Risk prediction models for diabetic peripheral neuropathy:a systematic review and Meta-analysis

ZHAO Sisi; ZHANG Chunling; CHEN Lu; ZHU Lei; LUO Zhiqin; WANG Xinghui; YANG Xuan; YAN Huimin; XIANG Jin; ZHAO Yue

doi:10.12102/j.issn.2095-8668.2025.01.002

Abstract：Objective:To systematically evaluate the quality of risk prediction models for diabetic peripheral neuropathy(DPN)and provide a reference for improving existing models and establishing new models.Methods:Literature related to the construction of risk prediction models for DPN were retrieved from PubMed,the Cochrane Library,EMbase,Web of Science,CBM,CNKI,VIP,and WanFang Database from the establishment of the database to April 1,2024.The quality of the included prediction model studies was assessed using the PROBAST tool,and the key predictors were Meta-analyzed using RevMan 5.4 software.Results:A total of 12 articles were included in this study involving 23 380 patients,and 19 DPN risk prediction models were constructed.Among them,11 articles were calibrated and 9 articles were validated.The PROBAST evaluation results showed that all 12 articles were at high risk of bias.The predictive factors mainly included basic factors,disease factors,and laboratory test indicators.Meta-analysis results showed that glycosylated hemoglobin(OR=1.54,95％CI 1.38-1.72,Z=7.55,P＜0.001),age(OR=1.10,95％CI 1.00-1.20,Z=2.06,P=0.040),and diabetic course(OR=1.22,95％CI 1.14-1.30,Z=5.95,P＜0.001)were predictive factors for DPN.Conclusions:current evidence shows that the prediction model of DPN is still in the exploratory stage.In the future,a prediction model with low bias risk and strong applicability should be selected or developed according to the actual clinical situation,and its prediction effect should be continuously monitored in practical applications.At the same time,medical staff should pay high attention to the patient's glycated hemoglobin level,age,and course of disease,in order to detect DPN earlier and take targeted prevention and treatment measures.