预测食管癌调强适形放疗致放射性肺损伤的列线图模型的建立与验证

Establishment and Verification of Nomogram Model in Predicting Radiation-induced Lung Injury Caused by Intensity-modulated Conformal Radiotherapy for Esophageal Cancer

[目的]建立预测食管癌调强适形放疗致放射性肺损伤(radiation-induced lung toxicity,RILT)的列线图模型,并对模型的预测效能进行验证。[方法]选取2015年12月至2017年12月间行放射治疗的207例食管癌患者作为研究对象,采用简单交叉验证法随机将患者分为训练集(n=153)和验证集(n=54),回顾性分析训练集患者的临床资料,使用Logistic回归分析发生≥2级RILT的独立危险因素,并建立相关列线图预测模型。[结果]同期化疗、V5≥60%、V10≥40%、V20≥28%、V30≥20%、肺Dmean≥13Gy是食管癌调强适形放疗致≥2级RILT的独立危险因素(P<0.05)。基于以上6项独立危险因素建立预测食管癌调强适形放疗致≥2级RILT的列线图模型,并对该模型进行验证。训练集和验证集的C-index分别为0.831(95%CI:0.794~0.882)和0.784(95%CI:0.737~0.824),校正曲线均与标准曲线拟合较好,受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under curve,AUC)分别为0.917和0.890,表明该列线图模型具有良好的预测能力。[结论]基于危险因素建立的列线图模型具有良好的预测能力,可为临床筛查高风险患者和进一步改善患者预后提供参考依据。

[Objective]To establish a nomogram model in predicting radiation-induced lung toxicity(RILT)induced by intensity-modulated conformal radiotherapy(IMRT)for esophageal cancer,and to verify the predictive efficacy of the model.[Methods]The patients with esophageal cancer who underwent radiotherapy between December 2015 and December 2017 were selected as the research objects.The patients were randomly divided into training set(n=153)and verification set(n=54)by simple cross validation method.The clinical data of the patients in the training set were retrospectively analyzed,and Logistic regression was used to analyze the independent risk factors of RILT≥grade 2,and the related nomogram prediction model was established.[Results]Simultaneous chemotherapy,V5≥60%,V10≥40%,V20≥28%,V30≥20%,lung Dmean≥13 Gy were independent risk factors for≥grade 2 RILT caused by IMRT for esophageal cancer(P<0.05).Based on the above 6 independent risk factors,a nomogram model was established to predict RILT of≥grade 2 caused by IMRT,and the model was validated.The results show that the C-index indexes of the training set and the validation set are 0.831(95%CI:0.794~0.882)and 0.784(95%CI:0.737~0.824),respectively.The calibration curve fits well with the standard curve.The area under curve(AUC)of receiver operating characteristic(ROC)curve was 0.917 and 0.890,respectively,indicating that the nomogram model had good predictive ability.[Conclusion]The nomogram model established in this study based on risk factors has good predictive ability and can provide a reference for clinical screening of high-risk patients and further improve the prognosis of patients.