Decision trees tailor airway tube sizes for dogs

A decision-tree analysis of 580 dogs suggests that body weight and tracheal diameter can guide the selection of endotracheal tubes, with separate pathways reflecting the narrower airways often found in brachycephalic breeds. The retrospective, hospital-based study analyzed dogs that underwent anesthesia or emergency care at the Northern Regional Veterinary Healthcare and Learning Center at Chiang Mai University in Thailand during 2024. Its models identified body weight as the strongest predictor of tube size and measured internal tracheal diameter as the second most important variable, while age and sex contributed little to the predictions in most cases.

Endotracheal intubation secures the airway during anesthesia, supports oxygenation and delivery of volatile anesthetics, and can reduce aspiration of regurgitated material. Tube size presents a trade-off: a tube that is too narrow can increase airway resistance and the work of breathing, while one that is too wide can injure the larynx or trachea. Because dogs vary considerably in skull shape and airway anatomy, the study grouped 446 dolichocephalic and mesocephalic dogs together as Group DM and compared them with 134 brachycephalic dogs in Group B. At the same body weight, the brachycephalic group generally required tubes with smaller internal diameters, consistent with anatomical constraints associated with breeds such as English Bulldogs, French Bulldogs, Pugs and Pekingese.

To build and test the models, the researchers used medical records and pre-anesthetic lateral thoracic radiographs from dogs that had been intubated between January and December 2024. The final dataset followed the exclusion of duplicate or erroneous records, non-canine patients, cases without thoracic radiographs, and animals examined radiographically because of severe illness. Recorded variables included breed, sex, age, body weight, the internal diameter of the cuffed tube used, and the internal tracheal diameter at the thoracic inlet, defined as the region immediately cranial to the first rib. One investigator measured the tracheal diameter three times for each dog and used the mean in subsequent analyses.

The repeated measurements were highly consistent. The mean within-dog standard deviation was 0.1899 millimeters, while the intraclass correlation coefficient was 0.997 for a single measurement and 0.999 for the average of three measurements. These results indicate excellent repeatability by the same observer, although the study did not test whether different observers would obtain equally consistent measurements. Within each anatomical group, the data were divided randomly into training and testing subsets, accounting for 70% and 30% of the observations, respectively. The decision trees were tuned with 10-fold cross-validation repeated 10 times and then pruned using cost-complexity methods.

Variable-importance analysis demonstrated that body weight and internal tracheal diameter were the dominant predictors in both groups. Together, they accounted for 80% and 57.9% of predictive importance in Group DM and for 68.7% and 36.7% in Group B. Age and sex had little influence in the non-brachycephalic group, at 0.7% and 2.3%, respectively. In the brachycephalic group, age reached 12.8% importance, compared with 0.7% for sex. A separate random-forest analysis produced the same ranking of body weight first and tracheal diameter second, indicating that the result was not an artifact of one modeling approach.

For non-brachycephalic dogs, the Group DM tree began with a body-weight split at 10.4 kilograms. Dogs below 4.68 kilograms were assigned a predicted tube internal diameter of 3.8 millimeters, which the investigators rounded to the commercially available 4.0-millimeter size. For dogs weighing 4.68 to 10.4 kilograms, the model combined weight with internal tracheal diameter: a diameter below 7.77 millimeters indicated a 4.5-millimeter tube, while a diameter of at least 7.77 millimeters indicated a 5.5-millimeter tube. Above 10.4 kilograms, the pathways relied on weight alone, moving through 6.5-, 7.5-, 8.0- and 8.5-millimeter tubes before suggesting a 9.5-millimeter tube for dogs heavier than 33 kilograms. That largest recommendation was limited in practice because the hospital’s maximum available tube had a 10.0-millimeter internal diameter.

The Group B tree followed a different structure, beginning with a body-weight threshold of 6.95 kilograms. Dogs below 3.93 kilograms were assigned 4.0-millimeter tubes, and those between 3.93 and 6.95 kilograms were assigned 4.5-millimeter tubes. For brachycephalic dogs above 6.95 kilograms, the model used tracheal diameter: an internal diameter below 9.62 millimeters indicated a 5.0-millimeter tube, 9.62 to 10.7 millimeters indicated a 5.5-millimeter tube, and a diameter above 10.7 millimeters indicated a 6.5-millimeter tube. The smaller tube recommendations at similar weights highlight why a single weight-based rule can be inadequate for dogs with brachycephalic airway anatomy.

Performance was stronger for the larger, non-brachycephalic model. Its coefficient of determination was 0.868 in the training subset and 0.861 in the independent test subset, with test-set mean absolute error of 0.507 millimeters and root mean square error of 0.723 millimeters. Within the predefined tolerance of plus or minus 0.5 millimeters, 56.72% of the Group DM test predictions matched the clinician-selected tube size; 16.42% were more than 0.5 millimeters too large and 26.87% were more than 0.5 millimeters too small. The brachycephalic model performed less consistently: its testing R-squared fell to 0.265, mean absolute error rose to 0.524 millimeters and mean absolute scaled error reached 0.747, indicating weaker generalization than in the training data. Even so, 75.61% of Group B test predictions fell within the stated tolerance.

The researchers also compared predicted tube outer diameter with measured internal tracheal diameter as an indirect safety check. The predicted outer diameter did not exceed the measured tracheal diameter for 91.0% of all dogs, including 91.9% of Group DM and 88.1% of Group B. These associations do not establish that every tube fit safely: the study lacked routine post-intubation imaging, cuff-pressure monitoring and formal minimal-occlusive-volume measurements, and the “hiss test” used to identify substantial leakage did not quantify airway injury. The retrospective records also came from multiple anesthetists at one teaching hospital, creating possible differences in tube-selection practice. The limited brachycephalic sample and large anatomical variation were further constraints, while the absence of inter-observer testing means that the radiographic measurement process remains to be validated beyond the single investigator who performed it.

Tube outer diameter remains a decisive factor in anatomical compatibility, and the wall thickness of the polyurethane and polyvinyl chloride tubes used in the study could vary by 2 to 4 millimeters between manufacturers. Internal diameter, which largely determines resistance to gas flow, therefore does not fully describe whether a tube will pass atraumatically. On this basis, the authors recommend preparing the predicted tube together with one size smaller and one size larger, corresponding to an internal-diameter difference of 0.5 millimeters, so that a veterinarian can adjust the choice during intubation. The findings offer a practical decision-support tool for novice anesthetists, but the retrospective single-center design means that larger multicenter studies, direct measurements of cuff inflation, post-intubation imaging and assessments of airway resistance, ventilation and long-term complications are needed before the algorithms can be treated as universal clinical guidelines.

Subject of Research: Retrospective development of decision-tree models for selecting endotracheal tube sizes in dogs by body weight, tracheal diameter and skull conformation.

Article Title: Decision tree models to guide endotracheal tube selection in dogs

Article References: Ngamlertwong, N., Punyapornwithaya, V., Siriyakhun, S., Seymour, C., & Seki, S. (2026). Decision tree models to guide endotracheal tube selection in dogs. Frontiers in Veterinary Science, 13, Article 1894775. https://doi.org/10.3389/fvets.2026.1894775

Image Credits: AI Generated

DOI: 10.3389/fvets.2026.1894775

Keywords: veterinary anesthesia, endotracheal tubes, decision trees, dogs, airway anatomy, machine learning, brachycephalic breeds, body weight, tracheal diameter, retrospective study, Decision, tree