Hepatologist‐Driven Screening for Obstructive Sleep Apnea in Patients With MASLD Using a Modified STOP‐Bang Questionnaire: The SOMA Model
Naruyasu Kakita, Shin Satake, Midori Jono, Hiroshi Tsumuro, Shuko Iwatani, Kenji Aoi, Masami Yamazaki, Toshio Naka, Kazuhiro KatayamaABSTRACT
Aim
To evaluate the feasibility, diagnostic yield, and downstream clinical linkage of a hepatologist‐driven obstructive sleep apnea (OSA) screening pathway using a modified STOP‐Bang questionnaire in patients with metabolic dysfunction‐associated steatotic liver disease (MASLD).
Methods
From July 2025 to March 2026, 555 outpatients with MASLD were screened using the Screening for OSA in MASLD (SOMA) model. Home sleep apnea testing (HSAT) referral relied on the modified STOP‐Bang questionnaire, Epworth Sleepiness Scale (ESS), and clinical judgment. Among 276 HSAT‐evaluated patients, 258 with valid recordings were analyzed. Clinically significant OSA was defined as a respiratory event index (REI) ≥ 15 events/hour.
Results
Among HSAT‐evaluated outpatients with MASLD, OSA (REI ≥ 5 events/hour) and clinically significant OSA occurred in 253 (98.1%) and 203 (78.7%), respectively. The modified STOP‐Bang questionnaire outperformed the ESS (area under the curve: 0.616 vs. 0.485, p = 0.021). A modified STOP‐Bang score ≥ 3 showed high sensitivity (93.6%) but low specificity (7.3%), whereas a score ≥ 5 showed more balanced sensitivity (59.1%) and specificity (54.5%). ESS score ≥ 11 showed poor sensitivity (10.8%). Among patients with REI ≥ 15 and ≥ 30 events/hour, full polysomnography was completed or scheduled in 64.0% and 73.0%, respectively. OSA severity was not significantly associated with Fibrosis‐4 index, Mac‐2 binding protein glycosylation isomer levels, or shear wave elastography findings.
Conclusions
Clinically significant OSA was frequent among selected HSAT‐evaluated MASLD outpatients. The SOMA model may represent a feasible hepatologist‐driven screening/referral pathway integrating structured triage, clinical judgment, and HSAT. However, these findings should not be interpreted as prevalence estimates or evidence of long‐term clinical benefit.