Comparison of cardiac output of both 2 and 3 dimensional transoesophageal echocardiography with transpulmonary thermodilution during cardiac surgery

Abstract

Transpulmonary thermodilution (TD) is the most used method for cardiac output (CO) monitoring during cardiac surgery. Although 2D transoesophageal (TOE) measurement of CO using spectral Doppler correlates with TD, 3D TOE more accurately measures left ventricular outflow tract (LVOT) and aortic valve (AV) area, which are used to calculate CO. The aim of this study is to compare the precision of CO measurement between 2D and 3D TOE against TD. After ethics approval, 50 patients aged over 18 years scheduled for on-bypass cardiac surgery were recruited prospectively at two institutions. Exclusion criteria included more than mild valvular regurgitation, and atrial fibrillation. CO was measured simultaneously by TD and TOE before sternotomy and after cardiopulmonary bypass. CO was calculated using TOE by the product of either the LVOT or AV area, the velocity-time integral (VTI) of flow at the same site and heart rate. The LVOT area was assumed circular and calculated using the LVOT diameter for 2D but with planimetry using 3D TOE. The AV area was estimated with planimetry by both 2D and 3D TOE, and VTI with continuous wave Doppler. Both modal and outer edge traces of LVOT VTI were performed with the cursor 0.5 cm from the annulus. Measurements were averaged from 3 consecutive beats by 2 observers. Deming model II regression was used to assess fixed and proportional bias of agreement with TD. Bland-Altman technique was used to assess closeness of fit. CO was measured at 94 time-points (50 before sternotomy, 44 after chest closure) in 50 patients. The 3D methods had better agreement than 2D with TD, with the best agreement with 3D planimetry of the AV (bias -0.04 Lmin-1, SD of difference between the mean 1.37 Lmin-1), followed by 3D LVOT area planimetry (0.14, 1.41), 2D AV area planimetry (0.28, 1.3) and 2D LVOT VTI outer edge trace (-0.59, 1.29). 3D LVOT area planimetry had the best correlation (slope 1.39, 95%CI 0.97-1.82), followed by 2D AV planimetry (1.15, 0.81-1.48), 3D AV planimetry (1.07, 0.76-1.39), and 2D LVOT VTI outer edge trace (0.92, 0.33-1.52), all of which had no proportional or fixed bias. The VTI modal trace model had the worst agreement (-1.59, 1.37) and correlation (0.40, 0.08-0.69) and had proportional but not fixed bias. However the limits of agreement were similar amongst all methods. Tracing the edge of the LVOT VTI rather than the modal line reduced the bias (-0.59, 1.29), and improved correlation (0.92, 0.33-1.52). For 2D measurements, tracing around the LVOT VTI rather than through the modal line improved precision. The other 2D and 3D measurements of CO showed absence of fixed or proportional bias, and reasonable agreement with thermodilution as the reference method.