A comparison of noninvasive bioreactance with oesophageal Doppler estimation of stroke volume during open abdominal surgery

An observational study

Abstract

CONTEXT 

The anaesthetist must maintain tissue perfusion by ensuring optimal perioperative fluid balance. This can be achieved using less invasive cardiac output monitors such as oesophageal Doppler monitoring (ODM). Other less invasive cardiac output monitors using bio-impedence technology (noninvasive cardiac output monitoring, NICOM) may have a role in monitoring the circulation and informing fluid management decisions.

OBJECTIVE 

To compare estimates of stroke volume from ODM with those from NICOM, a noninvasive monitor using bioreactance, a modification of transthoracic bio-impedence.

DESIGN 

An observational study.

SETTING 

Manchester Royal Infirmary, UK. Data collected in 2011 and 2012.

PARTICIPANTS 

Twenty-two patients scheduled for major, open abdominal surgery. Reasons for noninclusion: atrial fibrillation; heart failure; oesophageal disease; lack of capacity; and known sensitivity to colloid.

INTERVENTION 

All patients had oesophageal Doppler cardiac output monitoring as a standard element of anaesthesia care. We placed NICOM Bioreactance electrodes and recorded stroke volume estimates from both devices. Fluid challenges were given by the anaesthetist and the haemodynamic responses were recorded.

MAIN OUTCOME MEASURE 

Stroke volume during surgery. The Bland–Altman method was used to compare bias and limits of agreement for stroke volume and cardiac output. Fluid responders were defined as patients who increased stroke volume by at least 10% after fluid loading. The precision of each device was calculated during periods of haemodynamic stability.

RESULTS 

We made 788 acceptable measurements of cardiac output. The bias was −6.9 ml and the limits of agreement were −22.9 to 36.8 ml. The percentage error was 57%. Average precision for both the ODM and NICOM were similar, 8.5% (SD 5.4%) and 8.7% (SD 3.2%). The concordance for the stroke volume change following fluid challenge was 90.5%. Both devices produced unacceptable readings with electrical diathermy.

CONCLUSION 

Simultaneous stroke volume estimations made by noninvasive Bioreactance (NICOM) and oesophageal Doppler showed bias and limits of agreement that are not clinically acceptable. The measurements made by these two devices cannot be regarded as interchangeable.

Introduction

Optimising the intravascular volume status of individual patients undergoing surgical procedures has become an essential part of modern anaesthetic practice. The use of minimally invasive cardiac output monitors to guide perioperative administration of intravenous fluid and vasoactive drugs has been shown to improve outcome by reducing postoperative complications. Monitors such as the oesophageal Doppler, which estimates blood flow in a dynamic situation such as surgery, are gaining popularity compared with static pressure based monitoring systems.^1–4 The National Institute for Health and Clinical Excellence (NICE), a UK governmental body, recommends using oesophageal Doppler monitoring (ODM) in patients undergoing major surgery for whom a clinician would consider using invasive cardiovascular monitoring (http://www.nice.org.uk/MTG3). However, ODM has limitations: the probe requires regular adjustment; it cannot be used in patients having oesophageal or head and neck surgery; and, it is difficult to use in awake patients. Less invasive alternatives to the pulmonary artery catheter have been proposed for patients undergoing major surgery.⁵ One alternative is bioimpedence, which measures changes in transthoracic electrical flow and impedence to estimate blood flow noninvasively. However, previous attempts to use bioimpedance clinically have been unsuccessful due to signal interference.⁶ The Noninvasive Cardiac Output Monitor (NICOM, Cheetah Medical, Portland, Oregon) is an innovative bioimpedance device processing and analysing the electrical signal using the novel technique of bioreactance that takes into account changes in conductivity of blood during the cardiac cycle. Using NICOM Bioreactance, derived cardiac output can potentially overcome problems associated with ODM. Previous validation studies of NICOM in cardiac patients have shown it to be comparable to continuous thermodilution, bolus thermodilution and uncalibrated pulse contour analysis even in situations in which haemodynamic physiology is rapidly changing.^7–9 These studies did not look at patients undergoing abdominal surgery or receiving perioperative intravenous fluid challenges. Haemodynamic changes following intravenous fluid bolus during surgery are well established and have been used to evaluate cardiac output monitors.^10,11 The aim of the present study was to compare stroke volume monitoring with NICOM Bioreactance with those of ODM in open abdominal surgery with intravenous fluid challenges.

Materials and methods

The protocol used in the present study was approved on by Greater Manchester South Research Ethics Committee.

We obtained written, informed consent from patients undergoing open abdominal surgery, if they fulfilled the inclusion criteria of good preoperative fitness level (>4 metabolic equivalents) or adequate cardiorespiratory reserve estimated by cardiopulmonary exercise testing performed by the preoperative anaesthetic assessment team; sinus rhythm on preoperative electrocardiograph; individualised stroke volume optimisation planned by the anaesthetist using ODM; capacity and ability to provide informed consent. Patients with atrial fibrillation on preoperative ECG; congestive cardiac failure or poor cardiorespiratory reserve as measured by cardiopulmonary exercise testing; haemodynamically significant valvular disease; oesophageal disease; hypersensitivity to colloid solution; allergy to ECG electrode; or, no capacity or ability to provide informed consent were excluded from the study.

Anaesthesia and monitoring techniques

Anaesthesia was induced with propofol, rocuronium and either fentanyl or remifentanil. The tracheas of all patients were intubated and all patients were commenced on mechanical ventilation (Primus Draeger, Lubeck, Germany) with a tidal volume of 6 to 8 ml kg⁻¹ and positive end expiratory pressure (PEEP) of 1 to 5 cmH₂O. The NICOM system was calibrated prior to induction of anaesthesia. NICOM stickers consist of four dual electrodes that are placed on the four quadrants of the anterior chest wall. Within each sticker, one electrode is driven by an electric generator to inject the high frequency current across the body, whereas the other electrode connects to a voltage amplifier. Signals are applied to, and recorded from, both the left and right side. To improve the signal, we shaved body hair at the site of placement, cleaned with a chlorhexidine 2% sponge and placed a waterproof dressing over each electrode. The NICOM was calibrated to average the stroke volume estimates over 30 s.

Following induction of anaesthesia and tracheal intubation, the CardioQ-ODM probe (Deltex Medical Limited, Chichester, UK) was placed by the anaesthetist in charge of the case. The ODM was calibrated to average stroke volume over 30 s. Every 5 min, we recorded simultaneous snapshots of stroke volume displayed by each device. Other variables such as cardiac output, heart rate and mean arterial pressure (MAP) were also recorded every 5 min. The start and end times of fluid challenges, defined as a bolus of 250 to 500 ml of colloid solution, were noted. We anticipated a monitoring duration of 2 to 4 h. The NICOM electrodes were removed from the patient shortly after the end of surgery.

An independent consultant anaesthetist, blinded to the type of monitor, evaluated signals that appeared to be artefact. We disregarded measurements designated by the monitors as ‘weak’, allowing only ‘adequate’ or ‘strong’ signals to be recorded. The patient's response to a fluid challenge was determined by recording the stroke volume measurements of ODM and NICOM 1 min prior to the fluid infusion, and then recording stroke volume measurements each minute following infusion with a rise of more than 10% in stroke volume, suggesting that the patient was fluid responsive.

Statistical analysis

Scatterplots were constructed from ODM and NICOM recordings and linear regression and correlation calculated for ODM stroke volume against NICOM stroke volume including within-individual correlation analysis; ODM cardiac output against NICOM cardiac output including within-individual correlation analysis; MAP against ODM stroke volume; and MAP against NICOM stroke volume. Within-individual or intra-class correlation was estimated using equality of covariance. We used the Bland–Altman method, corrected for repeated measures, to calculate bias, limits of agreement and percentage error. We predefined an acceptable percentage error as less than 30%, as described by Critchley and Critchley.^12,13 We then conducted a posthoc analysis of within-individual regression of the Bland–Altman plot for stroke volume to describe the 95% prediction interval. We calculated precision during 10-min periods of haemodynamic stability by analysing 20 stroke volume readings for each patient.

We measured the time delay between the start of fluid bolus and the point when the maximum stroke volume was obtained. Amplitude response for positive fluid challenges was calculated as the difference between the baseline stroke volume at the start of a fluid challenge and the maximum stroke volume recorded following the challenge. We assessed direction of change by drawing a four-quadrant plot to display the degree of agreement of change in stroke volume readings from both devices. We refined the four-quadrant plot by applying a 10% exclusion zone for small changes in stroke volume. We then calculated the percentage agreement for changes more than 10%.

For all comparisons, P value of 0.05 or less was considered significant. Statistical analysis was performed using SPSS software v13 (SPSS Inc, Chicago, Illinois, USA) and StatsDirect v2.7.8 (StatsDirect statistical software, Cheshire, UK).

Results

We made 788 acceptable measurements of cardiac output and stroke volume from 22 patients (17 men and five women) undergoing bowel resection (five), Whipple's procedure (five), abdominal aortic graft (two), prostatectomy (six), cystectomy (2), nephrectomy (one) and splenectomy (one). Patient characteristics and fluid management are described in Table 1. Figure 1 illustrates stroke volume measurements from one patient. After an initial baseline period, a fluid challenge was given and both the NICOM and ODM display an increase in stroke volume of more than 10%. For one patient having aortic surgery, the Cardio-Q ODM produced un-physiological cardiac output estimates during cross-clamping, whereas the NICOM Bioreactance maintained acceptable readings as judged by an independent anaesthetist. We disregarded measurements made just before, during or after aortic cross-clamping from further analysis.

Table 1: Characteristics for the 22 patients enrolled in the study undergoing open abdominal surgery

Fig. 1: No captions available.

For all patients, equality of covariance returned an estimate of within-individual correlation of 0.44 for stroke volume (R² = 0.19) and 0.52 (R² = 0.27) for cardiac output for ODM and NICOM (Figs. 2 and 3), respectively. No correlation between stroke volume and MAP was demonstrated by either device, with R values of 0.06 for the Doppler and −0.008 for NICOM. Bland–Altman analysis for the stroke volume measurements (Fig. 4) showed a bias of −6.9 ml and the limits of agreement were −22.9 to 36.8 ml. The percentage error was 57%. Bland–Altman analysis for the cardiac output measurements (Fig. 5) resulted in a bias of −0.46 l min⁻¹ and the limits of agreement were −1.64 to 2.54 l min⁻¹. The percentage error was 59.6%. Precision of stroke volume measurements for ODM and NICOM was similar, 8.5% (SD 5.4%) and 8.7% (SD 3.2%), respectively.

Fig. 2: No captions available.

Fig. 3: No captions available.

Fig. 4: No captions available.

Fig. 5: No captions available.

We analysed 59 fluid challenges, having discarded the results of nine challenges due to signal interference at some point during the challenge. The directional response of each device to fluid challenges showed concordance of 90.5% with stroke volume changes of more than 10% (Fig. 6). Following 59 discrete fluid challenges of which 24 were positive, the NICOM agreed with the Cardio-Q ODM on 18 occasions with a sensitivity of 0.75 and specificity of 0.69 for those fluid challenges in which stroke volume increased by more than 10%. We noted three occasions in which NICOM stroke volume fell when the Cardio-Q ODM suggested a rise of more than 10%. The average time to maximum stroke volume from the start of the fluid challenge was 15 min for ODM and 14 min for NICOM. The average amplitude in stroke volume change for fluid challenges wherein ODM and NICOM stroke volume both increased more than 10% was 56.6 and 51.5%, respectively. Posthoc analysis of the Bland–Altman plot for a within-individual regression produced a sloped prediction interval line and no change in the width of the limits of agreement (Fig. 7).

Fig. 6: No captions available.

Fig. 7: No captions available.

Discussion

In this study of patients undergoing major abdominal surgery, simultaneous stroke volume estimations made by noninvasive Bioreactance (NICOM) and oesophageal Doppler showed bias and limits of agreement that are not clinically acceptable. The measurements made by these two devices cannot be regarded as interchangeable.

In order to compare continuous cardiac output monitors, different statistical tests have been proposed, yet consensus has not been reached.¹⁴ In a recent review, Squara, Cecconi and colleagues devised a strategy to overcome the assumption that the reference technology a priori should be superior to the new technique. The authors highlight five areas when evaluating cardiac output monitors: accuracy; precision; short response time; accurate amplitude response; and the ability to detect significant cardiac output directional changes.¹⁵ Critchley and Critchley suggest that new methods of cardiac output measurement should have a similar precision to the reference method, whereby a percentage error less than 30% is acceptable when comparing any new device to transpulmonary thermodilution.¹³ This approach has been challenged; for example, Columb has suggested that researchers predefine a clinical tolerability interval. Extreme misclassification is regarded as unacceptable, whereas misclassification within the tolerability interval is acceptable.¹⁶ Previous reports of ODM suggest that it under-estimates stroke volume compared with the reference standards of thermodilution or aortic flow measurement.^17,18

The two monitors we studied demonstrated poor within-individual correlation with a low intraclass correlation coefficient. This could be related to the influence of another explanatory variable, such as timing of measurement or haemodynamic changes during surgery. The scatterplots of stroke volume and cardiac output (Figs. 2 and 3) suggest a skew away from the zero point. This may occur if the ODM is underestimating stroke volume whilst NICOM overestimated stroke volume compared with ODM. Posthoc analysis of prediction intervals revealed a slope in the Bland–Altman plots such that at higher stroke volumes, the difference between the estimates from each device also increased, although the width of the limits of agreement remained the same.

The NICOM Bioreactance monitor uses noninvasive bio-impedance signals to calculate radiofrequency phase shift when high frequency currents cross the chest, enabling estimation of stroke volume.¹⁹ Validation studies compared NICOM with transpulmonary thermodilution and arterial pulse contour analysis in patients on cardiac and ICUs.^7–9,19 They found that the bias between continuous thermodilution and NICOM was less than 0.2 l min⁻¹, the limits of agreement were ±2 l min⁻¹ and the NICOM reduced response time by 50% compared with continuous thermodilution.²⁰ This agreement is comparable with studies of continuous thermodilution and bolus thermodilution and between transpulmonary thermodilution and ODM.^21,22 Unlike the NICOM validation studies, we investigated the usefulness of Bioreactance in major abdominal surgery, with volume loading as a part of an ODM-guided stroke volume optimisation technique.^23,24

The precision of haemodynamic measurements is best assessed from segments of steady-state readings. Encountering such conditions during abdominal surgery is probably more challenging than in cardiac ICU patients. The NICOM validation studies demonstrated a precision of between 5.6 and 12%.^7,9,19 The precision of oesophageal Doppler has been reported as 8% in ICU patients.²² We measured precision at least 15 min following a fluid challenge because this is a period of haemodynamic stability, with less variability due to mechanical ventilation.^25,26 In comparison, a continuous thermodilution technique has been reported to have a precision of 9 to 11% in experimental conditions rising to 20% in clinical practice or when using bolus thermodilution techniques.^27,28 Recently, with very careful use of the thermodilution technique, Metzelder et al.²⁹ report a precision of 2.6% using a pulmonary artery catheter as the reference device. As the NICOM demonstrated a precision of 8.7%, this may lead us to accept a higher percentage error when comparing devices.

A number of methods of predicting fluid responsiveness during surgery have been proposed. The stroke volume optimisation technique used in this study incorporates the concept wherein a stroke volume change of more than 10% following an intravenous fluid bolus suggests that the patient is ‘fluid responsive’.^3,4,23,24 Static indices such as central venous pressure (CVP), pulmonary artery occlusion pressure and on-off measurements of cardiac output are not good predictors of the response to a fluid challenge.³⁰ Other dynamic predictors of fluid responsiveness are stroke volume variation (SVV) and pulse pressure variation (PPV). Both SVV and PPV have been shown to be superior to static measurements such as CVP.³⁰ NICOM, but not the ODM used at that time, calculated SVV, so we could not compare the two devices in this respect. Not all patients in our study received invasive arterial blood pressure monitoring, so we were unable to assess PPV.

In clinical practice, anaesthetists value the ability of continuous, minimally invasive monitoring to detect directional trends in stroke volume at key moments during surgery. The validation studies of NICOM demonstrated appropriate directional change in response to changes in PEEP during lung recruitment manoeuvres; 20 cmH₂O of PEEP was associated with a rapid reduction (31 ± 15%) in cardiac output.⁹ Lung recruitment manoeuvres producing rapid, large haemodynamic changes are rarely part of intraoperative care; therefore, we looked at the effect of a fluid challenge on stroke volume estimated by each device. We found a high degree of concordance following fluid challenge, suggesting that the directional response of NICOM and ODM is similar during abdominal surgery. These directional changes can be more useful than absolute stroke volume measurements in managing intraoperative fluid therapy. In the NICOM validation study, Squara et al.⁹ measured time response as the delay between the recruitment manoeuvre and the point at which the minimum cardiac output was obtained. For NICOM, the time response was 3 to 4 min, which was comparable to pulse contour analysis.⁹ In our study, the time response from the beginning of fluid challenge and the point at which maximum stroke volume was obtained was very similar between the two devices, as was the measured amplitude response for positive fluid challenges.

We only used data from the devices when the signal was deemed acceptable, discarding data associated with a poor signal. Both devices demonstrated susceptibility to interference by electrical diathermy during surgery. We found that oesophageal Doppler was prone to develop a poor signal during periods of patient positioning, which would require re-focussing of the probe. The NICOM Bioreactance maintained an adequate signal during these periods. We found erroneous cardiac output estimates with ODM during aortic cross-clamping. Systematic cardiac output bias during aortic cross-clamping was demonstrated by Lafanechère et al.³¹ using a different oesophageal Doppler device. Neither ODM nor the NICOM demonstrated any correlation between cardiac output or stroke volume and MAP during surgery. Thus, both devices appear to distinguish flow from pressure within the circulation. These results are in contrast to studies of widely available uncalibrated pulse contour analysis devices, which suggest that cardiac output estimated by these devices can be influenced by MAP.^32,33

The present study has a number of limitations. ODM cannot be regarded as a gold standard for cardiac output measurement due to the known bias with reference methods. Gold standard references such as the aortic flow probe and transpulmonary thermodilution with a pulmonary artery catheter would be regarded by most anaesthetists as excessively invasive for elective abdominal surgery.^1,34,35 We chose to compare ODM and NICOM, as trends in cardiac output can be useful to clinicians, despite concerns regarding interchangeability of absolute values.³⁶ Despite adjusting for repeated measures, the total number of measurements made during anaesthesia averaged 35 per patients, which may have introduced a degree of bias due to multiple measurement. When following trends, the precision of a technique assumes great importance, as the variability around a trend slope should be minimal when interpreting changes in stroke volume and cardiac output. We investigated the usefulness of NICOM in patients undergoing major abdominal surgery. This study population included open bowel surgery, urological surgery and aortic vascular surgery. With such heterogeneity, haemodynamic changes relating to different surgical interventions, particularly aortic surgery, may well be different despite each patient acting as their own control. It was usual for the anaesthetist to give an early fluid challenge, often before surgery had commenced. However, many of the readings were taken whilst other changes such as surgical incision, visceral manipulation or retraction, and alterations in anaesthetic drugs were taking place, sometimes simultaneously. It may be that these other changes produced changes in the quality of the signal rather than a true haemodynamic change. We discarded signals that were not deemed acceptable or seemed physiologically wrong. Both devices were unable to produce reliable signals during electrical diathermy.

In summary, the moderate agreement between NICOM Bioreactance and oesophageal Doppler would not be clinically acceptable. The measurements made by these two devices cannot be regarded as interchangeable. Both devices demonstrated similar directional trends in stroke volume and cardiac output following fluid challenge during abdominal surgery.

Acknowledgements

Assistance with the study: the authors would like to thank Dr M. Columb for statistical advice and Dr M. Nirmalan for reviewing drafts of the manuscript.

Financial support and sponsorship: the purchase of the NICOM monitor and disposables was entirely funded by the generosity of the Anaesthesia Research Society and was awarded following an open competitive application to the National Institute of Academic Anaesthesia Small Project Grant Round 2010.

Conflicts of interest: Dr Conway has received travel expenses from Deltex Medical Ltd to attend educational meetings. Dr Conway acted as an expert advisor to NICE Medical Technology Advisory Committee regarding oesophageal Doppler (MTG3) claiming no expenses or payment. Dr Gall and Dr Hussain have no conflicts of interest to declare.

Presentation: Preliminary results presented at Euroanaesthesia Congress, Paris, June 2012.