摘要
目的探讨特发性肺纤维化(IPF)进展的影响因素并构建预测模型。方法本研究为回顾性队列研究。收集2019年1月至2023年12月在解放军总医院第四、第六、第八医学中心住院接受治疗的220例IPF患者的基线临床资料。按7∶3划分为训练集(153例)与测试集(67例), 并依据随访结果分为进展组与稳定组, 进而比较各亚组间的基线临床资料。通过4种不同的方法(最小绝对值收敛和选择算子回归、随机森林、梯度提升机、单因素logistic回归分析)对变量进行初步筛选, 将4种方法共同选中的变量, 以及结合文献报道与临床经验知识纳入自变量, 进行多因素logistic回归分析, 寻找IPF进展的独立影响因素, 并构建IPF进展的列线图预测模型。在训练集和测试集中分别采用受试者操作特征(ROC)曲线和校准曲线评估预测模型的准确性和一致性, 采用临床决策曲线评估预测模型的临床效用。结果共纳入病例220例, 男162例(73.6%), 女58例(26.4%), 年龄(67.2±10.0)岁。训练集153例(69.5%), 其中进展组84例(54.9%), 稳定组69例(45.1%);测试集67例(30.5%), 其中进展组37例(55.2%), 稳定组30例(44.8%)。在训练集和测试集中, 与稳定组相比, 进展组吸烟、合并胃食管反流病(GERD)、合并肺动脉高压患者均占比更高, 改良英国医学研究委员会呼吸困难量表(mMRC)评分更高, 用力肺活量占预计值百分比(FVC%pred)、肺一氧化碳弥散量占预计值百分比(DLCO%pred)和白蛋白水平更低, 差异均有统计学意义(均P<0.05)。多因素logistic回归结果显示, 基线高水平DLCO%pred(OR=0.942, 95%CI:0.929~0.980, P<0.001)及FVC%pred(OR=0.956, 95%CI:0.910~0.973, P<0.001)是IPF进展的独立保护因素, 吸烟(OR=10.040, 95%CI:3.970~27.634, P<0.001)、合并GERD(OR=6.801, 95%CI:1.655~38.184, P=0.015)、高mMRC评分(OR=2.088, 95%CI:1.175~3.710, P=0.012)是IPF进展的独立危险因素。预测模型在训练集ROC曲线的曲线下面积为0.92(95%CI:0.87~0.96), 灵敏度为0.87, 特异度为0.88, 在测试集的曲线下面积为0.92(95%CI:0.84~0.99), 灵敏度为0.86, 特异度为0.80, 说明该模型具有很好的区分力。训练集及测试集所绘制的校准曲线与标准曲线均基本接近, 提示该模型有较好的校准能力。决策曲线显示训练集及测试集在较大的阈值内均有较高的临床净收益, 说明该模型有较好的临床实用性。结论基线高水平DLCO%pred及FVC%pred是IPF进展的独立保护因素, 吸烟、合并GERD、高mMRC评分是IPF进展的独立危险因素。基于这些指标构建的列线图预测模型具有较好的预测效能。
ObjectiveTo identify risk factors for progression in idiopathic pulmonary fibrosis(IPF)and to develop and validate a predictive model.MethodsThis retrospective cohort study collected baseline clinical data from IPF patients hospitalized at the Fourth,Sixth,and Eighth Medical Centers of the Chinese PLA General Hospital from January 2019 to December 2023.A total of 220 patients were split into a training set(n=153)and a test set(n=67)in a 7∶3 ratio.Patients were classified as progressive or stable based on follow-up outcomes.Baseline characteristics were compared across groups.Candidate predictors were preselected using four methods-least absolute shrinkage and selection operator regression,random forest,gradient boosting,and univariable logistic regression-and then refined by clinical judgment.Variables selected by consensus and clinical relevance were entered into a multivariable stepwise logistic regression to identify independent predictors of IPF progression.A nomogram was constructed from the final model.Model discrimination was evaluated by receiver operating characteristic(ROC)curves and area under the curve(AUC);calibration was assessed with calibration plots;and clinical utility was assessed using decision curve analysis.ResultsA total of 220 patients were enrolled(162 men[73.6%]and 58 women[26.4%];mean age 67.2±10.0 years).The cohort was split into a training set of 153 patients(69.5%)-84(54.9%)progressive and 69(45.1%)stable-and a test set of 67 patients(30.5%)-37(55.2%)progressive and 30(44.8%)stable.In both the training and test sets,compared with stable patients,those with progression had significantly higher rates of smoking history,comorbid gastroesophageal reflux disease(GERD),and comorbid pulmonary hypertension(PH);higher modified Medical Research Council(mMRC)dyspnea scale scores(more patients with scores of 1 and 2);and lower median FVC%pred,D LCO%pred,and albumin(all P<0.05).Multivariable stepwise logistic regression showed that higher baseline D LCO%pred(OR=0.942,95%CI:0.929-0.980,P<0.001)and FVC%pred(OR=0.956,95%CI:0.910-0.973,P<0.001)were independent protective factors against IPF progression.Smoking(OR=10.040,95%CI:3.970-27.634,P<0.001),comorbid GERD(OR=6.801,95%CI:1.655-38.184,P=0.015),and higher mMRC score(OR=2.088,95%CI:1.175-3.710,P=0.012)were independent risk factors for IPF progression.A predictive model was constructed from the multivariable logistic regression.In the training set the model achieved an AUC of the ROC curve of 0.92(95%CI:0.87-0.96),with sensitivity 0.87 and specificity 0.88.In the test set the AUC was 0.92(95%CI:0.84-0.99),with sensitivity 0.86 and specificity 0.80,indicating strong discriminative performance.Calibration plots for both the training and test sets closely approximated the ideal diagonal,indicating good calibration.Decision curve analysis demonstrated substantial net clinical benefit across a wide range of threshold probabilities in both sets,supporting the model's clinical utility.ConclusionHigher baseline D LCO%pred and FVC%pred were independently associated with reduced risk of IPF progression,while smoking,comorbid GERD,and higher mMRC score were independent predictors of progression.A nomogram based on multivariable stepwise logistic regression showed good discrimination and calibration,indicating satisfactory predictive performance.
作者
张丽丽
崔俊昌
韩志海
赵铁梅
孟激光
丁毅伟
王雪
丁静
张春阳
Zhang Lili;Cui Junchang;Han Zhihai;Zhao Tiemei;Meng Jiguang;Ding Yiwei;Wang Xue;Ding Jing;Zhang Chunyang(Department of Respiratory and Critical Care Medicine,the Sixth Medical Center,Chinese PLA General Hospital,Beijing 100048,China;Department of Respiratory and Critical Care Medicine,the Eighth Medical Center,Chinese PLA General Hospital,Beijing 100091,China;Department of Respiratory and Critical Care Medicine,the Fourth Medical Center,Chinese PLA General Hospital,Beijing 100048,China)
出处
《国际呼吸杂志》
2025年第12期1070-1081,共12页
International Journal of Respiration
基金
国家重点研发计划(2023YFC2507101)
解放军总医院第六医学中心创新培育基金(CXPY202309)。
关键词
肺纤维化
特发性肺纤维化
危险因素
进展
预测模型
Pulmonary fibrosis
Idiopathic pulmonary fibrosis
Risk factors
Progression
Predictive models
