[Construction and validation of clinical prediction model of tongue base collapse under drug-induced sleep endoscopy in OSA patients].

ABSTRACT

Objective: To analyze the correlation between drug-induced sleep endoscopy (DISE), results, polysomnography (PSG) indicators, and clinical parameters in patients with obstructive sleep apnea (OSA), and to establish and validate a predictive model for tongue base plane obstruction. Methods: This retrospective study analyzed 117 OSA patients diagnosed via PSG and treated at the Department of Otolaryngology-Head and Neck Surgery, Nanfang Hospital, Southern Medical University, between October 2014 and March 2022. The cohort comprised of 114 males and 3 females, with an age range of 20 to 54 years (mean age 38.1±8.4 years). Data on DISE results, PSG results, and clinical indicators were collected for all 117 patients. Logistic regression analysis was performed to identify relevant indicators, and a predictive model for tongue base plane obstruction was constructed and internally validated using the R programming language. Results: Univariate logistic regression analysis identified four independent risk factors for predicting tongue root plane obstruction tonsil grading, N2, N3, and rapid eye movement sleep(REM) stage [OR0.412(0.260~0.652),1.045(1.012~1.079),0.943(0.903~0.984),0.961(0.925~0.998),P <0.05]. Multivariate logistic regression analysis confirmed tonsil grading and N3 sleep stage (12.48±12.22%) as significant predictors. A nomogram model incorporating these factors demonstrated good predictive performance, with an area under curve(AUC) of 0.82 (95%CI 0.548-1.000), an optimal cutoff of 0.519, a specificity of 80.0%, and a sensitivity of 86.7%. Internal validation of the model in the validation cohort yielded an AUC of 0.751 (95%CI 0.625-0.876). Conclusions: Tongue base plane obstruction observed during DISE in OSA patients is associated with tonsil grading and N3 sleep stage duration. The predictive model developed for tongue base plane obstruction based on DISE demonstrates good efficacy, as evidenced by its internal validation.