Abstract Book

S322

ESTRO 37

harmonize and improve radiotherapy QA in clinical trials. Dosimetric intercomparisons are one important tool in this process. The purpose of this intercomparison was to understand the differences between the QA methods used by participating national QA groups Material and Methods Three IMRT plans were calculated on four different QA phantoms using Pinnacle TPS: ACDS 2D system (slab phantom + PTW 2D array 1500), RTTQA (ArcCheck), TROG (VESPA-EPID panel) and NPL (Octavius 4D + PTW 2D array 1500). Plans were delivered on the NPL Elekta linac (VERSA HD). The results of gamma analysis passing rates (γPR) from the different QA systems and different versions of the same analysis tool were compared. Measurement data were converted to Verisoft format (ACDS) in order to eliminate one source of variation (gamma analysis algorithm). Results A summary of the results of the γPR (%) for global planar gamma analysis with 3%/3mm and 20% threshold is shown. Overall, the lowest γPR were for ACDS 2D system and ArcCheck (average of 93.7 and 93.9% resp.). For Octavius 4D, there were minor differences between a given plane global 3%/3mm γPR in comparison with the γPR averaged over all planes. There were also differences (up to 1%) between the 3D γPR per plane and the 3D volume analysis. Overall, and as expected, for each acceptance criteria, the 3D γPR was higher than the 2D γPR. Systems that reconstruct 3D dose based on planar 2D measurement (Octavius 4D and VESPA) gave a γPR 2D >2% higher than the system that performs only 2D analysis based on similar planar 2D measurements (2D system). From nineteen fields delivered to the EPID and analysed by VESPA, 100% had a γPR at the optimal level as established by the ACDS 2D system (≥97.5% at 3%/3mm). For the fields delivered to the 2D system and analysed with Verisoft 6.2 only 85% of the fields comply with the ACDS audit outcome criteria (74% for Verisoft 6.1). ANOVA statistical analysis, run for each plan and all audits groups, when their own g analysis or the ACDS system was used, does not show statistically significant differences (F Conclusion The use of different measurement systems from different audit groups led to results that differed, sometimes substantially, even when assessing the same data- sets/TPS/linac combination. This can make it complex to interpret each group’s results. This intercomparison exercise reinforces the need for harmonization. OC-0611 Modelling the clinical impact of machine specific dose variations on outcome using national data M. Bolt 1 , C. Clark 1 , T. Chen 2 , A. Nisbet 3 1 Royal Surrey County Hospital, Medical Physics, Guildford, United Kingdom 2 University of Surrey, Chemical & Process Engineering, Guildford, United Kingdom 3 University of Surrey, Physics, Guildford, United Kingdom Purpose or Objective The accuracy of dose delivery during radiotherapy treatments depends upon a number of patient and equipment factors, including the current treatment

machine calibration and output. Modern treatment techniques and improved image guidance reduce the variation in delivered dose due to planning and setup. A result of this is that control of treatment machine calibration is of greater significance in the overall delivery uncertainty. The impact of calibration errors and uncertainties in daily dose delivery on clinical outcomes in radiotherapy has been modelled using evidence based The dose variation introduced by 3D conformal planning and IMRT planning has been assessed through extraction of dose statistics from the PARSPORT head and neck trial [1]. A dataset containing >24,000 measurements of beam output from 204 UK radiotherapy treatment machines [2] was combined with a national dataset containing dosimetry audit measurements [3] to assess the potential dose variability for patients treated on individual machines, within a single centre or nationally. Radiobiological modelling was completed for prostate and head and neck cancers incorporating the daily dose variability. Results IMRT planning techniques within PARSPORT not only reduced dose delivered to the OARs, but also the variation in dose to the PTV across the trial population. The D99 standard deviation reduced by around 2/3 to 3% when IMRT techniques were used whilst also increasing PTV coverage (fig 1). This is of the same magnitude as dose variations due to beam calibration. Radiobiological modelling predicts that variation in beam calibration and output introduces a 10% variation in predicted TCP (2 year survival) and 6% in NTCP (xerostomia) for patients in the PARSPORT trial. For prostate cancer this results in variations of TCP (bPFS) and NTCP (grade 1-2 proctitis) of 3% and 4% respectively. This variation is seen to be possible within a single treatment centre. radiobiological models. Material and Methods

Conclusion Control of fundamental dosimetry within the radiotherapy chain remains highly significant even with modern treatment technologies. Variation in the control of beam calibration provides a potential source of uncertainty which could lead to measurably different outcomes for patients treated on different machines. Whilst fundamental dosimetry measurement has improved over time, the advances in treatment technology now mean that the impact of fundamental dosimetry is more significant than ever. As technology improves, the tolerances should also be reduced appropriately to best make use of these advances. This variation may have a direct effect both on small local trials as well as large scale clinical trials and should be included in their considerations of uncertainties. References [1] Nutting et al 2011 Lancet Oncol. 2011 Feb;12(2):127- 136. [2] Bolt et al Phys. Imaging Radiat. Oncol. [Under review] [3] Thomas et al 2017 Phys. Imaging Radiat. Oncol. 3 21–7