A total of 704 PBRT patients treated by a single surgeon (DHC) for uveal melanoma (1996-2010) were reviewed for NVG in our prospectively maintained database. All patients received 56 GyE in 4 fractions. Median follow-up was 58.3 months. Analyses included the Kaplan-Meier method to estimate NVG distributions, univariate log-rank tests, and Cox's proportional hazards multivariate analysis using likelihood ratio tests to identify independent risk factors of NVG among patient, tumor, and dose-volume histogram parameters.
The 5-year PBRT NVG rate was 12.7 % (95 % confidence interval [CI] 10.2 % -15.9 % ). The 5-year rate of enucleation due to NVG was 4.9 % (95 % CI 3.4 % -7.2 % ). Univariately, the NVG rate increased significantly with larger tumor diameter (P<.0001), greater height (P<.0001), higher T stage (P<.0001), and closer proximity to the disc (P=.002). Dose-volume histogram analysis revealed that if >30 % of the lens or ciliary body received ¡Ý50 % dose (¡Ý28 GyE), there was a higher probability of NVG (P<.0001 for both). Furthermore, if 100 % of the disc or macula received ¡Ý28 GyE, the NVG rate was higher (P<.0001 and P=.03, respectively). If both anterior and posterior doses were above specified cut points, NVG risk was highest (P<.0001). Multivariate analysis confirmed significant independent risk factors to include tumor height (P<.0001), age (P<.0001), % disc treated to ¡Ý50 % Dose (<100 % vs 100 % ) (P=.0007), larger tumor diameter (P=.01), % lens treated to ¡Ý90 % Dose (0 vs >0 % -30 % vs >30 % ) (P=.01), and optic nerve length treated to ¡Ý90 % Dose (¡Ü1 mm vs >1 mm) (P=.02).
Our current PBRT patients experience a low rate of NVG and resultant enucleation compared with historical data. The present analysis shows that tumor height, diameter, and anterior as well as posterior critical structure dose-volume parameters may be used to predict NVG risk.