To compare non-invasive hemodynamic measurements obtained in pregnant and postpartum women using two automated cardiac output monitors against those obtained by two-dimensional (2D) transthoracic echocardiography (TTE).
To compare non-invasive hemodynamic measurements obtained in pregnant and postpartum women using two automated cardiac output monitors against those obtained by two-dimensional (2D) transthoracic echocardiography (TTE).
This was a cross-comparison study into which we recruited 114 healthy women, either with normal singleton pregnancy (across all three trimesters) or within 72 hours following delivery. Cardiac output estimations were obtained non-invasively using two different monitors, Ultrasound Cardiac Output Monitor (USCOM®, which uses continuous-wave Doppler analysis of transaortic blood flow) and Non-Invasive Cardiac Output Monitor (NICOM®, which uses thoracic bioreactance), and 2D-TTE. The performance of each monitor was assessed relative to that of TTE by calculating bias, precision, 95% limits of agreement and mean percentage difference (MPD). Intraobserver repeatability was assessed for both monitors and interobserver reproducibility was assessed for USCOM, NICOM being operator-independent.
Following exclusions due to poor-quality results of a monitor or TTE, or for medical reasons, our analysis included 98 women (29 in the first trimester, 25 in the second and 21 in the third, and 23 postpartum). For cardiac output estimation, when compared with TTE, USCOM had a bias ranging from 0.4 to 0.9 L/min. The MPD of USCOM was 29% in the third-trimester cohort. NICOM had a bias ranging from −1.0 to 0.6 L/min, with a MPD of 32% in the third-trimester group. There was limited agreement between the cardiac output monitors and TTE in the first and second trimesters, with a MPD of 38% for USCOM in both first and second trimesters, and 71% and 61% for NICOM in first and second trimesters, respectively. For cardiac output estimation using USCOM, we found excellent intraobserver repeatability (intraclass correlation coefficient (ICC), 0.97; 95% CI, 0.95–0.98) and interobserver reproducibility (ICC, 0.90; 95% CI, 0.81–0.94), and the repeatability for NICOM was comparable (ICC, 0.95; 95% CI, 0.93–0.97).
We found good agreement of both USCOM and NICOM when compared with 2D-TTE, specifically in the third trimester of pregnancy. Both devices had good intraobserver repeatability and either had good interobserver reproducibility or were operator-independent. Future studies should take into account the significant differences in the precise maternal hemodynamic values obtained by these devices, and consider developing device-specific reference ranges in pregnancy and the postpartum period. Copyright © 2016 ISUOG. Published by John Wiley & Sons Ltd.
Pregnancy is associated with significant changes in maternal hemodynamics, increasingly so as gestation advances. In pathological conditions, these changes can be even more profound and are of increased clinical significance, both during and following pregnancy[1-4]. In routine clinical practice, reliance is placed on maternal heart rate (HR) and brachial blood pressure as surrogate markers to provide information on cardiac indices such as maternal stroke volume (SV) and cardiac output (CO). Despite being relatively crude, these proxy markers are used routinely to guide clinical management and decisions regarding cardiovascular resuscitation and restoration of hemodynamic homeostasis. In previous decades, changes in maternal hemodynamics were investigated using dye dilution techniques and pulmonary artery catheterization (PAC). CO measurement with pulmonary artery catheterization using the bolus thermodilution method became the gold standard and the reference method against which to compare novel, non-invasive technologies[7, 8]. Such invasive techniques are undesirable in current obstetric practice due to the significant risk of complications such as infection, vascular injury and cardiac arrhythmias. Furthermore, validation of non-invasive methods in the obstetric population against such invasive techniques is neither practically nor ethically feasible.
Transthoracic echocardiography (TTE) is now a widely accepted methodology for CO estimation in pregnancy due to its non-invasive nature and absence of ionizing radiation; its use in expectant mothers is considered entirely safe and acceptable. It has been validated in pregnancy against thermodilution and dye dilution techniques, and has been reported to be an accurate method for CO estimation[9, 10]. However, access to TTE in labor requires both costly equipment and clinical expertise, thereby limiting its availability.
More recently, a plethora of non-invasive cardiac monitors have become readily available for clinical use, providing an opportunity to assess easily and accurately maternal hemodynamic status in the peripartum period. Two such commercially available devices are the UltraSound Cardiac Output Monitor (USCOM®, USCOM Ltd, Sydney, Australia), which uses continuous-wave Doppler analysis, and the Non-Invasive Cardiac Output Monitor (NICOM®, Cheetah Medical, Boston, MA, USA), which uses thoracic bioreactance analysis, to estimate hemodynamic indices in simple-to-use platforms. USCOM is an operator-dependent device, potentially subject to inter- and intraobserver variation. Proficient use of this device is associated with an individual learning curve and a requirement to attend a training session and carry out approximately 30 test cases before attempting to obtain data for clinical or research purposes. As obtaining a hemodynamic profile using USCOM requires access to the suprasternal notch (and some extension of the subject's neck), it is not practically feasible to use it for continuous hemodynamic evaluation (e.g. intraoperatively) or intrapartum. NICOM is entirely operator-independent and therefore not subject to interobserver variation. NICOM electrodes can be placed on the thorax and continuous hemodynamic variables can be obtained twice a minute; it can therefore be utilized during an operative procedure (e.g. during Cesarean section) with minimal disruption to the patient or the medical team. A drawback, however, of NICOM is the ongoing cost of consumables (skin surface electrodes).
The aim of this study was to compare hemodynamic measurements obtained in pregnant and postpartum women by these two automated CO monitors against those obtained by two-dimensional (2D) TTE.
Pregnant women aged 16 or over with a healthy, singleton pregnancy were recruited from various (booking, scanning and routine antenatal) clinics at our tertiary center. Women were recruited across all three trimesters and grouped accordingly, i.e. into first- (< 14 weeks), second- (14–26 weeks) and third-trimester (> 26 weeks) groups. In the postnatal group, all study participants were recruited within 72 hours of delivery. Women receiving antihypertensive medication, those with a known history of congenital or acquired heart disease and those with an incidental finding of a structural abnormality on echocardiography were excluded. Women who had HR > 100 bpm or mean arterial pressure > 125 mmHg were excluded. Written informed consent was obtained from all study participants and local research ethics committee approval (12/LO/0810) was sought prior to data collection. The study obstetrician examined all women and both maternal and fetal wellbeing were confirmed prior to obtaining any measurements.
All non-invasive monitor and TTE studies were performed in the same room, under standardized conditions and by the same operators for the entire cohort. A single operator (D.V.), following a 5-min period of patient inactivity, obtained non-invasive measurements from the two automated monitors simultaneously, immediately after which another operator (O.P.) performed the TTE examination. Both assessors were blinded to each other's recordings. Patients < 24 weeks' gestation were in a semi-recumbent position, and those ≥ 24 weeks were assessed in a left lateral position in order to avoid aortocaval compression by the gravid uterus.
USCOM employs continuous-wave Doppler, with a non-imaging probe placed in the suprasternal notch, to obtain velocity time integrals (VTI) of transaortic or transpulmonary blood flow, at the left or right ventricular outflow tract, respectively. Using an anthropometric algorithm, which correlates the outflow tract diameter with the patient's height, USCOM uses the VTIs to compute SV, CO and a complete hemodynamic profile. USCOM tracings were obtained in flowtracer mode (automated tracing of each Doppler profile), and the single operator analyzed all images, excluding poor-quality Doppler profiles. Each acquisition used for analysis had a minimum of two consecutive Doppler profiles (Figure 1). Prior to recruiting patients to the study, the operator received formal training in the use of USCOM and performed measurements in over 50 test cases.
All repeatability and reproducibility studies were performed in pregnant study participants (not in the postnatal cohort). Intraobserver repeatability of USCOM was assessed by each patient having two separate, consecutive Doppler acquisitions obtained within 5 min of one another. Interobserver reproducibility of USCOM was assessed with another set of Doppler acquisitions performed within the next 10 min by a third operator, a research midwife who had received training from the study investigator (D.V.) and formal training from an USCOM representative, and who had measured the recommended 30 test cases prior to collecting the data used for the reproducibility analysis. Operators were blinded to one another's recordings.
NICOM uses thoracic bioreactance technology, which is based on analysis of thoracic voltage amplitude changes in response to a high-frequency injected current. The NICOM device analyzes the variations in frequency spectra (relative phase shifts) after delivering a transthoracic alternating current and uses several assumptions (about thoracic shape and fluid volumes) and algorithms to compute SV, CO and an array of hemodynamic parameters. NICOM surface electrodes were placed on the participant's thorax (posterior aspect) prior to their lying supine or in the left lateral position, depending on gestational age. Following calibration, two separate complete sets of data were collected for each patient, simultaneously to USCOM readings. The first set of data was used for comparison analysis; the second was used for repeatability analysis.
Immediately following USCOM and NICOM readings, 2D-TTE was performed by a single, qualified (adult and pediatric/fetal) cardiac sonographer (O.P.), using a GE Vivid E9 (GE Healthcare, Little Chalfont, UK) ultrasound machine equipped with a single crystal phased-array M5Sc transducer. Participants were scanned in the left lateral position and multiple recordings of three views were obtained (parasternal long axis, apical four-chamber and apical five-chamber views). A three-lead electrocardiogram was used in order to enable image gating. Analysis was performed using dedicated software (EchoPAC, GE Healthcare). All measurements were performed according to recommendations of the American Society of Echocardiography.
After cardiac structural normality was confirmed, two methods of CO estimation by TTE were employed.
This method determines the SV as a product of the left ventricular outflow tract (LVOT) cross-sectional area (CSA) and the LVOT-VTI obtained by pulsed-wave (PW) Doppler. The diameter of the LVOT was measured in the parasternal long-axis view, at the level of the aortic valve annulus in early systole (Figure 2). The measurement was obtained from inner edge to inner edge of the aortic cusp insertion. Due to the inherent tendency of the tomographic plane to result in underestimation of the annulus diameter, we obtained three to five recordings of the aortic valve dimensions, and used their average in subsequent calculations. Assuming a circular shape of the LVOT cross-section, its area was calculated using the formula CSA = 0.785 × D2, where D is its diameter.
Measurements of the LVOT-VTI were obtained from the apical five-chamber view. A PW Doppler sample was positioned at the center of the LVOT, 3–5 mm proximal to the aortic valve (Figure 3). If the aortic valve opening click was present in the Doppler recording, the sample volume was withdrawn slightly into the outflow tract. The sample volume length was 2–5 mm, in order to narrow the spectral broadening of the PW Doppler signal. Care was taken to ensure that the ultrasound beam was parallel to the direction of blood flow in the LVOT. Multiple PW Doppler recordings were obtained in order to avoid underestimation of the LVOT velocities. The LVOT-VTI envelope was traced along the leading velocity (outer edge of the densest envelope) with at least three beats measured and averaged. The SV was calculated as LVOT-VTI × LVOT-CSA (Figure 3). HR was obtained by measuring the R-R interval using the electrocardiogram and then multiplying this interval by 60. CO was calculated using the formula: CO = SV × HR. Cardiac index was also calculated, by correcting CO for body surface area.
For the single-plane Simpson rule method, left ventricular end-diastolic volume (LVEDV) and end-systolic volume (LVESV) were calculated automatically by tracing the endocardial border in diastole and systole, respectively, in the apical four-chamber view (Figures S1 and S2). Volumes were calculated automatically by mathematically dividing the ventricle along its long axis into a series of discs of equal height. The ventricular volume was then represented by the sum of the volume of each of the discs. The ventricular SV was calculated as SV = LVEDV – LVESV. As for the LVOT-VTI method, CO was calculated using the formula: CO = SV × HR.
Statistical analysis was carried out using IBM SPSS Statistics version 21. The Shapiro–Wilk test was performed on all datasets to assess whether the data were distributed normally. We assessed the linearity of the relationship between the two methods using Pearson's (for normally distributed data) or Spearman's (non-normally distributed data) correlation coefficients. Accuracy and precision statistics included bias (mean difference between two methods), precision (SD of differences), 95% limits of agreement (LOA) (= bias ± 1.96 SD) and mean percentage difference (MPD) (= LOA/mean of two methodologies). Bland–Altman plots were constructed. Intraobserver repeatability and interobserver reproducibility were evaluated using intraclass correlation coefficient (ICC) with 95% CI, calculated using a two-way mixed-effects model. It has been proposed that a MPD of 30% between two methodologies is clinically acceptable for CO estimation.
Between January and October 2015, there were 114 participants recruited to the study (Figure 4). Following exclusions, there were 98 patients with complete datasets of measurements obtained by USCOM, NICOM and TTE included in the final analysis. Seven of the exclusions were for data of insufficient quality; in five cases this was due to poor views on TTE (predominantly in third-trimester women with increased body mass index (BMI)) and in two it was due to USCOM Doppler acquisitions that were unsuitable for analysis (in both cases because acquisitions were felt, upon quality control prior to analysis, not to represent the VTI obtained at the aortic valve). There were no failures to obtain hemodynamic variables using NICOM. Table 1 shows the basic demographic data of the study cohort.
|First trimester(n = 29)||Second trimester(n = 25)||Third trimester(n = 21)||Postpartum(n = 23)|
|Maternal age (years)||31 ± 6.3||33 ± 3.5||34 ± 5.5||30 ± 5.1|
|Nulliparous||18 (62)||23 (92)||7 (33)||N/A|
|Caucasian||22 (76)||21 (84)||18 (86)||13 (57)|
|Maternal weight at booking (kg)||65.3 ± 9.6||68.6 ± 14||70.4 ± 14||71.2 ± 15.9|
|Maternal BMI at booking (kg/m2)||23.1 ± 5.7||24.3 ± 7.3||26.3 ± 5.3||26.4 ± 5.4|
|Maternal MAP at booking (mmHg)||83 ± 18||85 ± 8||93 ± 20||86 ± 10|
|Maternal weight at assessment (kg)||65.5 ± 10||74.5 ± 15||85.4 ± 13.3||76.6 ± 16|
|Maternal BMI at assessment (kg/m2)||24.2 ± 3.8||27.3 ± 5.5||31.9 ± 4.9||28.3 ± 5.3|
|Maternal MAP at assessment (mmHg)||85 ± 19||84.3 ± 9||96.7 ± 8||89 ± 14|
HR correlation of each of the CO monitors with TTE was strong (USCOM: r = 0.866, P < 0.01; NICOM: r = 0.846, P < 0.01). The HR correlation between the two monitors was also strong (r = 0.920, P < 0.01). Comparison of USCOM relative to TTE showed a bias of 0.3 bpm, precision of 7 bpm, 95% LOA of −13 to +14 bpm and MPD of 17%. Comparison of NICOM relative to TTE revealed similar agreement, with a bias of 0.5 bpm, precision of 7 bpm, 95% LOA of −14 to +15 bpm and MPD of 17%. On comparison of USCOM relative to NICOM, the HR analysis of the monitors demonstrated a bias of −0.9 bpm, precision of 4 bpm, 95% LOA of −9 to +7 bpm and MPD of 10%.
Comparison of SV results obtained by USCOM and by TTE showed moderate correlation (r = 0.330, P < 0.05), with further analysis of USCOM relative to TTE demonstrating a bias of 11.4 mL, precision of 16.4 mL, 95% LOA from −20 to +43 mL and MPD of 45%. Comparison between NICOM and TTE showed weak correlation (r = 0.272, P < 0.05), with further analysis of NICOM relative to TTE revealing a bias of −1.9 mL, precision of 19.9 mL, 95% LOA from −40.8 to 36.9 mL and MPD of 60%.
The correlation coefficients and precision analyses for USCOM and NICOM compared with TTE using the LVOT-VTI method for CO estimation in each trimester and in the postpartum group are presented in Table 2. Equivalent results for TTE using the single-plane Simpson's method are presented in Table S1. Cardiac index findings are given in Tables S2 and S3. Bland–Altman plots are provided in Figures S3–S18.
|First trimester||Second trimester||Third trimester||Postpartum|
|Correlation coefficient vs TTE||0.366||0.364||0.569*||0.176|
|95% LOA (L/min)||−1.304; +2.779||−1.242; +3.034||−0.683; +2.407||−2.221; +3.099|
|Correlation coefficient vs TTE||0.048||−0.086||0.664*||0.610*|
|95% LOA (L/min)||−4.12; +2.18||−3.403; +2.696||−1.099; +2.261||−2.593; +2.575|
Intraobserver repeatability was assessed in 67 subjects, using three USCOM Doppler acquisitions of acceptable quality per patient and two complete NICOM profiles per patient. The ICC for CO estimation with USCOM was 0.97 (95% CI, 0.95–0.98) and that with NICOM was 0.95 (95% CI, 0.93–0.97). The interobserver reproducibility of USCOM was assessed in 40 subjects, across all three trimesters. ICCs varied from 0.90 (95% CI, 0.79–0.95) in the first trimester (27 cases) to 0.97 (95% CI, 0.88–0.99) in the second trimester (10 cases) and 0.96 (95% CI, 0.18–0.99) in the third trimester (three cases). The ICC for CO estimation across all three trimesters was 0.90 (95% CI, 0.81–0.94) (40 cases). As NICOM is a user-independent device, interobserver reproducibility was not assessed.
USCOM and NICOM demonstrated good agreement with TTE in the third trimester, with that between USCOM and TTE meeting the recommended level of clinical acceptability. The MPD for USCOM ranged from 28.8% in the third trimester to 43.8% postpartum; at earlier gestations it was 38%, comparable to previously published data in non-pregnant patients. NICOM had near-clinically acceptable agreement in the third trimester (MPD of 32%) and no agreement outside the third trimester.
Table 3 summarizes recent validation studies for USCOM and NICOM. A meta-analysis of USCOM validation studies in non-pregnant women reported a mean bias of −0.39 L/min, precision of 1.27 L/min, 95% LOA of −2.879 L/min to +2.099 L/min and MPD of 42.7%. In another study, comparing USCOM to three-dimensional TTE in advanced pregnancy, the authors reported a bias of +0.4 L/min, precision of 1.0 L/min, 95% LOA of −1.4 L/min to +2.3 L/min and MPD of 31.4%, comparable to the findings in our study. The level of agreement in the current study was higher than has been reported previously in the non-obstetric population.
NICOM in the third trimester had a MPD approaching 30%, showing the potential to be used intrapartum and at advanced gestations. NICOM has many positive attributes, including its simplicity of use, operator-independence and ability to provide continuous hemodynamic profiles in an intraoperative/intrapartum situation. The levels of agreement we observed would indicate that NICOM cannot be used interchangeably with TTE at earlier gestations or postnatally. Two previous NICOM validation studies[25, 26] displayed highly variable levels of accuracy and precision (Table 3). Early validation studies reported strong correlations (r = 0.84–0.90) between bioreactance and pulmonary artery catheterization-derived CO estimates; however, Bland–Altman analysis was lacking. Importantly, the differences between NICOM, USCOM and TTE are acceptable as long as different technology-specific reference ranges are employed.
|Reference||Year||Patient population||Reference method||Device||Bias (L/min or L/min/m2)||Precision (L/min or L/min/m2)||95% LOA (L/min or L/min/m2)||MPD (%)|
|Knirsch||2008||Pediatric cardiology||PAC therm||USCOM||−0.13||1.34||−1.47; +1.21||36.4|
|Tan||2005||ICU||PAC therm||USCOM||−0.18||0.82||−1.43; +1.78||35.7|
|Thom||2009||ICU||PAC therm||USCOM||−0.09||1.47||−3.01; +2.83||51.7|
|Wong||2008||Liver transplant||PAC therm||USCOM||−0.39||0.93||−1.47; +2.25||25.6|
|Kober||2013||Cytoreductive surgery in ovarian carcinoma||PAC therm||NICOM||0.26||0.85||−1.39; +1.92||50.7|
|Kupersztych-Hagege||2013||ICU||PAC therm||NICOM||−0.09||2.55||−2.20; +4.10||82.0|
Strengths and limitations of the various means of assessing CO should be considered. CSA, used to deduce SV for TTE and USCOM, can be obtained by different methodologies. The strength of using TTE is that the LVOT is measured in individual patients. USCOM employs an algorithm, based only on the patient's height, to provide the outflow tract diameter. Measurement of flow in the region of the aortic annulus, as performed by TTE, provides the highest accuracy, compared with other methods of estimating CO from the LVOT[15, 16]. Whilst the USCOM estimation provides an easy and reproducible method, enabling its bedside application, it does not factor in weight or body surface area, a potential source of error in CO estimation. Pregnancy is associated with significant changes in body surface area, which could be associated with changes in aortic dilatation[17-19].
TTE and USCOM utilize different Doppler modalities. PW Doppler, used by TTE, enables a recording of velocity at any specific point within the cardiac anatomy; this is in contrast to continuous-wave Doppler, used by USCOM, which cannot ascertain the precise location at which a velocity has been obtained. Training in the correct application of USCOM is important in order to obtain the transaortic VTI. Inaccurate recordings (e.g. recordings obtained instead in the ascending aorta) can be a significant source of error.
Finally, it is plausible that the hyperdynamic circulation will result in a certain amount of background noise, which will be picked up by the USCOM continuous-wave Doppler spectrum. This may result in ‘over-reading’ of the VTI, and hence produce a falsely elevated CO estimation compared with that provided by PW Doppler techniques. This may explain the positive bias in our obstetric population, as compared with the negative bias reported in the non-obstetric population.
Our data demonstrate that USCOM and NICOM show good agreement with TTE in advanced gestation. In clinical practice, their application will be in the management of critical illness in advanced pregnancy (e.g. septic shock, hemorrhage or pre-eclampsia). Both technologies have favorable characteristics that make them suitable for bedside application. They are simple to use, with measurements obtained within minutes by both medical and non-medical personnel, with minimal training required to achieve operative proficiency.
This cross-comparison study, particularly the results at earlier gestations, leads us to question the accuracy and precision of monitors such as USCOM and NICOM, and the interpretation of their findings, when used outside the third trimester; hence, thorough validation should be performed prior to interpreting published maternal hemodynamics data. Importantly, the indices obtained using such devices will not be directly comparable with those of TTE and hence device-specific reference ranges need to be constructed. One of the most relevant applications of this technology will be to display trends in hemodynamics. The inherent assumptions made by the machine algorithms should be of little relevance when assessing trends within a patient, as the ‘error’ is likely to remain constant. Future studies should evaluate the ability of these devices to assess accurately hemodynamic trends in both routine and pathological obstetric cases.
In conclusion, there is an unquestionable need for non-invasive, simple but accurate hemodynamic monitoring in the management of critically ill obstetric patients and for goal-directed therapy. However, the need and desire to utilize non-invasive technology should not be compromised by poorly reproducible, inaccurate or unvalidated measurements. Our findings suggest that both USCOM and NICOM perform well and have excellent repeatability and reproducibility in advanced pregnancy.
We would like to thank both USCOM and NICOM for the loan of their invaluable equipment in order for us to conduct vital research studies in this field. We would also like to thank Sophie Bowe, Research Associate, who participated in the interobserver analysis, and the research midwives and sonographers at St George's Hospital who helped us recruit our study participants.