Breast cancer is a complex but increasingly well-understood disease. Clearly, multiple alterations from normal mammary cells are required to achieve a transformed phenotype. Furthermore, there may be several possible alterations within broad categories that will produce the transformations leading to the malignant state. The specific set of alterations within a given cancer may thus provide necessary information about how it is unique and how it may best be treated. Several of the newer biologic markers of breast cancer may provide very specific treatment information. erbB-2 may predict for improved response to doxorubicin, rather than CMF. hsp 27 may predict for failure of doxorubicin. pS2 or EGFR may provide supplemental information predicting response to hormonal therapy. Each of these variables has strong evidence to support its use in this manner, but that evidence has been obtained on limited numbers of patients treated in a limited number of ways. The most established markers, with multiple studies indicating their prognostic benefit, are erbB-2, cathepsin D, and proliferation markers. Of the several proliferation markers there may be no one choice that is best. However, very clearly, any marker must be carefully assessed for appropriate cut-off values, and cut-off values established by one cohort of patients should be verified against another cohort of patients. The oncoproteins associated with cell cycle regulation (cyclin D, p53, Rb, and c-myc) have shown strong promise of providing important prognostic information. The limited studies to date indicate that these markers are independent of one another. Cell cycle regulation may be an area in which any defect may serve to deregulate the cell, and therefore several defects in one cell would be unlikely. The specific nature of the defect in a given cancer may be very important. With the advent of immunohistochemical methods to measure most of the markers, more information may become available. Finally, the burgeoning area of tumor-stromal interactions is replete with potentially important markers of cancer prognosis. The growth factors, which are marginally a part of this area owing to the probable importance of paracrine effects on cancer cell growth, have progressively developed a body of literature supporting their prognostic potential. However, they have rarely been studied in conjunction with the other aspects of tumor-stromal cooperation. The markers of metastatic potential, nm23 and angiogenesis, have been shown in small cohorts to have considerable prognostic import.(ABSTRACT TRUNCATED AT 400 WORDS)