Separately, other researchers found that several new tumor profiling tests for breast cancer, including two already in general use, are similar in accuracy and should allow many women to avoid unnecessary chemo.
The lung cancer one needs far more testing, but is "breakthrough research" building on years of work to develop personalized cancer treatments, said Dr. Len Lichtenfeld, deputy chief medical officer of the American Cancer Society.
Its accuracy so far — about 80 percent — is "better than what we have, but it's not as good as we would like," said Lichtenfeld, who had no role in the research.
Both studies were reported in Thursday's New England Journal of Medicine.
Treatment guidelines for cancer have been relatively crude — based on a tumor's size, whether it has spread, and other characteristics.
As a result, many women with early breast cancer get chemo even though the vast majority would do fine without it. It is the opposite with early-stage lung cancer: Even though about a third of them will get worse and die, few get chemo because doctors can't tell which ones will benefit, and the treatment itself carries risks. Chemo can damage the liver, heart and other organs, and in some cases can kill.
Doctors hope that growing knowledge about the genes fueling these cancers will lead to better ways to tell who really needs chemo. The newly emerging tumor profile tests are tools to let them do that.
To develop the lung cancer one, Duke University researchers examined 198 tumor samples and analyzed 133 genes whose activity correlated with how aggressive the cancer was.
"It's a fingerprint unique to the individual patient (that) predicts survival chances," said lead researcher Dr. Anil Potti.
Patients were scored as having a low or a high risk of recurrence based on the test, and results were compared to what actually happened to them.
The test was 93 percent accurate on the half of patients whose tumor samples came from Duke and 75 percent accurate on the rest. Current best tests to estimate risk based on tissue characteristics are about 60 percent accurate, Potti said.
Several of the researchers are part owners of, or have other financial ties to, a genetic testing lab established by Duke.
A larger study of 1,200 lung cancer patients will start in January to further evaluate the Duke test. After surgery to remove the initial tumor, patients will get chemo or not depending on their test score, and then will be followed for a few years to see how they do.
That study will reveal whether the test really works, said Dr. David Johnson, a Vanderbilt University lung cancer specialist and a former president of the American Society of Clinical Oncology.
"It hardly is a slam dunk," Johnson said. "It is not ready for prime time" yet but could turn out to be "a huge breakthrough for patients."
In the other study, researchers at the University of North Carolina at Chapel Hill compared how five gene profiling tests performed at predicting outcomes of 295 breast cancer patients. Although the tests used different sets of genes, four were remarkably similar in accuracy and better than tests used now that are based on tumor characteristics.
"They agreed 80 percent of the time," indicating they all are ultimately tracking the same biological processes affecting tumor growth, said lead researcher Charles Perou.
Two of the four have been widely available for about two years, and two large international studies have been launched to establish their ultimate accuracy and usefulness.
The tests can be automated and standardized for wide use, eliminating the variability that exists when pathologists have to evaluate the appearance of tumor cells under a microscope, Dr. Joyce O'Shaughnessy of the Baylor Sammons Cancer Center noted in an editorial.
Such tests in the future could allow 30 percent to 50 percent of women to skip chemo, she wrote.