Obesity and age increase hypoxia risk during esophagogastroduodenoscopy under monitored anesthesia care: risk stratification and evaluation of a procedural oxygen mask

Deep sedation under monitored anesthesia care (MAC) during esophagogastroduodenoscopy (EGD) has a significant risk of hypoxia in obese patients. Recently high flow nasal cannula studies were conducted with good results for obese patients, however, data on Procedural Oxygen Mask (POM) are extremely scarce. The objectives of this study were to determine the incidence of hypoxia and impact of obesity during EGD, to describe our experience with POM, and to evaluate the risk factors and derive a prediction model using the machine-learning (ML) models. We reviewed the records of consecutive patients with BMI below 50 kg/m² who underwent EGD alone or both EGD and colonoscopy at our outpatient endoscopy center under MAC from 7/31/2023 to 1/31/2024. Multivariable logistic regression models were used to assess factors associated with hypoxia. In addition, we used our previously reported case series ( n = 559) (Supplementary Table 1) and current data ( n = 1294 without POM) to analyze the risk factors for hypoxia using machine-learning (ML) techniques. The median age of the cohort was 58 years (range: 19–90). The median Body Mass Index (BMI) was 25 kg/m² (range: 15–45 kg/m²) with 302 obese (BMI > = 30 kg/m²). Intra-procedural hypoxia developed in 205 cases (15.4%). The incidence of hypoxia was significantly greater among those with obesity (84 of 302, 27.8%) compared to those without obesity (121 of 1021, 11.8%, p < 0.0001) with a difference of 16.0% (95% confidence interval [CI] 10.5–21.4%). No hypoxia was observed in the 29 POM-managed patients of whom 20 were obese. In multivariable analysis, obesity and age were associated with hypoxia (odds ratio [OR] 2.78, 95% CI 2.01–3.85, p < 0.0001, and OR 1.02, 95% CI 1.01–1.03, p = 0.001, respectively). Seven ML-based models were constructed which demonstrated modest area under the receiver operating characteristic curve of 60–65% in predicting hypoxia. Our study found that obesity and age are risk factors for hypoxia during EGD while the performance of ML-based models was modest. The current exploratory data of POM warrant further evaluation in the high-risk population.