Article In Brief
Leading Alzheimer's disease researchers look at why BACE inhibitors have failed in clinical trials but they say there are reasons why they're not giving up on the drug class.
LOS ANGELES—The news has not been good for beta-site amyloid precursor protein cleaving enzyme 1 (BACE) inhibitors. After a series of trials were terminated early after patients receiving treatment had worse cognitive decline than those getting placebo, the Alzheimer's disease (AD) research community finds itself pondering: How are these drugs, time and again, seemingly doing what we want them to do—limiting new amyloid-beta (Abeta) production—but having the exact opposite of the desired effect on cognition? And where does the field go from here?
A group of researchers who participated in the failed trials—along with others in the Alzheimer's research community—came together here at the Alzheimer's Association International Conference in July to brainstorm solutions to these questions and more. The session grew out of a task force on BACE inhibitors—for beta-site amyloid precursor protein cleaving enzyme—assembled by the Alzheimer's Association.
“This has been an amazing, frustrating road,” Barry Greenberg, PhD, professor of neurology at Johns Hopkins, told the panel during the audience participation segment. Despite the frustration, the consensus was that the drug class is worth further exploration, particularly in lower doses and earlier in the disease course.
The two trials of the Novartis drug, umibecestat, also known as CNP520, were the latest to be halted. In the second week of July, an unblinded review of data found that some measures of cognitive function had worsened in the treatment groups compared with placebo. Participants were cognitively unimpaired but positive for the apolipoprotein E4 that predisposed them to AD. The trials were conducted with Amgen and Banner Alzheimer Institute.
“I thought I would be in a different place than I am right now,” said Ana Graf, MD, senior global program head of Alzheimer's disease programs at Novartis, as she took the podium for a presentation focused on morphological changes in rats and mice treated with the drug. Data on cognitive function probably won't be presented until at least the middle of next year, she said, adding blinded follow-up visits will be continued through then.
“We do think it will be an important contribution to the field to keep the blind and to have these follow up visits,” she said.
The halting of the umibecestat trial came on the heels of disappointing results for lanabecestat in patients with mild cognitive impairment due to AD and mild AD dementia, atabecestat for the cognitively unimpaired with pathologic amyloid levels, and verubecestat for prodromal AD.
The lanabecestat findings in AstraZeneca and Eli Lilly's AMARANTH trial showed that despite “substantial reduction” of cerebrospinal fluid Abeta levels and “robust target engagement,” worsening was seen for the two dose groups on a battery of tests, which assess a variety of cognitive functions. But there seems to have been improvement for the treatment groups in letter and category fluency.
Researchers on Janssen's atabecestat trial—stopped early due to problems with liver function but in which similar cognitive worsening was seen in the treatment groups—found that cognitive function improved after treatment was stopped. Those in the 25 mg group, for example, had a worsening of 1.23 points in the Preclinical Alzheimer Cognitive Composite scores from baseline to the end of treatment, but improved by 1 point from last treatment out to six months off treatment.
Data from the trials sometimes showed cognitive worsening that happened quickly. In Merck's verubecestat trial, both the 12 mg and the 40 mg treatment groups had significantly worse scores than placebo at 13 weeks on the Composite Cognition Score-3 Domain measure (CCS-3D) and Alzheimer's Disease Assessment Scale-Cognitive (ADAS-Cog) measure.
Where There Is Hope
Hope remains—though perhaps tempered by the recent disappointments—for Eisai's elenbecestat, a BACE inhibitor selective for beta-secretase 1. In non-clinical studies, it reduced amyloid-beta in the cerebrospinal fluid without a loss in dendritic spine density—a feature related to cognitive impairment in AD seen with other drugs in the class.
In a phase 2 trial in patients with mild cognitive impairment due to AD and mild to moderate AD, the drug has produced less worsening than placebo on Clinical Dementia Rating-Sum of Boxes (CDR-SB) and ADCOMS—a measure compiling results from four cognitive assessment scales—although these differences weren't found to be significant through 79 weeks (p= 0.55 and 0.38, respectively).
The worsening in cognition has typically been seen with targets of reducing amyloid by 50 percent or more.
But Reisa Sperling, MD—director of the Center for Alzheimer Research and Treatment at Brigham and Women's Hospital, who co-chaired the session—said she may pursue a lower target dose as a lead investigator in the prevention trials of the Alzheimer's Clinical Trials Consortium, which recently announced that it will be evaluating elenbecestat in the trials.
“For me, seeing all these data, it's critical to understand whether there is a safe dose if we go lower—so I actually think we should be trying a 30 to 35 percent reduction based on all of the data,” she said. “I'd already wanted to test a low dose and this pushes me even more in that direction.”
Even more essential, she said, is the question of what is causing the greater cognitive deficits in these treatment groups.
“We absolutely need to understand whether this is related to the lowering of amyloid or something off target,” she said. “I'm still not sure exactly how to do that but I absolutely think that's the most important next step.”
Mike Egan, MD, vice president in neuroscience for global clinical development at Merck, said there might be reason to doubt that a safe lower dose will be found. He said that their own pharmacokinetics and pharmacodynamics analyses have found that lower exposures have not been better, although researchers with the company haven't actually dosed for a target below 50 percent Abeta lowering.
“I think that's something to be concerned about,” Dr. Egan said. “Maybe you won't see a deleterious effect at 40 percent or 30 percent, but we're at 50 to 60 percent and we're seeing a deleterious effect.”
Robert J. Vassar, PhD, director of the Northwestern Alzheimer's Disease Core center and scientific director of behavioral neurology in the department of neurology, said the best dose is still unknown.
Higher doses bring about “significant cognitive worsening, but around the 50 percent level it's only a trend,” he said.
“What we don't know is what level is going to be safe enough,” he said. “We know that Abeta lowering of around 20 to 30 percent is probably over the long term going to delay the onset of amyloid deposition.... But what levels can be safe is still an open question.”
Dr. Greenberg of Hopkins said that the genetics on amyloid precursor protein (APP) and Abeta that impacts disease relates to substrate specificity, while BACE inhibitors—and before them, gamma secretase inhibitors—involve enzyme specificities.
“Maybe the amyloid hypothesis is fully valid and still has not been tested because we're inhibiting enzymes that are impacting on a wide range of substrates and we're only looking at one of them,” he said.
Dr. Vassar said that there are at least a few dozen substrates that are cut by BACE. APP is actually fairly minor, he said, while others—including seizure protein 6, which is involved in synaptic spine formation—are probably more important.
“I think it's really important for us as a scientific community to investigate further the role of these substrates and how can we manage to get some substrate selectivity out of these compounds by lowering the dose,” he said.
Looking for a Broader Picture
Robert Stern, PhD, director of the clinical core at the Boston University Alzheimer's Disease Center, said that looking at results from composite tests doesn't give a full picture of what is happening with patients on BACE inhibitor treatment, such as improvements on category and letter fluency but worsening on other measures for patients on lanabecestat.
“Both of those fluency measures do not require incidental learning,” he said. “They don't require speed of information processing. They don't require attention. What they require is speed of semantic access and generativity, which is one component of executive function. One thing I'd want to caution is, when trying to interpret any of these findings, when we start at a global measure, that is going to have the good stuff, the bad stuff, all combined—and you might not be able to detect any differences.”
The researchers agreed that continuing to share information that normally might be kept under wraps has been the best approach to tackling the problem—the Novartis trial might not have been stopped if they hadn't been alerted to heighten monitoring as a result of the task force's work, Dr. Sperling said.
She said she hopes that the pharmaceutical companies will share samples from their trials so that spinal fluid and blood could be analyzed for even more markers that might correlate with cognitive worsening.
“Often it's hard to get companies to share data, especially about negative effects,” she said. “So I think this BACE inhibitor task force has been a positive thing.”