NSCLC患者发生免疫检查点抑制剂相关性肺炎的危险因素分析和预测模型构建

Risk factors and prediction model for immune checkpoint inhibitor-related pneumonitis in non-small cell lung cancer patients

目的探讨接受免疫治疗的非小细胞肺癌(NSCLC)患者发生免疫检查点抑制剂相关性肺炎(CIP)的危险因素,并运用相关危险因素构建Nomograms预测模型。方法本研究为观察性研究,采用非随机抽样的方法选取2019年12月1日至2021年12月1日于武汉大学人民医院接受单药免疫治疗或免疫治疗联合化疗的NSCLC患者411例为研究对象。采取简单随机抽样的方法将411例NSCLC患者按照6∶4分为训练组(247例)和验证组(164例)。根据是否发生CIP将训练组NSCLC患者分为CIP组(28例)和非CIP组(219例)。比较训练组和验证组、CIP组和非CIP组NSCLC患者的人口学特征和临床资料,包括性别、年龄、吸烟史、饮酒史、白细胞计数、是否发生CIP、病理类型、既往基础疾病史、肺部放射性肺炎史、既往肺部疾病史和治疗方案。多因素logistic回归分析NSCLC患者发生CIP的独立危险因素。根据独立危险因素构建Nomograms预测模型,并验证其内部准确性和外部实用性。结果训练组NSCLC患者男性比例高于验证组[61.1%(151/247)比48.8%(80/164),χ^(2)=6.11,P=0.013],其余临床资料2组间比较差异均无统计学意义(均P>0.05)。CIP组NSCLC患者男性、≥59岁者、有肺部放射性肺炎史者、有既往肺部疾病史者、免疫治疗联合化疗者比例均高于非CIP组[82.1%(23/28)比58.4%(128/219),85.7%(24/28)比65.8%(144/219),39.3%(11/28)比2.7%(6/219),46.4%(13/28)比15.5%(34/219),78.6%(22/28)比58.9%(129/219),均P<0.05]。男性、有肺部放射性肺炎史、有既往肺部疾病史、免疫治疗联合化疗是训练组NSCLC患者发生CIP的独立危险因素(OR值分别为3.124、10.039、8.831、3.131,均P<0.05)。根据4个独立危险因素构建Nomograms预测模型,训练组中,Hosmer-Lemeshow检验结果P=0.956,受试者工作特征曲线的曲线下面积为0.890(95%CI:0.830~0.950),校准曲线中,校准前后的两条曲线与理想预测曲线吻合较好,上下波动范围较小,其决策曲线分析曲线在3%~96%的阈值概率范围内。验证组中,曲线下面积为0.793(95%CI:0.709~0.876),校准前后两条曲线在理想预测曲线上下的波动范围较大,决策曲线分析曲线中可获得正向净收益的阈值概率范围为4%~66%。结论男性、有肺部放射性肺炎史、有既往肺部疾病史、免疫治疗联合化疗是接受免疫治疗的NSCLC患者发生CIP的独立危险因素。根据这4种独立危险因素建立的Nomograms预测模型对于CIP的发生具有一定的预测能力。

ObjectiveTo investigate the risk factors for immune checkpoint inhibitor-related pneumonitis(CIP)in patients with non-small cell lung cancer(NSCLC),and to construct a Nomograms prediction model using the related risk factors.MethodsIt was an observational study involving 411 NSCLC patients who received monotherapy of immunotherapy or immunotherapy combined with chemotherapy in Renmin Hospital of Wuhan University from December 1,2019 to December 1,2021 selected by non-random sampling.They were divided into the training group(247 cases)and validation group(164 cases)according at a ratio of 6∶4 by simple random sampling.NSCLC patients in the training group were divided into CIP group(28 cases)and non-CIP group(219 cases)according to the presence or absence of CIP.Demographic characteristics and clinical data of NSCLC patients in the training group versus validation group,and CIP group versus non-CIP group were compared,including gender,age,smoking history,alcohol consumption history,white blood cell count,CIP,pathological type,history of underlying disease,history of lung radiation pneumonia,history of lung diseases,and treatment regimen.Multivariate logistic regression analysis was performed to identify independent risk factors for CIP in NSCLC patients.A Nomograms prediction model was constructed based on independent risk factors and the internal accuracy and external practicality were evaluated.ResultsThe proportion of male NSCLC patients in the training group was significantly higher than that of the validation group(61.1%[151/247]vs 48.8%[80/164],χ^(2)=6.11,P=0.013).There were no significant differences in other clinical data between groups(all P>0.05).The proportions of males(82.1%[23/28]vs 58.4%[128/219]),≥59 years of age(85.7%[24/28]vs 65.8%[144/219]),history of lung radiation pneumonitis(39.3%[11/28]vs 2.7%[6/219]),history of lung diseases(46.4%[13/28]vs 15.5%[34/219])and immunotherapy combined with chemotherapy(78.6%[22/28]vs 58.9%[129/219])in CIP group were significantly higher than those of non-CIP group(all P<0.05).Male gender,history of lung radiation pneumonitis,history of lung diseases and immunotherapy combined with chemotherapy were independent risk factors for CIP in NSCLC patients of the training group(OR=3.124,10.039,8.831 and 3.131,respectively;all P<0.05).A Nomograms prediction model was constructed based on the four independent risk factors.In the training group,the Hosmer-Lemeshow test result showed p-value of 0.956.The area under the curve(AUC)was 0.890(95%CI:0.830-0.950).After adjustment,the calibration curves of the training dataset were in a good agreement with the ideal prediction curve,with a small fluctuation range.The decision curve analysis(DCA)curve showed the range of the threshold probability from 3%to 96%.In the validation group,the AUC was 0.793(95%CI:0.709-0.876).After adjustment,the calibration curves of the validation dataset greatly fluctuated above and below the ideal prediction curve.DCA curve showed the threshold probability range for the net benefit of 4%-66%.ConclusionsMale,history of lung radiation pneumonia,history of lung diseases and immunotherapy combined with chemotherapy are independent risk factors for CIP in NSCLC patients receiving immunotherapy.The Nomograms prediction model established according to the four independent risk factors has a certain ability to predict the occurrence of CIP.