A minimally invasive monitoring system of cardiac output using aortic flow velocity and peripheral arterial pressure profile.

BACKGROUND

In managing patients with unstable hemodynamics, monitoring cardiac output (CO) can provide critical diagnostic data. However, conventional CO measurements are invasive, intermittent, and/or inaccurate. The purpose of this study was to validate our newly developed CO monitoring system.

METHODS

This system automatically determines peak velocity of the ascending aortic flow using continuous-wave Doppler transthoracic echocardiography and estimates cardiac ejection time and aortic cross-sectional area using the pulse contour of the radial arterial pressure. These parameters are continuously processed to estimate CO (CO(est)). In 10 anesthetized closed-chest dogs instrumented with an aortic flowprobe to measure reference CO (CO(ref)), hemodynamic conditions were varied over wide ranges by infusing cardiovascular drugs or by random atrial pacing. Under each condition, CO(ref) and CO(est) were determined. Absolute changes of CO(ref) (ΔCOref) and CO(est) (ΔCO(est)), and relative changes of CO(ref) (%ΔCO(ref)) and CO(est) (%ΔCO(est)) from the corresponding baseline values were determined in each animal. We calibrated CO(est) against CO(ref) to obtain proportionally scaled CO(est) (CO(est)(N)).

RESULTS

A total of 1335 datasets of CO(ref) and CO(est) were obtained, in which CO(ref) ranged from 0.17 to 5.34 L/min. Bland-Altman analysis between CO(ref) and CO(est) indicated that the limits of agreement (the bias ± 1.96 × SD of the difference) and the percentage error (1.96 × [SD of the difference]/[mean CO] × 100) were from -1.01 to 1.13 L/min (95% confidence interval, -1.76 to 1.88 L/min) and 43%, respectively. The agreement between CO(ref) and CO(est)(N) was improved, with limits of agreement from -0.53 to 0.49 L/min (95% confidence interval, -0.62 to 0.59 L/min) and the percentage error of 20%. Polar plot analysis between ΔCO(ref) and ΔCO(est) indicated that mean ± 1.96 × SD of polar angle was -2° ± 22°. Four quadrant plot analysis indicated that %ΔCO(est) correlated tightly with %ΔCO(ref) (R(2) = 0.93). The %ΔCO(est) and %ΔCO(ref) changed in the same direction in 95% of the datasets. Reliability of this system was well preserved under conditions of random atrial pacing and also in a continuous manner.

CONCLUSION

Over a wide range of hemodynamic conditions, irrespective of cardiac beat irregularity, this system may allow minimally invasive monitoring of CO with a good trending ability. The present results warrant further research and development of this system for future clinical application.