Accurate estimation of skeletal muscle mass by comparison of computed tomographic images of the third lumbar and third cervical vertebrae in Japanese patients with oral squamous cell carcinoma

Objectives We evaluated the accuracy of estimating the cross-sectional area (CSA) at the third lumbar vertebra (L3) based on the CSA at the third cervical vertebra (C3) using computed tomographic images, and we identified the sources of error and bias using the evaluation of absolute reliability in 89 Japanese patients with oral squamous cell carcinoma. Methods Skeletal muscle CSA was measured at the C3 and L3 on pretreatment computed tomographic images. We used the CSA at the C3 to estimate CSA at the L3 in an existing prediction formula. Correlation coefficients were used to evaluate the relative reliability of the estimate, and Bland–Altman analysis and minimum detectable change (MDC) were used to evaluate its absolute reliability. Results Estimated and actual CSAs at L3 were strongly correlated ( r = 0.885, p < 0.001). The mean difference between the estimated and actual CSAs was − 1.0887 cm 2 , the 95% confidence interval was − 4.09 to 1.91 cm 2 ( p = 0.472), and the 95% limits of agreement were − 29.0 and 26.8 cm 2 . The MDC at the 95% level of confidence in estimated and actual CSAs was 27.9 cm 2 . Conclusions The estimation of CSA at the L3 from the existing prediction formula with the CSA at the C3 had no systematic biases, but it did have random errors. Random errors resulted from measurement errors and biological variation. Usefulness of the existing formula is limited by physical differences in populations.