Factor Analysis And Model Investigation For Predicting Radiation Pneumonitis In Lung Cancer

Background and Objectives:The mortality rate of lung cancer is the highest of all malignant tumors in the worldand about2/3patients need definitive radiotherapy （RT）. With the rapid development ofradiotherapy technology, radiation lung injury （RLI） is becoming a prominent question. Theincidence of acute radiation lung injury after thoracic irradiation is from5%to36%according over-seas literature. Respiratory failure is one of the main causes of death causedby pulmonary toxicity, which include acute radiation pneumonitis （RP） and chronicradiation pulmonary fiberosis （RPF）, and once occurred we have no satisfying remedialmeasure at present. Therefore, it has become an important topic to find factors predictingRP.Meta-analysis based on the data from pooled patients provided us a novel avenue toevaluate the risk factors of RP. We analyzed the correlative factors and established themodel to predict the incidence of RP in the treatment of three-dimensional conformalradiotherapy （3D-CRT） for lung cancer patients based on the meta-analysis results. The aimwas to find prognostic parameters and provide reference standard for the clinical besttreatment plan.Methods:1. Articles were searched in February2012from PubMed, EMBASE, CochraneLibrary and Chinese Journal Fulltext Database （CNKI） using the Medical Subject Headings,“lung cancer” and “radiation pneumonitis” or “radiation lung injury”, and references citedin the retrieved articles were hand-searched. The outcome was the RP incidence. Thequality of eligible trails was assessed by Newcastle-Ottawa scale and Consort2010statement. We pooled the data using RevMan5.1software. Relative Risk （RR）, Odds Radio（OR） or mean difference （MD） were calculated from the reports of the studies available andthen combined together to get estimate of efficiency. 2. Lung cancer patient who received lung radiation in our hospital for the first time anddidn’t undergoing thorax radiation in other site during January2009to October2011wasscreened. RP was defined by the imaging examination and RTOG standard after a period ofsix months for follow-up at least. Analyze the relationship between radiation pneumonitiswhich was equal or greater than2grade and the factors described above. We pooled thedata using SPSS16.0software to analysis the relationship between clinical factors ordose-volume parameters and RP, and establish a model to predict the incidence of RP.Results:1. From a total of73selected papers, we included the following factors: age, gender,race, weight loss, smoking history, chronic lung disease, performance status, pre-radiationtherapy （RT） pulmonary function, disease clinical stage, histological type, tumor location,pre-RT surgery, RT combined with chemotherapy （RCT）, RT/RCT combined withamifostine, plasma end-RT/pre-RT TGF-β₁ratio and dose-volume histogram （DVH）parameters. Analyzed by36papers, the significant risk factors for RP≥grade2were:patients with chronic lung disease, tumor located in the middle or lower lobe, RCT andgross tumor volume （GTV）. But higher level of normal tissue complication probability（NTCP）, mean lung dose （MLD） of total lung tissue and V₅、20maybe risk predict factors forRP. The following factors were identified significant for RP incidence, including tumorlocated in the middle or lower lobe, smokers, combined with chronic lung diseases ordiabetes mellitus, low pre-RT pulmonary function （FEV1<2L）, RCT and no amifostinecombined RT/RCT, plasma end-RT/pre-RT TGF-β₁ratio≥1and dose-volume parameters,which include V30＞18%, V20＞25%and the average of GTV and V5、10. But higher levelof NTCP, MLD of total lung tissue and V20、30maybe risk predict factors for RP.2. RP group with a relatively short local progression-free surial than non-RP group.The data analysis suggest that many parameters (obstructive pneumonia before RT, plantarget volume（PTV）, exposed fieldsnumbers and diameters, MLD and V13、15、20、25、40、45ofdiseased lung and bilateral lung, V5of contralateral lung and V₅、10、30、50、55of diseased lungwere significantly associated with RP equal and greater than grade2（P<0.05）. Binarylogistic analysis revealed that exposed field’s numbers were associated with RP. These twofactors combined with V₃0、45of diseased lung entered the predictive model which with asensitivity of22.6%. Conclusion:1. Our meta-analysis considered that smokers with chronic lung disease or diabetesmellitus, low pre-RT pulmonary function （FEV₁<2L）, tumor located in the middle or lowerlobe, lack of pre-RT lung tumor surgery, RCT and no amifostine during RT, plasma TGF-β₁increased after RT and dose-volume parameters, which include V₃0＞18%, V₂0＞25%andhigher GTV and V₅、10were risk predict factors for RP.2. RP can influence patient by shorten the local progression-free survival. Obstructivepneumonia before RT and DVH parameters were significantly associated with RP, whichwas similar with our meta-analysis.3. The predictive model with a high specificity and low sensitivity that need to beexaminated and improved in our future work. In our own validation, the actual RPincidence was similar with forecast by the model. Thus it has a certain clinical referencevalue.