糖尿病周围神经病变诊断预测模型的建立和验证

Establishment and validation of a predictive model for the diagnosis of diabetic peripheral neuropathy

目的:建立糖尿病周围神经病变(diabetic peripheral neuropathy,DPN)的预测模型并验证其效能。方法:回顾性收集2017年1月—2021年10月在揭阳市人民医院住院的260例2型糖尿病患者的临床资料。根据纳入时间排序,将患者按照7∶3的比例分为训练集(183例)和验证集(77例)。训练集中男性72例,女性111例,年龄(61.9±9.1)岁,验证集中男性27例,女性50例,年龄(63.1±8.6)岁。所有患者分为DPN组和非DPN组。通过Lasso回归和多因素logistic回归分析筛选DPN的相关因素,构建临床因素联合肌电图评分列线图预测模型。采用受试者操作特征曲线下面积(area under curve,AUC)评价模型的区分度,采用校准曲线评估模型预测概率与实际结果的一致性,采用决策分析曲线评估模型的临床实用性。结果:logistic多因素分析显示吸烟(OR=7.851,95%CI:2.624~23.489)、病程(OR=1.016,95%CI:1.010~1.022)、收缩压(OR=1.018,95%CI:1.001~1.035)、血红蛋白(OR=0.974,95%CI:0.953~0.995)是DPN的相关因素。该模型的AUC在训练集和验证集分别为0.957、0.944,校准曲线表明模型预测和实际风险的一致性良好,决策分析曲线表明该模型具有临床应用价值。结论:构建的临床因素联合肌电图评分列线图预测模型有较高的准确度和临床适用性,有助于DPN的早期诊断。

Objective:To develop a predictive model for diabetic peripheral neuropathy(DPN)and validate its efficacy.Methods:Clinical data of 260 patients with type 2 diabetes mellitus hospitalized in Jieyang Peoples Hospital from January 2017 to October 2021 were retrospectively collected.The patients were divided into training set(183 cases)and validation set(77 cases)in a ratio of 7∶3 according to the order of inclusion time.There were 72 males and 111 females in the training set,aged(61.9±9.1)years,and 27 males and 50 females in the validation set,aged(63.1±8.6)years.All patients were divided into DPN and non-DPN groups.Relevant factors of DPN were screened by Lasso regression and multivariate logistic regression analysis,and the clinical factors combined with electromyography score nomogram prediction model was constructed.The area under curve(AUC)of receiver operating characteristic was used to evaluate the discrimination of the model,the calibration curve was used to assess the consistency of the models predicted probability with the actual results,and the decision analysis curve was used to assess the clinical utility of the model.Results:Multivariate logistic regression analysis showed that smoking(OR=7.851,95%CI:2.624-23.489),disease duration(OR=1.016,95%CI:1.010-1.022),systolic blood pressure(OR=1.018,95%CI:1.001-1.035),hemoglobin(OR=0.974,95%CI:0.953-0.995)were correlates of DPN.The AUC of the model was 0.957 and 0.944 in the training set and validation set,respectively,and the calibration curve showed good agreement between the model prediction and the actual risk,and the decision216 analysis curve indicated that the model had clinical application.Conclusion:The constructed nomogram prediction model of clinical factors combined with electromyography score has high accuracy and clinical applicability,which helps in the early diagnosis of DPN.