American Journal of Law & Medicine

Demanding individually safe drugs today: overcoming the cross-labeling legal hurdle to pharmacogenomics.

If new refrigerators hurt 7% of customers and failed to work for another one-third of them, customers would expect refunds ... manufacturers would be strictly liable for the injuries, and there would be implied warranties even if the manufacturer made no guarantees. (1)


What if physicians could use genetic tests to tailor prescriptions to their patients' individual genotypes? Physicians and pharmaceutical companies can use pharmacogenomics to decrease the number of adverse drug reactions, increase drug efficacy, and lower health care costs. Unfortunately, cross-labeling rules serve as both legal and policy hurdles for these advances, hurdles the FDA has the power to remove. Part I explains pharmacogenomics and why it currently has a narrow application. Part II discusses the FDA's regulatory approach to pharmacogenomics. Part III explains the legal and policy hurdles of cross-labeling and how they impede the more widespread use of pharmacogenomics. Part IV examines ways to dear the legal cross-labeling hurdles while Part V examines ways to clear the policy cross-labeling hurdles. Finally, Part VI discusses some of the many other complex legal and policy issues that lawmakers, regulators, and the industry will need to resolve in order to realize the full potential of pharmacogenomics.



Drug therapies are never safe for every individual patient. A frequently cited meta-analysis published in 1998 found that serious adverse drug events cause two million hospitalizations and 100,000 deaths in the U.S. annually, making adverse drug reactions the fifth leading cause of death in America. (2) Even when drugs are generally safe, they are not always effective. A 2001 study found that efficacy rates for drugs used to treat most diseases typically range between 50% and 75%. (3)

This seemingly low efficacy rate stems from the great variability in the way that individuals respond to particular drugs. After it is administered, a drug's plasma concentrations vary as much as thirty to fifty-fold between patients, even if they receive equal doses. (4) With warfarin, an anticoagulant, there can be as much as a twenty-fold difference in dosages required to achieve the desired therapeutic effect in individual patients. (5)

Numerous environmental and genetic factors cause this variability. Non-genetic factors include age, organ dysfunction or other disease, pregnancy, lactation, drug-drug interactions, and diet. (6) Genetic factors are primarily individual differences in gene expression in the construction of gene products that metabolize drugs. (7) These gene products include drug-metabolizing enzymes, drug transporters, drug targets, and downstream signal transduction molecules. (8) Single-nueleotide polymorphisms (SNPs) in the genes that encode gene products cause these individual differences. (9) Pharmaeogenomics is the study of how these individual genetic variances affect the body's response to drugs. (10)

Traditionally, physicians used trial-and-error to determine the optimal dosages for individual patients. (11) Pharmacogenomics can change this. If researchers can find specific associations between genetic variations and drug reactions and can develop diagnostic tests for those genetic variations, they can tailor prescriptions to individual patients. These diagnostic tests reveal a patient's genotype (12) even though they are not all genetic tests; some ascertain genotype by testing for the presence of gene products. (13) Researchers refer to the gene, protein, or chemical that they test for as a biomarker. (14) There already are a significant number of valid biomarkers, which the FDA defines as ones "measured in an analytical test system with well-established performance characteristics" and for which there is "an established scientific framework or body of evidence that elucidates the physiologic, pharmacologic, toxicologic, or clinical significance of the test results." (15) Researchers have also identified many exploratory markers, but currently lack the evidence to label them as valid. (16)

For example, three families of CYP genes (CYP1, CYP2, and CYP3) are the major genes that code for enzymes used in oxidative metabolism. (17) Polymorphisms in these genes affect the activity level of the enzymes they produce. (18) Some variations cause the enzymes to break medications down too quickly before they can work, while other variations cause them to break medications down too slowly, allowing the medication to accumulate to dangerous levels. (19) The CYP2C9*2 and CYP2C9*3 alleles of the CYP2C9 gene are associated with slow metabolization of warfarin, which increases the risk of over-anticoagulation and serious bleeding. (20) Studies estimate that 30% of the population carries alleles causing warfarin sensitivity. (21) Thus, a physician can use a genetic test to determine the presence of the CYP2C9*2 and CYP2C9*3 alleles and he can adjust the normal dosage downwards if the results are positive, thereby avoiding unnecessary adverse reactions, and upwards if the results are negative, thereby preventing strokes that might result from under-medication. (22) Genetic tests for CYP2C9 genotypes are currently available, (23) and in January 2005, the FDA approved Roche's AmpliChip CYP450 Array, which tests for the CYP2D6 and CYP2C19 genotypes, (24) two genes that encode enzymes that metabolize 25% of all prescription drugs. (25)

Herceptin provides another example of pharmacogenomics. In some breast cancer patients, the HER2 gene over-expresses itself, causing over-expression of the HER2 receptor protein. (26) Hereeptin targets this protein (27) and has demonstrated a response rate of 35% for patients with HER2 over-expression (roughly one third of breast cancer patients). However, Herceptin is ineffective for the two-thirds of patients without over-expression of HER2. (28) Clinicians are currently utilizing this targeted therapy and are achieving positive results. (29)

Thus, pharmacogenomics can improve drug therapies in a variety of ways. Physicians can use diagnostic tests to adjust dosages, prescribe alternative drugs, or prescribe no drug at all depending on a patient's genotype. (30) They can also use them to decide whether to prescribe a drug with a narrow therapeutic index that will only be effective in a subset of patients. (31) In sum, pharmacogenomics can optimize physician prescribing and drug effectiveness.


   [U]pon further examination of your test results, your doctor finds 
   that you would benefit greatly from a new drug on the market, and 
   that there would be little likelihood that you would react 
   negatively to it. A day like this will be coming to your doctor's 
   office soon, brought to you by pharmacogenomics. (32) 

Commentators and stakeholders have been touting the anticipated benefits of pharmacogenomics for many years. (33) They cite the benefits described above, as well as its potential to provide advance screening for diseases, improved vaccines, quicker and cheaper drug discovery and development (with more drugs making it to and staying on the market), and decreases in total health care costs. (34)

The AEI-Brookings Joint Center recently published a study estimating that if physicians performed a genetic test with 95% sensitivity and specificity on every new warfarin patient in the U.S., it would reduce serious bleeding events caused by the drug by 85,000 and strokes caused by under-medication by 17,000 annually, resulting in annual net health care savings of $1.1 billion. (35) The standard of care does not yet require physicians to genetically test warfarin patients, but this study makes a strong case that it should. (36)

Warfarin of course is only one drug. Adverse drug events cause millions of hospitalizations annually, and many of the expensive drugs that we purchase fail to deliver any therapeutic result. (37) Some estimate that the U.S. spends up to $65 billion annually on drugs that produce no therapeutic effect at all. (38) One study estimated that the cost of treating adverse drug reactions (morbidity and mortality) is $177 billion annually. (39) In another, the authors estimated that 50% of the 2,227 adverse drug events surveyed were likely associated with genetic factors. (40) A third study found that if physicians monitor drug therapy more closely, they can prevent 76% of geriatric drug-related hospital readmissions. (41) Thus, pharmacogenomics has enormous potential for improving health care quality and reducing its cost.


   Rational economic calculations on the part of pharmaceutical 
   companies do not always translate into sound social policy. (42) 

Recently, commentators have noted that the value of pharmacogenomics has not been realized as soon as some had hoped. (43) The examples cited above are part of the small group of pharmacogenomic applications that have actually reached clinical practice. (44) As of September 2006, Roche's AmpliChip had failed to achieve widespread clinical use or reimbursement. (45) Although physicians (46) and insurers are partly responsible for the fact that pharmacogenomics is not the current standard of care (by failing to use and reimburse these tests), pharmaceutical companies are also responsible. Pharmaceutical companies play an important role in creating pharmacogenomic data, developing diagnostic tests, and encouraging physicians to use the tests by referencing them in drug labeling.

Pharmaceutical companies may not have enough incentive to support widespread pharmacogenomics. Suppose that a brand prescription drug is effective in 60% of patients, (47) has no effect in 30% of patients, and produces an adverse reaction in 10% of patients. Suppose further that the drug is likely to be approved for the general population or is already approved. Without pharmacogenomics, physicians will be unable to tell which patients will benefit from the drug and which will not; 100% of these patients might be prescribed the drug. If physicians can identify the non-responders and highly sensitive patients, however, they will prescribe an alternative drug or no drug at all. If the pharmaceutical company that sells this drug develops this test or allows a diagnostic company to develop it, it will lose 40% of the drug's market. "Companies with high levels of sales to nonresponding individuals seemingly have the most to lose. Commercial incentives for research may be weakest precisely where improved targeting is most needed." (48)

Pharmaceutical companies may not be ready to let go of the "blockbuster" model just yet. (49) Drugs that only small subgroups of the population will use are simply a less attractive investment. (50) "In some big pharma companies, the use of pharmacogenomics is being driven by marketing departments whose priority is screening out anything that is not pharmacogenomically 'clean." (51) Further, if fewer patients use a given drug, unit prices will increase and revenues may decline. (52) Thus, pharmaceutical companies are likely to develop only targeted drugs that treat the most common and profitable genotypes and subpopulations. (53)

Pharmaceutical companies have some incentives to embrace pharmacogenomics. First, they can use it to improve the efficiency of the drug approval process. Pharmaceutical companies could obtain faster FDA approvals by maximizing their trials' safety and efficacy endpoints. (54) They can use existing pharmacogenomic data to adjust inclusion and exclusion criteria in clinical trials so that study groups include only those patients most likely to benefit without adverse events. (55) They can use pharmacogenomic data from early clinical trials to similarly adjust inclusion and exclusion criteria for later trials. (56) This could enable them to reduce samples sizes thereby saving time and money. (57) Finally, they can use pharmacogenomic data to screen out compounds that are not likely to obtain approval before incurring the cost of clinical trials. (58)

However, these cost savings may be offset by the limited patient population approved for the drug. Drugs tested on patients with certain genotypes will only be approved for use on such patients. (59) Gathering and analyzing pharmacogenomic data and developing and obtaining approval for the diagnostic tests is also costly. (60) Pharmacogenomic data discovered during clinical trials may require companies to redesign the trials or conduct additional trials. (61) A pharmaceutical company will only have incentives to incur these costs if it believes the FDA will not approve the drug for general use and that the drug will be profitable enough in the narrower market to outweigh these costs. (62)

Pharmaceutical companies that use pharmacogenomics may "save" drugs that the FDA otherwise would not approve for the general population or drugs that the FDA would force them to withdraw from the market due to serious adverse events. (63) An estimated 50% of Phase III trials fail. (64) However, very few examples of these kinds of drug rescues exist. (65) It is difficult to draw valid statistical inferences from failed trials concerning safety and efficacy in the group that did not experience adverse reactions. This is because earlier clinical trials were not designed to test for safety and efficacy in that subgroup. (66) "Drug rescue is not free, in the sense that a targeted therapy can simply be plucked from the sea of data already on hand." (67) Despite the expenditures that a pharmaceutical company may have already sunk into a drug, the cost of rescuing it via pharmacogenomics may be too great, especially if rescue will only result in the drug's approval for a narrow subpopulation. (68)

Another potential incentive is that pharmaceutical companies may be able to undercut brand drugs by introducing and marketing competing drugs that are safer and more effective for a subpopulation by virtue of the drug having been designed or tested specifically on that subpopulation. (69) If the size of the subpopulation is small, however, it may not represent enough revenue to justify the manufacturer's investment. (70)

Pharmaceutical companies can also use pharmacogenomics to avoid products liability. To recover on a defective design theory, a plaintiff must show that the drug's foreseeable risks were so great compared to the therapeutic benefits that no reasonable physician would prescribe the drug for any class of patients. (71) Plaintiffs struggle to meet this standard of proof where the FDA has approved a drug and therefore determined that its benefits outweigh its risks. (72) Thus, if a pharmaceutical company has reliable data showing safety in the general population, it may not be concerned enough about products liability to develop a pharmacogenomic test. Unless it commits some other error or wrongdoing, failure to develop a pharmacogenomic test alone will not lead to products liability.

Thus, pharmacogenomics is an important public health issue. Pharmaceutical companies sell products that save life and health in the aggregate, but can often be unsafe and ineffective for individuals. Pharmacogenomics has the potential to change this, but pharmaceutical companies may not have the incentives to zealously pursue it. The FDA cannot force pharmacogenomics on the industry and only has limited ability to shift incentives. An independent bio-tech or diagnostic testing firm might, however, have the right incentives to develop and sell a diagnostic test for use with a pharmaceutical company's drug if the pharmaceutical company could not use cross-labeling to stop them. This article focuses on the legal hurdles associated with cross-labeling.


   The FDA is also responsible for advancing the public health by 
   helping to speed innovations that make medicines and foods more 
   effective, safer, and more affordable, and helping the public get 
   the accurate, science-based information they need. (73) 

The FDA's approach to pharmacogenomics "steers a path between aggressive regulation of a new field, which might have a real chilling effect on the development of that field, and a hands-off approach to new science that allows that science to develop without interaction with regulators." (74) In 2002, the FDA held a workshop with pharmaceutical industry groups including PhRMA. (75) In 2003, it published draft guidance for the industry on pharmacogenomic data submissions and received public comments. (76) The FDA learned that the industry was reluctant to introduce genomic studies into its drug development plans because of concerns that the FDA would take the data out of context and request additional clinical trials, put trials on hold, or limit a drug's indication to certain subgroups. (77) The industry was also concerned that FDA staff lacked expertise in interpreting pharmacogenomic data. (78)

The FDA responded in March 2005 by issuing final nonbinding guidance on pharmacogenomic data submissions. (79) The guidance presents recommendations and illustrates the agency's theory on when existing regulations require submission of pharmacogenomic data relevant to safety and efficacy. (80) The FDA noted at the time that there were only a few known valid biomarkers, (81) and that most pharmacogenomic associations that researchers had studied were not "well enough established scientifically to be appropriate for regulatory decision making." (82)

The FDA's guidance distinguishes between "known valid biomarkers," (83) "probable valid biomarkers,"(84) and exploratory biomarkers. (85) It requires submission of pharmacogenomic data that the drug sponsor generates or possesses if the biomarker is known or probable valid, or if the sponsor intends to use the data to make decisions concerning the structure of a clinical trial. (86) In all other cases, submission of data is voluntary. (87) The guidance assures sponsors that the FDA will not use data contained in voluntary submissions for regulatory decision-making on investigational new drug applications (INDs), new drug applications (NDAs), or biologics license applications (BLAs). (88) Instead, it states that the FDA will review the data only to educate and familiarize its staff with pharmacogenomic studies. …

Log in to your account to read this article – and millions more.