女性原发性干燥综合征患者嗅觉障碍危险因素及预测模型构建

Risk factors and prediction model construction of olfactory dysfunction in female patients with primary Sjögren′s syndrome

目的探讨女性原发性干燥综合征(pSS)患者嗅觉障碍(OD)的危险因素并构建预测模型。方法纳入2021年1月至2023年1月湖州市第三人民医院就诊的女性pSS患者252例,根据嗅觉功能分为有OD组(pSS-OD组)144例和无OD组108例(无pSS-OD组),比较2组各因素的差别,以多因素logistic回归分析筛选OD发生的独立危险因素做为预测变量构建风险列线图预测模型。依靠受试者工作特征(ROC)曲线、Hosmer-Lemeshow拟合优度及决策曲线分析(DCA)检验验证和评价模型的区分度、校准度及临床有效性。结果pSS患者中OD发生率为57.1%(144/252),OD以嗅觉减退为主(97.2%,140/144)。单因素分析显示pSS-OD组、无pSS-OD组在病程(t=-2.05,P=0.040)、RF(t=2.90,P=0.004)、血清IgG(t=4.41,P=0.001)、补体C3(t=5.47,P=0.001)、干燥综合征患者报告指数(ESSPRI)(t=2.55,P=0.011)和病情评估指数(ESSDAI)评分(t=3.80,P=0.001)、汉密尔顿抑郁量表(HAMD)评分(t=3.38,P=0.001)上差异均有统计学意义。二元多因素logistic回归分析显示高滴度RF、高血清IgG、低补体C3、高ESSDAI评分和高HAMD评分是pSS-OD发生的独立危险因素[OR值(95%CI)分别为1.01(1.01,1.03)、1.21(1.06,1.38)、0.98(0.96,0.89)、1.52(1.20,1.92)、1.13(1.03,1.23),P值分别为0.016、0.005、0.004、<0.001、0.007]。基于上述独立危险因素构建的临床预测模型,建模组和验证组的ROC曲线下面积分别为0.83(0.77,0.89)和0.69(0.57,0.82),Hosmer-Lemeshow拟合优度检验表明预测模型具有较好的校准能力(P=0.083),DCA曲线显示模型具有较好的临床价值。结论本研究建立的风险列线图临床预测模型可以用来评估pSS-OD发生的风险,有助于早期发现和及时干预,改善pSS-OD患者的生活质量。

Objective To investigate the risk factors of olfactory dysfunction(OD)of female patients with primary Sjögren′s syndrome(pSS)and a prediction model was constructed.Methods A total of 252 female pSS patients in the Third Municipal Hospital of Huzhou,from January 2021 to 2023 were recruited.According to the olfactory function,they were divided into OD group(144 cases)and without OD group(108 cases).The independent risk factors of OD were evaluated and the clinical prediction efficiency of the model was analyzed by receiver operating characteristic(ROC)curve,Hosmer-Lemeshow goodness-of-fit test and the decision curve analysis(DCA)curve.Results The rate of OD was 57.1%((144/252))in female pSS patients.Hyposmia was the main olfactory disorder in OD group(140/144,97.2%).Univariate analysis showed that there were significant differences between OD and without OD groups,including disease course(t=-2.05,P=0.040),RF(t=2.90,P=0.004),IgG(t=4.41,P=0.001),C3(t=5.47,P=0.001),ESSPRI(t=2.55,P=0.011),ESSDAI(t=3.80,P=0.001),HAMD(t=3.38,P=0.001).Logistic regression analysis showed that high serum RF,IgG and low complement 3,high scores of ESSDAI and HAMD high scores were independent risk factors for pSS patients with OD patients[OR(95%CI)=1.01(1.01,1.03),1.21(1.06,1.38),0.98(0.96,0.89),1.52(1.20,1.92),1.13(1.03,1.23)](P value were 0.016,0.005,0.004,<0.001 and 0.007 respectively).Based on the above independent risk factors,we constructed the prediction model nomogram and performed the internal data validation.The ROC curve AUC(95%CI)of the modeling group was 0.83(0.77,0.89)and validation group was 0.69(0.57,0.82),the prediction model was well differentiated.Hosmer-Lemeshow goodness-of-fit test showed that the prediction model had good calibration ability(P=0.083),and the DCA curve display model is clinically valuable.Conclusion The clinical prediction model developed in this study can be used to assess the risk of pSS-OD development,which is helpful for early detection and timely interention so can improve the quality of life of pSS-OD patients in turn.