Categorizing Cancers: Gene activity predicts leukemia outcome

Two studies show that patterns of gene activity can be used to anticipate the prospects of patients who have a common form of leukemia. Doctors could someday use such gene patterns to make decisions about treatments, some researchers say.

In acute myeloid leukemia, or AML, bone marrow cells that make white blood cells go out of control. Chemotherapy initially eliminates the cancerous marrow cells in a majority of patients, but the disease often reappears with deadly consequences. Marrow transplanted from a healthy, matched donor can cure the cancer, but the procedure and its aftermath introduce serious risks.

Abnormally organized chromosomes and some specific DNA mutations have been linked to high or low risk of cancer recurrence after chemotherapy. Doctors use such clues in deciding whether transplantation is warranted. But many patients don’t show these signs and so fall into a murky category of intermediate risk.

“There’s no real consensus on how aggressively to treat these patients,” says Jonathan R. Pollack of Stanford University. Profiling the activity of many genes could be useful for choosing a therapy for that group, he says. Such profiling has produced insights to other blood cancers and certain tumors (SN: 9/14/02, p. 171: Targeted Therapies).

In one of the new studies, Peter Valk of Erasmus University Medical Center in Rotterdam, the Netherlands, and his colleagues analyzed blood or marrow samples from 285 people with AML. Using tools known as microarrays, the researchers determined which of about 13,000 genes were abnormally active or inactive in each patient. They then used a computer to group patients having similar gene-activity profiles.

That effort identified 16 groups with unique patterns of aberrant gene activity. Among patients in one group, only 18 percent survived at least 5 years, compared with 30 to 60 percent of patients in other groups. This group’s members had a range of chromosomal abnormalities, so under current clinical practices, they might have been given different treatments. Instead, they could all belong in a single treatment category, Valk and his colleagues say in the April 15 New England Journal of Medicine.

In a separate study reported in the same journal issue, Pollack and his collaborators used microarray analyses of 26,000 genes to select the gene activities that best reflected how 59 patients fared. From the activities of the 133 chosen genes, the team built a mathematical model of patient outcomes and used it to predict the fate of an additional 57 patients.

The model sorted patients into high-risk and low-risk categories that corresponded to those defined by current tests. Within the group of patients that traditional analysis had lumped into the intermediate-risk category, the gene-activity model distinguished some patients as either high or low risk. These classifications were largely borne out over 4 years of patient data. Similar data could guide doctors’ treatment decisions for future patients, according to Pollack’s team.

The two studies’ parallel findings indicate that gene-activity profiling can predict AML patients’ outcomes, says David Grimwade of Guy’s, King’s, and St. Thomas’ School of Medicine in London. However, with thousands of genes examined in each study but far fewer patients, dozens of expression patterns that currently appear predictive might result solely from chance, he cautions.

Once researchers home in on a set of genes that consistently predicts AML outcomes in multiple studies, microarrays will be a practical tool, Grimwade says. “Come 5-years’ time, I think we’ll see smaller versions of these arrays used to tune treatment decisions,” he forecasts.

More Stories from Science News on Health & Medicine

From the Nature Index

Paid Content