ESTRO 2022 - Abstract Book

S299

Abstract book

ESTRO 2022

Our early experience suggests that MRgART for liver cancer is feasible and safe with acceptable dosimetric parameters and treatment time. We continue to collect data and evidence on patient and clinician reported outcome, safety and tolerability of MRgART.

PD-0335 Dosimetric differences between field and volume-based regional nodal RT in the POSNOC breast trial

H. Price 1 , Z. Nabi 1 , R. Butt 1 , R. Mir 2 , P. Diez 1 , D. Dodwell 3 , S. Sadiq 4 , M. Lewis 4 , A. Goyal 5

1 National Radiotherapy Trials Quality Assurance Group, Mount Vernon Cancer Centre, Radiotherapy Physics, Northwood, United Kingdom; 2 Mount Vernon Cancer Centre, Clinical Oncology, Northwood, United Kingdom; 3 Oxford University, Nuffield Department of Population Health, Oxford, United Kingdom; 4 University of Nottingham, Nottingham Clinical Trials Unit, Nottingham, United Kingdom; 5 Royal Derby Hospital, Oncoloplastic Breast Surgery, Derby, United Kingdom Purpose or Objective POSNOC (NCT02401685) is a randomised, multicentre, non-inferiority trial comparing adjuvant RT to adjuvant RT plus axillary treatment in patients with early-stage breast cancer. Field-based RT techniques were used by most centres upon opening. Subsequent adoption of a nodal volume-based approach has been slow. Dosimetric differences between field- based and volume-based nodal plans in patients recruited to POSNOC are reported. Materials and Methods A left-sided breast cancer patient with axillary RT from 20 institutions planned with field-based techniques was evaluated for dose coverage to axillary lymph node levels 1-4 (L1-4). Breast specialists retrospectively delineated L1-4 in line with ESTRO consensus guidelines (Offersen et al, 2015); contours were checked independently by a radiation oncologist. The PTV was defined as a 5mm expansion of the CTV, then grown by 6mm to account for penumbra, allowing for direct comparison with conventional field placement. POSNOC pre-defined planning objectives: Field-based plans aimed for >90% dose to the supra-clavicular fossa (SCF) and >80% dose to mid axilla; volume-based used PTV V 90% ≥ 90%. In addition, coverage of L1-4 by field-based techniques was evaluated by suitability of field border placement and PTV V 80% ≥ 90% . Results 14/20 centres met POSNOC field-based objectives. The cranial border extended beyond L3 and L4 by 18 mm, on average. Field borders were short caudally, laterally and medially by a mean of 7, 2 and 10 mm, respectively. The PTV V 80% objective was consistently achieved for L2 and L3 (96.5% and 92.9%, respectively), however not so for L1 and L4 (88.3% and 78.8%, respectively), where results were variable (Figure 1). Loss of L1 coverage was mostly in the cranial aspect due to short lateral axillary borders; L4 coverage was compromised medially where borders were shortened to reduce dose to lung and spinal cord. Use of oblique fields at some centres improved coverage. V 90% ≥ 90% was met only by 4/20 centres; although possible with field-based techniques, absence of volumes made it difficult. One patient was identified as an outlier (>2 standard deviations from the mean) but was included in the analysis. The centre was informed.

Conclusion POSNOC aimed to treat the SCF and axillary regions, whereas ESTRO consensus guidelines extend nodal inclusion to the caudal aspect of L4. As such, the POSNOC-defined field borders failed to encompass L4 in its entirety. A field-based approach for left-sided axillary RT provided acceptable dose coverage in terms of V 80% for all combined nodal levels and for L2 and L3, specifically. However, this technique resulted in under-coverage of L1 and L4. In addition, excess normal tissue was irradiated cranially. V 90% ≥ 90% was often not met. Field-based planning for breast and axillary treatment may fail to fully encompass L1-4. A volume-based approach would provide a target for which parameters such as field size, beam orientation, and energy could be optimised to achieve adequate coverage of the nodal region.

PD-0336 Evaluation of a deep-learning segmentation software in thoracic organs at risk: an early analysis