| Part ⅠComprehensive Pneumonitis Profile of Thoracic Radiotherapy Followed by Immune Checkpoint Inhibitor and Risk Factors for Pneumonitis in Lung CancerObjective:Whilst survival benefits of thoracic radiotherapy(TRT)followed by immune checkpoint inhibitor(ICI)have been reported in patients with lung cancer,the potential high risk of treatment-related pneumonitis remains a concern.Asians may be more sensitive to lung toxicity than other races.This retrospective study intended to provide a comprehensive pneumonitis profile of TRT followed by ICI and investigate the risk factors from a Chinese cohort of lung cancer.Methods:From January 2016 to July 2021,196 patients with lung cancer who received TRRT prior to ICI were retrospectively analyzed.Treatment-related pneumonitis,including checkpoint inhibitor pneumonitis(CIP),radiation pneumonitis(RP),and radiation recall pneumonitis(RRP),were recorded and graded through medical records and chest computed tomography.Characteristics predictive of pneumonitis were assessed using logistic regression models,and the receiver operating characteristic analyses were performed to identify optimal cut points for quantitative variables.Results:With a median follow-up of 18 months,a total of 108 patients(55.1%)developed treatment-related pneumonitis during ICI therapy,with an incidence of 25.5%for grade 2 or higher(G2+)and 4.1%for G3+.The overall rates of CIP,RP and RRP were 8.2%(n=16),46.9%(n=92)and 7.1%(n=14),respectively.With a total mortality rate of 1.5%,vast majority of the patients recovered from pneumonitis or remained stable.No patients died of RRP.Half of the patients with G2+RP who withheld ICI therapy restarted ICI safely after resolution of RP.On univariate analyses,history of chronic pulmonary diseases(36%vs.21.9%,P=0.049),ICI consolidation(62%vs.42.5%,P=0.017)and interval between TRT and ICI less than 3 months(70%vs.45.2%,P=0.002)were significantly associated with G2+treatment-related pneumonitis when compared with the grade 0-1 group,but did not reach significance for G3+group when compared with the grade 0-2 group.Interval between TRT and ICI less than 3 months was an independent predictor for G2+treatmentrelated pneumonitis in a multivariate model(Odds ratio OR=2.787,P=0.004).The history of chronic pulmonary diseases(P=0.05),mean lung dose(MLD,P=0.038),percent volume of lung receiving≥5 Gy(V5,P=0.012)and percent volume of lung receiving≥20 Gy(V20,P=0.030)predicted the occurrence of RRP in univariate analyses.Conclusions:Treatment-related pneumonitis,especially RRP,is acceptable and manageable in the setting of TRT followed by ICI in this Asian population.Dosimetric parameters MLD,V5 and V20 may improve the predictions of RRP in clinical practice.Part ⅡRadiographic Features and Patterns of Treatment-related Pneumonitis in Lung Cancer Patients Received Thoracic Radiotherapy and Immune Checkpoint InhibitorsObjective:Patients with lung cancer may developed treatment-related pneumonitis after thoracic radiotherapy(TRT)and immune checkpoint inhibitors(ICI).This study aimed to investigate the computed tomography(CT)features and radiographic patterns of treatment-related pneumonitis of the combination therapy.Methods:From January 2016 to July 2021,lung cancer patients who received ICI within 3 months before or after TRT and developed treatment-related pneumonitis after the combination therapy at two medical institutions were retrospectively analyzed.CT characteristics were analyzed for the association with pneumonitis severity,and were compared between radiation pneumonitis(RP)and checkpoint inhibitor pneumonitis(CIP).Results:One hundred and thirty seven eligible patients with 145 treatment-related pneumonitis were included in the study.Seventy three patients had grade 1 pneumonitis,sixty grade 2,eleven grade 3 and one grade 4.Organizing pneumonia(OP,59.3%)was the most common radiographic pattern of treatment-related pneumonitis.Concurrent TRT with systemic drugs was significantly associated with grade 2 or higher(G2+)pneumonitis(P=0.02).When compared with grade 1 pneumonitis,G2+pneumonitis was more likely to present as hilar/mediastinal non-metastatic lymph node enlargement(P=0.006),groundglass opacity(P=0.025),patchy hypoattenuated lobules(P=0.026).And G2+pneumonitis was more likely bilateral(P=0.001),involving all lobes(P=0.011)and had diffused distribution(P<0.001).Diffused distribution was an independent predictor for G2+pneumonitis(odds ratio:14.4,P<0.001)in multivariable analyses adjusting for clinical variables.Grade 3 or higher(G3+)pneumonitis involved larger area of the lung and had more diverse CT features than grade 1 to 2.One hundred and twenty two patients had RP,13 had CIP and 11 had RP+CIP.CIP was more likely to present as ground-glass opacity(P=0.044)while bronchial wall thickening(0.002)was more often in RP.CIP was more likely to involve all lobes(P=0.016),have diffused distribution(P=0.003)and have fuzzy border(P=0.044)than RP.Conclusions:OP pattern was the most common radiographic pattern for treatment-related pneumonitis after TRT and ICI.Diffused distribution was associated with severe treatmentrelated pneumonitis.RP and CIP also exhibited distinct spatial features on CT.Part ⅢMulti-omics prediction for symptomatic radiation pneumonitis through machine learning methods in patients with lung cancerObjective:Radiation pneumonitis(RP)is the major dose-limiting toxicity of thoracic radiotherapy.Besides the clinical and dosimetric risk factors,previous studies reported that genotype information,such as single nucleotide polymorphisms(SNP),was associated with the development of symptomatic(grade 2 or higher)RP.This study aimed to developed a multi-omics prediction model for symptomatic RP based on genomics and dosiomics data through machine learning methods.Methods:From 2009 to 2016,lung cancer patients who received thoracic radiotherapy and had samples of peripheral blood were screened on the SNP selection phase.Patients with grade 3 or higher RP were matched with grade 0 to 1 RP with the following factors:age,lower lobe irradiation(yes or no),concurrent chemoradiotherapy(yes or no),percent volume of lung receiving≥20 Gy(V20)and mean lung dose(MLD).Whole exome sequencing(WES)was used and Fisher’s exact test were performed to compared the differences of SNP genotypes between the groups.SNP with P<0.015 were selected.On the modeling phase,lung cancer patients received thoracic radiotherapy and had samples of peripheral blood from 2016 to 2020 were enrolled.All patients were genotyped via MassARRAY platform for selected SNP from the selection phase,and randomly divided into training group and testing group(4:1)stratified by the symptomatic pneumonitis.Fractionation machine(FM)method was used to build the SNP-FM model for symptomatic pneumonitis based on SNP genotyping data.And ySNP was calculated through SNP-FM model.The mRMR algorithm and bootstrap method were adopted to select and extract dosiomics features.A comprehensive prediction model based on ySNP and selected dosiomics features for symptomatic pneumonitis was built through logistic regression(LR).The models were trained in the training group and tested in the testing group.Clinical factors were separately assessed by general statistical tests.Results:On the SNP selection phase,peripheral blood samples of 28 patients with grade 3 or higher RP and the matched 28 patients with grade 0 to 1 RP were sequenced.Eightyone SNP with P<0.015 and another 21 radiation-toxicity-related SNP reported previously were selected for next step.On the modeling phase,400 eligible patients(including 108 symptomatic RP)were enrolled and genotyped for selected SNP.The SNP-FM model included 19 SNP for the prediction of symptomatic RP,and achieved an area under the curve(AUC)of 0.76 in the testing group.Four dosiomics features were selected and combined with ySNP in the LR model,the AUC of the comprehensive model increased to 0.81 in the testing group.Conclusions:By FM method,this study integrate complicated genotype information into ySNP,and developed a multi-omics model based on ySNP and dosiomics features to predict symptomatic RP. |
PDF Full Text Request