Development and validation of nomograms for the prediction of low muscle mass and radiodensity in gastric cancer patients

Background: The skeletal muscle mass index (SMI) and skeletal muscle radiodensity (SMD) are important components of sarcopenia and malnutrition. However, their assessment requires additional resources in cancer patients, which is inconvenient for the early detection of sarcopenia and malnutrition.

Objectives: This study aimed to develop and validate nomograms for the prediction of low muscle mass and muscle radiodensity and to examine the application value of the nomograms in the diagnoses of sarcopenia and malnutrition.

Methods: A total of 1315 patients diagnosed with gastric cancer between July 2014 and May 2019 were included. Random resampling with an 80/20 split ratio was performed to obtain a training cohort (n = 1056) and a validation cohort (n = 259). Nomograms were separately constructed for low SMI (LSMI) and low SMD (LSMD) in the training cohort based on prospectively collected preoperative data. The performance of the nomograms was assessed using the AUC, calibration curve, and Hosmer-Lemeshow test. The application values of the nomograms in the diagnoses of sarcopenia and malnutrition were also evaluated.

Results: Age, BMI, hemoglobin concentration, and gait speed were included in the nomogram for LSMI predictions. These variables, in addition to sex, were included in the nomogram for LSMD predictions. The diagnostic nomograms exhibited good discrimination, with AUCs of 0.818 (95% CI, 0.791-0.845) for the LSMI nomogram and 0.788 (95% CI, 0.761-0.815) for the LSMD diagnostic nomogram in the training cohort. Calibration was also excellent. The agreement ratios between the nomograms and actual observations in the total population were 92.3% and 95.6% for sarcopenia and malnutrition, respectively. Prognostic nomograms exhibited similar performance in the validation cohort.

Conclusions: Diagnostic nomograms consisting of preoperative factors can successfully predict LSMI and LSMD. These models facilitate early identification and timely interventions for at-risk populations.