# Development of a tongue ultrasound-based predictive model for hypoxemia during painless gastroscopy in ASA I-II patients

## TL;DR

This study develops a predictive model using tongue ultrasound to identify patients at risk of hypoxemia during painless gastroscopy.

## Contribution

A novel predictive model for hypoxemia during gastroscopy using tongue ultrasound indicators is developed and validated.

## Key findings

- A nomogram integrating BMI, Mallampati score, tongue thickness, and propofol dose showed good predictive accuracy (AUC 0.833).
- The model demonstrated favorable calibration and decision curve analysis results, indicating clinical utility.
- Tongue ultrasound-based indicators were found to be reliable predictors of hypoxemia in this patient population.

## Abstract

The risk of hypoxemia in painless gastroscopy has been widely recognized, but reliable predictors are still lacking. Tongue ultrasonography has been shown to facilitate the identification of difficult airways. In this study, we hypothesize that tongue ultrasonography may predict hypoxemia during painless gastroscopy, and aim to develop a predictive model for hypoxemia based on its prognostic value.

This study included 304 patients underwent painless gastroscopy. Common and tongue ultrasound indicators were used for the prediction, including body mass index (BMI), Mallampati test score, tongue thickness (TT) and hyomental distance. Univariate and multivariate logistic regression were used to identify independent predictors of hypoxemia. Nomograms were constructed to predict hypoxemia based on the logistic regression analysis results and established risk factors documented in prior literature. Receiver operating characteristic (ROC) curves were used to evaluate the accuracy of the nomograms. The nomogram was internally validated.

BMI, Mallampati score, TT, and popofol dose were integrated for hypoxemia nomogram. The areas under the ROC curves were 0.833 (95% confidence interval (CI) [0.762–0.904]). The calibration curve and decision curve analysis of the prediction model indicated that the model could have favourable predictive ability.

Nomograms based on tongue ultrasonography could be a reliable tool in predicting hypoxemia during painless gastroscopy.