Predictive Factors Of Radiation-induced Lung Injury In Lymphoma Patients Treated With Intensity Modulated Radiation Therapy

Background:Lymphoma is a common hematological tumor.Radiotherapy plays an important role in the treatment of both Hodgkin and non-Hodgkin lymphoma.In lymphoma patients receiving supraphrenic irradiation,radiation-induced lung injury is one of the complications that restrict the dosage of radiotherapy.The dosimetric restricting criterion of lung that currently applies is based on the study of lung cancer.Given the difference between lymphoma and lung cancer patients,unique dosimetric constraints might be necessary for this population.Objective:This study is designed to analyze the predictive factors of radiation-induced lung injury(RILI),to unveil the relationship between dose-volume histogram(DVH)parameters and RILI,to establish predictive model of RILI,as well as to explore the potential threshold value of related DVH parameters.Methods:We retrospectively reviewed clinical data and radiation plan of 309 lymphoma patients treated with intensity modulated radiation therapy,among which 196 cases had complete DVH information.151 Hodgkin lymphoma and 158 non-Hodgkin lymphoma patients were included.Among the non-Hodgkin lymphoma group,83.5%were diffuse large B-cell lymphoma.All patients received chemotherapy followed by involed-field radiotherapy.5 patients received autologous stem cell transplant.RILI was graded according to CTCAE version 4.0(pneumonitis).Smooth hazard ratio model,Lyman-Kutcher-Burman model(LKB model)and multivariate logistic model was constructed to reveal the relationship between mean lung dose(MLD)and Vx(defined as the percentage of normal lung tissue receiving dosage greater than x Gy)and RILI.Receiver operating curve(ROC)was used to identify the threshold values.Area under cuve(AUC)was used to evaluate the predictive ability of each model.Results:The incidence of RILI was 3.4%,with symptomatic radiation-induced lung injury(SRILI)to be 3.4%.Univariate analysis suggested ECOG?2,elevated lactic dehydrogenase level,any chemotherapy given after radiotherapy and all DVH parameters to be the potential risk factors for RILI,but elevated lactic dehydrogenase level lost significance in multivariate analysis.B symptom as well as V15-V40 was significantly associated with SRILI.Under current dose prescription strategy for lymphoma,the probability of RILI appeared to be linearly correlated with DVH parameters,and threshold value unidentifiable.As for SRILI,low-dosage area(V5-V15)was poorly related to SRILI,and MLD>15.1 Gy,V20>28%,V25>20%as well as V30>15%seemed to be reasonable threshold.High-dosage area(V30-V40)once demonstrated linear correlation with SRILI.All the DVH parameters were equally competent in the case of predicting RILI,but SRILI was significantly more sensitive to median-high dosage area.Low-dosage area seemed to rather irrelevant with SRILI.Smooth hazard ratio model,LKB model and multivariate logistic model paralleled in the sensitivity and specificity of prediction.Conclusion:Optimal lung dosage restriction ought to be based on the balance between each Vx parameter.Maximum reduction of RILI depends on the minimal irradiation of lung.As for SRILI,priority should be given to minimize medina-high dosage area.Moderate increase of low-dosage area might be acceptable.