UNLABELLED We compared survival experiences for three groups with uveal melanoma: (1) 120 patients treated by proton beam irradiation from 1975 to 1981; (2) 235 patients treated by enucleation from 1953 to 1973 (enucleation group 1); and (3) 161 patients treated by enucleation from 1975 to 1981 (enucleation group 2). The following variables were predictive of melanoma-related deaths. TREATMENT rate ratio (ratio of the rate of death for one category of the variable relative to the rate of death for another category) of 6.32 (95% confidence interval 1.70 to 23.51) for enucleation group 1 vs proton beam, 3.06 (0.81 to 11.54) for enucleation group 2 vs proton beam, and 2.07 (1.05 to 4.07) for enucleation group 1 vs enucleation group 2. Size: rate ratio of 3.61 (1.51 to 8.64) for large (16+ mm) vs small (less than 11 mm) tumors. LOCATION rate ratio of 3.19 (1.56 to 6.53) for tumors anterior vs posterior to the equator. When death from all causes was evaluated as the outcome, the estimated rate ratios were similar. When melanoma metastasis was evaluated as the outcome, the estimated rate ratios were reduced in magnitude.