Dosimetric Comparison based on Consensus Delineation of Clinical Target Volume for CT- and MR-Based Brachytherapy in Locally Advanced Cervical Cancer Akila N. Viswanathan, MD, MPH Brigham and Women’s Hospital Dana-Farber Cancer Institute Harvard Medical School ASTRO Conference /15/14

Objective To compare dosimetric results based on RTOG atlas consensus contours for computed tomography (CT) and 3 Tesla magnetic resonance (MR) image-based cervical-cancer brachytherapy (BT)

Material and Methods 23 gynecologic radiation oncology experts contoured 3 cervical-cancer brachytherapy cases: Case 1: tandem and ovoid Case 2: ovoid with needles Case 3: tandem and ring Contours were analyzed for consistency and clarity of target delineation using expectation maximization algorithm for simultaneous truth and performance level estimation (STAPLE) with kappa statistics Contours were reviewed and modified by experts in order to build an online atlas for use in clinical trials Using these consensus contours, dosimetric calculations of D90 and D2cc to the bladder, rectum and sigmoid were performed using: Point A calculations were compared to optimized and optimized CT compared to optimized MR 1) CT and MR images normalized to standard point A 2) CT and MR images optimized to ensure that CT or MR CTV D90 was maximized while reducing doses to organs at risk (OAR)

Results Table 1: Mean DVH values (cGy) for 3 cervical-cancer cases based on consensus brachytherapy contours StandardManual Optimization CTMRCTMR CT-CTV D SD: 0.4SD: 3.1 MR-CTV D SD: 0.3SD: 15.6 Rectum D2cc SD: 52.8SD: 32.4 Bladder D2cc SD: 0.6SD: 15.1 Sigmoid D2cc SD:13.4SD: 0.5 Of the 23 physicians contouring, 50% reported using MR for BT contouring routinely ‒ 90% use CT for BT planning either in addition to or instead of MR routinely Despite a consistently larger volume on CT than MR for all 3 cases (p<0.001), there was no statistically significant difference in D90 or D2cc OAR comparing CT to MR Case 1 (large tumor, intermediate response) had the largest difference between CT and MR volumes and also had the largest difference between CT - and MR-optimized D90

Table 2: CT and MR imaged-based contours for each method from two different cases Normalized to Point A Optimized CT MR Conclusion and Future Direction MR-contoured volumes were consistently smaller than CT volumes, particularly in cases with parametrial extension These differences did not, however, translate into significant dosimetric differences in the D90 tumor or D2cc of the rectum, sigmoid or bladder for optimized plans (all p >0.05) These findings will be validated by a large group of physicists creating plans. Future clinical trials may implement CT- or MR- based BT planning Case 1: (Top) Axial view (Bottom) Sagittal view