2型糖尿病患者周围神经病变预测模型的构建

Development of a prediction model for diabetic peripheral neuropathy in type 2 diabetes mellitus

目的比较不同检测方法基于局部热充血的2型糖尿病周围神经病变预测模型性能,为临床早期筛查提供参考。方法纳入174例2型糖尿病患者,在试验局部加热至44℃并使用激光多普勒血流仪测量血流变化率。同时进行临床5项筛查、震动感觉阈值测定和神经传导检查,根据各自的诊断标准,采用多因素logistic回归分析探讨糖尿病周围神经病变的影响因素。以神经传导检查为“金标准”,筛选最优模型构建列线图预测模型。结果3种检测方法均提示,血流变化率、糖尿病病程、糖尿病肾病、糖化血红蛋白是糖尿病周围神经病变的影响因素(均P<0.05)。根据临床5项筛查结果建立的logit模型性能最好,据此构建的列线图模型具有较好的准确度(Hosmer-Lemeshow检验χ^(2)=11.147,P>0.05)和区分度(AUC=0.872)。结论血流变化率、糖尿病病程、糖尿病肾病和糖化血红蛋白是糖尿病周围神经病变的影响因素。以临床5项筛查结果构建的列线图模型具有良好的诊断效能,可为临床筛选、识别糖尿病周围神经病变患者提供参考。

Objective To compare the performance of different detection methods for constructing a prediction model for diabetic peripheral neuropathy(DPN)in type 2 diabetes patients based on local thermal hyperemia,so as to provide reference for early clinical screening.Methods A total of 174 patients with type 2 diabetes were included in the study.Local skin heating was applied to induce local thermal hyperaemia with a temperature of 44℃,and laser Doppler flowmetry was used to measure dermal blood flow change.Simultaneously,clinical examination(pressure perception assessed by 10-g Semmes-Weinstein monofilament,vibration perception by 128-Hz tuning fork,discrimination by pin prick,thermal perception by Tip-therm,and reflexes by Achilles tendon reflexes),vibration perception threshold measurement,and nerve conduction study were performed to explore the influencing factors of DPN using multivariate logistic regression analysis based on their respective diagnostic criteria.Using nerve conduction study as the"gold standard"for DPN diagnosis,the optimal model was selected to construct the nomogram prediction model.Results All three detection methods indicated that the rate of dermal blood flow change,duration of diabetes,diabetic nephropathy,and glycosylated hemoglobin were the influencing factors for DPN(all P<0.05).The performance of the logit model resulting from clinical examination was the best,and the nomogram model constructed based on the results of clinical examination showed good accuracy(Hosmer-Lemeshow testχ^(2)=11.147,P>0.05)and discrimination(AUC=0.872).Conclusion The rate of dermal blood flow change,duration of diabetes,diabetic nephropathy,and glycosylated hemoglobin are influencing factors for DPN.The nomogram model constructed based on the results of clinical examinationhas good diagnostic efficiency,which can be used for clinical screening and identification of DPN in patients with type 2 diabetes.