DNA microarrays in breast cancer: the promise of personalised medicine

COMMENTARY See page 1590 Breast cancer is clinically heterogeneous, with varying natural history and response to treatment. Despite much effort to identify clinical measures of risk, methods to accurately predict an individual's clinical course are lacking. Whilst lymph-node status at diagnosis is the most important measure for future recurrence and overall survival, it is a surrogate that is imperfect at best. About a third of patients with no detectable lymph-node involvement, for example, will develop recurrent disease within 10 years. 1 The clinical heterogeneity of breast cancer is probably due to the genetic complexity of individual tumours, which have multiple somatic mutations and epigenetic changes that influence the expression of many genes that drive tumour growth, invasion, and metastasis. Different breast tumours, moreover, may arise from distinct cell-types. 2 This complexity has been difficult to study with traditional methods which are best suited to studying one gene at a time. The advent of DNA microarray technology, however, has recently enabled the quantitative measurement of complex multigene expression-patterns in human cancer. 3 Gene-expression profiling by DNA microarrays uses nucleic acid polymers, immobilised on a solid surface, as probes for gene sequences. DNA microarrays are relatively easy to use, yield gene-expression measurements for thousands of genes simultaneously, and can be used in large numbers of samples in parallel. 4 The results can be used to accurately diagnose and molecularly classify tumours, 5–7 assess their propensity to metastasise, 8 and predict response to combination chemotherapy. 9 Thus there is keen interest in defining the gene-expression profiles of all human tumours to create a new generation of clinically useful cancer diagnostics. There is also great hope that genetic information from these studies will lead to a deeper mechanistic understanding of the molecular pathways that cause cancer. Breast cancer has been particularly fertile ground for exploring the diagnostic usefulness of microarrays. Recent studies suggest that gene-expression patterns of primary tumours are better than available clinicopathological methods for determining the prognosis of individual patients. 6,10,11 In this issue of The Lancet, Erich Huang and colleagues extend these observations by using microarray-derived gene-expression profiles to classify individual breast tumours by their likelihood of having associated lymph-node metastases at diagnosis and by 3-year recurrence risk. These investigators first used unsupervised learning to cluster about 12 000 genes into groups based on similarity of gene expression across breast cancer samples. They then used singular-value decomposition to determine a " …