Bias, trending ability and diagnostic performance of a non-calibrated multi-beat analysis continuous cardiac output monitor to identify fluid responsiveness in critically ill patients

Objective

To evaluate the accuracy of non-calibrated multi-beat analysis continuous cardiac output (CCO_MBA), against calibrated pulse-contour analysis continuous cardiac output (CCO_PCA) during a passive leg raise (PLR) and/or a fluid challenge (FC).

Design

Observational, single-centre, prospective study.

Setting

Tertiary academic medical intensive care unit, Lyon, France.

Participants

Adult patients receiving norepinephrine, monitored by CCO_PCA, and in which a PLR and/or a FC was indicated.

Main outcome measures

CCO_MBA and CCO_PCA were recorded prior to and during the PLR/FC to evaluate bias and evaluate changes in CCO_MBA and CCO_PCA (∆%CCO_MBA and ∆%CCO_PCA). Fluid responsiveness was identified by an increase >15% in calibrated cardiac output after FC, to identify the optimal ∆%CCO_MBA threshold during PLR to predict fluid responsiveness.

Results

29 patients (median age 68 [IQR: 57–74]) performed 28 PLR and 16 FC. The bias between methods increased with higher CCO_PCA values, with a percentage error of 64% (_95%confidence interval: 52%–77%). ∆%CCO_MBA adequately tracked changes in ∆%CCO_PCA with an angular bias of 2 ± 29°. ∆%CCO_MBA during PLR had an AUROC of 0.92 (P < 0.05), with an optimal threshold >14% to predict fluid responsiveness (sensitivity: 0.99, specificity: 0.87).

Conclusions

CCO_MBA showed a non-constant bias and a percentage error >30% against calibrated CCO_PCA, but an adequate ability to track changes in CCO_PCA and to predict fluid responsiveness.