Risk Factors Analysis And Predictive Model Construction Of Treatment-related Pneumonia Induced By Thoracic Radiotherapy Combined With Immune Checkpoint Inhibitors

Objective:In the treatment of thoracic tumors,radiation therapy combined with immune checkpoint inhibitors can induce a variety of treatment-related pneumonia,including radiation pneumonia,immune checkpoint inhibitor-related pneumonia and some special pneumonia like radiation recall pneumonia.The occurrence of these treatment-related pneumonia often interrupts the process of treatment and affects the prognosis of patients.In this study,we retrospectively analyzed the clinical data of patients who received thoracic radiotherapy and ICIs,explored the risk factors of treatment-related pneumonia in the context of combination therapy and try to conduct a predictive model in order to identify high-risk patients early in clinical practice and take preventive measures to improve the prognosis of patients.Methods:According to the inclusion and exclusion criteria,the electronic medical records of117 patients who were treated by chest radiotherapy combined with immune checkpoint inhibitors in Shanxi Cancer Hospital and Shanxi Bethune Hospital from January 1,2020 to September 1,2022 were collected and analyzed.The data collected included the general condition of the patient(age,sex,smoking history,lung disease history,physical status score),radiotherapy parameters(total dose,dose division,target volume,mean lung dose,Total lung V5,Total lung V20,Total lung V30)and immune checkpoint inhibitors(type of treatment drugs,number of treatment cycles,combination with radiotherapy)and so on.Patients were divided into symptomatic treatment-related pneumonia group and negative group according to whether ≥ Grade 2 treatment-related pneumonia occurred.The least absolute shrinkage and selection operator(LASSO)algorithm was used to initially screen the study variables,and the independent risk factors were identified and the clinical prediction model was constructed by multi-factor Logistic regression.The Bootstrap method was used to internally verify the prediction model.The Nomogram was used to visualize the constructed prediction model,the receiver operating characteristic curve was drawn and the Area under the curve was calculated to evaluate the discrimination of the prediction model.The Hosmer-Lemeshow test was used and the model calibration curve was plotted to evaluate the calibration of the prediction model,and Decision Curve Analysis was plotted to evaluate the clinical utility of the prediction model.Results:Finally,117 patients met the inclusion criteria and were included in the study according to the sodium exclusion criteria.Of the patients included in the study,48 eventually developed Grade ≥2 treatment-related pneumonia,with an incidence rate of41.03%.Univariate analysis revealed statistically significant differences among groups in the occurrence of ≥ Grade 2 treatment-related pneumonia among smoking history(χ~2=6.703,P=0.01),MLD(T =-4.491,P=0.031),total lung V5(T =-4.602,P=0.018),total lung V20(T =-3.835,P=0.043),and total lung V30(Z =-5.262,P<0.001).After the initial selection of variables by LASSO regression,a multivariate Logistic regression model was established,which finally determined that smoking history(B=1.383,P = 0.006 95% CI: 1.49-10.71),total lung V5(B=0.094,P=0.004,95% CI: 1.03-1.17),and total lung V30(B=0.191,P=0.004,95% CI: 1.06-1.38)were independent risk factors for ≥ Grade 2 treatment-related pneumonia.The maximum AUC value of the prediction model established by the above three indicators after verification by the ROC curve was 0.827(95% CI: 0.747-0.906).Under this condition,the sensitivity and specificity of the model were 85.5% and 68.8%,respectively.The results of Hosmer-Lemeshow test for the prediction model suggested that p = 0.148 >0.05,and the calibration curve was similar to that of the ideal model.Analysis of the decision curve suggested that when the threshold probability of ≥ Grade 2treatment-related pneumonia ranged from 11% to 87%,clinical decisions made based on the predictive model could benefit patients.Conclusion:(1)In the background of chest radiation therapy combined with immune checkpoint inhibitor therapy,the patient’s smoking history,whole lung V5 and whole lung V30 were the independent risk factors for treatment-related pneumonia.(2)The prediction model established based on the independent risk factors of treatment-related pneumonia has good identification ability,calibration ability and clinical application value,which can provide a reference for clinicians to assess the risk of symptomatic treatment-related pneumonia in the process of clinical practice.