Aspiring to Reasonableness in Accelerated Approval: Anticipating and Avoiding the Next Aducanumab

Abstract

The U.S. Food and Drug Administration’s (FDA) decisions about drug approval—though guided by science, as well as relevant statutes, regulations, and guidance documents—reflect normative judgments about how the agency should exercise its discretion. This is particularly true in the context of the “accelerated approval” pathway, where the agency must balance speeding to market drugs for patients with unmet needs before they have been proven to work and ensuring confidence about the benefits and risks of those drugs. A key challenge in evaluating normative judgments such as these is that reasonable people can disagree, rendering it difficult to proclaim with certainty that a particular decision is right or wrong. Therefore, we propose that it is preferable to ask whether a decision is reasonable. A decision is reasonable when it transparently, comprehensively, and fairly balances the interests of affected parties, within the parameters of the decisionmaker’s legal authority. If a decision achieves these three qualities, it can be viewed as legitimate and worthy of trust and confidence, regardless of whether one agrees with the particular outcome. We recommend that FDA adopt procedural protections to promote reasonableness in four domains affecting accelerated approval decisions: pathway gatekeeping, endpoint selection, stakeholder engagement, and deliberation. This should aid the agency in minimizing controversies, such as that surrounding the 2021 approval of aducanumab (Aduhelm; Biogen).

1. Introduction

Accelerated approval is a regulatory pathway, now three decades old, that allows the U.S. Food and Drug Administration (FDA) to grant a type of early marketing approval for drugs that fill an unmet medical need for serious conditions, contingent on required post-approval trials to confirm benefit.¹ Compared to traditional approval, which relies on either direct evidence of clinical benefit or validated surrogate endpoints, accelerated approvals rely on unvalidated surrogates (i.e., markers or measures deemed reasonably likely to predict clinical benefit but not yet known to do so) or intermediate clinical endpoints (i.e., measures of therapeutic effect that can be assessed earlier than an effect on irreversible morbidity or mortality (IMM) and considered reasonably likely to predict effect on IMM or other clinical benefit).² Given that these alternative endpoints can typically be studied more quickly or in smaller trials, accelerated approval represents an attractive bargain to many desperate patients.³ But its central tradeoff between speeding drugs to market and ensuring confidence about their benefits and risks remains contentious.

When making approval decisions, FDA’s task appears deceivingly straightforward: determine whether a drug is safe and effective for its intended use.⁴ That demands assessing the adequacy of evidence and then weighing evidence of risks and benefits. For accelerated approval, FDA faces an added and challenging task: determining whether weaker endpoints are adequate predictors of meaningful patient benefit.⁵ Science informs each of these tasks, while relevant statutes, regulations, and guidance documents provide constraints and guideposts. Yet substantial discretion inevitably remains. FDA’s final approval decisions therefore reflect a normative judgment about how to exercise this discretion. A key challenge in evaluating normative judgments is that reasonable people can disagree about them, rendering it difficult to definitively proclaim a decision right or wrong.

There are many proposals to improve accelerated approval, ranging from strengthening the speed and design of confirmatory trials to paying less for these drugs until benefit is proven. We endorse many of these suggestions. However, apart from calls to impose firm expiration dates on accelerated approvals,⁶ existing suggestions tend to overlook the discretionary nature of FDA’s judgments, thereby missing an important additional opportunity for change. Here, we propose a new approach to both improving and assessing FDA decision making around new drugs: holding FDA to a standard of reasonableness. This approach would help assure that even those who disagree with FDA’s decisions can nonetheless recognize them as legitimate. To be clear, achieving “upfront” reasonableness will not negate the importance of evaluating the downstream impact of FDA’s drug approval decisions, including on such outcomes as patient harm and costs borne by both patients and society more broadly. But especially for accelerated approval decisions, which are rooted in the inherent acceptance of some uncertainty, striving for reasonableness until more certainty can be gained will be an important step forward.

We use FDA’s June 2021 decision to grant accelerated approval to aducanumab (Aduhelm; Biogen) as a case study for our analysis. This is the first drug approved for treatment of Alzheimer’s disease (AD) since 2003, and its approval marks the first time FDA used the accelerated approval pathway for a drug directed at a common disease of aging. We identify features of unreasonableness in FDA’s decision, and explain why that decision sets a bad precedent for other diseases of aging, patient communities, and health care systems. We then propose a reasonableness framework with four domains—pathway gatekeeping, endpoint selection, stakeholder engagement, and deliberation—that should be incorporated into accelerated approval decisions going forward. A key ethical insight of our proposal is that procedural fairness along these domains of reasonableness provides legitimacy for difficult, highly discretionary, subjective decisions; as long as decision makers adopt reasonable procedures, their decisions should be regarded as reasonable, even if people disagree with them. Striving for reasonableness may allow us to retain a key advantage of the accelerated approval mechanism—early access to potentially beneficial therapies for devastating diseases—while addressing current shortcomings in its implementation.

2. Aducanumab and Beyond

The accelerated approval of aducanumab exemplifies the high stakes tradeoff between speeding drugs to market and ensuring confidence about their benefits and risks, as well as the difficulty of evaluating normative judgments. In the wake of aducanumab’s approval, FDA officials emphasized that many people with AD and their families “are willing to accept the trade-off of some uncertainty about clinical benefit in exchange for earlier access to a potentially effective drug.”⁷ Meanwhile, numerous clinicians and regulatory experts decried aducanumab’s approval as “the worst drug approval decision in recent U.S. history,” given scant evidence of clinical benefit, safety concerns, and fear of setting a dangerous regulatory precedent in AD and beyond.⁸

Unfortunately, many of the problems with FDA’s aducanumab decision—from concerns about underlying data to worries about resulting costs to healthcare systems—are not unique. Because aducanumab has magnified attention to growing concerns about how FDA should balance evidence and access, it offers a critical lens to acknowledge and address broad challenges inherent to the accelerated approval pathway.

2.1. Aducanumab’s Path to Accelerated Approval

On March 21, 2019, Biogen announced the discontinuation of a pair of Phase III clinical trials, ENGAGE and EMERGE, after a futility analysis indicated the studies were unlikely to meet their primary objective—that is, to evaluate the effectiveness of aducanumab on a clinical endpoint, slowing cognitive impairment as measured by changes in the Clinical Dementia Rating-Sum of Boxes (CDR-SB) score.^9,10 However, in October 2019, Biogen reversed course, announcing that, after reviewing a larger dataset and conducting additional analyses (not published in the peer-reviewed literature until March 2022),¹¹ it planned to pursue FDA approval.^12,13

FDA called a meeting of its Peripheral and Central Nervous System Advisory Committee, a group of external experts, in November 2020. At the meeting, Dr. Billy Dunn, Director of FDA’s Office of Neuroscience, strongly advocated for aducanumab in disregard of an FDA statistician’s concerns, as well as concerns raised by advisory committee members.¹⁴ In light of his behavior, which included answering questions directed to the Biogen executive presenting the company’s data, Dunn was described as “a cheerleader for the company.”¹⁵ After concluding that Biogen had failed to prove effectiveness in treating AD, members of the advisory committee, with a single abstention, unanimously voted against aducanumab’s approval. Notably, the advisory committee did not discuss the possibility of accelerated approval, having been explicitly told by Dunn that FDA was not considering aducanumab’s demonstrated reduction in amyloid-β plaques.⁸ Although amyloid-β is a recognized biomarker of AD, there is fierce scientific debate over whether amyloid-β causes AD¹⁶ and whether clearing these plaques will meaningfully impact outcomes for people living with AD. In other words, reduction in amyloid-β plaques is an unvalidated surrogate endpoint about which there is substantial controversy over its likelihood of reasonably predicting clinical benefit, the accelerated approval standard.

Following the advisory committee meeting, the Alzheimer’s Association, a patient advocacy organization that receives some financial support from industry, organized a January 2021 discussion between senior FDA officials and patients, families, and Association representatives, based on concerns that “the voices of those diagnosed with Alzheimer’s and their caregivers had not truly been heard during the November 2020 advisory committee meeting.”¹⁷ Notably, FDA has not provided an agenda, list of attendees, transcript. or summary report of this meeting. However, it has been reported that the “session included testimonials from six patients and two caregivers” and that the group was asked about “the amount of risk they could tolerate when taking a new treatment, whether they would take a treatment with limited benefits … , and the value they placed on a treatment that could slow disease progression and extend their lives.”¹⁷

In June 2021, without ever returning to the advisory committee, FDA announced the accelerated approval of aducanumab based on the drug’s reduction of amyloid-β plaques.^18,19 The decision was made even though aducanumab had been tested against clinical endpoints, but failed to meet them in pre-specified analyses—particularly slowing cognitive impairment as measured by the CDR-SB. The drug also poses substantial risks of brain bleeds and swelling referred to as amyloid-related imaging abnormalities (ARIA).^20,21 European and Japanese regulators have rejected aducanumab outright and, in April 2022, Biogen pulled its application for marketing approval in Europe.^22,23

Following the approval, three members of the FDA advisory committee resigned in protest,²⁴ while senior FDA officials publicly defended the agency’s decision.^7,25–27 These officials conceded that evidence from the clinical trials “did not meet the standard for ‘regular’ approval” but argued that “accelerated approval was designed for situations in which there is residual uncertainty about clinical benefit.”²⁵

Many aspects of FDA’s decision have subsequently been criticized, leading to some changes already. First, the initial FDA-approved label stated that aducanumab was for “the treatment of Alzheimer’s disease,” although the trials had included only patients with (1) mild cognitive impairment (MCI) or mild dementia and (2) elevated levels of amyloid-β. Even champions of aducanumab deemed this problematic.²⁸ FDA quickly re-issued the label, specifying that the drug should be prescribed only for MCI or mild dementia.²⁹ However, the label still does not recommend testing for amyloid-β, nor does it recommend APOE testing, even though a patient’s APOE genotype influences risk for ARIA.³⁰

Second, FDA did not require agreement on confirmatory trial design as a condition of aducanumab’s approval and set a 9-year deadline for reporting confirmatory trial results.³¹ Facing pressure over the leniency of these conditions, Biogen announced in December 2021 that it planned to complete its confirmatory trial by 2026³² and submitted the protocol to FDA in March 2022.³³

Third, the poor evidence supporting aducanumab’s approval—coupled with a $56,000 per year price tag—generated substantial pushback. Various health systems and payers indicated they will not offer or cover aducanumab.^34,35 In November 2021, Medicare announced it was hiking “Part B” premiums to cover anticipated costs, with implications for all Medicare beneficiaries, not only those who might receive aducanumab.^36,37 In December 2021, with coverage on the line, Biogen announced that it would reduce aducanumab’s price by roughly half.³⁸ Yet, the new price remains well in excess of the $3,000 to $8,400 per year that the Institute for Clinical and Economic Review (ICER) concluded would render the drug cost-effective.³⁹

Perhaps the most important development since aducanumab’s accelerated approval has been the Centers for Medicare and Medicaid Services (CMS) decision to rely on the rarely utilized lever of “coverage with evidence development” (CED)⁴⁰ as part of a national coverage determination (NCD).⁴¹ In April 2022, after receiving more than 10,000 public comments, CMS announced that monoclonal antibodies directed against amyloid for the treatment of AD— including aducanumab and other drugs in this class that have not yet received FDA approval—are nationally non-covered unless provided within an FDA-approved randomized controlled trial (RCT), CMS approved study, or study supported by the National Institutes of Health (NIH).⁴² Signaling its concern with the weakness of the available data supporting aducanumab, this decision represents the first time CMS has imposed an RCT requirement (rather than an observational study or registry) for coverage of a drug through CED; it also limited coverage to patients with confirmed presence of amyloid, a tighter requirement than aducanumab’s labeled indication.

Importantly, the CMS decision distinguishes between drugs in this class that receive accelerated approval and those that receive traditional approval. The former will be covered exclusively in an FDA-approved RCT, whereas the latter may be covered in CMS-approved prospective comparative studies, including registries. According to CMS administrators, this was “an attempt to provide … important limits on Aduhelm’s coverage, while not necessarily consigning future anti-amyloid monoclonal antibody drugs to similar restrictions.”⁴³ CMS’s Chief Medical Officer clarified, however, that the decision was specific to this drug class and “should not be viewed as setting a new direction on therapies that receive FDA accelerated approval.”⁴⁴

Critics have highlighted numerous additional concerns about FDA’s decision, including conflicts of interest (COI) and regulatory capture. For instance, the Alzheimer’s Association submitted an October 2020 letter supporting aducanumab to FDA, without disclosing that it received between $250,000 and $499,999 in financial support from both Biogen and its partner, Eisai, in FY20.^45,46 It was also revealed that private, off-the-record meetings took place between Dunn and Biogen in spring 2019, seemingly in violation of FDA protocol.⁴⁷ By June 2019, contrary to statements made to the advisory committee, officials in the FDA Office of Neuroscience had proposed accelerated approval as a means of “saving” aducanumab from failed trials. According to reporting, “[t]he move stunned even Biogen’s top executives, who had considered that out of the question for a host of reasons, including the fact that the FDA had never used the pathway for an Alzheimer’s treatment.”⁴⁷ These issues, and others, led to a Congressional investigation (ongoing as of April 2022), as well as an overarching assessment of how FDA implements the accelerated approval pathway led by the U.S. Department of Health and Human Services (HHS) Office of the Inspector General (OIG).^48,49

2.2. Problems and Precedents

Accelerated approval is meant to help desperate patients, but its effects are not limited to those who ultimately use the products that come to market under this pathway. Aducanumab’s accelerated approval sets a dangerous regulatory precedent, with potential negative consequences for similar drugs targeting diseases of aging, patients with other serious illnesses, and the U.S. health care system more broadly.⁵⁰

Unvalidated surrogate endpoints inherently carry uncertainty, but the aducanumab decision risks stretching reliance on these endpoints beyond sound scientific practice in future approvals. Commonly used surrogate measures in oncology, where accelerated approval is most common, include response rate, time to tumor progression, and disease-free survival. Compared to clinical endpoints, these surrogates may be more prone to measurement bias or poorly associated with meaningful patient benefit.⁵¹ Yet problems with the “amyloid hypothesis”—that is, “the assumption that accumulation of the peptide amyloid-β is the main cause” of AD—run far deeper than criticism of typical oncology surrogates. Perhaps the biggest challenge to the amyloid hypothesis is that multiple Phase III trials of anti-amyloid therapies have failed; further, researchers note that the number of amyloid deposits in the brain is not well correlated with patients’ degree of cognitive impairment.¹⁹ Although the amyloid hypothesis dominated the field for years, it has more recently been described as thwarting scientific progress.^19,52

In approving aducanumab, FDA gave the amyloid hypothesis new life. This will likely reverberate across AD clinical practice and research far into the future, as companies predictably refocus their efforts on reduction of amyloid-β plaques over other endpoints that may be more promising. Indeed, sponsors are already seizing this opportunity. In late 2021, Biogen and Eisai announced they were pursuing accelerated approval for lecanemab, which works in a similar manner to aducanumab.⁵³ Eli Lilly announced plans to ask FDA to grant accelerated approval to donanemab based on the drug’s ability to clear amyloid-β plaques,⁵⁴ though the company pushed back the timeline for submitting its application after CMS indicated its plans for the NCD.⁵⁵ Shifting the research terrain back to the amyloid hypothesis could negatively affect millions of AD patients and their families hoping for disease-modifying therapies, although the stringent terms of the Medicare CED decision should help disincentivize that approach. However, FDA’s novel use of accelerated approval for AD drugs has the potential to extend reliance on unvalidated surrogates into entirely new and expansive therapeutic areas, including other diseases of the aging body, such as frailty (sarcopenia).

The approval of aducanumab has also attracted the interest (and frustration) of other patient groups with similarly serious, unmet medical needs. People with amyotrophic lateral sclerosis (ALS), for example, have urged FDA to use its regulatory flexibility to approve drugs quickly rather than demanding more evidence in the face of uncertainty. The ALS community has expressed confusion and anger as to why aducanumab was seemingly ushered through approval on the basis of patient desperation while drugs with equal or greater promise for ALS were initially pushed off or not as well supported by FDA at advisory committee meetings.⁵⁶ Some in the rare disease community worry that backlash following aducanumab’s approval will lead FDA to overcorrect, leading to more conservative decisions going forward that may inhibit access to promising drugs for their diseases.^57,58

FDA’s decision might also substantially expand the financial impact of the accelerated approval pathway, which has long been a source of concern. More than 6 million Americans currently live with AD, and that number is expected to reach nearly 13 million by 2050.⁵⁹ While drug costs are not within FDA’s jurisdiction and have no legal role in its approval decisions, they are a source of anxiety for patients, public and private insurers, and taxpayers.^60,61 Before the price reduction, it was estimated that aducanumab would generate hundreds of billions of dollars in health care spending.⁶² CMS’s decision to limit coverage of aducanumab to RCTs was based on lack of evidence to demonstrate the drug is reasonable and necessary for the Medicare population, not on the basis of price or cost-effectiveness, but it will have the salutary incidental effect of dramatically limiting the immediate financial impact of aducanumab’s price on the health care system. Nonetheless, taxpayers could be on the hook to cover the cost of the drug in the trial Biogen is required to conduct as a condition of accelerated approval, other payers may decide (or be required) to cover the drug more expansively, and patients may also decide to pay out of pocket. If FDA’s decision signals the acceptance of ever weaker approaches to surrogate endpoints—opening the door to other expensive drugs with limited evidence of efficacy—fewer research and health care resources may be available for other critical societal needs, unless payers push back more than they have traditionally. Unfortunately, CMS has taken pains to emphasize that its aducanumab decision is exceptional and not the new normal.

3. If Not Right, Then Reasonable

FDA’s aducanumab decision highlights the many ways accelerated approval can lead down a troubling road, although we do not suggest that the pathway should be abandoned altogether. To the contrary, we accept the foundational premise undergirding accelerated approval, namely that it is sometimes appropriate to prioritize speed over certainty for seriously ill patients without other treatment options. Yet specifying the proper balance between speed and certainty does not lend itself to assessment on a concrete, objective scale of either right or wrong. Instead, we must take a more qualitative approach, asking whether FDA’s normative judgments are reasonable. Notably, the same is true beyond accelerated approval, when FDA is making its typical discretionary judgments about the balance between a drug’s risks and benefits. A decision is reasonable when it transparently, comprehensively, and fairly balances the interests of affected parties and, as a result, can be viewed as legitimate and worthy of trust and confidence, regardless of whether one agrees with the outcome.

Reasonableness standards are common and serve an important purpose in both law and ethics. For example, when courts decide negligence cases, they ask whether the defendant behaved reasonably given the circumstances; even if greater care could have avoided the harm in question, the standard is not perfection. When considering approaches to prioritizing access to scarce resources, ethicists ask whether the selected policy was reasonable from the public’s perspective, recognizing that there may have been other acceptable ways to address competing considerations. In these cases, what counts as reasonable will depend heavily on procedural protections around the decision-making process, such as giving stakeholders voice (i.e., a chance to express concerns), sharing information and reasons transparently, avoiding COIs, and promoting consistency. The process by which the decision is reached is what makes it reasonable. In public policy, where there is often no single approach on which all stakeholders agree, adherence to procedural protections is often the best we can do to assure decisions are fair, trustworthy, and legitimate.

Given its extensive discretion, FDA can draw the line between speed and certainty in several different places – but the way it draws these lines can be more or less reasonable. To promote the former and minimize the latter, we enumerate four aspects of FDA’s accelerated approval decision-making process where procedures should be formalized to improve reasonableness. First is pathway gatekeeping reasonableness, which refers to FDA’s decision to allow the accelerated approval pathway to be used for a given drug. Second is endpoint reasonableness, which concerns the level of evidentiary support for unvalidated surrogate endpoints. Third is stakeholder engagement reasonableness, which refers to the practices by which FDA comes to understand the values and preferences of stakeholders within the scope of its legal authority who have an interest in the approval decision. Fourth, deliberative reasonableness encompasses how FDA weighs and explains the various considerations to reach its final decision. In what follows, we situate both existing and new proposals to improve accelerated approval—and the agency’s normative judgments more broadly—across this novel reasonableness framework. Although no single action across these four domains will suffice to render a decision reasonable per se, each can increase the degree of reasonableness, and thus legitimacy.

3.1. Pathway Gatekeeping Reasonableness

As gatekeeper of the accelerated approval pathway, FDA should ensure that drug companies access this pathway via appropriate “on-ramps” and are directed to “off-ramps” as necessary. Lack of upstream mechanisms that prevent unreasonable access to accelerated approval can result in serious downstream consequences.

Furthest upstream, sponsors should be expected to declare to FDA their intention to pursue accelerated approval at the point of designing pivotal trials. This will avoid circumstances in which accelerated approval is used to “rescue” drugs—like aducanumab—that are tested against, but fail to meet, the clinical endpoints needed for traditional approval. This could be a one-way declaration, in the sense that companies who anticipate pursuing accelerated approval could later switch to seeking standard approval, if desired, but would be barred from pursuing accelerated approval if they failed to make advance arrangements with FDA. This approach would allow early discussion of appropriate endpoints, as well as pre-planning for post-approval confirmatory trials. FDA guidance already encourages “sponsors to communicate with the Agency early in development concerning the potential eligibility of a drug for accelerated approval, proposed surrogate endpoints or intermediate clinical endpoints, clinical trial designs, and planning and conduct of confirmatory trials.”⁶³ What we propose is to solidify this guidance into a more formal legal expectation—one that would need to include protection of FDA’s discretion to reject a company’s declared intention to pursue accelerated approval if traditional clinical endpoints would be more appropriate.

There are also several gatekeeping protections that should be adopted at the point of considering a company’s New Drug Application (NDA). At the very least, if no accelerated approval declaration is required prior to initiating pivotal trials, one should be expected by the time of any FDA advisory committee meeting at which the drug will be discussed, in order to allow these experts to weigh in on selected surrogates and related factors. Additionally, to avoid delay in confirmatory trials, FDA should not grant accelerated approval until the design of required post-approval trials is settled and preparations are in place to begin them immediately upon approval, if not earlier.^64,65 Current FDA guidance points in this direction, calling for confirmatory trials to be underway at the time a marketing application is submitted, or at least agreement on trial design before approval,⁶³ but this was not followed for aducanumab. Along these lines, FDA’s FY23 budget request seeks legislative authority to require as a condition of accelerated approval that sponsors first demonstrate their confirmatory study is adequately designed and can be completed in a timely manner,⁶⁶ with proposed legislation already introduced along the same lines.⁶

Further downstream, after accelerated approval, FDA’s gatekeeping obligations remain substantial. At this stage, relevant procedural protections should include clear milestones for completing confirmatory trials and more automatic processes for withdrawal of approval if requirements are not satisfied. Designing post-approval trials in the pre-approval period should entail setting forth the specific outcomes, including remaining uncertainty due to inconclusive results, that will result in either loss of approval with the need for further study or transition to full approval.^67,68 Minimizing FDA’s discretion may also be desirable, for example, by subjecting accelerated approvals to an automatic sunset date specified by Congress. Each of these improvements to help advance reasonableness in FDA’s accelerated approval decisions are part of recent legislative proposals, including frequent progress reports to FDA on the status of confirmatory trials and accelerated approval expiration dates no later than one year after the target trial completion date or five years after approval.⁶

3.2. Endpoint Reasonableness

A second type of reasonableness centers on the selection of appropriate endpoints for accelerated approval. Unvalidated surrogate or intermediate clinical endpoints necessarily entail uncertainty, but they can enjoy stronger or weaker evidentiary support. Though we acknowledge some conceptual overlap between endpoint reasonableness and other domains of reasonableness discussed below, we suggest that endpoint reasonableness deserves its own category due to the centrality of these data to accelerated approval decisions.

Grading systems have been proposed to assess the robustness of and build consensus for endpoints used in the accelerated approval pathway.⁶⁹ Grading a surrogate endpoint prior to conducting a trial would allow FDA and independent advisors to better evaluate trial results and might also improve the likelihood that accelerated approval would convert to full approval based on clinical criteria.^51,70

Further, prior to endorsing a new surrogate endpoint for accelerated approval, FDA could publish a preliminary justification of its decision, grade the proposed surrogate, and invite public comments.⁶⁸ A notice-and-comment approach could help approximate scientific peer review of which surrogates should be accepted for which disease areas (i.e., disease-specific, not drug specific). If this model were adopted, FDA could also be expected to engage with comments and explain why it is either following them or taking a different approach. To reduce delay, considering that time is of the essence in accelerated approval contexts, FDA could engage in this process first via its usual approach to issuing guidance and then solidify its decision via the formal rulemaking approach required for new regulations.

Another aspect of endpoint reasonableness could include holding companies to tighter standards in both the design and completion of confirmatory trials the more controversial the surrogate relied upon for accelerated approval. Moreover, reasonableness calls for special scrutiny where surrogate endpoints have been tested alongside clinical endpoints and the results fail to align. In these cases, there should be a default presumption against reliance on that surrogate, which may be overcome only by a strong and specific reason to the contrary.

3.3. Stakeholder Engagement Reasonableness

Drug approval decisions are judgment calls that affect many different stakeholders. This requires FDA to consider their various—and potentially divergent—perspectives and interests. Patients have the greatest stake in drug approval decisions, and at Congress’s direction,⁷¹ FDA has been actively working over the past several years to engage patients and “gain insights that might inform the risk-benefit assessments they make in evaluating new therapies.”⁷² But the voices of others, including caregivers, clinical experts, payers, and all members of society that stand to be impacted by FDA approval decisions are also relevant. Reasonableness demands that FDA balance these competing interests in deciding what evidence should suffice for accelerated approval, what benefits are enough, and what risks are too high.

Even within a single disease, patient opinions about FDA decisions are often varied, making it essential to avoid generalizations about patient preferences.⁷³ FDA must seek out and listen to a diversity of patients rather than assuming that those who submit comments, speak at public meetings, or serve in advisory and advocacy roles are representative. Patient surveys, interviews, and focus groups can be helpful to understand perspectives on contentious approval decisions and critical tradeoffs between risk, uncertainty, and speed inherent to accelerated approval.

Additionally, reasonable FDA judgments will need to look beyond the needs of current patients to also consider those of future patients, even if no one is formally speaking on their behalf.⁷⁴ While patients today may understandably prioritize speed, patients tomorrow (i.e., those who are not yet sick or may not yet even be born) have an interest in avoiding being forced into the same tradeoffs due to lack of progress—a challenge recognized in the earliest days of accelerated approval by some AIDS advocates who worried that they were giving up too much in favor of early access.³ The preferences of future patients, if they were in a position to articulate them, would likely lean towards encouraging more certainty about benefit and safety.

FDA’s judgments also need to consider the needs of the public more broadly, within the constraints of the agency’s authority, as there are distinct rationales for patient and public involvement.⁷² Members of the public are affected by FDA’s approval decisions, even for drugs they will never take. This is most obvious with regard to the impact of approval decisions on health care costs, but those considerations are outside the scope of FDA’s remit. Beyond these, the public is affected by weakened evidentiary standards for FDA approval to the extent they establish informal precedents and reduce incentives for companies to produce the information we all need to make important treatment decisions. Additionally, when FDA makes unreasonable decisions, the resulting reputational harm can undermine public trust in the agency⁷⁵ and compound blows from, for example, advocates of the Right to Try movement, who seek to weaken FDA oversight.⁷⁶

Public trust can also be impacted if FDA decisions are influenced by stakeholders bearing COIs.⁷⁷ Relevant procedural protections to promote reasonable stakeholder engagement should therefore include efforts to discern the potential influence of industry on messages delivered by patients or advocacy groups;^78,79 to require disclosure and management of financial interests and relationships; and to avoid favoring select patient advocacy organizations with privileged access or outsized influence in FDA’s decision making.

Another important procedural protection for reasonably engaging stakeholders, from companies to patients, is that all meetings with stakeholders should be on-the-record, formal, and documented through minutes and other records that could be made available to oversight bodies such as Congress and the HHS OIG, as needed.

3.4. Deliberative Reasonableness

Finally, FDA should be reasonable in its deliberations about and justifications for accelerated approval. Already, FDA and other regulatory agencies are legally subject to an “arbitrary or capricious” standard, which precludes them from willfully and unreasonably acting without consideration for or in disregard of facts or law. However, this is a low bar that should be further specified in a number of ways.

One is to consider how FDA weighs the perspectives of its career scientists and statisticians, who engage deeply with trial data submitted as part of an application for approval. If those employees have reservations about approval that are overruled by senior leaders, reasonableness demands that the justifications for doing so be clearly, transparently, and publicly articulated.

Similarly, although advisory committee recommendations should not be viewed as binding on FDA, since that would inappropriately delegate important government decisions to those outside the agency, advisory committees should be given substantial deference to the extent that they offer expert peer review. This is especially true when advisory committees are unanimous in their recommendations. As above, a decision to ignore such a recommendation should be clearly, transparently, and publicly justified. In the same vein, a decision not to return to an advisory committee with new material information, such as a change in anticipated approval pathway, should be treated as a contravention of default expectations, the reasons for which should be clearly stated.

Overall, FDA officials should cultivate neutrality, approaching approval decisions by asking “is this product safe and effective for patients?” rather than “how can we get this product approved?” This approach conceptualizes FDA in the role of a judge neutrally adjudicating a case, tasked with considering the strongest arguments both for and against approval. Companies are well positioned to advance the case for approval, but at present, no one is formally tasked with presenting the latter case. We would not necessarily propose an adversarial process, similar to litigation between opposing parties; yet, it could be useful to have a better mechanism to ensure that all angles of an approval decision are fully considered. For example, this might be a role explicitly given to external advisory committees, or an official within FDA could be tasked with making sure that counterarguments to approval are being given adequate consideration.

Deliberative reasonableness can also be promoted through efforts to improve consistency in FDA’s decisions. It has been noted that “FDA has no mechanism to find or tradition to cite similar cases when weighing evidence for approvals, resulting in standalone, bespoke decisions.”⁸⁰ Yet especially when making normative judgments rooted in highly discretionary standards, as is the case for accelerated approval, precedent is valuable both for the guidance it provides, the confidence it engenders, and the legal support it provides. The more FDA can ground its accelerated approval decisions in clear, strong reasons based on similar prior decisions, the more reasonable those decisions will be. The same is true with regard to giving reasons why a current case should be distinguished from prior decisions. Although there is some concern about entrenching problematic precedent, especially if the precedent resembles the aducanumab decision, if the procedural protections described herein are followed, those precedents should at least be reasonable.

4. Conclusion

It is essential to remember that FDA’s job is not to approve drugs. Rather, it is to make the best decisions on drug applications, which sometimes means rejecting them or calling for more evidence. What counts as the “best” decision, however, is a matter of normative judgment, meaning that reasonable people are likely to disagree. In these cases, which are particularly pronounced for accelerated approval—where standards are especially discretionary and stakes are high—we can only demand that FDA make reasonable decisions.

To do that, we recommend that FDA adopt procedural protections to promote reasonableness in pathway gatekeeping, endpoint selection, stakeholder engagement, and deliberation (Figure 1). These domains of reasonableness help identify the ways in which FDA’s decision to grant accelerated approval to aducanumab was not necessarily objectively wrong, but was clearly unreasonable. Measured against the proposals outlined herein, the accelerated approval of aducanumab falls short. Had FDA more closely hewn to these procedural safeguards, it is possible that it would have nonetheless reached the same outcome, but that outcome would be more legitimate and may have been able to avert deep reputational harms to FDA.

Figure 1.

Overview of the Four Domains of Reasonableness. A decision is reasonable when it can be viewed as legitimate and worthy of trust and confidence, regardless of agreement with the outcome.

Key Points.

• The U.S. Food and Drug Administration (FDA) enjoys significant discretion when making approval decisions.

• For FDA’s decisions to be viewed by stakeholders as legitimate and worthy of trust, FDA should adopt procedural protections, especially in the context of accelerated approval.