Establishment and validation of a risk prediction model for sarcopenia in gastrointestinal cancer patients: A systematic review and meta-analysis-based approach

Objective: The study aimed to develop a model to predict the risk of sarcopenia in gastrointestinal cancer patients. The goal was to identify these patients early and classify them into different risk categories based on their likelihood of developing sarcopenia. Methods: This study evaluated risk factors for sarcopenia in patients with gastrointestinal cancers through a systematic review and meta-analysis. The natural logarithm of the combined risk estimate for each factor was used as a coefficient to assign scores within the model for risk prediction. Data from 270 patients with gastrointestinal cancers, collected between October 2023 and April 2024, was used to assess the predictive performance of the scoring model. Results: The analysis included 17 studies that included 9405 patients with gastrointestinal cancers, out of which 4361 had sarcopenia. The model identified several significant predictors of sarcopenia, including age (OR = 2.45), sex (OR = 1.15), combined diabetes (OR = 2.02), neutrophil-to-lymphocyte ratio (NLR) category (OR = 1.61), TNM stage (OR = 1.61), and weight change (OR = 1.60). Model validation was performed using an external cohort through logistic regression, resulting in an area under the curve (AUC) of 0.773. This model attained a sensitivity of 0.714 and a specificity of 0.688 and ultimately selected 16.5 as the ideal critical risk score. Furthermore, an AUC of 0.770 was obtained from Bayesian model validation; the optimal critical risk score was determined to be 19.0, which corresponds to a sensitivity of 0.658 and a specificity of 0.847. Conclusions: The model of risk prediction developed through systematic review and meta-analysis demonstrates substantial for sarcopenia in patients with gastrointestinal cancers. Its clinical usability facilitates the screening of patients at high risk for sarcopenia. (c) 2024 Published by Elsevier Ltd.