糖尿病周围神经病变风险预测模型的系统评价  

Risk Prediction Models for Diabetic Peripheral Neuropathy:a Systematic Review

作　　者：李旭霞 王晓云 李丽霞[1] 张苗苗 寇婷媛 郑彭燕 刘三娇 吕婷 LI Xuxia;WANG Xiaoyun;LI Lixia;ZHANG Miaomiao;KOU Tingyuan;ZHENG Pengyan;LIU Sanjiao;LYU Ting(School of Nursing,Shanxi University of Chinese Medicine,Jinzhong 030619,China;Endocrinology Department,Shanxi Provincial People's Hospital,Taiyuan 030012,China;School of Nursing,Shanxi Medical University,Taiyuan 030001,China)

机构地区：[1]山西中医药大学护理学院,山西省晋中市030619 [2]山西省人民医院内分泌科,山西省太原市030012 [3]山西医科大学护理学院,山西省太原市030001

出　　处：《实用心脑肺血管病杂志》2025年第5期110-115,共6页Practical Journal of Cardiac Cerebral Pneumal and Vascular Disease

基　　金：山西省教育厅2024年度研究生实践创新项目(2024SJ338);山西中医药大学2024年研究生创新实践项目(X2024SJ024);山西省科技战略研究专项(202104031402129)。

摘　　要：目的系统评价国内外现有的糖尿病周围神经病变风险预测模型的偏倚风险、适用性。方法计算机检索英文数据库(PubMed、Embase、Web of Science、Cochrane Library)及中文数据库(中国知网、万方数据知识服务平台、维普网、中国生物医学文献服务系统)中关于糖尿病周围神经病变风险预测模型的文献,检索时限为建库至2024年11月。由两位研究人员严格按照文献纳入与排除标准独立进行文献筛选,由两位研究人员根据系统评价的批判性评估和数据提取清单(CHARMS)进行资料提取,再由两位研究人员采用预测模型偏倚风险评估工具评价纳入模型的偏倚风险和适用性。结果最终纳入21篇文献,共构建38个糖尿病周围神经病变风险预测模型,本研究仅选择各文献中性能表现最佳的模型。21个模型的候选预测变量数为6~86个;17个模型每个自变量的事件数(EPV)均<20;在模型性能评估结果方面,2个模型未进行性能评估;在模型校准方法方面,6个模型未进行校准;在模型验证方法方面,2个模型未进行验证。21个模型的整体偏倚风险均为高风险,整体适用性均为低风险。结论现有的糖尿病周围神经病变风险预测模型多数具备一定的区分能力,整体适用性较好,但整体偏倚风险较高。Objective To systematically evaluate the risk of bias and applicability of existing risk prediction models for diabetic peripheral neuropathy.Methods Studies on diabetic peripheral neuropathy risk prediction models in English database(PubMed,Embase,Web of Science,Cochrane Library)and Chinese database(CNKI,Wanfang Data,VIP and CBM)were searched by computer,and the search time was from the database establishment to November 2024.Two researchers independently screened the literature in strict accordance with the inclusion and exclusion criteria,and two researchers conducted data extraction according to the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies(CHARMS),and two researchers evaluated the bias risk and applicability of the included models by the Prediction Model Risk of Bias Assessment Tool.Results Finally,21 articles were included,and a total of 38 diabetic peripheral neuropathy risk prediction models were constructed,and only the best-performing models in each article was selected for this study.The number of candidate predictor variables of the 21 models ranged from 6 to 86,and the events per variable(EPV)of 17 models was lower than 20.In terms of model performance evaluation results,2 modes were not evaluated.In terms of model calibration method,6 modes had not been calibrated.In terms of model verification method,2 models had not been verified.The overall risk of bias for all 21 models was high,and the overall risk of applicability was low.Conclusion Most of the existing diabetic peripheral neuropathy risk prediction models have certain distinguishing ability and good overall applicability,but their overall bias risk is high.

关 键 词：糖尿病血管病变 糖尿病周围神经病变 预测模型 系统评价 

分 类 号：R587.23[医药卫生—内分泌]