Abstract
A thorough understanding of the dose‐response of individual organs‐at‐risk is essential for being able to choose the best radiotherapytreatment for a patient. Therefore, the spatial distribution of dose to the rectum, the anal canal and the parotid glands and its role for radiation‐induced late toxicities were assessed by performing a retrospective analysis of data from the randomised controlled trials MRC RT01 and PARSPORT. Interpretable geometrical features were introduced to quantify the spatial distribution of dose and correlations between the dose to the anorectal wall and seven clinically relevant late rectal toxicities were assessed. Novel risk‐factors based on the spatial distribution of dose could be identified for different endpoints. It was demonstrated that the type of dose‐response relationship differed considerably between different endpoints. Rectal bleeding was more strongly correlated to the lateral extent than to any other measure, including the dose‐surface histogram (DSH). The dosimetric measure with the strongest correlation to loose stools and proctitis were longitudinal extent and DSHs respectively. Based on this statistical analysis,dosimetric constraints taking spatial information into account were derived. An interpretable, non‐linear normal‐tissue‐complication‐probability model (NTCP model) explicitly taking spatial information into account was presented and its predictive power was evaluated and compared to NTCP models used in current clinical practice. Our novel NTCP models predicted late rectal toxicities significantly better than conventional models. Using the novel NTCP models 3D dose patterns related to a low risk of complications have been identified. These dose patterns may be useful for optimising and assessing treatment plans, potentially yielding better treatments with a reduced risk of late rectal toxicities. Furthermore, the role of spatial information and clinical risk factors for xerostomia after head‐and‐neck radiotherapy was assessed. Multivariate NTCP models taking the shape and the location of the dose to the parotid glands into account were derived based on data from the PARSPORT trial. Results were externally validated using data from two independent patient cohorts. A “directed” bath‐and‐shower effect was identified, indicating that a high relative concentration of the dose in the medial‐inferior part of the ipsi‐lateral gland and a high skewness of the dose in cranio‐caudal direction in the superficial lobe of the ipsi‐lateral parotid are beneficial for patients in terms of reducing the risk of xerostomia. Our novel NTCP model taking the “directed” bath‐and‐shower effect into account can be used to rank treatment plans more reliably than standard mean‐dose models. We have also shown that it is possible to implement this model in an inverse‐treatment‐planning system and take morphological information into account when generating IMRTtreatment plans. Learning Objectives: 1. Understand issues related to quantifying the spatial distribution of dose to organs‐at risk 2. Understand issues related to generating and validating NTCP models based on these risk‐factors 3. Understand issues related to identifying risk factors related to the spatial distribution of dose
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have