Summary The variation in survival of women with clinically similar breast cancers may lead to difficulty in clinical management so it is important to recognise factors which indicate the prognosis. Immunoperoxidase staining patterns of primary breast tumours using monoclonal antibody have been shown to relate to overall survival (Ellis et al., 1985) but the results have not been reproducible in other centres. In this study paraffin sections of 483 primary breast cancers were stained with NCRC-11 and 3E1.2 using an immunoperoxidase system. The tumour staining patterns were compared with overall survival using life tables and tested for relative prognostic significance by Cox's multivariate analysis. NCRC-I (Fisher et al., 1983). Factors relating to histological differentiation and hormone receptor status have also been found to have prognostic significance (Elston, 1984;Bryan et al., 1986) but this in the former is observer dependent (Gilchrist & Kalish, 1985) and in the latter probably only applies in lymph node positive cases (Williams et al., 1987). There is therefore a need to discover other prognostic factors which will help in patient management and in understanding more of the biology of breast cancer. This paper describes the investigation of the prognostic value of monoclonal antibodies (MoAbs) NCRC-11 and 3E1.2 in 483 women with primary breast cancer. Both of these MoAbs are of the anti-Epithelial Membrane Antigen (EMA) type and their characterisations and clinical uses have been described elsewhere (Ellis et al., 1984(Ellis et al., , 1985Stacker et al., 1985Stacker et al., , 1988Stacker et al., , 1989. The synthesis of EMA by mammary acini and ducts is thought to be a specialised function and therefore tumours which contain EMA in abundance may be better differentiated and hence have a better prognosis than those in which it is less plentiful. In two studies using NCRC-1 and an immunoperoxidase method a relationship was demonstrated between well stained tumours and a favourable prognosis (Ellis et al., 1985. In a smaller study this relationship could not be confirmed (Angus et al., 1986) and no such study has been undertaken for 3E1.2. By using a large series of unselected patients followed-up for between 5 and 10 years, this study attempts to clarify the prognostic worth of both NCRC-11 and 3E1.2.