Comparing of different normal tissue complication probability models for plan evaluation of radical radiotherapy for esophageal cancer

Introduction: Chemoradiation is a well-established protocol for management of localized esophageal cancer. However, some early and late toxicities are associated with this therapy. Plan evaluation in term of normal tissue complication probability (NTCP) may help clinicians to tailor this therapy more individually. Given the variety of present radiobiological models, the choice of a model that is suitable for this purpose remains an issue. The purpose of this study was to compare different NTCP models and parameters for plan evaluation of radical radiotherapy for esophageal cancer, in terms of mean values and ranges of predicted NTCPs for four different radiation-induced normal tissue endpoints. Materials and Methods: Eight patients with localized esophageal cancer were included in the study. The patients had been treated in two phases and received 54 Gy in 30 fractions. Dose volume histogram data for heart, lungs and spinal cord were extracted from the treatment planning system and were considered as the main organs-at-risk for NTCP evaluation. Three NTCP models, namely, Lyman-Kutcher-Burman (LKB), Relative Seriality (RS) and Parallel with their corresponding sets of published parameters were used for plan comparisons for the following endpoints: radiation induced pericarditis, ischemic heart disease, pneumonitis and myelopathy. Results: For radiation induced pericarditis and ischemic heart disease, the predicted NTCP value ranges were 5%-50% and 10%-40% by the LKB and RS models, respectively. For pneumonitis, both the LKB and RS models predicted similar NTCPs (1-5%) depending on the parameters used. The Parallel model also gave similar results. For radiation myelopathy, the LKB model predicted NTCPs of <5%. However, there were no RS and Parallel model parameters available for prediction of radiation myelopathy. Conclusion: For the treatment technique studied, the patient plans showed greater mean values and range of NTCPs for radiation-induced pericarditis and ischemic heart disease compared to pneumonitis and myelopathy. The four models showed general agreement. However, availability of reliable model parameters still remains a challenge to routine use of radiobiological model in clinical decision making, which requires further studies.