Diabetes Mellitus Predicts Severe Respiratory Failure After Aortic Arch Replacement

Objective: The lung is a target organ of diabetes mellitus (DM) via glycation of pulmonary elastin and collagen. In addition, hyperglycemia facilitates hypercapnic respiratory failure, which is instead mitigated by low-carbohydrate high-lipid (LC-HL) nutrition. We sought to determine the association between diabetes and severe respiratory failure (SRF, i.e., need of reintubation\tracheostomy) after open aortic arch replacement (AAR).

Methods: Machine learning algorithms were evaluated for precision and recall (F2 score) and for clinical applicability to predict SRF. Conditional regression evaluated independent predictors of SRF after 1:2 propensity-score matching.

Results: Information on diabetic status was available in 1275 patients undergoing AAR (1997–2023). Although support vector machine presented the highest F2 score (F2 = 0.337), conditional inference trees provided the most clinically applicable algorithm: diabetes was the best predictor of SRF, which occurred in 20/150 diabetics (13.3%) vs. 40/1125 nondiabetics (3.6%) (p < 0.01). In diabetics, left ventricular ejection fraction (LVEF) was the next best predictor of SRF, which occurred in 8/20 diabetics with LVEF ≤ 30% (40%) vs. 12/130 diabetics with LVEF >30% (9.2%) (p = 0.02). In nondiabetics, chronic obstructive pulmonary disease (COPD) was the next best predictor of respiratory failure, which occurred in 14/177 nondiabetics with COPD (7.9%) vs. 26/948 nondiabetics without COPD (2.7%) (p = 0.01). In the matched sample, diabetes was independently able to predict SRF (OR 2.20, 95% CI 1.10 | 4.42).

Conclusions: DM was the best predictor of SRF after AAR. Strict glycemic control and LC-HL nutrition should be evaluated as measures to reduce postoperative SRF in diabetic patients.