基于多参数的NSCLC放射性肺炎预测模型的构建及验证

Construction and validation of multi-parameter-based NSCLC radiation pneumonia prediction model

目的基于临床多种参数数据,对接受放射治疗的非小细胞肺癌(non small cell lung cancer,NSCLC)患者发生放射性肺炎(radiation pneumonia,RP)的相关因素进行分析,构建预测模型识别高危人群。方法回顾性分析2018年1月-2021年12月收治并接受放射治疗的NSCLC患者,并进行7∶3随机分为建模组及验证组。在建模组中,利用单因素及多因素Logistic分析发生RP的影响因素并建立预测模型,由验证组患者对预测模型进行效能验证。使用受试者工作特征曲线(receiver operating characteristic curve,ROC)、校准曲线(calibration curve,Cal)和决策曲线分析法(decision curve analysis,DCA)验证模型的准确性及临床获益度,使用列线图进行模型可视化。结果共纳入121例NSCLC患者,其中85例患者为建模组,36例患者为验证组。单因素及Logistic回归分析结果显示,合并高血压,周围型病灶,腺癌,GTV>60cm 3是发生RP的主要因素。建模组及验证组的ROC曲线下面积(AUC)分别为0.731和0.717。两组的校准曲线均显示出高效诊断的高稳定性。此外,DCA曲线表明该模型具有令人满意的正净获益。结论合并高血压,周围型病灶,腺癌及GTV>60cm 3联合可用于建立晚期NSCLC患者放射性肺炎的预测模型,本研究构建的预测模型可为临床早期筛选高风险患者提供参考。

Objective To analyze the factors associated with the development of radiation pneumonia(RP)in patients with non-small cell lung cancer(NSCLC)treated with radiation therapy based on clinical data of various parameters and to construct a prediction model to identify high-risk groups.Methods Patients with NSCLC admitted and treated with radiation therapy from January 2018 to December 2021 were retrospectively analyzed and randomized 7:3 into the modeling and validation groups.In the modeling group,the factors influencing the occurrence of RP were analyzed using univariate and multifactorial logistic analysis and a prediction model was developed,and the efficacy of the prediction model was validated by patients in the validation group.The accuracy and clinical benefit of the model were verified using receiver operating characteristic curve(ROC),calibration curve(Cal),and decision curve analysis(DCA).The model was visualized using line plots.Results A total of 121 patients with NSCLC were included,85 of which were in the modeling group and 36 cases in the validation group.Univariate and logistic regression analysis showed that combined hypertension,peripheral lesions,adenocarcinoma,GTV>60 and intensity-modulated radiotherapy were the main factors for the occurrence of RP.The area under the ROC curve(AUC)was 0.735 and 0.786 in the modeling and validation groups,respectively.Calibrated curves for both groups showed high stability for efficient diagnosis.In addition,the DCA curves indicated a satisfactory positive net benefit of the model.Conclusion The combination of hypertension,peripheral lesions,adenocarcinoma,GTV>60 and intensity-modulated radiotherapy can be used to build a predictive model for radiation pneumonia in patients with advanced NSCLC,and the predictive model constructed in this study can provide a reference for early clinical screening of high-risk patients.