Hemodynamic Response to the End-expiratory Occlusion Test to Titrate Fluid Challenge in Operating Room.

Overview

Personalzing intraoperative anesthetic fluid management may help in preventing fluid accumulation and related complications. Fluids are gine as boluses in operating room (the so-called FC). The response to the FC is due to several physiological conditions related to the "preload dependency" (i.e. the intrinsic ability of the heart of increasing the stroke volume - SV - in response to fluid administration). The minimal volume required to appropriately "challenge" the cardiovascular system is 4 ml/kg of fluid, but higher volumes (up to 6 ml/kg may be needed). Predicting the response to FC administration may be possible by applying a physiological test (called functional hemodynamic test), such as the end-expiratory occlusion test, consisting in interrupping the mechanical ventilation and hence promoting venous return and consequente SV changes. The percentage of SV increase associated to EEOT may predict fluid responsiveness to the FC (patients responders will increase SV to a bigger extent, as compared to non-responders)

A two-step statistical approach will be used to define the best model to predict the fluid responsiveness. 1. A univariable logistic regression model to test the association of the considered hemodynamic variables provided by the hemodynamic monitoring with the primary outcome (fluid responsiveness at 10th minute). Then a multivariable analysis incorporating the variable in univariable analysis with a p \< 0.2, after testing the colinearity and interactions. Significance threshold for multivariable analysis will be set to 0.05 2. A Hosmer and Lemeshow test was calculated to evaluate goodness of fit for the logistic regression model, the Informative criterion metrics, such as Akaike Infromation Criterion, AIC) and the receiver operating characteristic (ROC) curve \[standard error, (SE)\] analysis evaluated the performance of predictive items for FC response (i.e.Y = dependent variable =SVI increasedby ≥ 10%) 10 minutes (Y10) after FC infusion. The absence of a significant increase in the likelihood value afteromission of each of the remaining variables was checked. To define the best model to predict the amount of fluid in responder group a machine-learning approach will be considered where Y = dependent variable = total amount of crystalloids to obtained SVI ≥ 10% after FC infusion, X = matrix of parameters. The final model decision will be made among following commonly used regression algorithms: linear reagression, Lasso Regression or Ridge Regression. The model performance assessment will be made using metrics like Mean Squared Error (MSE) or R-squared. K-fold cross-validation technique will be applied to get a more robust estimate of the model's performance. The hemodynamic values of responders and non-responders at each step of the protocol are analyzed with a one -way analysis of variance for repeated measurements (ANOVA) and Geisser -Greenhouse (G-G) correction as ajustement for lack of sphericity if needed. Post-hoc pairwise multiple comparisons analysis are performed using Tukey's test to control familywise error. To understand whether hemodynamic changes after EEOT could help in the prediction of minimal dose of FC the study will enroll 2-year evaluable patients, and the final numer will be foreseen in about 300, with aroud 500 fluid challenge. The sample size will allow us to perform three step: 4\. an initial step of 50 patients to understand which variables of hemodynamic changes after EEOT will affect the minimal dose of FC. In this initial part a variable will be considered interesting if the relative p value will be under 0.1 5. A second step including 200 patients' data to create the model, considering also variables interactions 6. A finel step of 50 patients to validate the model