Invasive versus non-invasive assessment of valvuloarterial impedance in severe aortic stenosis

Figures

• 

  Figure 1. A scatter plot of ZVA measurements. (A) Fick versus TTE, (B) TD versus TTE, and (C) TD versus Fick. If agreement between methods was strong, the measurements would be clustered tightly along the diagonal (red line). (A) TTE overestimates ZVA when compared with Fick (points are mostly above the diagonal) and (B) underestimates ZVA when compared with TD (points are mostly below the diagonal). The spread of the scatter indicates that the variability between the TTE and invasive methods (A, B) is much larger than between the BHC methods (C). For risk rule ZVA >4.5 (dashed lines), Fick tends to underestimate the risk relative to TTE, and TD tends to overestimate the risk relative to both TTE and Fick; see shaded points off in the upper and lower off-diagonal quadrants, respectively. The r number in each plot is the estimated correlation coefficient. BHC, bilateral heart catheterisation; TD, thermodilution; TTE, transthoracic echocardiography; ZVA, valvuloarterial impedance.

• 

  Figure 2. Tukey’s mean difference plots for (A) TTE versus Fick, (B) TTE versus TD, and (C) TD versus Fick with estimated mean differences (red lines) and their variability (dashed lines; L and U correspond to mean difference ±2 SD). Strong agreement between the methods would be characterised by small mean difference and small variability. Here, the mean differences for TTE versus BHC methods (A, B) are much smaller than between Fick and TD (C); however, the latter differences between are less variable than the former. BHC, bilateral heart catheterisation; TD, thermodilution; TTE, transthoracic echocardiography; ZVA, valvuloarterial impedance.

• 

  Figure 3. (A) Kappa statistic as a function of threshold z0 for risk stratification rule of ZVA>z0. Small kappa values indicate poor agreement in risk classification between TTE and BHC over the range of threshold values. Notably the kappa values are higher for classifications based on BHC methods. (B) Brier scores as a function of z0 for classification models to identify high-risk patients per TTE from BHC measurements (red and green colours); the lower scores the higher the agreement between classifications. The scores were calculated for 31 equidistant threshold values z0 between 3.0 and 6.0. For each z0, patients at risk were identified as ZVA>z0, separately from TTE and BHC, and the scores were computed for the resulting classifications as detailed in the methods. The score values are plotted against the corresponding thresholds with a local polynomial overlay for each model. The results indicate poor agreement between risk stratifications from BHC and TTE. In contrast, recovering TD risk classification from Fick measurements appears to be more accurate (blue colours). BHC, bilateral heart catheterisation; TD, thermodilution; TTE, transthoracic echocardiography; ZVA, valvuloarterial…

• 

  Figure 4. Kappa statistic as a function of threshold z0 for risk stratification rule of ZVA>z0 showing the effect of SAP and CO on classification agreement. Classification statistics based on ZVATTE and ZVA based on BHC are shown in black. Using SAP on BHC does not seem to improve the classification agreement relative to Fick and shows some improvement for z0 <3.5 for TD; see kappas and the smoother plotted in blue. Using CO based on BHC improves the agreement in risk classifications, with kappa values being above 0.5 for 75% of threshold cut-offs. BHC, bilateral heart catheterisation; CO, cardiac output; SAP, systolic aortic pressure; SVI, stroke volume index; TD, thermodilution; TTE, transthoracic echocardiography; ZVA, valvuloarterial impedance.