Correlation, agreement and concordance of cardiac output estimated by transthoracic ultrasound and transesophageal Doppler with pulmonary artery thermodilution in anesthetized cats

Abstract

Objective: To characterize the correlation, agreement and concordance of cardiac output (CO) measured with transthoracic ultrasound and the correlation and concordance of aortic blood flow (ABF) minute distance (MD) measured by transesophageal Doppler with CO measured by pulmonary artery thermodilution (PATD) in cats.

Study design: Experimental study.

Animals: A group of six healthy male neutered cats, aged 2-8 years and weighing 5.3 ± 0.3 kg.

Methods: Cats were anesthetized with isoflurane in oxygen. CO was measured by PATD (CO_PATD) and transthoracic echocardiography (CO_ECHO). ABF MD was measured using an esophageal Doppler flow probe aligned with descending ABF. All measurements were made under three conditions: dexmedetomidine (20 μg kg^-1) intravenously; atipamezole (200 μg kg^-1) intramuscularly and atropine (20 μg kg^-1) intravenously as needed to achieve a minimum heart rate of 140 beats minute^-1; and dopamine (20 μg kg^-1 minute^-1) intravenously in that order. Correlation between CO_PATD and CO_ECHO, and CO_PATD and Doppler MD was evaluated using repeated measures correlation. Agreement between CO_PATD and CO_ECHO was evaluated using Bland-Altman method. Differences between consecutive pairs of CO measurements were calculated for concordance analysis.

Results: Correlation between CO_PATD and CO_ECHO and between CO_PATD and MD was significant (p < 0.001), with correlation coefficients greater than 0.92. A bias of > 27% and upper limits of agreement of 66% were found between CO_PATD and CO_ECHO. Concordance rate with CO_PATD was 76-80% for CO_ECHO and 72% for MD.

Conclusions and clinical relevance: Echocardiographic methods for the measurement of CO showed poor agreement and concordance with PATD. MD showed poor concordance with PATD. As such, these methods cannot be used as an alternative to PATD nor can they appropriately track changes in CO in anesthetized cats.

Keywords: Doppler; cardiac output; cats; echocardiography; thermodilution; ultrasound.

Figure 1

Representative recording of the velocity time waveform obtained by the esophageal Doppler monitor. The top arrows indicate identification by the software of the peak velocity and the lower arrows represent the flow time (duration of time of the flow from the left ventricle during systole). The actual signal is represented by the blurry filling and the lines delineating each waveform are traced by the software and used for calculation of stroke distance as the area under the curve (velocity time integral).

Figure 2

Illustration depicting echocardiographic images and assessment of cardiac output and fractional area change. Representative echocardiographic illustrations for the measurement of aortic cross-sectional area (left) and velocity time integral (VTI; right) in the right parasternal long-axis (a) and subcostal (b) views for the calculation of cardiac output (CO; mL minute⁻¹) is provided. The VTI was always obtained from the subcostal view. The position of the sample gate for measurement of VTI is illustrated in yellow. Illustrative echocardiographic images in the right parasternal short-axis view (c) depicting diastolic (left) and systolic (right) left ventricular (LV) area measurements for the calculation of LV fractional area change (FAC%) is presented.

Figure 3

Bland–Altman plots showing the agreement between cardiac output (CO) measured using pulmonary artery thermodilution (PATD) and ultrasound parasternal long-axis view (US AortPS, left) or ultrasound subcostal long-axis view (US AortSC, right). Cardiac output was measured in six cats (each represented by a different symbol) anesthetized with isoflurane in oxygen, following administration of dexmedetomidine [20 μg kg⁻¹, intravenously (IV)], then atipamezole (200 μg kg⁻¹ intramuscularly) and atropine (20 μg kg⁻¹ IV) as needed to produce a heart rate of 140 beats minute⁻¹ or higher, and lastly during IV administration of dopamine (20 μg kg⁻¹ minute⁻¹). The order of treatments was not randomized. Three CO measurements were obtained at 10 minutes intervals after each drug administration with each measurement technique. Limits of agreement are defined as 1.96 times the standard deviation of the differences in measurement between the two techniques. Bias is the average difference in CO between the two measurement techniques. CI, confidence interval; LLA, lower limit of agreement; ULA, upper limit of agreement.

Figure 4

Four quadrant plots showing the concordance between pulmonary artery thermodilution (PATD, considered to represent the true direction of change in cardiac output) and ultrasound parasternal long-axis view (US AortPS, top left), ultrasound subcostal long-axis view (US AortSC, top right), ultrasound fractional area change (US FAC) and transesophageal Doppler minute distance. Cardiac output (CO) was measured in six cats (each represented by a different symbol) anesthetized with isoflurane in oxygen, following administration of dexmedetomidine [20 μg kg⁻¹ intravenously (IV)], then atipamezole (200 μg kg⁻¹ intramuscularly) and atropine (20 μg kg⁻¹ IV) as needed to produce a heart rate of 140 beats minute⁻¹ or higher, and lastly during IV administration of dopamine (20 μg kg⁻¹ minute⁻¹). The order of treatments was not randomized. Three CO measurements were obtained at 10 minute intervals after each drug administration. The vertical and horizontal lines divide the plot in quadrants where the lower left quadrant contains data points where both methods showed an increase in CO, the upper right quadrant contains data points where both methods showed a decrease in CO, the upper left quadrant contains data points where the reference method showed an increase in CO and the alternative method showed a decrease in CO and the lower right quadrant contains data points where the reference method showed a decrease in CO and the alternative method showed an increase in CO. Measurements showing concordance are contained in the lower left and upper right quadrants. Data on the oblique line would represent equal changes with the two methods. The rectangle shows a 15% change exclusion (data contained in the rectangle were not used to calculate the rate of concordance as the magnitude of change is considered too small for meaningful interpretation). Δ: change. Negative Δ values reflect an increase in CO in pairs of successive measurements and positive Δ values reflect a decrease in CO in pairs of successive measurements.