ROC curves and evaluation of radiation-induced pulmonary toxicity in breast cancer

PURPOSE: To study clinical, radiologic, and physiologic pulmonary toxicity in 128 women after adjuvant radiotherapy (RT) for breast cancer in relation to dosimetric factors. METHODS AND MATERIAL: The patients underwent pulmonary function testing before and 5 months post-RT. Similarly, computer tomography of the chest was repeated 4 months post-RT and changes were scored with a semiquantitative system. Clinical symptoms were registered and scored according to Common Toxicity Criteria. All patients underwent three-dimensional dose planning, and the ipsilateral lung volume receiving > or = 13 Gy (V13), V20, and V30 were calculated. Multiple logistic or regression analyses were used for multivariate modeling. The relation between the dosimetric factors and side effects was also analyzed with receiver operating characteristic (ROC) curves. RESULTS: V20 was, according to multivariate modeling, the most important variable for the occurrence of the three studied side effects (p < 0.01). Age was also related to symptomatic and radiologic pneumonitis. Reduced pre-RT functional level was more common in patients developing symptomatic toxicity. The ROC areas for symptomatic pneumonitis in relation to V13, V20, and V30 were 0.69, 0.69, and 0.67, and for radiologic pneumonitis 0.85, 0.85, and 0.81. CONCLUSIONS: Our results support the use of three-dimensional planning aimed at minimizing the percent of incidentally irradiated lung volume to reduce pulmonary toxicity. Age was also correlated with post-RT side effects. According to ROC analysis, V20 could well predict the risk for radiologic pneumonitis for the studied semiquantitative model.