The most appropriate systemic therapy for a population of patients with breast cancer is determined from clinical trial data. However, the heterogeneity of breast cancer is such that within a population individual patients derive variable benefit. There is therefore a need for predictive molecular factors in order that treatment can be individualised. This review describes the roles of HER-2, epidermal growth factor receptor (EGFR), oestrogen receptor (ER)/progesterone receptor (PgR), Ki67, Bcl-2, p53 and gene expression profiling in predicting responses to endocrine, cytotoxic and biological therapies. ER and PgR remain the only well-established predictive markers of responses to endocrine therapy, although HER-2/neu has an emerging role in this area and in choice of adjuvant chemotherapy. There are considerable methodological difficulties in identifying useful predictive factors but on the basis of current evidence other biomarkers add little additional information. The development of targeted therapies means that the molecular targets themselves may become useful predictive factors for directing use of these therapies. HER-2 already has an established role in this area, but the role of EGFR requires further elaboration. The use of DNA microarrays to assess gene expression profiles may revolutionise our ability to predict responses to therapy.