The test could help women make better choices about how to treat their disease, said federal officials who helped fund the study and conduct it. Half could safely skip chemo, they said.
"It outperformed all known prognostic factors for predicting response," said Dr. Soonmyung Paik, a University of Pittsburgh physician who presented the research Friday at a breast cancer conference in Texas. Part of it also was published online by the New England Journal of Medicine.
Most breast cancer cases that haven't spread to lymph nodes will not recur after surgery, but doctors have no good ways to predict which ones will. They guess, based on the size of a tumor and the woman's age, but they know that they are giving many women chemotherapy who don't need it.
The new test, made by California-based Genomic Health, measures the activity of 21 genes to classify women as high, low or medium risk for recurrence.
Previous studies have established the test's ability to predict this. With this latest study, doctors extended those findings to show that chemotherapy helps high-risk women a lot and low-risk ones not at all.
The researchers looked at 651 women with early-stage breast cancer that had not spread to lymph nodes and whose growth was affected by the hormone estrogen -- a situation that applies to nearly half of all cases diagnosed in the United States. Currently, chemotherapy is recommended for nearly all such women. Some of the women studied got chemotherapy plus the drug tamoxifen; the rest got just tamoxifen.
Among the women classified as high-risk on the genetic test, 88 percent of those who got chemotherapy were alive 10 years later without their cancer spreading to other parts of the body, versus 60 percent of those who got tamoxifen alone. Low-risk women saw hardly any difference -- about 95 percent had no spread.
Chemotherapy is not as grueling as it used to be decades ago but still often causes hair loss and nausea and should be avoided if it won't do a woman any good, said Dr. Larry Norton, breast cancer chief at Memorial Sloan-Kettering Cancer Center, who had no role in the study or ties to the company making the test.
"I'm going to use it for patients on the borderline that should get chemotherapy, to put people into the chemotherapy category rather than keep them out," he said.
Sparing women unnecessary chemotherapy also will help many avoid becoming infertile from the treatments.
Christina Koenig, who works for the advocacy group Y-ME, said she wished such a test were available when she had breast cancer several years ago.
"I was 39 when I was diagnosed. I told my doctor at the time I didn't want any more kids, give me what you've got," the 42-year-old woman said.
The test is expensive -- $3,460 (euro2,623) -- but so is chemotherapy, which costs $37,000 (euro28,050) or more depending on the drugs and length of time it is given.
In an editorial in the New England Journal, Drs. Robert Bast and Gabriel Hortobagyi of the University of Texas' M.D. Anderson Cancer Center call the test "an exciting work in progress" that will need to evolve as new breast cancer drugs such as aromatase inhibitors come into wider use.
Genomic Health, which helped pay for the research, makes the only such test licensed in the United States. A test by Dutch researchers is expected to be available in January, though doctors say it is technically more difficult to use and has not proved itself for predicting chemotherapy response yet.