Treatments buy AZD4547 were delivered with 15 MV photon beam generated by a Clinac 2100 CD Varian accelerator, equipped with Millennium MLC (120 leaves). Toxicity evaluation Caspase inhibitor rectal toxicity was assessed using the Radiation Therapy Oncology Group (RTOG) scale [13], every six months for the first three years after the end of treatment and afterwards every year. The incidence of ≥ G2 late rectal toxicity as a function of time (months from the end of treatment) was evaluated by Kaplan-Meier curves using MedCalc software (Version 8.1.0.0, Mariakerke, Belgium). The log rank test was performed to establish if

any statistically significant difference exists between the two arms. Radiobiologic calculations Cumulative dose-volume histograms (DVHs) have been first evaluated for the two arms,

independently. Then, to compare the two different treatment schemes, DVHs for both arms have been corrected converting the physical dose in the i-th volume fraction to the biologically equivalent total dose normalized to the standard fraction of 2 Gy (NTD2), as described in appendix 1 (A.5). The Lyman-Burman-Kutcher (LKB) model was used to predict the NTCP for late rectal toxicity. The learn more ≥ G2 late rectal toxicity was assumed as primary end point in the NTCP calculations. The original model parameters are n, m and TD50 and they determine the volume dependence of NTCP, the slope of NTCP vs. dose and the tolerance dose to the whole CHIR-99021 organ leading to a 50% complication probability, respectively (appendix 1). The α/β parameter was then introduced in the model by the NTD2 to take into account for altered fractionaction schemes, as illustrated also by other authors [14, 15]. At first, the values n = 0.12, m = 0.15 estimated by Burman et al. [10] and the value TD50 = 80 Gy evaluated by Emami et al. [16] were involved in the calculation of the NTCP distributions for conventional and hypofractionated arms. To minimize

the deviation between the clinical and the predicted complication incidences, the best parameters estimation of the model was performed by the maximum likelihood method [17]. For binomially distributed data such as the NTCP data, the log-likelihood for the entire data set is given by: where N is the total number of patients, R i is equal to 1 for patients who did experience ≥ G2 late rectal toxicity or 0 for patients who did not. The optimization of all the four model parameters was initially run but, because of the large resulting 95% confidence intervals (CI) due to the limited number of patients experiencing ≥ G2 late toxicity, the results were not reported. Consequently, it was decided to reduce the number of degrees of freedom by keeping fix the n and m parameters at the original values proposed by Burman et al. [10].