Risk Factors Analysis for Clinical Important Postoperative Nausea and Vomiting

Overview

Postoperative nausea and vomiting (PONV) is a distressing and common complication after surgery. The concept of clinical important PONV (CI-PONV) assesses the impact of PONV on patient-reported outcomes. This research aims to conduct an analysis of the risk factors contributing to CI-PONV utilizing the least absolute shrinkage and selection operator (LASSO) and stepwise regression techniques. All 1154 patients participating in the FDP-PONV trial are included in this study and categorized into two groups: the CI-PONV group and the non-CI-PONV group. CI-PONV is defined as the occurrence of PONV with a simplified PONV impact scale score of 5 or higher within 24 hours after surgery. The LASSO method is employed to identify the most relevant variables from an initial set of 56 related variables and stepwise regression is used to further refine the selection of the ultimate predictors.A logistic regression model was developed based on the selected factors influencing CIPONV. A nomogram was developed as a tool for clinical application.

Drawing from prior studies, we conducted a sample size calculation for a predictive model using the website https://mvansmeden.shinyapps.io/BeyondEPV/. By setting the number of candidate predictors to 9, the events fraction to 0.14, and the criterion value for rMPSE to 0.04, we determined that a minimum total sample size of 900 is required, with a minimally expected event per variable of 13.9. All patients were classified into either the CI-PONV group or the non-CI-PONV group. All 56 perioperative clinical features, encompassing baseline characteristics, preoperative conditions, and intraoperative information, were considered as potential predictive factors. In the quest to uncover potential predictive factors associated with CI-PONV, we employed the least absolute shrinkage and selection operator (LASSO) to sift through clinically significant variables. Subsequently, we utilized stepwise regression based on the Akaike Information Criterion (AIC) to further refine the selection of the ultimate predictors. Finally, a logistic regression model was developed based on the selected factors influencing CIPONV. The discrimination of the model was assessed by the ROCAUC and the goodness of fit of the model was evaluated using the Hosmer-Lemeshow test and calibration plots. A nomogram based on the logistic regression model output was developed as a tool for clinical application.