Seven years ago, the approval of the breast cancer drug Herceptin created a stir both in the medical community and the popular press. It helped only about one in five women who took it. But those women turned out to have a mutation in their tumor cells that clearly differentiated them from non-responders. When California-based biotech giant Genentech began selling the drug along with a diagnostic test that could determine which patients would benefit from it, Herceptin became more than just another cancer drug. It quickly became the poster child for a new age in medicine, in which treatment could be tailored to an individual patients genetic profile.

Since then, however, few such drugs have become available. "Herceptin, I believe, was just a lucky punch," says Heino von Prondzynski, CEO of Swiss company Roche Diagnostics, which is developing gene-based diagnostics. "Advances in life sciences are providing very exciting discoveries, but people underestimated the time until it will bring tangible results." Although the tools and technology to make such drugs and diagnostics exist, experts say, the framework for how companies should develop them and how physicians should use them is still in the making.

Scientifically, the biggest challenge for applying genetics to medicine (a field called pharmacogenomics) lies in identifying biomarkers--genes, or the proteins they express--that predict how well a particular drug will work. "That's not easy," states Donna Mendrick, scientific fellow of toxicogenomics at Gene Logic in Maryland. Analyzing the profiles of thousands of genes for thousands of patients produces so much data, that researchers simply do not know what to do with the results. There is no agreed upon procedure for turning a lone companys hunch about a protein into a clinically validated and accepted marker, Mendrick explains. "There are lots of biomarkers in use now that have never been through stringent trials," she notes. "We in industry are setting up a process that was never used before."

The lack of precedent has also reflected a regulatory void: until recently the FDA did not have an established system for approving pharmacogenomic products. Although many companies had used biomarkers for in-house research for several years, they feared that because they had not been validated, the FDA would refuse to approve products based on that research. "There's been a concern that if they conduct those studies, the FDA will use this against them," says Carol Reed, vice president of medical affairs at Genaissance, a Connecticut company that is developing gene-based medicines and diagnostics.

In March, however, the FDA produced a long-awaited guidance document clearly specifying what kinds of genomic data it will require for drug approvals, and encouraging companies to submit preliminary data voluntarily in order to help build a scientific basis for interpreting pharmacogenomic experiments. The agency is now working with companies to develop another set of guidelines to define the process for validating biomarkers.

But regulatory issues aside, the pharmaceutical industry has also had a strong profit motive in avoiding personalizing medicine. Part of their reluctance has come from the desire to not limit their market--a test identifying who will (and who won't) respond to a drug means that fewer patients will buy it. "Companies have been worried about what we perceive as fear of the loss of the blockbuster model of drug development," Reed notes.

Although the path or developing new drugs is becoming clear, personalized medicine has an equally important role to play in drugs already on the market. While companies have focused on discovering new pharmacogenomic medicines, academic researchers have begun to apply genomics to the well-understood cytochrome P450 (CYP450) group of genes, which encode a family of liver enzymes. These enzymes metabolize as many as a third of the medicines that are already on the market, from the blood thinner warfarin to codeine to many antidepressants. Mutations in these genes can make patients metabolize these drugs either more slowly or more quickly than normal, which for them means that a drug's standard dose will be either too small to be effective or too large, causing side effects or even over-doses. They also regulate how different drugs will interact with each other--a crucial factor for the many patients who take multiple medications. In the last few years, many academic medical centers have developed home-brewed tests for these mutations for research, and are now using them in patients. In January, Roche Diagnostics received the FDA's approval to market its version of a test for three CYP genes commercially to physicians.

Yet despite their availability, the tests have been slow to catch on in the clinic, comments Howard McLeod, a professor of medicine at Washington University. They, too, require a lot more validation, even though the basic biology is known. Although many studies have shown the relationship between CYP450 genes and specific drugs, very few have actually correlated the relationship with dosing guidelines that physicians can use, McLeod explains. What's more, because these drugs are long-time staples of clinical practice, physicians are used to the trial-and-error approach for prescribing them. "We really have to prove that making decisions with the test is better than making decisions without it," he says.

But conducting the prospective studies to establish that dosing is gene-dependent "takes a lot of effort, a lot of time and a lot of money," McLeod reflects, and big pharma has not been particularly motivated to fund them. "There's no sugar-daddy for 40-year-old drugs." Although the FDA can demand such studies for new drugs and diagnostics, it is powerless to ask companies to conduct more safety studies when a new technology becomes available.

Still, observes Roland Valdes, professor of clinical chemistry at the University of Louisville, theres enough data to start educating physicians on how to incorporate pharmacogenomic testing in their practice. "In all cases, you start by using the tests, and then match that up with the studies. If you wait for the studies it will never happen." In some clinical areas testing is gradually gaining acceptance. Psychiatrists, for example, are particularly interested, not just because up to 80 percent of psychiatric medicines are metabolized by CYP450 genes, "but also because there are some guidelines emerging for what you do about it" as researchers conduct more trials, McLeod remarks.

Reframing medicine to include pharmacogenomics will take some time, and experts note that the process will not always be pretty. "I hate to say it, but one of the biggest friends of pharmacogenomics [will be] litigation attorneys," McLeod warns. "Realizing that there are tests to predict toxicity--that's low-hanging fruit." But the benefits--not only for patient safety, but also for drug company profits--are only getting clearer. "I think companies are going to be encouraged to do this," Gene Logic's Mendrick predicts, recalling the flurry of problems with drug safety over the past year. "Think of Vioxx-- they could have saved that compound if they could tell which people would really be helped."