Introduction

In non-small cell lung cancer (NSCLC), body composition metrics, such as skeletal muscle (SKM) volume, adipose tissue, and aortic calcification, were reported to be associated with baseline physiological resilience, survival and treatment-related side-effects [1], [2], [3], [4]. These non-invasive biomarkers may help identify patients who were more vulnerable, support pre-treatment optimisation, guide supportive care and follow-up.

Body composition is often approximated using body weight or body mass index, but these measures did not distinguish muscle from fat. Computed tomography (CT)-based body-composition analysis provided a more accurate assessment by directly quantifying tissue compartments [5]. This is particularly relevant in patients with NSCLC receiving radiotherapy, where skeletal muscle loss had been shown to predict worse overall survival (OS) [2], [6]. Automated AI methods enable fast, reproducible, population-scale assessments without time-consuming manual segmentation [7].

Beyond muscle and fat, other CT-derived metrics may carry prognostic value. A recent study suggested that a significant reduction in left ventricular (LV) mass is associated with poorer functional status and increased all-cause mortality in advanced cancer patients [8]. Only a few studies have used routine non-electrocardiogram (ECG)-gated CT scans for survival analysis [9], [10]. A high cohort-level correlation between LV mass measured on gated and ungated scans has been reported (r = 0.95) [9], but individual-level agreement and potential systematic bias remained uncertain.

Most existing studies either associated radiation dose metrics with anatomical changes [11] or associated those changes with patient survival, typically using only two time points [12], [13]. This approach limits characterisation of longitudinal anatomical patterns across multiple time points.

We hypothesised that longitudinal changes in body composition, including skeletal muscle (SKM) and adipose tissue, and in the irradiated heart over months to years would be associated with overall survival in patients with NSCLC undergoing radical radiotherapy. This study extends previous research by evaluating body composition changes over a longer follow-up period and by including an exploratory assessment of left ventricular (LV) mass using ungated CT. In addition, we investigated treatment-related and dose–volume predictors of adverse longitudinal changes, including severe skeletal muscle loss, LV myocardial atrophy, and LV myocardial hypertrophy.