ESTRO 37 Abstract book

ESTRO 37

S483

Improvements in feasibility score correlate well with reductions in the area under DVH curves. The most significant change is for ipsi-ilateral kidney, the area under the DVH curve decreases from 1076 Gycc with RP v15.5 to 641 Gycc with MCO v15.5 (p = 0.0003). Changes in the area under OAR DVH curves correlate well to changes in Feasibility with a Pearson correlation of 0.9. Conclusion Improved PlanIQ feasibility scores and DICE coefficients with MCO v15.5 in comparison to RP v15.5 indicates that MCO generated plans are closer to the hypothetical ideal plan and the achieved DVHs better match the shape of idealized DVH curves. This is consistent with MCO Pareto surface planning concepts. PO-0905 Data mining for pre-plan evaluation of organ at risk priority D. Christophides 1 , A. Gilbert 2 , A. Appelt 2 , J. Lilley 1 , D. Sebag-Montefiore 2 1 St. James's University Hospital, Leeds Cancer Centre, Leeds, United Kingdom 2 University of Leeds, Leeds Institute of Cancer and Pathology, Leeds, United Kingdom Purpose or Objective Radiotherapy aims to provide dose coverage of the planning target volume (PTV) while adhering to fixed dose limits for organs at risk (OARs). Developments like volumetric modulated arc radiotherapy and multi-criteria optimization (MCO)[1] can be used to create a large number of plans per patient that can satisfy these constraints. A recent study[2] has demonstrated the use of MCO to create clinically acceptable plans that prioritised the sparing of different OARs. In this work we investigate the feasibility of using normal tissue complication probability (NTCP) models and data mining to establish such OAR priorities prior to planning. Material and Methods Data from 326 patients treated with pelvic radiotherapy were made available from a study (DRF-2012-05-201) on optimising patient outcomes. Patient-reported outcome measure (PROM) scores were reported on a four-level scale on dyspareunia (“Did you have pain or discomfort during intercourse?”) by 72 out of 221 female patients, as well as scores from 249 patients on bowel urgency (“When you felt the urge to move your bowels, did you have to hurry to get to the toilet?”). Multivariate NTCP models were developed to predict the PROM score for both symptoms using clinical parameters and principal components (PCs) from dose volume histograms (DVHs)[3]. For the dyspareunia model, age and vagina PC5 were significant, whereas the body mass index (BMI) and mid-rectum PC2 were significant for bowel urgency (Wald test p-values < 0.05). Population-level DVHs were estimated at the 10%, 30%, 50%, 70% and 90% distribution ranges from the DVHs available from the vagina (N=221) and mid-rectum (N=143) structures. In combination with individual patient risk factors, they were used to predict a range of potential NTCP values for PROM score ≥ 2 (≥ ‘Quite a bit’) for dyspareunia and bowel urgency prior to treatment planning. Results A visual assessment of the population level DVHs demonstrated the differences between the doses received by the vagina and mid-rectum structures (Fig. 1). Using these data, assessment charts were constructed for each patient to decide which OAR to prioritise.

An example is shown (Fig. 2) where patient 1 (age:60 years, BMI:39.4) was compared to patient 2 (age:25 years, BMI:23.2). In this scenario, patient 1 had a lower predicted risk of dyspareunia compared to the risk of bowel urgency, thus the mid-rectum could be set to a higher priority compared to the vagina. Conversely, for patient 2, a higher priority could be placed on the vagina structure compared to the mid-rectum.

Conclusion The developed method can be used to provide pre-plan estimates on complication risk on a patient-to-patient