Clinical features, 8 histological features, 7 nuclear morphometric variables and 2 mitotic indices were entered in a Cox's model to assess their independent predictive power in a series of 517 breast cancer patients followed up for over 10 years. The volume-corrected mitotic index (M/V index) (p less than 0.001), axillary lymph-node status (p = 0.002), the shortest nuclear axis (p = 0.006) and the degree of tubule formation (p = 0.02) predicted independently the recurrence-free survival. In N- tumours (n = 293), the M/V index (p = 0.005), the degree of tubule formation (p = 0.016) and tumour size (p = 0.023) were independent prognostic predictors, whereas in N+ tumours (n = 224), only the M/V index (p = 0.004) and the maximum nuclear axis (p = 0.004) had independent prognostic value. The corrected survival was predicted independently by the axillary lymph-node status, degree of tubule formation, M/V index, tumour size (p less than 0.001), age (p = 0.002) and year of treatment (p = 0.008). In N- tumours, the degree of tubule formation (p = 0.005) and intraductal growth pattern (p = 0.015) exhibited independent predictive value. In N+ tumours, patient survival was related to the M/V index (p less than 0.001), tumour size (p = 0.005) and patient age (p = 0.005). The results show that the assessment of the M/V index, axillary lymph-node status, tumour size, intraductal growth pattern and tubule formation are reliable factors in predicting the prognosis of breast cancer. The conventional mitotic activity index (MAI) and histological grading should be replaced by the M/V index in histological assessment of malignancy in breast cancer. The prognostic scores combining the independent variables reflecting the proliferative rate and metastatic potential of the tumours are more accurate predictors of the recurrence-free survival and overall survival (p less than 0.0001) than the single variables used alone.