ICHNO-ECHNO 2022 - Abstract Book

S38

ICHNO-ECHNO 2022

In conclusion, the use of OE-MRI in patients with HNSCC on a clinical MR scanner is feasible, well tolerated and offers potential advantages over other tumour hypoxia measuring techniques. Further work is underway to correlate scan results to treatment outcomes in a larger patient cohort. Funded by a grant from Nottingham Hospitals Charity (APP2361/N0379).

Poster: HPV or EBV related cancers

PO-0069 Augmenting Inter-rater Concordance of Radiologic Extranodal Extension in HPV+ Oropharyngeal Cancer

F. Hoebers 1 , E. Yu 2 , B. O’Sullivan 3,6 , A. Postma 4,5 , W. Palm 4 , E. Bartlett 7 , J. Su 8 , W. Xu 8 , S.H. Huang 9,6

1 Maastricht University Medical Centre, Department of Radiation Oncology (Maastro), Maastricht, The Netherlands; 2 Princess Margaret Cancer Centre, University of Toronto, Department of Neuroradiology and Head and Neck Imaging, Toronto, Canada; 3 Princess Margaret Cancer Centre, University of Toronto, Department of Radiation Oncology, Toronto, Canada; 4 Maastricht University Medical Centre, Department of Radiology, Maastricht, The Netherlands; 5 Maastricht University , School for Mental Health and Sciences, Maastricht, The Netherlands; 6 Princess Margaret Cancer Centre, University of Toronto, Department of Otolaryngology – Head & Neck Surgery, Toronto, Canada; 7 Princess Margaret Cancer Centre, University of Toronto, Department of Neuroradiology and Head and Neck Imaging, Toronto, Canada; 8 Princess Margaret Cancer Centre, University of Toronto, Department of Biostatistics, Toronto, Canada; 9 Princess Margaret Cancer Centre, University of Toronto, Department of Radiation Oncology, Toronto, Canada Purpose or Objective The presence of radiologic extranodal extension (rENE) has been associated with poor survival in HPV+ oropharyngeal cancer (OPC) patients in several retrospective single center studies using single readers to review diagnostic images. This study aims to assess inter-institutional concordance and identify methods to augment rENE ascertainment in HPV+ OPC. Materials and Methods The rENE assessment comprised two-phases. In phase I, pre-treatment CT/MRI of randomly selected 20 cases were reviewed by 4 radiologists (2 from each of 2 institutions) based on individual a priori appreciation of literature criteria concerning presence/absence of rENE, rENE pattern (pattern 1, 2a, 2b, and 3) and level of certainty (<50%, 50-75%, and >75%) for rENE declaration. The results were verbally deliberated and reconciled, and experiences shared among the 4 radiologists. Operational definitions and methods of assessment were consolidated. In phase II, CT/MRI of a subsequent random 30 cases were reviewed by the same radiologists. Inter-rater Fleiss Kappa values and percentage (%) of agreement in rENE assessment among the 4 radiologists were calculated using >50% and >75% certainty respectively. The results were linked to disease recurrence data. Results In phase I, the interrater Fleiss’ kappa for rENE+ at >50% certainty level was 0.38 (95 % confidence interval 0.20-0.56) and improved to 0.52 (0.34-0.70) for certainty >75%. In phase II, inter-rater kappa for >50% and >75% certainty level improved substantially to 0.84 (0.74-0.94) and 0.93 (0.83-1.00) (see table). The % of agreement at certainty of >50% and >75% among the observers increased from 70%-90% and 70%-95% in phase I to 87-97% and 93-100% in phase II, respectively. Of 13 cases with recurrence (4 regional and 9 distant) in Phase II, the presence of rENE was identified by all 4 radiologists. During group discussion, we re-reviewed discrepant cases to develop qualitative methods to augment rENE ascertainment. This included assessing general quality of the images, e.g. patient movement during scan; importance of multiple radiologic planes to assess infiltration into surrounding tissues, or the presence of coalescent nodes (i.e. loss of internodal fat planes); focusing on the center of the metastatic node rather than proximal or distal extent to minimize partial volume effects, since rapid contour change at the extreme ends of nodal capsules might create the illusion of ill-defined nodal border. Lastly, in case of uncertainty about presence of rENE, it was agreed to down-classify to rENE negative.

Conclusion Our study shows that rENE can be reliably ascertained across institutions and improved after consolidation of rENE operational definitions. Final kappa values were indicative of substantial agreement, especially for higher levels of certainty concerning diagnosis of rENE. Our study demonstrates a learning curve of rENE assessment and proposes a strategy to augment the reliability of rENE ascertainment.