A prediction model for skeletal muscle evaluation and computed tomography-defined sarcopenia diagnosis in a predominantly overweight cohort of patients with head and neck cancer

PURPOSE: This study investigates the feasibility of computed tomography (CT)-defined sarcopenia assessment using a prediction model for estimating the cross-sectional area (CSA) of skeletal muscle (SM) in CT scans at the third lumbar vertebra (L3), using measures at the third cervical level (C3) in a predominantly overweight population with head and neck cancer (HNC). METHODS: Analysis was conducted on adult patients with newly diagnosed HNC who had a diagnostic positron emission tomography–CT scan. CSA of SM in CT images was measured at L3 and C3 in each patient, and a predictive formula developed using fivefold cross-validation and linear regression modelling. Correlation and agreement between measured CSA at L3 and predicted values were evaluated using intraclass correlation coefficients (ICC) and Bland–Altman plot. The model’s ability to identify sarcopenia was investigated using Cohen’s Kappa (k). RESULTS: A total of 109 patient scans were analysed, with 64% of the cohort being overweight or obese. The prediction model demonstrated high level of correlation between measured and predicted CSA measures (ICC 0.954, r = 0.916, p < 0.001), and skeletal muscle index (SMI) (ICC 0.939, r = 0.883, p < 0.001). Bland–Altman plot showed good agreement in SMI, with mean difference (bias) = 0.22% (SD 8.65, 95% CI − 3.35 to 3.79%), limits of agreement (− 16.74 to 17.17%). The model had a sensitivity of 80.0% and specificity of 85.0%, with moderate agreement on sarcopenia diagnosis (k = 0.565, p = 0.004). CONCLUSION: This model is effective in predicting lumbar SM CSA using measures at C3, and in identifying low SM in a predominately overweight group of patients with HNC.