44 UK radiotherapy practice in volume delineation–data from the APPROACH trial (ISRCTN:1339049) radiotherapy quality assurance (RTQA) programme benchmark outlining case

Abstract

Abstract Introduction Interobserver variation in radiotherapy contouring can affect protocol compliance and the robustness of clinical trial outcomes. The APPROACH (Analysis of Proton vs Photon Radiotherapy in Oligodendroglioma and Assessments of Cognitive Health) trial protocol mandates delineation of target volumes and organs-at-risk (OARs), including the hippocampi and hypothalamus, to enable correlation of delivered dose with late toxicity. Here we assess centres’ previous experience in contouring OARs and performance in the benchmark outlining case. Methods The APPROACH radiotherapy quality assurance (RTQA) programme included a pre-accrual facility questionnaire (FQ) designed to document local radiotherapy practices and a benchmark outlining and planning case in accordance with the APPROACH Radiotherapy Guidance using 1 mm-slice dedicated planning MRI. Contours were compared against a gold standard generated by expert consensus. Interobserver variability was quantified using the Jaccard conformity-index (JCI), discordance- index (DI), geographical-miss-index (GMI), and mean distance to conformity (MDC) both volumetrically and slice-by-slice to assess regional variation. Results All centres returned FQs; variation in standard OAR contouring is demonstrated. 6/25 routinely contoured the hippocampus and one proton centre routinely contoured the hypothalamus. Benchmark contours were evaluated from 20 centres. Target volumes showed good conformity (GTV-FLAIR JCI 0.87, CTV 0.91). OARs demonstrate lower agreement, including right and left hippocampi (median JCI 0.62 and 0.59) and right and left hypothalamus (both 0.51). Overall, OAR JCI ranged from 0.91-0.18. Slice-by-slice analysis demonstrates variability was greatest at superior and inferior slice extents, with both under- and over-contouring observed. Conclusions The outlining benchmark case highlights significant interobserver variation exists in delineation of OARs, despite atlas-based guidance. Reasons for poor conformity identified from examination of the contours include misidentification of structures in areas of poor anatomical definition, and different local contouring practice.