# The dose-effect and clinical prediction of the longest apnea duration driving the decrease of blood oxygen: a large sample OSA study with 34-second and 52-second cut-off values was established

## TL;DR

This study finds that longer apnea durations in sleep apnea patients lead to lower blood oxygen levels and establishes thresholds to predict severe oxygen decline.

## Contribution

The study quantifies the dose-response relationship between apnea duration and oxygen saturation and establishes clinical thresholds for prediction.

## Key findings

- For every 1-second increase in apnea duration, blood oxygen saturation decreases by approximately 0.18-0.28%.
- Critical thresholds of 34.20 seconds and 52.07 seconds for apnea duration correspond to oxygen saturation levels of 85% and 80%, respectively.
- A predictive model with high accuracy (AUC = 0.93-0.96) was developed to identify patients at risk of severe oxygen decline.

## Abstract

To quantify the relationship between the longest apnea duration (LAD) and the lowest oxygen saturation (LSaO2) in patients with obstructive sleep apnea (OSA) and to develop a predictive model for the risk of LSaO2 decline.

A total of 1716 OSA patients were enrolled and grouped by severity (236 non-OSA, 395 mild, 365 moderate, and 720 severe). Multiple linear regression was used to analyze the dose-effect relationship between LAD and LSaO2. A logistic regression model was developed to predict LSaO2 grade, with the dataset partitioned into a training set (n = 1,372) and a testing set (n = 344) using random sampling.

(1) For every 1-s increase in LAD, LSaO2 decreased by 0.280% (95% CI: −0.291%∼-0.269%) in a univariate model and still decreased by 0.183% (95% CI: −0.197%∼-0.170%) after adjusting for sex, age, BMI, and AHI; (2) Critical points were identified: LSaO2 was 85% when LAD was 34.20 s and 80% when LAD was 52.07 s; (3) The predictive model showed excellent identification performance for severe (AUC = 0.93) and moderate-severe LSaO2 (AUC = 0.96).

The study first quantifies the dose-response relationship between LAD and LSaO2 and establishes relevant clinical thresholds. The developed model can accurately identify patients at risk of severe and moderate-severe hypoxia, offering a new tool for individualized intervention.