Personalized Selection of Hyperangulated vs. Macintosh Videolaryngoscope Blades: Randomized Controlled Trial and Development of an Integrated Predictive Algorithm
Personalized Selection of Hyperangulated vs. Macintosh Videolaryngoscope Blades: Randomized Controlled Trial and Development of an Integrated Predictive Algorithm
This study aims to determine whether hyperangulated blades have a higher intubation success rate than Macintosh blades in patients undergoing general anesthesia, and to identify patient subgroups in which hyperangulated blades are particularly advantageous for tracheal intubation.
Randomized controlled trial
This study was designed as a randomized controlled trial. A total of 524 patients were randomly allocated in a 1:1 ratio to either the hyperangulated blade group (H group) or the Macintosh blade group (M group).
The primary endpoint was first-attempt intubation success. The secondary endpoints included the overall tracheal intubation success rate, the number of intubation attempts, time for the tracheal tube to pass through the vocal cords (T5), time to ventilation (T6), and the use of adjunctive airway maneuvers (e.g., external laryngeal manipulation [ELM]).
Between-group comparisons were performed according to the characteristics of each endpoint. Categorical variables were analyzed using the χ² test or Fisher's exact test, as appropriate, and results were presented as risk ratios (RRs) with 95% confidence intervals (CIs). Continuous variables were assessed for normality using the Shapiro-Wilk test. Normally distributed variables were compared using the independent-samples t-test, whereas non-normally distributed variables were analyzed using the Mann-Whitney U test. Results were reported as mean or median differences with 95% CIs.
All study participants experienced both videolaryngoscope blades. However, actual tracheal intubation was performed only with the second blade, while the first blade was used solely for assessment of glottic view and approach times. To minimize learning-related order effects, each attempt was performed by a different investigator, and each investigator was restricted to using only the blade assigned to them. In addition, within-patient comparisons of POGO scores, approach times, and visualization difficulty between the first and second blades were conducted to analyze blade-specific visualization characteristics. If significant differences were observed, potential order effects were further evaluated during the analysis.
Within-patient comparison (paired evaluation)
Before actual tracheal intubation, glottic visualization up to just before the vocal cords was assessed using both blades in all participants. The order of blade use was balanced according to randomization: patients in the M group underwent visualization assessment in the H→M sequence followed by intubation with the Macintosh blade, whereas patients in the H group underwent visualization assessment in the M→H sequence followed by intubation with the hyperangulated blade. Accordingly, within-patient comparisons were conducted using visualization data obtained in the predefined sequence for each group, thereby minimizing order effects.
The evaluated parameters included:
(i) time to obtain a laryngeal view (T1/T3), (ii) time for the tracheal tube tip to reach the vocal cords (T2/T4), (iii) VIDIAC score, (iv) percentage of glottic opening (POGO) score, (v) operator-reported subjective difficulty score, and (vi) complications (e.g., SpO₂ < 90%, dental or soft-tissue injury).
Differences in continuous paired outcomes between the hyperangulated and Macintosh blades were analyzed using the paired t-test or Wilcoxon signed-rank test, as appropriate. When classifying each outcome according to which blade was superior (hyperangulated-favored, Macintosh-favored, or equivalent), McNemar's test was used for binary classifications and the Stuart-Maxwell test for three-category classifications.
Machine learning-based exploratory analysis
Finally, data obtained from the randomized controlled trial and paired analyses were further leveraged to explore, in an exploratory manner, which blade may be more advantageous under specific anatomical and demographic conditions. Outcomes that demonstrated between-group differences in the primary or secondary analyses were selected as target variables. All covariates were included as input features, and an XGBoost-based ensemble tree model was trained.
Model performance was evaluated using the area under the receiver operating characteristic curve (AUROC) with 95% CIs for binary outcomes and root mean squared error (RMSE) and mean absolute error (MAE) for continuous outcomes. SHapley Additive exPlanations (SHAP) analysis was performed to visualize feature importance and interaction patterns, and decision curve analysis (DCA) was used to assess the clinical net benefit of the predictive models.
Inclusion Criteria:
Exclusion Criteria: