Questions of reproducibility and efficacy of histologic malignancy grading relative to alternative proliferation index measurements for outcome prediction remain unanswered. Microsections of specimens from the Cooperative Breast Cancer Tissue Resource (CBCTR) were evaluated by seven pathologists for reproducibility of grade and classification. Nuclear figure classification was assessed using photographs. Grade was assigned by the Bloom-Richardson method, Nottingham modification. Proliferation index was evaluated prospectively by deoxyribose nucleic acid precursor uptake with thymidine (autoradiographic) or bromodeoxyuridine (immunohistochemical) labeling index using fresh tissue from 631 node-negative breast cancer patients accessioned at St Luke's Hospital. A modified Nottingham-Bloom-Richardson grade was derived from histopathologic data. Median post-treatment observation was 6.4 years. Agreement on classification of nuclear figures (N ¼ 43) was less than good by kappa statistic (j ¼ 0.38). Grade was moderately reproducible in four trials (N ¼ 10,10,19, 10) with CBCTR specimens (j ¼ 0.50-0.59). Of components of Bloom-Richardson grade, agreement was least for nuclear pleomorphism (j ¼ 0.37-0.50), best for tubularity (j ¼ 0.57-0.83), and intermediate for mitotic count (j ¼ 0.45-0.64). Bloom-Richardson grade was a univariate predictor of prognosis in node-negative St Luke's patients, and was improved when mitotic count was replaced by labeling index (low, mid, or high). When labeling index was added to a multivariate model containing tumor size and vessel invasion, grade was no longer a significant predictor of tumor-specific relapse-free or overall survival. Mitotic index predicted best when intervals were lowered to 0-2, 3-10, and 410 mitotic figures per ten 0.18 mm 2 highpower fields. We conclude that Nottingham-Bloom-Richardson grades remain only modestly reproducible. Prognostically useful components of grade are mitotic index and tubularity. The Nottingham-BloomRichardson system can be improved by lowering cutoffs for mitotic index and by counting 20-30 rather than 10 high-power fields. Measurement of proliferation index by immunohistochemically detectable markers will probably give superior prognostic results in comparison to grade.