ESTRO 38 Abstract book

S614 ESTRO 38

S. Cox 1 , E. Miles 2 , J. Staffurth 3 , S. Gwynne 1 1 South West Wales Cancer Centre, National Radiotherapy Trials QA RTTQA Group, Swansea, United Kingdom ; 2 Mount Vernon Hospital, National Radiotherapy Trials QA RTTQA Group, London, United Kingdom ; 3 Velindre Cancer Centre, National Radiotherapy Trials QA RTTQA Group, Cardiff, United Kingdom Purpose or Objective The international radiotherapy community has recognised that non‐adherence to RT protocols can influence trial endpoints. However this conclusion is based on studies predominantly assessing the impact of deviations in dosimetric or treatment delivery protocol parameters rather than target volume delineation (TVD). This systematic review evaluates the assessment of TVD within RTQA programmes and the clinical impact of TVD protocol deviations. The implications for RTQA programmes are discussed. Material and Methods MEDLINE, PreMEDLINE, Embase, Cochrane Library, Web of Science, OpenGrey, WHO International Clinical Trials Registry Platform portal and ClinicalTrials.gov were searched for ‘target volume delineation’, ‘interobserver variation’, ‘radiotherapy trials quality assurance’ and ‘protocol non‐adherence’. Full‐length articles and conference abstracts (January 2005‐January 2018) were included to avoid publication bias. Eligibility criteria included: 1. RTQA assessment of TVD in phase II‐III trials; 2. impact of deviations assessed & categorised; 3. impact correlated with patient outcome. Results 5864 abstracts were screened for relevance; 94 full‐length articles were reviewed. 5 relevant trials were identified (4 journal papers and 3 conference abstracts) (table 1). Various classification systems were used to assess protocol deviations; ‘unacceptable’ or ‘major’ deviations in TVD occurred in 2.9‐13.4% of assessed RT plans (when reported). It was often not possible to establish deviation rates specifically related to TVD as these were frequently combined with other types of protocol deviations including OAR contouring and dosimetric or treatment delivery parameters. The details on what specific variations in TVD were deemed to be ‘unacceptable’ by RTQA teams was also not routinely reported and the difficulty in establishing a ‘consensus’ for appropriate TVD for on‐trial patients was highlighted. Results suggest that deviations in TVD were associated with poorer outcomes for overall survival, local control and treatment‐related toxicity however the data was hetereogenous (table 1). RTQA of TVD was retrospective and feedback on the quality of TVD to recruiting centres was not standard.

and autosegmentation server. Fully automated delineation process has the following steps: 1) CT‐scanner exports CT slices to pseudonymization gateway on which Conquest service is configured to pseudonymize these CT slices and export them to Ensemble integration platform. 2) Ensemble integration platform uses SFTP protocol to transfer CT‐slices to autosegmentation server. 3) On the server, an autosegmentation service (MIM Software inc.) will automatically start the segmentation for all new data it receives. This atlas‐based autosegmentation consists of 20 atlases, which were contoured according to the combination of the ESTRO and Danish Breast Cancer Cooperative Group guidelines. 4) After the segmentation is completed, Ensemble integration platform will forward delineated structures to the hospital network where Conquest service removes pseudonymization and forwards the delineated structures to TPS.

Results This fully automated autosegmentation process has been implemented for thorax area including both left and right sided breast cancer patients, respectively (Fig 2). The automated transfer has been realized so far with two RT centers and the concept has proven to work. The atlas‐ based autodelineations of the target volumes and critical structures have reduced manual work regarding contouring, although the validation of the segmentation is underway.

Conclusion Target delineation is one of the biggest errors in RT. Interobserver variations can be reduced by guidelines, training and autocontouring tools. We have implemented fully automated contouring service available for Finnish RT centers, which should harmonize the national target delineation process. [1] Segedin B, Petric P. Uncertainties in target volume delineation in radiotherapy ‐ are they relevant and what can we do about them? Radiol Oncol. 2016; 50:254‐262

Conclusion Non‐adherence to TVD protocols may have negative clinical consequences including worse overall survival, local control and treatment‐related toxicity rates. Unacceptable TVD deviations should be clearly defined at the time of protocol development to minimise

PO-1105 Impact of deviations in target volume delineation - time for a new RTQA approach?