The focus these days is on genes and their expression. A lot of attention has been paid in the media to "breast cancer genes," the ones known as BRCA1 and BRCA2. However, breast cancer is rarely inherited in this way as a single-gene defect. It is essentially polygenic. In a useful summary of progress in breast-cancer genetics, subtitled "It’s not just about BRCA genes anymore," writer Ricki Lewis in The Scientist (17[3]: 24, 10 February 2003) included among the interviewees Prof. Rene Bernards from the Netherlands Cancer Institute, Amsterdam, a coauthor of a paper currently hovering outside the Science Watch Top Ten.

M.J. van de Vijver, with Bernards and other colleagues (New Engl. J. Med., 347[25]: 1999-2009, 19 December 2002; #14 this period, #16 last time; total cites to date 203) took 295 consecutive breast cancer patients with stage I or II disease and used a technique called microarray analysis to identify their prognostic profile across a package of no fewer than 70 genes. The profile separating poor and good prognosis had been previously determined on a different set of patients (L.J. van T’ Veer, Nature, 415[6871]: 530-6, 2002). One hundred eighty women had a poor-prognosis signature and the other 115 had a good one. Ten-year survival rates were 54.6% and 94.5%, respectively, and the probabilities of being free of metastases at that time were 50.6% and 85.2%. This gene profile was deemed to be a more powerful predictor of outcome in young patients (under 53 years) than systems based on clinical and histologic criteria.

In August of this year a group from Sweden’s Lund University published a paper suggesting that "good old" clinical markers might be as powerful prognostic indicators as microarray gene expression profiles (P. Eden, et al., Eur. J. Cancer, 40[12]: 1837-41, 2004). Coauthor Prof. Carsten Peterson assures Science Watch that this does not mean that his team is opposed to genomic analyses. Far from it; they do them all the time. However, what is needed is a "sober" evaluation before technologies such as the 70-gene microarray are adopted into routine clinical practice. Microarray profiling scans the entire genome, which is especially important in breast cancer, where "our understanding of the basic events leading to metastasis is insufficient," Prof. Peterson says. "Then the question is, of course, how many genes are needed to get correct predictions?"

Although the mathematical techniques used to tease out the potential of gene profiling can seem like mysterious "black-box" tools, genes that are especially important can be identified. In the Dutch work "it is the genes that are associated with the ER pathways that play a leading role," says Peterson, and work from Lund (S. Guvberger, et al., Cancer Res., 61(16): 5979-84, 2001) has shown that "the ER pathway is so ‘deep’ that ER status defines more or less different phenotypes."

This line of work is being vigorously pursued by several groups. Peterson notes the launch of a Breast International Group trial in which patients will be randomized and treated according to the risk evaluation as determined by the 70-gene classifier and the St. Gallen criteria (but "not our ‘good old’ clinical marker classifier or the Nottingham Prognostic Index, as far as we know," he adds).

So, as Science Watch asked Prof. Peterson, does the need for "sober" appraisal still stand? "It most certainly does! We have not yet seen a single case where gene classifiers outperform predictors based upon clinical markers. More publications are on the way," he replied. However, once microarray technology has matured, he believes that it will be a successful classifier, "presumably as a hybrid with the ‘good old’ variables."