原发性干燥综合征合并间质性肺病患者预后不良的预警模型构建和验证

Construction and validation of an early warning model for poor prognosis in patients with primary Sjögren's syndrome complicated with interstitial lung disease

摘要:

目的 探究原发性干燥综合征（pSS）合并间质性肺病（ILD）患者预后不良的影响因素，构建预警模型并进行验证。方法 回顾性选取2019年3月—2024年3月开滦总医院收治的243例pSS合并ILD患者，按随机化原则，根据8∶2分配定律将其分为训练集（194例）与验证集（49例）。对患者进行至少1年的随访，根据有无发生疾病进展或死亡观察患者预后情况。收集患者相关资料并进行单因素分析，采用多因素逐步Logistic回归模型分析pSS合并ILD患者预后不良的影响因素，并以此构建列线图预警模型，绘制受试者工作特征（ROC）曲线进行预警模型的验证。结果 243例患者中死亡22例、疾病进展62例，预后不良总发生率为34.57%（84/243），其中训练集预后不良66例、验证集预后不良18例。预后不良组的年龄、CA19-9水平、KL-6水平、LDH水平、铁蛋白水平、低蛋白血症率和HRCT蜂窝影率均高于预后良好组，预后不良组FVC%pred水平、DLCO%pred水平、25（OH）D水平均低于预后良好组（P <0.05）。多因素逐步Logistic回归分析结果显示：年龄高［O^R =1.184（95% CI：1.073，1.306）］、DLCO%pred水平低［O^R =0.901（95% CI：0.856，0.950）］、KL-6水平高［O^R =1.009（95% CI：1.004，1.015）］、25（OH）D水平低［O^R =0.795（95% CI：0.710，0.890）］、低蛋白血症［O^R =4.751（95% CI：1.451，15.555）］、HRCT蜂窝影［O^R=23.963（95% CI：5.714，100.494）］均是患者预后不良的危险因素（P <0.05）。基于多因素逐步Logistic回归分析结果构建的列线图，C-index指数0.842，校正与理想曲线趋近（P <0.05）。训练集ROC曲线下面积（AUC）为0.876，特异性为0.836、敏感性为0.803；验证集ROC曲线的AUC为0.865，特异性为0.811、敏感性为0.759。结论 基于年龄高、DLCO% pred水平低、KL-6水平高、25（OH）D水平低、低蛋白血症、HRCT蜂窝影构建的预警模型可有效预测患者预后。

Abstract:

Objective To explore the influencing factors of poor prognosis in patients with primary Sj?gren's syndrome (pSS) complicated with interstitial lung disease (ILD), and to construct and validate an early warning model.Methods A total of 243 patients with pSS complicated with ILD admitted to Kailuan General Hospital from March 2019 to March 2024 were retrospectively selected. According to the 8:2 allocation rule based on the randomization principle, they were divided into the training set (194 cases) and the validation set (49 cases). All patients were followed up for at least 1 year, and the prognosis was observed based on the occurrence of disease progression or death. Relevant clinical data of patients were collected for univariate analysis. Multivariate stepwise Logistic regression model was used to analyze the influencing factors of poor prognosis in patients with pSS complicated with ILD, based on which a nomogram early warning model was constructed. Receiver operating characteristic (ROC) curve was plotted to validate the early warning model.Results Among the 243 patients, 22 cases died and 62 cases had disease progression, with a total incidence of poor prognosis of 34.57% (84/243), including 66 cases of poor prognosis in the training set and 18 cases in the validation set. Compared with the good prognosis group, the poor prognosis group had significantly higher age, carbohydrate antigen 19-9 (CA19-9) level, Krebs von den Lungen-6 (KL-6) level, lactate dehydrogenase (LDH) level, ferritin level, hypoproteinemia rate and high-resolution computed tomography (HRCT) honeycomb shadow rate, while significantly lower forced vital capacity as a percentage of predicted value (FVC%pred) level, carbon monoxide diffusion capacity as a percentage of predicted value (DLCO%pred) level and 25-hydroxyvitamin D [25(OH)D] level (all P < 0.05). The results of multivariate stepwise Logistic regression analysis showed that advanced age [O^R = 1.184 (95% CI: 1.073, 1.306) ], low DLCO%pred level [O^R = 0.901 (95% CI: 0.856, 0.950) ], high KL-6 level [O^R = 1.009 (95% CI: 1.004, 1.015) ], low 25(OH)D level [O^R = 0.795 (95% CI: 0.710, 0.890) ], hypoproteinemia [O^R = 4.751 (95% CI: 1.451, 15.555) ] and HRCT honeycomb shadow [O^R = 23.963 (95% CI: 5.714, 100.494) ] were all risk factors for poor prognosis in these patients (all P < 0.05). The C-index of the nomogram constructed based on the results of multivariate stepwise Logistic regression analysis was 0.842, and the calibration curve was close to the ideal curve (P < 0.05). The area under the ROC curve (AUC) of the training set was 0.876, with a specificity of 0.836 and a sensitivity of 0.803; the AUC of the validation set was 0.865, with a specificity of 0.811 and a sensitivity of 0.759.Conclusion The early warning model constructed based on advanced age, low DLCO%pred level, high KL-6 level, low 25(OH)D level, hypoproteinemia and HRCT honeycomb shadow can effectively predict the prognosis of patients with pSS complicated with ILD.