Importance of the reference volume in assessing outlining performance for the purpose of training and revalidation
Session type: Poster / e-Poster / Silent Theatre session
Assessment of outlining performance will form an increasingly important part of future training and revalidation for clinical oncologists, measured by comparison to a pre-defined reference volume. Definition of the latter can be informed by experience gained within the SCOPE 1 trial.
The pre-trial RTTQA programme of SCOPE 1 (UK phase 2/3 randomised controlled trial of chemoradiotherapy for oesophageal cancer) included an outlining assessment of a mid oesophageal tumour. 50 investigator GTV (i-GTV) outlines were compared both to the original pre-defined gold standard (gs-GTV) (defined by the chief investigator and an upper GI radiologist) and a consensus volume (c-GTV) (included the gs-GTV plus i-GTVs of three clinical oncology members of the Trial Management Group using the STAPLE algorithm).
CERR was used to calculate the conformity of each i-GTV against the gs-GTV and c-GTV using the Jaccard Conformity Index (JCI). JCI values of <0.5 and ≥0.7 were taken to represent poor and excellent conformity respectively. All 50 i-GTVs were considered acceptable by the RTTQA team.
Length and volume of gs-GTV and c-GTV were 7.8cm and 8.1cm and 39.15cc and 46.77cc respectively. Median JCI and the percentage of i-GTVs achieving a JCI of ≥0.7 against the gs-GTV and c-GTV was 0.67 and 0.74 (Wilcoxon signed rank test p=<0.0001) and 28% and 81% respectively. Only one i-GTV achieved a JCI <0.5 against the gs-GTV but none against the c-GTV.
In the SCOPE 1 pre-trial test case the method of creation of the reference volume increased the number of i-GTVs meeting the criteria for excellent conformity by 53%. In the context of training and revalidation, correctly identifying individuals with excellent and poor conformity is critical. This will require careful definition of a consensus volume to act as a reference volume and development of robust criteria for acceptable JCI values.