Purpose: New knowledge on normal tissue response requires updated NTCP models. Based on accumulated experience translating such models into treatment planning optimization goals, this study derived evidence-based dose-volume criteria from the recently published RTOG 0617 overall survival (OS) model.
Methods: The published model included Atria D45%, Mean Lung Dose (MLD), Pericardium MOH55%, and Ventricles MOH5%. This study focused on reducing Pericardium MOH55% since MLD is already constrained internally (MLD<20Gy), Atria and Ventricles are included in the Pericardium, and the strongest association with OS was observed with Pericardium dose. Objective function input in commercial treatment planning systems is limited to Dx and Vx. Hence, MOH55% was most strongly represented by a combination of D15 and pericardium mean dose (PMD, Spearman’s rho: 0.93, 0.99). These two variables were also among the best OS predictors in RTOG 0617. Dose-volume goals were D15<6Gy and PMD<5Gy, which represented low risk in the OS model. Nine of 105 randomly selected internal patients previously treated to 60Gy in 30 fractions with concurrent chemo-radiation and durvalumab were re-planned incorporating D15 and PMD goals using standard planning practice. Pericardium was generated from our open-source deep-learning algorithm.
Results: Compared to the original plans, D15 was on average reduced from 43Gy to 29Gy; PMD was reduced from 20Gy to 13Gy; the normalized relative reduction ranged between 6-72% and 7-64%, respectively. Only 2/9 patients met D15<6Gy and PMD<5Gy; the majority (7/9 patients) were reduced to an intermediate-moderate risk (D15<49Gy and PMD<20Gy). Pericardium sparing was less in patients with a larger (>8%) Pericardium-PTV overlap; median relative reduction: 18% vs. 58% for D15 and 20% vs. 48% for PMD.
Conclusion: This study has demonstrated a practical treatment planning-based approach to translate a published NTCP model that included multiple and new organs and novel dose parameterizations to form dose-volume criteria for use in treatment planning.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by the MSK Cancer Center support grant/core grantP30 CA008748 and NCI 5 R01 CA198121-04
Not Applicable / None Entered.
Not Applicable / None Entered.