摘要
背景 糖尿病肾病(DN)是糖尿病患者常见的并发症,对其发生风险进行预测与验证,有助于提前识别高风险患者并采取干预措施,以避免或延缓肾脏疾病的进展.目的 分析影响 2 型糖尿病(T2DM)患者并发DN的风险因素,构建T2DM患者发生DN风险的预测模型并进行验证.方法 选取 2016 年 1 月—2021 年 6 月在新疆医科大学第一附属医院住院的 5 810 例T2DM患者为研究对象,根据是否并发DN将患者分为DN组(481 例)和非DN组(5 329 例).对其中 481 例DN患者和非DN患者依据性别、年龄(±2 岁)进行 1∶1 病例对照匹配,将匹配后的962 例T2DM患者根据 2∶1 比例随机分为训练组(n=641)和验证组(n=321).收集患者的基础数据,如临床特征、实验室检查结果及其他相关数据.采用LASSO回归优化筛选变量,利用多因素Logistic回归分析建立列线图预测模型.分别采用受试者工作特征(ROC)曲线、Hosmer-Lemeshow校准曲线和决策曲线分析(DCA)评价预测模型的区分度、校准度以及预测模型的临床有效性.结果 DN组与非DN组患者性别、年龄、BMI、糖尿病病程、白细胞计数、总胆固醇、三酰甘油、低密度脂蛋白胆固醇、血清肌酐、高血压、收缩压、舒张压、糖化血红蛋白载脂蛋白B、24 h尿微量总蛋白、定性尿蛋白比较,差异有统计学意义(P<0.05).采用LASSO回归分析方法,筛选出 5 个与T2DM患者发生DN风险相关的预测变量,结合多因素Logistic回归分析结果显示,糖尿病病程、总胆固醇、血清肌酐、高血压、定性尿蛋白是T2DM患者并发DN的危险因素(P<0.05).训练组DN发生风险的ROC曲线下的面积(AUC)为0.866(95%CI=0.839～0.894),验证组DN发生风险的AUC为 0.849(95%CI=0.804～0.889).Hosmer-Lemeshow校准曲线拟合度较好(训练组P=0.748;验证组P=0.986).DCA显示当患者的阈值概率为 0.15～0.95 时,使用列线图预测模型预测T2DM患者发生DN风险更有益.结论 本研究发现糖尿病病程、总胆固醇、血清肌酐、高血压、定性尿蛋白可能是T2DM患者并发DN的危险因素,建立了包含该 5 个危险因素的列线图预测模型,可用于预测T2DM患者发生DN的风险.
Abstract
Background Diabetes nephropathy(DN)is a common complication of diabetes patients.The prediction and validation of its risk will help identify high-risk patients in advance and take intervention measures to avoid or delay the progress of nephropathy.Objective To analyze the risk factors affecting the complication of DN in patients with type 2 diabetes mellitus(T2DM),construct a risk prediction model for the risk of DN in T2DM patients and validate it.Methods A total of 5 810 patients with T2DM admitted to the First Affiliated Hospital of Xinjiang Medical University from January 2016 to June 2021 were selected as the study subjects and divided into the DN group(n=481)and non-DN group(n=5 329)according to the complication of DN.A 1∶1 case-control matching was performed on 481 of these DN patients and non-DN patients by gender and age(±2 years),and the matched 962 T2DM patients were randomly divided into the training group(n=641)and validation group(n=321)based on a 2∶1 ratio.Basic data of patients,such as clinical characteristics,laboratory test results and other related data,were collected.LASSO regression was applied to optimize the screening variables,and a nomogram prediction model was developed using multivariate Logistic regression analysis.The discriminability,calibration and clinical validity of the prediction model were evaluated by using the receiver operating characteristic(ROC)curve,Hosmer-Lemeshow calibration curve,and decision curve analysis(DCA),respectively.Results There were significant differences in gender,age,BMI,course of diabetes,white blood cell count,total cholesterol,triacylglycerol,low-density lipoprotein cholesterol,serum creatinine,hypertension,systolic blood pressure,diastolic blood pressure,glycosylated hemoglobin,apolipoprotein B,24-hour urinary micro total protein,qualitative urinary protein between the DN and non-DN group(P<0.05).Five predictor variables associated with the risk of DN in patients with T2DM were screened using LASSO regression analysis,and the results combined with multivariate Logistic regression analysis showed that duration of diabetes,total cholesterol,serum creatinine,hypertension,and qualitative urinary protein were risk factors for the complication of DN in T2DM patients(P<0.05).The area under the ROC curve(AUC)for the risk of DN in the training group of the model was 0.866(95%CI=0.839-0.894),and the AUC for predicting the risk of DN in the validation group was 0.849(95%CI=0.804-0.889)based on the predictor variables.The Hosmer-Lemeshow calibration curve fit was good(P=0.748 for the training group;P=0.986 for the validation group).DCA showed that the use of nomogram prediction model was more beneficial in predicting DN when the threshold probability of patients was 0.15 to 0.95.Conclusion The nomogram prediction model containing five predictor variables(diabetes duration,total cholesterol,serum creatinine,hypertension,qualitative urinary protein)developed in this study can be used to predict the risk of DN in patients with T2DM.
基金项目
新疆维吾尔自治区自然科学基金资助项目(2022D01A311)
新疆维吾尔自治区自然科学基金资助项目(2022D01C184)
国家科技期刊平台
NSTL
维普
万方数据