2026/07/21 by Wu Yj, Zhenyi Yan, Z Li · 1 voice
Medicine · #Airway Management and Intubation Techniques #Infection Control and Ventilation #Nasal Surgery and Airway Studies
paper · doi:10.1111/anae.70295
openalex publication_date 2026/07/21 · openalex created_date 2026/07/22 · openalex updated_date 2026/07/22
We read with interest the study by Wünsch et al., who evaluated pre-operative three-dimensional facial scanning for predicting difficult facemask ventilation [1]. The study is a useful step towards making airway assessment more objective and reproducible. The authors assessed the DIFFMASK score [2] in a different surgical population and identified interpretable facial features, particularly those involving the nose; lower mandible; neck region; and facial convexity. These features have plausible anatomical links with mask seal and upper airway patency. We suggest, however, that the clinical value of this approach should ultimately be judged not only by model discrimination but also by whether it identifies reproducible airway difficulty and prompts actionable changes before induction. The primary outcome was difficult facemask ventilation recorded as a health record alert by the airway operator after difficulty had occurred. This outcome is pragmatic and reflects routine clinical practice but it also depends on individual judgement, local documentation habits and the threshold for issuing an airway alert. A model trained against this endpoint might therefore predict documentation behaviour as much as the physiological or technical difficulty of ventilation. Future studies could report the alert alongside more operational outcomes, such as two-handed mask ventilation; use of an oral or nasal airway; jaw thrust; conversion to manual ventilation; senior anaesthetist takeover; inability to generate end-tidal carbon dioxide; hypoxaemia; emergency supraglottic airway insertion; or unplanned tracheal intubation [3, 4]. Such endpoints would bring the prediction target closer to events that directly affect airway safety. The gain in discrimination also needs translation into a clinical decision. Adding three facial shape features to DIFFMASK increased the optimism-corrected area under the receiver–operator curve from 0.73 to 0.76. This finding suggests incremental information, but it remains unclear at what risk threshold the scan should alter management. Calibration, likelihood ratios at clinically plausible thresholds, decision-curve analysis and estimates of net benefit would help determine whether scanning improves decisions rather than model fit alone [5]. A high-risk scan should ideally correspond to specific actions, such as enhanced pre-oxygenation; high-flow nasal oxygen; two-person mask ventilation readiness; early airway adjunct use; delayed neuromuscular blockade; immediate availability of a supraglottic airway; or consideration of awake airway management. Finally, the settings in which a pre-operative warning might be most useful differ from the development cohort. The study was done in a single-centre, mainly involved elective head and neck surgery and did not include patients planned for rapid sequence induction or awake tracheal intubation, or patients who were pregnant. These exclusions are reasonable for an initial development study, but external validation should prioritise populations in whom difficult facemask ventilation is both more consequential and more likely to change the airway plan, including patients with severe obesity or obstructive sleep apnoea, and those having emergency or obstetric surgery. Wünsch et al. have provided an important step towards digital airway phenotyping. Linking three-dimensional facial scanning to standardised airway outcomes, clinically meaningful thresholds and validation in high-risk settings would make this promising approach more persuasive as a tool for improving airway safety.