An investigational Alzheimer’s drug significantly slowed patients’ cognitive decline in a large clinical trial, unexpected and unprecedented results that could move the treatment’s makers to roll the dice on an early approval.

In the study, the highest-tested dose of a drug from Biogen and Eisai was markedly better than placebo at delaying the memory-destroying effects of Alzheimer’s. After 18 months, patients who got the drug, called BAN2401, performed 30 percent better on a cognitive test than those getting placebo.

The data, presented here at the Alzheimer’s Association International Conference, came as a positive surprise to neurologists, who have spent years watching once-promising therapies fail in clinical trials.

But there are caveats. The Phase 2 trial, which employed multiple statistical measures, failed its primary goal. Four doses of BAN2401 didn’t outperform placebo, and the high dose was tested on just 161 patients. Furthermore, the metric Biogen and Eisai used to measure mental acuity is a homegrown composite that has never before been used to win Food and Drug Administration approval.

“I’ll remain cautiously optimistic,” said Dr. Ronald Petersen, director of Mayo Clinic Alzheimer’s Disease Research Center. “I think the data are intriguing. The effect sizes sound reasonable, the drug seems safe, and on the biological side of it, the drug seems to be working.”

But, he added, “you’d really want to see a Phase 3 to replicate those results.”

Whether he’ll get one remains an open question.

Despite persistent questions from Wall Street, Biogen and Eisai have not said whether they will take up a larger study to tease out BAN2401’s benefits or go straight to the FDA with the Phase 2 results. Earlier this month, Dr. Lynn Kramer, chief medical officer of Eisai’s neurology division, said the company considered the Phase 2 trial to be “late stage” and that it would discuss the possibility of using it to win approval when it meets with regulators later this year.

It’s unclear whether the FDA will lower the traditionally high bar set for Alzheimer’s therapies, which have in the past needed two Phase 3 trials to merit approval.

If BAN2401 pans out, it could end a generation of bitter disappointment in Alzheimer’s, which has not seen a new approved treatment in more than 15 years. Time and again, therapies have showed promise in early testing only come up empty in the final stages, frustrating patients, doctors, and scientists. Meanwhile, the disease only gets more prevalent, affecting more than 5.7 million Americans and costing the country more than $250 billion a year, according to the Alzheimer’s Association.

The BAN2401 results will be warmly received by investors. Biogen’s share price has risen by nearly 30 percent over the past month as analysts predicted modestly positive data from the Phase 2 trial. The actual results outstripped Wall Street’s consensus, putting Biogen on a path for further growth.

BAN2401 is part of a cadre of Alzheimer’s treatments predicated on the idea that toxic plaques called amyloids are responsible for the disease’s corrosive effect on the brain. The so-called amyloid hypothesis is supported by reams of genetic data, but no drug aimed at those plaques has ever managed to significantly affect the disease in a large trial.

That’s what made BAN2401 such a surprise. In the Phase 2 trial, the highest dose of the drug had a marked effect on amyloid buildups in the brain, with 81 percent of patients going from amyloid-positive to amyloid-negative. And in addition to its effect on Eisai’s proprietary measure of cognition, the dose registered a 47 percent reduction in cognitive decline compared with placebo on a well-regarded metric called ADAS-cog.

“That’s the thing that really makes me feel like they’re onto something, that this is a real effect that we’re seeing,” said Dr. Jim Hendrix, director of global science initiatives at the Alzheimer’s Association.

The next step, he said, should be to enroll another, larger trial that isolates the best BAN2401 dose and tests whether it can replicate its promise.

Republished with permission from STAT. This article originally appeared on July 25, 2018