Wouldn't we all like the power to predict the future? For breast cancer patients, the accuracy with which the progress of their disease can be mapped can make the difference between whether they are overtreated or undertreated and can also affect predictions of their survival. Current criteria used for prognosis in breast cancer include age, the size of the tumour, axillary-node status, histological type, pathological grade and hormone-receptor status. Van de Vijver and colleagues now add a gene-expression profile to the list of useful prognostic indicators, as they report in the 19 December issue of New England Journal of Medicine.

The authors used a 70-gene prognosis profile that was first described by the same research team earlier in 2002 to classify 295 patients with primary breast cancer — which included 61 of the 78 patients, all lymph-node negative, from the first study — into poor-prognosis and good-prognosis groups. The prognosis signature assigned 115 tumours to the good-prognosis category and 180 to the poor-prognosis category. There was a strong correlation between the probability of remaining free of metastases and of surviving, and the good-prognosis signature.

Interestingly, the prognostic profile was independent of lymph-node status — the prognosis signature was also highly predictive of the risk of distant metasases in the lymph-node-positive subgroup. This could be useful for identifying patients with lymph-node-positive disease who have an unexpectedly good prognosis and indicates that lymph-node metastases develop independently of distant metastases.

The gene-expression profile was also predictive of overall survival — 94.5% of patients in the good-prognosis group survived for 10 years compared with 54.6% in the poor-prognosis group. The hazard ratio for distant metastases in the poor-prognosis signature group compared with the good-prognosis signature group was 5.1, and the hazard ratio for overall survival was 8.6. Van de Vijer et al. compare their classification with that made by standard consensus criteria, and suggest that the current criteria misclassifies many patients, which results in some patients getting unnecessary treatment and others missing out on adjuvant treatment that would be beneficial. This has important implications for the future of breast cancer management, especially if the prediction technique is validated in older patients with more advanced disease.