Tumor Control Probability and Time-Dose-Response Modeling for Stereotactic Radiosurgery of Uveal Melanoma.

ABSTRACT

PURPOSE: Uveal melanoma (UM), although a rare malignancy, stands as the most prevalent intraocular malignancy in adults. Controversies persist regarding the dose dependency of local control (LC) through radiation therapy. This study sought to elucidate the significance of the prescription dose by employing time-dose-response models for patients with UM receiving photon-based stereotactic radiosurgery (SRS). METHODS AND MATERIALS The analysis included patients with UM treated between 2005 and 2019. All patients underwent single-fraction SRS. Datapoints were separated into 3 dose groups, with Kaplan-Meier analysis performed on each group, from which time-dose-response models for LC were created at 2, 4, and 7 years after SRS using maximum-likelihood fitted logistic models. RESULTS: Outcomes from 594 patients with 594 UMs were used to create time-dose-response models. The prescribed doses and the number of patients were as follows 17 to 19 Gy (24 patients), 20 Gy (122 patients), 21 Gy (442 patients), and 22 Gy (6 patients). Averaged over all patients and doses, LC rates at 2, 4, and 7 years were 94.4%, 88.2%, and 69.0%, respectively. Time-dose-response models for LC demonstrated a dose-dependent effect, showing 2-year LC rates of more than 90% with 20 Gy and 95% with 22 Gy. For 4 years and a LC of 90%, a dose of approximately 21 Gy was required. After 7 years, the 21 Gy prescription dose was predicted to maintain a LC above 70%, sharply declining to less than 60% LC with 19 Gy and less than 40% with 18 Gy. CONCLUSIONS: In contrast to prior findings, the time-dose-response models for UM undergoing photon-based SRS emphasize the critical role of the prescription dose in achieving lasting LC. The dose selection must be carefully balanced against toxicity risks, considering tumor geometry and individual patient characteristics to tailor treatments accordingly.