By Vibhu Sharma, MD
Associate Professor of Medicine, University of Colorado, Denver
SYNOPSIS: In this secondary analysis of the BOUGIE trial, patient and operator characteristics may affect successful intubation with either technique despite no difference in outcomes in the primary analysis.
SOURCE: Seitz KP, Spicer AB, Casey JD, et al. Individualized treatment effects of bougie versus stylet for tracheal intubation in critical illness. Am J Respir Crit Care Med 2023;207:1602-1611.
This secondary analysis of the Bougie or Stylet in Patients Undergoing Intubation Emergently (BOUGIE) trial used a machine learning (ML) algorithm to assess whether patient or operator-level differences affect successful intubation with either technique. The original trial showed no differences in first-pass success in the overall population (n = 1,102) or in any prespecified subgroup with respect to stylet or bougie upfront. The authors hypothesized that an ML model may identify whether a specific category of patient benefits from use of a bougie upfront or vice versa.
The authors used a causal forest ML method originally developed by economists to predict a small or large effect of an intervention. Causal ML builds on regression equations wherein an effect (outcome) can be predicted based on multiple prediction variables. The algorithm is able to account for and better detect non-linear relationships between variables: stylet vs. bougie, multiple causes of difficult airways, operator experience with either tool (measured as: yes/no, number done prior to intubation by operator), patient body mass index (BMI), and presence or absence of sleep apnea. The treatment effect/outcome was success of first-pass intubation. A large number of covariates and interaction terms can quickly outnumber the number of observations in a dataset, and an ML algorithm is able to partition for and account for these limitations. The algorithm used the first half of the dataset as a “training” cohort wherein the ML algorithm “learned the dataset,” creating a predictive model, and used the second half to validate the model.
The results of the causal forest plot suggested that operator and patient characteristics did affect successful first-pass intubation depending on whether a stylet or a bougie was used. This is in contrast to prior traditional subgroup analyses, which did not find any differences between stylet or a bougie. The ML algorithm predicted bougie use would be more likely to succeed in the presence of difficult airway characteristics specified a priori, such as higher BMI, higher Acute Physiology and Chronic Health Evaluation (APACHE) score, and operator experience with a bougie. Stylet use was more likely to succeed with attempts in patients with lower BMI, operators with less prior experience with a bougie, or in patients with a lower APACHE score.
COMMENTARY
The BOUGIE trial showed a first-pass success rate of approximately 80% in either group (stylet or bougie).1 This trial also did not show a difference in first-pass success rate with a bougie when a video laryngoscope (VL) was used compared with a direct laryngoscope (DL). Operator comfort with a specific intubation technique is important in clinical practice. The authors suggested that the ML algorithms used in this study are complex, and “external validation in a separate randomized clinical trial of bougie versus stylet would provide a higher level of confidence.”
The BOUGIE trial was done in academic medical centers and, therefore, may not be generalizable to other centers. Based on the results of the DEVICE2 trial and in my practice (an academic medical center with low overall bougie use for first-pass intubation), upfront VL with a stylet, with a bougie as a back-up device if the first attempt fails, is the preferred sequence. I favor an awake intubation with ketamine and/or topical anesthesia if multiple predictors of a difficult airway are present (e.g., obesity, Mallampati class III/IV, cervical spine fracture or mobility concerns, blood or vomitus in the oral cavity, or severe hypoxia). This secondary analysis of the BOUGIE trial does not alter my practice but reinforces the need to work through methodically the probability of a difficult airway prior to intubation and ensure that a bougie is available immediately if there is an unexpected difficult airway despite predictors to the contrary.
REFERENCES
- Driver BE, Semler WE, Self WH, et al. Effect of use of a bougie vs endotracheal tube with stylet on successful intubation on the first attempt among critically ill patients undergoing tracheal intubation: A randomized clinical trial. JAMA 2021;326:2488-2497.
- Prekker ME, Driver BE, Trent SA, et al. Video versus direct laryngoscopy for tracheal intubation of critically ill adults. N Engl J Med 2023;389:418-429.
