S1538
Interdisciplinary - Quality assurance and risk management
ESTRO 2026
Digital Poster Highlight 4098
most frequently observed for liver dose (26 instances), planning target volume (PTV) coverage (12), and lung dose (11) (Fig. 1).
Automated planning as a tool for radiotherapy trials quality assurance: protecting the patient and the clinical trial Geraint J Lewis 1,2 , Sarah Gwynne 3,2 , Sarah Bridges 4 , Lisette Nixon 4 , Tom Crosby 5 , Philip Wheeler 1 1 Radiotherapy Physics, Velindre University NHS Trust, Cardiff, United Kingdom. 2 National RTQA, Radiotherapy Trials Quality Assurance (RTTQA) group, London, United Kingdom. 3 Clinical Oncology, South- West Wales Cancer Centre, Swansea, United Kingdom. 4 Centre for Trials Research, Cardiff University, Cardiff, United Kingdom. 5 Clinical Oncology, Velindre University NHS Trust, Cardiff, United Kingdom Purpose/Objective: Radiotherapy Quality Assurance (RTQA) is a vital aspect of clinical trials, ensuring that radiotherapy delivery adheres to protocol specifications and that study outcomes are not confounded by variations in treatment quality. This study builds upon prior research undertaken to calibrate and clinically validate a Protocol-Based Automatic Iterative Optimisation (PB- AIO) model for automated radiotherapy planning in oesophageal cancer. The purpose of this work was to evaluate the application of this automated planning model as a comparator to support RTQA within a clinical trial setting. Material/Methods: A site-specific ‘autoplan’ protocol was developed and validated for clinical use. Automated plans were generated for a cohort of 108 patients from the SCOPE 2 clinical trial for oesophageal carcinoma. Residual values for each planning metric were calculated through subtraction of the dose values for each plan from the paired automated plan. Comparative analyses between automated and manually optimised plans were performed through analysis of residual values for a range of plan quality metrics. Planning metrics were highlighted as outliers where the calculated residual values were >1.5 times the inter- quartile range higher or lower than the first or third quartile of the dataset. Positive and negative outliers were defined this work as outliers which were superior or inferior, respectively, than the automated comparator plan. The outlier data were analysed for trends over time, by centre, by planning metric and by frequency to isolate trends in the data. Results: The PB-AIO model produced high-quality automated plans with notable improvements in target coverage and organ-at-risk sparing compared to manually- optimised submissions to the trial.Analysis of residuals revealed inter-centre variation in plan quality across the trial. Negative outliers—indicating poorer performance than the automated reference—were
On average, 0.6 negative and 0.2 positive outliers per plan were recorded, with a maximum of three negative outliers observed in plans from a single centre (Fig. 2). No consistent trend in plan quality improvement or decline over time was detected.
Conclusion: This study demonstrates that PB-AIO automated planning provides a robust, objective reference for identifying suboptimal radiotherapy plans in clinical trials. Using such models as comparators can enhance RTQA processes, enable targeted feedback to participating centres, and promote plan optimisation. Integrating automated planning into prospective QA review may improve efficiency, consistency, and ultimately contribute to better patient outcomes. References: Wheeler, P. A., et al. (2019a). ‘Utilisation of pareto navigation techniques to calibrate a fully automated radiotherapy treatment planning solution’ Phys
Imaging Radiat Oncol, 10 pp. 41-48. DOI: 10.1016/j.phro.2019.04.005 Available at:
https://www.ncbi.nlm.nih.gov/pubmed/33458267.Helb row, J., et al. (2025). ‘Radiotherapy quality assurance in the scope2 trial: What lessons can be learned for the next uk trial in oesophageal cancer?’ Clin Oncol (R Coll Radiol), 38 p. 103735. DOI:
Made with FlippingBook - Share PDF online