The Alzheimer's R&D Odyssey Comes to Another Fork in the Road
With Phase III failures starting to offer hard lessons, the R&D field starts to refocus on ever-earlier stages of the disease. And the FDA is helping.
From a distance, the failure of Merck’s BACE drug verubecestat in back-to-back Phase III studies may have looked like just another setback in a 15-year blur of clinical dead ends, missteps and blunders.
Get up close, though, by mixing among the researchers who did the work, and you’ll get a much different perspective — a view that reflects the latest distinct shift in the gradual evolution of Alzheimer’s research.
I talked with Matt Kennedy, Merck’s head of early discovery for neurosciences, who has been working on Merck’s BACE efforts since shortly after he joined the company 19 years ago. By the time the Merck group was ready to start two Phase III studies with verubecestat, he says, they felt they not only had the right drug, but also the right technology for selecting the right trial patients and tracking their responses.
They also believed they were using the right approach. Their philosophy: “If we invest in the right clinical trial — which we feel we conducted — it would provide unequivocal evidence of whether the drug would work,” Kennedy says.
Without doubt, he says, the drug worked as planned in patients with mild to moderate Alzheimer’s, as shown in the first of verubecestat’s Phase III trials. By moving upstream of the toxic clusters of amyloid beta that clog the brains of many — though not all — patients with more advanced disease, the researchers were able to throw a monkey wrench into the mechanics of its pathology.
As one researcher in the field told me, it was a way of turning off the faucet steadily dribbling into a tub.
The drug inhibited the development of plaques in the brains of patients. But it did not in any significant way stop the trajectory of the disease. Perhaps because the tub was already full.
If that sounds like just another failure, however, you’re not paying close enough attention. The verubecestat trial helped chart a new course for Alzheimer’s researchers.
“It provides very clear data so that we can now move and evolve our clinical strategy,” Kennedy says, “and others can as well.”
About five years ago we saw the first distinct shift of Alzheimer’s research programs into earlier diagnoses for the disease. If drugs had little chance of treating a disease once it had started causing obvious signs of brain damage, maybe they’d have a shot if therapy began early on, when patients could still be evaluated using standard tests for cognition and daily function. Turns out — as Merck discovered — research may need to move even further upstream.
Some are doing just that. The drumbeat of negative clinical data that continued with Merck’s first verubecestat trial is pushing everyone in the field toward combination approaches and pre-symptomatic Alzheimer’s disease. And it’s not just researchers. The FDA itself is gradually developing new standards for late-stage trials that will assist the ongoing sea change in research.
Not only are most major drug development players still refusing to quit, but at least one big outfit is making a calculated leap into the game. And the arrival of an upstart biotech company has grabbed the spotlight as it takes advantage of the shift to better technologies in clinical research.
While researchers are finally beginning to learn clear lessons from failure, Alzheimer’s disease continues to ruin millions of lives and threatens millions more. No drug on the market has anything but a temporary and transient effect on its symptoms. Alzheimer’s is both a blight and a beacon — a perplexing affliction with no clear trigger on hand, and a shot at a multibillion-dollar market waiting for any drug that can even partially derail its progress.
And if Phase III work in Alzheimer’s looks as if it has become the long-shot Powerball lottery game of biopharma, you’ll see why we likely have a long way to go before any major progress can be declared for patients.
More R&D failures to come
There will be no second tries at Merck on using a BACE drug to effect disease modification in Alzheimer’s for mild to moderate patients. The earlier, prodromal set of patients in the second Phase III study — those who are starting to see their memories wink out but still function normally — also failed to improve significantly with verubecestat, forcing Merck to call it off well ahead of schedule.
Kennedy hasn’t spent the time he needs to fully analyze the trial data. So he’s not ready to put a stake in its heart — yet.
But some researchers are.
David Michelson spent more than a decade in Merck’s neurosciences group, playing a big role in the Phase III study of verubecestat. Over that time, he saw plenty of evidence of poor pharmacology in Alzheimer’s studies, where nothing was proven one way or the other. The Merck trial may have proven a failure for patients, but echoing his colleagues still at the company, he says it was not a failure of design or execution.
“I think there’s one trial that’s definitive — the study we did with BACE. We showed target engagement, pharmacology, the near complete knockdown of BACE and amyloid beta production,” Michelson said. “The placebo group progressed as expected.”
“We showed no treatment effect,” says the researcher. “In patients who are symptomatic, prodromal or mild-to-moderate, it is not going to be a benefit for those patients.”
That doesn’t mean, however, that the drug won’t work in patients who have yet to show signs of the disease, says Michelson. For patients with symptoms, “[i]f you suppress production (through BACE), that doesn’t change the trajectory of the disease,” he says. “Where you have to be careful is to say that there is no intervention that will stop disease progression in asymptomatic patients.”
He’s not saying new drugs will work in that group, you just can’t rule it out without the hard data needed to test the hypothesis.
Michelson, who recently retired from Merck and opened a new chapter in his career at a biotech called Proclara, is also eager to find out whether combination approaches can now begin to demonstrate evidence of disease modification in patients with symptoms.
In the meantime, Merck’s failure raises some major implications for the BACE drugs that are following up behind verubecestat in the clinic.
One of the big BACE — beta-site APP cleaving enzyme1 — studies belongs to Eli Lilly and AstraZeneca. More than three years ago, Lilly paid AstraZeneca $50 million to partner on AZD3293. Lilly handed over $100 million more when they moved the drug into late-stage testing for early Alzheimer’s and launched a new Phase III for mild forms of the disease.
The same year Lilly and AstraZeneca joined forces on BACE, Biogen and Eisai allied to develop elenbecestat (E2609). That med is also in a Phase III program, testing the drug in patients with early Alzheimer’s symptoms.
Both of these programs follow the path Merck blazed. If Merck is right about its data, the chances of success for these follow-up contenders are extremely poor, and the two partnerships are shelling out hundreds of millions of dollars despite dismal odds of success.
But it’s not over yet.
BACE drugs just might work in a smaller group of patients with specific genetic characteristics, and that’s where the focus of the BACE solo theory is starting to shift. Companies are testing their drugs in asymptomatic patients with one or two copies of the risk-indicating APOE4 gene, or other genetic markers. Johnson & Johnson, for instance, is trialing JNJ-54861911 — which followed in the wake of bapi’s crash-up several years ago — in the DIAN-TU study involving patients with a genetic trigger for early-onset Alzheimers as well as the EARLY and A3 studies in elderly at-risk groups. Novartis/Amgen’s CNP520 is engaged in two GENERATION studies looking at the same population, and also in combination with an amyloid beta peptide-targeting vaccine CAD106, codeveloped by Novartis and Cytos. Merck may well follow their lead with a new effort on verubecestat.
The amyloid beta conundrum
Before verubecestat and the latest trials, the approach to patient selection and PET scans to ID plaques was notoriously imprecise. Phase III studies didn’t provide clear results. Did the drugs really fail, or did they litter each study with the wrong patients? Did the therapy hit the target? Did it even get into the brain?
Now there’s a renewed focus on even bigger questions.
Is amyloid beta a legitimate target for the disease?
The current generation of setbacks arose from a widespread commitment to the amyloid beta theory. After all, the plaques are a hallmark of the disease. It made sense that fighting those plaques would hinder it.
The first big studies weren’t enough to give that approach a thumbs-up or thumbs-down, thanks to uncertainty and fudged results. Meanwhile, drugmakers knew Alzheimer’s represented a vast and growing market of millions of patients who only have a few options, and those drugs only temporarily slow the progressive symptoms. It was almost irresistible to try again and again with drugs that appeared to be interfering with those toxic proteins.
Roche is doing that now with gantenerumab. Investigators stubbornly vowed, back at an international Alzheimer’s conference in 2015, that amping up the dosage of Roche’s amyloid beta antibody would deliver a real treatment effect, improving cognition and function. They had fired up two studies for early and prodromal patients.
That means they’re following the same path that Eli Lilly took on solanezumab, which failed three straight late-stage studies last year. And the litany of trial failures is beginning to shake analysts’ confidence in the second and third attempts at success for failed drugs.
The trend toward failure is disheartening.
“We’re ending up with a negative announcement every day now,” AC Immune CEO Andrea Pfeifer told me in the wake of the Merck failure. “The field is extremely discouraged.”
AC Immune is collaborating with Genentech on crenezumab, a monoclonal antibody that works differently from BACE drugs; it primarily blocks oligomers in the brain. Genentech advanced crenezumab into a Phase III study despite its failure in Phase II. Investigators in the Phase II were able to track a boost to cognitive and functional scores for patients diagnosed in the earliest stages — but here again some researchers in the field shake their heads at what they see as sketchy data at best.
Till the crenezumab Phase III data debut, the jury is still out, of course. But the series of BACE faceplants is already calling the target into question. “The whole story about BACE puts the amyloid beta hypothesis again on the discussion,” Pfeifer says.
One of the few amyloid beta Phase III drugs that still whips up analysts and researchers is Biogen’s aducanumab, but even there the growing doubts about any one drug’s ability to bend the curve of the disease by targeting amyloid beta is raising fresh doubts. Biogen decided to pursue its late-stage Alzheimer’s program for aducanumab despite mixed Phase II data and evidence of safety issues with ARIA. And the company managed to shake up everyone on Wall Street just by adding hundreds of patients to its pivotal program, to help improve the odds of coming up with data that could beat the p-value barrier.
The Merck setback didn’t surprise Pfeifer, as she and her team had already concluded that you had to start early in the pathology of the disease — taking a preventive course — for a BACE drug to work. For Pfeifer, the failure at Merck shows just how important it is for the field to get a better handle on earlier diagnosis, with a more careful approach to the early- and mid-stage research studies needed to determine whether a drug is working before launching a full-scale pivotal trial.
Eli Lilly learned this the (very) hard way. It followed a series of hopeful signs on solanezumab over the Phase III amyloid beta cliff on three separate occasions. Now it’s also finally given up on disease progression in early-stage patients and moved to a presymptomatic study to see if they can slow the arrival of full blown disease.
Other Alzheimer’s disease targets have flopped, too, leaving little room for doubt about their ineffectiveness. Before Merck completed its back-to-back flops for verubecestat, Axovant went down spectacularly in what is likely to be among the last attempts to influence Alzheimer’s symptoms by following the once-promising 5HT6 pathway. Vivek Ramaswamy convinced investors to back Axovant with hundreds of millions of dollars to try again with a drug that flopped badly at GlaxoSmithKline in 4 studies. And they all but gave it away for a $5 million upfront.
Lundbeck also helped sound the death knell for 5HT6 with three straight failures announced just ahead of the Axovant disaster. And Pfizer fired the warning shot with its decision to scrap a Phase II study for PF-05212377 two years ago after a monitoring committee saw nothing encouraging in the data.
Meanwhile, Takeda, in pursuit of a theory that a low dose of their diabetes drug Actos could prevent the progression of disease symptoms, gave up two months ago after an interim analysis offered little chance of success.
Described as the biggest Phase III in Alzheimer’s to date when it was fully enrolled in early 2016, Takeda’s TOMORROW study recruited more than 3,500 patients at 50 sites, using a genetic-based biomarker risk assignment algorithm from the small US biotech Zinfandel Pharmaceuticals, a Duke spinout, to identify the best subjects.
And a few weeks ago Boehringer Ingelheim discarded its Alzheimer’s program for a drug called BI 409306, another effort at modulating glutamate, which is one of the neurotransmitters that can run amuck in dementia.
In the most dramatic example of stepping away from failure, Pfizer — a company that years ago struck a $725 million deal to back a Russian antihistamine called Dimebon for Alzheimer’s, now failed — walked away from the disease and neurosciences in general, laying off 300 staffers and shutting down a unit that promised little and cost much. The motivation wasn’t cost-cutting. Pfizer just wanted to spend more on fields where it had a better shot at contributing to the bottom line.
But the companies still in the field are hanging on to amyloid beta. Nothing can beat Eli Lilly for sheer obstinacy in the face of compelling proof it was on the wrong track with solanezumab.
Indeed, most scientists at the pharma giants forging ahead are sticking with the science of neurodegeneration. If you close one research door, you open another.
Hence the combo-drug strategy, as well as the earlier-in-the-disease approach. How about targeting tau and neuroinflammation as well? And also go ahead of the symptoms?
No validated targets
Anyone who comes into Alzheimer’s R&D for the first time now might quickly determine that industry standards for driving a Phase III trial were tossed aside long ago.
“We don’t understand Alzheimer’s disease,” sums up Lon Schneider, a researcher at USC’s Keck School of Medicine and one of the most outspoken critics of bad trial design and execution. “We don’t have validated targets.”
There’s no lack of consensus, he adds, provided you break it into competing groups. Quite a few researchers have said with certainty it’s amyloid beta, the most popular culprit; others point to tau.
Yet Alzheimer’s is a heterogeneous, complex disease with a complex phenotype, says Schneider. There are probably multiple factors that lead to a common pathway.
“You can’t get around that people can have the pathology of Alzheimer’s disease and not have clinical symptoms,” he adds. “They can have a head full of amyloid and be at risk of developing dementia,” but then never experience symptoms of the disease.
He adds: “All of this speaks to highly imperfect knowledge and difficulty targeting the disease.”
The latest series of clinical defeats has emphasized that the old, haphazard approach is being called out as a nonstarter more quickly. And companies that could once attract and retain tremendous investor interest with the shakiest of theories built on the weakest data — a roll of the dice for vast gains with minimal investment — are being forced to account for higher levels of skepticism and rising standards for what a biopharma company can prove at Phase II — the new decisive hurdle for researchers in the field.
If headlines drive investment trends, they’re up against some steep odds.
The biotech approach
Axovant’s implosion, quickly followed by David Hung’s decision to depart from the biotech along with his team, underscores the particularly perilous journey that small biotechs face when taking on Alzheimer’s.
Forum Pharmaceuticals, backed by the Fidelity group that created F-Prime, was shuttered in the summer of 2016 after it took a shot at Alzheimer’s, only to be shut down by the FDA as researchers turned up safety issues with its lead drug.
Declarations of success are met with increased skepticism — and for good reason. A little over a year ago, San Diego-based Acadia Pharmaceuticals asserted that its drug Nuplazid (pimavanserin), a 5HT2A-targeting drug approved for Parkinson’s, hit its primary endpoint in a mid-stage study focusing on Alzheimer’s disease psychosis.
Critics quickly jumped on the fact that patients’ improvement in NPI-NH Psychosis scores at week 6 didn’t sustain those improvements to week 12 — and that’s a standard measure of success. A year later, when the company published its Nuplazid results, it was clear the drug had flunked every secondary and primary endpoint at week 12. After the data were all in, USC’s Schneider wrote off Acadia’s “success” as the result of nothing other than a temporary worsening in the placebo arm at week 6 in the more symptomatically severe subgroup.
Acadia shifted gears and is now studying the drug for disease-related psychosis in a group of ailments, with Alzheimer’s among them.
Despite the hazards, smaller biotechs are continuing to plug away, undaunted by the wreckage — and even undeterred by the lack of success in symptomatic disease.
Michelson retired from Merck not long ago, but he didn’t retire from Alzheimer’s research. His new job is chief medical officer at Cambridge, MA-based Proclara, and neither Michelson or Proclara are giving up on coming up with a treatment for patients who already exhibit symptoms of Alzheimer’s.
The company is developing a fusion protein that can simultaneously target the accumulated misfiled proteins of amyloid beta and tau. And they’re testing it in an early-stage study of patients with probable Alzheimer’s.
“There are no studies on stopping both amyloid and tau,” says Michelson, “so it’s premature to say we can’t help patients who are symptomatic.”
It’s not just small biotechs in the mix, either. Even as a major player like Pfizer was joining a retreat that long ago took GlaxoSmithKline out and seriously reduced efforts at AstraZeneca, at least one big new player has cautiously stepped up its activities.
Celgene started out small in neurodegeneration, but R&D chief Rupert Vessey recently highlighted its plans in a chat with analysts. Some experts in the field have been paying close attention as they’ve made contacts and explored new paths of research.
“[W]e have formed a small internal team that’s very expert in neuroscience and … we’ve been focusing on leveraging our strengths in protein homeostasis,” Vessey noted. “Last year, you’re aware, we formed a relationship with Evotec that was looking at that angle for neurodegenerative disease.
“And then also we really see ourselves as an immunology-expert company, for obvious reasons. … [N]euroinflammation is a key cause of neurodegenerative disease, and we’re in the process of evaluating approaches for that angle of therapy as well.”
Then, in back-to-back deals, Celgene started to ramp up a pipeline of neurodegeneration drugs. In particular, the big biotech kicked off a discovery alliance with Prothena with a $150 million up front that includes a program on tau.
The four ‘rights’
Axovant’s failure will make it infinitely harder for any other biotechs to try the same strategy of plucking a failed drug from pharma’s back-shelf duds. That was more than apparent when Alzheon recently filed for an $80 million IPO, with plans to use the money to pay for a Phase III study of a drug that was earlier discarded by Neurochem. Trial experts and analysts quickly trounced the idea on Twitter, raising a steep hurdle for the biotech to clear.
Investors, though, are still willing to take a chance when they feel that professionals are making a strong effort to try something new. Denali Therapeutics went from a record IPO in 2017 to a market cap of $2 billion, driven by a crew of ex-Genentech investigators who started out by equipping themselves with the latest technologies for piercing the blood-brain barrier and tracking results for the drugs they’re developing.
In a key exhibition of its blood-brain barrier tech, Denali recently spotlighted a non-human primate study where they were able to measure reduction of cerebral spinal fluid amyloid beta in cynomolgus monkeys after dosing anti-BACE1 antibody. Whether or not that is still a viable strategy in symptomatic patients, it helped illustrate how its antibody transport vehicle tech — ATVs — can work in getting a therapy where it’s needed.
It’s those kinds of basics that will be crucial for any company that wants to make a mark in Alzheimer’s.
Denali also just got its first human study underway for a RIPK1 inhibitor dubbed DNL747. RIPK1 is an enzyme that stokes inflammation and amps up levels of amyloid beta, making it one of the culprits that could gang up on patients to trigger the disease. And Denali is pursuing a variety of pathways to multiply their chances of a combination success.
For Denali CEO Ryan Watts and Carole Ho, the CSO, there are four key “rights” that you have to keep in mind when trying to treat the disease.
There’s the right target, the right time, the right place and the right patient. And it simply hasn’t been done before.
“We’ve taken triggers like amyloid beta and we’re not testing it in the right time and right patients,” says Watts, especially if you consider that the disease pathology is in place for 15 years before classic symptoms begin to emerge.
“I think tau is one of the most interesting targets,” Watts tells me. “It correlates with disease progression and cognitive decline, well ahead of cognitive deficits.”
Denali has a BACE/tau bispecific in hand that Takeda took an option on when it signed a billion-dollar development deal with Denali. Other combinations and genetically defined subpopulations are being studied to line up the best shot for specific patient groups, sizing up the four rights in each.
“We are balancing risk across areas to find probability of success,” says the CEO. “In Alzheimer’s, you have to break down the disease.”
There probably isn’t going to be a solid clinical success in two years, say the two Denali execs. But it will be possible to focus on biomarkers of the disease with more sensitive tracking of cognition. And they’re excited that the FDA is signaling that they are open to that approach.
Won’t be fooled again
At this point, the growing weight of evidence that any single therapeutic approach to Alzheimer’s can’t offer a definable benefit for all patients who fit the same rough definition for the disease has triggered a sea change in the way Alzheimer’s drugs of the future will hit the clinic.
Biomarkers that reflect the pathology of the disease are growing increasingly important. And the FDA is keeping pace with the steady drive toward research in even earlier forms of the disease. The agency has recommended new guidelines — still in the works — that point developers to refine their approach to cognition, and leave function aside as an endpoint.
The FDA is now gathering responses to those guidelines, ahead of a formal adoption. They’re the first new set of guidelines for Alzheimer’s R&D to come along since 2013, when regulators mapped preclinical and prodromal stages of the disease and laid out a mix of cognitive and functional endpoints that could be used for an approval.
Those earlier guidelines have never even been tested by developers, though, because none of the data were good enough to give it a try.
Now the FDA, which has adopted a spirited approach to helping developers design successful studies and win approvals, is moving to a new set of rules that will do away with the preclinical/prodromal definition in favor of a three-stage approach to early research. Here are the definitions of the disease stages FDA suggests:
• Stage one: Biomarkers for the disease point to Alzheimer’s, but no cognitive or functional impairment is observable.
• Stage two: It’s possible to identify early cognitive decline in patients, with no functional impairment in terms of daily living.
• Stage three: The first signs of functional issues start to arise in patients with cognitive deficits.
Thus, at the earliest stages of the disease, regulators are willing to approve a drug based on cognition alone. At the preclinical level, developers could even use relevant biomarkers to win a green light — but there’s a big catch. No biomarkers are now recognized as indicators of decline. The FDA could grant accelerated approval with a requirement that follow-up data show the drug delays disease symptoms.
“The document acknowledges the shift toward very early intervention in trials that target the pathobiology of AD, and confirms FDA willingness to consider accelerated or even full approval on the basis of a convincing treatment effect on cognition alone in early trials,” notes Paul Aisen, a high-profile USC investigator who championed Lilly’s solanezumab at one point, in written remarks in the online Alzheimer’s Forum.
It’s more of a regulatory fine-tuning than a major transition, however, Schneider notes. And the guidelines present some steep hurdles that need to be addressed.
“For stage 1, their guidance for accelerated approval is to find a biomarker of the pathophysiology of Alzheimer’s whose change is reasonably likely to predict clinical benefit (good luck with this),” he writes.
Plus, as Schneider notes, “Alternative endpoints for stage 2 and stage 1 diagnoses are to follow trial participants into their next stage, but this risks prohibitively long trials.”
Merck’s Kennedy is hopeful that regulators will remain open to using biomarkers to demonstrate drug activity among asymptomatic patients — essentially, that biomarkers alone can show a drug altering the course of the disease.
That can be an important change, given the fact that Alzheimer’s can run its course for 10 to 20 years before symptoms appear. The Merck team wants to take that guidance and apply it, confident that they are using better tools than ever for defining its pathology.
Before amyloid imaging tech came along, early Alzheimer’s researchers worked with patient groups where anywhere from 20% to 30% of the subjects didn’t even have the plaque deposits they were trying to target. Since then, technology has come a long way, but it’s still a work in progress.
“I absolutely believe our appreciation of biomarkers, a panel of biomarkers, is ahead of where it was,” says David Miller, the clinical vice president of Bracket, a tech provider which specializes in Alzheimer’s research. “There’s been an improvement in our understanding of how these biomarkers work together. It doesn’t mean we are where we need to be, but we are getting closer.”
“We’re looking at how to move the field forward and going earlier and with no symptoms,” Miller went on to say. “I think that’s where we need to go.”
Understanding of neuroinflammation, a trigger for cognitive impairment, is clearly getting better. And Miller agrees that failures like Merck’s are helping the next round of studies.
“With every study that fails, we need to learn from that and do it better,” says Miller. “We need to figure out ways to do measurements better.” Those measurements need to be “sensitive to earlier stages of the disease,” he says. Researchers need to be “looking at cognition and function for more sensitive ways of doing it.”
Miller is particularly intrigued by the potential of a combination of markers for amyloid, tau and neurodegeneration for future studies — along with a more sensitive approach to evaluating a drug’s effect on early symptoms. Patient-reported outcomes could become significantly more influential.
But, he adds, “there has to be uniformity of doing that. You’re looking at studies that take place across the globe. It has to be done consistently across the globe, not just what you measure, but how you measure it, that’s important.”
This editorial Deep Dive is sponsored by Bracket.