Radiation pneumonitis after breast cancer irradiation: Analysis of the complication probability using the relative seriality model

Background: Toxicity of the respiratory system is quite common after radiot herapy of thoracic tumors; breast cancer patients represent one of the grou ps for which there is also a long expected survival. The quantification of lung tissue response to irradiation is important in designing treatments as sociated with a minimum of complications and maximum tumor control. Methods: The study population consisted of 68 patients who received irradia tion for breast cancer at Stage II. radiation pneumonitis was retrospective ly assessed on the basis of clinical symptoms and radiological findings. Fo r each patient, a measure of the exposure (i.e., the lung dose-volume histo gram [DVH]) and a measure of the outcome was available. Based on these data , a maximum likelihood fitting to the relative seriality model was performe d. The uncertainties of the model parameters were calculated and their impa ct on the dose-response curve was studied. The optimum parameter set was th en applied to 5 other patient groups treated for breast cancer, and the nor mal tissue complication probability (NTCP) was calculated. Each group was i ndividuated by the radiotherapy treatment technique used; the dose distribu tion in the lung was described by a mean DVH and the incidence of radiation pneumonitis in each group was known. Lung radiosensitivity was assumed to be homogeneous through all of the calculations. Results: The relative seriality model could describe the dataset, The volum e effect was found to be relevant in the description of radiation pneumonit is. Age was found to be associated with increased risk of radiation pneumon itis. Two distinct dose-response curves were obtained by splitting the grou p according to age, The impact of the parameter uncertainties on the dose-r esponse curve was quite large, The parameter set determined could be used p redictively on 3 of the 5 patient groups. Conclusion: The complication data could be modeled with the relative serial ity model, However, further independent datasets, classified according to t he same endpoint, must be analyzed before introducing NTCP modeling in clin ical practice. (C) 2000 Elsevier Science Inc.