We compared the ability of original nuclear grades from surgical pathology reports and grades reviewed by a urologic pathologist to predict death due to renal cell carcinoma (RCC) for 2,042 patients treated with radical nephrectomy between January 1970 and December 1998. Reviewed grade I tumors had small, round nuclei with inconspicuous nucleoli visible at x400; grade 2 contained round to slightly irregular nuclei with mildly enlarged nucleoli visible at x200; grade 3 had round to irregular nuclei with prominent nucleoli visible at x100; grade 4 contained enlarged pleomorphic or giant cells. Predictive abilities were compared using R2 values from Cox proportional hazards models. There were 1,733 (84.87%) clear cell, 222 (10.87%) papillary, and 87 (4.26%) chromophobe tumors. Reviewed grades were more predictive of death due to RCC than original grades for clear cell (R2, 21% vs 16%), papillary (R2, 16% vs 13%), and chromophobe (R2, 39% vs 27%) RCC. Among patients with clear cell and papillary RCC, this difference was apparent even after adjusting for the 1997 TNM stage. Standardized nuclear grades were more predictive of death due to RCC than nonstandardized grades for all subtypes studied.