Many people on aducanumab (trade name Aduhelm) develop the amyloid-related imaging abnormalities (ARIA) that mark fluid retention and microhemorrhages in the brain. In the November 22 JAMA Neurology, researchers led by Stephen Salloway at Brown University in Providence, Rhode Island, formally published the incidence and severity data from the aducanumab Phase 3 EMERGE and ENGAGE trials.

  • Overall, about one-third of people taking Aduhelm developed ARIA-E.
  • Of these, a quarter were symptomatic, 3 percent were serious.
  • Having brain microhemorrhage at baseline, or APOE4, increased risk.

Overall, 41 percent of people taking the highest dose developed ARIA, compared to 10 percent of controls. Most cases were asymptomatic, and most symptoms were mild. However, 1 percent of participants experienced serious side effects, i.e., those requiring hospitalization or causing long-lasting impairment. Some of these data were previously reported at the 2021 Alzheimer’s Association International Conference (Aug 2021 conference news). 

Adam Brickman at Columbia University, New York City, cautioned that the ARIA profile might be worse in clinical practice, where a broader range of patients are likely to be treated and monitoring may be less stringent. Brickman believes the trials’ lack of diversity leaves unclear how the results would generalize to a typical clinic population.

Others have expressed similar concerns, with one recent study noting that 92 percent of Medicare patients would not have qualified for the Phase 3 trials due to comorbidities such as cerebrovascular disease and other exclusions (Oct 2021 news). Additional reports of adverse events and a death potentially linked to aducanumab have surfaced (Nov 2021 conference news). 

In these two Phase 3 studies, the incidence of ARIA rose with dose. At 10 mg/kg—the dosage approved for clinical use—35 percent of participants developed the edema known as ARIA-E. This compared to an incidence of 2.7 percent on placebo. Notably, ARIA-E sometimes popped up in the same person more than once, with 10 percent of the high-dose group having recurrent episodes.

About 40 percent of people with ARIA-E also had the microhemorrhages known as ARIA-H. ARIA-H without ARIA-E was less frequent, at about 8 percent of the cohort, which was similar to the control group. About 6 percent of the high-dose cohort discontinued treatment due to ARIA-E or -H.

Three-quarters of all ARIA cases were asymptomatic and detected only on MRI. Of the symptomatic cases, two-thirds were mild, with the most common symptoms being headache, confusion, dizziness, and nausea. Most cases resolved within 16 weeks.

Factors such as age, sex, disease stage, or use of blood thinners did not affect ARIA-E incidence in these trials. In contrast, the presence of microhemorrhages at study baseline nearly doubled a person’s risk. APOE genotype likewise had a large effect, with ARIA-E incidence being 20 percent in noncarriers, 36 percent in heterozygotes, and 66 percent in homozygotes. For ARIA-H risk, APOE status and baseline microhemorrhages had no effect, but older age did.

As have others, Brickman bemoaned the fact that the efficacy data remain unpublished 2.5 years after the two trials were halted. “Scientists … should have the opportunity to scrutinize both the efficacy and the safety data not only in the form of peer-reviewed publications, but also via access to the raw data for re-analysis. An important question with respect to the newly reported findings is whether ARIA or baseline cerebrovascular status moderates clinical efficacy,” Brickman wrote.

Separately, Brickman wondered whether ARIA might ultimately hasten cognitive decline. “It’s hard to put a positive spin on the neuroimaging abnormalities … we simply do not know the long-term consequences,” he wrote to Alzforum (full comment below).

Meanwhile, in Washington, D.C., the Centers for Medicare and Medicaid Services is deliberating whether to cover Aduhelm, or go for a coverage with evidence development program. On December 10, Senator Ron Wyden (D-OR), called on Department of Health and Human Services Secretary Xavier Becerra to hold off on a scheduled 2022 increase in Medicare premiums for the country’s elderly population until after the CMS has rendered its decision, which is expected to come around January 12 (Wyden letter to HHS). When Medicare rejects coverage of an FDA-approved drug, state Medicaid programs pick up the cost; alas, resistance seems to be forming among some state Medicaid directors. Today, Stat reported that Oregon is preparing a request to be released from this requirement.

On December 3 in Science magazine, Holly Fernandez Lynch, University of Pennsylvania, and Christopher Robertson, Boston University, joined a growing chorus of researchers demanding that accelerated approval be buttressed with better, faster confirmatory trials (Lynch and Robertson, 2021).—Madolyn Bowman Rogers


  1. It’s hard to put a positive spin on the neuroimaging abnormalities reported in the EMERGE and ENGAGE Phase 3 trials of aducanumab. So-called amyloid-related imaging abnormalities (ARIA) of edema (ARIA-E) and hemorrhage (ARIA-H) were observed in about 41 percent of enrolled participants. Increased likelihood of experiencing ARIA was related to APOE genotype, older age, and baseline evidence of hemorrhage. Most abnormalities were described as mild, transient, and generally asymptomatic, although a fair proportion of participants had severe enough ARIA to discontinue treatment.

    The FDA approval of aducanumab for the treatment of Alzheimer’s disease has been controversial. Peer-reviewed data regarding safety are critical to inform the discussion among scientists, clinicians, advocacy groups, and patients and their families. While myriad points could be made about these important new data, I would like to highlight four non-mutually exclusive thoughts that come to mind to inform future research and clinical considerations.

    First, we simply do not know the long-term consequences of ARIA-H or ARIA-E. Both small hemorrhages (microbleeds) and evidence of brain edema in other contexts are documented to predict long-term negative clinical outcomes even in the absence of frank clinical correlates or in the presence of mild symptoms. So, while it is somewhat comforting that most imaging side effects were considered mild and transient, they may have longer-term negative impacts on cognitive trajectories, which may obscure or exceed any clinical benefit of aducanumab itself. Along these lines, it is important to highlight that the FDA provided accelerated approval for aducanumab because there was “an expectation of clinical benefit” based on biomarker evidence of amyloid removal, not based on unambiguous evidence of clinical efficacy. Using the same logic and considering the literature on brain edema and vascular sources of cognitive impairment, might we not conclude similarly that there is “an expectation of” longer-term negative clinical consequences due to these biomarker data of edema and hemorrhage even in the absence of obvious contemporaneous clinical correlates?

    Second, it is difficult to contextualize the reported neuroimaging abnormalities without a full understanding of the efficacy data in the EMERGE and ENGAGE trials. The results of these trials were presented in high-profile paper sessions at scientific conferences but have not appeared in a peer-reviewed publication. Especially in light of the safety concerns, potential financial conflicts, high cost, public health impact of the disease itself, and public skepticism over the process of approval, we need to see the efficacy data fully fleshed out. Scientists not involved directly with the aducanumab trials should have the opportunity to scrutinize both the efficacy and the safety data not only in the form of peer-reviewed publications but also via access to the raw data for re-analysis. An important question with respect to the newly reported findings is whether ARIA or baseline cerebrovascular status moderates clinical efficacy. 

    Third, the participants in EMERGE and ENGAGE did not represent the population of patients who would ultimately be eligible for treatment. As my colleagues Jennifer Manly and Maria Glymour pointed out recently in JAMA Neurology, only 0.6 percent (n=19) of participants who self-identified as black or African American were enrolled in EMERGE and ENGAGE, and of those, only six received the dose that was approved by FDA. Given the higher frequency of baseline cerebrovascular disease and APOE ε4 allele among black adults, we could speculate that ARIA would be more severe in this group because both factors were identified as predictors of these safety outcomes. On the other hand, while the APOE ε4 allele is more common in black people, it confers lower risk for development of Alzheimer’s disease in them, so there may be important effect modification by race. Without inclusive trials, we cannot draw reasonable inferences about safety outcomes for all prospective patients who could potentially benefit (or be differentially harmed) by the medication.

    Finally, the report raises concerns of how to implement accurate safety monitoring in a real-world clinical setting. At the very least, the findings suggest that prior to receiving treatment with aducanumab, patients should be screened with MRI for microbleeds, receive genetic testing for APOE status, and, once treatment commences, should receive frequent monitoring with MRI. Accurate detection of baseline (and incident) microbleeds is dependent on magnet field strength, scan resolution, specific sequence, protocol followed, and operator expertise. Outside of the tightly controlled setting of a clinical trial, it is likely that microbleeds will be under-reported, potentially putting patients at particular risk for ARIA. Genetic testing for APOE not only would inform a patient’s potential for developing ARIA with treatment but has implications for first-degree family members and their own respective risk for developing Alzheimer’s disease. Medical comorbidities are more common in “real-world patients” than in those who are enrolled in clinical trials, and there are likely comorbid conditions that would increase risk for ARIA that were not appreciated in a clinical trial setting.

    In summary, this report provides important insights into the safety profile of aducanumab. They need to be examined in the long term, integrated with efficacy data, studied in more diverse populations, and carefully implemented in real-world settings.

  2. Aduhelm was dropped on the public having failed the minimal FDA standard for regular approval, i.e., substantial evidence of effectiveness, as clearly demonstrated by the FDA’s independent statisticians. Seemingly as a work-around, it was given accelerated approval wherein the key criterion was FDA’s opinion that plaque reduction is “reasonably likely” to predict future clinical benefit. Since then, the hunt has been on to find even slight correlation between plaque reduction and clinical change that FDA hopes will persuade aducanumab nonbelievers that plaque lowering is a substitute for clinical benefit.

    The JAMA Neurology paper on ARIA is as expected of a well-written clinical trials report, except for the lack of clinical outcomes. By reporting only ARIA, the authors stoke the great mystery of 2021: Where in the world are the Phase 3, peer-reviewed clinical outcomes? It has been nearly three years since the trials ended.

    The paper reads like the ghost of a full clinical trials report, reporting trial methods, eligibility, CONSORT diagrams, dosing, discontinuations, and adverse events. The previously confidential protocol and statistical analysis plan are revealed for the first time. Indeed, “everything” is here except for any clinical benefit.

    The problem with this report is that we cannot fairly assess the safety of a treatment outside the context of its efficacy. A medication without clinical benefit is not safe at any level of adverse events. With aducanumab, there is a 41 percent incidence of ARIA-E in APOE4 carriers, 70 percent of whom have lesions larger than 5 centimeters. These study participants with “moderate” or “severe” ARIA were required by protocol to at least temporarily discontinue aducanumab. The authors casually state that 98 percent of ARIA resolves and no deaths in the trials were due to ARIA.

    They conclude by recommending that prescribers follow the label, get MRIs at the times stated, or when there are symptoms, and that the MRIs should be carefully assessed. They tell us that best practices and characterization of ARIA are continually being updated. This is hardly reassuring, as it sounds like “we’re still learning.” Tolerating high rates of edema might be OK if we were dealing with an effective medication as we can weigh the risks against benefits. So far, however, we can identify the individuals harmed by aducanumab; but we have yet to identify even one person who was helped by aducanumab.

    The expectation is extraordinary that a health system can monitor all patients, carefully assess, and protect them. The 76-year-old woman who died after developing severe cerebral ARIA had completed a Phase 3 trial, was in the open-label study, and had the best of clinical trials surveillance. Yet apparently an MRI showing moderate edema was missed by both local and central readers, and she received three more infusions (Business Insider). If this can happen in a clinical trial, then it can happen in regular care settings. There seems little margin for error here.

Make a Comment

To make a comment you must login or register.


Therapeutics Citations

  1. Aduhelm

News Citations

  1. Aducanumab: Will Appropriate-Use Recommendations Speed Uptake?
  2. Aduhelm Administration Remains a Trickle, ARIA a Concern
  3. Aduhelm Lowers Tau; Registry to Track Real-World Performance

Paper Citations

  1. . Challenges in confirming drug effectiveness after early approval. Science. 2021 Dec 3;374(6572):1205-1207. Epub 2021 Dec 2 PubMed.

External Citations

  1. Wyden letter to HHS
  2. Stat reported

Further Reading

Primary Papers

  1. . Amyloid-Related Imaging Abnormalities in 2 Phase 3 Studies Evaluating Aducanumab in Patients With Early Alzheimer Disease. JAMA Neurol. 2022 Jan 1;79(1):13-21. PubMed.