BARBARA J. EVANS
I. INTRODUCTION
The United States health care industry is sometimes characterized as a system of disease care that sits on its hands until people fall ill and then mounts a costly, crisis-oriented response, often after the patient is beyond help (Brigham and Johns 2012; Topol 2012). The Patient Protection and Affordable Care Act of 2010 modestly alters the way health care is financed but does very little to alter this essence of U.S. health care. After decades of investing more than other Organisation for Economic Co-operation and Development (OECD) nations spend per capita on health care, Americans qualify as some of the sickest, shortest-lived rich people in the world. A growing body of literature decries this “disease-oriented, reactive, and sporadic approach to care” and calls for a shift to “prospective medicine” that would help healthy people stay well (Snyderman and Yoediono 2008). This shift presents unique challenges for the U.S. Food and Drug Administration’s (FDA’s) regulation of medical products. This chapter explores these challenges. It describes how the Food and Drug Administration Amendments Act of 2007 (FDAAA) helps address the challenges and inquires whether FDA is making full use of its new powers under FDAAA. Although it is too early to appraise the agency’s overall success in implementing FDAAA, there are early signs that the agency may be underconstruing its statutory powers in ways that may impede the progress of twenty-first-century prospective medicine.
II. THE CHALLENGES OF PROSPECTIVE MEDICINE
Prospective medicine is sometimes described as “personalized, predictive, preventive, and participatory” (Snyderman and Yoediono 2008). Each of these attributes strains FDA’s traditional regulatory paradigm. Predictive technologies involve testing gene variants or other biomarkers to identify future health risks or to forecast treatment responses, as in pharmacogenetic testing (Bookman et al. 2006; Institute of Medicine 2008). The ultimate goal is to be able to respond to predictive or presymptomatic testing with preventive interventions that can avert the development of disease or at least minimize burdens of illness and treatment.
Predictive and preventive technologies imply a shift from the curative model that now dominates health care. Rapid advances in diagnosis and treatment made it possible for twentieth-century doctors to diagnose and treat manifest disease, but prognosis—which had held an important place in medical practice from ancient times until late in the nineteenth century—fell into disrepute (Christakis 1997; Rich 2002). Predicting health outcomes at the level of individual patients was seen as ethically questionable and unscientific because twentieth-century science treated individual health outcomes as a product of random chance (Christakis and Iwashyna 1998; Evans 2010). Recent advances in the life sciences are giving prediction and prognosis an evidentiary basis that they previously did not have and are “catalyzing a revolution in healthcare focused around an informational view of medicine” (Institute for Systems Biology 2010).
As Richard Merrill has remarked, FDA’s product-oriented regulations furnish an “odd-fitting framework for regulating what is basically an information service” (Merrill 2000), which is an apt characterization of predictive tests. To complicate matters further, participatory medicine envisions a “far greater role for patient involvement” (Snyderman and Yoediono 2008), which means patients will need greater access to health information than they traditionally have enjoyed. Even when a test is analytically valid, it may take many years to establish that the test results have clinical validity and utility, and when relationships between biomarkers and clinical conditions are “well established,” this often implies a statistical rather than deterministic relationship (Bookman et al. 2006). Even when motivated by worthy health and safety objectives, governmental efforts to restrict flows of information can raise First Amendment questions as FDA has learned in a series of bruising encounters with the courts in recent years (see e.g., Washington Legal Foundation v. Friedman 1998; Pearson v. Shalala 1999; Thompson v. Western States Medical Center 2002; Whitaker v. Thompson 2002; United States v. Caputo 2003; Alliance for Natural Health US v. Sebelius 2010; United States v. Caronia 2012). It is one thing for FDA to deem a product insufficiently safe and effective to be allowed onto the market. It is more disturbing when a government agency declares that information is not safe and effective and thus must be suppressed (Evans 2014).
Personalized medicine negates the pivotal simplifying assumption of FDA’s twentieth-century drug regulatory framework—that population-average statistics for safety and effectiveness are informative when treating individuals (Evans 2010; Topol 2012). Personalized medicine accepts that this assumption is a fiction and aspires to use medical products in ways that will be safe and effective at the individual level. “Over the past half century, biomedical science has developed randomized, controlled clinical-trial methods that can distinguish treatment effects from the noise of human variability” (Woodcock and Lesko 2009). In personalized medicine, unfortunately, the eliminated “noise” is precisely what is interesting. FDA’s premarket clinical trials are too small, too brief in duration, and too unreflective of actual patients to detect and explain the differences in individual responses to medical products. Today’s medical products routinely come to market trailing clouds of uncertainty about how patients will respond and with little or no practical information to help clinicians optimize individual health outcomes.
This problem only grows worse in a world of preventive medicine. Unlike today’s medical products that regularly exhibit late-emerging safety problems, tomorrow’s products will have late-emerging issues with efficacy as well as safety. Preventive medical products deliver their benefits over long time scales—from twenty to fifty years—that make it impractical to assess effectiveness using FDA’s traditional premarket clinical trials (Evans 2010). The impossibility of proving “preventive effectiveness” in a compact premarket trial implies that many preventive interventions will be off-label uses of products that already are on the market for therapeutic indications, and there may be little data to support the preventive use of the product. Products that do receive FDA approval for a preventive indication often will have been approved using surrogate endpoints of effectiveness. Assessing whether preventive interventions actually work will require ongoing postmarketing studies—not only of safety but of effectiveness. Validating the effectiveness of preventive interventions is critical because prospective medicine envisions putting drugs into healthy people who, although susceptible to illness, are presently asymptomatic. Safety risks that would be acceptable when treating manifest disease may be entirely unacceptable unless there is clear evidence that the preventive intervention works (Evans 2010).
III. NEW REGULATORY POWERS UNDER FDAAA
FDAAA was a forward-looking statute that gave FDA crucial powers to regulate new kinds of medical products that are emerging in the age of prospective medicine. FDAAA represented the most profound modernization of FDA’s drug regulatory framework since the 1962 Drug Amendments (Pub. L. No. 87-781, 76 Stat. 780), which first empowered FDA to require evidence of safety and effectiveness before approving new drugs. FDAAA did not change FDA’s premarket drug approval process but supplemented it with an array of new powers after drugs are approved (Evans 2010). To summarize, FDAAA accepted that premarket clinical trials cannot fully answer questions about new drugs and expanded FDA’s authority to require manufacturers to conduct postmarketing studies and clinical trials (21 USC § 355(o)(3)). It recognized that making drugs safe in clinical use requires ongoing, shared effort that draws on numerous methodologies, including large-scale observational studies to monitor how approved drugs perform in actual clinical use. It authorized FDA to develop a large-scale data infrastructure in the form of a 100-million-person postmarketing “risk identification and analysis system” (21 USC § 355(k)(3)). This data infrastructure, known as the Sentinel System, relies on administrative data (such as insurance claims information) and clinical records (Behrman et al. 2011; Robb et al. 2012). FDA’s Sentinel pilot project, known as Mini-Sentinel, is already operating and responding to drug safety queries from FDA (Curtis et al. 2012; Platt et al. 2012).
FDAAA establishes a hierarchy of approaches for investigating safety issues with approved drugs (21 USC § 355(o)(3)(D)(i)-(ii)). First, FDA must study the problem itself using its traditional adverse event reporting system and the new Sentinel System. If these approaches are insufficient to answer the drug safety question, FDA may order the drug manufacturer to do a § 505(o)(3) “study,” which is defined in a way that includes investigational methods other than clinical trials, for example, laboratory and animal studies and observational approaches (Food and Drug Administration 2011). The agency can order the manufacturer to conduct a randomized, controlled clinical trial during the postmarketing period only as a last resort if all of these other methods are deemed insufficient to answer the safety question.
Crucially, FDAAA lets FDA take concrete regulatory actions in response to new evidence developed after drugs are approved. The agency can require safety-related labeling changes after certain procedural steps (21 USC § 355(o)(4)), and it can distribute risk information to patients and physicians through a new Internet-based communication system (21 USC § 355(r)(2)(a)). FDAAA’s Risk Evaluation and Mitigation Strategy (REMS) program offers a way for FDA to condition the sale of drugs on specific measures to manage their risks (21 USC §§ 355(p), 355-1). FDA can require a REMS at the time of a new drug’s approval or later in the drug’s life if emerging evidence shows that a REMS is necessary to ensure that the drug’s benefits outweigh its risks (21 USC § 355-1(a)). Every REMS must provide for ongoing evaluation of the drug’s risks and also may include additional discretionary elements (21 USC § 355-1(d)-(e)). These may be as simple as requiring a medication guide, patient package inserts, or warning letters to health care providers.
A REMS can include more stringent elements if a drug is effective but has known risks so serious that the drug otherwise would not be approvable (or, if already approved, would need to be removed from the market). Under these circumstances, FDA can condition sales of the drug on restrictions known as “elements to ensure safe use” (21 USC §§ 355-1 (f)(1)(B), (3)). FDAAA identifies six such elements, including one that allows the agency to require that patients undergo laboratory tests before they can receive the drug (21 USC § 355-1(f)(3)(D)). To date, FDA has used this power to require pregnancy testing in conjunction with prescriptions for teratogenic drugs, but this provision also would lend itself to requiring pharmacogenetic tests in situations where a drug, without the testing, does not have an acceptable risk–benefit ratio. This provision, in effect, gives FDA a way to achieve the functional equivalent of cross-labeling the drug and a companion diagnostic. It thus could help resolve legal and practical problems that have made it hard to cross-label drugs and tests (Evans 2010). This is just one example of the ways FDAAA helps address barriers to the clinical translation of pharmacogenomics and personalized medicine; a more-detailed discussion has been provided elsewhere (Evans 2010).
IV. REGULATING EFFECTIVENESS DURING THE POSTMARKETING PERIOD
FDAAA’s § 505(o)(3) lets FDA order postmarketing studies and clinical trials only if the agency has concerns about a drug’s safety (21 USC § 355(o)(3)(B)). Safety concerns also are the trigger for requiring labeling changes and REMS elements to assure safe use. FDAAA equips FDA to address postmarketing safety questions, but what about efficacy questions that loom so large in prospective medicine? FDA can only order postmarketing investigations of drug safety issues, but this focus is not as narrow as it initially sounds.
FDAAA’s § 505(o)(3) (21 USC § 355(o)(3)) adopts a set of definitions that include “failure of expected pharmacological action” as a type of serious drug safety risk, if the failure entails serious medical consequences for the patient (21 USC §§ 355-1(b)(1)(E), (b)(4), and (b)(5); 355(o)(2)(C)). For example, if an antibiotic fails to work because germs are now resistant to it, and if patients who take the drug are seeing their infections progress to gangrene, the drug’s efficacy failure would qualify as a serious safety risk under FDAAA’s definitions. FDAAA’s approach of viewing failures of expected efficacy as potential safety problems is consistent with FDA’s longstanding practice of framing “safety” as a favorable ratio of benefits and risks (Food and Drug Administration 2011). If a drug fails to deliver its expected pharmacological action, then obviously its ratio of benefits and risks is diminished, and the drug is not as “safe” as it was originally thought to be.
FDA’s power to address unexpected efficacy failures is an outgrowth of FDAAA’s definition of serious drug safety risks. Specifically, the statute defines the term adverse drug experience as including events FDA traditionally has counted as drug-related adverse events plus “any failure of expected pharmacological action of the drug” (21 USC § 355-1(b)(1)). This last provision, for example, seems to count therapeutic nonresponse as an adverse drug experience. Nonresponse happens when a patient is neither poisoned nor helped by a drug. For such patients, the expected pharmacological action, which was inferred from population-average data from premarket clinical trials, fails to materialize.
Traditionally, FDA regarded nonresponse as an efficacy problem but not a safety problem. The risk–benefit methodology FDA employs when approving new drugs distinguishes harms that are directly caused by the drug itself (drug toxicity) from consequential harms (such as worsening of the patient’s illness that occurs because the drug failed to produce its expected therapeutic response) (Food and Drug Administration 2004; Food and Drug Administration 2005). These consequential harms can be real and serious, for example, when patients have chronic or progressive diseases that worsen while the patient is taking an ineffective drug. The safety dimensions of treatment failure are all the more sobering because an estimated 30 to 60 percent of the prescriptions written in the United States result in nonresponse (Peakman and Arlington 2001; Henderson and Reavis 2008).
FDAAA defines a “serious” adverse drug experience as one that kills or puts the patient at immediate risk of death, that causes persistent or significant incapacity or substantial disruption of ability to conduct normal life functions, or that requires medical or surgical intervention to prevent such outcomes (21 USC § 355-1(b)(4)(A)). FDAAA equates “serious risk” to risk of a serious adverse drug experience (21 USC § 355-1(b)(5)). These definitions supply statutory authority for FDA to address nonresponse using the same powers the agency has for addressing problems with drug toxicity. Thus, FDAAA would allow FDA to order postmarketing studies and clinical trials, labeling changes, or REMS use conditions (such as ordering a pharmacogenetic test) to address unexpected efficacy failures. These powers potentially include the power of FDA to address heterogeneity of treatment effects, the phenomenon that causes individual responses to depart from the population-average effectiveness data that are the basis for FDA’s new drug approvals. This new statutory authority to address unexpected efficacy failures in approved drugs is one of FDAAA’s most crucial features as FDA adapts to an era of predictive, preventive, and personalized medicine.
V. IS FDA UNDERCONSTRUING ITS POWERS UNDER FDAAA?
In its 2006 Future of Drug Safety report, the Institute of Medicine remarked on FDA’s record of poor follow-through on proposed initiatives (IOM 2006). Perhaps attentive to that criticism, the agency has displayed admirable speed in implementing key aspects of FDAAA such as the Sentinel System. There are isolated areas of concern, however, one of which involves the postmarket efficacy issues discussed in the previous section. While it is still too early to draw conclusions, FDA appears to be turning its back on the crucial statutory authority that FDAAA gave the agency to address efficacy failures during the postmarket period.
The agency published final guidance on Postmarketing Studies and Clinical Trials (“Guidance”) in April 2011 (Food and Drug Administration 2011). This Guidance clarifies how the agency intends to exercise its powers to order manufacturers to conduct postmarketing studies and clinical trials under FDAAA’s § 505(o)(3). As just discussed, the statute envisions that FDA can use this power to generate evidence about serious adverse drug experiences and defines this italicized phrase as meaning:
any adverse event associated with the use of a drug in humans, whether or not considered drug related, including—
(A) an adverse event occurring in the course of the use of the drug in professional practice;
(B) an adverse event occurring from an overdose of the drug, whether accidental or intentional;
(C) an adverse event occurring from abuse of the drug;
(D) an adverse event occurring from withdrawal of the drug; and
(E) any failure of expected pharmacological action of the drug.
—(21 USC § 355–1(B)(1))
FDA’s April 2011 Guidance did considerable violence to this definition, shortening it to:
Adverse drug experience includes an adverse event in the course of the use of the drug in professional practice.
—(FOOD AND DRUG ADMINISTRATION 2011)
This rendering appears in an appendix to the Guidance that purports to provide definitions from 21 USC § 355-1(b), yet it completely jettisons prongs (B)–(E) of the statutory definition of adverse drug experience. Elsewhere, this same appendix faithfully renders the complete statutory definitions of other terms such as serious adverse drug experience and serious risk. Omitting subsections (B)–(E) eliminates the crucial clause at section 355-1(b)(1)(E) that mentions “any failure of expected pharmacological action of the drug.” This is the clause that authorizes FDA to address efficacy failures during the postmarketing period.
This omission does not appear to have been a mere accident or an abbreviation of longer statutory text. In the main text of the Guidance, FDA affirmatively disclaims its authority to address unexpected efficacy failures during the postmarketing period. The Guidance asserts that FDA cannot require clinical trials
in which the primary endpoint is related to further defining efficacy, designed to:
• Evaluate long-term effectiveness or duration of response
• Evaluate efficacy using a withdrawal design
• Evaluate efficacy in a subgroup
—(FOOD AND DRUG ADMINISTRATION 2011)
Yet evaluating efficacy in a population subgroup is precisely the type of investigation that the statute at 21 USC § 355-1(b)(1)(E) does appear to authorize. In effect, the April 2011 Guidance rewrites the statute in a way that eliminates a power that the statute, in fact, appears to confer. Some of the pharmaceutical industry’s comments on the proposed Guidance called for postmarketing efficacy studies to be classified as “commitments” that companies could agree to do rather than “requirements” that FDA could order them to do (see, e.g., Baranello 2009). The Guidance may have been trying to allay industry concerns that the agency will exploit FDAAA’s definitions to order wide-ranging postmarketing investigations of efficacy-related issues. The agency disclaimed any intent to require randomized, controlled clinical trials (RCTs) in which the primary endpoint is an efficacy question. This seemingly includes trials of efficacy in a suspected nonresponding population subgroup (Food and Drug Administration 2011).
Subgroup efficacy studies are very important in the age of pharmacogenomics. The agency approves drugs based on average safety and efficacy statistics that may badly misrepresent a drug’s actual clinical performance in people who happen to be cursed with nonaverage genes. FDAAA’s definition of “serious risks” seemingly would allow the agency to require postmarketing investigations of subgroup efficacy if the agency chose to do so. The April 2011 Guidance suggests that FDA has no wish to do so. It is true that FDA’s authority to order investigations under § 505(o)(3) is discretionary in nature, and federal agencies can choose not to regulate to the maximal extent that their statutes theoretically would allow. The April 2011 Guidance modestly suggests that clinical trials of subgroup efficacy might go forward as voluntary commitments between drug manufacturers and FDA. In contrast, the statute envisions that the agency can require such studies.
Federal agencies are allowed to exercise enforcement discretion, but they are not allowed to rewrite their statutes. A familiar example conveys this crucial distinction: A police officer is not required to ticket a motorist who is driving sixty miles per hour (mph) in a thirty-mph zone. The officer can choose to let the driver go with a warning, and this would amount to an exercise of the officer’s enforcement discretion. On the other hand, the officer cannot declare that the speed limit is sixty mph when it is, in fact, thirty mph. That would be amending the statute, which only the legislature can do. When the April 2011 Guidance affirmatively disclaimed the agency’s power to order clinical trials of subgroup efficacy effects, it arguably crossed the line into rewriting FDAAA.
FDA did not, however, completely wash its hands of addressing efficacy issues that pose potential risks to patient safety. The agency stood its ground when public comments asked it to delete certain types of efficacy-related investigations from the list of § 505(o)(3) studies (that is, investigations other than RCTs) that FDA can require. For example, public comments asked FDA to remove studies that explore mechanisms of drug resistance that might cause a drug to fail to deliver its expected level of efficacy (Baranello 2009). The final Guidance kept those studies in the list of postmarketing investigations that FDA can require under § 505(o)(3).
At the risk of overinterpreting what may have been inadvertent language in the April 2011 Guidance, it appears that FDA may be willing to require a § 505(o)(3) study (such as an observational study or a laboratory study) to investigate failures of a drug’s expected pharmacological action, but the agency will stop short of ordering postmarketing RCTs to investigate such failures. Under the rubric of § 505(o)(3) studies, FDA could, for example, require a prospective registry study to identify subgroups that have failed to respond to a drug, or FDA could require a laboratory analysis of biospecimens obtained with the consent of such patients in the hope of identifying a gene variant that predicts their nonresponse to the drug. However, FDA apparently would stop short of requiring a clinical trial that randomizes patients from a discrete, genetically defined subpopulation to receive various therapies to determine which treatment works better for them.
Actual practice in coming years will clarify how actively FDA intends to use its powers under § 505(o)(3) to foster pharmacogenetic discovery to improve drug safety for all people, including those in atypical genetic subgroups. Although the April 2011 Guidance shrank from requiring clinical trials of subgroup efficacy, it did not slam the door shut on other valuable pharmacogenetic investigations, such as laboratory and observational studies that could be carried out as § 505(o)(3) studies.
The same is true of § 505(o)(3) observational studies to assess long-term efficacy and duration of treatment response—the April 2011 Guidance did not rule them out. It merely ruled out RCTs for assessing long-term efficacy. Observational studies may, in fact, be the superior alternative for assessing long-term efficacy and duration of treatment response. Long-term RCTs, apart from their cost, raise troubling ethical issues (for example, keeping patients unaware whether they are taking a test or control drug during a double-blind RCT that continues for many years) and logistical problems (e.g., high drop-out rates in prolonged RCTs; Evans 2010). In light of the many problems they pose, FDA may not have sacrificed a practical option by renouncing RCTs of long-term efficacy.
VI. CONCLUSION
FDAAA imbued FDA with important powers for adapting the agency’s medical product regulations to the needs of twenty-first-century prospective medicine. To succeed, however, the agency will need to embrace its new powers and make skillful, aggressive use of them.
REFERENCES
Alliance for Natural Health US v. Sebelius, 714 F. Supp. 2d 48 (DDC 2010).
Behrman, Rachel E. et al. 2011. “Developing the Sentinel System—A National Resource for Evidence Development.” New England Journal of Medicine 364:498–99.
Bookman, Ebony B. et al. 2006. “Reporting Genetic Results in Research Studies: Summary and Recommendations of an NHLBI Working Group.” Am. J. Med. Genet. Part A (May 15):1033–40.
Brigham, Kenneth and Michael M. E. Johns. 2012. Predictive Health: How We Can Reinvent Medicine to Extend Our Best Years. New York: Basic Books.
Christakis, Nicholas A. 1997. “The Ellipsis of Prognosis in Modern Medical Thought.” Social Science and Medicine 44(3):301–15.
Christakis, Nicholas A. and Theodore J. Iwashyna. 1998. “Attitude and Self-Reported Practice Regarding Prognostication in a National Sample of Internists.” Archives of Internal Medicine. 158:2389–95.
Curtis, Lesley H. et al. 2012. “Design Considerations, Architecture, and Use of the Mini-Sentinel Distributed Data System.” Pharmacoepidemiology and Drug Safety 21(Supplement S1):23–31.
Evans, Barbara J. 2010. “Seven Pillars of a New Evidentiary Paradigm: The Food, Drug, and Cosmetic Act Enters the Genomic Era.” Notre Dame Law Review 85:419–524.
Evans, Barbara J. 2014. “The First Amendment Right to Speak About the Human Genome.” University of Pennsylvania Journal of Constitutional Law 16:549–636.
Institute for Systems Biology. 2010. Annual Report: Pushing Boundaries. Seattle: Institute for Systems Biology.
Institute of Medicine, Committee on the Assessment of the U.S. Drug Safety System. 2006. The Future of Drug Safety. Washington, D.C.: The National Academies Press.
Institute of Medicine, Roundtable on Translating Genomic-Based Research for Health. 2008. Diffusion and Use of Genomic Innovations in Health and Medicine: Workshop Summary. Washington, D.C.: The National Academies Press.
Merrill, Richard A. 2000. “Genetic Testing: A Role for FDA?” Jurimetrics 41(1):63–66.
Pearson v. Shalala, 164 F.3d 650, 658 (D.C. Cir. 1999)
Platt, Richard et al. 2012. “The U.S. Food and Drug Administration’s Mini-Sentinel Program: Status and Direction.” Pharmacoepidemiology and Drug Safety 21(Supplement S1):1–8.
Rich, Ben A. 2002. “Prognostication in Clinical Medicine: Prophecy or Professional Responsibility?” Journal of Legal Medicine 23:297–358.
Robb, Melissa A. et al. 2012. “The U.S. Food and Drug Administration’s Sentinel Initiative: Expanding the Horizons of Medical Product Safety.” Pharmacoepidemiology and Drug Safety 21(Supplement S1):9–11.
Snyderman, Ralph and Ziggy Yoediono. 2008. “Perspective: Prospective Health Care and the Role of Academic Medicine: Lead, Follow, or Get Out of the Way.” Academic Medicine 83(August):707–14.
Thompson v. Western States Medical Center, 535 U.S. 357 (2002).
Topol, Eric. 2012. The Creative Destruction of Medicine: How the Digital Revolution Will Create Better Health Care. New York: Basic Books.
United States v. Caputo, 288 F. Supp.2d 912, 920–21 (S.D.N.Y. 2003).
United States v. Caronia, 703 F.3d 149 (2d Cir. 2012).
Washington Legal Foundation v. Friedman, 13 F. Supp. 2d 51 (DDC 1998).
Whitaker v. Thompson, 248 F. Supp. 2d 1, 2, 7 (DDC 2002)
Woodcock, Janet and Lawrence J. Lesko, 2009. “Pharmacogenetics: Tailoring Treatment for the Outliers.” New England Journal of Medicine 360(8):811–13.
