Continuous Estimation of Cardiac Output in Critical Care: A Noninvasive Method Based on Pulse Wave Transit Time Compared with Transpulmonary Thermodilution

Abstract

Purpose: Estimation of cardiac output (CO) and evaluation of change in CO as a result of therapeutic interventions are essential in critical care medicine. Whether noninvasive tools estimating CO, such as continuous cardiac output (esCCOTM) methods, are sufficiently accurate and precise to guide therapy needs further evaluation. We compared esCCOTM with an established method, namely, transpulmonary thermodilution (TPTD). Patients and Methods. In a single center mixed ICU, esCCOTM was compared with the TPTD method in 38 patients. The primary endpoint was accuracy and precision. The cardiac output was assessed by two investigators at baseline and after eight hours.

Results: In 38 critically ill patients, the two methods correlated significantly (r = 0.742). The Bland-Altman analysis showed a bias of 1.6 l/min with limits of agreement of -1.76 l/min and +4.98 l/min. The percentage error for CO_esCCO was 47%. The correlation of trends in cardiac output after eight hours was significant (r = 0.442), with a concordance of 74%. The performance of CO_esCCO could not be linked to the patient's condition.

Conclusion: The accuracy and precision of the esCCOTM method were not clinically acceptable for our critical patients. EsCCOTM also failed to reliably detect changes in cardiac output.

Figure 1

Comparison between CO_TPTD and CO_esCCO. (a) Correlation between CO_TPTD and CO_esCCO estimates at baseline. Each dot represents one patient. Correlation between the two estimates was significant (r = 0.742, p < 0.001). The regression coefficient was 0.52, and the intercept was 2.21 l/min (CO_esCCO = 0.52 × CO_TPTD + 2.21 l/min) indicating that at low CO, the esCCO was overestimated and at high cardiac output, underestimated CO as compared to the TPTD method. (b) The Bland–Altman plot of CO_TPTD and CO_esCCO, Each dot represents a pair of simultaneous cardiac output measurements by esCCO and TPTD of the same patient. The midhorizontal line marks the average difference between CO_TPTD and CO_esCCO (bias; 1.61 l/min). The upper and lower horizontal lines represent the 95% confidence interval of the difference between CO_TPTD and CO_esCCO (limits of agreement; −1.76 and +4.98 l/min). (c) The four quadrant plot of the correlation between ΔCO_esCCO and ΔCO_TPTD. Each dot represents the change of cardiac output over an 8-hour period (M_8 h − M_{mean baseline}) assessed by transpulmonary thermodilution (TPTD) and continuous cardiac output (esCCO). The regression fitted (ΔCO_esCCO = 0.35 × ΔCO_TPTD − 0.09 l/min) supported that a change in cardiac output overtime was underestimated by the esCCO as compared to the TPTD method.

Figure 2

Comparison between CO_TPTD and CO_esCCO in patients with sepsis (a–c) and patients with neither sepsis nor trauma (d–f). (a) In patients with sepsis, correlation between COTPTD and COesCCO estimates at baseline was significant (r = 0.813, p=0.001), regression coefficient was 0.59, and the intercept was 1.82 l/min (COesCCO = 0.59 × COTPTD + 1.82 l/min); (d) in patients with neither sepsis nor trauma, correlation was significant (r = 0.669, p=0.002), regression coefficient was 0.41, and the intercept was 2.75 l/min (COesCCO = 0.41 × COTPTD + 2.75 l/min). In patients with sepsis (b) comparing COTPTD and COesCCO yielded a bias of 1.52 l/min and limits of agreement of −2.08 and +5.12 l/min; in patients with neither sepsis nor trauma (e), bias was 1.77 l/min and limits of agreement of −1.81 and +5.34 l/min. Estimates of the change in CO over the first 8 h (M8 h – Mmean baseline) was analyzed by four quadrant plots with regressions fitted. In patients with sepsis (c), ΔCOesCCO = 0.24 × ΔCOTPTD – 0.27 l/min, and for patients with neither sepsis nor trauma (f), ΔCOesCCO = 0.35 × ΔCOTPTD + 0.02 l/min.