1055 - Implementation of Knowledge-Based Planning as a Quality Assurance Feedback Tool for a Multicentre Prostate Clinical Trial (TROG 18.01 NINJA)
Presenter(s)
J. M. Martin1, M. Sidhom2, P. Banyer3, S. Porter3, A. J. Moore4, and O. M. Cook5; 1Radiation Oncology Calvary Mater Newcastle Hospital, Newcastle, Australia, 2Liverpool Cancer Therapy Centre, Sydney, Australia, 3TROG Cancer Research, NEWCASTLE, NSW, Australia, 4University of Newcastle, Newcastle, Australia, 5TROG Cancer Research, Newcastle, Australia
Purpose/Objective(s): Quality assurance (QA) of clinical trial radiation therapy (RT) treatment and planning is essential to ensure protocol compliance and trial quality. Knowledge based planning (KBP) uses machine learning to generate a knowledge based dose-volume histogram (DVH) estimation model. TROG18.01 NINJA aims to compare two emerging schedules of RT in the treatment of intermediate/high-risk prostate cancer. A novel approach to plan QA was implemented, using KBP, in addition to standard RT QA peer review processes. The study aimed to assess the impact and feasibility of prospective plan quality feedback using KBP for the NINJA trial, with a secondary study outcome being the rate of replanning following KBP feedback to centers.
Materials/Methods: The RTQA program utilized real time review (RTR), and all centers were required to submit RT treatment plans for trial participants in DICOM format to central QA seven days prior to treatment. Across 243 patients recruited from 12 participating centers, 100 (41.2%) underwent pre-treatment (real-time) KBP feedback and 86 (35.4%) were retrospectively analyzed. 57 (23.5%) sampled cases were unable to be analyzed using KBP due to incompatible treatment methods, import errors or unconventional gantry configurations. A feedback report comparing the KBP generated DVH versus the initial plan was collated using a customized software script and sent to sites within 24 hours. Protocol compliance was assessed including dose constraints for the CTV, PTV, rectum, bladder, urethra, penile bulb, femoral heads, sigmoid and small bowel. Centers reviewed the report and determined whether they would amend their clinical plan. Sites were then asked to complete a survey providing feedback on the KBP tool.
Results: High clinical plan quality was observed, with only 3 (3%) clinical plans amended following RTR KBP feedback. It was observed that the replan submissions resulted in improved OAR dosimetry, most notably for the rectum and bladder. Across all cases, there was a high level of agreement between the KBP generated plan and the clinical plan. Of the organ at risk (OAR) parameters assessed, 56% (1,690) demonstrated no variation between the KBP and the clinical DVH, 25% (731) showed the clinical plan could be improved and in 19% (580) the submitted plan was preferable. Feedback was gathered via a questionnaire and email correspondence. KBP feedback helped identify dosimetric outliers and confirm plan quality. Key reasons for not modifying clinical plans included confidence in existing plans, reluctance to increase modulation, time/resource constraints, and viewing DVH differences as clinically minor.
Conclusion: The use of KBP real-time plan quality feedback was successfully implemented for the NINJA trial. The KBP model generated robust plans when compared to the clinical site submitted plan. This novel approach provided a valuable plan quality benchmark in which to assess the clinical plan.