Low CT-Derived Thoracic Muscle Area and Composite Weaning Failure in Mechanically Ventilated ICU Patients with Pneumonia: An Exploratory Analysis by Age Group

Background/Objectives: Whether the prognostic contribution of low skeletal muscle mass varies with age in ventilated patients is untested. We assessed whether the association between low computed tomography (CT)-derived thoracic muscle area and ventilator-liberation outcomes differs by age. Methods: This single-center retrospective cohort included 607 adults with pneumonia requiring invasive ventilation in the intensive care unit (ICU) between 2016 and 2021. Thoracic muscle area was quantified by deep-learning segmentation of the whole thoracic musculature (first to twelfth thoracic vertebra) on the chest CT obtained closest to ICU admission and normalized to the length of the thoracic spine (median offset 0 days; 86% within ±7 days, before or after); low thoracic muscle area was the sex-specific lowest quartile. The primary outcome was composite weaning failure (extubation failure, tracheostomy, or ventilation >21 days). Multivariable logistic regression stratified at 75 years tested a muscle-area-by-age interaction. Results: Composite weaning failure occurred in 290 patients (47.8%). In fully adjusted complete-case models (n = 539), low thoracic muscle area was associated with composite weaning failure in younger patients (adjusted odds ratio 3.63; 95% CI 1.67–7.87) but not significantly in older patients (1.68; 0.91–3.13); the age interaction was not significant (p = 0.061) and was removed by height-normalization and by correction for multiple comparisons. A graded association across muscle-area quartiles was present in both strata (lowest vs. highest quartile: younger 4.79, older 2.53). Low thoracic muscle area was not associated with ICU mortality in either stratum, although an exploratory landmark analysis showed higher 7-day mortality in younger patients with low muscle area (20.6% vs. 10.8%). Conclusions: Low CT-derived thoracic muscle area was associated with composite weaning failure in both age strata, with a larger point estimate in patients younger than 75 years, although the age interaction was not significant in the fully adjusted model and was removed by height-normalization and by correction for multiple comparisons. The graded association in both age strata is hypothesis-generating and requires prospective validation.