Derivation of anthropometric-based equations to predict lean body mass composition of cancer patients

Background Lean body mass (LBM) composition of cancer patients is a predictor of chemotherapy-related adverse events and overall cancer survival. However, clinicians lack validated algorithms that can be applied to measure the LBM of cancer patients to facilitate accurate chemotherapy dosing. Our goal was to develop LBM predictive equations using routinely measured anthropometric measures among cancer patients. Methods We leveraged the 1999-2006 National Health and Nutrition Examination Survey (NHANES) data cycles containing information on self-reported cancer diagnosis, LBM measures based on dual-energy x-ray absorptiometry (DXA) and several anthropometric and demographic factors. We restricted our analysis to participants who had been diagnosed with cancer at the time of surveys. The data was randomly split to 75%:25% to train and test predictive models. Least absolute shrinkage and selection operator (LASSO) models were used to predict LBM based on anthropometric and demographic factors, overall and separately among sex and sex-by-race/ethnic subgroups. LBM measured directly with DXA served as the gold standard for assessing the predictive abilities (correlations [R2] and the Root Mean Square Error [RMSE]) of the derived LBM-algorithms. We further compared the correlations between both DXA-based LBM and predicted LBM and urine creatinine levels, a known biomarker of muscle mass. Results We identified 1,777 cancer patients with a median age of 71 (interquartile range [IQR]: 60-80) years. The most parsimonious model comprised of height and weight, which accurately predicted LBM overall (R2=0.86, RMSE =2.26). The predictive abilities of these models varied across sex-by-race/ethnic groups. The magnitude of correlations between derived LBM-algorithm and urine creatinine levels were larger compared to those measured between DXA-based LBM and urine creatinine levels (R2=0.30 vs. R2=0.17) Conclusions We successfully developed a simple sex-specific and sex-by-race/ethnicity-specific models to accurately predict the LBM of cancer patients by using only height and weight. The simplicity and high accuracy of these models make them inexpensive alternatives to measuring the LBM of cancer patients. Data on the LBM of cancer patients could help guide optimal chemotherapy dose selection among cancer patients.