Abstract Book

S1096

ESTRO 37

Purpose or Objective Delivered dose to the rectum in prostate radiotherapy differs from planned dose due to interfraction motion. By developing independent dose calculation systems based on daily megavoltage image guidance scans, we have achieved a better estimate of delivered dose to the rectum [1]. Here we have developed two Normal Tissue Complication Probability (NTCP) models for rectal bleeding to compare discriminatory ability between delivered dose and planned dose. Material and Methods The study comprised 200 prostate cancer patients treated with TomoTherapy. Cumulative incidence of rectal bleeding ≥ Grade 2 (CTCAEv4.03) at 1 year was prospectively assessed (n=15, 7.5%). By assuming equal biologically effective dose between treatment schedules of 60Gy/20 fractions (n=86) and 74Gy/37 fractions (n=114) [2], an α/β ratio of 2.123 was derived for the rectum and used to combine datasets. Dose surface maps (DSMs) of the rectal wall were pixel-corrected to equivalent dose in 37 fractions. DSMs enable preservation of geometric information during dose accumulation, overcoming a fundamental limitation associated with dose volume histograms. The dosimetric parameters investigated were Equivalent Uniform Dose (EUD), and ‘DSM-dose widths’ (a spatial surrogate for the lateral dimension of a given isodose) [1]. Univariate analyses were used to assess clinical prognostic factors (Fisher’s Exact, Mann-Whitney U) and dosimetric parameters (independent t-test) for inclusion in a multivariable prediction model. Odds ratios were calculated for independent variables. Dose parameters from both delivered and planned DSMs were tested for collinearity and removed if the variance inflation factor was > 10. Univariate associations with p > 0.15 were also removed. Results Two binomial logistic regression models were developed and optimised using delivered and planned DSM parameters respectively. Common predictors of rectal bleeding were 35, 45, 55, and 65 Gy DSM dose-widths, alongside pre-treatment status of previous pelvic or abdominal surgery, and presence of rectal bleeding at baseline. Model performance was internally validated and corrected using bootstrapping. NTCPs for risk of rectal bleeding were calculated based on variables from each model. Discriminatory ability, assessed using the area under the receiver operating characteristic curve (AUC), indicated that the delivered-DSM derived model (AUC = 0.809, CI: 0.711-0.907) was a superior predictor of rectal bleeding than the planned-DSM derived model (AUC = 0.782, CI: 0.662-0.902) (Figure 1).

Conclusion Rectal toxicity prediction can be improved by incorporating parameters of delivered DSMs into NTCP modelling. Ultimately, we anticipate that accumulating delivered dose throughout the course of treatment will facilitate online risk prediction for decision making in adaptive radiotherapy. [1] Shelley, L.E.A., et al., Radiother Oncol. 2017 Jun; 123(3): 466–471. [2] Dearnaley, D., et al., Lancet Oncol. 2016 Aug ; 17(8) : 1047-1060. EP-2011 Accuracy of NTCP models for rectum and bladder toxicity in prostate cancer patients E. Gargioni 1 , S. Domanski 1 , R. Schwarz 1 1 University Medical Center Hamburg - Eppendorf UKE, Department of Radiology and Radiotherapy, Hamburg, Germany Purpose or Objective The use of radiobiological models for predicting toxicity after radiotherapy, for optimizing IMRT treatment plans, and for allowing plan adaptation during the course of treatment is being implemented in modern treatment planning systems. However, the reliability of such models and the accuracy of their predictions need to be thoroughly investigated before they can be used in the clinic. In this work, the treatment plans of prostate cancer patients that underwent helical tomotherapy were retrospectively analyzed in order to compare the observed incidence of gastro-intestinal and genito-urinary toxicity and the NTCP model predictions. Material and Methods This work is based on a group of 139 prostate cancer patients that were treated with helical tomotherapy between January 2007 and December 2010. The dose prescription to the PTV varied between 72 Gy and 76 Gy, with dose escalation to the GTV up to 80 Gy for high-risk patients according to the D’Amico classification. The dose volume-histograms (DVH) of bladder and rectum were retrospectively analyzed and the NTCP values were calculated for late gastro-intestinal (GI) and late genito-urinary (GU) toxicities of grade 2 and grade 1, respectively. The Lyman-Kutcher-Burman (LKB) model was used with parameters taken from the literature. The observed incidence of side effects was then compared to the model predictions for each risk group. Finally, possible