Abstract
Purpose
While changes in weight are often observed in patients with colorectal cancer (CRC), they provide limited information regarding how or if body composition changes might occur. We used artificial intelligence (AI) to analyse computed tomography (CT) images from patients’ lumbar region to investigate how body composition differs across clinical disease stages of CRC.
Methods
A retrospective analysis was performed on patients diagnosed with CRC and treated at Western Health, Melbourne (2013–2023). Three‐dimensional measures of skeletal muscle (SM), visceral adipose tissue (VAT), subcutaneous adipose tissue (SAT) volumes and radiodensities were automatically segmented from staging CT scans (L1–L5) using a pretrained AI model. Disease stage was determined using the Australian Clinicopathological Staging (ACPS) system. Body composition values were adjusted for physiological variation (age, sex and height) and compared across ACPS stages.
Results
A total of 1138 individuals were evaluated (68.2 ± 13.0 years; 59.8% male), with 24% classified as Stage A, 30.5% as Stage B, 17.5% as Stage C, and 28% as Stage D. No significant differences were observed between Stages B and C, and these groups were therefore combined (Stage B/C) for further analysis. VAT volume was significantly lower in Stage B/C versus Stage A (
p
= 0.004) and declined further in Stage D (
p
= 0.003). In contrast, SM and SAT volumes were significantly lower only in Stage D compared with earlier stages. Radiodensity measures showed that SAT and VAT radiodensity increased progressively from Stage A to Stage D, whereas SM radiodensity increased only in Stage D.
Conclusion
CRC patients were shown to exhibit distinct body composition changes at diagnosis depending on disease stage. Adipose tissue loss and density changes occurred earlier than muscle changes. Longitudinal studies are warranted to understand these trajectories but suggest that nutritional and supportive interventions could be targeted based on CRC stage at diagnosis.