The Alzheimer's Divide

Issues of clinical trial design are at the forefront in Alzheimer's drug development. Researchers and investors are weighing the impact of late-stage failures, even as significant unknowns remain about disease mechanism and the correlation of biochemical markers and clinical effects. A fundamental question hangs over AD drug development, which is unique in many respects: in deciding how to move from Phase II into Phase III, what can you know and how can you know it?

• The tone of the recent ICAD Alzheimer’s Disease meeting was sober, appropriate for a field still working its way through the basic science underlying the disease and facing up to several late-stage clinical failures offset by only modest—and early--successes.
• At the same time, some compounds are moving forward quickly in development, driven by the compelling need for better treatments and the tremendous market opportunity.
• In many cases, Phase II trials of these drug candidates are not aiming to establish efficacy. Rather, the sponsors are content with using biomarkers to confirm that the drugs work as they are intended.
• A fundamental question hangs over AD drug development, which is unique in many respects: in deciding how to move from Phase II into Phase III, what can you know and how can you know it?

Speaking to a group of investors following the July 2008 International Conference on Alzheimer’s Disease (ICAD), Lon Schneider, MD, of the Keck School of Medicine at the University of Southern California compared the current odds of success in late-stage Alzheimer’s Disease drug development to a Major League baseball player at the plate: he is unlikely to get a hit but that doesn’t mean he won’t get one.

The recent Phase III clinical trials in AD have been strikeouts: tarenflurbil (Flurizan), from Myriad Genetics Inc., failed in Phase III in June 2008, as did tramiprosate (3APS, Alzhemed), from the Canadian company Bellus Health Inc. (formerly Neurochem), last November. Added to those, bapineuzumab, an antibody co-developed by Elan Corp. PLC’s Elan Pharmaceuticals Inc. and Wyeth [See Deal] and perhaps Wyeth’s highest-profile pipeline compound, is at best behind in the count: the companies announced in June that a Phase II trial failed to meet its endpoint while simultaneously expressing optimism about the data. Despite the companies’ spin, a highly anticipated explanatory presentation at ICAD did nothing to assuage doubters.

Indeed, the mood at ICAD was sober and thoughtful. There was virtually none of the "this approach is really exciting" brand of enthusiasm so often heard from clinicians at medical meetings during post-presentation Q&As. Instead, programs were being praised for their structure rather than their positive outcomes. The Flurizan Phase III, for example, the second-largest Phase III study and the largest ever completed in AD, was repeatedly cited as "a failed trial in terms of the compound, but extremely successful in terms of conduct."

That tone was appropriate for a field still working its way through the basic science underlying the disease, yet also moving forward quickly in development because of the need for better treatments and the tremendous market opportunity. Researchers at ICAD were also keenly aware of the long road ahead before any of the current biochemical markers of drug activity in AD could be used with confidence to also gauge clinical response. In short, they were struggling with a fundamental issue that will dog the field for some time to come: are there any shortcuts that can be taken in an area where patient populations are diverse and clinical trials readouts lengthy? What assumptions are reasonable and what evidence is needed in deciding where to focus clinical and financial resources?

Amyloid under Attack

The presumptive mechanism underlying the majority of experimental AD drugs—and all but one or two of those in late-stage development--is through interference with the so-called amyloid cascade. The amyloid hypothesis posits that deposition of amyloid beta (Abeta) peptides and the formation of amyloid plaques in the brain are early events that trigger subsequent events such as neurodegeneration and the formation of the neurofibrillary tangles that can destroy nerve cells and neurons. "Most of us believe that loss of synapses and neurons per se are what cause the signs and symptoms of the disease," says Steve Paul, MD, EVP, Science and Technology, and President of Eli Lilly & Co.’s Lilly Research Laboratories. Among the Big Pharmas, Lilly, along with Wyeth, has invested heavily in Abeta approaches. Lilly currently has two AD drugs based on Abeta on the cusp of Phase III—a gamma secretase inhibitor [GSI] and an anti-Abeta antibody.

The amyloid hypothesis has been modified over the past several years to suggest that it’s not the deposition of amyloid that’s critical. Rather, the production of certain oligomeric species of Abeta are toxic—they interfere with synaptic transmission and cause memory impairment as well as some of the other signs and symptoms of AD.

Nuances to the hypothesis nothwithstanding, Abeta is very much under fire.

On top of the late-stage failures and a lack of any offsetting Abeta-related clinical success stories, a July 2008 research paper in The Lancet cast further doubt on the mechanism when it detailed a post-mortem analysis of patients from trials of an earlier failed AD drug, AN-1792, an Abeta-targeting antibody vaccine from Elan and Wyeth whose development was stopped in 2002 after 6% of the patients developed meningoencephalitis. AN-1792 had shown no major differences in cognitive performance when antibody responders were compared with the placebo group after a year, and the new paper indicated that several patients in the trial who had clinically deteriorated to severe dementia had near complete plaque removal at autopsy. An accompanying editorial in the journal asserted that "the continued clinical deterioration and eventual death of patients after effective vaccination clearly diminishes the hope that anti-amyloid therapies will bring symptomatic Alzheimer’s disease to a dramatic halt."

The Flurizan episode, a month earlier, also led to attention-grabbing headlines worldwide. The Financial Times, for example, claimed "Myriad’s failure in Alzheimer’s causes industry to rethink amyloid hypothesis," despite the fact that there was scant mechanistic evidence showing that Flurizan was even working in the manner in which the investigators suspected--to reduce the production of Abeta peptide.

By most accounts, a majority of Alzheimer’s researchers continue to have faith in amyloid, at least to some extent. "It’s too early to be nihilistic," notes Sam Gandy, MD, PhD, Professor of Alzheimer’s Research at New York’s Mount Sinai School of Medicine, located at Mount Sinai Medical Center. "We know there are mutations in APP [the amyloid precursor protein]," he says, that lead to familial forms of the disease. "Those rare forms of AD provide pretty compelling evidence that you can cause the whole syndrome beginning with amyloid." The question is whether in more common forms of AD there is some inciting factor that both causes the amyloid pathology and is also independently neurotoxic: oxidation, for example, or calcium formation (a calcium regulatory gene has been linked to common forms of Alzheimer’s). Moreover, Gandy stresses that none of the studies of the failed Abeta drugs required investigators to take baseline measurements of amyloid and monitor whether that burden increased or decreased with time—tools now available that could have provided the fodder for an analysis of the mechanism, even if they would have said little about clinical impact.

Martin Tolar, MD, PhD, CBO at CoMentis Inc., is less sanguine. "[Abeta] might play a role," he says. "But the question is how big a role, because none of the Abeta treatments and interventions to date have shown an impact on cognition, or on behavior, that we would like."

The harsh reality is that the underlying etiology of AD remains a mystery—and that presents unique challenges to drug developers. In the absence of biochemical markers or imaging techniques that correlate with clinical responses defined by cognitive, functional, and behavioral tests, pharmaceutical companies must decide how much seasoning to give their drugs before bringing them up to the Big Show; namely, into pivotal studies. Or, to borrow from the lingo of the Watergate investigation, in deciding how to move from Phase II into Phase III, drug developers must ask two basic questions: what can you know and how can you know it?

Toward Disease Modification

When draft guidelines for the approval of anti-dementia drugs were written in the 1990s for the US, Europe, and different countries in Asia, they called for a statistically significant showing of benefit based on a psychometric test and a global measure—some clinical measure of symptomatic relief. "Without that, you can’t say you have an anti-dementia drug," explains Rachelle Doody, MD, PhD, of Baylor University’s Baylor College of Medicine. The first wave of therapies—basically, the five current AD drugs--were all neurotransmitter-based drugs: the cholinesterase inhibitors tacrine (Cognex); donepezil (Aricept); galantamine (Razadyne); and rivastigamine (Exelon); and the NMDA antagonist memantine (Namenda). And every Phase II study for them, she points out, was designed around those guidelines.

Symptomatic drugs appear to have hit their limit, however. "A product profile for the next generation of symptomatic relief would have to have significantly better efficacy," explains Martin Tolar, who was formerly at Pfizer Inc., which markets Aricept along with its developer, Eisai Co. Ltd. [See Deal] Aricept is not very effective but it’s fairly safe, with only 15% GI side effects. The question is can you push the cholinergic system much further beyond what you do with the current inhibitors? Medivation Inc.’s dimebon, an antihistamine claimed also to have neuroprotective effects and now in Phase III, is the only one that looks promising, he suggests. (On September 3, as this article was going to press, Pfizer licensed rights to dimebon, paying $225 million upfront, up to $500 million in development milestones, and assuming 60% of the development costs and commercialization expenses in exchange for a 60% share of the profits. [See Deal])

AD drug developers are now much more attracted to the notion of disease modification (DM), toward drugs that don’t necessarily improve symptoms—at least in the short term—but which could halt the disease over time. (See Exhibit 1.) But these approaches, which include Abeta, present their own challenges. For one thing, trials of such drugs are likely to be lengthy and costly. That’s because a majority of AD patients in clinical trials have mild to moderate forms of the disease. They decline slowly: a substantial minority of patients—40% or so—don’t decline at all over the course of 12-18 months. This, coupled with the heterogeneous nature of the disease, means it will likely take a large study and more than a year to differentiate the effect of a DM drug on treated patients as compared to the baseline of a placebo control group. Not too surprisingly, as questions concerning trial design and cost have arisen, it has sparked a more intense desire to develop methods that allow developers to get an early read of potential efficacy.

"If you want to demonstrate a true disease modification signal in Phase II, you have to have a very long study," explains Jim Kupeic, MD, PhD, executive director of neuroscience development at Pfizer Central Research. And then in essence you have to re-do that study in a much larger sample size for as long a duration as the Phase II study, so the Phase III will also have to be very long. "The total duration of the development program is therefore going to be very long for any compound that is hoping to definitively prove modification of the underlying pathology," he says.

But it’s difficult to prove disease modification in Phase II because there is no clearly accepted surrogate marker of efficacy in AD. Thus, the new notion is to say: "We don’t need efficacy, we understand our drug and its mechanism and all we want to show is the pharmacodynamics. Based on that, we know that it’s going to work." In that case, however, where there’s only a small Phase II study--potentially a shorter Phase II study than would be typical in order to justify going into Phase III--a company is going into Phase III based on limited data. "The question is whether there’s enough data—biomarker data or safety data or some hint of efficacy--to justify the fairly substantial decision to move into Phase III," says Kupiec. "It’s a high-risk position. That’s the key question every company has to grapple with in this area." Still, because of the tremendous unmet medical need and the large amount of competition moving into late-stage development in AD, pharmaceutical companies appear willing to take the extra risk.

Is Abeta Therapy Destined to be too Little too Late?

As it turns out, the actual formation of amyloid-containing plaques appears to be a very early event and probably occurs in people who don’t even have AD: dissolving it, researchers now suspect, might be too little too late.

"Even for an effective anti-amyloid therapy, it’s possible that when you already have Alzheimer’s Disease you may have as much amyloid as you’re ever going to get," acknowledges Lilly’s Steve Paul. "Perhaps it’s a bit late to hope that an effective anti-amyloid treatment would work well at that point." On the other hand, there may still be enough amyloid deposition occurring to have the opportunity to reduce Abeta oligomers and beneficially impact the course of the disease so that survival curves between treated and untreated groups will separate. In the bapineuzimab Phase II trial, for example, using brain imaging (volumetric MRI) as an endpoint, there did seem to be evidence of a survival difference, at least in the subgroup of patients – the approximately one-third of patients in the trial who did not carry the APOE4 gene mutation, a known risk factor for AD. (Those analyses were done post hoc, but a year ago, after an early look at the Phase II data, Elan and Wyeth decided to separate out APOE4 carriers versus noncarriers as part of its Phase III trial design.)

What Lilly and Elan-Wyeth are therefore hoping is to see a difference in the progression of the disease over the course of maybe 18-24 months. Given current estimates of placebo decline, "This would be sufficient to slow the progression to the point where patients were being completely incapacitated or needed to be institutionalized, by 6, 8, 12 months," Paul estimates. "That would be a very substantial contribution and would also argue for trying to treat patients earlier in the course of the disease, and eventually, I believe, even try to detect the disease process itself presymptomatically."

The counter argument is that companies are no longer expecting to improve the patient, as opposed to stopping the disease progression. "They have a disadvantage with respect to the usual criteria of showing a drug-placebo difference because they don’t have any bump over baseline for the treated patients," says Baylor’s Rachelle Doody. Doody is well aware of the ironies. "We’ve come to a time," she contends, in which many thoughtful researchers are saying "OK, we’ve got a better drug because it impacts the disease but it’s not going to improve anybody, so we have to invest more to show a smaller difference," and ensure a decline in the placebo group.

Paul, of course, sees it differently. He and others analogize the Abeta therapy situation to that of treating heart disease with lipid/cholesterol-lowering agents. In patients with coronary artery disease and mild symptomology, like angina, and who also have high cholesterol, statins are effective: they reduce the morbidity due to subsequent heart attacks and mortality due to heart attacks or strokes. But if you wait until someone’s already got CHF, already enough infarctions so they don’t have a lot of heart muscle left, statins are probably not going to be very effective at that point. And they may be most effective in people who are treated much earlier, even before they have symptoms. "I think this may be what turns out to be the case with Alzheimer’s disease," Paul says. "The question is whether we’ll be able to see a signal -- enough improvement in the Phase III studies, which have to be done over a long period of time to be clinically meaningful. That would allow us, like with the statins and HMG co-reductase inhibitors, to come back and do studies earlier in the disease, potentially in people who are presymptomatic but at high risk of getting AD."

The reasoning is sound if, as Paul says, the trials run long enough to show the anticipated clinical improvement—especially since surrogate markers for the disease are lacking and there’s no inkling that effective ones will be available anytime soon. As Eric Siemers, MD, medical director at Lilly, noted at ICAD, data from use of the current biochemical markers, which range from measurements of Abeta or phospho-tau in cerebrospinal fluid or blood to imaging using PET or MRI, "suggests these probably will not be surrogate biomarkers." (The neurofibrillary tangles in the brain that are also characteristic of AD contain phosphorylated tau protein: it is another proposed mechanism for the physiopathology of the disease.) (See "TauRx’s Alzheimer’s Drug: A Web of Tangles," The IN VIVO Blog, August 1, 2008.)

What biomarkers can tell you is if a drug is doing what you want it to do mechanistically. There are also transgenic mouse models of Abeta, which show that modulating it does correlate with behavioral improvements in the animals. Thus, even if the link between Abeta and a human clinical improvement has yet to be demonstrated, the most advanced disease-modifying AD programs are Abeta focused. "One of the reasons that our most advanced compounds are amyloid compounds is that the tools to study them are so much better," notes Ron Black, MD, assistant VP, neuroscience, at Wyeth. For instance, the triple transgenic mouse models that accumulate amyloid have been around since early in the decade, allowing researchers to move ahead in developing drugs that are effective around amyloid. At the same time, "we are looking at a lot of different targets," insists Black. Both Elan-Wyeth and Lilly are cautious when discussing the prospects for proving the Abeta hypothesis, despite the investments they’ve made. "If there was one mechanism we were sure would work, everybody would be working on it," sums up Tom Megerian, MD, PhD, executive director, clinical research, at Epix Pharmaceuticals Inc.

Two Views of Managing the Clinical Risk

With bapineuzumab, Elan and Wyeth decided to conduct a lengthy (18-month) Phase II trial. The main focus was dosing and safety – they were already aware, for example, that the drug might cause vasogenic edema, a side effect not seen in the placebo group and to different degrees in APOE4 carriers versus non-carriers. But it was also set to look at efficacy. "Given the limitations of the study – it was a small study with very limited power – we are very encouraged by what we saw," Black says. In effect, the investigators were assessing four different doses in parallel for descriptive signals of safety and efficacy, with the intention of averaging them together. But what they found was that the APOE4 carriers did not show much of a benefit from therapy as measured on cognitive or behavioral scales, except for the small number out of the total enrollment who completed the full course. The APOE4 completers were "more positive" and the efficacy signal "may be of interest," noted Sid Gilman, MD, director of the Michigan Alzheimer’s Disease Research Center at the University of Michigan, the chairman of the independent safety monitoring committee for bapineuzumab who presented the Phase II data at ICAD. Moreover, there was "quite substantial" improvement in the APOE4 non-carriers and "absolutely dynamite" data in the APOE4 non-carrier-completers.

Seeing the data thus parsed, and realizing that it entailed asking elderly AD patients to come back for costly infusions of a drug that may not work except after a long period of time, Wall Street’s reaction was swift and negative. But in the context of the drug’s development, Black is quick to point out that "the basic trial design we chose a year ago, based on what we looked at when we finally unblinded the full study, looks like the right decision. When you try to accelerate a program because of its importance to the company and to the medical community, you take those kinds of risks. You don’t always get things right. Yet at least for me personally, when we unblinded it, it looks like we made the right decisions."

Lilly, on the other hand, did not necessarily seek to establish efficacy with either its GSI or its antibody before deciding to move them into Phase III. In their case, they decided to do shorter-term Phase II trials--its antibody trial was only 12 weeks in duration--that would give them some indication of drug activity, but not necessarily efficacy, before advancing into Phase III. And in the case of its antibody drug, because it binds only soluble Abeta and not aggregated Abeta or plaque (as does bapineuzumab), there was no a priori concern about vascular inflammation. For both compounds, biomarker results encouraged Lilly sufficiently to begin pivotal studies. However, aware of significant clinical risks associated with both programs, the pharma did also seek to lessen its financial risk by offloading pivotal trial expenses for both programs to the investment firm TPG-Axon and NovaQuest , the partnering arm of Quintiles Transnational Holdings Inc. [See Deal] (See "Alzheimer’s Development Financing: Risky Therapies Call for Innovative Deals," IN VIVO, September 2008 (Also see "Alzheimer's Development Financing: Risky Therapies Call for Innovative Deals" - In Vivo, 1 Sep, 2008.).) This risk-sharing move has done little to assuage Wall Street, which is even more skeptical of the drugs’ potential utility since Lilly is willing to give up significant up-side associated with the medicines.

It’s clear that Lilly is bucking the odds. In terms of probability of technical success, the industry average for molecules getting into Phase III to also get to launch is about 50%. "You should only really be losing molecules in Phase III for relatively rare adverse events that you didn’t pick up in Phase II," says Steve Paul, with the notable exception of those where you’re actually testing a hypothesis, such as the amyloid cascade hypothesis. "I don’t know where the Alzheimer’s molecules fall but they are certainly not near the top of probability of technical success," he adds. "Clearly this is a very legitimate hypothesis that we’re testing with fairly good drugs. Having said that, it is a higher risk hypothesis to be testing in Phase III, there’s no doubt about that."

Perhaps the greatest cause of excitement at this year’s ICAD centered on Medivation’s dimebon, which in a Phase II study conducted in Russia was safe and appeared to improve the clinical course of patients with mild to moderate AD. The drug’s potential in AD was first identified by screening known compounds for dual activity against the cholinesterase and NMDA receptors—its mechanism of action is still being debated--and dimebon is now lined up to begin a pivotal Phase III study. If the new study meets certain endpoints, the FDA has said it would accept an application for approval with the completed Phase II as a second pivotal study—a testament to that study’s design and conduct.

That’s encouraging news for the field, as is the amount Pfizer is willing to pay to round out its AD portfolio, which also includes PFE360365, an Abeta-binding antibody via its acquisition of Rinat Neuroscience Corp. in 2006 and a dual-mechanism antagonist of RAGE (receptor for advanced glycosylation products), licensed from vTv Therapeutics Inc. the same year. [See Deal] [See Deal] But at the same time, the serendipitous nature of dimebon’s discovery as an AD drug has led some to lament that, in terms of discovery approaches, it may be the best they can do. No wonder Pharma is hedging its bets.