Aduhelm, the Newly Approved Medication for Alzheimer Disease: What Epidemiologists Can Learn and What Epidemiology Can Offer

Abstract

Alzheimer disease (AD) is a progressive disorder common among older adults and culminating in profound cognitive impairments and high mortality risk. The US Food and Drug Administration (FDA) recently provided accelerated approval for Aduhelm, a medication for AD treatment. Aduhelm (Biogen Inc., Cambridge, Massachusetts) has been described as the first disease-modifying treatment for AD but has not been demonstrated to improve patients’ cognitive or functional outcomes. In this commentary, we describe why Aduhelm approval was controversial and aspects of the current evidence of special pertinence to epidemiologists. The FDA decision was based primarily on 2 randomized controlled trials (RCTs), both terminated early, with conflicting findings about the cognitive benefits of Aduhelm. Both RCTs showed important adverse effects of the medication. The FDA cited the documented reduction in brain amyloid, an AD biomarker hypothesized as a surrogate outcome, to justify accelerated approval. Despite lack of racial/ethnic diversity in the RCT participants, concerns about health disparities have been invoked to argue for public funding of this expensive medication. The Centers for Medicare and Medicaid Services recently made a “Coverage with Evidence Development” determination for Aduhelm and similar medications. We end by describing how innovative study designs could accelerate postapproval research and evaluate the proposed surrogate outcomes.

Abbreviation

AD

Alzheimer disease

CMS

Centers for Medicare and Medicaid Services

FDA

Food and Drug Administration

RCT

randomized controlled trial

Editor ’s note: The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the American Journal of Epidemiology.

About a third of older adults in the United States will develop dementia in their lifetimes (1), experiencing progressive and profound memory and cognitive losses. Alzheimer disease (AD) is considered the most common cause of the syndrome of dementia, and there are no medications established to slow its progression. In June 2021, the US Food and Drug Administration (FDA) provided accelerated approval for Aduhelm (aducanumab; Biogen Inc., Cambridge, Massachusetts), a new medication to treat AD (2). The FDA’s decision sparked controversy because evidence that Aduhelm improves cognitive or functional outcomes is not convincing. Aduhelm was touted as the first “disease-modifying” drug ever approved for AD, in contrast to drugs that aim to relieve symptoms. The FDA’s approval letter: 1) cited the effect of Aduhelm on reducing brain amyloid, an AD biomarker, arguing that such biomarker changes were reasonably likely to predict clinical benefits; and 2) mandated a new randomized clinical trial (RCT) to be reported by 2030 verifying the efficacy of Aduhelm for clinical outcomes. Biogen initially priced Aduhelm at $56,000 annually (3) and subsequently revised the price to $28,200; additional costs will be incurred to evaluate eligibility, administer the medication, and monitor safety. In this commentary, we outline features of the controversy, aspects of the current evidence, and future research opportunities to efficiently deliver convincing evidence.

ADUHELM RCTS SHOW UNCLEAR CLINICAL BENEFIT BUT CLEAR ADVERSE EFFECTS

Although both available Aduhelm phase 3 RCTs convincingly showed reduced brain amyloid, they were terminated early due to a futility analysis indicating that substantial cognitive benefit was unlikely to be established. When more data accumulated after study termination, data from one trial showed a modest slowing of cognitive decline among treated participants; the other identically designed trial showed no cognitive benefit (4). Although analyses for such RCTs are typically prespecified to avoid inflation of type-1 errors due to multiple comparisons, trial termination creates ambiguity about when to freeze the data for analysis. Conventionally, FDA approval requires 2 RCTs demonstrating efficacy (5). Unfortunately, evaluation of successful RCTs need not consider other, potentially unsuccessful RCTs of the same medications. Failed RCTs of similar medications are also not necessarily considered, increasing the likelihood of cherry-picking trials that are “successful” by chance. The 2 major phase 3 Aduhelm trials followed many unsuccessful trials of medications targeting similar biological mechanisms (6).

Aduhelm caused brain bleeding (microhemorrhages) or swelling (edema) in about 35% of treated patients (7). The high rate of adverse events occurred although trial participants were selected to be at low risk for such events, which may thus be even more common in general clinical populations. Although side effects typically appeared to resolve when medication was stopped, medication cessation reveals treatment assignment to participants. Such unblinding may bias trial results if outcomes are subjectively reported. The Aduhelm trials specified primary outcomes including subjective caregiver assessments of patient functioning, such as the Clinical Dementia Rating Scale—Sum of Boxes (CDR-SB).

In the United States, dementia disproportionately affects Black/African-American older adults and some Latino populations (1, 8, 9). We consider racial/ethnic identity as social conventions, adopted to reflect and maintain race-based social privileges. Lifelong exposures to these inequalities are manifest in people’s health and may result in racial/ethnic differences in the safety and efficacy of medications. Non-White individuals were extremely underrepresented in Aduhelm trials as in much of the research motivating the trials (10): the 2 phase 3 RCTs randomized only 6 Black participants to the high-dose treatment approved by the FDA (11). Given elevated stroke risk among Black compared with White people in the United States, the severe adverse events observed in the Aduhelm RCTs might be even more common and more harmful in Black adults (12). Although statistical methods for extending RCT evidence to other populations are promising, no statistical wizardry can overcome n = 6.

THE SURROGATE OUTCOME USED TO JUSTIFY ADUHELM APPROVAL IS NOT VALIDATED

To date, AD drug development has focused primarily on removal of brain amyloid. Brain amyloid is a biomarker that defines Alzheimer pathological change in the dominant, but controversial (13, 14), disease framework (15). The Aduhelm approval was premised on the assumption that reducing brain amyloid predicts clinical benefit. Already, 2 additional medications have been placed in the pipeline for FDA approval based on established impact on amyloid, despite ambiguous evidence for clinical benefit. If amyloid reduction is a valid surrogate for clinical benefit, it could expedite AD research. If it is not, we risk approving clinically useless medications.

No study has demonstrated that amyloid reduction leads to clinical improvements. Ackley et al. (6) evaluated 14 RCTs (40 randomization arms) of medications hypothesized to remove amyloid, applying an instrumental variable (IV) analysis to account for variation in amyloid reduction achieved in each trial arm. This analysis found no evidence that amyloid reduction improved cognition, and the meta-analysis provided narrow confidence intervals around the null (6). This backdrop of repeatedly null findings for amyloid-targeting therapies raises the possibility that the single successful Aduhelm trial was a chance finding and calls into question amyloid reduction as a valid surrogate (16).

Defenders of the importance of amyloid argue that amyloid has many different biological forms (fibrils, monomers, oligomers), that different drugs target different amyloid forms (17), and that past medications may have failed to provide cognitive benefit because they targeted the wrong form of amyloid. Although not the most straightforward interpretation of the evidence, this plausible critique amounts to a consistency violation in the measure of amyloid (18).

Secondary analyses of donanemab, a drug hypothesized to operate via the same mechanism as Aduhelm, also call into question the relevance of amyloid reduction as a surrogate. The largest donanemab RCT randomized 257 individuals to placebo or donanemab. In the donanemab arm, brain amyloid was dramatically lower, and the Integrated Alzheimer’s Disease Rating Scale (iADRS) (the prespecified primary outcome) was 0.25 standard deviations (95% confidence interval: 0.01, 0.48; P = 0.04) better. The overall result was encouraging despite many caveats. However, the extent of reduction in brain amyloid—the putative mechanism of action—was not associated with cognitive outcomes (19, 20). Although not formal mediation models, if analyses were adequately powered and well-conducted, the lack of association between reductions in brain amyloid and cognitive outcomes suggests that amyloid reduction is not a valid surrogate for cognitive benefit.

EARLY DIAGNOSIS, HEALTH DISPARITIES, AND HEALTH-CARE ACCESS ARE BEING USED AS MARKETING STRATEGIES

Biogen launched a major advertising campaign in the United States as well as partnerships dressed up as efforts to facilitate early AD diagnosis. The campaign offers a brief symptom quiz for memory difficulties. All respondents, even individuals who report never experiencing any of the broad, vaguely defined symptoms, are informed “Even if you answered ‘Never’ or ‘Almost Never’ to all questions, talk to your doctor about any concerns you may have and ask if cognitive screening is right for you” (21). Encouraging cognitive screening seems innocuous, but this is a familiar playbook: Industries that stand to benefit from treating a disease work to increase the number of people diagnosed with that disease (22). These efforts erode the boundaries of disease and grow the pool of potential candidates for treatment.

The excess risk of dementia in Black individuals has elicited more attention as Biogen and advocates consider access to the medication. Shortly before the FDA decision, the Alzheimer’s Association—which actively campaigned for Aduhelm approval—released a special report entitled “Race, Ethnicity and Alzheimer’s in America” (23). Remarkably, this report gave little attention to the unequal burden of AD in communities of color; why these communities are disproportionately affected; or effective interventions to improve well-being and quality of life of people living with AD and their caregivers (24, 25). The report instead emphasized unequal access to health care, particularly difficulty with payment and lack of insurance coverage.

Because Aduhelm RCTs lacked racial/ethnic diversity, we have little evidence that Aduhelm is safe or effective for people of all racial/ethnic groups. Biogen nonetheless invoked issues of health inequities as justifications for Medicare coverage of Aduhelm, arguing in public comments: “Any unnecessary undue rigidity in coverage requirements or delays in access will…exacerbate health inequities.” (26) Medicare coverage decisions will simultaneously address Aduhelm and other amyloid-targeting medications approved in the future. If Medicare agrees to cover Aduhelm, the federal government promises to pay Biogen 80% of the cost for millions of Americans with mild cognitive impairment (MCI) or AD (27). If only 10% of the 20 million US residents estimated to be living with MCI or AD pursue treatment, the annual cost would total $45 billion. For comparison, the total 2021 National Institutes of Health budget was $43 billion, with about $3 billion designated for AD research (28). The Aduhelm price tag is shocking because RCTs demonstrate that it does not cure AD or even stop cognitive decline. At this price, for every research dollar the federal government spends to identify effective strategies for AD, it would pay Biogen around $15 for a potentially ineffective medication.

Unequal access to health care is a grave injustice, and attention to this priority—including commitments from pharmaceutical companies such as Biogen—is always welcome. Biogen could improve access to Aduhelm by reducing the price.

PRAGMATIC POSTAPPROVAL STUDY DESIGNS COULD ACCELERATE PROGRESS TOWARDS EFFECTIVE TREATMENTS

The FDA specified that a new phase 4 RCT of Aduhelm must be reported by 2030. Even this timing may be optimistic: FDA-mandated postapproval studies routinely miss specified deadlines with no repercussions. Fielding an RCT of a medication that is being prescribed in clinical settings will be difficult, and Biogen’s consumer-targeted marketing efforts may undermine RCT enrollment.

Much sooner than 2030, urgent concerns about Aduhelm must be resolved, including overall efficacy, adverse effects, generalizability to real-world patients (29), and whether amyloid reduction is an appropriate surrogate for clinical improvement in AD. The Centers for Medicare and Medicaid Services (CMS) offered a step in this direction in a “Coverage With Evidence Development” decision for amyloid-targeting therapies. This decision offered coverage of amyloid-targeting therapies approved by the FDA on the basis of biomarkers rather than clinical evidence, such as aducanumab, only for participants in qualified RCTs. Delays in treatment access were a major argument against the CMS coverage decision (30).

Pragmatic randomized study designs could provide rigorous evidence more quickly than traditional RCTs (31). There are 2 essential features of a rigorous study design: 1) a data source to define a population eligible to receive treatment and monitor outcomes for those individuals; 2) a process that increases the likelihood that some identifiable group of people receives treatment, although those individuals are otherwise unlikely to have different outcomes (i.e., people must be effectively randomized so the probability of receiving treatment is unrelated to their potential cognitive outcomes). Two critical insights of instrumental variable methods are that the process for achieving exchangeability can arise in numerous “quasi-random” ways and quasi-randomization can be useful even if only a moderate proportion of people adhere to assigned treatment.

Leveraging ongoing data collection platforms will expedite the research. Possible data sources include existing observational studies, which would take advantage of prior investments in recruitment, data collection, neuroimaging, and already planned or ongoing monitoring of participants. Pragmatic study designs often require large sample sizes, thus prioritizing passive follow-up data sources that can quickly accrue large samples (e.g., medical records in large health-care networks or Medicare data).

A simple method to create pragmatic quasi-experiments for evaluating safety and efficacy in real-world populations draws on lottery-based distribution or timing. Many versions of such lotteries are feasible, including individual-level randomization as in an RCT; randomized timing of when clinics first offer treatment; randomly assigning dates for active outreach to patients to assess their interest in the medication; or even a sequence based on a patient characteristic that is otherwise unrelated to AD outcomes, such as odd versus even day of birth. Timing-based lotteries could be fielded nationally (e.g., by CMS) or within a large health-care network. Given the timing of medication effects, the lottery would need to assign individuals to roughly a 0- to 12-month delay in treatment initiation.

Arbitrary thresholds or discontinuities offer another pragmatic study design. Health-care systems could define sharp thresholds for treatment initiation, such as a Mini-Mental State Examination score below 26 or diagnosis of mild cognitive impairment prior to a specific date, creating an arbitrary discontinuity in the likelihood of receiving treatment.

None of these designs are perfect, and each would require tailoring to the context, data source, operational considerations, and priorities of patients and clinicians. All designs entail limitations, such as lack of blinding or provider-shopping. Design elements to ameliorate potential biases are important in pragmatic research, and pragmatic studies would not replace the importance of a phase 4 confirmatory RCT. However, pragmatic designs—coordinated by CMS as part of a Coverage With Evidence Development designation, or by provider networks, health-care systems, or even research networks—could deliver convincing evidence within 2 years. Faster evidence may also help agencies in other countries—where Aduhelm has not yet been approved—make wise decisions.

CONCLUSION

Effective treatments for AD are urgently needed. The evidence available to date is rife with methodological problems. Rigorous scientific evidence from experimental or quasi-experimental studies of Aduhelm is necessary to justify widespread adoption.