Comparative Evaluation of Treatment Planning Systems Used in Lung SBRT

Abstract

The effectiveness of lung SBRT depends on the choice of treatment planning system (TPS) used for dose calculations. Whereas Monte-Carlo (MC) based TPS calculate the most accurate dose, conventional planning systems cannot adequately account for dose variation caused by heterogeneous lung tissue. We aim to evaluate the performance of several commercial TPS used in lung SBRT. We hypothesize that the use of MC-based TPS will result in more accurate and consistent delivered doses in lung SBRT. SBRT plans for 30 early-stage lung cancer patients (31 targets) were retrospectively evaluated in this multi-institutional study. Patient data with structure sets were anonymized and sent to four participating institutions to plan on five different TPS: one pencil-beam (PB), two convolution-superposition (CS1 & CS2), and two Monte Carlo (MC1 & MC2) systems. A test case with homogenous slab phantom was also included to allow benchmarking of each TPS. Patients were grouped according to PTV size: small (<25cc), medium (25 – 60cc), large (>60cc); and location: island, adjacent to chest-wall/mediastinum, and mixed. All patients were initially planned with MC1TPS to deliver 50Gy PTV dose in 5 fractions using 7-12 non-coplanar, conformal, 6 MV fields. Treatment plans were optimized for > 95% PTV dose coverage as per the Radiation Therapy Oncology Group guidelines. Using the same beam parameters, calculation geometry, and MUs, all patient plans were re-computed on other TPS. Dose-volume-histograms (DVH) for PTV and organs-at-risk (OARs) including lung, esophagus, trachea, heart, and spinal cord were evaluated. Two-tailed student t-test was used for statistical analysis. Average PTV was 53.5cc (range: 8.6 – 149.4cc). The largest dose difference was found between MC1 and PB for small island targets. Compared to MC1, average values of PTV Dmean for all patients were higher by 12.5%, 3%, 0.5% and 2% respectively in PB, CS1, CS2 and MC2 TPS (p < 0.01). Larger dose differences were seen for PTV Dmin: 21.8%, 4.6%, 10.6% and 12.8% (p < 0.01), indicating greater sensitivity of dose calculation algorithms in the target periphery due to density changes. Whereas the prescription dose conformality (CI) was similar in CS and MC TPS, dose spillage (D2cm) was 6.2%, 4.8%, 1.5% and 0.9% higher respectively for PB, CS1, CS2 and MC2 (p< 0.01). Compared to MC1, average lung V20 showed an 11.4% increase in PB, but was 2.4%, 2.4% and 4.4% lower in CS1, CS2 and MC2 TPS respectively (p < 0.01). Other OARs showed similar trend with MC1 having closest agreement with MC2, followed by CS2, CS1 and PB. Conventional TPS for lung SBRT result in significant dose variation depending on the calculation algorithm. MC based TPS provide the greatest accuracy and consistency in delivered doses, thereby allowing easier plan comparison among different institutions. In light of observed dose differences, current RTOG dose guidelines may require further evaluation or revision.