Recent evidence suggests that cardiac output (CO) should be monitored and goal-directed therapies applied in patients undergoing major abdominal surgery.1 , 2 3
Results from Bland-Altman style validation studies that compared the NICOM with reference CO have yielded mostly indifferent results with percentage errors above the 30% threshold for clinical acceptance.4–6 7 n = 9) showed good correlation (r = 0.87) and trending on time and 4-quadrant plots. These authors also showed good trending against thermodilution (r = 0.9) in postoperative patients. Heerdt et al.8 n = 5). What has yet to be shown is whether the good trending ability of NICOM translates to reliable intraoperative clinical use.
One of the major obstacles of validating CO technology in the clinical setting has been the lack of a suitable reference standard. Our group at the Chinese University of Hong Kong has recently shown that Doppler monitoring can be used to reliably measure trends in CO intraoperatively, despite any lack in accuracy.9
Preliminary results by our group when using the NICOM showed that certain surgical interventions, such as placement of retractors in the upper abdomen for open surgery, insufflation with carbon dioxide gas for laparoscopy, and head-down tilt, all caused a shift in calibration of NICOM readings against those from the 2 Doppler methods. Therefore, a validation study was proposed that would compare CO changes measured by the NICOM and the 2 Doppler methods used in tandem. The study was designed to look at the effects of intraoperative surgical interventions such as (1) retractor placement in the upper abdomen, (2) laparoscopic insufflation of the peritoneal cavity, and (3) positioning the patient in a steep head-down tilt for robotic surgery.
METHODS
Ethics and Patient Recruitment
Approval from the Joint Chinese University of Hong Kong-New Territories East Cluster Research Ethics Committee was granted for the study, and written informed patient consent was obtained. Adult ASA physical status I, II and stable III patients scheduled for major surgery were recruited. Patients were selected as they presented for routine surgery at the Prince of Wales Hospital and if their surgery was suitable for inclusion into 1 of 4 study groups: (a) controls having lower abdominal or peripheral surgery and 3 intervention subgroups; (b) laparoscopic surgery (i.e., use of abdominal cavity insufflation); (c) open upper abdominal surgery (i.e., use of large multiblade metal retractors in the upper abdomen); and (d) robotic surgery (i.e., patient in steep head-down Trendelenburg position with abdominal cavity insufflation).
Anesthetic Technique
All patients were anesthetized by using a general anesthesia technique of induction with propofol, muscle relaxation with atracurium, maintenance with sevoflurane in oxygen-enriched air, and analgesia with opioid, including remifentanil infusion. The lungs were mechanically ventilated to mild hypocapnia via an endotracheal tube. None of the patients received a regional block. Standard patient monitoring was used plus an arterial line where indicated.
Equipment—NICOM Cheetah
The NICOM is a transthoracic bioimpedance device that uses frequency modulation (i.e., phase shift) of the impedance signal to measure aortic blood flow and thus CO. It passes a harmless high-frequency (i.e., 75 kHz), low-amplitude (i.e., <4 mA), alternating current through the thorax of the patient, and it measures a continuous variable from the thoracic impedance signal called “phase shift” (Ф) from which stroke volume is calculated. CO is derived by multiplying by heart rate. The transit route, or flux, of alternating current across the thorax is through the conductive blood-filled centrally positioned heart and great vessels, and its transit is slightly delayed by the inductive and the capacitive impedance of the blood that causes the lag in phase. As the amount of blood volume lying within the path of the current flux varies with the cardiac cycle, so does the phase shift measured by the NICOM. From the differential phase shift waveform (1) a surrogate of peak systolic blood flow in the aorta [i.e., d(Ф)/dt(max)], and (2) the duration of left ventricular systole (i.e., left ventricular ejection time) are measured; these 2 derived variables are used to calculate stroke volume. A third calibration factor is also needed and measured from the overall impedance of the thorax (i.e., Zo). The NICOM is calibrated by inputting the patient’s age, sex, height, and weight and measuring thoracic impedance. Although a continuous monitor of CO, the NICOM displays new readings every 30 seconds. The NICOM monitor is attached to the patient by 4 proprietary dual skin electrodes that are placed over the 2 clavicles, left and right, and lower thorax, left and right, at the level of the diaphragm. Numerical and graphic data are displayed by selecting 1 of the 3 screen displays.
Equipment—In Tandem Doppler Monitoring
“In tandem” means using the 2 Doppler monitors together to provide the most reliable CO data possible by cross-checking readings with respect to scan quality and erroneous flow profiles. The USCOM is an external Doppler device that can be used to measure CO intraoperatively. A handheld probe placed in the sternal notch is used, which insonates the blood flow across the aortic valve.10 11 9 , 12 , 13 14 15
Study Protocol
The NICOM was attached to the patient using the 4 dual electrodes with the lower electrodes fixed in position with waterproof transparent dressings. It was calibrated and set to “Run” mode before induction of anesthesia. The USCOM was used at intervals via the suprasternal route, except in one case where pulmonary valve data were used instead. The CardioQ probe was inserted via the nose into the lower esophagus after induction of anesthesia. The scan data from the 2 Doppler monitors were assessed for acceptable quality, without which the study could not proceed. For USCOM, a score of 5 or above was considered acceptable.15
Study Size
The original sample size submitted to the ethics committee was 17 patients, based on achieving 90% power to detect a CO difference of 0.5 L/min between USCOM and NICOM readings. This represented a 10% difference in CO readings, assuming that the average CO for an adult was 5 L/min. In most goal-directed studies, >10% to 15% is set as the threshold for clinically significant change. However, it quickly became apparent from preliminary time plots that major surgical interventions affected NICOM readings relative to Doppler and that these effects should become the focus of the study. Thus, the protocol was extended to include 3 groups of 6 patients for each type of major abdominal intervention and 9 control cases where there was no upper abdominal intervention. As the shifts in NICOM readings were 0.5 to 1.0 L/min/m2 , the study was still adequately powered.
Data Analysis
Data collected during the study were compiled using Excel spreadsheets (Microsoft, Redmond, WA). The CO data were indexed (CI) to body surface area. Patient CI data were first analyzed within individuals by (1) drawing time plots that compared the 3 sets of CO data and (2) performing within-individual regression analysis. Correlation coefficients were generated by the linear regression and correlation function, which is provided by Excel and uses Pearson method. An R 2 > 0.8 was set to show good trending of CO between methods and R 2 <0.6 unreliable trending.16
The CI data were also analyzed across patient groups for the 3 CO devices. The average and range of CI for each set of patient data were calculated. Comparisons were then performed on these summary data and R 2 correlations using Student t test or analysis of variance, as appropriate.
Scatter plots with regression lines were drawn using data from all patients with USCOM data as the independent variable. Correlation coefficients were calculated.
The Bland-Altman analysis was performed with limits of agreement generated using MedCalc (MedCalc Software bvba, Ostend, Belgium), which corrects for repeated measures. The Bland-Altman method was used.17
Percentage error was also calculated.18
Trend analysis using all the data was performed using (a) the change in CI (ΔCI) that was calculated (i.e., CIa − CIb ) from serial CI readings. The ΔCI data were first plotted on a 4-quadrant plot and concordance analysis performed using a 15% exclusion zone of 0.5 L/min/m2 . An acceptance level for good trending between monitors was set at >92%.16
; n = number of data points and p = proportion in agreement.
StatView version 5 for Windows (SAS Institute Inc., Cary, NC) was used for the remaining statistical analysis. P < 0.05 was considered significant. Graphs were drawn using Sigma plot version 7.1 (Systat Software Inc., San Jose, CA).
RESULTS
Patient Data
Between June 2013 and March 2014, 28 patients scheduled for major surgery at the Prince of Wales Hospital were recruited. NICOM data were successfully obtained from them all. The USCOM failed to provide satisfactory data because of poor quality scans with a Cattermole score of <5 of 12 in 3 patients (Tables 1 and 2 ). Acceptable USCOM scans with median (range) Cattermole scores of 8 (6–11) were achieved in 86% of study cases. In 1 patient, CardioQ data were used instead (i.e., case 27) and in another pulmonary valve USCOM data were used instead (i.e., case 8). The CardioQ failed to provide satisfactory data in 6 cases because of failure to locate a reasonable quality signal (n = 4) or cable failure (n = 2). In several patients (n =8), CardioQ data collection was incomplete because of delays with probe insertion or failure to obtain a satisfactory flow signal. In 1 patient, neither of the 2 Doppler monitors could be used, and this patient was excluded from any further analysis. The Doppler scans were performed by 2 experienced operators.
Table 1: Demographic and Correlation Data for Control Group Patients
Table 2: Demographic, Scan, Correlation, and Bias Data from Intervention Group Patients
Individual patient details are summarized in Table 1 for control group cases and Table 2 for intervention group cases. The mean (range) age of all the patients was 58 (32–78) years, sex 12 males and 15 females, ASA physical status (I/II/III) 7/16/4, weight 62 (41–95) kg, height 161 (148–185) cm, and body mass index 24 (16–32) kg/m2 . Three hundred ninety sets of CO readings were collected, which included 372 sets of USCOM data and 274 sets of CardioQ data. Mean (range) duration of data collection in each patient was 248 (98–667) minutes or 4 (1½–11) hours. Accordingly, sets of readings per patient were mean (range) 14 (7–27).
Summary of CI Data
The average CI for each study was (mean [SD]) 3.2 (0.7) L/min/m2 for NICOM, 3.5 (1.0) for USCOM, and 3.2 (0.6) for CardioQ, which were all similar (P = 0.2500). The range of CIs for each study was (mean [SD]) 1.9 (0.5) to 3.9 (0.8) L/min/m2 for NICOM, 2.3 (0.6) to 4.1 (0.9) for USCOM, and 2.1 (0.6) to 3.6 (1.2) for CardioQ.
Within-Individual Correlation Data
The within-individual correlations between the 3 monitoring systems are presented in Tables 1 and 2 . In control group patients, the correlations between the NICOM and the USCOM readings (mean [range]) was R 2 = 0.89 (0.69–0.97), which was similar to that between the USCOM and CardioQ readings of R 2 = 0.90 (0.70– 0.97) (P = 0.7000). In the intervention groups, the correlations between NICOM and USCOM readings were R 2 = 0.43 (0.03–0.71), which was significantly less than that between CardioQ and USCOM readings, which was R 2 = 0.87 (0.60–0.97) (P < 0.0001) and indicated that the trending ability of NICOM was less reliable than the Doppler methods in these groups of patients.
Grouped Data Analysis Figure 1: Scatter and Bland-Altman plots comparing NICOM and CardioQ readings with those from the USCOM. CI = cardiac index.
When NICOM and Doppler readings from all 27 patients (i.e., all USCOM readings with exception of case 27) were plotted together on a scatter plot, there was a wide dispersion of data pairs shown by a poor correlation coefficient of R 2 = 0.28 (Fig. 1, upper left plot). The accompanying Bland-Altman plot had wide limits of agreement with a percentage error of 57 (52–62) % and bias between methods of 0.37 (0.27–0.47) L/min/m2 (Fig. 1, upper right plot). In comparison, the CardioQ readings were more highly correlated with USCOM readings, R 2 = 0.50, and had better Bland-Altman agreement with a percentage error of 42 (39–45) % and bias of 0.14 (0.03–0.25) L/m/m2 (Fig. 1, lower plots).
Grouped Data Trend Analysis
The trending capabilities of the NICOM and CardioQ against the USCOM were further analyzed using ΔCI data on a 4-quadrant plot (i.e., concordance analysis). A central exclusion zone for small CI changes was set at 0.5 L/min/m2 , which was based on 15% of the mean CI for the study, which was 3.3 L/min/m2 .16
The concordance rate (confidence intervals) between the NICOM and the USCOM was 82 (77–88) % (Fig. 2, upper left plot), which was well below the 92% threshold for good trending capability.16
Figure 2: Four quadrant plots showing ΔCI data for NICOM or CardioQ comparisons with USCOM (upper plots) and NICOM control and intervention groups compared with USCOM (lower plots). CI = cardiac index.
The concordance rates between the NICOM and the USCOM for each patient group were (a) controls 95 (90–100) %, (b) laparoscopy 67 (53–83) %, (c) open surgery using retractors 75 (67–88) %, and (d) robotic surgery with head-down tilt 85 (74–96) %. Only in the control group with a concordance rate of 95% was good trending between the 2 monitors shown (Fig. 2, lower plots).
Time Plots Figure 3: Annotate time plot from case 17 showing the trend relations between the 3 CI monitors. Regressions plots of NICOM and CardioQ against USCOM provided (right). CI = cardiac index.
Figure 4: Annotated time plots from 3 further cases 2, 15, and 25. Correlation coefficients for CardioQ and NICOM readings against USCOM are shown (lower right corner).
Within-individual CI readings were also plotted against time to show their trend relationships (Figs. 3 and 4). In most cases, and when the data were available, the plots showed that USCOM and CardioQ readings tracked each other closely as CO varied. However, NICOM readings did not always track USCOM and CardioQ readings, and in the intervention groups, there were obvious deviations from the USCOM and CardioQ trend lines. In 13 of 18 patients (72%), the NICOM readings decreased relative to Doppler during the intervention, but in 5 of 18 (28%) the readings increased. The magnitude of the changes in calibration (mean [range]) was ±0.9 (0.6–1.4) L/min/m2 (Table 2 ).
DISCUSSION
The main findings from this study were that in the intervention groups the correlations between NICOM and USCOM readings were R 2 = 0.43 (0.03–0.71); significantly less than that between CardioQ and USCOM readings, which were R 2 = 0.87 (0.60–0.97) (P < 0.0001) and which indicated that the trending ability of NICOM was less reliable than the Doppler methods in these patients. Surgical interventions that affected the upper abdomen, such as open retractor placement, laparoscopic insufflation, and head-down tilt, were shown to be the main cause, and these shifts reverted back on cessation of the intervention (Figs. 3 and 4).
The shifts in NICOM readings are a novel finding as previous authors had not addressed the issue of trending by the NICOM in such a diligent manner. Most of the published validation work on the NICOM has involved comparisons against another CO method when applying some test intervention, such as goal-directed fluid therapy or passive leg raise.19 , 20 4 R 2 = 0.27 and their Bland-Altman percentage error was 60%, which were results similar to the present findings of R 2 = 0.43 and percentage error = 57% for NICOM versus Doppler comparisons. Conway et al.4 4 21 21
In the present study, similar examples were found of shifts in NICOM calibration. Figure 3 shows surgical retractor placement. The CI readings from the 2 Doppler monitors are seen to track each other closely for the duration of surgery and with minimal deviations for 5 hours. A regression plot of the USCOM against CardioQ data (i.e., lower right plot) demonstrates their alignment with a close correlation of data points of R 2 = 0.93. Although the readings from the NICOM still tracked the Doppler trend, there was a major downward deviation of its trend line during the period when a surgical retractor was used. The case was partial liver resection, or hepatectomy (i.e., case 17, Table 2 ), which involved the placement of a large multiblade retractor in the upper abdomen to provide surgical access. The correlation coefficient of R 2 = 0.65 between the NICOM and the USCOM readings (i.e., upper right plot) is not as good as that between the CardioQ. However, on closer inspection, there are 2 sets of points that are better aligned and correlated, which correspond to with and without retractor data.
Figure 4 shows 3 more examples. Case 15 was selected from the laparoscopy group as a typical patient who underwent laparoscopic sigmoid colectomy (Table 2 ; Fig. 4 middle plot). NICOM readings are seen to diverge downward from the USCOM trend line. After 30 minutes, the abdominal cavity is insufflated, which resulted in a decrease of NICOM readings relative to USCOM trend line by 0.7 L/min/m2 and lasted for the duration of abdominal insufflation (R 2 = 0.36). The CardioQ readings followed the USCOM trend line without any major deviations (R 2 = 0.89). Case 25 was selected from the robotic surgery group as a typical patient who underwent a total mesorectal excision of a rectal cancer (Table 2 ; Fig. 4 lower plot). In addition to abdominal insufflation, the patient required a steep head-down tilt. These interventions caused the NICOM readings to deviate downward and to fall below the USCOM trend line by 0.9 L/min/m2 , which lasted for the duration of the interventions (R 2 = 0.20). The CardioQ readings did not deviate from the USCOM trend line; however, they were offset upward with readings 0.5 to 1.0 L/min/m2 higher than those of the USCOM (R 2 = 0.91).
Case 2 was selected from the control group where no upper abdominal intervention was performed. The patient underwent total abdominal hysterectomy with bilateral salpingo-oophorectomy and excision of the omentum. The surgery was confined to the lower abdomen and was associated with major blood loss of over 3 L (Table 1 ; Fig. 4 upper plot). NICOM readings are seen to follow the USCOM trend line with a correlation of R 2 = 0.96. CardioQ readings are shown to also follow the USCOM trend line with a correlation of R 2 = 0.94, but with a downward offset, or bias, of approximately 1 L/min/m2 . When analyzing data from this study, the authors also looked at other factors that could cause shifts in NICOM calibration, such as blood volume changes, hemoglobin levels, and peripheral vascular resistance changes. Resistivity of blood is a factor in calculating bioimpedance CO and is related to the hematocrit, even though most bioimpedance algorithms assume that it remains constant.22 23
The present study confirms research published previously using bioimpedance CO during surgery (i.e., the BioMed),24 24
It is easy to assume that abdominal insufflation, head-down tilt, or insertion of a retractor displaces the diaphragm upward, and therefore alters flux pathways. Furthermore, one would expect the associated change in NICOM readings relative to Doppler to be similar in direction. However, in the present study, the NICOM readings showed a paradoxical response, with shifts downward in 72% of cases, but upward in 28% (Tables 1 and 2). Thus, the physiological explanation seems more complicated than just upward displacement of the diaphragm. Possibly, changes in shape and position of the heart are also involved.
Doppler measurements could also have been affected by the surgical interventions. The angle that the esophageal probe makes with the descending aorta may have been changed, and this possibility has been suspected by the authors in several cases involving the upper abdominal aorta and kidneys.9 25 2 ) between the 2 Doppler methods, which would support an explanation of change in probe alignment.
Another possible explanation is the positioning of the lower NICOM electrodes. In the present study, the electrodes were placed laterally on the chest wall, but the actual dermatome level was never recorded. High or low placement of these electrodes could have influenced current flux pathways and determined whether NICOM readings shifted upward or downward after surgical interventions.
The study had some limitations; (1) an accepted CO reference method, such as thermodilution, was not included in the protocol. However, thermodilution CO is no longer used routinely in anesthesia practice, and thus its use in the present study would have been difficult to justify ethically. Furthermore, to collect the sets of simultaneous thermodilution, CO readings every 15 to 30 minutes would have been difficult, if not impossible, to achieve. (2) The measurement of Doppler CO required a high level of psychomotor skill, and this inevitably would have introduced some bias. The Doppler trend lines upon which the analyses of the study were based were shown to be highly consistent by the R 2 data. (3) Fluid responsiveness was not assessed, which limited the clinical applications of the outcomes of the study.19 , 20
In conclusion, abdominal surgical procedures decrease the ability of the NICOM to track changes in CO reliably. However, interventions that alter the geometry of the upper abdomen can alter NICOM readings by >1 L/min/m2 , and the direction of these shifts in calibration is unpredictable. Using Doppler CO monitoring in tandem was a key step in establishing a valid trend line of CO changes, which helped elucidate the study findings. However, the electrophysiological reasons for the shifts in NICOM calibration could not be determined. The clinical relevance of the study is that anesthesiologists should be aware of the possibility of calibration shifts, as they can be large (i.e., >1 L/min/m2 ) and should anticipate their occurrence whenever the NICOM is used, especially during abdominal surgery.
DISCLOSURES
Name : Huang Li, MB BS, PhD.
Contribution : This author helped design the study, collected and analyzed the data, and prepared the manuscript.
Attestation : Huang Li approved the final manuscript.
Name : Lester A. H. Critchley, MD, FFARCSI, FHKAM.
Contribution : This author supervised the study, helped design the study, collected and analyzed the data, and prepared the manuscript.
Attestation : Lester A. H. Critchley approved the final manuscript.
Name : Zhange Jie, MB BS.
Contribution : This author helped recruit patients and collect data.
Attestation : Zhang Jie approved the final manuscript.
This manuscript was handled by: Maxime Cannesson, MD, PhD.
REFERENCES
1. Pearse RM, Harrison DA, MacDonald N, Gillies MA, Blunt M, Ackland G, Grocott MP, Ahern A, Griggs K, Scott R, Hinds C, Rowan KOPTIMISE Study Group. OPTIMISE Study Group. . Effect of a perioperative, cardiac output-guided hemodynamic therapy algorithm on outcomes following major gastrointestinal surgery: a randomized clinical trial and systematic review. JAMA. 2014;311:2181–90
2. 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;112:1392–402
3. Marik PE. Noninvasive cardiac output monitors: a state-of the-art review. J Cardiothorac Vasc Anesth. 2013;27:121–34
4. Conway DH, Hussain OA, Gall I. A comparison of noninvasive bioreactance with oesophageal Doppler estimation of stroke volume during open abdominal surgery: an observational study. Eur J Anaesthesiol. 2013;30:501–8
5. Kober D, Trepte C, Petzoldt M, Nitzschke R, Herich L, Reuter DA, Haas S. Cardiac index assessment using bioreactance in patients undergoing cytoreductive surgery in ovarian carcinoma. J Clin Monit Comput. 2013;27:621–7
6. Kupersztych-Hagege E, Teboul JL, Artigas A, Talbot A, Sabatier C, Richard C, Monnet X. Bioreactance is not reliable for estimating cardiac output and the effects of passive leg raising in critically ill patients. Br J Anaesth. 2013;111:961–6
7. Keren H, Burkhoff D, Squara P. Evaluation of a noninvasive continuous cardiac output monitoring system based on thoracic bioreactance. Am J Physiol Heart Circ Physiol. 2007;293:H583–9
8. Heerdt PM, Wagner CL, DeMais M, Savarese JJ. Noninvasive cardiac output monitoring with bioreactance as an alternative to invasive instrumentation for preclinical drug evaluation in beagles. J Pharmacol Toxicol Methods. 2011;64:111–8
9. Huang L, Critchley LA. An assessment of two Doppler-based monitors to track cardiac output changes in anaesthetised patients undergoing major surgery. Anaesth Intensive Care. 2014;42:631–9
10. Critchley LA, Huang L. USCOM-window to the circulation: utility of supra-sternal Doppler in an elderly anaesthetized patient for a robotic cystectomy. J Clin Monit Comput. 2014;28:83–93
11. Singer M. Oesophageal Doppler. Curr Opin Crit Care. 2009;15:244–8
12. Critchley LA, Peng ZY, Fok BS, Lee A, Phillips RA. Testing the reliability of a new ultrasonic cardiac output monitor, the USCOM, by using aortic flowprobes in anesthetized dogs. Anesth Analg. 2005;100:748–53
13. Phillips RA, Hood SG, Jacobson BM, West MJ, Wan L, May CN. Pulmonary artery catheter (PAC) accuracy and efficacy compared with flow probe and transcutaneous Doppler (USCOM): an ovine cardiac output validation. Crit Care Res Pract. 2012;2012:621496
14. Huang L, Critchley LA, Lok RL, Liu Y. Correlation between supra-sternal Doppler cardiac output (USCOM) measurements and chest radiological features. Anaesthesia. 2013;68:1156–60
15. Cattermole GN LP, Tang CO, Smith BE, Graham CA, Rainer TH. A new method to score the quality of USCOM scans. Hong Kong J Emerg Med. 2009;16:288
16. 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;111:1180–92
17. Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat. 2007;17:571–82
18. 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
19. Waldron NH, Miller TE, Thacker JK, Manchester AK, White WD, Nardiello J, Elgasim MA, Moon RE, Gan TJ. A prospective comparison of a noninvasive cardiac output monitor versus esophageal Doppler monitor for goal-directed fluid therapy in colorectal surgery patients. Anesth Analg. 2014;118:966–75
20. Benomar B, Ouattara A, Estagnasie P, Brusset A, Squara P. Fluid responsiveness predicted by noninvasive bioreactance-based passive leg raise test. Intensive Care Med. 2010;36:1875–81
21. Squara P, Rotcajg D, Denjean D, Estagnasie P, Brusset A. Comparison of monitoring performance of Bioreactance vs. pulse contour during lung recruitment maneuvers. Crit Care. 2009;13:R125
22. Quail AW, Traugott FM, Porges WL, White SW. Thoracic resistivity for stroke volume determination in impedance cardiography. J Appl Physiol Respir Environ Exerc Physiol. 1981;50:191–5
23. Schlöglhofer T, Gilly H, Schima H. Semi-invasive measurement of cardiac output based on pulse contour: a review and analysis. Can J Anaesth. 2014;61:452–79
24. Critchley LA, Leung DH, Short TG. Abdominal surgery alters the calibration of bioimpedance cardiac output measurement. Int J Clin Monit Comput. 1996;13:1–8
25. Singer M, Clarke J, Bennett ED. Continuous hemodynamic monitoring by esophageal Doppler. Crit Care Med. 1989;17:447–52
