Dec. 29, 2008 By Andrew Pollack for The New York Times

For more than two years, Jody Uslan had been taking the drug tamoxifen in hopes of preventing a recurrence of breast cancer. Then a new test suggested that because of her genetic makeup, the drug was not doing her any good.

“I was devastated,” said Ms. Uslan, 52, who stopped taking tamoxifen and is now evaluating alternative treatments. “You find out you’ve been taking this medication for all of this time, and find out you are not getting benefit.”

Ms. Uslan’s situation is all too common — and not just among the hundreds of thousands of women in this country taking tamoxifen.

Experts say that most drugs, whatever the disease, work for only about half the people who take them. Not only is much of the nation’s approximately $300 billion annual drug spending wasted, but countless patients are being exposed unnecessarily to side effects.

No wonder so much hope is riding on the promise of “personalized medicine,” in which genetic screening and other tests give doctors more evidence for tailoring treatments to patients, potentially improving care and saving money.

Many policy experts are calling for more studies to compare the effectiveness of different treatments. One drawback is that such studies tend to be “one size fits all,” with the winning treatment recommended for everybody. Personalized medicine would go beyond that by determining which drug is best for which patient, rather than continuing to treat everyone the same in hopes of benefiting the fortunate few.

The colon cancer drugs Erbitux and Vectibix, for instance, do not work for the 40 percent of patients whose tumors have a particular genetic mutation. The Food and Drug Administration held a meeting this month to discuss whether patients should be tested to narrow use of the drugs, which cost $8,000 to $10,000 a month.

And a genetic test might help doctors determine the optimal dose of warfarin, a blood thinner used by millions of Americans. Tens of thousands of them are hospitalized each year because of internal bleeding from an overdose or a blood clot from an inadequate dose.

“If you save one hospitalization for every 100 new warfarin users, you more than offset the cost of testing all 100,” said Dr. Robert S. Epstein, the chief medical officer of Medco Health Solutions, which manages prescription plans for employers. The test typically costs $100 to $600.

Dr. David Flockhart discovered how the body converted the medication tamoxifen.
Dr. David Flockhart discovered how the body converted the medication tamoxifen.Credit…A.J. Mast for The New York Times

For all the potential, experts see some formidable obstacles on the path to the promised land of personalized medicine.

“It’s going to take 20 to 30 years for all this to fall into place,” said Dr. Gregory Downing, who heads efforts by the Department of Health and Human Services to spur personalized health care.

The hurdles include drug makers, which can be reluctant to develop or encourage tests that may limit the use of their drugs. Insurers may not pay for tests, which can cost up to a few thousand dollars. For makers of the tests, which hope their business becomes one of health care’s next big growth industries, a major obstacle is proving that their products are accurate and useful. While drugs must prove themselves in clinical trials before they can be sold, there is no generally recognized process for evaluating genetic tests, many of which can be marketed by laboratories without F.D.A. approval.

Genentech, a developer of cancer drugs, petitioned the F.D.A. this month to regulate such tests. It warned of “safety risks for patients, as more treatment decisions are based in whole or in part on the claims made by such test makers.”

A cautionary case is Herceptin, a Genentech breast cancer drug that is considered the archetype of personalized medicine because it works only for women whose tumors have a particular genetic characteristic. But now, 10 years after Herceptin reached the market, scientists are finding that the various tests — some approved by the F.D.A., some not — can be inaccurate.

Moreover, doctors do not always conduct the tests or follow the results. The big insurer UnitedHealthcare found in 2005 that 8 percent of the women getting the drug had tested negative for the required genetic characteristic. An additional 4 percent had not been tested at all, or their test results could not be found.

Tamoxifen, the drug Ms. Uslan took, illustrates the promise and current limitations of genetic testing. In 2003, more than 25 years after tamoxifen was introduced, researchers led by Dr. David A. Flockhart at Indiana University School of Medicine figured out that the body coverts tamoxifen into another substance called endoxifen. It is endoxifen that actually exerts the cancer-fighting effect. The conversion is done by an enzyme in the body called CYP2D6, or 2D6 for short.

But variations in people’s 2D6 genes mean the enzymes have different levels of activity. Up to 7 percent of people, depending on their ethnic group, have an inactive enzyme, Dr. Flockhart said, while another 20 to 40 percent have an only modestly active enzyme.

The implications were “scary,” Dr. Flockhart said. Many women were apparently not being protected against cancer’s return because they could not convert tamoxifen to endoxifen.

The economic implications could be just as scary to big pharmaceutical companies.

Tamoxifen, now a generic drug, costs as little as $500 for the typical five-year treatment. But most patients in the United States are currently treated with a newer, much more expensive class of drugs, called aromatase inhibitors, that cost about $18,000 over five years. Those drugs — made by AstraZeneca, Novartis and Pfizer — performed better than tamoxifen in clinical trials before the role of 2D6 was generally understood.

If only women with active 2D6 had been assessed, tamoxifen might have worked as well or better than the newer drugs, according to researchers at the Dana-Farber Cancer Institute in Boston.

After Jody Uslan had taken tamoxifen for more than two years, a genetic test suggested that it was not doing her any good.
After Jody Uslan had taken tamoxifen for more than two years, a genetic test suggested that it was not doing her any good.Credit…Marissa Roth for The New York Times

But proving these suppositions and having them incorporated into medical practice have not been easy.

The F.D.A., in its meeting this month, said clinical trials were the ideal way to validate a test. But many test developers argue that trials would be too costly and time-consuming, so many tests are validated by reanalyzing patient data from old trials.

In the case of tamoxifen, Dr. Matthew P. Goetz of the Mayo Clinic and colleagues went back to an old trial and used stored tumor samples to test the 2D6 genes of each patient. The researchers reported in 2005 that 32 percent of the women with inactive 2D6 enzyme had relapsed or died within two years, in contrast to only 2 percent of the other women.

But while some subsequent studies have backed those conclusions, two had contradictory results. That leaves many experts hesitant to use the test, which costs about $300.

There are other complications. Dozens of variants of the 2D6 gene exist, and laboratories can differ in their interpretation of test results. And it is not always clear how to act upon the information the test provides.

Ms. Uslan, who lives in the Woodland Hills neighborhood of Los Angeles, is in a predicament since she stopped taking tamoxifen. The newer alternative, aromatase inhibitors, work only for postmenopausal women and she has not yet completed menopause. To take an aromatase inhibitor, she must have her ovaries removed or take a drug to induce menopause. Because both options are unattractive, many experts say there is no point testing premenopausal women for 2D6.

Such complexities are not confined to tamoxifen testing. The labels of about 200 drugs now contain some information relating genes to drug response, said Lawrence J. Lesko, the F.D.A.’s head of clinical pharmacology. But in many cases, he said, doctors are not told specifically enough what to do with the test results, such as how much to change the dose.

Despite all the obstacles, personalized medicine is coming. Even the drug companies, which have been worried that testing would reduce their sales, are starting to realize that their medicines might not be approved or paid for without better evidence that they work.

Last year, for instance, European regulators said Amgen’s colon cancer drug Vectibix did not provide enough benefit to patients to be approved.

So Amgen reanalyzed the data from its clinical trial. After the results showed Vectibix worked better in patients whose tumors did not have a mutation in a gene called KRAS, the drug was approved for those patients only.

As for tamoxifen, an F.D.A. advisory panel recommended two years ago that the 2D6 test be mentioned in the drug’s label. But the agency itself was not persuaded there was enough evidence until just recently, Dr. Lesko said. “There’s no ‘one size fits all’ for evidence that everybody buys into.”

This article was originally published in The New York Times