Most cancer types, and breast cancer is no exception, can be subcategorised by clinical stage and pathological subtype. These categories can be correlated with survival data, which allows the prediction of disease natural history and, to a lesser extent, treatment response and benefit for a given patient. This is fundamental to therapeutic decision-making in oncology and increasingly allows treatment to be tailored on an individual patient basis.
However, the standard methods for subtyping breast cancers remain relatively crude. Clinical staging and routine pathology are the principle indices used to identify those individuals at risk of developing metastases and who should therefore be considered for adjuvant chemotherapy. These features have been incorporated into various standardised scoring algorithms (Galea et al, 1992; Goldhirsch et al, 1998; Eifel et al, 2001). Their application, however, results in the ‘over-treatment’ of many patients in whom cure would have been achieved without chemotherapy or possibly even endocrine treatment. This is illustrated by the Oxford Overviews of systemic treatments that demonstrate a significant proportion of long-term survivors in the untreated arms (Early Breast Cancer Trialists' Collaborative group, 1998a).
Another group of patients who receive treatment from which they will not benefit are those who will develop metastatic disease despite adjuvant cytotoxic treatment. Although staining for the oestrogen and HER-2 receptors are powerful individual predictors of response (and benefit from) tamoxifen and herceptin, respectively, clinicians lack a marker that predicts those who will benefit from chemotherapy.
Unlike standard methodologies that rely on a few pathological features and immunohistochemical markers, molecular profiling allows tumours to be defined by the expression pattern or genomic alteration of thousands of genes simultaneously. With these techniques comes the prospect of defining individual genes or combinations of genes whose expression level(s) can discriminate efficiently between clinically significant subtypes of breast tumours requiring different treatment strategies.
Gene expression microarrays have been used extensively to study breast cancer. The technical aspects of these approaches have been reviewed extensively in the scientific literature (Schulze and Downward, 2001). Here, we will consider the role of expression microarray profiling in the definition of existing and novel categories of breast cancer. In particular, we will address how it may reduce the large number of breast cancer patients who receive inappropriate, yet toxic treatments.