METHODS: This study used the Systematic Reviews and Meta-Analysis statement (PRISMA) for systematic review.
METHODS: PubMed, Web of Science, Embase, Cochrane Library and Medline databases up to September 2023.
METHODS: Risk of bias (ROB) was evaluated using the Prediction model Risk Of Bias ASsessment Tool (PROBAST). Random effect models were calculated due to high heterogeneity identified.
RESULTS: Fifteen models from twelve studies were analyzed. All studies had high ROB and three of them posed a high risk in terms of applicability. The pooled AUC, sensitivity, and specificity were 0.715, 0.583 and 0.656 respectively. The diagnostic criteria (P=0.0046), country (P=0.0046), and study design (P=0.0087) were significant sources of the heterogeneity. Analysing purely from the data perspective, grouping by diagnostic criterias, the AUC and specificity [(0.773, 95% CI 0.12-0.99, (0.652, 95% CI 0.641-0.664)] of the Asian Working Group for Sarcopenia (AWGS) group was lower than the European Working Group on Sarcopenia in Older People (EWGSOP) group [(0.859, 95% CI 0.12-1.00), (0.874, 95% CI 0.803-0.926)]. Grouping by styles of research, the AUC, sensitivity, and specificity in development group [(0.890, 95% CI 0.16-1.00), (0.751, 95% CI 0.697-0.800), (0.875, 95% CI 0.854-0.895)] were all higher than validation group [(0.715, 95% CI 0.09-0.98), (0.550, 95% CI 0.524-0.576), (0.617, 95% CI 0.604-0.629)].
CONCLUSIONS: Moving forward, there is a critical need to create low-ROB, high-applicability, and more accurate sarcopenia prediction models for MHD patients, customized for diverse global populations.
方法:本研究采用系统评价和Meta分析(PRISMA)进行系统评价。
方法:PubMed,WebofScience,Embase,截至2023年9月,Cochrane图书馆和Medline数据库。
方法:使用预测模型偏差风险评估工具(PROBAST)评估偏差风险(ROB)。由于确定的高度异质性,计算了随机效应模型。
结果:分析了来自12项研究的15个模型。所有研究都有高ROB,其中三项在适用性方面存在高风险。合并的AUC,灵敏度,特异性分别为0.715、0.583和0.656。诊断标准(P=0.0046),国家(P=0.0046),和研究设计(P=0.0087)是异质性的重要来源。纯粹从数据角度分析,按诊断标准分组,亚洲肌肉减少症工作组(AWGS)组的AUC和特异性[(0.773,95%CI0.12-0.99,(0.652,95%CI0.641-0.664)]低于欧洲老年人肌肉减少症工作组(EWGSOP)组[(0.859,95%CI0.12-1.00),(0.874,95%CI0.803-0.926)]。按研究风格分组,AUC,灵敏度,和发展组的特异性[(0.890,95%CI0.16-1.00),(0.751,95%CI0.697-0.800),(0.875,95%CI0.854-0.895)]均高于验证组[(0.715,95%CI0.09-0.98),(0.550,95%CI0.524-0.576),(0.617,95%CI0.604-0.629)]。
结论:展望未来,迫切需要创造低ROB,高适用性,和更准确的MHD患者的肌肉减少症预测模型,为不同的全球人口定制。