We did a multivariate analysis of 1262 patients with operable, invasive ductal breast carcinoma to assess the prognostic value of the Scarff-Bloom-Richardson (SBR) histologic grading system. Nodal metastasis and SBR were the two most important factors for metastasis-free survival (MFS), P = 10-9 and P = 10-5, respectively, for total study time. In patients who were node negative, the SBR and International Union Against Cancer (UICC) stages were the most important for MFS (P = 4 X 10-4 and P = 0.03). In order to try to improve the SBR prognostic value, we first studied the three components of the SBR separately: ductoglandular differentiation proved the least predictive and nuclear pleomorphism and mitotic index the most predictive. A rearrangement of the two nuclear scores alone produced higher risk values and better risk separation of patient subpopulations than SBR, and eliminated the SBR from the multivariate model. This rearrangement, modified SBR (MSBR), defined five new risk subgroups with statistically different risk ratios for MFS (P = 3 X 10-8). SBR grade II (55% of patients) was separated into three MSBR groups significantly different according to MFS (P = 0.008). In the patients who were node negative, MSBR replaced the SBR and was the most important factor for prediction of relapse of MFS (P less than 0.00001). The MSBR is more accurate and predictive than the standard SBR grade and is particularly useful when the nodal status of the patient is negative or unknown.