Reliable prognostic factors are needed to improve the stratification of patients with lymph node-negative breast cancer to different therapy modalities. We investigated the prognostic value of quantitative histopathology in a retrospective study of 98 "low-risk" breast cancer patients (T1+2N0M0) with a median follow-up of 9 years. An interactive video system and stereological and morphometric techniques were used to obtain estimates of four nuclear features (mean volume, mean profile area, volume fraction, and profile density), and two mitotic counts [mitotic profile frequency (MF) and mitotic profile density (MD)]. All measurements were performed in fields of vision sampled systematically from the whole tumour area of a routine histological section. Histological grade, histological type, and oestrogen receptor (ER) status was reassessed, whereas tumour diameter and age at diagnosis were recorded from the files. We found that all quantitative histopathological variables and ER status were correlated with histological grade. Single-factor prognostic analyses showed a highly significant value of MF (2p = 0.001) and a marginally significant value of MD (2p = 0.09), whereas no other variable approached statistical significance (2p > or = 0.25). In a multivariate Cox analysis, MF was the only parameter of significant independent prognostic value (2p = 0.03). Thus, the prognostic value of nuclear features found in previous studies could not be reproduced, whereas the marked value of mitotic counts for prediction of the outcome in patients with breast cancer was confirmed. Mitotic counts are easily obtained and may be of clinical value for identification of high-risk cases among patients with lymph node-negative breast cancer.