ESTRO 2020 Abstract Book

S350 ESTRO 2020

affecting the patient’s quality of life, e.g. xerostomia and dysphagia. These models combine dosimetric predictors with clinical and treatment-related risk factors. While these mostly logisitic regression-based NTCP models contain actionable dosimetric predictors, they often have suboptimal accuracy measures, with AUCs typically around 0.7. The limited discrimination performance currently leads to false positive and false negative predictions, and thus to suboptimal optimization objectives and/or patient selection. It could be hypothesized that the reason for the dosimetric suboptimality of these classical NTCP models is that DVH points or summarizing metrics as the mean doses of organs at risk (e.g. the parotid glands) were used as input to the modeling studies. As DVHs condense 3D planning dose maps into a 1D curve, clearly regional information is lost. This regional dependence hypothesis was underpinned by new evidence from small animal studies showing that suborgan radiosensitivity exists in the parotid gland. More specifically, the subregion of the major salivary gland containing stem cells was most radiosensitive. This finding was corroborated by human data, where imaging studies quantifying radiation-induced damage showed similarly significant regional differences. Parotid functioning on MR images and salivary flow rate on SPECT imaging showed a dose-related regional dependence, and were associated to xerostomia endpoints. This 3D dose-response knowledge could pave the way to improve the NTCP model accuracy. A voxel-wise statistical modeling strategy using the 3D dose map and taking into account regional information, was previously described. This technique could highlight the dose level of importance for the endpoint in association with a subregion of the parotid gland. Another method that could be relevant to study in this context is to characterize the 3D shape of the parotid isodoses into a shape feature signature. Ultimately, these techniques might result in a higher accuracy NTCP model giving 3D guidance for sparing the relevant regions from the relevant dose level. In conclusion, this lecture will give an overview of the existing NTCP models in head and neck cancer with their dosimetric limitations. The small animal and human data on sub-parotid radiosensitivity will be discussed. This data will bring us to potential improvements of NTCP models using 3D dose map-based prediction modeling. SP-0620 Management of Radiation Toxicity in Head and Neck Cancers E. Forde 1 St James Hospital Trinity Centre for Health, Dublin, Ireland

Symposium: Preventing radiation-induced toxicity in head and neck cancer patients

SP-0617 Clinical approaches to preventing toxicity in head and neck cancer patients

Presentation cancelled

SP-0618 Automated treatment planning for prevention of radiation-induced toxicity M. Hoogeman 1 1 Erasmus MC, Radiation Oncology, Rotterdam, The Netherlands Abstract text The definite treatment of head and neck cancer is associated with a considerable risk of moderate and sever acute and late side effects induced by radiation. These side effects deteriorate the patient’s quality of life. Modern radiation therapy techniques such as intensity- modulated radiotherapy (IMRT) and volumetric modulated arc therapy (VMAT) combined with image-guidance were introduced to reduce these adverse effects. The introduction of IMRT and VMAT, however, added considerable complexity to the treatment planning process. The planner has to tweak the optimization parameters in order to properly balance the often conflicting constraints and objectives of the target volumes and organs at risk. This complexity results in considerable variation in the quality of the treatment plans. Automated treatment planning has been advocated as a means for reducing this variation and improving the quality of the treatment plans. The latter could result in reduced doses to organs at risk, which may lower radiation-induced side effects. In this presentation, literature will be critically reviewed on how automated treatment planning may contribute to the prevention of radiation-induced toxicity. The limitations of automated treatment planning will be discussed as well. Furthermore, we will address how automated treatment planning can be used as a decision tool to select the optimal treatment modality for an individual patient. This approach will be compared with decision tools that are not based on a full optimization of a treatment plan. Finally, in-silico studies that make use of automated treatment planning for example to determine the impact of treatment accuracy on treatment planning, the dose to organs at risk, and patient selection will be discussed. SP-0619 Insight in sub-organ radiosensitivity profiles may improve prediction models of toxicity G. Defraene 1 1 KU Leuven - University of Leuven, Department of Oncology - Experimental Radiation Oncology, Leuven, Belgium Abstract text Accurately predicting radiation-induced toxicity outcomes is a crucial step towards a precision medicine approach tailoring radiotherapy treatments to the individual patient. An informed clinical decision could then be based on the weighting of the risks from different normal tissue complication probability (NTCP) models. In practice, this could result in sparing the relevant organs at risk or referring the patient to the most effective treatment modality. In head and neck cancer, several NTCP models have been published, predicting the risk of relevant endpoints

Abstract not available

Symposium: Multidisciplinary treatment of hepatocellular cancer

SP-0621 Multidisciplinary treatment of hepatocellular cancer: the radiation oncologist's point of view V. Vitolo 1 1 Fondazione CNAO, Radiation Oncology, Pavia, Italy Abstract text The Role of RT in management of inoperable HCC is recently emerging. Since few years ago, RT has been considered one of the possible treatment for patients not eligible to other local approaches. The reasons have to be found in the lack of consistent data from the literature: