Corneal Substructure Dosimetry Predicts Corneal Toxicity in Patients With Uveal Melanoma Treated With Proton Beam...

This study examines the relationship between dose to corneal substructures and incidence of corneal toxicity within 6 months of proton beam therapy (PBT) for uveal melanoma. We aim to develop clinically meaningful dose constraints that can be used to mitigate corneal toxicity.

Ninety-two patients were treated with PBT between 2015 and 2017 and evaluated for grade 2+ (GR2+) intervention-requiring corneal toxicity in our prospectively maintained database. Most patients were treated with 50 Gy (relative biological effectiveness [RBE]) in 5 fractions, and all had complete six-month follow-up. Analyses included Mann-Whitney, χ², Fisher exact, and receiver operating curve tests to identify risk factors for GR2+ toxicity. Bivariate logistic regression was used to identify independent dose-volume histogram (DVH) predictors of toxicity after adjustment for the most important clinical risk factor.

The 6-month PBT GR2+ corneal toxicity rate was 10.9%, with half of patients experiencing grade 2 toxicity and half experiencing grade 3 toxicity, with no grade 4 events. Patients with anterior chamber tumors had a higher risk (58.3%) for toxicity than those with posterior tumors (0%) or posterior tumors extending past the equator (25%, P < .0001). On univariate analysis, larger size according to Collaborative Ocular Melanoma Studies was associated with increased toxicity rate (P < .004). DVH analysis revealed that cutoffs of 58% for V25, 32% for V45, 51.8 Gy (RBE) for maximum dose, and 32 Gy (RBE) for mean dose to the cornea separated patients into groups experiencing and not experiencing toxicity with 90% sensitivity and ≥96% specificity. Bivariate logistic regression indicated that corneal V25, V45, and mean dose independently predicted for toxicity after adjusting for tumor location.

Patients receiving PBT for anterior uveal melanomas experience a high rate of GR2+ corneal toxicity because of increased corneal dose. Anterior location and corneal DVH parameters independently predict toxicity risk. We propose dosimetric constraints to facilitate treatment planning and toxicity mitigation.