The accurate measurement of SV and CO has important applications in both clinical and research areas. Although CO provides an excellent global perspective of cardiovascular function, hemodynamic monitoring of critically ill patients has become increasingly more sophisticated, with higher standards of care and goal-directed therapy. Goal-directed therapy involves targeting adequate tissue oxygenation by modifying cardiac variables such as preload, afterload, and contractility. This has been associated with improved outcomes in people hospitalized for severe sepsis, and it has been advocated in veterinary medicine.1,2
Invasive pulmonary arterial catheterization to measure CO by use of thermodilution has become the de facto reference method for cardiac monitoring in human medicine.3 However, results of recent large trials have raised questions about the role of invasive catheterization because of associated morbidities, especially when less invasive methods are increasingly available.4 The overall incidence of adverse effects of pulmonary artery catheterization is relatively low in large-scale human studies, but additional concerns have been raised about the acquisition of reliable data and appropriate interpretation of information.4 Other minimally invasive alternatives are available (eg, lithium dilution or transpulmonary pulse contour analysis); however, both of those alternatives require systemic arterial and venous catheterization.5,6 Therefore, safe, reliable, accurate, and noninvasive alternatives for the measurement of CO are needed for clinical and research purposes.
The widespread availability and recent advances in ultrasound technology are reasons that echocardiography should be considered as a noninvasive hemodynamic monitoring tool. Transthoracic and transesophageal echocardiography have been used to estimate CO in humans,7 dogs,8 and horses9,10 via both pulsed-wave Doppler and non-Doppler echocardiography methods. Pulsed-wave Doppler echocardiography techniques measure velocities across the pulmonary or aortic valve to calculate the area under the curve, which is known as the velocity time integral. The velocity time integral provides stroke distance (ie, distance blood travels during 1 heartbeat), and stroke distance times area yields the SV. In addition, SV times HR yields CO.
Alternatively, non-Doppler echocardiography estimates of CO can be obtained by estimating EDV and ESV and calculating SV by use of the following equation: SV = EDV – ESV. The estimation of LV volumes with 1-D and 2-D echocardiography is based on equations that use geometric shape assumptions, most commonly the Teichholz formula and modified Simpson MOD, respectively. The Teichholz method assumes the LV has an ellipsoid shape and extrapolates volume from a single end-diastolic and end-systolic measurement of the internal dimensions of the minor axis of the LV from an M-mode image. In the MOD, the endocardial border of the LV is traced on a 2-D image, and the lumen is divided into slices and correspondent volumes. The MOD assumes the LV has a cylindrical configuration.
The recent development of matrix array transducers and RT3DE has enabled quantification of LV volumes by use of echocardiography without the need for geometric shape assumptions. Matrix array transducers have thousands of ultrasound elements that provide pyramidal volume information. Semiautomated and fully automated software for detection of the endocardial border produce a 3-D cast of the LV cavity and calculate LV volumes throughout the cardiac cycle. Studies, including meta-analyses, in human cardiology have revealed superior accuracy of RT3DE, compared with results for 2-D echocardiographic methods, for LV volume quantification.11,12
Imaging methods such as CMRI and MDCT also provide valuable insight into cardiac structures and function because both have excellent spatial resolution. Cardiac MRI is the current criterion-referenced standard for LV volume and ejection fraction in humans, with cardiac-gated CT considered a viable alternative for patients with mechanical prostheses or pacemakers. There are conflicting data regarding overestimation or underestimation of CT volumes, compared with results for CMRI; however, the measured differences typically are not significant.13 In veterinary medicine, the necessity for general anesthesia of patients, longer acquisition times, lack of availability of equipment, and greater costs associated with CMRI make MDCT a more practical option. Multidetector CT technology has advanced and has enabled clinicians and researchers to rapidly acquire cardiac images with high temporal resolution and create motion-free images of a beating heart. These images can be analyzed by use of postprocessing software to semiautomatically calculate EDV and ESV or to generate attenuation-time curves from the linear relationship between contrast material and x-ray attenuation in the main pulmonary artery or aorta.
Veterinary investigators have researched cardiac function by use of advanced imaging modalities. In 1 study,14 investigators evaluated EDV, ESV, and ejection fraction by use of 1-, 2-, and 3-D echocardiography, compared with results for CMRI, and found that echocardiographic variables significantly underestimated LV volumes unless the real-time triplane technique was used. Investigators of another study15 assessed various analyzing programs for real-time 3-D echocardiographic LV volume and function, compared with results for CMRI, and found that there was a small but systematic underestimation for the 3-D echocardiographic data, compared with the CMRI data. In yet another study,16 investigators evaluated LV geometry in dogs with degenerative valve disease by use of RT3DE methods, compared with results for 1- and 2-D methods, and found that there was significant LV volume expansion in dogs with advanced chronic degenerative valve disease. Cardiac function for a cohort of healthy dogs was evaluated by use of standard echocardiographic methods (1-D and 2-D) and compared with results for MDCT and CMRI, which revealed that LV volumes and systolic function were not significantly different between methods.17 This is in contrast with results for another study18 of healthy dogs in which investigators compared LV volumes determined by use of MDCT and CMRI and found that EDV was significantly greater for MDCT than for CMRI. Investigators of that study18 did not evaluate LV volumes by use of echocardiography.
The aforementioned studies were conducted to explore cardiac variables of function such as SV and ejection fraction; however, few veterinary comparative studies have been performed to specifically evaluate noninvasive measurement of CO. To the authors' knowledge, no studies have been performed to evaluate 3-D imaging modalities, including 3-D echocardiography or MDCT, to specifically assess CO in dogs. Therefore, the objective of the study reported here was to evaluate the accuracy of CO estimation by noninvasive methods (MDCT and 1-, 2-, and 3-D echocardiography), compared with results for invasively measured CO obtained by use of thermodilution. We hypothesized that echocardiographic CO determined by use of 2- and 3-D methods would systematically underestimate CO, compared with CO determined by use of MDCT and thermodilution. We also hypothesized that CO determined by use of a 1-D echocardiographic method would correlate poorly with CO obtained by use of thermodilution.
Materials and Methods
Animals
Six sexually intact (3 females and 3 males) purpose-bred hound-cross dogs were enrolled in the study. All dogs were 13 months old and weighed between 19.5 and 23.8 kg (median, 22.4 kg). The dogs had been purchased by another service of the veterinary teaching hospital at Oregon State University for use in an unrelated unpublished study; dogs were enrolled in the present study following conclusion of that unrelated study. Dogs were assessed as healthy on the basis of results of a physical examination, CBC, biochemical analysis, and complete echocardiographic examination. The dogs were housed at the laboratory animal center at Oregon State University. The study was approved by the Institutional Animal Care and Use Committee at Oregon State University.
Anesthetic protocol
All dogs were anesthetized for the ECG-gated contrast-enhanced MDCT, echocardiographic, and invasive transpulmonary artery thermodilution measurements. Dogs were sedated with butorphanol tartrate (0.1 to 0.2 mg/kg, IM). A catheter was then inserted into a cephalic vein, and anesthesia was induced with propofol (3 to 5 mg/kg, IV, titrated to effect). An endotracheal tube was placed, and anesthesia was maintained with isoflurane delivered in oxygen. Lactated Ringer solution (10 mL/kg/h, IV) was administered for the duration of anesthesia. To minimize possible bias caused by duration of anesthesia on ventricular function, 3 dogs were arbitrarily assigned to first undergo ECG-gated MDCT and then echocardiography, and the other 3 dogs were assigned to first undergo echocardiography followed by ECG-gated MDCT. Thermodilution was performed last in each dog.
Electrocardiogram-gated MDCT
Contrast-enhanced MDCT images were acquired by use of a 64-row detector CT scanner.a Dogs were placed in sternal recumbency with ECG electrodes attached to the forelimbs and left hind limb. The MDCT scans were obtained by use of the following scanning parameters: 0.5-mm collimation, 0.5-mm reconstruction interval, 512 × 512 matrix, 163 to 213 mm display field of view, pitch factor of 0.829, gantry rotation speed of 350 milliseconds, 120 kV, 200 to 500 mA, and tilt of 0°. The retrospective ECG-gated MDCT scan was performed with a 3-phase injection protocol for iodinated contrast mediumb administered by use of a dual-barrel power injector.c The 3-phase injection consisted of an initial fast flow rate of contrast medium (5 mL/s), a second slower flow rate of contrast medium (2 mL/s), and a final fast flow rate of sterile saline (0.9% NaCl) solution (5 mL/s); total volume of contrast agent was 1 mL/kg, and total volume of saline solution was 0.5 mL/kg. Automated bolus tracking was initiated when 180 Hounsfield units were detected in the ascending aorta. Tracking of the Hounsfield units in the aorta began 3 seconds after start of the IV injection of the iodinated contrast medium.
The MDCT images were transferred to a server for offline analysis on a workstation with LV functional analysis software.d Images were reconstructed in 10% increments of the R-R interval. Multiplanar reconstruction images were created to display standard long- and short-axis views of the LV. The functional analysis software displayed the 4-chamber, 2-chamber, and short-axis views of the LV (Figure 1). The 4- and 2-chamber axes were manually adjusted to the mitral valve annulus hinge points dorsally and LV apex ventrally. The region of interest was manually adjusted on the long-axis view to include the LV from the mitral valve annulus to the LV apex. The region of interest for the short-axis view was manually adjusted to ensure inclusion of the entire LV lumen. Papillary muscles were excluded from the LV volume by the analysis software. After adjustments were completed, the software computed the LV volume by use of the Simpson MOD for SV estimation at each 10% interval and displayed the EDV, ESV, SV, and ejection fraction. Calculation of CO was obtained by use of the following equation: CO (L/min) = SV (L/heartbeat) × HR (beats/min). The HR was automatically provided by the analysis software. All MDCT measurements were obtained separately by 2 observers (NLL and KFS).
Images of the 4-chamber plane (A), 2-chamber plane (B), and short-axis plane (C) and the LV endocardial cast (D) used for MDCT quantification of the LV (salmon-colored area).
Citation: American Journal of Veterinary Research 78, 7; 10.2460/ajvr.78.7.818
Echocardiography
A comprehensive echocardiographic examination was performed on each dog by use of a protocol that consisted of M-mode echocardiography, 2-D echocardiography, pulsed-wave Doppler echocardiography, and RT3DE with continuous ECG monitoring on an ultrasound unite (Figure 2). All examinations were performed by the same board-certified veterinary cardiologist (NLL). All echocardiographic data were obtained by use of a 3- to 8-MHz phased-array transducer or 1- to 5-MHz matrix transducer. Data were sent to an image storage server for offline analysis on a workstation with commercial software.f The CO was calculated by use of the following equation: CO (L/min) = SV (L/heartbeat) × HR (beats/min). The SV was calculated as EDV – ESV, and HR was recorded as the mean HR for the 5-beat cine loop or still images. Each echocardiographic measurement was obtained in triplicate, and the mean value was calculated for use in statistical analysis. For all echocardiographic modalities, end diastole was defined as the first frame with complete mitral valve closure at the beginning of the QRS complex, and end systole was defined as the last frame before mitral valve opening. All echocardiographic measurements were obtained separately by 2 observers (NLL and KFS).
Echocardiographic images used to assess SV from the right parasternal imaging window for 1-D (A), 2-D (B), and 3-D (C) echocardiography. A—An M-mode image with end-diastolic and end-systolic internal dimensions of the LV denoted for use in the Teichholz formula. Tick marks are at intervals of 10 mm. B–Images of the MOD with the largest diastolic (top) and smallest systolic (bottom) frames. Tick marks are at intervals of 10 mm. C—The 3-D quantification of the LV in the 4-chamber (upper left panel), 2-chamber (upper right panel), and short-axis (lower left panel) planes and the LV endocardial cast (lower right panel). Notice in the various panels the intersection of the 2-chamber (red box), short-axis (blue box), and 4-chamber (green box) views. L = Lateral.
Citation: American Journal of Veterinary Research 78, 7; 10.2460/ajvr.78.7.818
1-D echocardiography—The LV lumen was measured by use of M-mode in the right parasternal short-axis view at the level of the papillary muscles in accordance with conventional methods.19 End-diastolic and end-systolic measurements were obtained over 3 consecutive heart beats, and the mean value was calculated. The CO was estimated by use of the Teichholz formula as follows: SV = ([7 × LVIDd3]/[LVIDd + 2.4]) – ([7 × LVIDs3]/[LVIDs + 2.4]), where LVIDd is the LV internal dimension during diastole and LVIDs is the LV internal dimension during systole.
2-D echocardiography—For each 2-D imaging plane, a loop was recorded during 5 consecutive cardiac cycles while optimizing the field of view for the LV in the right parasternal long-axis and left apical imaging windows. Manual tracing of the LV endocardial border was performed from the lateral to septal mitral valve annulus and included the papillary muscles within the LV cavity. Length of the LV was measured from the midpoint of the mitral valve annulus to the LV endocardial apex. Both EDV and ESV were automatically calculated by the workstation software by use of the Simpson MOD.
Pulsed-wave Doppler echocardiography—Pulsed-wave Doppler echocardiography–based calculation of CO was assessed by use of data for both the pulmonic and aortic valves. The 2-D echocardiographic images of the pulmonic and aortic valves were obtained with the right parasternal short-axis view and the right parasternal long-axis 5-chamber view, respectively. Linear measurements for both valves were obtained at the valve annulus during systole by use of the trailing edge-to-leading edge method.20 Pulsed-wave spectral Doppler echocardiographic recordings were obtained for the pulmonic and aortic valves from the right parasternal short-axis view and left apical 5-chamber view, respectively. Investigators were careful to obtain parallel alignment with blood flow, and the Doppler gate was placed at the center of the valve annulus. The velocity time integral was acquired by tracing the spectral signal with electronic calipers. Cross-sectional areas of the pulmonary artery and aorta were calculated manually by use of the following equation: CSA = π × (D/2)2, where CSA is the cross-sectional area, and D is the diameter of the valve annulus. Three consecutive cardiac cycles were measured for each dog, and the mean value for HR was calculated. The SV was calculated by use of the following equation: SV = velocity time integral × CSA.
3-D echocardiography—Images were acquired over 4 consecutive cardiac cycles to produce a complete data set in wide-angle full-volume acquisition mode for both the right parasternal long-axis and left apical imaging windows. Acquisition was initiated by recognition of the R wave in the ECG. The entire LV was included in the 3-D pyramidal data set, and 4 wedge-shaped subvolumes were obtained for each complete cardiac cycle. Images obtained were stored digitally and analyzed offline on a workstation with commercial software.g Pyramidal data displayed in the 4-chamber, 2-chamber, and short-axis planes were adjusted manually to ensure orthogonal planes and visualization of the entire LV apex. Five reference points were identified (4 around the mitral valve annulus and 1 at the LV apex) at both end diastole and end systole. The software then automatically detected the endocardial border to create a cast of the LV throughout the cardiac cycle. Manual adjustments were made to this region of interest in each frame, and the EDV, ESV, SV, and ejection fraction were subsequently calculated automatically by the software.
Thermodilution
Dogs were placed in left lateral recumbency. A 7F introducer catheter was percutaneously placed in the right jugular vein of each dog by use of the modified Seldinger technique. A 7F Swan-Ganz catheter was inserted through the introducer catheter and advanced into the pulmonary artery with fluoroscopic guidance. The distal thermistor tip of the Swan-Ganz catheter was confirmed to be in the right or left pulmonary artery on the basis of results for fluoroscopy and by characteristic pressure waveforms. The proximal hole of the Swan-Ganz catheter was positioned in the right atrium. An aliquot (5 mL) of ice-cold sterile saline solution was injected via the proximal port of the Swan-Ganz catheter into the right atrium, and the catheter was attached to a workstation.h The CO was determined by use of the area-under-the-curve method (3 measurements with < 10% variation and minimal changes in HR), and then the mean CO was calculated. The HR was obtained from the simultaneous ECG recording for each measurement, and the mean HR was calculated.
Statistical analysis
Statistical analysis was performed with a commercially available software package.i Data were assessed for normality by use of the Kolmogorov-Smirnov normality test. All data were normally distributed and reported as mean ± SD. Agreement between CO determined by invasive and noninvasive methods was assessed by use of the methods of Bland and Altman.21 The mean difference of all methods against the CO for thermodilution was determined (defined as bias), and the SD of those differences was used to determine the limits of agreement (± 1.96•SD). The relationship between CO determined with invasive and noninvasive methods was analyzed by use of linear regression analysis, and correlation was assessed by use of the Pearson product-moment correlation coefficient. A 1-way ANOVA with Tukey multiple comparisons test was used to assess differences in HR as well as differences in CO obtained by use of thermodilution versus echocardiographic and MDCT methods. The mean coefficient of variation was used to evaluate interobserver variability by use of the following equation: coefficient of variation (percentage) = (SD of the measurements/mean of the measurements) × 100. Statistical significance was set at P < 0.05.
Results
The MDCT images were of excellent quality except for 1 dog in which the images were of diagnostic quality but had lower contrast enhancement of the blood volume of the heart, compared with images for the other ECG-gated MDCT examinations, which limited the precise evaluation of LV for CT measurements. This was secondary to late initiation of the MDCT scanner; therefore, the MDCT CO values for that dog were excluded from statistical analysis. Diagnostic-quality results of complete echocardiographic examinations were obtained for all dogs. The HR data were normally distributed, and HR did not differ significantly (P = 0.25) during MDCT, echocardiography, or thermodilution. Mean values for CO and HR obtained by use of MDCT, echocardiography, and thermodilution methods were determined (Table 1). There were no significant differences in CO and HR between noninvasive methods and thermodilution. Teichholz (1-D echocardiography) and Doppler echocardiography–based methods of estimating CO had higher variance, compared with variance for the non-Doppler echocardiography methods. Overall, CO values derived by use of MDCT, 2-D echocardiography, and 3-D echocardiography were quantitatively less than, but not significantly different from, CO derived by use of thermodilution.
Mean, SD, and range of CO and HR obtained by use of MDCT, echocardiography, and thermodilution methods for 6 healthy dogs.
| CO (L/min) | HR (beats/min) | |||
|---|---|---|---|---|
| Measurement technique | Mean ± SD | Range | Mean ± SD | Range |
| MDCT | 1.60 ± 0.30 | 1.20–2.00 | 69.8 ± 18.1 | 50–100 |
| Teichholz | 2.13 ± 0.77 | 1.46–3.39 | 74.3 ± 20.4 | 49–96 |
| 2-D RLA MOD | 1.59 ± 0.31 | 1.09–1.93 | 71.8 ± 24.5 | 37–97 |
| 2-D LAP MOD | 1.51 ± 0.24 | 1.26–1.85 | 68.3 ± 18.4 | 45–94 |
| Aorta Doppler | 2.10 ± 0.66 | 1.59–3.37 | 82.2 ± 28.2 | 46–130 |
| PA Doppler | 2.07 ± 0.51 | 1.39–2.55 | 76.8 ± 20.8 | 52–97 |
| 3-D RLA | 1.94 ± 0.42 | 1.46–2.55 | 82.9 ± 16.7 | 58.67–98.33 |
| 3-D LAP | 1.70 ± 0.35 | 1.20–2.07 | 70.2 ± 17.0 | 47.67–94.00 |
| Thermodilution | 2.17 ± 0.36 | 1.73–2.60 | 63.8 ± 16.8 | 44–91 |
LAP = Left apical view. PA = Pulmonary artery. RLA = Right parasternal long-axis view.
Bland-Altman plots and linear regression graphs were created for CO measured by use of MDCT, echocardiography, and thermodilution (Figures 3 and 4). Evaluation of the Bland-Altman plots indicated a systematic underestimation of CO values derived from MDCT, 2-D echocardiography, and 3-D echocardiography measurements. Also, there were relatively wide limits of agreement for CO obtained by use of the Teichholz and Doppler echocardiography methods. The MDCT CO values were moderately correlated (r2 = 0.64; P = 0.01) with CO derived by use of thermodilution. A strong correlation was observed between CO determined by use of thermodilution and CO measured by use of 2-D echocardiography MOD (r2 = 0.97; P = 0.01) from the left apical window as well as 2-D echocardiography MOD (r2 = 0.81; P = 0.01) and 3-D echocardiography (r2 = 0.85; P = 0.01) from the right parasternal long-axis window. Conversely, a weak correlation (r2 = 0.46; P = 0.14) was found between CO determined by use of thermodilution and CO determined by use of 3-D echocardiography from the left apical window. There was a weak relationship between CO derived by use of thermodilution and Doppler echocardiography–based CO obtained by aortic (r2 = 0.03; P = 0.75) or pulmonic (r2 = 0.02; P = 0.78) outflow measurements. Estimation of CO by use of the Teichholz formula was not significantly correlated (r2 = 0.01; P = 0.91) with CO determined by use of thermodilution.
Bland-Altman plots (A, C, E, and G) and the corresponding linear regression graphs (B, D, F, and H) for CO obtained by use of MDCT (A and B), the Teichholz formula (C and D), 2-D right parasternal long-axis (RLA) MOD (E and F), and 2-D left apical (LAP) MOD (G and H), compared with CO determined by use of thermodilution (TD). Each circle represents results for 1 dog. For the Bland-Altman plots, the mean bias (dashed lines) and limits of agreement (bias ± 1.96•SD; dotted lines) are indicated. For the linear regression graphs, the regression line (solid lines) and confidence limits (dashed lines) are indicated.
Citation: American Journal of Veterinary Research 78, 7; 10.2460/ajvr.78.7.818
Bland-Altman plots (A, C, E, and G) and the corresponding linear regression graphs (B, D, F, and H) for CO obtained by use of aorta Doppler echocardiography (A and B), pulmonary artery (PA) Doppler echocardiography (C and D), 3-D echocardiography from the right parasternal long-axis (RLA) window (E and F), and 3-D echocardiography from the left apical (LAP) window (G and H), compared with CO determined by use of thermodilution (TD). See Figure 3 for remainder of key.
Citation: American Journal of Veterinary Research 78, 7; 10.2460/ajvr.78.7.818
Interobserver variability for the MDCT and echocardiographic estimations of CO was assessed (Table 2). There was acceptable interobserver variability for all methods, except for volumes obtained by use of 3-D echocardiography from the left apical window.
Interobserver variability for estimations of CO determined by use of MDCT and echocardiography for 6 healthy dogs.
| Measurement technique | Coefficient of variation (%)* |
|---|---|
| MDCT | 8.91 |
| Teichholz | 9.20 |
| 2-D RLA MOD | 10.61 |
| 2-D LAP MOD | 5.38 |
| Aorta Doppler | 8.94 |
| PA Doppler | 9.75 |
| 3-D RLA | 10.27 |
| 3-D LAP | 14.26 |
Coefficient of variation was calculated as follows: (SD of the measurements/mean of the measurements) × 100.
Discussion
Results of the study reported here supported the hypothesis that MDCT, 2-D echocardiography, and 3-D echocardiography can be used to estimate CO as an alternative to CO obtained by use of invasive thermodilution. With regard to the modalities evaluated, results for the present study suggested that the 2-D and 3-D echocardiographic CO obtained from the right parasternal long-axis window had little systematic bias with narrow limits of agreement and acceptable repeatability between observers. We suspect this was attributable to the good spatial and excellent temporal resolution afforded by imaging the LV from the right parasternal long-axis window, coupled with the lack of requisite geometric modeling by use of the 3-D echocardiographic methods. The 2-D and 3-D data from the left apical window also had little bias and relatively narrow limits of agreement, although interobserver variability and the correlation between CO derived from the left apical 3-D data and CO derived by use of thermodilution were suboptimal.
Analysis of the data also suggested that both the Teichholz and Doppler echocardiography–based methods for estimating CO had wide limits of agreement, which would limit the use of such methods in clinical or research settings. The wide limits of agreement for the Teichholz and Doppler echocardiography–based methods are consistent with those for a study9 of horses. Furthermore, as of 2015, the American Society of Echocardiography no longer recommends measuring LV volumes of people via the Teichholz method.20 The authors suggest that on the basis of data for the study reported here, the Teichholz method may also have limited applicability for use in echocardiographic determination of CO in veterinary patients.
Overall, few veterinary studies5,10,22 have been conducted to compare echocardiographic determination of CO with measurement of CO by use of invasive methods, and most of the existing research has been focused on equids. The results of that research are not necessarily relevant to dogs because there is a species-specific challenge for horses attributable to the difficulties of imaging the entire LV in a single window and parallel alignment with the aorta or pulmonary artery blood flow when obtaining Doppler echocardiography measurements. Few studies8,23 on CO measurement in small animals have been published; moreover, those studies were published > 10 years ago, which is relevant because of the important advances in 2- and 3-D echocardiography that have occurred during the past 10 years.
Findings for the present study also supported a moderate correlation between MDCT CO and CO determined by use of thermodilution, although MDCT CO values were overall less than those obtained by use of thermodilution. Studies24,25 that involved the use of conventional CT in dogs revealed a strong linear correlation between CT-derived and thermodilution-derived CO. In humans, there is a strong correlation between end-diastolic and end-systolic LV volumes for 16- to 128-slice MDCT, 2-D echocardiography, and 3-D echocardiography.13,26–28 Results of studies27,28 suggest a slight underestimation of 2-D and 3-D echocardiographic measurements of LV volumes, compared with CT-based LV volumes. This relationship was not supported by the results of the present study, for which SV determined by use of thermodilution was closest to 3-D echocardiographic SV, followed by 2-D echocardiographic and MDCT volumes. One possible reason for the lower MDCT volumes may have been related to differences in measurement techniques. The papillary muscles were excluded from LV volume measurement by the MDCT software, whereas the papillary muscles were included for echocardiography. Alternatively, this finding may have been attributable to the small sample size, which prevented the authors from drawing absolute conclusions and warrants investigation with a larger data set.
Results of the present study also suggested that the interobserver variability was acceptable for all methods, except 3-D volumes obtained from the left apical window. The suboptimal reproducibility could have represented a statistical error attributable to the limited number of data points. Alternatively, this finding may have been attributable to the fact that ideal alignment of the LV from the left parasternal window can be technically challenging in dogs, which, combined with a low frame rate during 3-D echocardiography, may have resulted in observers being poorly able to discern the LV border for 3-D analysis. Interestingly, there was nominal variability between observers for 2-D–based measurements from the left apical window. An explanation for this discrepancy could have been the higher frame rate afforded by 2-D echocardiography, whereby the LV endocardial border can be more readily estimated when more frames are available to track the endomyocardium despite limitations in image quality.
Although thermodilution is the most commonly used method in human and veterinary medicine, there is no universally accepted criterion-referenced standard for CO determination. Thermodilution is a practical criterion-referenced standard used for comparison of newer methods for determination of CO.6 Nonetheless, thermodilution has an inherent error rate of 10% to 20%.29 It has been documented that CO measurements obtained by use of thermodilution can differ considerably, even in hemodynamically stable patients.30 Factors that can contribute to this variability include indicator error (eg, temperature of injectate, amount of injectate, and catheter dead space), errors in signal processing in the computer system, and cyclic changes in right ventricular output associated with venous return and respiration.30,31 Investigators of a previous study31 found no significant difference between CO obtained at end respiration or end inspiration; however, the variability for CO values obtained at random phases of the respiratory cycle was always > 10%.31 Measurement of CO in humans by use of thermodilution has a bias of ± 0.48 L/min with limits of agreement of approximately 1 L/min or ± 20%.32 Therefore, it has been proposed that the limits of agreement for alternative methods of CO estimation should ideally be < 30% to reflect a similar intrinsic error to the current standard.33
The study reported here had several limitations. Probably the most important limitation was the logistical inability to measure CO by use of thermodilution and echocardiography simultaneously. Cardiac output is affected by rapid changes in neuroendocrine control, and method agreement may have been weakened by asynchronous measurement. Stroke output can differ as much as 50%, depending on the phase of the respiratory cycle.30,31 We attempted to mitigate these differences with a stable anesthetic depth, similar loading conditions, stable HR, and determination of the mean value for several measurements with < 10% variation. Furthermore, the postprocessing software we used did not have calipers for individual ECG intervals, so the HR data from the echocardiographic portion of the study was a mean of 5 cardiac cycle cine loops. Also, thermodilution measures pulmonary blood flow, which is not necessarily the same as aortic blood flow, even in the absence of a shunting lesion. This is attributable to phasic changes associated with respiration with right-sided preload that is less pronounced on left-sided chambers. The ability to accurately measure right ventricular CO may yield an even better correlation with CO measured by use of thermodilution. Also, anesthesia in the present study was induced with propofol and maintained with an inhalation gas that has known cardiac depressant effects, which would preclude translation of the data to awake unsedated dogs. Another limitation was the small number of dog without major variations in age or body weight, which did not allow us to make generalized definitive conclusions. The methods for data collection and analysis require verification in a larger prospective study to delineate the exact benefits and constraints of these methods.
Results of the present study suggested certain echocardiographic techniques available for CO estimation were comparable to invasively measured CO obtained by use of transpulmonary artery thermodilution. Specifically, the 2-D and 3-D echocardiographic measurements from the right parasternal long-axis window and the 2-D echocardiographic measurements from the left apical window had strong agreement with CO determined by use of thermodilution, acceptable repeatability, and little variance. The MDCT method also had moderate agreement with thermodilution. Both Doppler echocardiographic methods and Teichholz estimation of CO were poorly correlated with thermodilution and had unacceptable variance. The correlation between methods indicated a close relationship between selected techniques, although they cannot be used interchangeably. Cardiac output was likely slightly underestimated by use of the noninvasive techniques, compared with CO determined by use of thermodilution. The differences between thermodilution and noninvasive methods may have been attributable to errors of the measurement method, subject SV variability, HR fluctuation, or inaccuracies of thermodilution.
Acknowledgments
Supported by research funds allocated by the Department of Clinical Sciences, College of Veterinary Medicine, Oregon State University.
Presented as an oral presentation at the American College of Veterinary Internal Medicine Forum, Denver, June 2016.
The authors thank Darcy Palmer, Amy Berry, Robyn Panico, Cynthia Viramontes, and Jason Wiest for technical assistance.
ABBREVIATIONS
| CMRI | Cardiac MRI |
| CO | Cardiac output |
| EDV | End-diastolic volume |
| ESV | End-systolic volume |
| HR | Heart rate |
| LV | Left ventricle |
| MDCT | Multidetector CT |
| MOD | Method of disks |
| RT3DE | Real-time 3-D echocardiography |
| SV | Stroke volume |
Footnotes
Toshiba Aquilion 64 CT, Toshiba America Medical Systems Inc, Tustin, Calif.
Isovue 370, Empower CTA, Bracco Diagnostics Inc, Princeton, NJ.
Empower CTA, Bracco Diagnostics Inc, Princeton, NJ.
Vitrea workstation, software version 6.7.4, Vital Images Inc, Minnetonka, Minn.
Philips Medical Systems, Andover, Mass.
Excelera, Philips Medical Systems, Andover, Mass.
Excelera, QLAB, 3DQ-Advanced, Philips Medical Systems, Andover, Mass.
Mac-Lab, General Electric Medical Systems, Little Chalfont, Buckinghamshire, England.
GraphPad Prism, version 6.0h, GraphPad Software Inc, La Jolla, Calif.
References
1. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med 2001; 345: 1368–1377.
2. Butler AL. Goal-directed therapy in small animal critical illness. Vet Clin North Am Small Anim Pract 2011; 41: 817–838.
3. Reuter DA, Huang C, Edrich T, et al. Cardiac output monitoring using indicator-dilution techniques: basics, limits, and perspectives. Anesth Analg 2010; 110: 799–811.
4. Bernard GR, Sopko G, Cerra F, et al. Pulmonary artery catheterization and clinical outcomes: National Heart, Lung, and Blood Institute and Food and Drug Administration Workshop Report. Consensus statement. JAMA 2000; 283: 2568–2572.
5. Shih A, Giguere S, Sanchez LC, et al. Determination of cardiac output in neonatal foals by ultrasound velocity dilution and its comparison to the lithium dilution method. J Vet Emerg Crit Care (San Antonio) 2009; 19: 438–443.
6. Pugsley J, Lerner AB. Cardiac output monitoring: is there a gold standard and how do the newer technologies compare? Semin Cardiothorac Vasc Anesth 2010; 14: 274–282.
7. Brown JM. Use of echocardiography for hemodynamic monitoring. Crit Care Med 2002; 30: 1361–1364.
8. Uehara Y, Koga M, Takahashi M. Determination of cardiac output by echocardiography. J Vet Med Sci 1995; 57: 401–407.
9. McConachie E, Barton MH, Rapoport G, et al. Doppler and volumetric echocardiographic methods for cardiac output measurement in standing adult horses. J Vet Intern Med 2013; 27: 324–330.
10. Young LE, Blissitt KJ, Bartram DH, et al. Measurement of cardiac output by transoesophageal Doppler echocardiography in anaesthetized horses: comparison with thermodilution. Br J Anaesth 1996; 77: 773–780.
11. Dorosz JL, Lezotte DC, Weitzenkamp DA, et al. Performance of 3-dimensional echocardiography in measuring left ventricular volumes and ejection fraction: a systematic review and meta-analysis. J Am Coll Cardiol 2012; 59: 1799–1808.
12. Jenkins C, Bricknell K, Hanekom L, et al. Reproducibility and accuracy of echocardiographic measurements of left ventricular parameters using real-time three-dimensional echocardiography. J Am Coll Cardiol 2004; 44: 878–886.
13. Salm LP, Schuijf JD, de Roos A, et al. Global and regional left ventricular function assessment with 16-detector row CT: comparison with echocardiography and cardiovascular magnetic resonance. Eur J Echocardiogr 2006; 7: 308–314.
14. Meyer J, Wefstaedt P, Dziallas P, et al. Assessment of left ventricular volumes by use of one-, two-, and three-dimensional echocardiography versus magnetic resonance imaging in healthy dogs. Am J Vet Res 2013; 74: 1223–1230.
15. Eskofier J, Wefstaedt P, Beyerbach M, et al. Quantification of left ventricular volumes and function in anesthetized Beagles using real-time three-dimensional echocardiography: 4D-TomTec analysis versus 4D-AutLVQ analysis in comparison with cardiac magnetic resonance imaging. BMC Vet Res 2015; 11: 260.
16. Ljungvall I, Hoglund K, Carnabuci C, et al. Assessment of global and regional left ventricular volume and shape by real-time 3-dimensional echocardiography in dogs with myxomatous mitral valve disease. J Vet Intern Med 2011; 25: 1036–1043.
17. Drees R, Johnson RA, Stepien RL, et al. Quantitative planar and volumetric cardiac measurements using 64 MDCT and 3T MRI vs. standard 2D and M-mode echocardiography: does anesthetic protocol matter? Vet Radiol Ultrasound 2015; 56: 638–657.
18. Sieslack AK, Dziallas P, Nolte I, et al. Comparative assessment of left ventricular function variables determined via cardiac computed tomography and cardiac magnetic resonance imaging in dogs. Am J Vet Res 2013; 74: 990–998.
19. Teichholz LE, Kreulen T, Herman M V, et al. Problems in echocardiographic volume determinations: echocardiographic-angiographic correlations in the presence of absence of asynergy. Am J Cardiol 1976; 37: 7–11.
20. Lang RM, Bierig M, Devereux RB, et al. Recommendations for chamber quantification: a report from the American Society of Echocardiography's Guidelines and Standards Committee and the Chamber Quantification Writing Group, developed in conjunction with the European Association of Echocardiography, a branch of the European Society of Cardiology. J Am Soc Echocardiogr 2005; 18: 1440–1463.
21. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999; 8: 135–160.
22. Shih A. Cardiac output monitoring in horses. Vet Clin North Am Equine Pract 2013; 29: 155–167.
23. Gueret P, Meerbaum S, Zwehl W, et al. Two-dimensional echocardiographic assessment of left ventricular stroke volume: experimental correlation with thermodilution and cineangiography in normal and ischemic states. Cathet Cardiovasc Diagn 1981; 7: 247–258.
24. Herfkens RJ, Axel L, Lipton MJ, et al. Measurement of cardiac output by computed transmission tomography. Invest Radiol 1982; 17: 550–553.
25. Garrett JS, Lanzer P, Jaschke W, et al. Measurement of cardiac output by cine computed tomography. Am J Cardiol 1985; 56: 657–661.
26. Vieira ML, Nomura CH, Tranchesi B Jr, et al. Real-time three-dimensional echocardiographic left ventricular systolic assessment: side-by-side comparison with 64-slice multi-detector cardiac computed tomography. Eur J Echocardiogr 2010; 11: 257–263.
27. Sugeng L, Mor-Avi V, Weinert L, et al. Quantitative assessment of left ventricular size and function: side-by-side comparison of real-time three-dimensional echocardiography and computed tomography with magnetic resonance reference. Circulation 2006; 114: 654–661.
28. Bansal D, Singh RM, Sarkar M, et al. Assessment of left ventricular function: comparison of cardiac multidetector-row computed tomography with two-dimension standard echocardiography for assessment of left ventricular function. Int J Cardiovasc Imaging 2008; 24: 317–325.
29. Stetz CW, Miller RG, Kelly GE, et al. Reliability of the thermodilution method in the determination of cardiac output in clinical practice. Am Rev Respir Dis 1982; 126: 1001–1004.
30. Synder JV, Powner DJ. Effects of mechanical ventilation on the measurement of cardiac output by thermodilution. Crit Care Med 1982; 10: 677–682.
31. Stevens JH, Raffin TA, Mihm FG, et al. Thermodilution cardiac output measurement. Effects of the respiratory cycle on its reproducibility. JAMA 1985; 253: 2240–2242.
32. Mackenzie JD, Haites NE, Rawles JM. Method of assessing the reproducibility of blood flow measurement: factors influencing the performance of thermodilution cardiac output computers. Br Heart J 1986; 55: 14–24.
33. 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.
