Correlation Between Upper Airway Endoscopy and Physiological Traits of Obstructive Sleep Apnea

Correlation Between Upper Airway Endoscopy and Physiological Traits of Obstructive Sleep Apnea: a Retrospective Study

OBJECTIVES

To train and test a mathematical model to predict complete concentric collapse at the level of the palate (CCCp, primary) and other sites of upper airway collapse (secondary) during drug-induced sleep endoscopy (DISE) using the data captured during a diagnostic polysomnography (PSG).

HYPOTHESIS

The site, pattern and degree of upper airway collapse is associated with distinct flow features as captured during a baseline PSG.

STUDY DESIGN

Retrospective trial.

STUDY POPULATION

200 patients with moderate to severe obstructive sleep apnea (OSA, AHI ≥ 15/h) who underwent both a DISE and a diagnostic PSG at the Antwerp University Hospital (UZA) between January 2018 and December 2020.

OUTCOME MEASURES:

Raw data as captured during a diagnostic PSG, including electroencephalography (EEG), flow, electrocardiography (ECG), electromyography (EMG), oxygen desaturation and breathing effort.

SAMPLE SIZE / DATA ANALYSIS

Data of 200 patients will be retrospectively included into this study protocol. Different machine learning techniques will be adopted to select features, train the model and test the model.

TIME SCHEDULE

January 30, 2021 - November 30, 2021

Inclusion Criteria:

• 18 years or older.
• Diagnosis with symptomatic moderate to severe OSA (AHI ≥ 15/h).
• Underwent DISE at UZA as part of the clinical pathway for non-continuous positive airway pressure (CPAP) therapies
• Underwent a baseline diagnostic PSG at UZA max 2 years before DISE

Exclusion Criteria:

• Medication use related to sleeping disorders
• Central Sleep Apnea Syndrome.
• Medical history of known causes of tiredness by day or severe sleep disruption (insomnia, periodic leg movements, narcolepsy).
• Seizure disorder.
• Known medical history of intellectual disability, memory disorders or current psychiatric disorders (psychotic illness, major depression, or acute anxiety attacks as mentioned by the patient).
• Pregnancy
• <4 h of data recording during the baseline PSG night