PURPOSE: This study reports local tumor control and survival after proton beam radiotherapy (PBRT) of uveal melanoma. It identifies the risk factors for local tumor-control failure and for ocular tumor-related death. It presents the improvements implemented to increase the rate of local tumor control, and compares the survival rate of patients with locally controlled tumors to those of patients who had to receive a second treatment. PATIENTS AND METHODS: We have treated 2,435 uveal melanomas with PBRT between March 1984 and December 1998. Data were analyzed as of September 1999. Patients' age ranged from 9 to 89 years; there were 1,188 men and 1,247 women. The largest tumor diameter ranged from 4 to 26 mm, and tumor thickness from 0.9 to 15.6 mm. Median follow-up time was 40 months. RESULTS: Local tumor control probability at 5 years was improved from 90.6 +/- 1.7% for patients treated before 1988, to 96.3 +/- 0.6% for patients treated between 1989 and 1993, and became 98.9 +/- 0.6% for patients treated after 1993. Among 2,435 treated patients, 73 (3%) had to receive a second treatment because of tumor regrowth. Cause-specific survival at 10 years was calculated to 72.6 +/- 1.9% for patients with controlled tumors compared to 47.5 +/- 6.5% for those with recurrent tumors. CONCLUSION: Reduced safety margins, large ciliary body tumors, eyelids within the treatment field, inadequate positioning of tantalum clips, and male gender were identified to be the main factors impairing local tumor control. The improvement of local tumor control rate after 1993 is attributed to changes implemented in the treatment procedure. Our data strongly support that the rate of death by metastases is influenced by local tumor control failure: improvement of the local tumor control rate results in a better survival rate.