Is It Time to Approve Drugs Based on Amyloid Removal?

Just as Eisai notches wins in its campaign for lecanemab approval around the world—last month in the European Union, this week in Mexico—a dozen research and clinical leaders called on the U.S. FDA and its fellow agencies elsewhere to reconceptualize their deliberations themselves.

On November 13, Dennis Selkoe at Boston’s Brigham and Women’s Hospital, with peers in the U.S., Japan, and Europe, offered a proposal. They ask regulators to acknowledge four decades of evidence showing that cerebral Aβ aggregation causes Alzheimer’s disease by focusing the licensing approval process squarely on amyloid clearance. The authors advocate for their position in a Perspective article in the journal Alzheimer’s and Dementia (Aisen et al., 2024).

The time has come, they say, to set aside the current regulatory requirement for cognitive/clinical change to prove an investigational drug’s efficacy. Instead, the FDA could grant standard, traditional approval to anti-amyloid antibodies that safely remove most aggregated Aβ from a person’s brain. Especially if a drug at hand also changes the biomarkers downstream of amyloid deposition that reflect AD pathobiology, then this drug interferes with the central disease pathway and ought to be made available. Cognitive, clinical, and quality-of-life data can still be gathered during the drug’s subsequent clinical use, the authors suggest.

Their main points are:

  • A large and varied body of published literature on human genetics, longitudinal cohorts, as well as mechanistic exploration in animal and other model systems, have established cerebral Aβ amyloid aggregation as a cause of AD. Substantial consensus exists on a subsequent pathogenic process toward neurodegeneration and dementia that can be characterized with increasing precision by change on imaging and blood-based biomarkers.
  • Three FDA-approved immunotherapies plus gantenerumab all remove brain amyloid; the more robustly an antibody does so, the more strongly it slows downstream biomarker and cognitive worsening.
  • The field’s formal, published knowledge base is sufficient to predict with confidence that amyloid removal interferes with the pathobiology of the disease. This renders amyloid removal a surrogate marker for regulatory considerations.
  • Therefore, new agents using a similar mechanism as lecanemab and donanemab deserve approval if they safely and substantially remove brain amyloid.
  • Forthcoming results of ongoing secondary prevention trials of approved amyloid immunotherapies will likely ease this proposed regulatory change.
  • Outside of AD, lowering aggregated molecular deposition in atherosclerosis and in transthyretin cardiac amyloidosis slows failure of the affected organ and clinical symptoms in these diseases.
  • In 1987, the FDA approved lovastatin based on LDL-cholesterol biomarker reduction. Other statins soon followed, years before their clinical benefit in preventing myocardial infarcts was formally proven. Like AD, atherosclerotic cardiovascular disease is complex and multifactorial; still, this regulatory strategy proved correct.

According to the authors, the past 50 years of debate about whether Aβ deposition causes AD, and whether its removal would benefit patients, have drawn to a close with the Phase 3 lecanemab and donanemab results. Throw in the less successful trials of aducanumab, gantenerumab, and previous anti-amyloid agents, they say, and you can see that those that were proven to have engaged, and then robustly reduced, their target did affect the disease process. In fact, the more completely an agent removed aggregated Aβ, the more beneficial it was. “This reasoning represents a conceptual shift: we should consider in our evaluation of such amyloid-targeting trial outcomes that the agent demonstrably inhibited a causative biological process,” they write.

The lecanemab and donanemab Phase 3 trials were large, statistically strong, concordant, peer-reviewed, and published—essentially lifting the field out of the realm of debate and onto settled terrain. With that, the time has come for the FDA to update its framework. The agency can now consider evidence of amyloid lowering below 15 to 20 centiloids—supported by evidence of a lessening of tau, microglial inflammation, astrocytosis, or neurodegeneration biomarker abnormality—not only for accelerated approval, as the agency currently does, but also as a surrogate biomarker for future cognitive benefit and grounds for traditional approval.

Dear reader, what do you think? Is this proposal timely? Premature? Overdue?

For context, after decades of effort but little progress, an age of transformation in AD clinical research and care has finally arrived. Even as pharma companies, hospitals, payors, neurologists, and nurses grapple with the challenges of rolling out these first disease-modifying treatments, self-injected and more potent second-generation immunotherapies are already coming down the pike, and DNA-based anti-amyloid drugs are entering early stage trials.

As 2025 fast approaches, is the evidence sufficient for regulators to speed up this medical transition by agreeing that standard approval can rest on “significant changes in a mechanistically linked and biologically meaningful biomarker coupled with reasonable safety data,” as the authors put it?

Would this best serve the 60 million people who live with early symptomatic AD? What about the 300 million or so who are presymptomatic?

The authors hope their argument will “stimulate discussion and debate among regulators, clinicians, scientists, and members of the lay public concerned about accelerating access to disease-modifying treatments for sporadic AD.”

Read their open-acess article, and consider it in light of what you know and see in your own work, in your own life. Then contribute your perspective into the comment box below, or email contact@alzforum.org.—Gabrielle Strobel

Comments

  1. While Aisen et al. argue for regulatory approval of amyloid-lowering therapies based on biomarker evidence, we currently caution against prioritizing biomarker outcomes over robust clinical endpoints in Alzheimer’s disease therapeutic trials for the following main reasons:

    1) Amyloid Load Remains an Unvalidated Surrogate Endpoint

    As we argued previously (Planche and Villain, 2021), demonstrating a causal link between amyloid clearance (or residual amyloid load) and clinical benefit is critical but currently lacking. Without robust mediation analyses to prove that amyloid reduction drives cognitive improvement, using this biomarker for regulatory approval is premature. Other fields, such as multiple sclerosis, have adopted surrogate endpoints like MRI T2 hypersignal load only after thorough validation (Sormani and Bruzzi, 2013). 

    2) Undefined Thresholds and Efficacy Concerns HERE Aisen et al.'s endorsement of regulatory approval of amyloid-lowering therapies based on biomarker evidence does not address the necessity for specific thresholds or rates of clearance for treatment efficacy. The authors reference “robust and significant amyloid removal (potentially targeted to < 20 CL by amyloid PET, a level considered to be in the normal range.” This is an ambiguous benchmark. This oversight could result in the approval of drugs like gantenerumab, which, despite significant amyloid clearance in the Phase 3 GRADUATE trials (approximately -48 to -57 centiloids), showed no cognitive benefits and raised safety concerns with a 25 percent incidence of ARIA-E and 5 percent symptomatic ARIA (Bateman et al., 2023). 

    The data on amyloid-negative individuals in the anti-amyloid trials remain difficult to interpret. In the GRADUATE I trial, 28.0 percent of participants receiving gantenerumab achieved amyloid-negative status (≤24 centiloids) compared to 2.4 percent in the placebo group. In the GRADUATE II trial, this was 26.8 percent for gantenerumab and none for placebo (Bateman et al., 2023). In contrast, in the lecanemab trial, 68 percent of the lecanemab group were amyloid-negative (< 30 centiloids) versus 16 percent in the placebo group (Bateman, 2022), while in the donanemab trial, approximately 76.5 percent of the combined donanemab group were amyloid-negative (< 24.1 centiloids) compared to ~0.5 percent in the placebo group (Sims et al., 2023). 

    To standardize these findings across studies, accounting for high variability in the placebo groups, risk ratios can be calculated as recommended by the Cochrane Handbook for meta-analysis of trials with dichotomous outcomes. These RR should be interpreted as surrogates of Hazard Ratios in the absence of survival curves. Calculations yield the following:

    • Lecanemab (CLARITY-AD): RR = 4.23 (95 percent CI 3.05-5.86)
    • Gantenerumab (pooled GRADUATE I + II): RR = 25.4 (95 percent CI 3.52-182.8)
    • Donanemab: RR = 176.1 (95 percent CI 56.87-545.0)

    Given these results, what should the approval threshold for "robust and significant amyloid removal" be? Should we focus on the percentage of the treatment group independent of the placebo group or consider the Hazard Ratio of occurrence of negative amyloid PET? Should the statistical significance of HR be considered enough?

    Approving drugs solely based on amyloid reduction risks introducing ineffective therapies into clinical practice, potentially harming patients. The field should instead emphasize rigorous validation of surrogate endpoints to address the urgent need highlighted by Aisen et al. for expedited drug development in this critical public health issue. Once validated, establishing consensus on threshold levels will be essential.

    See also: Bateman, R. J. (2022). Imaging, Plasma, and CSF Biomarkers Assessments from Clarity AD. 15th Conference Clinical Trials Alzheimer’s Disease. 23743084

    References:

    Aisen P, Bateman RJ, Crowther D, Cummings J, Dwyer J, Iwatsubo T, Kosco-Vilbois M, McDade E, Mohs R, Scheltens P, Sperling R, Selkoe D. The case for regulatory approval of amyloid-lowering immunotherapies in Alzheimer's disease based on clearcut biomarker evidence. Alzheimers Dement. 2024 Nov 13; Epub 2024 Nov 13 PubMed.

    Bateman RJ, Smith J, Donohue MC, Delmar P, Abbas R, Salloway S, Wojtowicz J, Blennow K, Bittner T, Black SE, Klein G, Boada M, Grimmer T, Tamaoka A, Perry RJ, Turner RS, Watson D, Woodward M, Thanasopoulou A, Lane C, Baudler M, Fox NC, Cummings JL, Fontoura P, Doody RS, GRADUATE I and II Investigators and the Gantenerumab Study Group. Two Phase 3 Trials of Gantenerumab in Early Alzheimer's Disease. N Engl J Med. 2023 Nov 16;389(20):1862-1876. PubMed.

    Planche V, Villain N. US Food and Drug Administration Approval of Aducanumab-Is Amyloid Load a Valid Surrogate End Point for Alzheimer Disease Clinical Trials?. JAMA Neurol. 2021 Nov 1;78(11):1307-1308. PubMed.

    Sims JR, Zimmer JA, Evans CD, Lu M, Ardayfio P, Sparks J, Wessels AM, Shcherbinin S, Wang H, Monkul Nery ES, Collins EC, Solomon P, Salloway S, Apostolova LG, Hansson O, Ritchie C, Brooks DA, Mintun M, Skovronsky DM, TRAILBLAZER-ALZ 2 Investigators. Donanemab in Early Symptomatic Alzheimer Disease: The TRAILBLAZER-ALZ 2 Randomized Clinical Trial. JAMA. 2023 Aug 8;330(6):512-527. PubMed.

    Sormani MP, Bruzzi P. MRI lesions as a surrogate for relapses in multiple sclerosis: a meta-analysis of randomised trials. Lancet Neurol. 2013 Jul;12(7):669-76. Epub 2013 Jun 3 PubMed.

    View all comments by Nicolas Villain View all comments by Vincent Planche

  2. I want to start by congratulating the EMA CHMP for their courageous and forward-thinking decision to revise the verdict on the approval recommendation for Leqembi. This marks a critical step in addressing the immense challenge posed by Alzheimer’s disease in the EU. The decision reflects the growing body of evidence supporting amyloid-lowering treatments and their potential to alter the trajectory of the disease, which is brilliantly summarized in 12 key points in the article by Paul Aisen and other leading investigators in the field.

    The findings substantiating amyloid's central role in Alzheimer’s disease collectively solidify the argument that amyloid accumulation is not merely associated with, but indeed pivotal to, disease progression. Given this robust evidence, I wonder if the time has come to drop the term "hypothesis" from the “Amyloid Cascade Hypothesis”, especially in the contexts of DIAD and DS-AD, where its validity appears beyond reasonable doubt. Based on the results of trials studying amyloid lowering drugs and the observed effects in early sporadic AD populations, it might be time to accept this disease concept for all forms of AD.

    The article's key suggestion—to approve drugs based solely on their capacity to reduce amyloid burden and their safety profile—is compelling and pragmatic. Such a paradigm shift would expedite access to treatments for patients and lower both timelines and costs for the development of new drugs.

    However, this approach necessitates the establishment of registries (ideally global) to rigorously monitor the efficacy and safety of these treatments in real-world settings. An international disease registry of this nature would also allow for the identification of rare side effects and comparative analyses of various amyloid-lowering agents. This would be of great benefit to patients in need for effective therapies.

    Also, I think it is crucial to include small molecules, not only active and passive vaccination, in these considerations. These agents could play a significant role in making treatments accessible and affordable, especially for low- and middle-income countries.

    View all comments by Johannes Levin

  3. I strongly endorse the arguments put forward by Aisen and colleagues for regulatory approval of Aβ-targeting therapies based on imaging and biofluid biomarker evidence, exactly as happened for the approval in 1984 of lovastatin, based on its cholesterol-lowering efficacy.

    Central to these arguments is the continuing debate in some circles on what is the cause of AD? The etiology of AD is now clear: The proximal cause is the accumulation of Aβ-amyloid; there are multiple upstream causes which drive this accumulation of Aβ. Downstream modulators of the effects of Aβ-induced neurodegenerative changes—particularly tau/ptau accumulation, GFAP-associated gliosis, retrograde axonal degeneration manifested by white matter hyperintensities (WMH)—can also now be used as surrogate markers of Aβ-amyloid lowering.

    We are beginning to appreciate that Aβ also drives a-synuclein accumulation, which in turn may be used as a downstream marker of AD progression. These downstream events must be differentiated from true AD comorbidities—especially the frontotemporal dementia spectrum and small vessel disease, both of which are independent of the AD process—that currently cause confusion in evaluating the “clinical meaningfulness” of the Aβ-lowering immunotherapies.

    A major question, so far unanswered, is why do the Aβ-lowering immunotherapies not halt or reverse the disease in the more advanced stages? Much loose language is currently being used to describe “plaque” clearance and lowering the Aβ load to ”normal” levels at 15-20CL, but we lack postmortem evidence of just how much soluble and insoluble Aβ-amyloid remains at this 15-20CL level. The advent of next-generation PET tracers, better PET imaging equipment (with higher resolutions), and better reconstruction algorithms for quantitative evaluations of areas in the brain most affected should permit the estimation of 0CL as the true measure of Aβ-load clearance. Once this question has been clarified, we can begin to look at efficacies of therapies from both clinical and pathological perspectives.

    The need for primary and secondary prevention of Aβ-accumulation has never been greater. Treatments for established AD dementia are feasible, we just need to take advantage of all we have learned from the success of statins in controlling atherosclerotic diseases over the past 40 years.

    View all comments by Colin Masters

  4. Aisen et al. suggest the possibility of treating individuals with biomarker evidence of amyloid plaques with anti-amyloid treatments, regardless of whether these individuals have cognitive impairment, and they provide a strong scientific rationale for this approach based on the pathophysiology of AD. Importantly, they are clear that they are suggesting this possibility to encourage discussion within the scientific community.

    Based on our ever-increasing understanding of AD, including the key scientific findings described in this article, the authors are likely correct that clearing amyloid plaques, especially early in the course of AD when patients are still asymptomatic, will ultimately be beneficial. However, results from scientific research and clinical trials can be humbling—we sometimes learn that our hypotheses and assumptions are incorrect, or that the specific treatment regimen we are testing is suboptimal. Therefore, we believe that the use of these treatments in cognitively unimpaired persons must await the results of prevention trials that are designed to evaluate whether specific regimens for amyloid-lowering treatments delay or even prevent the onset of cognitive impairment.

    We appreciate that awaiting the results of prevention studies comes with an important drawback—some individuals may develop irreversible cognitive impairment that could have been forestalled if the treatments prove to be effective and had been provided earlier. On the other hand, these treatments are associated with significant risks, burden, and costs, and it would be highly unfortunate for individuals to bear those factors for years only to discover later that the treatment regimen provides no clinical benefit.

    An additional important issue is that providing these treatments based solely on biomarker results (regardless of cognitive symptoms) would require biomarker testing of many cognitively unimpaired older adults. There could be negative consequences associated with biomarker testing of cognitively unimpaired individuals, including possible loss of eligibility for certain forms of insurance as well as discrimination. For cognitively unimpaired individuals, the benefits of treatment must outweigh the potential risks of biomarker testing.

    Currently, we recommend focusing on increasing access for patients with early symptomatic AD to anti-amyloid treatments, because rigorously performed clinical trials have demonstrated that anti-amyloid treatments can benefit these individuals by slowing cognitive decline. We still have many challenges in providing these fully FDA-approved treatments to patients who may benefit, including those who are from minoritized groups. To address these challenges, we need to greatly increase our capacity for diagnostic testing, treatment, and monitoring. Building a high-capacity diagnostic and treatment clinic infrastructure that serves all groups will be extremely important as a platform to which the enormous demands of prevention therapy could eventually be added. 

    We are optimistic about the potential of amyloid-lowering treatments to be part of effective prevention strategies, but first these treatments must be rigorously demonstrated to delay or prevent the onset of symptomatic AD.

    View all comments by Suzanne Schindler View all comments by John Morris

  5. Aisen et al. propose that regulatory agencies should consider conditional regulatory approval for amyloid-lowering immunotherapies based on evidence of amyloid removal combined with clear-cut fluid biomarker evidence. Safety would be taken into account as usual. As the critical part of this proposal, positive outcome on clinical efficacy endpoints would not be required at this approval stage. Indeed, in the passive immunization studies in the prodromal and mild dementia stage of AD, the degree of amyloid lowering correlated with the size of the clinical effect. Hence, it would be relatively safe to assume that, in this population, other drugs with similar mechanisms of action that remove amyloid to the same degree would have similar clinical effects. However, the authors want to extend this reasoning to other segments of the Alzheimer population, in particular the asymptomatic Alzheimer stage. As a clinical scientist I find this problematic.

    To remove amyloid in the asymptomatic stage of Alzheimer disease should not be too difficult from a technical viewpoint, given what we now know from the prodromal and mild dementia stage studies. The authors propose that amyloid removal and its effect on standard biofluid markers would substitute for clinical endpoints. This argument critically relies on assumptions about the link between fluid biomarker levels and clinical benefit. In the asymptomatic stage, the evidence of a link between changes in fluid biomarkers and interventional benefit on clinical endpoints is not there at the moment. The strategy proposed may fail to resolve the essential scientific question: which regimen of amyloid removal in which asymptomatic individual at which timepoint makes a difference to the future of the individual in terms of cognitive and behavioral decline, and independence of living.

    The authors argue that the clinical effects can be determined post-conditional approval. That is an illusion. The type of evidence that the authors would omit is the most relevant but at the same time the most difficult to obtain. If it is not built into the original study design, endless debates may ensue. Stopping or delaying clinical Alzheimer’s disease obviously matters a lot. However, this requires studies of a long duration. In the asymptomatic phase, longer study duration may be more feasible than in the symptomatic stages. It would be a missed opportunity to replace this ambitious goal with a purely amyloid and fluid biomarker-defined criterion for positive outcome, which would be faster but less conclusive.

    However, there may be one exception as a biomarker. Tau PET levels have a strong association with current and future clinical disease expression. If an intervention in the asymptomatic stage removes amyloid and also affects tau aggregation as measured with tau PET, the likelihood that this would be beneficial for the clinical disease course is high and the strategy proposed by Aisen et al. may work very well.

    As a second problem I see in this proposal, for health technology assessment, evidence that an amyloid-lowering drug alters AD-specific biomarkers without proven clinical benefit in the study target population, may well be insufficient. Building effectiveness measures into the design of interventional studies is important. The size of the clinical effect allows for benchmarking compared to investments in other treatments and other diseases.

    Third, given the high prevalence of amyloid positivity in the asymptomatic population, the strategy proposed by the authors risks increasing the population eligible for the intervention a lot. Requiring also clinical effectiveness, instead, may well lead to more restricted, judicious use of the drug. For most healthcare systems worldwide, this is a more realistic and affordable way forward. Cost does not refer only to monetary cost: There is also a major concern about the medical intrusiveness into a healthy individual’s life if biomarkers only are deemed sufficient as criterion and endpoint, in the absence of proven clinical benefit during the life course of the individual.

    These are exciting times for Alzheimer’s clinical research. As scientists we are offered unprecedented opportunities to learn how the different elements of the Alzheimer’s jigsaw fit together, and how they can be rearranged for the good of our patients. In populations that differ from those that have been successfully targeted in the prior amyloid-lowering trials, the effect of amyloid removal on clinical endpoints is unknown. It can only be determined empirically. Argumentation based on theory cannot substitute for rigorous empirical proof.

    Conflicts of interest: RV was the global PI for the Phase 1 and 2 pivotal trials of 18F-flutemetamol and the PI for the first-in-human study of 18F-MK6240. RV’s institution has clinical trial agreements (RV as PI) with Alector, AviadoBio, Bristol Myers Squibb, Denali, J&J, and UCB. RV’s institution has consultancy agreements (RV as DMSB chair) with AC Immune.

    View all comments by Rik Vandenberghe

  6. This position paper proposes an interesting extension to evaluating outcomes of anti-amyloid immunotherapy: approving a drug based on safety and strong efficacy in amyloid clearance, and not requiring clinical outcome measures. Evidence for the amyloid hypothesis as an initiating factor in Alzheimer’s Disease is systematically summarized, as well as existing group-level data from anti-amyloid clinical trials in AD showing that antibodies that robustly lowered cerebral amyloid to < 20 CL were associated with slowing of clinical progression in symptomatic AD, and as-yet-unpublished data that a proportion of participants showed stabilization or even slight improvement of cognition in a long-term extension study of lecanemab.

    To some extent, the proposal is analogous to the well-established concept of therapeutic equivalency, in which outcomes from a new drug are compared against those from a similar drug. This concept could accelerate drug treatment options for AD. However, many important details of responses and outcomes of anti-amyloid immunotherapy have not yet been analyzed at an individual level. For example, the variability of responses to highly efficacious antibodies among patients has not yet been explained. In the donanemab Phase 3 trial, 69 percent of participants had an amyloid burden at 18 months of < 20 cL. Can we predict who the less-complete responders who did not achieve this threshold might be (e.g., older, APOE e4 carriers, people with co-pathology)? And how strongly was robust amyloid clearance associated with a significant clinical response of substantial slowing of progression or even stabilization and improvement? I think examining this level of data would help to inform a framework that might allow new amyloid-clearing drugs to be approved based on safety and amyloid clearance alone.  

    Many details would need to be ironed out. For example, what is an acceptable definition of safety? Would ARIA rates slightly higher than those of monoclonal Abs currently approved in the USA be considered safe? Would safety need to be assessed in relation to factors such as APOE genotype, in order to calibrate against existing data? How large a clinical trial would be needed and how long should be allowed? For example, the Aducanumab Phase 1b (PRIME) trial, in which dose-dependent reduction of amyloid PET burden was identified, was associated with clearance below SUVR threshold for significant amyloidosis in 69 percent of participants in long-term extension at 24 months (Chen et al, 2024). This compares less favorably with donanemab or lecanemab, which reported similar rates of amyloid clearance below detectable SUVR at 18 months. An assumption is that antibodies that remove amyloid, despite targeting markedly different epitopes, are likely to have similar outcomes, and that the clinical benefit—which in clinical trials to date has been slowing of cognitive progression on average—will be similar for every new antibody that shows safety and efficacy comparability. Perhaps animal model data and human neuropathology data could be used to support these claims.

    The proposal also suggests that regulatory approval based on safety and robust removal of amyloid should be applied to prevention trials in people without symptoms. This would markedly accelerate the availability of therapies for prevention. It is likely that anti-amyloid antibodies should clear amyloid more rapidly in these trials (such as the ongoing AHEAD A3-45 and Trailblazer-Alz 3 studies), because the amyloid burden at baseline is lower than in trials in early AD. However, without long-term follow-up from large clinical trials to determine at least some clinical outcomes, the proof of principle that amyloid clearance can delay or prevent emergence of symptomatic AD— or slow preclinical cognitive decline—does not yet exist. Emerging data from prevention trials will also allow blood-based biomarkers to be benchmarked as tools to monitor anti-amyloid treatment and track the recurrence of amyloid and other brain pathologies, but again will be most powerful if analyzed against the context of clinical outcome measures.

    While this proposal is forward-looking and has great potential to accelerate the availability of treatment options, I would like to see a stronger foundation of data to support the concepts and also to help clinicians to apply treatment to individual patients.

    References:

    Chen T, O'Gorman J, Castrillo-Viguera C, Rajagovindan R, Curiale GG, Tian Y, Patel D, von Rosenstiel P, von Hehn C, Salloway S, Hock C, Nitsch RM, Haeberlein SB, Sandrock A, Singhal P. Results from the long-term extension of PRIME: A randomized Phase 1b trial of aducanumab. Alzheimers Dement. 2024 May;20(5):3406-3415. Epub 2024 Apr 3 PubMed.

    View all comments by Douglas Galasko

  7. In this article, Aisen et al. argue that if new amyloid lowering immunotherapies (ALIs) were to come before the FDA with “clear-cut biomarker evidence” of amyloid lowering and acceptable safety profiles, approval should be granted prior to availability of evidence of clinical benefit from a fully powered analysis. The basis for the authors’ proposal is the promising but incomplete evidentiary record of lecanemab and donanemab. Here is why I believe this approach is misguided.

    My first objection is that while the group-level data for clinical benefits for the two drugs is convincing, individual-patient level data is almost completely lacking (which the authors acknowledge without any expression of curiosity or discomfort). Before concluding that the group-wise data is sufficient and the individual-level data not necessary, wouldn’t it be essential to obtain a thorough accounting of biomarker-clinical relationships at the individual patient level? For the record, I have asked for such information and been rebuffed, so far, by leadership from both sponsors.

    My second objection to their proposal is that the plan is silent on a key aspect of what constitutes “clear-cut biomarker evidence.” The authors state that clearance to 20 CL is their criteria but that is incomplete. In how many patients must full clearance occur? Merely 20 percent, which would have led to approval of gantenerumab? Or should it be 40 percent, to relitigate aducanumab’s accelerated approval? Or, instead greater than 60 percent to follow the model of lecanemab and donanemab? A straightforward interpretation of all of the ALIs would select the highest threshold, but Aisen et al. are silent on this point that they are surely cognizant of. 

    My third objection is their expectation that a pharma sponsor will report clinical outcomes in a timely and transparent manner. The authors acknowledge that evidence of clinical benefit would be required eventually, but they don’t specify when or how. What is to keep a pharma sponsor from dragging their feet on reporting disappointing findings?

    Then there is the slippery slope of variations in route of administration or dosing that are intended to reduce the logistical challenges of administering ALIs. How much leeway should be given? My guess is that some pharma sponsors will push the limits and beyond, to the point that the comparability of a new ALI to lecanemab and donanemab could be critically lost.

    But my main objection is that the authors’ well-intentioned goal of facilitating access to promising therapies will be counterproductive and put a new ALI in purgatory. One can imagine that pharma sponsors of future “me-too” ALI’s that chose to follow the Aisen et al. approach would perform trials that were minimally large enough to demonstrate requisite amyloid lowering and to meet regulatory safety requirements but underpowered for efficacy. Then what? The stakeholders, in contrast to the authors, would want proof of clinical benefit and would view an underpowered negative clinical outcome as a failure, regardless of the biomarker results. Why would patients choose to use an agent with unknown or unproved efficacy? Even with a gimmick that eases the logistical challenges and an agent with a better safety profile, what is the value to patients and sponsors of an accelerated approval of an ALI that cannot make an efficacy claim? Aducanumab proved to be a commercial failure because legitimate questions about its efficacy or lack thereof could not be papered over by the amyloid-lowering results or explained away with post hoc contrivances.

    As a treating physician, I acknowledge the allure of improved access of the Aisen et al. strategy; however, I am afraid it is an illusion. In Alzheimer therapeutics, the appropriate approach at this point remains the conduct of Phase 3 trials that simultaneously provide biomarker demonstration of target engagement in a majority of patients, yield adequate confidence in safety and tolerability, and prove clinical efficacy from a fully powered design. Ultimately, the transparent demonstration of clinical benefits is the gold standard that will drive improved access and allow patients, families and providers to make an informed decision on the merits of a new ALI.

    View all comments by David Knopman
    • User Profile Image Lon S. Schneider
      University of Southern California Keck School of Medicine
    • Posted:

    The authors state that lowering amyloid plaques, or attenuating their increase compared to placebo, should be considered a surrogate clinical outcome or marker for clinical trials of amyloid-lowering treatments. Specifically, for trials of amyloid-targeting antibodies for early symptomatic Alzheimer’s, preclinical Alzheimer's, and for primary prevention, evidence for clinical effectiveness could be inferred from differences in residual plaque load at the end of the trials and the antibody would be approved.

    Pragmatically, current and planned trials of lecanemab, donanemab, and amyloid-targeting antibodies in development would need to show only treatment-placebo differences in plaque load on PET to gain marketing approval. Clinical outcomes are optional.

    This position invites re-examination of the two Phase 3 gantenerumab trials and of the gantenerumab arm from the DIAN dominantly inherited AD trial for whether gantenerumab also should receive marketing approval despite not showing statistically significant effects on clinical outcomes. Notably, the gantenerumab trials showed about the same amount of plaque reduction as the aducanumab and lecanemab Phase 3 trials.

    The authors’ statement urging FDA approval based on plaques also requires, “the eventual provision of clinical data on efficacy.” This post-marketing requirement for the future demonstration of clinical efficacy makes the authors’ recommendations indistinguishable from FDA’s use of accelerated approval for aducanumab and lecanemab. (Accelerated approval is full marketing approval with post-marketing efficacy requirements.)

    A surrogate outcome is a substitute for the outcome of interest, which in Alzheimer’s disease is clinical outcomes or ratings. The main—if not only—reason to entertain a substitute is if the proposed surrogate is more efficient and precise and requires fewer research participants to gain a signal than the clinical outcome of interest. On its face, measuring plaques by PET seems to accomplish this, as it requires only a fraction of the 1,600 to 2,000 participants in a Phase 3 trial to show a difference in plaques when treating with amyloid targeting antibodies. Yet is that what we really want? Fewer participants exposed and smaller trials means less ability to examine subgroups and to determine differential effects. Or are we really seeking to oversize or overpower trials so that we have a plaque-density fallback if the clinical outcome isn’t significant?

    Elevated plaques may be definitional of Alzheimer pathology, yet change or new occurrence of plaques has not been meaningfully correlated or predictive of clinical change in the pivotal trials. FDA statisticians in their analyses for the aducanumab advisory committee shared that there is a very low correlation between lowering plaques and clinical change, correlation coefficients of around .09 to .12, indicating a shared variance of only 3 to 4 percent. Despite this lack of correlation, FDA officers approved aducanumab six months after the advisory committee based on their judgment that plaque reduction confers a likelihood of future benefit.

    The FDA background reports for the subsequent lecanemab and donanemab advisory committee meetings did not contain similar analyses that could serve to validate plaques as a marker. Neither have the pharma sponsors presented these analyses nor shared their trials data with others for exactly this purpose.

    The authors of this statement seemed to recognize this as well, writing, “[p]ending further analyses from the recent Phase 3 trials and their open-label extensions, we believe it is highly likely that the FDA approved Aβ antibodies to date have meaningful effects on the cognitive and functional trajectories of AD over time.”

    Belief and guesses are not adequate validators of efficacy. If we are going to advance biomarkers such as plaques as surrogate clinical outcomes in Alzheimer’s, then we must be willing to open science and share data to validate what are claimed to be meaningful outcomes.

    View all comments by Lon S. Schneider

  9. The authors present a case for accelerated regulatory approval of amyloid-lowering immunotherapies. Given current data in the field from anti-amyloid immunotherapies, they make several key points regarding anti-amyloid antibodies:

    1. Evidence of robust amyloid lowering combined with biomarker evidence of other biological benefits, for example, less soluble tau/phosphorylated tau (p-tau), less microglial inflammation, and less astrocytosis (GFAP), that could already contribute to accelerated approval by FDA criteria should, in our view, also support standard (traditional) approval based on robust plaque lowering (to < 15–20 CL) as a surrogate for future cognitive benefit.
    1. They argue that accepting such a regulatory paradigm shift would still require the eventual provision of clinical data on efficacy as well as safety, similar to the process in the FDA’s current accelerated approval pathway. Any regulatory approval based on robust amyloid lowering, improvement in other AD biomarkers, and safety would still involve the subsequent qualification of appropriate candidates for treatment by physicians knowledgeable about AD before any prescriptions are issued, and it would entail careful clinical follow-up.
    1. They also argue that amyloid lowering in secondary prevention trials of individuals who have amyloid plaques in the brain but are asymptomatic (as well as eventually forestalling amyloid plaque accumulation through primary prevention trials) should be accepted as compelling biomarker evidence of a drug-induced change in the fundamental biology of AD, providing the basis for regulatory approval and clinical follow-up.

    My view is that regarding points 1 and 2, I am in general agreement. However, the amount of data, the number of patients, safety requirements, and length of the trials that would show amyloid lowering below a centiloid value of 15-20 (as well as accompanying other biomarker effects) would need much greater definition. For example, I think that showing amyloid lowering in individuals with very mild or mild dementia due to AD (as in the trials of lecanemab of donanemab) would be required. Also, as safety issues are a concern with the current antibodies out to past six months, trials of reasonable length would be important, perhaps out to at least a year. It will also be important that any new agent being considered show no more and hopefully less side effects that are seen with currently approved antibodies. Some of the anti-amyloid antibodies have different mechanisms of brain entry that can be associated with other side effects (e.g., anemia and possibly other), so that would need to be taken into account. Provision of clinical data over a reasonable period of time (e.g., two to three years) would be very important for full approval.

    Regarding point 3, I think amyloid removal in individuals with preclinical AD (secondary prevention) and prevention of amyloid deposition in people prior to the onset of amyloid deposition is likely to show a much greater benefit than amyloid removal has been shown to date in individuals with very mild and mild dementia with AD. Two large secondary prevention trials, with lecanemab and donanemab, are ongoing. These should be completed. While it will take several years before they are completed, I think it is critical to see the outcome of these studies regarding efficacy and safety before any decisions are made to approve such agents based on amyloid removal at this time. If these studies are both positive, then I think the point the authors make—that lowering amyloid in secondary prevention trials, as well as forestalling amyloid plaque accumulation through primary prevention trials—should indeed be accepted as compelling biomarker evidence of a drug-induced change in the fundamental biology of AD, providing the basis for regulatory approval and clinical follow-up for future agents of similar class.

    View all comments by David Holtzman
    • User Profile Image Bruno Dubois
      Sorbonne University - APHP - Pitié-Salpêtrière Hospital
    • User Profile Image Nicolas Villain
      Sorbonne University - APHP - Pitié-Salpêtrière Hospital
    • Posted:

    The International Working Group has long contributed to refining the definition and diagnosis of Alzheimer’s disease (Dubois et al., 2007, 2024). Among its many efforts, two contributions remain essential:

    1) In clinical practice, AD should be diagnosed only in individuals with cognitive impairment. As clinicians confronted daily with patients, we cannot consider diagnosing such a serious disease in clinically unimpaired people who will probably never develop it.

    2) Regarding research, particularly clinical trials, we introduced and recently expanded the concept of presymptomatic AD subjects.

    This aligns with ethical and pragmatic imperatives. Ethical concerns: Current amyloid-targeting therapies, while achieving amyloid clearance, carry significant risks, including ARIA. These risks are acceptable in patients who have AD and make an informed choice, but they will be more difficult to accept in clinically normal people whose risk of progression to disease is uncertain. Their risk/benefit ratio substantially differs. Therefore, it appears more ethical to propose amyloid-removing drugs only to subjects with a very strong risk of developing the disease, i.e., to those who are already “on the tracks.” Pragmatically speaking, trials involving high-risk presymptomatic individuals are better poised to demonstrate therapeutic efficacy on clinical outcomes within reasonable timeframes.

    For these two reasons, it seems important to not go too fast in proposing regulatory approval of amyloid-lowering immunotherapies based on biomarkers in asymptomatic at-risk for AD. It is not proven that amyloid reduction drives cognitive improvement in cognitively unimpaired people. Instead, research should first focus on better defining presymptomatic AD subjects with a very high risk of progression to a clinical AD based on biomarkers profiles (among other factors), such as tau aggregates within the neocortex, very high amyloid load within APOE4 carrier, etc.

    We propose that specialized clinical research centers, so-called Brain Health Services (Frisoni et al., 2023), should develop algorithms to identify these high-risk individuals. This prediction involves multiple uncertainties, biomarker profiles being just one factor. Longitudinal studies of at-risk cohorts are essential for identifying additional relevant factors. These centers will also identify the different risk levels in asymptomatic individuals, which will determine tailored prevention strategies.

    References:

    Dubois B, Feldman HH, Jacova C, Dekosky ST, Barberger-Gateau P, Cummings J, Delacourte A, Galasko D, Gauthier S, Jicha G, Meguro K, O'brien J, Pasquier F, Robert P, Rossor M, Salloway S, Stern Y, Visser PJ, Scheltens P. Research criteria for the diagnosis of Alzheimer's disease: revising the NINCDS-ADRDA criteria. Lancet Neurol. 2007 Aug;6(8):734-46. PubMed.

    Dubois B, Villain N, Schneider L, Fox N, Campbell N, Galasko D, Kivipelto M, Jessen F, Hanseeuw B, Boada M, Barkhof F, Nordberg A, Froelich L, Waldemar G, Frederiksen KS, Padovani A, Planche V, Rowe C, Bejanin A, Ibanez A, Cappa S, Caramelli P, Nitrini R, Allegri R, Slachevsky A, de Souza LC, Bozoki A, Widera E, Blennow K, Ritchie C, Agronin M, Lopera F, Delano-Wood L, Bombois S, Levy R, Thambisetty M, Georges J, Jones DT, Lavretsky H, Schott J, Gatchel J, Swantek S, Newhouse P, Feldman HH, Frisoni GB. Alzheimer Disease as a Clinical-Biological Construct-An International Working Group Recommendation. JAMA Neurol. 2024 Dec 1;81(12):1304-1311. PubMed.

    Frisoni GB, Altomare D, Ribaldi F, Villain N, Brayne C, Mukadam N, Abramowicz M, Barkhof F, Berthier M, Bieler-Aeschlimann M, Blennow K, Brioschi Guevara A, Carrera E, Chételat G, Csajka C, Demonet JF, Dodich A, Garibotto V, Georges J, Hurst S, Jessen F, Kivipelto M, Llewellyn DJ, McWhirter L, Milne R, Minguillón C, Miniussi C, Molinuevo JL, Nilsson PM, Noyce A, Ranson JM, Grau-Rivera O, Schott JM, Solomon A, Stephen R, van der Flier W, van Duijn C, Vellas B, Visser LN, Cummings JL, Scheltens P, Ritchie C, Dubois B. Dementia prevention in memory clinics: recommendations from the European task force for brain health services. Lancet Reg Health Eur. 2023 Mar;26:100576. Epub 2023 Jan 31 PubMed.

    View all comments by Bruno Dubois View all comments by Nicolas Villain
    • User Profile Image Robert Alexander
      Alzheimer's Prevention Initiative, Banner Alzheimer's Institute
    • Posted:

    I think the proposal by Aisen and colleagues for approval of antibodies that remove Aβ based solely on imaging and supportive biomarker evidence is premature. While there is evidence from interventional and observational studies in support of an “amyloid threshold,” it remains a hypothesis and the required magnitude and rate of amyloid removal is not established. In particular, approving amyloid-lowering drugs for secondary prevention, in advance of the readouts of the AHEAD and Trailblazer-3 trials of lecanemab and donanemab, respectively, is not warranted without first establishing the link with delay in clinical progression.

    Accelerated approval of amyloid-lowering agents that have evidence of a differentiated safety profile, i.e., lower rates of ARIA, should be considered using biomarker evidence, but at the moment, lowering the bar for approval for all amyloid lowering agents would only result in more “me-too” therapies and risk the approval of ineffective agents.

    View all comments by Robert Alexander

  12. This is an extremely important discussion. I am homozygous for ApoE4, asymptomatic, a 15-year research participant in a longitudinal biomarker study, a research and AD advocate, and author of an award-winning book published by Harper-Collins of knowing my genetic status. The time is now that the patients' and research participants' voice be heard in a publication that is academically oriented. Lives and experiences in a non-academic voice need to be a part of the discussion as researchers decide the best way to proceed.

    I am grateful for the dedication and commitment the research community has for those of us who are at risk, and those patients and families who are horribly affected by AD. I am also disappointed that after 30+ years of AD research, we are not farther along in the process of decreasing progression of this disease. But before we move forward to treating asymptomatic patients, an important question needs to be answered to get a better understanding of why there are people who exist into old age with a large amyloid burden but never get the disease. We need not another hypothesis, but an actual answer before we move forward with treating asymptomatic patients. Ethics is the basis of all good research, and my question is: How ethical is it give an asymptomatic patient a DMT with possible harm, and a modest slowing of progression, when we don’t know if they will actually get the disease?

    Perception is reality. Sadly, the recent NYT article was misleading, but at the end of the day we have to recognize that ethics will be the most influential factor to ensure the trust of patients and participants.

    View all comments by Jamie Tyrone

  13. Let me add the perspective of a clinical trial participant. I entered a study of a mAb drug not knowing that having an APOE4 gene and a brain full of amyloid plaque would put me at higher risk for ARIA. I thought, being relatively young (65) and healthy, with only MCI, I would be at low risk. I was not told that I had both risk factors by the clinical staff who did know. After five subcu doses I developed ARIA-E with ocular migraines and visual disturbances, followed a few months later by a TIA, followed a few months later by seizures accompanied by ARIA-H. Also brain fog, confusion, medication, depression, loss of agency, loss of driver's license for eight months, and loss of confidence that goes along with brain injury. I will likely be on anti-epileptic medication for the rest of my life, according to my neurologist.

    I have read about the unfortunate side effects of the mAb drugs. If I could say that I knew my dementia had been delayed by a year or two, I might say it was worth it. But to have to say it got me nowhere is difficult. In fact, I suspect that my experience has brought me closer to dementia. If we multiply my experience by a small percentage of the number of trial participants, I fear we have damaged some brains in our efforts to heal them.

    In my reading about the clinical trials, I see the numbers of deaths, but I don't see the numbers of strokes, TIAs, or new seizures that are occurring. Who is keeping track of these? And who is monitoring the long-term effects of them? No one is monitoring me since the end of the phase of the study I was in. My recommendation, based on my experience and common sense, would be to not approve amyloid-lowering agents until we know more clearly that their benefits clearly outweigh their risks.

    View all comments by Kristine Shields

  14. I think it makes sense to concentrate on biomarker endpoints, but that doesn’t necessarily imply ignoring cognitive endpoints perpetually. Ultimately, that’s where the rubber meets the road!

    A broader point is that the Aβ immunotherapies are a relatively successful drug class, although this does not always seem to be broadly accepted yet.  I refer to Jicha et al., which compared Aβ immunotherapies to widely accepted biological (immunotherapeutic) remedies in the fields of multiple sclerosis, cancer, and rheumatology. The upshot is that the Aβ immunotherapies match up quite well in terms of efficacy, safety, and costs.

    References:

    Jicha GA, Abner EL, Coskun EP, Huffmyer MJ, Tucker TC, Nelson PT. Perspectives on the clinical use of anti-amyloid therapy for the treatment of Alzheimer's disease: Insights from the fields of cancer, rheumatology, and neurology. Alzheimers Dement (N Y). 2024;10(3):e12500. Epub 2024 Sep 18 PubMed.

    View all comments by Peter Nelson
    • User Profile Image Dennis Selkoe
      Co-Director, Brigham and Women's Hospital's Ann Romney Center for Neurologic Diseases
    • Posted:

    We are grateful for the numerous comments to our 12-author Perspective that have been posted on Alzforum through December 12. They provide thoughtful analyses and critiques of our provocative proposal of full regulatory approval of new amyloid-lowering immunotherapies based on amyloid PET and other AD biomarkers.

    While most commentators agreed with the general thrust of our piece, a frequent concern was distinguishing between trials in symptomatic AD patients and secondary prevention trials in presymptomatic individuals. Our group’s initial focus was on the former, but because the issues surrounding biomarker endpoints are closely similar, we included the latter. The sentiment of several commentators that one needs to await the completion of ongoing secondary prevention trials (AHEAD and Trailblazer 3) is something with which we fully agree. We do not intend to see any biomarker-based approvals for presymptomatic individuals until evidence of significant delay or prevention of clinical onset is observed. But we agree with those who opined that this is likely to occur, given the lower amyloid burdens and milder biological disease these people have.

    We acknowledge the critique that we did not yet specify details of what should be considered quantitatively “robust” amyloid lowering (perhaps to 15 CL or less) and significant decreases in related biomarkers (e.g., tau PET; plasma phosphotau and GFAP). In our view, general acceptance of the principal thrust of our Perspective, particularly as regards symptomatic individuals, should now be followed by consensus panels of experts in AD imaging and fluid biomarkers attempting to set these levels.

    This would include technical details of imaging parameters (tracers, scanners, centiloid cutoffs) and specific plasma and CSF assays. Obviously, each new pivotal trial will require individual design to set highly specific biomarker outcomes. Our point is that at least in symptomatic populations, robust lowering of amyloid and associated fluid biomarkers coupled with safety as good or better than current agents should allow approval and marketing to qualified patients before full clinical outcomes are in hand.

    We agree with those who called for regulators to require trial sponsors to provide clinical outcomes and further safety data post-marketing approval in a timely fashion. We also concur with the need for obligatory enrollment in a registry to collect such outcome data. And as suggested, the sponsors of the currently approved antibodies need to provide individual-level outcome correlations soon. Our principal goal is to accelerate access of symptomatic patients to new disease-modifying agents prior to full clinical validation but with required accrual of follow-up clinical and biomarker data. Regulators should be able to specify this, as in the FDA’s current accelerated approval pathway.

    As regards the challenges our field and society will face if one or more secondary prevention trials succeed on both biomarkers and symptom onset, we agree that the number of eligible recipients of amyloid-lowering agents may become very high, but biomarker-based prevention has been a successful approach in many other chronic diseases (e.g., ASCVD, hypertension, cancer, diabetes), and we have no choice but to follow these precedents. Why conduct intensive and expensive basic and clinical research on AD over four decades and not try to lower the number of people who advance to the late–and very expensive–stages of this devastating disease?

    Finally, among the numerous responses on Alzforum to date, none states that amyloid is not pathogenic or that its lowering is not critical. This contrasts with the egregious misinformation in a recent New York Times front-page article that claimed we know little about the cause of AD and the role of amyloid. In his posting, Colin Masters comments compellingly on the relationship of amyloid accrual to both upstream and downstream events in the Alzheimer’s cascade. It is no longer tenable to claim that “amyloid trials have failed” and that “we do not know what causes AD.” In our view, we must be bold in pushing ahead therapeutically even as new knowledge accrues, because waiting for ideal datasets risks leaving a great many victims behind.

    View all comments by Dennis Selkoe

  16. I was the director of FDA’s Division of Pharmacometrics before I left FDA in 2021. In July this year, I watched the presentation by Dr. Kevin Krudys, Associate Director, Office of Neuroscience, CDER, FDA, titled "Recent Examples of Reasonably Likely Surrogate Endpoint: Reduction in Amyloid Plaques Measured by PET in Alzheimer's Disease" (https://www.youtube.com/watch?v=luWpWCHAjKU starting at 25:40). Following this presentation, I made the following recommendation to my former FDA colleagues: It is time to convert amyloid plaque reduction to a validated surrogate endpoint to support traditional/full approval.

    I urged FDA to come up with a minimum threshold of amyloid plaque reduction to be clinically meaningful, so that the magnitude of amyloid plaque reduction can be translated to clinical benefit. That is the exact point of the CDR-sb vs SUVr group level mean plot in our aducanumab review, and of my publication Wang, 2023,  "An insider's perspective on FDA approval of aducanumab.

    I am glad Aisen et al finally published this paper to propose the same idea.

    References:

    Wang Y. An insider's perspective on FDA approval of aducanumab. Alzheimers Dement (N Y). 2023;9(2):e12382. Epub 2023 May 18 PubMed.

    View all comments by Yaning Wang

  17. A surrogate marker for clinical outcomes in Alzheimer’s therapeutics has long been regarded as the Holy Grail in the field. It could definitely make new drug development much easier and faster. I agree with the authors of this position paper that, based on biomarker evidence so far, amyloid lowering should be allowed regulatory approval as a surrogate marker to predict future clinical benefit in secondary prevention trials (for asymptomatic individuals) with anti-amyloid antibodies. When it comes to mildly symptomatic AD patients, however, it is a different story with me.

    At the end of the day, it is cognitive and functional outcomes that really matter to patients and families. Given the limitation of biomarkers for amyloid lowering, including apparent lack of consensus on "robust amyloid clearance'" definition by PET, and the variability in fluid biomarkers across racial and ethnic groups, it seems to me a bit too early to adopt amyloid lowering as a surrogate marker to substitute for clinical endpoints in early AD patients.

    From what I perceive, it is not determined yet how much change in fluid biomarkers (e.g., soluble Aβ42 or p-tau species) is equivalent to robust amyloid removal (e.g., 20 centiloid) by PET scan. In addition, there is an apparent discrepancy between robust amyloid lowering and clinical benefit depending on the stage of Alzheimer’s disease. For instance, early AD patients with high tau burden do not seem to respond to anti-amyloid immunotherapy as effectively as those with low tau burden. So, a surrogate marker of amyloid lowering alone would not suffice, and perhaps needs some other biomarkers combined, to substitute for clinical outcome in early AD therapeutic trials with amyloid lowering drugs.

    View all comments by Jae-Hong Lee

  18. It is interesting that the datasets, studies, and arguments presented in this perspective paper bring others to different conclusions, and with an equal degree of conviction.

    View all comments by Russell Swerdlow

  19. The Aisen et al. Perspective article argues that robust reduction of amyloid PET signal achieved by Aβ antibody-based therapies should serve as a surrogate biomarker of clinical response for approval of such treatments. The article provides a convincing summary of the evidence that, though there may be different causative mechanisms, accumulation of pathological forms of Aβ is the primary common pathway to initiation of AD. If one accepts this tenet, then several key questions remain, including:

    • At what point, and over what duration of the early stage time period, might pathological forms of Aβ, undetectable by amyloid PET or CSF Aβ measures, trigger molecular processes involving tau, glial, and/or other mechanisms contributing to synaptic degeneration?
    • Would such downstream processes be detectable with available biomarkers?  
    • At what point might such processes become largely self-sustaining and independent of Aβ?

    Accepting further that sufficient reduction of amyloid accumulation can have some degree of effect on slowing of cognitive and functional progression, fundamental findings in the field nevertheless argue against the use of amyloid PET as a surrogate biomarker for Aβ antibody-based therapies. As outlined in the Perspective article, there are no published reports of significant correlations between extent of decreases in amyloid PET measures and degree of slowing of cognitive and functional scores at the individual level—a fundamental requirement for a surrogate biomarker.

    This lack of correlation is consistent with the likelihood that, once triggered during early disease stages, other processes such as accumulation and intracellular cellular mislocalization of pathological forms of tau and/or various glial-based/inflammatory mechanisms make substantial contributions to ongoing synaptic failure and degeneration. An additional potential source of the lack of correlation between amyloid removal and clinical effects is the wide range of synaptic resilience to amyloid accumulation, reflected in maintenance of cognitive status in the face of significant amyloid loads in some individuals.

    The lack of correlation is also consistent with studies indicating that the best pathological correlation with decline in clinical status is the loss of synapses (e.g., Terry et al., 1991). Though Aβ pathology can drive the production of pathologic tau species, over time the influence of tau extends beyond locations in common with Aβ, and ultimately both synapse and tau status outperform amyloid accumulation in terms of cognitive correlations (Tanner et al., 2022) and thus may be viewed as more fundamental pathological processes underlying AD clinical status.  

    Though it is important to address target engagement, a surrogate biomarker should reflect pathological features of the disease that have been demonstrated to significantly correlate with clinical status. Given that synaptic failure and degeneration likely constitute the core and most proximate pathological features underlying the loss of cognitive function (Selkoe, 2002), superseding the influence of amyloid accumulation, surrogate markers in AD preferable to amyloid PET would be those capable of offering a measure of synaptic integrity.

    While further work is needed in this area, studies with synapse-based PET ligands or CSF or plasma synapse-related proteins measured in isolation or as part of panels emerging from proteomic studies (Lista et al., 2024) might lead to surrogate biomarkers that are more biologically fundamental to the core pathology of AD that is directly related to cognitive loss and thereby more accurate indicators of clinically meaningful therapeutic effects. This approach could then employ biomarkers that are treatment agnostic, but prioritize a measure of fundamental pathology that is as directly as possible related to, and correlated best with, cognitive and functional status.

    One cautionary note in allowing approval of new Aβ antibody-based therapies based on amyloid PET measures is the possibility that a given antibody approach might indeed substantially lower amyloid PET signal but not affect the distribution and/or composition of amyloid molecular species in a way that necessarily inhibits downstream pathological events to a sufficient extent. Different amyloid antibodies, while each reducing amyloid PET signal, could have distinct effects on tau species or other downstream events. Moreover, unexpected variations in the extent that amyloid lowering leads to reduction of downstream pathology might also occur across different populations. It is not yet clear whether we have sufficient downstream biomarkers to test for this possibility. For example, a recent concern has arisen that treatment with lecanemab, while successfully reducing amyloid PET signal in both female and male subjects, might have less clinical effect in female subjects (Lynch et al., 2024). 

    The possible approval of a therapy that reduces amyloid PET measures, but with minimal or no clinical effects overall or in specific populations, might have the net effect of slowing overall progress in the field toward creating more effective therapies, in addition to the concerning potential scenario of unrecognized undertreatment in those using the agent.

    Conflict of Interest. F. Longo is a board member, has equity interest in, and is a paid consultant for PharmatrophiX, a company focused on non-amyloid directed therapeutics for AD and other neurodegenerative diseases. F. Longo and S. Massa are listed as inventors of small molecules related to AD treatment and are entitled to related royalties.

    References:

    Terry RD, Masliah E, Salmon DP, Butters N, DeTeresa R, Hill R, Hansen LA, Katzman R. Physical basis of cognitive alterations in Alzheimer's disease: synapse loss is the major correlate of cognitive impairment. Ann Neurol. 1991 Oct;30(4):572-80. PubMed.

    Tanner JA, Iaccarino L, Edwards L, Asken BM, Gorno-Tempini ML, Kramer JH, Pham J, Perry DC, Possin K, Malpetti M, Mellinger T, Miller BL, Miller Z, Mundada NS, Rosen HJ, Soleimani-Meigooni DN, Strom A, La Joie R, Rabinovici GD. Amyloid, tau and metabolic PET correlates of cognition in early and late-onset Alzheimer's disease. Brain. 2022 Dec 19;145(12):4489-4505. PubMed.

    Selkoe DJ. Alzheimer's disease is a synaptic failure. Science. 2002 Oct 25;298(5594):789-91. PubMed.

    Lista S, Santos-Lozano A, Emanuele E, Mercuri NB, Gabelle A, López-Ortiz S, Martín-Hernández J, Maisto N, Imbimbo C, Caraci F, Imbimbo BP, Zetterberg H, Nisticò R. Monitoring synaptic pathology in Alzheimer's disease through fluid and PET imaging biomarkers: a comprehensive review and future perspectives. Mol Psychiatry. 2024 Mar;29(3):847-857. Epub 2024 Jan 16 PubMed.

    Lynch MA. A case for seeking sex-specific treatments in Alzheimer's disease. Front Aging Neurosci. 2024;16:1346621. Epub 2024 Feb 13 PubMed.

    View all comments by Frank Longo View all comments by Stephen Massa

  20. As a research neurologist based in Seoul, I would like to contribute my perspective on the use of biomarkers in Alzheimer’s disease diagnosis and treatment, particularly in light of the article shared here.

    Biomarkers versus neuropsychological tests:
    I fully agree with the vision of replacing neuropsychological tests with biomarkers in the long term. Biomarkers provide a more objective and scalable solution to tracking AD pathology. However, the current biomarker landscape, especially plasma biomarkers, may not yet be sufficient to fully replace existing diagnostic tools.

    For instance, in the Donanemab study, plasma pTau217 showed only a 23 percent reduction in slowing of disease progression, which is significantly less than the changes observed in amyloid PET or neuropsychological measures. This highlights the need to critically assess whether plasma biomarkers can adequately reflect disease progression, especially at the individual level.

    Limitations of plasma biomarkers:
    Plasma biomarkers face challenges with within-subject variability, which limits their effectiveness for personalized monitoring over time. While they show promise for group-level analysis, their application in individual treatment decisions remains constrained.

    Efficacy in low-pathology groups:
    I was particularly intrigued by the strong efficacy observed in patients with low amyloid or tau burden. However, placebo groups in these trials also show favorable outcomes, with approximately 50 percent experiencing no decline over 18 months and 25 percent even showing improvement. This phenomenon aligns with the IWG concept of "asymptomatic at risk" individuals, highlighting the need to better understand which patients genuinely require DMTs.

    Future directions:
    I strongly support the integration of biomarkers into companion diagnostics and clinical trial designs. However, to advance the field, we need plasma biomarkers capable of monitoring disease efficacy and identifying individuals who would benefit most from early intervention. Additionally, the development of preventive therapies for the preclinical stages of AD remains a critical priority.

    In summary, I believe biomarkers will play an increasingly central role in the future of AD treatment, but we must address their current limitations to maximize their utility in clinical practice and research.

    View all comments by Sang Won Seo

  21. Given the urgent need for primary or secondary prevention, safe interventions with negligible side effects are important. If a long-term intervention with negligible side effects results in a sustained reduction in Aβ accumulation, it should be approved by the FDA for early, including presymptomatic stage, AD treatment.

    View all comments by Yunfeng Zheng

  22. The main proposal in this perspective by Aisen et al. is that any new anti-amyloid antibody should be granted FDA approval if it is reasonably safe and lowers amyloid plaque burden substantially (under 15-20 centiloids is mentioned as a potential target), without requiring any initial proof of clinical efficacy. I will preface my comments by saying I agree, in the main, with the authors that the genetic evidence from early onset and late-onset Alzheimer’s disease points squarely at Aβ, in one form or another, as the likely causal agent in the pathogenic pathway. My comments here pertain specifically to amyloid plaque burden as a proposed outcome measure likely to predict clinical outcomes.

    The FDA’s decision in 2021 to approve aducanumab using the accelerated approval pathway was controversial. The FDA decided that amyloid plaque burden was a surrogate biomarker whose reduction was “reasonably likely to predict a clinical benefit.” The field, however, had long known from postmortem studies that the regions that harbored early, heavy amyloid plaque burden were not necessarily the regions considered to be dysfunctional early in AD (the medial prefrontal cortex is a good example of such a region). Conversely, regions like the medial temporal lobe—ground zero for tau pathology and linked functionally to memory loss, the classic first symptom of AD—do not show early deposition of amyloid plaque.

    This critical lack of spatial correlation between amyloid plaque deposition and “sick,” hypometabolic brain was beautifully redemonstrated with the advent of amyloid PET imaging. Work from the Rabinovici lab, for example, has shown that patients presenting with functionally distinct clinical AD syndromes (e.g., logopenic aphasia versus posterior cortical atrophy, aka PCA) have essentially indistinguishable amyloid PET scans whereas their tau PET and FDG PET scans faithfully reflect the functional neuroanatomy of their predominant symptoms (e.g., more tau and less metabolism in the left hemisphere of logopenic patients, more tau and less metabolism in the occipital lobes of PCA patients, etc. (Ossenkoppele et al., 2016). 

    Any lingering doubt that amyloid PET burden can serve as a surrogate biomarker “likely to predict clinical benefit” should now be fully dispelled in the light of four large, Phase 3 clinical trials of the three FDA-approved anti-amyloid antibodies, none of which has shown a significant correlation between the amount of amyloid plaque removed and the clinical outcome. That Aisen and colleagues would stick to the notion that amyloid plaque burden is a biomarker likely to predict clinical outcomes in the face of overwhelming evidence to the contrary (and derived from three distinct sources—postmortem studies, amyloid PET imaging, and clinical trials) is, in a word, surprising. In regard to the expected correlation between amyloid reduction and clinical outcomes, they appropriately state, “It should be noted that correlations reported to date are at a group level; there are few or no published reports of correlations in immunotherapy trials between decreases in amyloid PET and cognitive and functional scores at an individual level.”

    I think we can safely assume that there are no published reports of this correlation at the individual level in any of the four Phase 3 studies of approved antibodies because the correlation is close to zero and not significant. These data are notoriously hard to come by owing to the lack of data sharing and transparency on the part of the pharmaceutical companies involved. We managed to pluck some individual-level data for aducanumab’s Phase 3 studies from the appendix of the FDA’s aducanumab briefing document. In figure 1B of a recent commentary (Digma et al., 2024) we used these data to show that there was no significant correlation between amyloid plaque removal and clinical outcomes in either trial. There is a weak trend (r = 0.19, p = 0.075) for the ENGAGE study, but this was, ironically, the aducanumab trial that showed no clinical benefit.

    We were refused access to the data from the Phase 3 lecanemab or donanemab studies. Though they did not make the data available, I have since been informed by Eisai that there was no significant correlation between amyloid reduction and clinical outcomes in the Phase 3 lecanemab study (correcting for baseline amyloid levels, partial Pearson correlation, and partial Spearman correlation both < 0.1). Lilly has not, to my knowledge, provided any data or summary results on this question for the Phase 3 donanemab study, but I think one can safely assume that they also failed to see any correlation. In sum, four large Phase 3 studies of FDA-approved anti-amyloid antibodies have failed to show that the amount of amyloid plaque reduction is correlated with clinical outcomes, and yet Aisen et al. are suggesting amyloid plaque reduction is a biomarker likely to predict clinical outcomes. The proposal is seemingly at odds with the facts.

    I will close with some points about functional unblinding, since Aisen et al. address it in reference to our recent commentary (Digma et al., 2024). They state that functional unblinding is unlikely to account for an important share of the clinical benefit because 1) raters were blinded to ARIA, 2) sensitivity analyses were done to account for unblinding, 3) the clinical effect sizes in the lecanemab trial are larger than could be expected to result from unblinding due to ARIA. The attempt to blind raters is admirable, but in a given center word tends to get out about who developed ARIA. Furthermore, patients or their study partners could also incidentally inform the rater that, for example, their dosing was halted. More importantly, the CDR-SB depends on subjective feedback from the patient and the study partner so functional unblinding could impact scores even if raters remain perfectly blinded. Regarding sensitivity analyses, these are generally flawed in that statistical analyses correcting for ARIA invariably dilute the active treatment arm of longitudinal data from APOE4 carriers who tend to progress more quickly than non-carriers.

    Finally, in our commentary we focused on ARIA-E because it was the easiest to get individual-level data on from the studies. ARIA-H also triggered halts in the study protocol for these trials. What is really needed to understand the potential role of functional unblinding due to ARIA is the difference between active treatment and placebo groups in the percentage of patients with any ARIA (E or H) that triggered a change in the protocol. The authors’ point that only 2.8 percent of ARIA patients in the lecanemab trial were symptomatic likely has no bearing on functional unblinding, which we posit may occur when a patient and their study partner are told that dosing will be halted and/or more frequent safety MRIs will be required. Infusion-related reactions are also substantially more common in the active treatment arms of these studies than in the placebo arms and might also be expected to contribute to functional unblinding. It would be very easy to understand the impact of functional unblinding on outcomes if Eisai, Biogen, and Lilly would make the data available. Until it is examined explicitly and transparently, statements on the effect size of functional unblinding are merely speculative.

    References:

    Ossenkoppele R, Schonhaut DR, Schöll M, Lockhart SN, Ayakta N, Baker SL, O'Neil JP, Janabi M, Lazaris A, Cantwell A, Vogel J, Santos M, Miller ZA, Bettcher BM, Vossel KA, Kramer JH, Gorno-Tempini ML, Miller BL, Jagust WJ, Rabinovici GD. Tau PET patterns mirror clinical and neuroanatomical variability in Alzheimer's disease. Brain. 2016 May;139(Pt 5):1551-67. Epub 2016 Mar 8 PubMed.

    Digma LA, Winer JR, Greicius MD. Substantial Doubt Remains about the Efficacy of Anti-Amyloid Antibodies. J Alzheimers Dis. 2024;97(2):567-572. PubMed.

    View all comments by Michael Greicius

  23. This could accelerate the timeline for introducing antibody therapies, but societal costs, including high drug prices and resource utilization, must be carefully weighed against the clinical and societal benefits. Even if amyloid-clearing immunotherapy proves to be effective and necessary, there will inevitably be practical challenges to address.

    In the case of cognitively normal yet amyloid-positive individuals, it is often challenging to identify them, as they rarely seek medical attention under typical circumstances. To enable early detection and treatment for such individuals, the development and commercialization of blood-based biomarkers, such as plasma pTau-217, which can reflect early brain deposition, will be crucial. This approach could provide a cost-effective alternative to the high-cost imaging methods like amyloid PET scans.

    Another critical issue is the need to clearly communicate the potential side effects of amyloid immunotherapy, particularly ARIA. Given that FDA-approved drugs have been announced as disease-modifying therapies, expectations surrounding these treatments have grown significantly. Managing these expectations has become as crucial as managing dementia patients themselves.

    Although these therapies are disease-modifying, their efficacy is approximately 30 percent, and they carry the risk of ARIA. It is particularly important to emphasize caution for APOE4 carriers, as they may face higher risks. Public awareness must prioritize these aspects before promoting the use of such therapies.

    If it is announced that these treatments can be used for asymptomatic individuals with amyloid deposition, then older individuals with sufficient financial resources may strongly demand access to these therapies. To address such challenges, it is necessary to clearly define the eligibility criteria for treatment, provide transparent explanations of the current efficacy data, and emphasize the need for continued treatment. A more cautious and measured approach will be essential to ensure responsible use and appropriate management of these therapies.

    Lastly, there are concerns about prescribing an expensive drug like Leqembi, especially when its effectiveness for Asians and women remains unclear. Additionally, in Korea, as it has been approved by the KFDA but is not covered by insurance, there is a risk that access to treatment could exacerbate socioeconomic inequality. This raises the need for the drug to be supplied at a more affordable price. Notably, in Korea, the Korean Dementia Association is taking the lead in drafting guidelines and specific protocols for prescribing Leqembi. They are actively providing education to target clinicians and related healthcare professionals to ensure its proper use.

    View all comments by Inhee Mook-Jung
    • User Profile Image Karl Herrup
      University of Pittsburgh School of Medicine
    • Posted:

    The string of laudatory comments posted on this paper prompt me to weigh in and offer a bit of balance. My answer to the question in the title of this piece is an unequivocal “no.” Sometime later we can discuss whether the current crop of monoclonals should be prescribed to patients, but there is little to discuss with regard to using amyloid removal as the sole criterion, or even the major criterion, for approving a new drug or treatment. The Aisen et al. article makes 13 points in defense of this position. Twelve of them need to be dismissed; one is valid but weak. Taking them in order:

    Point 1: People with AD undergo progressive Aβ deposition.

    This is circular logic. Ever since the AT(N) “framework” was introduced, AD has been defined as dementia with amyloid. If a person with clinical AD does not have amyloid, they do not have AD by definition. That makes Point 1 nearly meaningless. It also conveniently ignores the large fraction of cognitively healthy people who have increasing amounts of amyloid in their brain, but no dementia. The comeback here is that they are just in the preclinical phase of AD (see Point 13 below), which is still more circular logic. You must believe in amyloid in spite of a considerable amount of conflicting data to get caught up in this spiral.

    Points 2-6 focus on the genetics of early onset (familial) AD

    Point 2: APP is on chromosome 21 and people with Down's syndrome get amyloid and dementia.

    Point 3: People with a microduplication of the APP region of chromosome 21 develop AD at an early age.

    Point 4: Mutations in APP or PSEN (1or 2) cause an early dementing illness

    Point 5: There is a linear correlation between age of AD onset and the Aβ 42/40 ratio.

    Point 6: A rare protective APP mutation slows Aβ production.

    These are all good points by themselves. They are consistent with a world in which Aβ, and in particular the 42/40 ratio, drives AD. But all the same data can be used to support different, equally valid models of AD. This is because the genetic data only point to the APP gene, not to β-amyloid. There is a substantial literature on the functions of the full-length APP protein, as well as its non-Aβ breakdown products—secreted APP (sAPP) and the APP intracellular domain (AICD). Many of these functions when compromised produce symptoms of dementia. Thus, if I were to argue that elevated levels of the AICD were the cause of AD, Points 2-4 and 6 could all be cited as evidence in favor. Point 4 deserves a special call-out. The PSEN fAD mutations are often used to argue that APP and its membrane processing needed to produce Aβ are key factors in early onset forms of dementia. As these proteases cut APP right at the C-terminus of the Aβ peptide, the PSEN mutations are consistent with an Aβ-driven model of AD. It is worth noting that they are also consistent with an AICD-driven model of AD. The problem is that if the presence of γ-secretase mutations is evidence in favor of an Aβ model, then the absence of α- or β-secretase mutations must be considered as equally strong evidence against it. There are additional flaws in Point 5 (the 42/40 ratio). It has been argued, for example, that it is the reduction of Aβ42 that is the true reason dementia sets in. Point 5 also sidesteps the finding that most of the PSEN mutations reduce rather than elevate the activity of the γ-secretase. Given the uncertainties, it is a bit of a stretch to use this evidence as an argument for a major policy change at the FDA.

    Point 7. APOE is involved in Aβ clearance. This is true, but APOE is also involved in a plethora of other cellular activities, including cholesterol transport in the brain. In the absence of a more solid mechanistic relationship, the finding that APOEe4 carriers have altered Aβ dynamics is currently just a correlation.

    Point 8. Injecting Aβ oligomers isolated from human AD brain into rat brains degrades memory. First, it’s curious that Aisen et al. chose to cite a crude brain extract as the active agent here. It may be that brain-derived amyloid preserves a novel peptide configuration but based on the methods in Shankar et al. (2009) it is equally likely that the active agent here is a small contaminant tagging along with Aβ. Second, lots of things injected into a rat brain will degrade memory, especially Tris-buffer extracts of human tissue. Third, degraded memory is not the full clinical picture of AD. There is a wide range of symptoms that are missing, including the relentlessly progressive nature of the clinical decline. As an argument in favor of an Aβ-based model, the data are weak. As support for an argument for the FDA to abandon clinical symptoms in evaluating a treatment, Point 8 is not persuasive.

    Point 9. Aβ lowering precedes other pathological signs in CSF during disease progression. This is true, but the emergence of gray hair also precedes many of the pathological signs of the disease. The temporal order of two events is not proof that the earlier event causes the later one. Loss of myelin integrity is also an early disease sign, but there has not been a groundswell of support for myelin-based treatment strategies. Similarly, if inflammation were the true cause of AD, the sequence of Aβ and tau appearance may simply reflect their differential sensitivities to chronic inflammatory signals; no causal relationship is needed.

    Point 10. The data from the three recent clinical trials of monoclonal antibody infusions targeting amyloid. We could argue this one till the cows come home, but instead I will allow that this point has some validity. I insist, however, that it is weak tea as the data we have from the trials are incomplete. In particular, the individual level data have not been released and, without them, the meaning of the modest slowing of the average rate of cognitive decline is difficult to interpret, especially since the proposal is to jettison these very symptoms as meaningful measures of the success or failure of a new treatment.

    Point 11: Lowering other amyloids protects other organs and …

    Point 12: Amyloids in other organs do damage.

    These points are largely irrelevant as support for the use of amyloid lowering as sufficient to decide the efficacy of a new AD treatment. The only amyloid that counts, and the only amyloid that will be measured, is APP-derived β-amyloid—assessed with either a PET ligand in situ or by ELISA in blood. Data from other amyloids belong in the discussion section of a manuscript, not in an argument before the FDA. Also, the finding that 30 percent of elderly individuals have brain amyloid burdens but no cognitive symptoms suggests that APP-derived amyloid in brain might be more benign than other amyloids in other organs.

    Point 13. People with amyloid, but no symptoms, have presymptomatic AD. This point makes for an excellent structure to the narrative in that the authors close their list of 13 points as they began it—with another powerful example of circular logic. If AD is defined by amyloid, then this concept of preclinical AD must be the case by definition, even if it is not true. It does not forward the authors’ case.

    The cell biology of AD does not support a role for the Aβ peptide as a major driver of Alzheimer’s disease. I spent most of an entire book discussing the reasons I take this position, and others have voiced more eloquent and equally powerful concerns. If the FDA were to adopt the position put forward in this article, our brains might be free of plaques, but our health care system would be flooded with clinically ineffective treatments.

    View all comments by Karl Herrup

  25. Where are we now? Where should we focus our efforts? These are two key questions for public health in diseases that affect large numbers of people, such as Alzheimer’s disease.

    The proposal discussed here, by Aisen et al., goes in this direction. The 13 points of the article make clear the role of amyloid in Alzheimer's disease and the impact its removal has on the disease.

    The change in clinical thinking about Alzheimer’s disease and its new diagnostic criteria has been driven by the emergence of biomarkers of the disease and by the works showing its preclinical pathophysiological development up to 20 years earlier. The approval of aducanumab was not good for its clinical impact but was a starting point for a conceptual change in the historical view of the disease, its pathophysiology, and its treatment.

    I think that two issues need to be separated. One is the diagnosis of Alzheimer's disease. My position is that you have Alzheimer's disease when you have symptoms, and in the preclinical stage you have a pathophysiology that leads to the disease, and that is why I prefer the IWG position on asymptomatic or presymptomatic patients at risk of the disease.

    This conceptualization is very important for Latin countries where, for patients, a diagnosis of asymptomatic Alzheimer's disease is different than a diagnosis of asymptomatic patients at risk of the disease. But at this point, we should not confuse this concept with the absence of treatment in asymptomatic subjects and waiting for the person with the pathophysiology to present symptoms to treat them.

    We need to move toward primary and secondary prevention in people at risk of Alzheimer’s disease, not in people with Alzheimer’s disease. For the latter, we need to look at biological outcomes. Otherwise, we cannot address the presymptomatic stage, and we neglect it. We need to think about treating the pathophysiology that leads to Alzheimer's disease. We need to learn from the history about the role of statins in controlling atherosclerosis and its consequences.

    It is essential, especially to be able to advance with these treatments in low- and middle-income countries (Custodio et al, 2024), to have simpler biomarkers to detect the disease; here, serum biomarkers will be key for these groups to access treatments. These fluid biomarkers of Alzheimer’s disease would allow us to predict future clinical benefits and wider access to patients, especially in low- and middle-income countries.

    I support the suggestion of Aisen et al. about the possibility of evaluating treatments based on amyloid reduction without clinical outcomes when we already know that it causes or leads to the disease, without forgetting what precautions or programs we should implement to include advances in LMIC.

    This is the only way to speed up access to disease-modifying treatments for millions of patients with a more diverse and inclusive perspective.

    References:

    Custodio N, Allegri R, Lopera F, Caramelli P. Need to adapt Alzheimer's disease criteria in Latin America. Alzheimers Dement. 2024 Nov;20(11):8206-8208. Epub 2024 Oct 6 PubMed.

    View all comments by Ricardo Allegri
    • User Profile Image Dennis Selkoe
      Co-Director, Brigham and Women's Hospital's Ann Romney Center for Neurologic Diseases
    • Posted:

    On behalf of the co-authors of our Perspective in Alzheimer’s & Dementia, I am taking the liberty of providing some responses to the interesting commentaries made here on Alzforum since December 12 (when I responded to prior comments). Given the logistical complexity of responding as a 12-author group, I have elected to provide my personal responses, which may differ in part from those of my colleagues. Other co-authors may choose to comment further in the future.

    I appreciate Yaning Wang’s reminder of his and his former FDA colleagues’ leadership in advocating amyloid plaque reduction as a validated surrogate endpoint to support traditional/full approval. Yaning and his co-workers helped pioneer the movement to analyze disease-defining AD biomarkers alongside clinical outcomes and safety in AD trials.

    We agree with Jae-Hong Lee and had emphasized in our Perspective that, in addition to lowering amyloid PET to normal levels and matching or improving existing safety, a new agent needs lowering of other well-defined AD biomarkers for approval, including certain plasma tau isoforms and GFAP. A combination of biomarkers of AD progression is needed.

    Frank Longo and Stephen Massa emphasize the importance of assessing biomarkers downstream of amyloid accumulation as well as cognitive endpoints. Our Perspective proposed that approval based on the surrogacy of amyloid PET should be accompanied by other biomarkers (plasma tau and GFAP and perhaps tau PET). This multi-biomarker requirement for approval addresses their point about antibody-mediated amyloid lowering that might not be associated with downstream benefits. Our central argument was that the 13 points of evidence (including causative genetics) supporting amyloid as a very early and ubiquitous feature of AD suggest that amyloid lowering is reasonably likely to predict future neurobiological benefits, including reducing tau accumulation and some sparing of synapses. Longo and Massa agree with us about the desirability of correlating amyloid loads with cognitive outcomes at the individual level, and we all hope that further data from the Phase 3 trials of lecanemab and donanemab may still address this. Longo and Massa seem to assume that individual-level correlations will not be forthcoming. In any event, the reported correlations at the group level suggest that amyloid lowering across many participants does correlate with both biomarker and cognitive outcomes; these group associations were the basis for the positive FDA advisory committee recommendations and standard approvals of lecanemab and donanemab.

    We agree that synaptic loss is a highly desirable outcome measure, and once such biomarkers are validated in trials, they may complement amyloid PET lowering. But again, the 13 points about amyloid in AD pathogenesis suggest that amyloid lowering is likely to predict synapse sparing in large populations. And the observed temporary resilience to amyloid accumulation in some hosts is not a reason not to try to reduce it in all patients. Longo and Massa’s concern that a newly approved amyloid-lowering immunotherapy might turn out not to be able to slow cognitive decline at all is tempered by our requirements for a) parallel tau, GFAP, and other downstream biomarkers; and b) post-marketing accrual of cognitive outcome data. Finaly, their focus on non-amyloid targets, while highly worthwhile, incurs the risk of allowing an upstream pathogenic influence in AD (continued accumulation of amyloid and its associated Aβ oligomers) to go unchecked against the benefits of such downstream treatments.

    We agree with Sang Won Seo for the need to strengthen the utility of current plasma biomarkers of AD progression. We believe that further progress on these, coupled with robust lowering of amyloid- PET signals (and safety), may soon become the basis for approval of immunotherapies based on combined surrogate biomarkers.

    Michael Greicius also raises the issue of amyloid PET and cognitive outcome correlations at the group vs. individual level. We agree that the sponsors of the Phase 3 studies should make the data available to analyze this unsettled question meaningfully. There were correlations at the group level in the lecanemab and donanemab trials, and the associated slower cognitive decline in the face of robust amyloid-PET lowering, coupled with lowering of some other biomarkers such as tau and GFAP and adequate safety, presumably motivated both FDA advisory panel unanimity and FDA full approval. Our authors emphasized that the extensive biological and biomarker evidence that amyloid accrual likely initiates AD pathogenesis and contributes to its progression, with which Greicius generally concurs, should be taken into account when considering the potential of robust amyloid lowering (to <15 CL) to predict future clinical benefit and thus enable access of such agents to wider populations at any earlier time point.

    Greicius also makes several thoughtful points about the remaining possibility that functional unblinding caused positive trial outcomes on a group level. We agree that ruling this out will require further detailed analysis of the full data set from the trials.

    Inhee Mook-Jung raises important concerns about public education on the complexities of administering disease-modifying amyloid-lowering agents, including future use in presymptomatic individuals if current secondary prevention studies show efficacy. We agree with the need for carefully qualifying, and then closely monitoring, AD patients treated with lecanemab, donanemab, and any follow-on immunotherapeutics. Detailed Appropriate Use Recommendations (AUR) from an expert panel are widely available and helpful in this effort to ensure proper utilization. Economic considerations are also of great importance, but our Perspective’s focus was on the biological and medical rationale for improving therapeutic access of future patients to amyloid-lowering agents. Public and private institutions, including patient advocacy organizations, are already hard at work to prepare for the wider availability of amyloid-based (and other) disease modifying treatments once approved, and the issues raised by Inhee Mook-Jung should and will be thoughtfully addressed.

    Finally, Ricardo Allegri’s comments are generally well aligned with our A&D Perspective. He agrees with our central message of the importance of regulatory approval based on AD biomarkers. He also emphasizes the relevance of blood biomarkers, especially for underserved populations in low- and middle-income countries. Such tests are rapidly growing in use and could substantially broaden access to AD diagnosis and the monitoring of disease-modifying treatments like amyloid-clearing antibodies in those populations. On one point I personally do not concur: Allegri’s preference for the IWG’s definition of presymptomatic pathophysiology as distinct from AD (symptomatic disease). I prefer the concept that AD is a long biological continuum that moves subtly from asymptomatic to symptomatic phases. Like atherosclerotic heart disease and cancer, AD pathobiology is in the organ for years or decades but may not yet have become noticeable to patient and physician. 

    On behalf of my co-authors, I am grateful for the numerous cogent commentaries on our provocative but, we believe, timely proposal. While we are unlikely to address future comments in this forum, we encourage the posting of other viewpoints and further public discourse of the important subject of biomarker-based regulatory approval in Alzheimer’s disease.

    View all comments by Dennis Selkoe

  27. Aisen et al. provide a succinct summary of the evidence that Aβ plays a central role in the pathogenesis of Alzheimer’s disease, as it is defined biologically. They also point to the encouraging results of trials of amyloid-lowering monoclonal antibodies. It is now clear that amyloid lowering is now possible to a level that is seen in healthy adults. Clinical benefits have been harder to demonstrate, but are beginning to appear.

    The question is whether or not we are now at the point at which drugs should be approved for use based on amyloid-lowering data alone. I believe that more clinical trial evidence is needed.

    The most convincing clinical benefits are seen when amyloid clearance is complete, yet some symptomatic trial participants, particularly those with one or two APOE ε4 alleles, are not able to tolerate sufficiently high doses of amyloid-lowering drugs to reach this point and are withdrawn from trials because of amyloid-related imaging abnormalities (ARIA).

    Should treatment be confined to asymptomatic people with positive biomarkers? This question requires an adequately powered double-blind randomized prevention trial of sufficient duration to achieve a robust result.

    Why are APOE ε4 carriers at higher risk of ARIA? Is this risk solely because they have higher levels of brain amyloid? Does it relate to cerebral amyloid angiopathy (CAA)? How can we assess the state of the cerebral vessels non-invasively before introducing these drugs?

    I note the comments of Jamie Tyrone and Kristine Shields. I think we need to wait for further studies, and follow-up of former trial participants, to refine our understanding of who can benefit from these medications, and who is at high risk of adverse effects, before approval based on biomarkers alone.

    View all comments by William Brooks

  28. While Aisen et al. clearly aim to stimulate discussion among regulators, clinicians, scientists, and the lay public, as a member of the latter group myself I feel some critical aspects of this argument remain underexplored and need further attention. In particular, I would like to raise a few points that, in my view, deserve deeper scrutiny.

    1. Access Barriers Beyond Regulatory Approval

    The central motivation behind the authors’ proposal is the urgent need to make effective treatments more accessible. However, I question whether the bottleneck in AD treatment access is truly regulatory approval. Take Leqembi, for example, the Aβ antibody recently approved in the U.S. Initial market experience reveals several significant challenges:

    •            Lack of infusion sites and infrastructure for patient monitoring are key limitations. Many hospitals and clinics are simply not equipped to handle the infusion demands of large-scale patient populations.

    •            Shortage of diagnostic tools like PET and MRI scans, which are required for surveillance of treatment-related risks, such as ARIA (amyloid-related imaging abnormalities), further complicate widespread access.

    •            Cost barriers remain a huge concern, particularly for U.S. patients. Despite Medicare coverage, patients still face significant out-of-pocket expenses, and it’s unclear whether the reimbursement models in Europe will be any more straightforward.

    Thus, the issue of access may lie less with regulatory approval and more with logistical, diagnostic, and financial barriers that need urgent attention.

    2. The Statin Analogy: A Risky Comparison

    The authors draw an analogy between AD drug approvals and the regulatory pathway for statins in cardiovascular disease. I believe this comparison is not entirely apt for two key reasons:

    •            Risk profiles differ substantially. Statins have a well-understood side effect profile, and their adverse effects are generally reversible upon discontinuation. Routine blood testing can easily monitor these risks. In contrast, amyloid-targeting antibodies, even with close surveillance, have been associated with serious side effects like ARIA-E and ARIA-H (brain swelling and bleeding), which can lead to long-term health complications or even fatalities. The implications of these risks are more complex and less easily monitored than those of statins, which complicates a simplistic approval process.

    •            The molecular target is far less clear. Statins target a single, well-defined enzyme—HMG CoA reductase—whereas Aβ plaques in AD are heterogeneous and consist of various Aβ species, including post-translational modifications that differ from patient to patient. These modifications have been suggested to influence synaptic toxicity and disease progression. Therefore, the heterogeneity of Aβ pathology across individuals raises questions about whether plaque removal alone can be a meaningful indicator of disease modification. Furthermore, the difference in binding mechanisms between drugs like lecanemab and donanemab underscores the complexity of the Aβ target and the potential for varying therapeutic outcomes based on specific interactions with different Aβ species and aggregation forms. Approving treatments based solely on Aβ plaque removal, without clear evidence of clinical efficacy, seems overly simplistic.

    3. Risk of Stifling Innovation in AD Drug Development

    Perhaps the greatest concern with this proposal is its potential adverse impact on innovation. By focusing solely on Aβ plaque removal as a primary approval criterion, we risk creating an environment where drugs are fast-tracked based on their plaque-clearing potential rather than their ability to truly modify the course of the disease. This could inadvertently divert attention away from novel, potentially more effective approaches, such as small molecules targeting upstream processes in AD pathology.

    Moreover, given the scientific controversy surrounding Aβ as a primary driver of AD, it is premature to base regulatory approval on plaque reduction alone, especially when clinical data for Aβ antibodies remains limited, particularly with respect to long-term outcomes. Current trials, mostly confined to 72 weeks of placebo-controlled treatment, are not yet sufficient to conclusively demonstrate the efficacy of plaque-targeting therapies in altering the course of the disease.

    In this context, the proposal for post-approval collection of clinical data might sound appealing, but it introduces additional uncertainties. Clinical outcomes like cognitive and functional improvements, which have long been the gold standard for AD drug approval, remain essential for evaluating true disease-modifying effects.

    4. The Need for a More Nuanced Approach

    A paradigm shift in regulatory approval is, of course, welcome if it ultimately leads to faster access to life-changing treatments. However, we must be cautious about oversimplifying the criteria for approval, particularly when the scientific understanding of AD is still evolving. The complexity of AD pathophysiology and the heterogeneity of patient responses to treatments must be considered when designing regulatory pathways that ensure both safety and efficacy.

    To conclude, as this fascinating and, at times, controversial discussion continues, I hope that regulators, scientists, and clinicians will carefully weigh the potential risks and benefits of adopting a new approval paradigm based on Aβ plaque reduction and safety. While faster access to treatments for AD patients is an urgent need, we must avoid taking shortcuts that might ultimately hinder progress and innovation in the field.

    View all comments by Ulrich Dauer

Make a Comment

To make a comment you must login or register.

References

Paper Citations

  1. Aisen P, Bateman RJ, Crowther D, Cummings J, Dwyer J, Iwatsubo T, Kosco-Vilbois M, McDade E, Mohs R, Scheltens P, Sperling R, Selkoe D. The case for regulatory approval of amyloid-lowering immunotherapies in Alzheimer's disease based on clearcut biomarker evidence. Alzheimers Dement. 2024 Nov 13; Epub 2024 Nov 13 PubMed.

External Citations

  1. open-acess article

Further Reading

No Available Further Reading

Annotate

To make an annotation you must Login or Register.