In a potential payoff from genomics research, cancer researchers are increasingly confident they can identify molecular "fingerprints" in tumors that will predict whether a given cancer is likely to spread quickly and lead to early death.

Such findings could eventually revolutionize the way many cancers are treated. Molecular tumor signatures may one day be useful in predicting which patients are most likely to respond to radiation or chemotherapy, allowing oncologists to personalize treatments to a much greater extent.

The work is "leading the way into a whole new approach to cancer," says Larry Norton, an oncologist at Memorial Sloan-Kettering Cancer Center in New York.

In a study to be published Thursday in the New England Journal of Medicine, a group of mostly Dutch researchers showed that a fingerprint derived from the activity of just 70 tumor genes can predict whether early-stage breast cancer is likely to metastasize, or spread aggressively.

These molecular fingerprints, however, are still some time away from general use in cancer treatment. Experts such as Todd Golub, a researcher at Dana-Farber Cancer Institute in Boston, warn that additional studies involving larger numbers of patients are necessary before the recent findings can be confirmed and integrated into standard treatment.

MORE COVERAGE   • See a list of cancer treatments1 in the development pipeline.   • 'Personalized' Chemo Treatment for Some Cancer Types Advances2 04/09/02   • For more health coverage, visit the Online Journal's new Health Industry Edition at wsj.com/Health3, and receive daily Health e-mails4.

"The issue is really being confident that you've got the right answer before you proceed," Dr. Golub says.

In part, such caution reflects the novelty, and uncertainty, associated with the genomic methods used in the latest studies. Known as gene-expression analysis, the technique involves the study of tens of thousands of genes in cancerous tissue from a variety of patients. Because most cancers are thought to result from genetic damage to cells, researchers figured a close look at variations in gene activity across tumors might help them identify aggressive cancers at an early stage.

Such analysis is possible thanks to a relatively new tool known as a microarray -- a sort of "gene chip" that can show which genes in a tumor cell are unusually active or unusually quiescent. Gene chips are layered with pieces of DNA from thousands of known human genes, most identified via the Human Genome Project. Washing the chips with serum extracted from a tumor-tissue sample leads to biochemical reactions that eventually cause the active tumor genes to fluoresce.

NEW TARGETS   Researchers have identified molecular 'fingerprints' in several types of cancers, among them:   Breast cancer Netherlands Cancer Institute Gastric cancer Stanford University B-cell lymphoma National Cancer Institute Adenocarcinomas* Dana-Farber Cancer Institute *Such as lung and ovary cancers   Source: WSJ reporting

Microarray studies, however, produce enormous quantities of data, much of it flecked with "noise," or stray data, from genes that cycle on and off for reasons having nothing to do with cancer. Some scientists initially despaired of extracting meaningful patterns from this sea of information.

The Dutch team in Thursday's report, led by Rene Bernards of the Netherlands Cancer Institute in Amsterdam, compared microarray data from tumor samples that had been stored from 295 women aged 52 and younger, then matched the results against these patients' medical records. Using the 70-gene fingerprint, previously plucked from tens of thousands of tumor-gene signals by sophisticated pattern-matching software, the scientists found that patients classified by the fingerprint technique as having a "good prognosis" had an 85% chance of living 10 years without metastasis, versus only a 51% chance for the "poor prognosis" group.

The results are still controversial because they suggest that a tumor's tendency to metastasize is inherent from an early stage. Cancer specialists have long believed that metastasis occurs later in the growth of a tumor, after it has accumulated mutations that help its cells survive in different parts of the body.

Yet a variety of scientists have produced findings that parallel those of Dr. Bernards's team. At Dana-Farber, Dr. Golub's group recently reported a 128-gene fingerprint that appears to predict whether a variety of tumors known as adenocarcinomas will metastasize. Last summer, researchers at the National Cancer Institute used microarrays to generate four signatures of 17 genes that could predict survival after chemotherapy in patients with large B-cell lymphoma, a type of blood cancer.

While those results are encouraging, researchers caution that it is far too soon to apply gene-chip analyses to conventional cancer treatment. For one thing, most of these studies are historical in nature; medical experts generally prefer forward-looking "prospective" studies before embracing new techniques. In addition, some critics worry that these studies, which have typically involved tumors from a few hundred cancer patients, may be too small to account for natural genetic variation across ethnically diverse populations.

Dr. Bernards is already planning a larger follow-up study to address such questions. In conjunction with Massachusetts General Hospital, his team will study the molecular signatures from as many as 5,000 women who have just been diagnosed with breast cancer in a study that will include both older and more ethnically varied patients.

Even once validated, however, it isn't immediately clear how these molecular signatures should be applied to treatment. Dr. Bernards argues that gene-chip tests could prevent "overtreatment" of many breast-cancer patients with less-aggressive tumors who are currently subjected to toxic chemotherapy, although other clinicians say they would be reluctant to take that risk without extremely strong evidence.

By itself, gene-based tumor fingerprinting also doesn't necessarily shed much light on the biological basis of cancer. Since the technique is almost purely statistical, it may highlight the role of many unrelated genes in addition to damaged genes that might be targets for less-toxic cancer drugs. Gene-chip enthusiasts such as Drs. Golub and Bernards acknowledge the difficulty, but counter that their studies are a useful first step toward identifying important cancer targets.

Ultimately, cancer specialists hope that the same techniques could help identify tumors that are most likely to respond to existing radiation treatment or chemotherapy. Such studies are under way, although no results have been published. Dr. Bernards also suggests that tumor fingerprinting could improve clinical trials of new drugs by helping companies test their experimental medications only in patients with aggressive forms of cancer.

Write to David P. Hamilton at david.hamilton@wsj.com⁵

Updated December 19, 2002