In a multicenter study, we employed automated volumetric body composition analysis from CT studies during initial lung cancer staging, focusing on the prognostic significance of abdominal adipose tissue, muscle, and bone volumes. Validation was independently conducted at University Hospital Münster (Hospital B) for findings from University Hospital Essen (Hospital A), aiming to lay the groundwork for personalized lung cancer treatment approaches and improved clinical decision-making.
Material und Methoden
Two lung cancer cohorts from Hospital A (n=3345, 86% NSCLC) and B (n=567, 84% NSCLC) underwent CT abdomen series analysis within +/- 60 days of diagnosis. Volumetric measures for the abdominal cavity, derived via Deep-Learning Segmentation Network, produced three markers: sarcopenia index (Muscle / Bone), myosteatostic fat index (IMAT / TAT), and abdominal fat index (VAT / SAT). Univariate and multivariate analyses, categorized by patient sex, explored their impact on overall survival. A multivariate regression model, trained on Hospital A’s data, was validated using Hospital B’s patient data.
Kaplan-Meier estimates revealed significant differences for sarcopenia and the myosteatotic fat index, with abdominal fat index significance observed exclusively among women in both cohorts. Cox regression affirmed the prognostic value of the first two indices, leaving the abdominal fat index uncertain. The multivariate regression model demonstrated consistent results in both training and validation sets.
The study establishes that CT-based body composition analysis (BCA) predicts overall survival in lung cancer patients across multicenter settings. Transferability of this impact to other sites was confirmed. Sex-specific patterns in visceral and subcutaneous adipose tissue call for additional research. Further investigation into how these characteristics influence intervention alternatives is warranted.