Medullary breast cancer (MBC) is a rare, diagnostically difficult, pathological subtype. Despite being high grade, it has a good prognosis. MBC patients have an excess of BRCA1 germ-line mutation and reliable identification of MBC could help to identify patients at risk of carrying germline BRCA1 mutations or in whom chemotherapy could be avoided. The aim of this study was therefore to improve diagnosis by establishing an MBC protein expression profile using immunohistochemistry (IHC) on tissue-microarrays (TMA). Using a series of 779 breast carcinomas ('EC' set), diagnosed initially as MBC, a double-reading session was carried out by several pathologists on all of the histological material to establish the diagnosis as firmly as possible using a 'medullary score'. Only MBCs with high scores, i.e. typical MBC (TMBC) (n=44) and non-TMBC grade III with no or low scores (n=160), were included in the IHC study. To validate the results obtained on this first set, a control series of TMBC (n=17) and non-MBC grade III cases (n=140) ('IPC' set) was studied. The expression of 18 proteins was studied in the 61 TMBCs and 300 grade III cases from the two sets. The global intra-observer concordance of the first reading for the diagnosis of TMBC was 94%, with almost perfect kappa (kappa) of 0.815. TMBC was characterized by a high degree of basal/myoepithelial differentiation. In multivariate analysis with logistic regression, TMBC was defined by the association of P-cadherin (R=2.29), MIB1 > 50 (R=3.80), ERBB2 negativity (R=2.24) and p53 positivity (RR=1.45)