Pharmacogenomics and the Brave New World of Personalized Medicine

March 8, 2005
FDA’s recent guidance on pharmacogenomics reduces manufacturers’ risks in evaluating genomic data. Will it spur development of more targeted therapies?

Last December, the era of personalized medicine inched closer to reality when the U.S. Food and Drug Administration (FDA) approved Nitromed’s (Lexington, Mass.) Bidil, a combination therapy designed to treat heart failure in African Americans. This segment of the patient population doesn’t typically respond well to the angiotensin-converting enzyme (ACE) inhibitors that are usually prescribed to treat this condition.

Bidil’s approval has stimulated more discussion, and debate, about personalized medicine. Although some drug manufacturers may have mixed feelings about developing targeted drugs with small markets, those in favor of personalized medicine say that it will reduce the billions of dollars lost each year to adverse drug events — lost income and production, as well as higher disability and health care costs.

“Personalized medicine will not result in treating fewer patients,” says Felix Frueh, Ph.D., associate director for genomics at FDA’s Center for Drug Evaluation and Research (CDER; Bethesda, Md.). “If you take all costs into consideration — not just sales, but the cost of treating patients unnecessarily, or patients who suffer serious adverse events, the economics work out,” he says. He calls fears of fractionating existing drug markets “absurd,” chiding doomsayers for accepting a system that “treats” patients but provides no benefit.

“Adverse drug events are costly, Frueh points out. According to the Center for Education and Research on Therapeutics, adverse drug events cost the U.S. economy $136 billion per year. “Preventing just a small percentage of this loss will make a big improvement in healthcare economics.” (For more statistics, visit

Genomics may not have played a part in Bidil’s development, but it promises to drive the personalized therapies of the future. The idea: to treat disease based on the identification of biomarkers associated with the likelihood of a drug’s efficacy or potential toxicity.

Genotype vs. phenotype

The notion of treating groups of patients with the same overt disease is based on the now-recognized differences between genotype and phenotype. A century ago, “consumption” was considered a disease, but we now recognize it as comprising more than twenty distinct ailments (genotypes). Similarly today, it may be tempting to speak of “cancer” or “schizophrenia” as single diseases when they are merely “phenotypes” — overt physical manifestations — of what may be many different underlying diseases.

Pharmacogenomics, or the study of how a patient’s genetic makeup affects responses to drugs, is the sine qua non of personalized medicine. It is only in the last few years — thanks to whole-genome sequencing and rapid gene characterization — that practical pharmacogenomics has been possible. As recently as 1998, Hoechst (now Aventis) was forced to withdraw Seldane, an allergy drug with $600 million annual sales, due to serious cardiotoxicity in less than 0.5% of potential patients. These patients carried a genetic defect, which, in the presence of co-therapy with erythromycin, blocked Seldane metabolism.

Today, a simple genetic test might identify patients at risk for this serious drug interaction. Whether insurance companies would view such a test as cost-effective, a critical factor in the success of personalized treatments, is another story. It is likely that some reasonably significant fraction of the roughly 90% of drugs that fail in clinical testing might be approved, if only for small patient populations, if an appropriate genetic test were available to predict safety and efficacy.

That is exactly what happened when, around the same time as the Seldane incident, Genentech (South San Francisco, Calif.) noted that its breast cancer drug Herceptin only worked in one-fourth of patients — those whose tumors over-expressed the HER2 gene. Herceptin went on to become the poster child for personalized medicine and the model for rescuing failed compounds. GlaxoSmithKline (London) has followed Genentech’s lead by linking its Ziagen HIV treatment with a pharmacogenomic test that identifies the 5% or so of patients at risk for a severe hypersensitivity reaction to the drug. GSK also co-markets a test for an enzyme defect alongside its Purineetho pediatric leukemia drug to spot the one in 300 children at risk for drug-induced bone marrow toxicity.

End of the blockbuster?

Some observers believe that personalized medicine will end the blockbuster drug model as we know it. According to this reasoning, pharmacogenomic tests that uncover non-responders or those at risk for serious side effects can only cause potential patient populations to shrink. This “glass half-empty” argument ignores the dynamism of the pharmaceutical marketplace, where dozens — sometimes hundreds—of drugs compete in the major therapeutic categories. Leading products vie not only on glitz and marketing, but on the basis of clinical experience.

It’s commonly thought that drugs are effective in anywhere from 20% to 80% of those who take them. The typical hypertensive patient tries several medications before settling on one that works. In essence, this patient and his or her physician have performed a type of experiment, which, in most cases, reaches the same conclusion as a putative pharmacogenomic test.

Pharmacogenomic assays have the potential to eliminate much of the uncertainty and serendipity inherent in prescribing medications. There will, of course, be winners and losers. Pharmacogenomics will undoubtedly uncover a few patients who will not respond to any drug at all, a situation that is certain to engage medical ethicists. But the likelihood of important drug markets shrinking is almost nonexistent.

Getting the right medications into the right bodies is not only good medicine, it’s good business, says Brian Spear, Ph.D., director of genomic and proteomic technology at Abbott Laboratories (Abbott Park, Ill.). “The opportunities for developing drugs that work for non-responders are tremendous,” he believes. Spear debunks the notion that personalized medicine will fractionate drug markets by identifying non-responders to whom certain drugs should not be prescribed. Instead, he views the marketing opportunities as a more-than-half-full glass. He points out that, especially for competitive indications like cardiovascular and oncology, market shares are small but patient numbers are huge.

Still, personalized medicine will force drug developers to look at potential markets that are small by today’s standards. Or as Pierre Cassigneul, CEO of genomic diagnostics firm XDx (South San Francisco, Calif.), notes, “There is a threshold above which a development drug becomes interesting.” In cases where drug developers do encounter very small patient populations for new drugs as a result of pharmacogenomic studies, FDA offers the usual incentives of orphan drug status. This designation, which applies to diseases with fewer than 200,000 U.S. patients, provides extended exclusivity, tax breaks, exemption from user fees, availability of grants to cover the cost of clinical trials and other benefits. Genzyme (Cambridge, Mass.), among other companies, has been very successful at developing orphan drugs, Cassigneul notes.

FDA has no plans to improve the “package” for orphan drugs any time soon. Rather, it is considering ways to educate industry in light of the current interest in personalized medicine. In March 2005, the agency issued guidance on pharmacogenomics in which it encouraged submission of pharmacogenomic data along with drug approval paperwork on a relatively no-risk basis. In other words, FDA promised it would not “hang” drug developers on the basis of this information.

When taken to its logical conclusion, the culture of personalized medicine could, through pharmacogenomics, streamline drug discovery and clinical development, and greatly improve the success rate for new chemical entities (NCEs). “Identifying patients who will be the most responsive during clinical trials will improve product efficacy and reduce the time spent in clinical trials,” says Cassigneul.

One possible carrot at the end of the stick is the ability to justify charging premium prices for drugs targeted to narrow patient groups. On the other hand, insurers may not view personalized treatments as cost-effective, and be reluctant to pick up the higher price tab, notes Diane Romza-Kutz, managing partner at the law firm of Epstein, Becker and Green (Chicago). “There’s a disconnect in the reimbursement scheme for these products,” she says.

Sorting out the issues

As the scientific questions surrounding personalized medicine are worked out, manufacturers and regulators will need to resolve myriad legal, privacy, regulatory, intellectual property and reimbursement issues. Many of these concerns arise from the added pharmacogenomic component — the test. Regulation of products like Herceptin is currently managed through FDA’s Office of Combination Products, which delegates regulatory authority to the appropriate drug or device offices, so that aspect appears to be under control. But the added cost of the test — and synchronizing approvals of chemical and pharmacogenomic components that are approved for use together—could cause large pharmaceutical manufacturers to think twice before entering personalized medicine ventures, perhaps opening the door to a renaissance for boutique drug companies.

Although timelines for diagnostics tend to be much shorter than for drugs, researchers may not have settled on a suitable biomarker before late Phase III testing. By then it may be too late to introduce both products simultaneously. This spells trouble in situations where the medicine has been approved or conceived as a combination product — that is, alongside a diagnostic. This is an area Abbott thinks about a great deal since it has a standalone diagnostics business. Spear says there are no quick fixes to this conundrum other than planning and communication between groups.

Nevertheless, the pharmaceutical industry and FDA appear to be in synch where personalized medicine is concerned. “They listened to industry in formulating their guidance,” says Spear, who testified before FDA’s Science Commission on the topic.

FDA’s March 2005 pharmacogenomics guidance (“Pharmacogenomic Data Submissions” — click the Download Now button below to download the 28-page .pdf document) clarified the agency’s position on the wave of genomics information that drug companies now routinely generate during drug discovery and development. Industry feared that by submitting those data, FDA might arrive at spurious conclusions that it would then use to scrutinize drug applications unfairly.

In its guidance, FDA assured pharma that such a scenario would not occur, that it encourages pharmacogenomics investigations and recognizes the exploratory nature of such work. Most important: While the Agency was eager to see such data, it would not base approval decisions on it — at least for now. Eventually, FDA plans to establish a separate group to review pharmacogenomic information.

Spear calls the guidance a win-win situation. “Since they’ve lowered the risk for carrying out this essential research, Abbott is encouraged to generate pharmacogenomic data,” he says. “What transpires from that will be educational for both sides.”

About the Author

Angelo De Palma | Ph.D.