2型糖尿病患者代谢相关脂肪性肝病列线图预测模型的建立与验证

Development and validation of a predicting nomogram for metabolic associated fatty liver disease in patients with type 2 diabetes mellitus

目的构建和验证2型糖尿病(type 2 diabetes mellitus,T2DM)患者代谢相关脂肪性肝病(metabolic associated fatty liver disease,MAFLD)的个体化预测模型。方法回顾性收集2020年1月至2022年12月上海市徐汇区大华医院收治的1592例T2DM患者的临床资料。根据超声/影像学检查结果将患者分为MAFLD组和非MAFLD组,比较两组患者的一般临床特征。采用二元Logistic回归构建和验证MAFLD的影响因素并建立列线图模型。采用受试者工作特征(receiver operating characteristic,ROC)曲线和一致性指数(concordance index,C-index)评价列线图模型区分度。采用bootstrap法验证列线图模型的拟合度和校准度,采用一致性曲线评估模型的预测效能,采用决策曲线分析(decision curve analysis,DCA)评价模型是否临床获益。结果纳入研究的1592例T2DM患者中,1085例(68.2%)患者为MAFLD,507例(31.8%)患者为非MAFLD。MAFLD组患者年龄(中位数:74岁比73岁)、有饮酒史患者比例[17.6%(191/1085)比13.0%(66/507)]、体重指数(中位数:22.5 kg/m^(2)比22.0 kg/m^(2))、24 h动态舒张压(中位数:76 mmHg比74 mmHg)、糖尿病持续时间[(14.9±10.2)年比(10.7±8.3)年]、糖化血红蛋白(中位数:8.5%比7.9%)、总胆固醇(中位数:5.2 mmol/L比4.8 mmol/L)、甘油三酯[(2.1±1.2)mmol/L比(1.8±1.2)mmol/L]和尿素氮[(7.1±3.1)mmol/L比(6.8±2.7)mmol/L]水平高于非MAFLD组,高密度脂蛋白胆固醇[(1.1±0.5)mmol/L比(1.2±0.4)mmol/L]、25-羟基维生素D(中位数:37.5 nmol/L比40.5 nmol/L)和血糖在目标范围内时间(中位数:75%比79%)低于非MAFLD组(P均<0.05)。多因素Logistic分析表明,饮酒史(OR=2.44,95%CI:1.56~3.83,P<0.001)、体重指数(OR=1.06,95%CI:1.01~1.12,P=0.045)、糖尿病病程(OR=1.05,95%CI:1.02~1.07,P=0.010)、糖化血红蛋白(OR=1.08,95%CI:1.01~1.17,P=0.032)、总胆固醇(OR=1.43,95%CI:1.23~1.65,P<0.001)、25-羟基维生素D(OR=0.91,95%CI:0.86~0.96,P=0.001)以及血糖在目标范围内时间(OR=0.98,95%CI:0.97~0.99,P<0.001)与T2DM人群发生MAFLD独立相关。将Logistic模型筛选的因素构建列线图模型,ROC曲线下面积为0.77(95%CI:0.71~0.83),C-index为0.76(95%CI:0.72~0.81)。一致性曲线显示预测概率与实际观察值一致性良好,DCA提示模型具有良好的临床获益。结论T2DM人群中,MAFLD患者更可能存在不良生活方式和代谢障碍性因素,更易存在25-羟基维生素D缺乏和血糖在目标范围内时间低。该模型可用于T2DM人群中MAFLD的预测。

Objective To construct and validate an individualized prediction model for metabolic associated fatty liver disease(MAFLD)in patients with type 2 diabetes mellitus(T2DM).Methods The clinical data of 1592 patients with T2DM from January 2020 to December 2022 in Shanghai Xuhui Dahua Hospital were retrospectively collected.Patients were divided into MAFLD group and non-MAFLD group according to ultrasound/imaging findings,and the general clinical characteristics of the two groups were compared.Multivariate Logistic regression was used to construct and validate independent risk factors for MAFLD and a nomogram model was established.The discrimination accuracy of the nomogram model was evaluated by receiver operating characteristic(ROC)curve and the concordance index(C-index).The fitting and calibration properties of the nomogram model were verified by the bootstrap method,and the predictive accuracy of the model was evaluated by the calibration curve.The decision curve analysis(DCA)was used to determine the clinical usefulness of the model.Results Of the included 1592 T2DM patients,1085 cases(68.2%)were diagnosed as MAFLD and 507 cases(31.8%)were diagnosed as non-MAFLD.Age(median:74 years vs.73 years),rates of alcohol consumption[17.6%(191/1085)vs.13.0%(66/507)],body mass 2 index(median:22.5 kg/m vs.22.0 kg/m^(2)),24-hour average diastolic blood pressure(median:76 mmHg vs.74 mmHg),duration of T2DM[(14.9±10.2)years vs.(10.7±8.3)years],glycated hemoglobin(median:8.5%vs.7.9%),total cholesterol(median:5.2 mmol/L vs.4.8 mmol/L),triglycerides[(2.1±1.2)mmol/L vs.(1.8±1.2)mmol/L]and blood urea nitrogen[(7.1±3.1)mmol/L vs.(6.8±2.7)mmol/L]of patients in MAFLD group were significantly higher than those in non-MAFLD group,while high-density lipoprotein[(1.1±0.5)mmol/L vs.(1.2±0.4)mmol/L],25-hydroxyvitamin D(median:37.5 nmol/L vs.40.5 nmol/L)and time in range based on the continuous glucose monitoring(median:75%vs.79%)were significantly lower(all P<0.05).Multivariate Logistic regression analysis showed that a history of alcohol consumption(OR=2.44,95%CI:1.56~3.83,P<0.001),body mass index(OR=1.06,95%CI:1.01~1.12,P=0.045),duration of T2DM(OR=1.05,95%CI:1.02~1.07,P=0.010),glycated hemoglobin(OR=1.08,95%CI:1.01~1.17,P=0.032),total cholesterol(OR=1.43,95%CI:1.23~1.65,P<0.001),25-hydroxyvitamin D(OR=0.91,95%CI:0.86~0.96,P=0.001)and time in range based on the continuous glucose monitoring(OR=0.98,95%CI:0.97~0.99,P<0.001)were independently associated with MAFLD in T2DM patients.A nomogram model was constructed using factors selected from the Logistic model,with an area under the ROC curve of 0.77(95%CI:0.71~0.83)and a C-index of 0.76(95%CI:0.72~0.81).The calibration curve indicated a good agreement between the predicted and observed probabilities,and DCA suggested that the model had significant clinical benefits.Conclusions Among individuals with T2DM,those with MAFLD were more likely to exhibit unfavorable lifestyle and metabolic disorders,as well as a higher likelihood of 25-hydroxyvitamin D deficiency and shorter time in range based on the continuous glucose monitoring.This predictive model can be utilized to identify the presence of MAFLD in T2DM population.