Abstract Background Single-slice computed tomography (CT) body composition has been studied extensively for prognostication in patients with cancer. New software packages can also provide multi-slice volumetric measurements, but the clinical utility of these remains under explored. This study aimed to evaluate the concordance between single- and multi-slice body composition analyses in patients with oesophagogastric cancer, and to explore the association between these measures and overall survival. Methods Consecutive patients with newly diagnosed oesophagogastric cancer were identified through the prospectively maintained regional database of the South-East Scotland Cancer Network across a 2-year study period. CT body composition analyses were undertaken using scans collected during routine clinical care. Single-slice (cross-sectional area at mid L3) and multi-slice (volume between T12-L4) measurements were obtained for skeletal muscle (SKM), subcutaneous adipose (SAT), visceral adipose (VAT) and intermuscular adipose (IMAT). Agreement between sex-stratified z-scores was quantified using Pearson correlation coefficients and Bland–Altman analyses. Cox proportional hazard modelling was used to estimate the effect of these measures on overall survival. Results Overall, 504 patients (67.9% male, median 72 years) were newly diagnosed with oesophagogastric cancer during the study period. Treatment was commenced with curative intent for 131 patients (26.0%). Single-slice area and multi-slice measurements were highly correlated for SKM (R: 0.97, p<0.001), SAT (R: 0.98, p<0.001) and VAT (R: 0.97, p<0.001), SKM radiodensity (R: 0.93, p<0.001) and IMAT (R: 0.88, p<0.001). Bias on Bland-Altman analysis was 0.00 for all tissue measurements. Limits of agreement (LoA) were narrowest for SAT (± 0.43), VAT (± 0.46) and SKM (± 0.48), but slightly wider for SKM radiodensity (± 0.73) and IMAT (± 0.96). Adipose tissue ‘outliers’ (those where agreement between single- and multi-slice z-scores was outside the LoA) had a higher median weight and body mass index (BMI), suggestive of poorer agreement in patients with obesity. Sensitivity analysis, excluding those with BMI >30, narrowed the LoA for SKM, VAT, SAT and IMAT. Direction and magnitudes of observed effect sizes for overall survival were all highly comparable, with hazard ratios for each tissue type varying by ≤0.04 between single- and multi-slice adjusted estimates. Conclusions Single-slice and multi-slice assessments of CT body composition provide highly correlated measurements for skeletal muscle, subcutaneous fat, visceral fat, and intra-muscular fat amongst patients with OG cancer. However, agreement between measurements is poorer in patients with obesity. The associations between these tissues measurements and overall survival are comparable across both types of body composition analysis.
Read full abstract