Risk prediction models for post-stroke sarcopenia: A systematic review and appraisal

Several risk prediction models for sarcopenia in stroke patients have been developed. However, their methodological quality and predictive performance remain uncertain. To systematically review and critically appraise the current risk prediction models for sarcopenia in stroke patients. A systematic review conducted in accordance with the PRISMA guidelines, with methodological quality assessed using the Prediction Model Risk of Bias Assessment Tool (PROBAST). PubMed, the Cochrane Library, Web of Science, Embase, CINAHL, ProQuest, Medline, CNKI, Wanfang, VIP, and CBM databases were systematically searched for studies indexed until May 8, 2025. Two reviewers independently screened studies, extracted data using the CHARMS checklist, and assessed risk of bias and applicability using PROBAST. Model performance was evaluated using metrics including the area under the curve(AUC). A total of 8 studies involving 11 models were included. The area under the AUC for all the models was > 0.6 (AUC range: 0.681–0.970) in development cohorts. However, 6 studies (75.0 %) were rated as having a high or unclear risk of bias, primarily due to issues in the analysis domain (e.g., insufficient sample size, lack of external validation). In contrast, the overall applicability of the models was generally favorable. Common predictors included age, BMI, nutritional status (MQSGA score), bone density, and the number of stroke episodes. Risk prediction models for sarcopenia in stroke patients generally have good applicability, although they present significant methodological flaws and a high risk of bias. Future research should focus on the transparent reporting of multivariable prediction models to develop and assess sarcopenia risk scores for stroke patients, and internal and external validation is needed in clinical practice. The review protocol was registered at PROSPERO (CRD420251036935).