Background: Muscle mass and density assessed from CT-images at the L3 level are prognostic for survival and predict toxicity in cancer patients. However, L3 is not always included on routine CT-scans. We aimed to investigate whether images at the Th4 level may be used instead.
Methods: Patients from three chemotherapy trials in advanced NSCLC were eligible (n = 1305). Skeletal muscle area (cm2), skeletal muscle index (SMI, cm2/m2) and skeletal muscle density (SMD) at Th4 and L3 levels were assessed from baseline CT-scans. SMI and SMD at the Th4 and L3 level were transformed into z-scores and the agreement between scores was investigated by Bland-Altman plots and estimated by intra-class correlation analyses. Linear regression was used to test if Th4 SMI and SMD z-scores predicted L3 SMI and SMD z-scores.
Results: CT-images from 401 patients were analysable at both levels. There was a moderate agreement between Th4 and L3 SMI z-scores with an intra-class correlation of 0.71 (95% CI 0.64-0.77) for men and 0.53 (95% CI 0.41-0.63) for women. Regression models predicting L3 SMI z-scores from Th4 SMI z-scores showed coefficients of 0.71 (95% CI 0.62-0.80) among men and 0.53 (95% CI 0.40-0.66) among women. R-squares were 0.51 and 0.28, respectively, indicating moderate agreement. A similar, moderate agreement between Th4 and L3 SMD z-scores was observed.
Conclusion: There was only moderate agreement between muscle measures from Th4 and L3 levels, indicating that missing data from the L3 level cannot be replaced by analysing images at the Th4 level.