House of Cards

The road to the new Alzheimer’s drugs was laid more than thirty years ago with the development of a theory that the disease is primarily caused by the buildup of a protein in the brain called amyloid beta. Although amyloid deposits are present in the brains of Alzheimer’s patients, some people with amyloid never become demented, and nobody has shown definitively that the protein causes the disease rather than being a byproduct of it.

Experts have poked holes in the so-called “amyloid hypothesis” for years. Recently, doubts about the theory ratcheted up when an ongoing investigation, detailed in the journal Science in 2022, found that technical images used in a key study lending support to the hypothesis had been doctored.

With little else to go on, the drug industry has sunk billions of dollars over the last two decades into researching and developing anti-amyloid drugs. From 2004 through 2021, companies pursued at least twenty-three of them, all of which proved useless or even dangerous. In some studies, patients’ cognition worsened; in others, they suffered serious side effects, including unrelenting seizures, encephalitis, and death.

Enter Aduhelm, Leqembi, and Kisunla. These drugs employ monoclonal antibodies to attack amyloid, much like the way the immune system attacks a virus. All three have proved to be highly effective at removing amyloid from the brain. Unfortunately, little else about them is so straightforward.

The first in the class, Aduhelm, was codeveloped by Biogen and Eisai. Biogen stopped its two pivotal studies of Aduhelm early, citing “futility.” In other words, the drug didn’t work. Later, the company reanalyzed the same data, this time yielding conflicting results: one study showed a slight slowing of dementia; the other, which was nearly identical in design, was still negative: the drug didn’t confirm any benefit to patients at all.

Buried in the nearly one thousand pages of FDA analyses of Aduhelm were some troubling findings. When the FDA reviewers compared the amount of amyloid reduction in patients on the drugs to their cognitive function, there was no correlation. In other words, patients could experience a significant reduction of amyloid with no slowing of memory loss.

After Biogen presented its data to the FDA, the agency convened an advisory committee meeting of external experts in 2020. (Such meetings are held when study results are conflicting or uncertain). FDA officials told the external review panel that the negative study was unreliable, telling them to focus on the positive study, which showed some slowing of dementia. Billy Dunn, the agency’s top neuroscience official, also informed them that the agency would not consider removal of amyloid as a basis for approval, saying, “We’re not using amyloid as a surrogate for efficacy.” Instead, the FDA wanted the advisors to evaluate the drug’s effect on slowing dementia.

The committee members protested that the studies were nearly identical in design and it was irrational to focus only on the positive study. They gave the drug a decisive thumbs down (ten people on the eleven-person committee said there was not enough evidence of the drug’s effectiveness; the eleventh said they were “uncertain”).

At the suggestion of Dunn, Biogen applied for accelerated approval of Aduhelm based on its ability to remove amyloid rather than whether or not it was actually shown to slow dementia down.

This was precisely the approach Dunn had told the advisers the FDA would not use. So when the agency announced in early 2021 that it was overruling its advisers and granting approval based on reductions in amyloid, three of the advisers quit in protest. One of them, Aaron Kesselheim, professor of medicine at Harvard and a leading expert on the FDA’s drug approval process, called it “probably the worst drug approval decision in recent U.S. history.”

The upshot was a public-relations disaster for the FDA and the drug. The press sniffed out a number of departures from agency protocol, including a plan that Biogen dubbed Project Onyx in which its scientists and FDA managers, including Dunn, worked together to get the drug approved. A congressional investigation concluded that Dunn was inappropriately close to Biogen and that the approval was “rife with irregularities.” Some prominent hospitals refused to provide Aduhelm to patients, and in early 2024, Biogen announced it was discontinuing the drug.

When the next drug in the class, Leqembi, came up for review in 2022, it looked like Aduhelm in many respects. Both drugs were developed by Biogen and Eisai (with Eisai taking the lead role for Leqembi). Like Aduhelm, Leqembi’s first clinical trial failed to confirm that patients taking the drug did better than those on the placebo, according to the company’s primary cognitive test. Eisai used secondary effects, including removal of amyloid, to apply for accelerated approval, and the FDA approved the drug in January 2023 — based on the results for just 152 patients.

At the same time, there was evidence the drug caused serious harm. The FDA detected complications associated with Leqembi, complications the company downplayed or hadn’t included in their summary of “adverse events.” The FDA noted that Eisai hadn’t reported symptoms caused by brain swelling in two patients (one had a seizure) and left out two cases of brain bleeds.

Conflicts of Interest

Before a second, larger study of Leqembi came up for review, the FDA replaced its eleven-member advisory committee, which had consisted mostly of independent academic physicians who had rejected Aduhelm. The new, six-member advisory committee included four physicians, a biostatistician, and a patient representative.

According to our analysis, three of the four physician advisers had financial conflicts of interest with Biogen, Eisai’s partner in developing Leqembi, including receiving consulting fees, speaking honoraria, and research funding from the company. A fourth adviser was the CEO of a drug-development consortium whose members included Eisai, Biogen, and Eli Lilly. The consortium receives funding from drugmakers and claims as one of its successes that it “drove crucial progress” for the approval of Aduhelm and Leqembi.

FDA regulations call for advisors to be as independent of drug companies as possible. While their financial ties don’t necessarily suggest malfeasance, studies have shown that researchers and physicians with such industry connections are more likely to view drugs favorably than experts with no conflicts of interest.

Since the newly constituted advisory panel gave a unanimous thumbs up to Leqembi, there was no spectacle of public resignations over its FDA approval, and the drug was not subjected to the same drubbing in the press as Aduhelm. In fact, coverage was celebratory. Time magazine anointed the drug as one of the “best inventions of 2023.”

At a March 7, 2024, news conference, Eisai announced a coordinated plan to boost sales of Leqembi, including the planned release of a new blood test for Alzheimer’s disease that would rapidly identify more patients. The blood test, along with the development of an injectable version of the drug that patients can use at home, could help move treatment into the hands of primary care providers.

But the biggest increase in sales could come from the results of a clinical trial testing Leqembi in people who have no memory problems but could be at risk for developing the disease because they have a genetic predisposition or signs of amyloid deposits in their brains. At the news conference, the company estimated that its potential market would roughly double to six million potential customers by 2032 if the FDA approves Leqembi for this “preclinical” use.

But the risks and effects of Leqembi (along with Kisunla, which was also widely praised when it came out this past summer) are broadly similar to those of Aduhelm, which is no longer on the market.

“All of the new Alzheimer’s drugs — Aduhelm, Leqembi, and Kisunla — are essentially the same,” George Perry, a neuroscience professor at the University of Texas at San Antonio and the editor in chief of the Journal of Alzheimer’s Disease, told the Lever. “They work by the same mechanism, and they demonstrate the same insignificant efficacy and serious harms.”

“Toxic and Potentially Lethal”

In 2019, a woman in her late sixties named Monique was living with her husband Richard on the outskirts of Paris when concerns about memory loss drove her to see a neurologist. (Richard requested we use only their first names.) She and the neurologist discussed a clinical trial of Leqembi, and she enrolled. After the eleventh infusion of the drug, Monique was taken to the hospital while experiencing a life-threatening string of seizures, which her neurologist determined were caused by brain swelling tied to Leqembi.

A cascade of new problems followed the seizures, keeping Monique in the hospital for months. When her family was allowed to see her, Richard recalls, “She was delirious, hallucinating, pulling on her medical equipment, her limbs tied to the bed rails, suffering, not recognizing us.” Once at home, Monique descended rapidly into dementia. The few moments when she is now lucid are difficult in a different way, Richard said, because she glimpses the reality of her condition.

Five months after Monique’s admission to the hospital, another Parisian, Nicole Nicolle, age seventy, was in the emergency room, having lost half her vision in each eye. A CT scan showed a massive cerebral hemorrhage. She too was in the Leqembi study and was receiving the drug. Within eighteen months, she had progressed from mild cognitive impairment to advanced dementia, a process that ordinarily can take up to a decade. Today her daughter Céline Marzin says her mother rarely recognizes the family. “She’s in her own world,” Marzin told us. “Nobody can get in.”

Both swelling and bleeding of the sort that Monique and Nicolle experienced are a result of the drugs doing their job, Northwestern’s Castellani says. Excess amyloid in the brain often lives in the walls of blood vessels, and when it’s removed, “it can be like tearing a scab off a wound,” he said. Blood vessels are shredded and become inflamed and leaky, which causes brain swelling and hemorrhage. In severe cases, damaged blood vessels begin to die. As Castellani put it, “The bottom line is you are dealing with a toxic and potentially lethal drug.”

Many patients on the drugs suffer smaller brain bleeds and less swelling than Monique and Nicolle. The manufacturers assert that these usually “resolve” or subside on their own. Some don’t cause symptoms. But nobody knows the longer-term consequences of these incidents — whether, for instance, they might speed up the progression of dementia — especially when they occur repeatedly. To know for sure, experts say researchers who aren’t being paid by the companies need information on patients who have experienced brain bleeds and swelling.

While Thambisetty was still at the National Institute on Aging, he and other researchers asked Eisai to release information on these complications, to no avail. The company said it will not provide patient data while the drug is under global regulatory review. But that’s “exactly when the data are needed,” said Perry of the Journal of Alzheimer’s Disease. Without it, regulators cannot review analyses by independent sources.

Even regarding the prevalence of Leqembi’s most dire potential side effect — death — there are uncertainties. In December 2022, Eisai published the results of its key clinical trial of Leqembi in the New England Journal of Medicine, in which the authors stated: “No deaths were considered by the investigators to be related” to the drug. They added, “Longer trials are warranted to determine the efficacy and safety of [Leqembi] in early Alzheimer’s disease.”

But a more thorough investigation of the drug’s complications didn’t appear to be a priority for Eisai; no autopsies were reported by the company on the six who died during the main phase of the study while taking the drug, out of the 898 total participants on the drug. Skipping an autopsy is common when the cause of death seems clearly unrelated to the drug in question, but it is potentially very important when a drug can suddenly damage the brain, and when death could be incorrectly attributed to something other than the drug.

According to Bryce Vissel, head of the neuroscience and regenerative-medicine program at St Vincent’s Hospital in Sydney, “Without an autopsy, there’s no way to exclude the possibility that the drug contributed to or even caused their deaths.” (Eisai does not routinely ask families if they would like to have an autopsy when a patient dies but says that it requests copies of any autopsy reports conducted by independent pathologists.)

Shortly after the main phase of the trial was completed, autopsy results emerged for two patients who died while taking Leqembi during the extension phase of the study. Both implicated the drug.

One of the patients was Genevieve Lane, the Florida patient who died six weeks after beginning infusions of the drug. Her autopsy was performed by a team of Vanderbilt University doctors led by neurologist Matthew Schrag, an assistant professor of neurology. He and eleven colleagues published their findings in the medical journal Nature Communications, documenting swelling and hemorrhaging, known effects of the drug. The paper included images and a video of the blood vessels in Lane’s brain.

Schrag sent his findings to Eisai and requested additional information. But the company’s response, he said, “was very slow” and failed to arrive in time. Based on documents Eisai submitted to the FDA, the agency stated that “firm conclusions cannot be made” about Lane’s cause of death, because the company told regulators it was unable to obtain “critical documents substantiating [the Schrag team’s] findings.”

In an email to the Lever, Eisai called Schrag’s conclusions from the autopsy an “opinion.” When asked what steps the company took to confirm or refute the findings, an Eisai spokesperson replied, “We generally do not interact directly with independent researchers as we do not want to be in a position to influence or bias their opinion.”

The findings of another autopsy, performed on Jean Terrien, an athletic sixty-five-year-old woman who died in late 2022, suggested the drug played a role in her death. Terrien was rushed to a Chicago emergency room, confused and unable to speak. When she died four days later, Castellani, the Northwestern neuropathologist, performed the autopsy. He reported that she died of extensive brain bleeds similar to those of Lane.

Eisai also called Castellani’s autopsy findings an “opinion,” suggesting that Terrien died of a stroke and brain hemorrhage caused by a clot-busting drug used to treat a presumed stroke. Castellani says, “She didn’t have a typical stroke; she had a stroke mimic” caused by Leqembi’s side effects. He told us her death resulted from a cascade of events triggered by the drug, adding, “If she hadn’t been given Leqembi, she wouldn’t have had the stroke mimic that led doctors to treat her with a clot buster.”

In May 2024, Eisai reported that four of the nine deaths during the extension phase were “possibly” related to the drug.

Patients also died during trials of Kisunla, the newest Alzheimer’s drug. Eli Lilly, the maker of the drug, submitted its data to the FDA in 2022. The agency noted an “imbalance” in deaths: seventeen patients (2.7 percent) taking Kisunla died during the clinical trials, compared with ten (1.4 percent) on a placebo. Furthermore, it turned out that Eli Lilly had stopped tracking 391 patients who dropped out of the 1,736-patient trial. People in clinical trials often drop out when they suffer side effects, and some may subsequently die. Failing to include those dropouts in the analysis of the data can make a drug look safer and more effective than it is.

The FDA turned down the company’s first attempt to get Kisunla approved, citing the large amount of missing data, and instructed it to track down the missing patients.

To do so, Eli Lilly turned to an unnamed third-party vendor, which dug into the missing patients. The vendor tracked down half of them, reporting two additional deaths among participants who took Kisunla and another five among patients on placebo. The findings somewhat narrowed the difference in deaths between the two groups (mortality was 2.5 percent with Kisunla and 1.9 percent with the placebo). However, several experts told the Lever that the lack of transparency about the third-party vendor and its methods make it impossible for independent researchers to assess the reliability of the new data.

Nevertheless, the FDA approved Kisunla last summer on the basis of its effect on cognition, with the proviso that Eli Lilly conduct a safety study after the drug was on the market. Such an arrangement is not uncommon. The agency has become increasingly willing to let companies prove their drugs’ safety and effectiveness after approval. Although the company has to file biannual safety reports with the FDA, the final report on Kisunla is not due until February 2037.

Diana Zuckerman, the president of the National Center for Health Research, a Washington, DC–based nonprofit, told us in an email that this timeline is “unacceptable.” Establishing whether a drug increases deaths and irreparable harm should take just a few years, she wrote, not thirteen.

But Do They Work?

What do Leqembi and Kisunla actually do for memory and mental capacity? The makers of the drugs do not claim that they improve these functions in Alzheimer’s patients, only that the drugs slow the rate of decline. But even using that standard, the evidence suggests the drugs do very little.

In assessing Leqembi during its trials, Eisai used a widely accepted test called the CDR-SB (short for “Clinical Dementia Rating — Sum of Boxes”) to measure changes in patients’ cognition. The test asks questions such as: “Do you remember your wedding date?” and “How well can you perform daily tasks like getting dressed?” Test scores can range from zero to eighteen points. The difference between patients who received Leqembi and those who were given a placebo after a year and a half of treatment was 0.45 points.

This is a small difference. In 2019, Eli Lilly published a study involving data from approximately 35,000 patients who were given the CDR-SB test annually for one year or more. That study concluded that for a drug’s effects to be noticeable to patients and their caregivers, the patients’ scores on the CDR-SB had to change by at least one full point for those with mild cognitive impairment and by 1.6 points for those with early Alzheimer’s. Since then, these figures have been accepted as the unofficial standards for “minimal clinically important difference.”

Leqembi failed to meet that standard. Kisunla, which produced a difference of 0.7 points, also failed.

That hasn’t stopped the drugmakers from touting seemingly large benefits. In a press release, Eisai says Leqembi slows the progression of dementia by 27 percent. Eli Lilly says Kisunla slows the decline by as much as 35 percent.

The math behind these claims can quickly become numbing, but an analogy may help. If a weight-loss drug were given to obese patients and those patients lost two pounds in a year while patients on a placebo lost one pound, the drugmaker could say that patients on the drug lost 100 percent more weight than those on a placebo. That same kind of math yields a 27 percent difference between patients who took Leqembi (in the single positive trial) and those taking a placebo, as well as a 35 percent difference in the Kisunla trial.

In the case of Leqembi, says Lon Schneider, a professor of psychiatry, neurology, and gerontology at the University of Southern California who consults for manufacturers of anti-amyloid drugs, 27 percent is “a meaningless number. It’s a mathematical game that misuses the data.”

As Jerome Hoffman, a professor emeritus at UCLA and an international expert in clinical trial analysis, told us, “Twenty-seven percent of almost nothing is still almost nothing.”

The drug manufacturers sometimes also cast the drugs’ effects in a temporal way, saying that the drugs slow the process of cognitive decline by months. But Alzheimer’s is a slow-progressing disease. And while those tiny statistical differences might be observed over a period of, say, six months, they are differences that are too small to be perceived by patients and caregivers. The serious side effects, however, are quite discernible.

Right before the FDA was scheduled to consider the approval of Leqembi, the Alzheimer’s Association, the largest nonprofit devoted to the disease, took action.

The Alzheimer’s Association lobbies on behalf of its members, supports research on Alzheimer’s, and provides education and referral services to Alzheimer’s patients and their families. The group has aggressively promoted all three of the new drugs.

In 2023, the nonprofit took in $6.4 million from thirty drugmakers, with donations from Eisai, Eli Lilly, and Biogen making up fully half of that amount. The association says donations from the medical industry amounted to only 1.6 percent of its $398 million annual revenues listed in its 2024 budget. However, the organization acknowledged in an email that approximately 30 percent of its $17.6 million budget for scientific conferences and meetings came from drug and medical device companies.

When early results from Leqembi began to emerge, the association assembled a panel to reconsider the accepted standard for a meaningful or “minimal clinically important” difference in CDR-SB test results. After reviewing early study results, the panelists concluded that in light of the “modest” benefit being produced by the drug, “our expectations . . . may need to be modified.”

They suggested that a change of as little as 0.5 points on the standard test, rather than one point, could be considered a meaningful difference. There was no new research conducted to explain the decision in the panel’s report, which was published just before the FDA was slated to make its final decision on Leqembi. Instead, the new standard was based on the impressions of nine physicians and researchers, seven of whom had financial ties to companies that are developing or marketing anti-amyloid drugs.

The FDA told the Lever that it took the new definition of a clinically meaningful difference into account while making its decision. Although Leqembi didn’t clear even the new, lower bar, the FDA approved the drug in July 2023.

“Tethered to the Hospital”

The international response to the new Alzheimer’s drugs has been mixed. Australia’s regulators and the European Medicines Agency, the regulatory authority for the European Union, both refused at first to approve Leqembi. The European Medicines Agency said the small benefit “does not counterbalance the risk of serious adverse events.” Eisai appealed that decision and, in November 2024, won approval for Leqembi, provided that doctors don’t prescribe it for the highest-risk patients. Kisunla, the newer drug, is still under review in Europe.

The United Kingdom’s National Health Service subsequently published its decision not to pay for Leqembi. Canada still has not made a decision about the drug. China, Israel, the United Arab Emirates, South Korea, and Japan have all approved the drug.

Beyond concerns about sometimes fatal side effects and minimal benefits for patients, the drugs are extremely expensive. In April, the Centers for Medicare and Medicaid Services estimated that Medicare would spend $3.5 billion on Leqembi in 2025.

All of this raises the question of why the drugs were approved in the first place.

Some of the FDA’s apparent eagerness to approve the new drugs may be chalked up to the fact that Alzheimer’s can be such a dreadful condition, affecting more and more Americans as the baby boomers age, and the agency is under pressure from all sides to address it.

The Alzheimer’s Association and other patient advocacy groups routinely complain that the FDA is dragging its feet, and they lean heavily on the agency to approve drugs faster and faster, even those that others see as problematic. Together with drug companies, many advocacy groups have lobbied Congress to pass legislation such as the 21st Century Cures Act of 2016, which has steadily weakened the FDA’s authority to require rigorous scientific standards before drugs can be approved.

From 2019 to 2021, when Aduhelm was under review at the FDA, Biogen doubled its lobbying expenditures from just under $2 million to more than $4 million a year. When Leqembi came up for review in 2022, Eisai’s lobbying expenditures jumped from $280,000 in 2021 to more than $1.7 million.

The drug industry has helped fuel public perception of the FDA as an impediment in part by spending millions of dollars a year on patient advocacy groups. Those millions influence the messages the groups provide to Congress and the press.

These various forces on the FDA not only produce questionable drug approvals, but they also close off other avenues of progress. The entrenchment of drugs that destroy amyloid plaques has guided funding decisions at pharmaceutical companies and the National Institutes of Health, leaving other possible causes of Alzheimer’s disease inadequately explored. Moreover, the money Medicare will spend on the new Alzheimer’s drugs can’t go toward easing the burden of caring for Alzheimer’s patients.

“Patients taking these drugs are tethered to the hospital,” says Vanderbilt University’s Schrag:

They have to undergo extensive testing, including a PET scan or spinal tap, to qualify for the treatment, then physically come to an infusion center every two to four weeks to receive the treatment and return multiple times in the first year for MRIs. If they develop side effects, which could range from headaches to disabling neurological symptoms, they may need even more testing.

This early phase of Alzheimer’s, says Schrag, is when doctors should be encouraging patients to address items on their bucket list. “Tell them, ‘If you want to travel or spend time with your grandkids, go do it now,’” he said. “‘This is a time when you are at your best. You’ll lose the opportunity to do other things if you are running back and forth to the hospital.’”

There is much we still do not know about the new Alzheimer’s drugs. Studies are ongoing, and assessments of both benefits and harms may change, for better or worse. Experts say open access to the manufacturer’s data sets would be a welcome development, allowing independent observers to study the drugs’ effects more closely.

But with billions of dollars in potential revenue at stake, drugmakers are not likely to release their data or abandon either the products or the questionable amyloid hypothesis. A business analytics firm projects that global sales of both Leqembi and Kisunla will total $5.5 billion in 2030.

Seven years after that, Eli Lilly’s final report on the safety of Kisunla will be due.