Drug Saf DOI 10.1007/s40264-016-0419-8

CURRENT OPINION

Ethical and Practical Considerations in Removing Black Box Warnings from Drug Labels James S. Yeh1

•

Ameet Sarpatwari1 • Aaron S. Kesselheim1

Ó Springer International Publishing Switzerland 2016

Abstract Boxed warnings—also known as ‘‘black box’’ warnings—can be a powerful tool in communicating drug risks to physicians and patients. The overall number of boxed warnings has grown in recent years as the US Food and Drug Administration (FDA) has approved more drugs on the basis of limited pre-marketing information and as new safety issues for marketed drugs have been identified. Two recent manufacturers’ petitions to remove boxed warnings on the drugs rosiglitazone (Avandia) and varenicline (Chantix) have led to divergent FDA decisions and revealed different considerations involved in boxed warning imposition and removal. For ethical and practical reasons, the FDA is justified in applying a higher standard for boxed warning removal than for imposition, as removal of a boxed warning may have unintended effects on physician and patient behavior. However, no guidelines on boxed warning removal currently exist. To promote safe use of approved prescription drugs, the FDA should adopt a uniform and transparent process governing decisions to impose or remove boxed warnings.

& James S. Yeh [email protected] 1

Program on Regulation, Therapeutics, And Law (PORTAL), Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, 1620 Tremont Street, Suite 3030, Boston, MA 02120, USA

Key Points Boxed warning may be imposed when safety concerns about a drug arises and such warnings are less commonly removed from drug labels when initial safety concerns turn out to be unfounded. Ethical and public health considerations necessitate different standards in boxed warning imposition and removal. The consensus for boxed warning removal should be uniform and transparent.

1 Introduction The US Food and Drug Administration (FDA) serves as an important gatekeeper of prescribing choices by determining the approval and ongoing availability of prescription drugs in the USA. After the FDA authorizes a drug, post-approval studies may reveal previously unappreciated drug safety problems. When this happens, the FDA has a number of tools at its disposal to mitigate patient harm, including rescinding marketing authorization, requiring the manufacturer to conduct additional safety trials, and communicating the new information to physicians and patients. Risk communication can entail ‘‘Dear Doctor’’ letters, in which physicians receive direct mailings about amendments to the drug label [1], and risk evaluation and mitigation strategies (REMS), which include various approaches to educate prescribers and patients about safe and appropriate use of the drug [2]. A REMS includes various approaches to reducing risk, depending on the level of the risk. It can highlight

J. S. Yeh et al.

relevant safety information in a drug brochure (medication guide) written for patients. It may involve inclusion of a communication plan to educate prescribers about safe and appropriate use of the drug. It may also require prescribers to undergo a formal training program to ensure proper prescribing of the drug. When particularly severe new risks arise for approved drugs, the FDA can recommend changes to drug labels, including adding boxed warnings—colloquially known as ‘‘black box’’ warnings—in a prominent section at the top of the drug’s label (see Box 1). The FDA can require a new warning to be added to a label when there is reasonable evidence of a clinically significant hazard. Although the primary responsibility for changing a drug’s label in response to such a hazard rests with the brandname drug manufacturer, the FDA also has the authority to initiate a label change by requiring the drug manufacturer to make the change when certain statutory requirements are met [3]. When the identified hazard poses a risk of death or serious injury, a boxed warning strategy may be warranted. First instituted in 1979, boxed warnings can be added at the time of drug approval or after a drug is already on the market to alert physicians and patients about safety issues relevant to an approved prescription drug [4, 5]. Such a move, in tandem with the media coverage it inevitably generates, can substantially reduce prescribing [6]. For example, when widely publicized disclosures of trial data from antidepressant manufacturers showed an association between antidepressant use and suicidality among children and adolescents, a boxed warning was added, and, in subsequent years, prescribing of these products in that population declined [7, 8]. However, drug risk information is fluid and subject to change based on additional experience with use or testing of a drug. In particular, identification of a risk that looks severe and warrants a boxed warning may later turn out to be unfounded, or the risk might turn out to be less severe. While there is substantial experience with adding boxed warnings among regulators and manufacturers in the USA, there is a limited precedent and no enumerated standards for removing them or downgrading the risk warning language to match the description of the drug’s other known risks. In recent years, controversy over how to manage boxed warnings that may require removal has emerged twice, in the cases of the anti-smoking drug varenicline (Chantix) and the diabetes drug rosiglitazone (Avandia).

2 Varenicline, Smoking Cessation, and Psychiatric Events Despite substantial progress in reducing cigarette-smoking rates, tobacco use remains the leading cause of preventable mortality in the USA, accounting for one in five deaths annually [9]. Numerous therapies are available to combat this

epidemic, including varenicline, a partial nicotine receptor agonist approved by the FDA in 2006. After varenicline’s approval, its manufacturer, Pfizer, began fielding reports of neuropsychiatric events, such as suicidality and changes in behavior and mood, among patients prescribed the drug. By 2009, sufficient safety concerns had prompted Pfizer to submit a supplemental new drug application to the FDA to revise varenicline’s label. The FDA, in turn, recommended that Pfizer add a boxed warning that highlighted the association of the drug with serious neuropsychiatric adverse events and recommended close monitoring of varenicline users (see Box 2). The drug safety plan (a REMS) also included development of a medication guide for patients [10] and organization of a randomized controlled trial to better estimate varenicline’s risks. Five years later, as additional data have emerged, the appropriateness of varenicline’s boxed warning has been called into question. In April 2014, Pfizer proposed changes to the drug label on the basis of five short-term (range 1–3 months) retrospective observational studies (four published and one unpublished), comparing varenicline use with nicotine replacement therapy or bupropion, which failed to detect significant associations between varenicline use and neuropsychiatric events [11–14]. Although these studies did not exclude high-risk patients, such as those with pre-existing mental illness, some of the study periods spanned the years after 2008, in which there was widespread awareness among prescribers of the potential risks of neuropsychiatric events associated with varenicline use. This awareness may have prompted selective prescribing of the drug to patients with lower risk. In October 2014, the FDA’s Psychopharmacology Drugs and Risk Management Advisory Committee members met jointly to review varenicline’s label [15]. The committee overwhelmingly voted to retain the boxed warning: 11 members supported no label changes, six favored minor changes, and only one supported the manufacturer’s request for boxed warning removal. Inherent methodological limitations, selective prescribing of varenicline for low-risk patients, and failure to include the full range of outcomes reported in post-approval surveillance may have led the submitted observational studies and meta-analyses to underestimate the true incidence of neuropsychiatric adverse events associated with varenicline use. The committee members anticipated subsequent reexamination of the evidence on varenicline’s neuropsychiatric risks after completion of the ongoing randomized trial.

3 Rosiglitazone, Diabetes, and Cardiovascular Events Rosiglitazone is a thiazolidinedione drug, which makes patients more sensitive to the action of insulin, and is used to treat individuals with type 2 diabetes mellitus. The FDA

Communicating Drug Risks Via Boxed Warning–Ethical and Practical Considerations

approved rosiglitazone in 1999 with a routine (i.e., nonboxed) warning of the risk of heart failure on the basis of pre-clinical studies. Substantial marketing by its manufacturer, GlaxoSmithKline, helped rosiglitazone become a blockbuster and one of the top-selling anti-diabetes drugs of the early 2000s in the USA, garnering $2.2 billion in sales in 2006 alone [16]. In 2007, Cleveland Clinic chief of cardiology Steven Nissen and his colleague Kathy Wolski published a meta-analysis of 42 mostly short-term clinical trials (mean duration 6 months) that showed an increased rate of ischemic cardiovascular events among users of rosiglitazone compared with placebo [17]. After investigating these data, the FDA recommended a boxed warning on the drug’s label and required conduct of additional safety studies. In 2010, at the FDA’s direction, GlaxoSmithKline instituted a REMS, further restricting access to the drug by requiring prescribing physicians to enroll their patients in a special registry. By 2014, follow-up studies called into question the association between rosiglitazone use and increased numbers of major cardiovascular events. The most compelling was a randomized open-label controlled trial (Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycaemia in Diabetes [RECORD]) with a follow-up time of 5–7 years, designed to evaluate the drug’s safety in comparison with standard treatments for diabetes (metformin and sulfonylureas) [18]. During its initial data review, the FDA expressed reservations about potential bias in the ascertainment of cardiovascular events and requested that GlaxoSmithKline obtain an independent review of the results. This review was conducted by the Duke Clinical Research Institute, which also concluded that rosiglitazone use was not associated with excess cardiovascular risk [19]. Though GlaxoSmithKline and Duke’s findings remain controversial [20], they motivated the FDA’s decision to remove the boxed warning.

4 Discussion As these cases show, the impositions of boxed warnings for varenicline and rosiglitazone were both based on information from short-term retrospective observational studies. In the case of varenicline, more recent retrospective observational studies proved insufficient to trigger FDA action. By contrast, RECORD trial data helped convince the FDA to remove the boxed warning for rosiglitazone. The varenicline and rosiglitazone cases reveal that not all boxed warnings are the same—i.e., the warnings may be based on different levels of evidence or may have different degrees of certainty supporting them. These characteristics of boxed warnings have important ethical implications. First, the FDA’s role as a public health agency requires that

it act with the beneficence of the patient population in mind [21]. When a new risk signal has emerged that warrants a boxed warning, acting expeditiously to change the label and disseminate the information to patients can help support safe use of the product and contribute to patients’ autonomous decision making. As a result, boxed warnings may be added before all of the data have been collected or when outstanding questions about the reliability of the safety association remain. Yet the fact that some boxed warnings may ultimately be disproven and retracted can lead to consumer confusion and mistrust in the ability of the FDA to safeguard public health. To avoid frequent flipflopping, we believe it is ethically justified for the FDA to require a greater burden of proof, or a greater level of certainty provided by the evidence at hand, to justify removal of a boxed warning than it did to impose one. Others might invoke the principle of equity to argue that it is unfair and inequitable that a manufacturer should be expected to provide greater justification to get a boxed warning removed than was initially invoked to apply the warning. However, given the current and scant understanding of how removal of boxed warnings would impact physicians’ prescribing preferences and patients’ decision making in weighing the benefits and risks of taking a drug, one might argue that the FDA should impose distinct standards on the two regulatory actions, as having a higher standard for boxed warning removal is the more prudent public health course of action. One way to resolve this ethical tension would be to provide greater clarity about the source of data used to impose a boxed warning and any ongoing studies suggested by the FDA to follow up on the safety details. With greater information, patients and physicians will be able to make a more autonomous decision about how much to take the boxed warning into account. Another strategy would be to provide greater clarity about the standards used to impose and remove boxed warnings. What type of evidence should be required to remove a boxed warning? Randomized double-blind controlled trials are the so-called gold standard for causal inference (i.e., a drug caused the harm) because randomization of an adequately large sample enables equal distribution of measured and unmeasured variables between the intervention and control groups, eliminating the potential for confounding, and knowledge of treatment assignment by either the investigator or the subject could influence behavior and impact reporting of outcomes. However, double or single blinding may not always be possible or advantageous, because of the complexity of the number of treatments the subjects may be taking or the laboratory measurements required for disease management. For example, comparison of novel oral anticoagulation treatments with warfarin would require measurement of international normalized ratio (INR) values for all trial participants to blind both the investigator and the

J. S. Yeh et al.

subject. Furthermore, ‘‘sham’’ INR values would also have to be generated for those taking novel oral anticoagulants. Such considerations would make a randomized blinded trial much more complicated [22, 23]. Similarly, randomized controlled trials—blinded or not—may not always be feasible or ethical to implement. If the population taking the drug is small, the outcome is rare, or the outcome takes a long time to occur, tens of thousands of patients and years of follow-up time may be needed to detect the effect of a drug. Additionally, randomized controlled trials are dependent on voluntary participants, who may not be representative of the population. The requirement for removal of boxed warnings should therefore not be predicated on the availability of data from randomized controlled trials alone. Well-designed observational studies using data sources such as the Sentinel system—a drug safety post-marketing monitoring program in which information on drug adverse events is generated from large claims databases collected by governmental and non-governmental organizations [24, 25]—can also provide confirmation of drug safety questions. Use of these observational data facilitates more efficient assessment and surveillance of safety signals, which can indicate higher than expected adverse event rates with use of a drug in millions of patients, rather than the smaller subset of highly selective patients in randomized controlled trials. Furthermore, these observational data can better detect safety signals in a pragmatic setting in which the direct and indirect effects of the drug are captured, as opposed to the idealized setting of a randomized double-blind controlled trial. A prospective cohort study—one of the observational study designs—can be used to emulate a randomized controlled trial by following a group of individuals in the database (cohort) and observing whether they develop the outcome. Whether a prospective cohort study, another observational study design, or a randomized controlled trial should be conducted to answer the drug safety question should be predicated on the severity of the concern, the treatment context, the frequency of the outcome of interest, and the feasibility of ascertaining drug harm while minimizing risk to participants. Randomized controlled trials are advantageous when the relative effect is small, while observational studies may be preferred when the relative risks are moderate or severe, and when the outcome is rare. What matters most is that the studies described above be well-designed, whether they be observational studies or randomized controlled trials. Some markers of high-quality design for post-approval studies justifying boxed warning removal include registering with a publicly available central repository with transparent protocols, having pre-determined safety outcomes, and providing accessible summary data. Active comparators should be included when consistent with routine use, the studies should measure clinically meaningful outcomes or well-validated

surrogate markers, and there should be sufficient long-term follow-up. In the case of observational studies, they should incorporate robust statistical methods to limit potential confounding. If the FDA harbors any doubts regarding the validity or generalizability of the study findings, it should defer the decision to remove boxed warnings until better data are available, or it should request independent confirmation of the safety signal. If the original safety concerns are upheld, then the previously implemented riskreduction actions can be maintained or enhanced by further limiting access to the drug or withdrawing the drug from the market completely. If these concerns are invalidated, then the previously implemented regulatory action leading to imposition of boxed warnings can be reversed. Moreover, the process for FDA evaluation of data from well-designed observational studies and randomized controlled trials should be standardized. At a minimum, the FDA should convene a public hearing on the merits of boxed warning removal in the particular case on the basis of the information submitted by the manufacturer, ensuring that all stakeholders have an opportunity to voice their opinions. There should be sufficient input from experts without conflict of interest who have a broad range of expertise in the various study methodologies and evaluation of drug safety to aid the FDA’s decision. The rationale for the FDA’s ultimate decision—with clear identification of points of contention and agreement, too—should be publicly released. Such transparency will help ensure consistency in boxed warning removal determinations, provide greater clarity to manufacturers on how to make a case for boxed warning removal, and enhance public trust in whatever decision is reached. It is notable that after the convoluted process leading up to the removal of the rosiglitazone warning, prominent physicians remained skeptical about whether the right decision was made [26]. As boxed warnings are intended to communicate drug risks, there is a complementary need to assess how physicians and patients interpret and react to boxed warnings. Although a few studies have shown that imposition of boxed warnings can have a substantial impact on prescribing based on the drug or drug class, the prescribing context, and the perception of risks, boxed warnings can also be ignored or undervalued [27]. More compellingly, we could find no data on how boxed warning removal might impact physicians’ prescribing preferences and patients’ perception and use of prescription drugs. We suspect that boxed warning removal might re-accelerate physicians’ prescribing and patients’ use of the drug, although the rate of prescribing is likely to remain below the level prior to the imposition of the boxed warning, as some prescribers and patients may continue to harbor skepticism about the safety of the drug. If the concern for drug safety was unfounded, then this lingering skepticism

Communicating Drug Risks Via Boxed Warning–Ethical and Practical Considerations

may potentially reduce optimal treatment of a disease by discouraging use of an effective and safe drug.

Box 2. Boxed warning added to the official prescription drug label for varenicline (Chantix) in 2009

5 Conclusion The issue of boxed warning removal will likely arise more often in the future as more drugs are being approved on the basis of limited pre-marketing data [28–30] and as larger observational databases and other tools for conducting postapproval safety studies help generate new evidence suggesting that previously placed boxed warnings are not necessary [31]. Logically, as more boxed warnings are being imposed because of faster approvals of new drugs by the FDA, more initial labels with boxed warnings will need to be adjusted in the future as post-marketing studies evaluate the safety of drugs. The FDA’s recent reviews of the boxed warnings for varenicline and rosiglitazone suggest that there are higher standards required for the removal of boxed warnings than for their imposition. Reaching a consensus on a uniform and transparent process in guiding boxed warning revisions may help promote effective risk communication, autonomous patient decision making, and safe use of these products. Compliance with Ethical Standards Funding Aaron Kesselheim’s work is funded by a Greenwall Faculty Scholar program grant, the Harvard Program in Therapeutic Science, and the Laura and John Arnold Foundation. Conflict of interest James Yeh, Ameet Sarpatwari, and Aaron Kesselheim have no conflicts of interest that are directly relevant to the content of this study.

Box 1. US Food and Drug Administration (FDA) requirement for inclusion of a boxed warning and adverse events of prescription drugs [21 CFR 201.57(c)(1)]

Boxed warning. Certain contraindications or serious warnings, particularly those that may lead to death or serious injury, may be required by the FDA to be presented in a box. The boxed warning ordinarily must be based on clinical data, but serious animal toxicity may also be the basis of a boxed warning in the absence of clinical data. The box must contain, in uppercase letters, a heading inside the box that includes the word ‘‘WARNING’’ and conveys the general focus of the information in the box. The box must briefly explain the risk and refer to more detailed information in the ‘‘Contraindications’’ or ‘‘Warnings and Precautions’’ section, accompanied by the identifying number for the section or subsection containing the detailed information.

‘‘Serious neuropsychiatric events have been reported in patients taking CHANTIX. Advise patients and caregivers that the patient should stop taking CHANTIX and contact a healthcare provider immediately if agitation, hostility, depressed mood, or changes in behavior or thinking that are not typical for the patient are observed, or if the patient develops suicidal ideation or suicidal behavior while taking CHANTIX or shortly after discontinuing CHANTIX.’’

References 1. Ricci JR, Coulen C, Berger JE, Moore MC, McQueen A, Jan SA. Prescriber compliance with black box warnings in older adult patients. Am J Manag Care. 2009;15(11):e103–8. 2. Cheung A, Sacks D, Dewa CS, Pong J, Levitt A. Pediatric prescribing practices and the FDA black-box warning on antidepressants. J Dev Behav Pediatr. 2008;29(3):213–5. 3. Center for Biologics Evaluation and Research. Safety labeling changes—implementation of section 505(o)(4) of the FD&C Act. Rockville: Center for Biologics Evaluation and Research; 2013. http://www.fda.gov/downloads/drugs/guidancecompliance regulatoryinformation/guidances/ucm250783.pdf. Accessed 14 Jan 2016. 4. Marks NS, Weiss K. Boxed warnings and other FDA communication tools. Am Fam Physician. 2010;81(3):259. 5. Cheng CM, Guglielmo BJ, Maselli J, Auerbach AD. Coverage of FDA medication boxed warnings in commonly used drug information resources. Arch Intern Med. 2010;170(9):831–3. 6. Dorsey ER, Rabbani A, Gallagher SA, Conti RM, Alexander GC. Impact of FDA black box advisory on antipsychotic medication use. Arch Intern Med. 2010;170(1):96–103. 7. Lu CY, Zhang F, Lakoma MD, Madden JM, Rusinak D, Penfold RB, et al. Changes in antidepressant use by young people and suicidal behavior after FDA warnings and media coverage: quasiexperimental study. BMJ. 2014;348:g3596. 8. Friedman RA. Antidepressants’ black-box warning—10 years later. N Engl J Med. 2014;371(18):1666–8. 9. US Department of Health and Human Services. The health consequences of smoking—50 years of progress. A report of the Surgeon General: executive summary. Atlanta: US Department of Health and Human Services; 2014. http://www.surgeongeneral. gov/library/reports/50-years-of-progress/exec-summary.pdf. Accessed 7 Mar 2016. 10. US Food and Drug Administration. Chantix and Zyban to get boxed warning on serious mental health events. Silver Spring: US Food and Drug Administration; 2009. http://www.fda.gov/ ForConsumers/ConsumerUpdates/ucm170356.htm. Accessed 5 Feb 2014. 11. Gunnell D, Irvine D, Wise L, Davies C, Martin RM. Varenicline and suicidal behaviour: a cohort study based on data from the General Practice Research Database. BMJ. 2009;339:b3805. 12. Meyer TE, Taylor LG, Xie S, Graham DJ, Mosholder AD, Williams JR, et al. Neuropsychiatric events in varenicline and

J. S. Yeh et al.

13.

14.

15.

16. 17.

18.

19.

20.

nicotine replacement patch users in the Military Health System. Addiction. 2013;108(1):203–10. Pasternak B, Svanstrom H, Hviid A. Use of varenicline versus bupropion and risk of psychiatric adverse events. Addiction. 2013;108(7):1336–43. Thomas KH, Martin RM, Davies NM, Metcalfe C, Windmeijer F, Gunnell D. Smoking cessation treatment and risk of depression, suicide, and self harm in the Clinical Practice Research Datalink: prospective cohort study. BMJ. 2013;347:f5704. US Food and Drug Administration. October 16, 2014: Joint Meeting of the Psychopharmacologic Drugs Advisory Committee and the Drug Safety and Risk Management Advisory Committee Meeting Announcement. Silver Spring: US Food and Drug Administration; 2014. http://www.fda.gov/AdvisoryCommittees/ ucm394876.htm. Accessed 14 Jan 2016. GlaxoSmithKline. Annual report 2006. GlaxoSmithKline; 2006. http://www.gsk.com/media/279918/annual-report-2006.pdf. Nissen SE, Wolski K. Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes. N Engl J Med. 2007;356(24):2457–71. Home PD, Pocock SJ, Beck-Nielsen H, Curtis PS, Gomis R, Hanefeld M, et al. Rosiglitazone evaluated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes (RECORD): a multicentre, randomised, open-label trial. Lancet. 2009;373(9681):2125–35. Mahaffey KW, Hafley G, Dickerson S, Burns S, Tourt-Uhlig S, White J, et al. Results of a reevaluation of cardiovascular outcomes in the RECORD trial. Am Heart J. 2013;166(2):240–9 e1. Nissen SE. The hidden agenda behind the FDA’s new Avandia hearings. Jersey City: Forbes.com LLC; 2013. http://www.forbes. com/sites/matthewherper/2013/05/23/steven-nissen-the-hiddenagenda-behind-the-fdas-avandia-hearings. Accessed 30 Apr 2015.

21. Hamburg MA, Sharfstein JM. The FDA as a public health agency. N Engl J Med. 2009;360(24):2493–5. 22. Beyer-Westendorf J, Buller H. External and internal validity of open label or double-blind trials in oral anticoagulation: better, worse or just different? J Thromb Haemost. 2011;9(11):2153–8. 23. Manja V, Lakshminrusimha S. Epidemiology and clinical research design, part 1: study types. NeoReviews. 2014;15(12):e558–69. 24. Behrman RE, Benner JS, Brown JS, McClellan M, Woodcock J, Platt R. Developing the Sentinel system—a national resource for evidence development. N Engl J Med. 2011;364(6):498–9. 25. Psaty BM, Breckenridge AM. Mini-Sentinel and regulatory science—big data rendered fit and functional. N Engl J Med. 2014;370(23):2165–7. 26. Burton TM. FDA removes marketing limits on diabetes drug Avandia. New York: Wall Street Journal; 2013. http://www.wsj. com/articles/SB10001424052702304465604579220271002055140. Accessed 14 Jan 2016. 27. Lasser KE, Seger DL, Yu DT, Karson AS, Fiskio JM, Seger AC, et al. Adherence to black box warnings for prescription medications in outpatients. Arch Intern Med. 2006;166(3):338–44. 28. Darrow JJ, Avorn J, Kesselheim AS. New FDA breakthroughdrug category—implications for patients. N Engl J Med. 2014;370(13):1252–8. 29. Frank C, Himmelstein DU, Woolhandler S, Bor DH, Wolfe SM, Heymann O, et al. Era of faster FDA drug approval has also seen increased black-box warnings and market withdrawals. Health Aff. 2014;33(8):1453–9. 30. Wolfe SM. Does $760m a year of industry funding affect the FDA’s drug approval process? BMJ. 2014;349:g5012. 31. Misbin RI. The phantom of lactic acidosis due to metformin in patients with diabetes. Diabetes Care. 2004;27(7):1791–3.