Given accelerated approval in 1996, the chemotherapy drug irinotecan (Camptosar) can attack metastatic colorectal cancers that don't respond to other drugs. However, by 2004 a series of studies had suggested that the drug was particularly toxic in a subset of patients homozygous for a polymorphism of
In November 2004, a Food and Drug Administration (FDA) subcommittee met with Pfizer to discuss the results, and the company is now in discussions with the regulatory agency and may update the drug's labeling.
In the future, such so-called pharmacogenetic data may be known long before a drug ever comes to market. The National Institutes of Health, through the Pharmacogenetics Research Network (PGRN), is promoting the discovery of new genes that affect drug metabolism, as well as stronger correlations between known polymorphisms and drug effects. And the FDA released new guidelines in March to make clear how and when drug companies must submit genomic data, and how the agency will evaluate it.
Many companies are exploring pharmacogenetics and pharmacogenomics as tools in guiding drug development, to eliminate ineffective or potentially toxic drugs early on, for instance. But uncertainty in the regulatory process had caused concern that the information could come back to haunt a company, possibly by hurting the chances of getting a drug approved, or by limiting its market.
Now FDA has clarified its position. Nadine Cohen, head of pharmacogenomics at Johnson & Johnson Pharmaceutical Research and Development (J&JPRD), says the FDA guidance "has been a very, very useful document, the first one we have from a regulatory authority on how to submit pharmacogenomics data, and in which format."
While it could be a decade or more before genetic testing is the rule rather than the exception in the clinic, a new generation of genetic tests should help. The first, approved in December 2004, is used to genotype patients at two cytochrome P450 genes involved in drug metabolism. Similar products are in the pipeline, and more will surely come to market.
But long-term studies will be needed to convince doctors that such genetic tests should become part of the standard of care. "It takes a while to do the prospective studies," says Howard McLeod, a pharmacogeneticist at Washington University in St. Louis. "That's where we are right now, [we're] really getting held back by the lack of clear evidence that the genetics is worth doing."
'SAFE HARBOR' FOR PHARMA
Although the terms are sometimes used interchangeably, pharmacogenetics is the study of how inherited DNA variations, typically in just a few genes, affect drug metabolism or toxicity. Pharmacogenomics is a broader term that encompasses all the technologies that can be used in high-throughput screening. Either way, the information is now too valuable to ignore.
Current estimates peg the cost of bringing a new drug to market at around $880 million over 15 years. "The goal of integrating pharmacogenomics into the research and development process," says Cohen, "is to make this process more efficient, and to come up with better drugs, to help in decision-making, and also to avoid withdrawal of the drug postmarketing due to an emerging side effect that we could not predict."
Pharmacogenetics could save drug companies as much as $335 million per drug, according to a 2001 report by the Boston Consulting Group, though the report projects the average savings will be closer to $80 million.1 J&JPRD tests every new drug to determine which metabolic enzymes process it, Cohen says. The data are then included in the drug's package insert as appropriate. Similarly, at Millennium Pharmaceuticals in Cambridge, Mass., every drug in the development pipeline has its own associated personalized medicine program, says Chris Webster, the company's director of regulatory strategy and intelligence.
Webster chairs the genomics group at Pharmaceutical Research and Manufacturers of America (PhRMA) and coordinated the trade group's comments to FDA on the new guidelines. The guidance document, he says, spells out procedures by which companies can voluntarily submit "nonvalidated" or exploratory genomics data to a newly formed interdisciplinary pharmacogenomic review group via a "safe harbor" process in which the data would not be used for regulatory decision-making. "That's a new process for the FDA," says Webster. "Previously, every scrap of data that a sponsor has submitted has been potentially available to regulators for them to use to make regulatory decisions."
IRESSA: A Pharmaceutical Case Study
When AstraZeneca announced last December 17 that Iressa, a lung cancer therapeutic approved by the FDA in 2003, failed to significantly improve survival in a placebo-controlled Phase III trial (Trial 709), the company's stock fell $3.50 per share (8.7%) overnight. Total prescriptions have dropped 49% and some consumer groups have called for the drug's withdrawal from the market.
Yet hope remains for Iressa. The drug targets the epidermal growth factor receptor (EGFR), which is overexpressed in non-small-cell lung cancers, but it works only in 10% to 20% of patients with such cancers. In 2004, two groups at Harvard identified polymorphisms in EGFR that strongly correlated with Iressa efficacy.12 Several small, prospective, investigator-led Phase II studies designed to test this directly have since been launched, says company spokesperson Mary Lynn Carver.
As Trial 709 was initiated before the biomarker's discovery, it "was not enriched for any genomics information," Carver says. Yet analysis of the trial data, reported at the March 4 meeting of the FDA's Oncologic Drugs Advisory Committee, shows that certain subgroups did receive statistically significant survival benefits. Asians, which represented about 20% of the trial's 1,692 patients, "did very well," says Carver, as did nonsmokers. Moreover, women and patients with adenocarcinoma "just missed" statistical significance.
Analysis of the literature suggests these four population groups are all enriched for the EGFR polymorphism, says Carver. Analyses of biopsy samples from the trial are now underway to measure EGFR expression and mutation status directly. Results are expected by June.
Validated biomarkers, which are known to be associated with toxicity or could be used to segregate patient treatment groups, must still be submitted to the agency for consideration in the drug-approval process. A biomarker is considered validated when clinical trials clearly show an association between a genetic variant and a drug's efficacy or toxicity in a group of patients.
But nonvalidated biomarkers, such as a gene that is associated with efficacy in only small studies or a microarray gene-expression pattern that correlates with clinical outcome, need not be submitted during the investigational new drug process, explains Larry Lesko, director of the FDA office of clinical pharmacology and biopharmaceutics. These markers "have potential to do good things like predict toxicities, but they're premature. The evidence basis [for a regulatory decision] is not there," he says.
Between 12 and 15 voluntary submissions have been made, Lesko says, with another four to five scheduled. He expects the agency to receive an average of two submissions per month.
Both the agency and companies stand to benefit from the process. Drug developers, says Webster, can get advice on how to hone their development efforts, as well as a sense of issues that may arise later in the drug-approval process. The agency, on the other hand, gets an early peek at the technologies drug developers are using.
More importantly, FDA can get a bird's eye view of the emerging drug-development landscape. "If multiple sponsors are going in and submitting voluntary data around, say, a particular class of drugs, by the very fact that FDA is able to accumulate data horizontally on multiple drug candidates, they may then see patterns emerging which no single company would be able to see by themselves," says Webster.
Europe has been holding similar discussions with industry since 2002, according to Marisa Papaluca-Amati, deputy head of sector for safety and efficacy of medicines, European Medicines Agency (EMEA). The EMEA's newly formed Pharmacogenetics Working Party issued draft guidelines for its briefing meetings in March.
The FDA is now working on what could be another useful pharmacogenetics document. At a meeting in Bethesda, Md., earlier this month, several industry groups, including the Biotechnology Industry Organization, the Drug Information Association, and PhRMA, met with the FDA to discuss guidelines for the development of drug/test combinations, such as the breast cancer drug, Herceptin, and its assay, HercepTest.
These combination products, sometimes called "theranostics," are particularly difficult to develop, says Lesko. "On a pragmatic basis it's complex because it means the strict coordination between developing a drug, which is difficult enough on its own, but also developing the evidence that the test would support patient selection for the use of the drug." The FDA has published a concept paper on drug/test codevelopment, as well as the pharmacogenomics data submission guidance document, at
Though it may be some time before physicians routinely use pharmacogenetics data, a small handful of tools are now available. In December 2004, the FDA approved its first such product, Roche Molecular Systems' AmpliChip Cytochrome P450 (CYP450) test. The microarray-based diagnostic system detects 29 polymorphisms in the CYP450 2D6 gene and two in 2C19. These enzymes play key roles in drug metabolism, and genetic variations in the genes can affect the rate of drug metabolism.
Although tests for such polymorphisms already exist and are in limited clinical use, these are typically batteries of assays based on PCR. The AmpliChip "reflects a lot of progress," says Julio Licinio, head of the Center for Pharmacogenomics and Clinical Pharmacology at the University of California, Los Angeles. "It predicts rates of metabolism for drugs based on genetic variants in the cytochrome p450 genes," and as such, the chip can help guide treatment, says Licinio, who is also editor of
To what extent this will happen, however, is still a matter of debate. No algorithm yet exists to tell physicians exactly how to adjust dosages based on test results. Moreover, such data are not always useful. Johnson & Johnson's antipsychotic drug, Risperdal, for example, is metabolized by 2D6. But, doctors don't need to adjust dosages based on 2D6 genotype, says J&JPRD's Cohen, because it has no effect on toxicity or efficacy. The same is true of Eli Lilly's attention deficit hyperactivity drug, Strattera. The antipsychotic drug thioridazine, however, can cause fatal cardiac arrhythmias in patients who have reduced 2D6 activity, or who are taking drugs such as Prozac that may inhibit 2D6.
"There's enough information out there to tell us that this is important, and not enough information out there to optimally guide us in using it yet," says Bruce Cohen, president and psychiatrist-in-chief at McLean Hospital, Belmont, Mass. The AmpliChip currently sells for €400 in Europe (US $515); US pricing has not yet been set.
PHARMACOGENETICS NETWORK SPURS COLLABORATION
HOW GOOD DRUGS GO BAD:
© PharmGKB 2004
This pathway shows the biotransformation of the chemotherapeutic, irinotecan, to its active metabolite, SN-38. SN-38 is primarily metabolized to the inactive SN-38 glucuronide by UGT1A1, but genetic variation in this gene, as well as in other metabolic enzymes and transporters shown here, could explain observed interpatient variability in response, both positive and negative, to the drug.
While cytochrome p450s are important, they are not the only genes to affect drug response. In addition to UGT1A1, there also is thiopurine methyltransferase (TPMT), which metabolizes the anticancer drug 6-mercaptopurine. TPMT-deficient patients are at risk of severe bone marrow toxicity at otherwise normal drug dosages. Now the NIH's PGRN is looking to expand the roster of genes that can predict how a patient will react to a drug.
Comprising 10 research groups that share resources, ideas, and a database called PharmGKB.org, PGRN was conceived in the late 1990s, at a time when NIH budgets were increasing and large-scale science was coming down the pike, says Rochelle Long, PGRN's program director. "The whole purpose of the initiative is to link genotype to phenotype: 'Where do genetic variations matter for drug responses, and where can we make robust research correlations and store that information in a knowledge base?'," she says.
Licinio heads up one of the PGRN networks; Howard McLeod leads another. An overriding theme in the PGRN, McLeod says, is honing pharmacogenetics research, to give scientists the tools they need to match gene variants with drug-response phenotypes. "The network gives you the resources, both financial and intellectual, to tackle these issues," says McLeod, who has been in the PGRN since 2001. Through the network, for instance, his 28-member team can tap into an enormous patient cohort (4,500 individuals in National Cancer Institute clinical trials) and screen their DNA for variants in the 140-odd genes his group has identified as critical to drug effects.
Richard Hockett, medical fellow, department of diagnostic and experimental medicine at Eli Lilly, says as many as 180 genes could affect drug metabolism, including metabolic enzymes, transporters, and other proteins. By his count those genes contain at least 2,000 different variants, and a truly comprehensive metabolic genotyping panel, he says, would have to test for all of them. In March, Eli Lilly and ParAllele BioScience of South San Francisco announced the development of just such a chip. Starting this summer, the MegAllele D-MET chip will be used to screen patients in Lilly's Phase I trials. ParAllele plans to file for FDA approval of the chip later this year.
Other AmpliChip products are in the pipeline at Roche, according to Walter Koch, vice president and head of research at Roche Molecular Systems, including a p53 resequencing array, a molecular classification tool for leukemia, and an HLA genotyping array.
But experts say it could be a decade or more before such genetic testing is the norm and not the exception. One limiting factor: It takes years to run the comprehensive clinical trials necessary to truly validate a biomarker.
"We think every publicly funded clinical trial should contain pharmacogenetics," says Mary Relling, a PGRN member who chairs the department of pharmaceutical sciences at St. Jude Children's Research Hospital, Memphis, Tenn. "We should be getting DNA and appropriate consent from patients on every trial that's supported by tax dollars," says Relling. "Otherwise, 20 years from now we will have made very little progress."