Today the U.S. Food and Drug Administration’s advisory committee gave the nod to donanemab, clearing the way for its marketing approval. All 11 members voted that the antibody was effective for people with mild cognitive impairment or mild dementia due to AD, and that its risk-to-benefit ratio was favorable overall.

Members expressed concern about a worse risk-to-benefit for APOE4/4 homozygotes, as well as a lack of data on underrepresented groups and on people who have amyloid plaques and MCI but almost no tangles. Nonetheless, the committee refrained from recommending that these groups not be treated. Rather, they stressed the importance of communicating limitations and uncertainties in the product label.

“The committee recognized, correctly I think, that the data supported a modestly favorable benefit to risk ratio for donanemab,” David Knopman at the Mayo Clinic in Rochester, Minnesota, wrote to Alzforum. Michael Greicius at Stanford University, Palo Alto, California, remains skeptical about anti-amyloid antibodies as a class, and today’s meeting did nothing to persuade him. “Fundamental questions about the efficacy of donanemab and lecanemab remain. It is remarkable that we have not seen a scatterplot showing the association between amyloid reduction and cognitive outcomes in the donanemab or lecanemab studies,” he wrote to Alzforum (comments below).

In good news for drug sponsor Eli Lilly, the committee argued against requiring tau PET scans before prescribing donanemab. The Phase 3 trial had selected patients by tau PET, enrolling only those above a cutoff of 1.10 SUVR. Those below this threshold were invited to join an open-label addendum trial that administered donanemab and tracked effects on biomarkers and safety, but did not measure cognitive/clinical outcomes. Despite the lack of this data, committee members nevertheless noted that this “no-tau” group notched similar biomarker benefits as did people in the Phase 3 trial. The committee believed requiring tau PET would impose too great a burden on patients and healthcare systems, strangling access to the drug.

In the donanemab trial, providers stopped treatment once plaque was gone. The AdComs liked this approach, saying it could motivate patients to complete treatment, and would free up infusion center chairs and provider time for a greater number of patients. However, the committee noted the dearth of data on what happens to cognitive status after stopping, and no data on when to restart treatment if a patient worsens. They felt these decisions should be left up to individual providers and patients.

During the meeting’s open forum, 18 people spoke in favor of donanemab, three against. The supporters comprised 10 patients or caregivers, three providers, and five representatives of advocacy or research groups.

If approved, donanemab will join lecanemab as the second anti-amyloid antibody currently on the market, after the earlier withdrawal of aducanumab. A more detailed story will appear in this space later this week.—Madolyn Bowman Rogers


  1. The advisory committee voted 11 to 0 on both questions in favor of the approval of donanemab. The committee recognized, correctly I think, that the data supported a modestly favorable benefit to risk ratio for donanemab. The committee acknowledged uncertainty around some of the subgroups, including the amyloid-positive, but low-to-no-Tau group, and seemed to be leaning toward being less restrictive rather than more restrictive.

    From the perspective of the role of an FDA approval, I agree with that approach. And yet, from the perspective of individual practitioners and patients who are considering the therapy, there needs to be a thoughtful discussion about the benefits in persons whose symptoms are very mild, those who are on the borderline of being too advanced, and those with two copies of the APOE e4 allele. Their benefit-to-risk ratios are less favorable.

    The committee had little enthusiasm for PET scanning despite acknowledging the unique information obtained from PET imaging. I am a bit disappointed that no one on the committee used the opportunity to call for greater availability of PET scanning. Its use does rationalize the use of donanemab.

    On the other hand, the committee did call for more education of potential patients and their families as well as clinicians, but that is a bit naïve considering how thin the dementia care expertise is in the U.S. except among dementia specialists. I have been concerned since the approval of lecanemab that the level of expertise for Alzheimer's disease and its clinical manifestations is very low in the community.

    Finally, although the committee was quite clear in wishing to improve access to donanemab, there was little acknowledgement of the considerable challenges in administering donanemab or lecanemab, from both a diagnostic and especially a safety perspective. The reality is that donanemab needs to be administered at a site with extensive multidisciplinary expertise, meaning that PET scanning might not be as inaccessible as the committee was worried about. Realistically, while the panel could not have required PET scanning, they could have emphasized in their comments that donanemab administration requires a high level of expertise in dementia neurology, neuropsychology, neuroradiology, and high levels of collaboration with nursing and with infusion center personnel. 

    But in the end, the advisory committee efficiently and unequivocally dealt with their charge from FDA.

  2. Fundamental questions about the efficacy of donanemab (and also lecanemab) remain.

    It is remarkable that we have not seen a scatterplot showing the association between amyloid reduction and cognitive outcomes in the donanemab or lecanemab studies. The proposed mechanism of action here—certainly for donanemab, which is supposed to be plaque-specific in its targeting—is that removing amyloid plaques will, somehow, result in better cognitive outcomes. If this is the case, then why have we not seen these basic scatterplots for donanemab and lecanemab? That would seem a critical piece of evidence to explore given the proposed mechanism. The unspoken understanding seems to be that there is no correlation, across subjects in a trial, between how much plaque is removed and how well patients do clinically. If there is no correlation between the biomarker reduction and the clinical outcome, then how do we explain the reported clinical efficacy?

    One explanation, which has not been adequately addressed, is the impact of functional unblinding due to ARIA. We have shown a correlation across trials such that trials with more ARIA tend to have better outcomes. This raises the concern that functional unblinding and the placebo effect may be contributing to the reported outcomes. The CDR-SB is particularly prone to inflation due to the placebo effect given its heavy reliance on subjective reports of how the patient is doing. Sensitivity analyses reported in the donanemab and lecanemab studies meant to address functional unblinding are inherently confounded by the much higher rate of ARIA among APOE4 carriers. Censoring data after ARIA will result in the removal of more longitudinal data from APOE4 carriers in the treatment arm. APOE4 carriers progress more quickly, and depleting your treatment group of them will make the treatment look better than the placebo.

    One simple plot would address both concerns. I encourage Lilly and Biogen/Eisai to provide a single scatter plot showing percent amyloid removed on one axis and clinical outcome on the other axis. The data points would be patients on active treatment. Patients who experienced a protocol-altering ARIA event (and who might have been unblinded) should be highlighted. This would let clinicians understand a) whether amyloid removal predicts outcome and b) whether patients who experience ARIA were more likely to have better outcomes (independent of the amount of amyloid removed).

    A second way to address the concern about functional unblinding due to ARIA would be for the companies to provide spaghetti plots showing individual patient trajectories on the CDR-SB over time with ARIA time-stamped in those patients who developed it. Will we see a post-ARIA bump supporting an unblinding/placebo effect? I bet that we would. Such an effect would also fit with the somewhat surprising finding that active and placebo arms begin to diverge by the first assessment (12 weeks in donanemab and three months in lecanemab), in line with much of the ARIA occurring after just two or three doses of drug.

    My humble request to the companies: Can you kindly provide 1) a scatterplot of amyloid removal vs outcome, with patients who experienced ARIA highlighted, and 2) spaghetti plots showing individual trajectories over time with ARIA events time-stamped. I would encourage my colleagues to urge the companies, at conferences and in papers, to make these basic figures available.

    I close by pointing to the dispiriting results of the donanemab tau PET study. Donanemab, boasting the largest reported effect size and the largest reported longitudinal tau PET sample (~400-600 per arm), had no effect on accumulation of tau PET signal at 76 weeks using two different regions of interest. How did donanemab achieve these clinical outcome results if there is no correlation between outcome and amyloid removal and if there is no impact of treatment on continued tau accrual? Functional unblinding remains a contender.

    The concerns raised above are elaborated on in our recent commentary (Digma et al., 2024). 


    . Substantial Doubt Remains about the Efficacy of Anti-Amyloid Antibodies. J Alzheimers Dis. 2024;97(2):567-572. PubMed.

Make a Comment

To make a comment you must login or register.


Therapeutics Citations

  1. Donanemab
  2. Leqembi
  3. Aduhelm

Further Reading