In the hunt for a drug to treat Alzheimer’s disease, even a whiff of success can be intoxicating. That helps explain why an experimental drug called BAN2401, which a few months ago seemed like it might join a growing heap of failed candidates, created so much buzz yesterday at the Alzheimer’s Association International Conference in Chicago, Illinois.
In a phase II trial, the drug had already failed to show the level of benefit that its developers—Biogen Inc. in Cambridge, Massachusetts; and Eisai Co. Ltd. in Tokyo—set as the study’s primary endpoint. But yesterday the companies presented a series of other analyses from the same trial that suggest BAN2401 might slow the pace of cognitive decline in Alzheimer’s patients, and reverse the buildup of a brain protein thought to drive the disease’s neurodegeneration. But the subset of patients who showed those benefits was relatively small—161 people—and an unexpected change to the way the study was randomized cast some skepticism on the results. For many, the findings are too preliminary to celebrate.
“If these results we saw today pan out in phase III clinical trials, then you’re looking at disease-modifying medication—the first one for Alzheimer’s disease,” says Keith Fargo, director of scientific programs & outreach at the Alzheimer’s Association in Naperville, Illinois. “But you don’t know whether they’re going to pan out until you actually do the phase III trial.”
Like several other Alzheimer’s drugs under development, BAN2401 targets ß-amyloid, the protein fragment that forms sticky plaques around neurons. Many scientists believe ß-amyloid is a key culprit in Alzheimer’s disease, blocking communication between neurons and eventually killing them. BAN2401 is an antibody that binds to and helps clear out ß-amyloid structures known as protofibrils that are in the process of clumping into plaques.
The trial included 856 patients with a mild, early form of Alzheimer’s, and relied on a new measure of cognitive function—the Alzheimer’s Disease Composite Score (ADCOMS)—designed to detect subtle changes in people with early-stage disease. The companies also decided on an innovative and complicated study design. The trial would be “adaptive”; instead of randomly assigning an incoming participant to one of five treatment groups, each getting a different dose of the drug, the assignment system increased the likelihood that a patient would receive the dose that seemed to be working the best at the time they entered the study.
The researchers also relied on a statistical method—known as Bayesian statistics—to analyze how the different groups of patients were responding to the drug while the trial was underway, rather than waiting until it was complete to do the statistical analysis. That approach was meant to give researchers an earlier hint about whether the drug was likely to work. This was the first use of Bayesian statistics in an Alzheimer’s trial, Fargo says. “I think the company felt very good about this drug,” he says, and thought Baysesian analysis could enable “a faster decision about whether or not to go to phase III.”
But the strategy seems to have backfired. That early analysis, 12 months into the treatment, was supposed to reveal at least an 80% probability that BAN2401 reduced the rate of cognitive decline by 25% or more, compared with a placebo. The drug didn’t clear that bar, Biogen and Eisai revealed in December 2017. Instead, the probability was judged to be 64%, explained Lynn Kramer, chief clinical officer of Eisai’s neurology division, in yesterday’s presentation—“relatively close,” he said.
But after following patients for another 6 months, the companies were much more optimistic. They announced earlier this month that in some patients, the drug had significant effects after all. And today’s presentation was the first glimpse at the size of those effects: After 18 months, the 161 patients getting the highest of five doses tested in the trial had 30% slower cognitive decline than those getting a placebo, as measured by the ADCOMS. That group saw a 47% slower decline than the placebo by another, more traditional cognitive measure, known as the Alzheimer’s Disease Assessment Scale-cognitive subscale. Brain imaging also revealed that the drug reduced levels of amyloid plaques in all the dosing groups, and that 81% of participants were judged to be “amyloid negative” in a brain scan after 18 months of treatment.
But Kramer also described an unanticipated and potentially problematic change to the trial. The investigators monitored patients for swelling of the brain, a potential safety risk of this and other antiamyloid antibodies—particularly in patients with a gene called APOE4, which increases risk of Alzheimer’s and is linked to faster cognitive decline. In July 2014, a regulatory agency outside the United States—the companies didn’t specify which one—requested that the trial stop assigning these APOE4 carriers to the highest dose (and potentially highest risk) arm of the trial. The company complied with that request.
That move not only reduced the size of the high-dose group, but it created a potentially confounding variable: Did that group show a slower mental decline because the drug was working, or because it contained fewer people genetically prone to decline quickly?
In part because of that change, “I think probably people aren’t quite as excited as they were beforehand,” Lawrence Honig, a neurologist at Columbia University’s Vagelos College of Physicians and Surgeons, says about yesterday’s presentation. He was not involved in the study, but is an investigator on other studies of drugs developed by both Eisai and Biogen. The change to the study’s randomization “does affect the interpretation a little bit,” Honig says, although, “even without these caveats, [the trial] would not be conclusive.” He notes that other Alzheimer’s drug candidates have looked promising in phase II and failed in a larger study.
It’s not yet clear what kind of study Biogen and Eisai will need to run next to prove the drug’s effectiveness to the U.S. Food and Drug Administration. Their follow-up will be one of a few closely watched efforts in this class of drug candidates. The companies are also jointly developing a different amyloid-targeting antibody, aducanumab, which catapulted straight into a large phase III study in 2015 after promising indications of benefit in just 125 patients. That study is expected to produce results in 2020.