Ki-67 labeling was quantified in 37 nonmalignant breast tissues and in 63 breast carcinomas by counting ten random high-power fields each in three section planes (RC) or by evaluation of the area with the highest labeling density (HDC). Both procedures proved to be highly correlated (rs = 0.94). Ki-67-positive fractions of the nonmalignant tissues (mean, 2.1% for RC and 4.1% for HDC) were significantly lower as compared with the carcinomas (mean, 14.5% for RC and 17.5% for HDC). In carcinomas the Ki-67 labeling was significantly associated with pT stage, axillary lymph node status, and tumor grading and inversely related to progesterone receptor status. Using the medians of both counting methods (12% for RC and 17% for HDC) as cutoff points, significantly different curves for overall and disease-free survival (median follow-up, 37 months) were obtained. However, Cox multivariate analysis failed to demonstrate an independent effect of Ki-67 labeling. In contrast, Ki-67 reactivity seems to be of independent prognostic value if a higher cutoff level was selected.