All quantities except ELV, EPBF and CvCO2 are directly obtained from breath-by-breath measurement. An overdetermined set of nine equations solved through a leastsquare-error optimisation of the fit between the left-handside and the right-hand-side of the equation calculates values for ELV, EPBF and CvCO2. A prerequisite for the mathematical model is the assumption that ELV, EPBF and CvCO2 remain constant during the complete measurement cycle consisting of a set of nine breaths (Fig. 2). The method is continuous because the measurement cycle is constantly updated with the most recent equation and expelling the last. The measured data using the capnodynamic method are compared on a breath-by-breath basis to an ideal one-lung compartment model. A correction for the shunt fraction is not included.

At the end of the protocol the animals were sacrificed by an injection of potassium chloride. A graphical explanation of the experiment is displayed in Fig. 3. 2.5 Statistics Data were checked for normal distribution by the D’Agostino and Pearson omnibus K2 test. Data are presented as mean (standard deviation, SD) or median (range) as appropriate. All statistical calculations except for the polar plots were done in Graph Pad Prism (version 6.0 for Windows, Graph Pad Software, San Diego, CA, US). For calculation of polar plots an excel sheet for conversion of Cartesian data to polar coordinates was used kindly provided by Critchley et al. [16] and displayed as graphs in Medcalc.

2.4 Experimental protocol 2.5.1 Precision After 30 min of stabilisation, baseline (BL) measurements of COTS, COEPBF, COPAC and COTPT were obtained at PEEP 5 cm H2O and a TV of 8 mL/kg. For the thermodilution-based monitors, COPAC and COTPT, an average of three injections of 10 mL cold saline was used to calculate CO at all interventions and BL. Measurements were then obtained at the following ventilatory interventions; PEEP 0, 12 cm H2O and BL followed by an increase in TV by 50 % to 12 mL/kg. CO manipulations were performed in the following order: Low CO: preload was reduced by the cava balloon inflation aiming at a decrease of 50 % of COTS at PEEP 5 and 12 cm H2O. High CO: under BL conditions (PEEP 5 cm H2O) an infusion of dobutamine was titrated to increase COTS to 150 %. Haemorrhage: controlled bleeding aiming at a mean arterial pressure (MAP) of 35 mmHg. After confirming stabilisation a BL measurement was performed before each CO manipulation. All measurements were accomplished at steady state conditions as judged by COTS. The recordings of COEPBF were obtained simultaneously with the COTS readings and were based on an average of approximately 10 s. Because the injections caused instability of the carbon dioxide signal in the measured expired gas these measurements were always performed after the COTS and COEPBF readings. Recording of data were randomly distributed relative the expiratory holds of the nine breathes manoeuvre cycle. Between all interventions the haemodynamic condition was allowed to stabilise.

123

Precision [defined as twice the coefficient of variation, CV (SD/mean CO of the reference method)] of COEPBF was calculated across all animals. Five values per animal, recorded with a frequency of one measurement per minute during BL, were included in the calculation [17]. The previously reported precisions for the different methods was: ±10 % for COTS [18], between ±8 and ±24 % for COPAC depending on ventilation/temperature combinations

Fig. 3 Haemodynamic data for all protocolled interventions. The haemodynamic events are displayed on the x-axis. The left y-axis displays cardiac output (L/min) as measured by the ultrasonic reference method (black line, COTS) and the capnodynamic method (blue line, COEPBF). Each haemodynamic challenge was followed by a baseline stabilisation (BL) period. Data are presented as mean (SD), (N = 8)

J Clin Monit Comput

and different haemodynamic conditions and positioning of the catheter per se [13, 19] and between ±6 and ±12 % for COTPT when an average of three boluses was used [20].

High CO state including inotropic stimulation with infusion of dobutamine (Table 1). 3.1 COEPBF

2.5.2 Absolute values 3.1.1 Overall agreement and trending ability For calculation of agreement the Bland–Altman methodology corrected for repeated measurements was used [21, 22]. Percentage error estimated the accuracy and was calculated as 2 SD/mean CO of the reference method for CO [23]. The mean difference between methods, i.e. bias, was used to evaluate trueness [24]. For calculation of precision, limits of agreement (LoA) was used [23]. A priori, we considered all three methods when compared to COTS interchangeable if percentage error was less than 30 % [23].

Bland–Altman analysis of all 141 paired CO measurements showed a bias (LoA) of 0.05 (-1.1 to 1.2) L/min and percentage error 36 % (Fig. 4). The overall trending ability was evaluated by including 61 paired delta values in fourquadrant and polar plot analyses. Concordance rate was 93 and 94 % respectively (Figs. 5, 6). The mean polar angle calculated by the polar plot methodology was 0.4 (95 % CI -3.7 to 4.5)° (Fig. 6). 3.1.2 Response time

2.5.3 Trending ability Two different statistical methods were used. The fourquadrant plot methodology which assesses the agreement of the direction of change between the test and reference method and the polar plot methodology also comparing the magnitude of change [16]. Agreement of the paired delta CO values were expressed as concordance rate for both methods i.e. for the four-quadrant plot the number of data points in the two quadrants of agreement divided by the total number of data points [16]. Concordance rate for the polar plot was calculated as the number of data points within the radial limits of agreement of ±30° divided by the total number of data points [16]. Because of the high precision of the reference method, \20 %, an exclusion zone of 10 % was used [25]. A priori we considered a concordance rate of [92 and [90 % calculated by the four-quadrant plot and the polar plot methodology respectively as good [26]. An angular bias smaller than ±5° indicated sufficient calibration between the test and the reference method [16, 26].

With a ventilatory pattern including nine cycles in the ongoing analysis the response time was approximately 15-20 s (Fig. 7). It takes one maneuver cycle (i.e. nine breaths) until an abrupt change in CO has been fully captured by COEPBF (which acts as a moving average of nine breaths). Into how many seconds this finally translates depends on the average respiratory rate, which in the present case was RR mean = 28 breaths/min (9 breaths = 19 s. 3.1.3 The effects on agreement by increased tidal volume The agreement between methods did not change significantly in response to an increase in TV from 8 to 12 mL/ kg. Bias (LoA) 0.1 (-1.1 to 1.4) L/min, percentage error 39 % and 0.2 (-0.9 to 1.4) L/min, percentage error 37 %, respectively. A four-quadrant plot showed that almost all of the data points were in the exclusion zone and the rest was distributed in all four quadrants indicating no consistent direction of change in response to changes in TV. 3.1.4 Trending ability in response to increased PEEP

3 Results All animals survived the protocol interventions that resulted in marked haemodynamic changes. Cardiac output varied between 1.3 and 8.0 L/min (mean 3.2 L/min). The calculated precision of COEPBF during steady state conditions at BL was ±14 %. Cardiac output data for agreement and trending capability for each CO method were divided in five different groups; (1) Overall; all CO data from the experiment. (2) and (3) CO values from BL and interventions at PEEP 5 and 12 respectively. (4) Low CO states including data from cava occlusion at both PEEP levels and haemorrhage. (5)

When the PEEP level was increased from 0 to 12 and 5 to 12 generated 15 paired delta values. Both four-quadrant and polar plot methodologies showed a concordance rate of 93 %. The mean polar angle was -4.1 (95 % CI -14.5 to 6.3)°. 3.1.5 Effects of different PEEP levels on agreement and trending ability The agreement at BL and during interventions at PEEP 5 showed a bias (LoA) of 0.1 (-0.9 to 1.2) L/min and remained largely unchanged at PEEP 12 cm H2O.

123

J Clin Monit Comput Table 1 The results from Bland–Altman analyses for the capnodynamic method (COEPBF), the pulmonary artery catheter (COPAC) and the transpulmonary thermodilution (COTPT), with the ultrasonic flow probe (COTS) as the reference method for cardiac output (L/min)

Overall

PEEP 5

PEEP 12

High CO

Low CO

COEPBF bias (LoA) PE

COPAC Bias (LoA) PE

COTPT bias (LoA) PE

0.05 (-1.1 to 1.2)

-0.2 (-1.6 to 1.2)

0.5 (-0.4 to 1.5)

36 %

45 %

29 %

(n = 141)

(n = 97)

(n = 109)

0.1 (-0.9 to 1.2)

-0.2 (-1.7 to 1.2)

0.5 (0.6 to 1.6)

32 %

38 %

29 %

(n = 70)

(n = 48)

(n = 54)

0.2 (-1.0 to 1.4)

0.1 (-1.3 to 0.1)

-0.5 (-0.3 to 1.2)

41 %

46 %

30 %

(n = 32)

(n = 21)

(n = 24)

-0.3 (-2.0 to 1.4)

0.09 (-1.4 to 1.6)

-0.05 (-2.1 to 2.0)

27 %

24 %

31 %

(n = 16)

(n = 7)

(n = 14)

-0.09 (-0.8 to 0.6)

-0.08 (-1.0 to 0.8)

0.43 (-0.2 to 1.1)

33 %

45 %

28 %

(n = 46)

(n = 21)

(n = 24)

The paired measurements are divided into overall, including all measurements, the two different PEEP steps 5 and 12 cm H2O, and high and low CO states. The bias, limits of agreement (LoA) and percentage error (PE) for each method are displayed in the corresponding column (N = 8)

Fig. 4 A Bland–Altman analysis of absolute values including 141 paired CO measurements from the reference method for cardiac output (COTS) and the capnodynamic method (COEPBF). Dotted lines indicate bias (mean difference), 95 % lower and upper limits of agreement

However, percentage error increased from 32 to 41 % (Table 1). Trending capability as assessed by concordance rate was 100 % for both PEEP levels regardless of plot technology. The mean polar angle was 1.4 (95 % CI -2.3 to 5.1)° and 1.8 (95 % CI -6.5 to 10.1)°, respectively. 3.1.6 High and low cardiac output states When separating data into high and low CO, bias and percentage error was low in both groups but LoA were wider in high CO states (Table 1). Trending ability before and after haemorrhage (n = 6) showed a concordance rate of 100 % for both methods assessing trending ability although the mean polar angle was wide 11.5 (95 % CI -0.4 to 23.4)°.

123

Fig. 5 A four-quadrant plot showing the 61 paired delta values displayed by the capnodynamic method COEPBF and the reference method for cardiac output COTS from all interventions including the PEEP alteration, tidal volume change and haemodynamic interventions. Because of the high precision of COEPBF an exclusion zone of 10 % was used

3.2 COPAC and COTPT 3.2.1 Overall agreement and trending ability For COPAC the overall bias (LoA) was -0.2 (-1.6 to 1.2) L/min and percentage error 45 % (Table 1). Concordance

J Clin Monit Comput

95 % respectively. The mean polar angle calculated by the polar plot methodology was 1.8 (95 % CI -1.4 to 5.0)°.

4 Discussion

Fig. 6 A polar plot displaying the 61 paired delta values displayed by the capnodynamic method COEPBF (L/min) and the reference method for cardiac output COTS (L/min) from all interventions including the PEEP and tidal volume change and haemodynamic interventions. The blue vertical line indicates the mean angular bias and the surrounding blue lines the confidence intervals of the angle. An exclusion zone of 10 % was used

rate for 55 paired delta values as assessed by the fourquadrant and polar plot methodology was 96 and 92 % respectively. The mean polar angle was -8.1 (95 % CI 4.7 to -11.4)°. For COTDT the overall bias (LoA) and percentage error was 0.5 (-0.4 to 1.5) L/min percentage error 29 %. The concordance rate for 58 paired delta values analysed by the four-quadrant and polar plot methodology was 97 and

In the current study we have evaluated the continuous capnodynamic methodology, COEPBF, using a modified ventilatory pattern based on cyclic variations in end-expiratory pauses, during significant changes in CO and ventilatory settings. In parallel we have also evaluated two thermo-dilution based methods for comparison. In this experimental setting the overall agreement and precision improved compared to our previous results when the ventilatory pattern was based on cyclic reoccurring inspiratory pauses [8]. Trending ability was preserved and no paradoxical rise in COEPBF was detected in response to an increased PEEP level. In comparison with COPAC, COEPBF showed a lower percentage error in all interventions except during high CO states. COTPT displayed the lowest percentage error in all groups except during high CO. A ventilatory pattern including inspiratory pauses is more likely to affect the pulmonary blood flow especially in hypovolemic states and higher PEEP levels as it introduced extended periods at maximal intrathoracic and airway pressures. Since the capnodynamic equation assumes that the pulmonary perfusion is constant during a measurement cycle, a pattern altering the pulmonary blood flow could be assumed to impair the agreement of the method. By the studied modification of the ventilatory pattern the agreement of COEPBF improved when an increased PEEP level of 12 cm H20 was used with a bias (LoA) and percentage error of 0.2 (-1.0 to 1.4) L/min and 41 %

Fig. 7 A typical response, from one animal, to inflation and deflation of the balloon in the inferior cava vein is displayed. The y-axis displays the reference method for cardiac output, the ultrasonic flow probe (COTS) and the capnodynamic method (COEPBF). Time is plotted on the x-axis

123

J Clin Monit Comput

compared to the results from our previous study 1.4 (-1.3 to 4.2) L/min and 90 % [9]. An additional advantage of the modification could be a less marked influence of the inspiratory pauses with subsequent depression of the circulation per se. This effect could be more pronounced during hypovolemic states, an unfavourable property of a CO monitor in the clinical setting. Our previous data also suggested that ten breathing cycles included in a measurement cycle could be reduced. This offered a possibility to modify the number of breaths included in the overdetermined set of equations in order to reduce the response time. Challenged by these findings the ventilatory pattern was redesigned with cyclic occurring end-expiratory pauses including nine breathing cycles that sufficed to induce the required alterations of mean alveolar carbon dioxide content. By reducing the number of cycles included in the ongoing analysis the response was shortened. In Fig. 7, this is exemplified by a typical example from one animal displaying COEPBF in comparison with CO during preload reduction. Available non-invasive continuous CO monitors seem to be less reliable when the circulation is unstable [27–30]. Haemodynamic instability with rapid changes in CO is a common problem in the perioperative period and requires a CO monitor that is fast and continuous with good abilities to detect trends to guide goal-directed administration of fluids, vasopressors and inotropic stimulation. We chose to induce large CO alterations in this study, as other noninvasive CO methods have been found unreliable during unstable haemodynamic situations [31]. It is highly important for a CO monitor to be reliable when the circulation unstable and the need for monitoring and guidance is most important. A meticulous evaluation of a new method requires severe haemodynamic interventions that cannot be easily evaluated in a human study. Therefore, in the current study, COEPBF was challenged with severe haemodynamic interventions. The fact that the percentage error increases during circulatory instability [25] must be taken into consideration when evaluating the agreement between methods. A definition of percentage error \30 % defining two methods as interchangeable could be considered highly conservative during these circumstances [23]. Some authors have, based on the performance of availably non-invasive CO methods, suggested a percentage error of \45 % as acceptable [32]. A limitation of the current study is that COEPBF, that does not include shunt flow, is compared to a reference method for CO. In the current experimental setting with healthy lungs shunt flow was low indicating COEPBF could be used as a proxy for CO. However, a small difference in

123

trueness is expected and could partially be explained by the shunt fraction, especially during high CO states. Since COEPBF is based on the analysis of changes in mean alveolar content of carbon dioxide, lung pathology with subsequent disturbances in gas exchange is most likely to affect the performance. A previous study with the original ventilatory pattern including reoccurring inspiratory holds during haemodynamic changes in a porcine model of lung lavage confirmed this [9]. Further evaluation of the modified COEPBF in a model of lung injury is warranted. In the high-risk perioperative setting with intubated and mechanically ventilated patients, usually with limited pulmonary pathology, a capnodynamic method requires no further invasiveness. When fully evaluated a future area for use of COEPBF could be in the non-cardiac perioperative setting. The number of patients going through high-risk surgery is increasing since older and more fragile patients are scheduled for extensive surgery. In this group, CO monitoring and haemodynamic optimisation might reduce postoperative complications [3–6]. Rapidly occurring haemodynamic changes like haemorrhage and subsequent volume loading infusion therapy is quite common and therefore the interest for the ability to detect CO trends has grown and been identified as an important property of a CO monitor. In conclusion, the performance of the capnodynamic method improved with a lower overall percentage error, 36 %, and bias of 0.05 L/min relative to an invasive gold standard measurement device, COTS, when the ventilatory pattern included reoccurring end-expiratory pauses compared to the corresponding results with percentage error 47 % and a bias of 0.2 L/min when inspiratory pauses were used. The trending ability was preserved [8]. Acknowledgments This study is a collaboration between Karolinska Institutet and Maquet Critical Care AB. The work was supported by grants from Maquet Critical Care AB, the regional agreement on medical training and research (ALF) between Stockholm County Council and the Karolinska Institutet and the HMT project (Health, Medicine and Technology) a collaboration project between the Stockholm County Council and the Royal Institute of Technology. Author’s contribution CH.S., A.O. and H.B.: Study design, data collection and analysis. Drafting of the manuscript. T.S.: Analysis of data and drafting of the manuscript. M.H., F.S.S. and M.W.: Study design, data collection, analysis and critical revision of manuscript. Compliance with ethical standards Conflict of interest Caroline Ha¨llsjo¨ Sander declares no conflict of interest. Thorir Sigmundsson declares no conflict of interest. Magnus Hallba¨ck is employed at Maquet Critical Care AB (MCC). Fernando Suarez Sipmann performs consultant activities for MCC. Mats Wallin is employed at MCC. Anders Oldner declares no conflict of interest. Ha˚kan Bjo¨rne has received a grant for research from MCC.

J Clin Monit Comput

References 1. Marik PE. Perioperative hemodynamic optimization: a revised approach. J Clin Anesth. 2014;26(6):500–5. doi:10.1016/j.jcli nane.2014.06.008. 2. Waldron NH, Miller TE, Gan TJ. Perioperative goal-directed therapy. J Cardiothorac Vasc Anesth. 2014;28(6):1635–41. doi:10.1053/j.jvca.2014.07.008. 3. Cecconi M, Corredor C, Arulkumaran N, Abuella G, Ball J, Grounds RM, Hamilton M, Rhodes A. Clinical review: goaldirected therapy-what is the evidence in surgical patients? The effect on different risk groups. Crit Care. 2013;17(2):209. doi:10. 1186/cc11823. 4. Hamilton MA, Cecconi M, Rhodes A. A systematic review and meta-analysis on the use of preemptive hemodynamic intervention to improve postoperative outcomes in moderate and high-risk surgical patients. Anesth Analg. 2011;112(6):1392–402. doi:10. 1213/ANE.0b013e3181eeaae5. 5. Grocott MP, Dushianthan A, Hamilton MA, Mythen MG, Harrison D, Rowan K, Optimisation Systematic Review Steering G. Perioperative increase in global blood flow to explicit defined goals and outcomes following surgery. Cochrane Database Syst Rev. 2012;11:CD004082. doi:10.1002/14651858.CD004082.pub5. 6. Gurgel ST, do Nascimento P Jr. Maintaining tissue perfusion in high-risk surgical patients: a systematic review of randomized clinical trials. Anesth Analg. 2011;112(6):1384–91. doi:10.1213/ ANE.0b013e3182055384. 7. Ramsingh D, Alexander B, Cannesson M. Clinical review: does it matter which hemodynamic monitoring system is used? Crit Care. 2013;17(2):208. doi:10.1186/cc11814. 8. Sander CH, Hallback M, Wallin M, Emtell P, Oldner A, Bjorne H. Novel continuous capnodynamic method for cardiac output assessment during mechanical ventilation. Br J Anaesth. 2014;112(5):824–31. doi:10.1093/bja/aet486. 9. Sander CH, Hallback M, Sipmann FS, Wallin M, Oldner A, Bjorne H. A novel continuous capnodynamic method for cardiac output assessment in a porcine model of lung lavage. Acta Anaesthesiol Scand. 2015. doi:10.1111/aas.12559. 10. Gedeon A, Forslund L, Hedenstierna G, Romano E. A new method for noninvasive bedside determination of pulmonary blood flow. Med Biol Eng Comput. 1980;18(4):411–8. 11. Capek JM, Roy RJ. Noninvasive measurement of cardiac output using partial CO2 rebreathing. IEEE Trans Biomed Eng. 1988;35(9):653–61. 12. Peyton PJ, Thompson D, Junor P. Non-invasive automated measurement of cardiac output during stable cardiac surgery using a fully integrated differential CO(2) Fick method. J Clin Monit Comput. 2008;22(4):285–92. doi:10.1007/s10877-0089131-2. 13. Yang XX, Critchley LA, Rowlands DK, Fang Z, Huang L. Systematic error of cardiac output measured by bolus thermodilution with a pulmonary artery catheter compared with that measured by an aortic flow probe in a pig model. J Cardiothorac Vasc Anesth. 2013;27(6):1133–9. doi:10.1053/j.jvca.2013.05.020. 14. Peeters Y, Bernards J, Mekeirele M, Hoffmann B, De Raes M, Malbrain ML. Hemodynamic monitoring: to calibrate or not to calibrate? Part 1—Calibrated techniques. Anaesthesiol Intensive Ther. 2015;47(5):487–500. doi:10.5603/AIT.a2015.0073. 15. Tusman G, Sipmann FS, Bohm SH. Rationale of dead space measurement by volumetric capnography. Anesth Analg. 2012;114(4):866–74. doi:10.1213/ANE.0b013e318247f6cc. 16. Critchley LA, Yang XX, Lee A. Assessment of trending ability of cardiac output monitors by polar plot methodology.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

J Cardiothorac Vasc Anesth. 2011;25(3):536–46. doi:10.1053/j. jvca.2011.01.003. Cecconi M, Rhodes A, Poloniecki J, Della Rocca G, Grounds RM. Bench-to-bedside review: the importance of the precision of the reference technique in method comparison studies–with specific reference to the measurement of cardiac output. Crit Care. 2009;13(1):201. doi:10.1186/cc7129. Transonic Systems Inc. Transonic precision flowprobes: perivascular flowprobe specifications. Technical Report 2011. vol RL-20a-ds Rev C 2–11. Kirkeby-Garstad I, Tronnes H, Stenseth R, Sellevold OF, Aadahl P, Skogvoll E. The precision of pulmonary artery catheter bolus thermodilution cardiac output measurements varies with the clinical situation. J Cardiothorac Vasc Anesth. 2014;. doi:10. 1053/j.jvca.2014.12.016. Monnet X, Persichini R, Ktari M, Jozwiak M, Richard C, Teboul JL. Precision of the transpulmonary thermodilution measurements. Crit Care. 2011;15(4):R204. doi:10.1186/cc10421. Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat. 2007;17(4):571–82. doi:10.1080/10543400701329422. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–10. 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(2):85–91. Squara P, Imhoff M, Cecconi M. Metrology in medicine: from measurements to decision, with specific reference to anesthesia and intensive care. Anesth Analg. 2015;120(1):66–75. doi:10. 1213/ANE.0000000000000477. Flammer SMA, Critchley LA, Weber A, Pirbodaghi T, Brinks H, Vandenberghe S. Reliability of lithium dilution cardiac output in anaesthetized sheep. Br J Anaesth. 2013;111(5):833–9. doi:10. 1093/bja/aet220. Critchley LA, Lee A, Ho AM. A critical review of the ability of continuous cardiac output monitors to measure trends in cardiac output. Anesth Analg. 2010;111(5):1180–92. doi:10.1213/ANE. 0b013e3181f08a5b. Reuter DA, Huang C, Edrich T, Shernan SK, Eltzschig HK. Cardiac output monitoring using indicator-dilution techniques: basics, limits, and perspectives. Anesth Analg. 2010;110(3): 799–811. doi:10.1213/ANE.0b013e3181cc885a. Pinsky MR, Vincent JL. Let us use the pulmonary artery catheter correctly and only when we need it. Crit Care Med. 2005;33(5):1119–22. Hadian M, Pinsky MR. Functional hemodynamic monitoring. Curr Opin Crit Care. 2007;13(3):318–23. doi:10.1097/MCC. 0b013e32811e14dd. Cecconi M, Dawson D, Casaretti R, Grounds RM, Rhodes A. A prospective study of the accuracy and precision of continuous cardiac output monitoring devices as compared to intermittent thermodilution. Minerva Anestesiol. 2010;76(12):1010–7. Schloglhofer T, Gilly H, Schima H. Semi-invasive measurement of cardiac output based on pulse contour: a review and analysis. Can J Anaesth. 2014;61(5):452–79. doi:10.1007/s12630-0140135-8. Peyton PJ, Chong SW. Minimally invasive measurement of cardiac output during surgery and critical care: a meta-analysis of accuracy and precision. Anesthesiology. 2010;113(5):1220–35. doi:10.1097/ALN.0b013e3181ee3130.

123