Establishment and Validation of a Predictive Model for Sarcopenia Based on 2-D Ultrasound and Shear Wave Elastography in the Medial Gastrocnemius Muscle

Abstract

ObjectiveTo develop and validate a predictive model for sarcopenia. MethodsA total of 240 subjects who visited our hospital between August 2021 and May 2023 were randomly divided by time of entry into a training set containing 2/3 of patients and a validation set containing 1/3 of patients. The muscle thickness (MT), echo intensity (EI), and shear wave velocity (SWV) of the medial gastrocnemius muscle were measured. Indicators that were meaningful in the univariate analysis in the training set were included in a binary logistic regression to derive a regression model, and the model was evaluated using a consistency index, calibration plot, and clinical validity curve. Diagnostic efficacy and clinical applicability were compared between the model and unifactorial indicators. ResultsFour meaningful variables, age, body mass index (BMI), MT, and SWV, were screened into the predictive model. The model was Logit Y = 21.292 + 0.065 × Age - 0.411 × BMI - 0.524 × MT - 3.072 × SWV. The model was well differentiated with an internally validated C-index of 0.924 and an external validation C-index of 0.914. The calibration plot predicted probabilities against actual probabilities showed excellent agreement. The specificity, sensitivity, and Youden's index of the model were 73.80%, 97.40%, and 71.20%, respectively, when using the diagnostic cut-off value of >0.279 for sarcopenia. The logistic model had higher diagnostic efficacy (p < 0.001) and higher net clinical benefit (p < 0.001) over the same threshold range compared to indicators. ConclusionThe logistic model of sarcopenia has been justified to have good discriminatory, calibrated, and clinical validity, and has higher diagnostic value than indicators.