Assessing the measurement error of different methods used to calculate Pulse Pressure Variation

Abstract

Objective: To assess the measurement error of different methods used to calculate Pulse Pressure Variation.

Background: Many studies have demonstrated the use of pulse pressure variation (PPV) as a predictor of fluid responsiveness as long as the limitations to its use are understood and respected. These limitations have proven a constraint in the use of PPV and various researchers have published methods of overcoming these constraints in daily practice. Different methods also exist to calculate PPV. This study aims to systematically calculate the measurement error of the different methods used to calculate PPV and compare them.

Methods: After approval of the institutional trial board and ethics committee of Ghent University Hospital, Ghent, Belgium, and registration with the local code number B670201629642 (intern:EC/2016/1113), postoperative measurements of invasive arterial pressure and ECG were simultaneously recorded over 1 minute between 29/11/2016 to 16/11/2018. Data was then analyzed using different methods, namely (i) individual PPV averaged over fixed number of respiratory cycles (iPPV family), (ii) pooled PPV over fixed number of respiratory cycles (pPPV family) and (iii) methods over fixed window in terms of time (Aboy and Lansdorp).The Taffe extension of the Bland Altman method was used to compare and determine the measurement error of these four different methods. IPPV1 was chosen as the common reference. Differential en proportional bias and precision are reported as the intercept and the slope respectively of the models studied.

Results: Data from 27 subjects were collected. The iPPV showed minimal bias and improved precision. pPPV showed increasing bias (0.879 - 1.999) with the increase in the respiratory cycles as well as precision (0.633-1.08). The Aboy algorithm model showed reduction in bias (-0.473 - -0.139) and precision (0.235-0.146) by the larger fixed windows. Bias increases from the smaller windows to the larger windows in the Lansdorp method. Precision improves over the same range.

Conclusions: Every method has its own measurement error. There is a proportionality in the measurement error in the methods we compared for calculating PPV. The bias is variable by each method we studied.