A retrospective study found that resectable non-small cell lung cancer (NSCL) can be significantly improved by integrating clinical tumour body composition (CTB) model with clinical tumour (CT) model.
The comprehensive model showed strong predictive capability for 1, 2, 3, and 5-year survival.
Researchers collected clinicopathological data for all patients (e.g. sex, height, weight, smoking status). 1,038 patients were randomly allocated to centres A and B in a 7:3 ratio into training and internal validation cohorts. Centre C was used for external validation of the survival prediction models.
All patients underwent contrast-enhanced chest CT, within a month before surgery. AutoPanoM segmented body composition, while PyRadiomics extracted features from tumours and segmented volumes.
Researchers derived tumour (T-score) and body composition (BC-score) risk scores using LASSO-Cox regression, then built stratified models to predict poor overall survival.
Patients were stratified into four phenotypes based on tumour and body composition scores, each with distinct prognoses.
The prediction accuracy improved from 0.725 with CT to 0.780 with CTB in the internal validation group and 0.726 to 0.779 in the external validation group.
The study also showed that the area under the curve for survival at 1, 2, 3, and 5 years were above 0.80 for the CTB model.
The researchers found that the tumour score and the body composition score were both independently linked to survival (hazard ratio: 2.72 and 2.03, respectively), suggesting that body composition provided unique information about survival.
The researchers concluded that combining tumour and body composition radiomics enhances predictive survival of NSCL, as it provides an intricate basis for prognostic assessment.
Integrating additional established prognostic factors, such as, epidermal growth factor receptor mutation status and programmed death-ligand 1 expression levels, to further refine predictive performance, could make survival prediction more accurate.
Reference
Huang Y et al. Automated three-dimensional radiomic body composition analysis enhances survival prediction in resectable non‑small cell lung cancer. European Radiology Experimental. 2026. 10(1):135.
Featured image: zapp2photo on AdobeStock
Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.
