Chien-Feng Lee,1,* Yunn-Jy Chen,2,* Wen-Chi Huang,3 Jen-Wen Hou,4 Yu-Ting Liu,5 Tiffany Ting-Fang Shih,6 Pei-Lin Lee,7– 10 Chong-Jen Yu1,8,9 1Department of Internal Medicine, National Taiwan University Hospital Hsin-Chu Branch, Hsin-Chu, Taiwan; 2Department of Dentistry, School of Dentistry, National Taiwan University Hospital, Taipei, Taiwan; 3Department of Computer Science and Information Engineering, National Taiwan University, Taipei, Taiwan; 4Graduate Institute of Electronics Engineering, National Taiwan University, Taipei, Taiwan; 5MediaTek Inc., Hsinchu, Taiwan; 6Department of Radiology and Medical Imaging, National Taiwan University Hospital, Taipei, Taiwan; 7Center of Sleep Disorder, National Taiwan University Hospital, Taipei, Taiwan; 8Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan; 9School of Medicine, College of Medicine, National Taiwan University, Taipei, Taiwan; 10Center for Electronics Technology Integration, National Taiwan University, Taipei, Taiwan*These authors contributed equally to this workCorrespondence: Pei-Lin Lee, Tel +886-2-2356-2755, Fax +886-2-2358-2867, Email leepeilin@ntu.edu.twObjective: Both continuous positive airway pressure (CPAP) pressure and polysomnographic phenotypes have been associated with mandibular advancement device (MAD) treatment response, but the precise relationship has not been fully elucidated. We hypothesized that utilizing CPAP pressure would predict the MAD response in treatment-naïve patients with moderate-severe obstructive sleep apnea (OSA), and the MAD response would be associated with two polysomnographic phenotypes, including sleep stage dependency and positional dependency.Methods: OSA treatment-naïve patients with an apnea-hypopnea index (AHI) ≥ 15/h who declined CPAP treatment and received MAD treatment for 3– 6 months were enrolled. The MAD treatment response was defined as 1) residual AHI under MAD (AHIMAD) < 5/h and 2) AHIMAD < 10/h. Logistic regression was applied to identify the association between CPAP pressure and MAD treatment responders. The predictability of the MAD responder status utilizing CPAP pressure was assessed with the area under the receiver operating characteristic (AUROC).Results: A total of 128 enrolled patients (AHI ≥ 30/h in 74.2%) were recruited, of whom 119 patients and 80 patients were included for analysis of sleep stage and positional dependency, respectively. REM-predominant OSA had lower AHI than stage-independent OSA, while the supine-predominant phenotype had lower anthropometrics than the nonpositional-dependent phenotype. The response rates for AHIMAD < 5/h and AHIMAD < 10/h were 25.8% and 48.4%, respectively. Lower anthropometrics, baseline AHI, and supine predominance were associated with the responder status, while CPAP pressure was an independent predictor. The AUROCs for the prediction of AHIMAD < 5/h and AHIMAD < 10/h responders were 0.635 and 0.664, respectively. Utilizing a CPAP level > 14 cmH2O as the cutoff to predict criterion 1 and 2 nonresponders, the sensitivity was 93.9% and 95.2%, respectively.Conclusion: In treatment-naïve patients with moderate-severe OSA, the supine-predominant phenotype and lower CPAP pressure were associated with the MAD response, while the sleep stage dependency phenotype was not. Utilization of a CPAP level > 14 cmH2O could be a sensitive measure to identify nonresponders.Keywords: continuous positive airway pressure, mandibular advancement device, polysomnography, phenotype, sleep apnea, obstructive