(from the January/February 2020 Issue)
Negative Space: Missing Pieces in the Aducanumab Story
It was the most eagerly awaited event of the Alzheimer’s holiday season: Biogen had promised to provide additional data buttressing its decision to remove aducanumab from the futility dustbin, and after a fresh coat of paint, submit it to the FDA for approval in the treatment of Alzheimer’s. CTAD’s annual meeting provided a suitable setting, and it was a packed house on December 5. Biogen’s Samantha Haeberlein provided an extensive review of aducanumab data and the rationale for claiming that disease-slowing had been demonstrated in the truncated Phase III trials. But it was not so much what was shown and said, but rather more what was missing, that was most striking. Among the omissions:
Dissent. The panel of discussants at the Biogen presentation were primarily investigators for the PhIII trial in question, plus Paul Aisen from USC, an amyloid-theory stalwart. They spoke movingly of their wish to have something to offer patients otherwise bereft of hope, and the importance of extending a patient’s capacity for autonomous functioning. Our background is clinical, we get it, but in this scientific setting, it came across as cheerleading based on a pre-specified wishlist, not clear-eyed science. There was no shortage of skeptics in the room, surely someone qualified could have been invited to participate as a spokesperson for the loyal opposition. Instead, the panel seemed to be parroting the party line.
Microphones. At CTAD, as at most large conferences, microphones are placed at various locations around the room so that questions can be asked of the presenter after the presentation has finished. For Biogen’s aducanumab talk, and for no other session, those microphones had been removed. In their place, attendees were invited to text questions to a moderator who selected a few softball queries, dutifully responded to by panel members. Oddly, when asked about regional differences, Biogen’s curt response was “we are not commenting on that” (why not?). This was a curated and controlled media moment, not the opportunity for scientific interchange for which CTAD would in theory have been an ideal locale.
The extended version of ARIA’s song. Biogen acknowledged the high rate of ARIA-E and H in the 10mg/kg, APOE4+ group, 50-54%, but was dismissive around the details, stating that only 20-29% of these cases were symptomatic, and “the majority” were able to eventually resume aducanumab. What happened to the plurality who were not able to resume treatment? In the PhII PRIME study, 22% of ARIA cases were severe (Biogen disclosed only that ‘78% were mild to moderate’, one had to do the math to figure out the rest). What percentage were severe in Phase III? And what happened to those patients? One of the panel members stated that he is comfortable ‘managing ARIA’, which may well be the case, backed up as he is by Brown University Medical School’s resources and imaging capabilities. That may not be so easily duplicated by general practice neurologists in the field. This is a safety issue not to be so cavalierly dismissed–the FDA will not ignore it.
Having seen one media report of an “electric atmosphere” following the Biogen presentation,we can report that NIR spoke to at least a dozen attendees about that presentation, and without exception, the reaction was one of disbelief and annoyance at having been subjected to an infomercial for aducanumab and the amyloid hypothesis. Or to put it another way, the atmosphere was more shocked than electric.
The Bottom Line:
Statistical purists may suggest that the announcement of the futility analysis results, and subsequent trial termination, irrevocably altered the framework for everything that followed, including the inclusion of later subjects and the post hoc revisiting of the ENGAGE trial. If so, it would then follow that neither study can be properly described as successfully documenting efficacy, and at most, the EMERGE trial puts forth a hypothesis to be tested. But if the FDA were hewing to this strict standard, and saw neither trial as constituting a valid pivotal study, we suspect that this would have been conveyed to Biogen during subsequent discussions. So for those keeping score at home, we will chalk up EMERGE as appearing to support aducanumab’s case. But on the other hand, all the post hoc genie-lamp rubbing cannot turn ENGAGE into anything other than a failed trial, indeed some numerical trends showed a worsening of the aducanumab group compared to placebo. So one trial was positive, one negative. Which was the outlier? Post hoc data mining cannot definitively answer that question.
What will happen now? There are a few possible scenarios:
A Tiebreaker is Required: If the FDA follows its own procedures and precedents, a third Phase III will be mandated. ENGAGE was not confirmatory, and those hoping that the Phase II results from the PRIME trial might substitute forget that the whole dose-response pattern in that trial was turned on its head. A tiebreaker should and will be the outcome, unless the following occurs:
Aducanumab Turns Into A Political Football: Given Alzheimer’s massive toll, well-meaning people may well decide that ‘something is better than
nothing.’ We can also imagine a political environment, with a Presidential election looming, and an impeachment trial that will not end with a conviction, where this would be seen as a potential crowd-pleasing ‘win’. Regulators will be under great pressure to be ‘flexible’, and there are examples of doing so in the past, single trial approvals in Duchenne Muscular Dystrophy and Parkinsonian psychosis come to mind. But this is different, in both the stakes and scale. The sheer size of the Alzheimer’s market dwarfs either of those, and there are safety issues with aducanumab not seen in those other contexts. To open the floodgates of demand for an Alzheimer’s treatment would mean millions of people could be exposed to a drug of unclear utility, wherein the impact of even a low-incidence adverse event will be greatly magnified by the size of the population using it. And in terms of reverberating effects: At best, aducanumab would be a mediocre treatment making the enrollment of trials for other, potentially superior therapies, nearly impossible, save for treatment failures unlikely to respond to any disease-modifier, or those with ‘unmanageable’ ARIA.
These concerns will not matter one iota to those who see this as potentially adding to the case for a Presidential re-election. In their view, a win is a win is a win, and the ability to claim that this administration ‘cured’ Alzheimer’s would be intoxicating. The most cynical scenario we can conjure up combines ‘‘It’s a nice little Agency you have here, it would be a shame if something happened to it’ and ‘I need a favor, though.’
The window of timing opportunity would be narrow, given that Biogen will file the NDA in early 2020, just months before the election. We can imagine immense pressure to act ASAP. However, there is also a new FDA Commissioner now in place, for whom the calculations could be both medico-ethical and political. We would like to think the former would rule the day, but if the latter is a factor, and if Stephen Hahn wants to be Commissioner longer than one year, he would have to consider the likelihood that if he is perceived as having ‘caved’ to Administration pressure, and the Democratic candidate then wins in November, his tenure will be short. This calculation would be inverted in the scenario of a Trump re-election.
Any eventual approval for aducanumab would have significant conditions attached. Given the ARIA issue, we expect a REMS requirement for quarterly MRI scans for the first eighteen months of a prescription, good news for the owners of imaging centers.
It seems almost bizarre to suggest that a decision that could affect millions of lives, cost billions of dollars in health care expenditures, and could constitute a major obstacle to the validation of other therapies in development, could be impacted by election-year maneuvering. But bizarre has become the new normal.
One Other Alzheimer’s Note
We do not question the sincerity of aducanumab’s supporters within the Alzheimer’s community. We do question the objectivity of their judgment. Aducanumab, as has been the case with previous high-profile projects, has become the poster child for cognitive dissonance theory, people seeing what they want to see in order to maintain consistency with long and deeply held beliefs. Paul Aisen proclaimed at CTAD that aducanumab is a “truly major advance”. We recall a previous CTAD where Aisen concluded that the solanezumab Phase III data set ‘proved’ the amyloid hypothesis. Neither statement was accurate, though we are sure that Aisen believes both to be true. The degree to which well-intentioned experts can twist their thinking in the service of cognitive congruency was exemplified by a later CTAD panel on BACE inhibitors, a class where the story has turned to the assessment and reversal of drug-induced cognitive worsening. Most presenters seemed to acknowledge that the therapeutic window for BACE inhibition, other than perhaps for rare autosomal dominant AD, is essentially nil. A couple of them suggested that, while BACE inhibitors impair memory, they seem to improve language fluency, a pyrrhic victory at best. Banner Health’s Eric Reiman took a contrarian view, proposing (as did Aisen the next day) that BACE inhibition should continue to be explored as a therapeutic option, particularly in APOE4+ patients without amyloid deposits, arguing that brain volume loss comes early and is neither ‘progressive’ nor clearly correlated with cognitive deterioration, which he stated might be both ‘tolerable’ and reversible. The notion that patients whose disease involves cognitive decline might be able to ‘tolerate’ treatment-induced impairment seems incongruous at best. While Reiman may have been intentionally provocative in his suggestions, they did illustrate the pretzeled logic that so often permeates the Alzheimer’s field.
Moments after their CTAD presentation, Biogen held a conference call for major investors/analysts. Where they fielded a number of questions in real-time; they were more open to investors than to their scientific peers. In that call, they provided more detail about their rationale for seeing ENGAGE as a net positive for aducanunab, but they still responded to a number of queries with ‘we’re not going to talk about that.‘ There continues to be a dearth of ARIA outcome information, other than bland assurances that it is ‘manageable.’ They insist they have no intention of running another pre-NDA trial. What do they know that we do not? Plenty, that’s a low bar. But Biogen’s parsing and filtering of information still strikes a sour, defensive note to our ears.
(from the November/December 2019 Issue)
Excerpt from the review of Parkinson’s
LRRK2, mutations of which are found in 1-3% of PD patients, has long been considered one of the more promising novel PD targets, given evidence that gain-of-function mutations in this kinase interfere with lysosomal function and are associated with neurodegeneration in Parkinson’s. This is true in the LRRK2 mutation population, but possibly in sporadic PD as well. However, LRRK2 has turned out to be a challenging target. In 2015 Genentech reported that two of its LRRK inhibitors were associated with lung tissue pathology in a preclinical model. A praiseworthy consortium of pharma companies with LRRK2 programs collaborated, under MJFF, in running preclinical testing of all the available LRRK2 candidates. The compounds from Genentech, Pfizer, and Merck were all associated with pneumocyte pathology at higher doses, indicating that this was an ‘on-target’ class effect. However, low doses of the Pfizer and Merck molecules did not cause lung vacuoles, even though they produced significant reduction in brain levels of LRRK2, and the lung changes incurred by the Genentech compound were completely reversed two weeks post-dosing.
The conclusion was that it might be possible to walk a dosing tightrope around the pulmonary safety concerns, though Genentech decided that they would not pursue the target, and outlicensed their compounds to Denali Therapeutics, as is discussed below. Merck is in early stage development with a replacement for MLi-02, Pfizer ended up spinning out PFE-360 to Cerevel, where it is not at the front of their developmental queue.
Denali has been the company that has pressed on most aggressively with LRRK2. DNL201 was the first candidate to finish PhI, and to then enter a PhIb safety/tolerability study, which was to finish in September. Similarly, they initiated (this past September) a PhIb study of DNL151, enrolling patients with and without LRRK2 mutations. Two doses are being tested for safety, in 24 patients, results are expected in early 2020. They will then select which molecule to enter into PhIII studies during 2020.
Denali has come to believe that LRRK2 inhibition may be applicable to idiopathic PD. A U.Pittsburgh group found that LRRK2 kinase-induced damage can be found in post-mortem brain tissue from sporadic PD patients. They proposed that this heightened kinase activity creates oxidative stress and contributes to a failure to clear alpha-synuclein, leading to toxic aggregation. However, in a classic intra-Pennsylvania dispute, a U.Penn group reported that LRRK2 inhibition, via Merck‘s MLi-2 compound, was not neuroprotective against alpha-synuclein pathology in a mouse model: The motor phenotype was unchanged, as was alpha-synuclein accumulation. They suggested that the role of LRRK2 inhibition as a tactic in idiopathic PD has not yet been validated.
Biogen has initiated PhI testing of an ASO (BIIB094) against LRRK2 developed by Ionis. 62 patients will be enrolled, including both LRRK2 mutation positive and negative subtypes. Results are expected in 2022.
A Harvard group (Isacson et al) is working preclinically with an antisense approach to LRRK2-mutation cases, seeking to achieve stable reductions in LRRK2 production by editing out two common mutations (R1441C and G2019S).
E-Scape is focused on the LRRK2 mutation population with a selective inhibitor (it does not inhibit wild-type LRRK2) that they expect will avoid the pulmonary pathology associated with these predecessors, and thus has a much broader therapeutic window. This is in discovery stage.
GSK is not a company that has earned much mention in these pages in recent years other than for the sake of snarky nostalgia, but GSK did sign a four-year agreement with 23 and Me last year that includes GSK’s preclinical LRRK inhibitors for PD. The deal gives GSK access to the 23 and Me database in the service of clinical trial patient selection and compound validation. The inference is that GSK intends, at some point, to go into the clinic with a LRRK inhibitor, if one passes muster. This would be the most tangible indicator of their long-rumored return to neuroscience.
Lundbeck/Zenobia Therapeutics worked together on identifying LRRK2 inhibitors, but the program failed, Zenobia folded. Ipsen Group partnered in 2012 with Oncodesign regarding LRRK2 inhibitors, this program was subsequently partnered with Servier.
It is worth noting that a 2019 paper linked LRRK2 to stress-induced impairment of GI motility, attributing this to impaired signalling in the hypothalamus. Separate from the putative involvement of LRRK2 in stress-related pathways, it must be wondered whether this could also be associated with the Parkinsonian hypomotility that has been deemed an early pathological sign of PD-induced changes in the gut-brain microbiome.
(from the September/October 2019 Issue)
Excerpt from the review of Alzheimer’s
The TauRx Blues
Only one alleged tau-targeting agent has produced clinical efficacy results–TauRx has received over $250 million from a Singaporean investor group, and returned nothing of value. The premise began with Rember, a purified form of the venerable staining agent methylene blue, which the CSO claimed was a tau-aggregation inhibitor. There are independent reports that Rember/methylene blue’s mechanisms involve antioxidant and MAO inhibitory properties, and does not promote tau clearance, thus this may not have been a actual test of the tau hypothesis.
TauRx originally came to attention when they claimed disease-modification in a 321pt Phase II trial, wherein the rate of decline over 50 weeks was reportedly reduced by 81%, a dose-response relationship cited with the two lower doses, 30mg and 60mg. The highest dose, 100mg, did not produce any benefit whatsoever, TauRx stated that the 100mg dose had chemically interacted with its gelatin capsule and was converted into a non-active form. Then, while the formal placebo-controlled period was 24 weeks, they proceeded to claim that the 100mg dose, administered for the next 60 weeks, functioned as a placebo. This was a convenient rationalization, since patients receiving Rember had blue urine, the actual placebo group did not, so the trial had been essentially unblinded from the start.
A second-generation compound (TRx0237/LMTX) went directly into two Phase III trials in Alzheimer’s and a Phase II in Frontotemporal Dementia, using a low dose (4mg) of TRx0237 as the ‘control’ group. TRx0237 failed in the first Phase III, an 891pt study in mild-moderate AD. TauRx assembled a specious post hoc supposition that the 15% of patients not on other Alzheimer’s drugs (essentially cholinesterase inhibitors, most patients had been treated with a ChEi as a monotherapy, some added memantine) separated robustly from placebo. TauRx speculated that ChEis might block (somehow) the binding of TRx0237 to some constellation of receptor sites. But at the same time, the investigators also suggested that the 4mg BID dose might actually have some clinical activity, that it was not biologically inert. This lysed whatever was left of their credibility, the same group cannot serve as a placebo and low-dose intervention at the same time.
A second Phase III was claimed to show that the 100mg dose, the active intervention, and the 4mg dose, intended as the control, both significantly improved ADAS-cog and ADCS-ADL deterioration rates when not accompanied by a ChEi or memantine. Thus any variance in outcome appeared to be largely a function of whether patients were receiving a symptomatic treatment, most probably reflecting differences in their rate of disease progression. TauRx tried to attribute the difference to TRx0237, rather than fundamental differences in the treatment populations, and even said they wanted to further develop the ostensible placebo-dose of 4mg as an active therapeutic.
This was reminiscent of the Affiris debacle in 2014. When a company seems confused as to which is a placebo and which is an active drug in a clinical trial, that is a red flag. Last year, we wrote that the “TauRx program appears to have finally lost the support of that Singaporean investor group, and is unlikely to be viable going forward“. This was incorrect, TauRx did somehow convince investors to fund a nine-month duration study for LMTX, utilizing a cognitive and functional endpoint cobbled together from the ADAS-cog and ADCS-ADL, using the 4mg BID dose they had deployed in the previous study as the placebo control group. The trial is aimed at enrolling 300 pts, plus an additional 75 patients being given 8mg dosing BID. The Company stated that positive results from this trial, in combination with previous efficacy data (which, it should be remembered used a dose 25 times higher, again, with the 4mg dose serving as the active arm in this trial having served as the placebo in that trial), would confirm the clinical utility of LMTX in AD. As we have noted in the past, since the placebo does not turn urine blue, this trial is not truly blinded, which is more than can be said for their investors.
(from the March/April 2019 Issue)
Excerpt from the review of Stroke
There is a ‘cytokine cascade’ of inflammatory factors that is precipitated in stroke. Interleukin-1(IL-1), interleukin-6 (IL-6), and Tumor Necrosis Factor-alpha (TNF-alpha) fuel the inflammatory process by recruiting leukocytes to the area, fueling apoptosis. The several companies which have targeted IL-1 and IL-6 in multiple sclerosis did not turn their attention to stroke. Xigen did make stroke one of the lead indications for its jnk inhibitor XG-102, and completed a 10pt Phase 1b study with XG-102, but then exited.
CCR5 (CC chemokine receptor 5) has been proposed by a UCLA group as a target for addressing neuroinflammation in stroke (and TBI), and to improve stroke rehabilitation outcomes. Stroke patients with a CCR5 gene mutation causing a loss-of-function had more improvement in sensorimotor and cognitive functioning at six months and a year post-stroke. In mice, blocking CCR5 expression led to a better functional outcome in stroke models,and showed improved dendritic spine density. The UCLA group has initiated a Phase II/III trial using the CCR5 blocker Selzentry/maraviroc, which is marketed for the treatment of HIV infection. The trial will enroll 60 patients who are within 4 weeks of their stroke (ischemic or subcortical hemorrhagic); their performance on two motor tests are the co-primary endpoints, other motoric measures of stroke recovery are among the secondaries. This trial is expected to complete around the end of this year.
MediciNova/Kyorin‘s ibudilast, a MIF and PDE4 inhibitor, has long been approved for the treatment of stroke in Japan and Korea, albeit for post-stroke dizziness in specific. US academic groups have been looking at its possible utility in MS or addictive disorders, no further development in stroke is likely.
Armagen developed AGT-110, a TNF decoy receptor fused with the transport antibody, that is intended to use the decoy receptor to reduce available levels of TNF-alpha, reducing its inflammatory effects. They claim that in a mouse model, AGT-110 reduces stroke volume by 62%, but their focus has been on rare diseases, this program has made no progress.
ArunA Biomedical, which had been developing stem cell therapeutics, has changed course somewhat, and is now developing exosomes (vesicles) obtained from neural stem cells as a means of treating stroke. Its lead exosome product, AB126, is currently being tested in pig models of stroke, following mouse model testing which showed decreases in infarct size and improved functional outcomes. Exosomes are a non-immunogenic vehicle for providing mRNA and cytokines that facilitate growth/repair–ArunA’s current working hypothesis regarding the as-yet-unproven mechanism, is that it operates via an anti-inflammation route, steering macrophages to an anti-inflammatory role and reducing pro-inflammatory T-cell generation.
A Columbia group initially cited complement component-1q (C1q) as a novel target in stroke, as C1q knockout mice show much less damage in a stroke model. The mechanism was not clear at that point, though the suppression of mitochondrial oxidation was suggested as a possibility. Based on work done since that time, ig appears that C1q operates as a key initiating factor for the classical complement pathway, which triggers microglial activation, the release of inflammatory cytokines, and the tagging of synapses for pruning. C1q is the major target for Annexon, which has raised $153 million in the past four years, arguably aided by the fact that they do not list stroke as an indication of interest (they do cite AD, HD, SMA, FTD). ANX005 is their lead C1Q mAb.
A University of Colorado/Medical University of South Carolina collaboration on complement system pathology has been spun out into AdMIRx, with stroke as the lead CNS indication (other indications include kidney disease and rheumatoid arthritis). They have developed a technology for inhibiting complement activation locally, in areas contiguous to a stroke infarct: Cerebral ischemia causes the production of a neuronal marker, an epitope, that triggers activation of complement C3d, which then tags those neurons for phagocytosis. AdMIRx’s B4Crry combines an antibody that recognizes that ischemic marker with to a complement inhibitor. This reduces local neuroinflammation associated with microglial activation and phagocytosis of neurons in the stroke penumbra. Because the marker is primarily found in the penumbra, this limits the scope of B4Crry’s impact to that vulnerable area. In animal models, a single infusion of B4Crry reduced C3d in the brain, reduced neuropathological symptoms and necrosis, providing for improved functioning 15 days later. B4Crry is preclinical.
There are other companies targeting complement system components as an anti-inflammation tactic, but InflaRx (C5a) is not pursuing stroke, nor is Apellis (C3), or the most clinically advanced complement-focused company, Alexion.
With the overall surge of research attention in neuroscience to the prominent role of neuroinflammation, there are a raft of yet-to-be-validated inflammation targets that have been suggested for study:
DSTN/destrin (actin depolymerizing factor) upregulation has been reported by a U. Pittsburgh group to decrease pro-inflammatory markers and reduce BBB leakage in a mouse model. They also claim improvement in spatial memory in those same mice. Another U. Pittsburgh researcher reported that increasing IL-33 decreases neuronal death in vitro, and reduces infarct volume in vivo. Work done at University of Texas has pinpointed IL-27 as biasing neutrophils towards an anti-inflammatory role in rodents, though the group has turned its focus towards downstream iron metabolism mediated by IL-27, stating that IL-27 itself might not have a sufficient safety margin as a target. Knocking down levels of TRAF6 (TNF receptor factor6) was reported by a Wuhan University group as reducing infarct size and reduced neurologic impairment in a mouse model.
Some of the molecular culprits pursued in other neurodegenerative disorders have also been proposed as playing a destructive role in stroke. A University of Wisconsin researcher reported alpha-synuclein knockout reduces infarct volume and improves survival/functioning in the MCA stroke model. Neither of the two major groups in human trials with alpha-synuclein mAbs (Biogen/Neurimmune and Roche/Prothena) have evidenced any interest in expanding into stroke. A U. South Wales team cites tau as a target for stroke therapy–knocking out tau reduces neurological deficits and time to recovery in mouse models.
An attempt to repurpose an MS drug turned out to be a road to nowhere for Biogen, developing natalizumab for stroke. They ran this integrin alpha(4) inhibitor through Phase IIb, hoping for an anti-inflammatory impact, but the trial failed to hit any of its endpoints, and the program was terminated.
2019 (from the January-February 2019 Issue)
Winter Is Coming: Rallying the Undead (Mechanisms)
There have been multiple sightings of resurrected mechanisms, long thought dead and gone, now alive, or more provisionally, undead. Whether they will be shown to be truly viable, or just products of overexercised fantasy, remains to be seen:
This glutamatergic receptor subtype was long heralded as a route to cognitive enhancement by Cortex Pharmaceuticals, Lilly, Servier, and Organon (amongst many), revisited by Takeda, and quietly, by Pfizer. While Takeda ran into tox concerns with TAK-653, Pfizer showed enough with PF-04958242 to catch Biogen‘s eye, not an easy catch, given that schizophrenia is the putative focus. Biogen acquired the compound, overcoming what had become an industry antipathy for AMPA in particular, and their own reticence about Psychiatry in general. Will the PhIIb bring AMPA fully back to life, for CIAS? This MOA now has its best shot ever.
Pheromones have not elicited a lot of attention in the industry, though one tiny firm did focus upon them–Pherin Pharmaceuticals. Pherin has been walking a narrow tightrope between corporate life and death for the past two plus decades, as they sought to develop intranasal drugs based on pheromones. Janssen, Organon, and Taisho had all taken turns partnering this work 15-25 years ago, all eventually departed. Columbia U. psychiatrist and Pherin consultant Michael Liebowitz has single-handedly kept PH94B (an intranasal spray intended for use by females in socially stressful situations) alive, overseeing three underfunded pilot studies from which significant impact on anxious distress in a Social Anxiety population was reported–the study cited as preparation for Phase III enrolled just 23 women. VistaGen has now partnered PH94B, they plan to go into PhIII.
3) PARP inhibitors were once seen as a viable intervention for neurodegeneration, but Guilford Pharmaceuticals failed with their program, others moved into oncology, where PARP has been a useful target, and toxicity less of a concern. A Johns Hopkins group has now shown that low-doses of approved (for cancer) PARP inhibitors (Rubraca, Lynparza), as well as veliparib, show in vivo and in vitro signs of efficacy in an alpha-synuclein model, reducing neuron loss and alpha-synuclein pathology, which points to a role for PARP in PD pathophysiology. Now we will see if AbbVie, AstraZeneca, or Clovis Oncology will choose to explore this–it is only a good fit for AbbVie.
4) NeuroHealing‘s subQ apomorphine for coma-awakening is finally going back into a clinical trial–in Belgium.
5) Feinstein Institute researchers just published that ‘rewiring’ can be imaged in patients who had received AAV2-GAD gene therapy in a 2008-10 Neurologix trial for Parkinson’s. Neurologix declared bankruptcy in 2012.
(from the Sept/Oct Alzheimer’s Review)
Passive Vaccines: the AB mAbs
The passive vaccine approach uses monoclonal antibodies to bind/deactivate beta-amyloid, and this has become the most popular route taken by late-stage programs, with ambiguous, mixed results at best. The hypotheses for why so many have failed include the stage of disease being too advanced, the brain penetrance for the antibodies too limited, or the beta-amyloid species targeted the wrong one.
Bapineuzumab: In Phase II, Elan/Wyeth‘s (eventually, Pfizer/JNJ‘s) bapineuzumab missed its endpoints for cognition, but they had already initiated a massive Phase III program, based on a posthoc analysis that showed a bifurcation based on the ApoE4 genotype: The 65% who were carriers showed no significant benefit from bapineuzumab, although their completers had a slightly (2.6 ADAS-cog points) better outcome than the placebo group. The non-carriers showed a difference of 5 ADAS-cog points, which did reach significance. This became an example par excellence of the problems presented by differing background rates of decline: The placebo group’s rate of decline was essentially the same as the drug group until they reached the final six months of the trial, when a nosedive took them towards the extreme end of the range in terms of the slope of cognitive decline. Vasogenic edema was a complicating factor, and the drug failed in Phase III.
Solanezumab: Lilly’s solanezumab/LY2062430 preferentially targeted soluble amyloid monomers rather than plaques and fibrils. In theory, this reduces vasogenic edema risk, since it spares plaque attached to the endothelial walls of the brain vasculature. But it also requires that the antibody neutralize an enormous ‘pool’ of amyloid. After a 52pt Phase II, covering only a twelve-week period, Lilly deemed solanezumab ready for two Phase III trials, that enrolled 1000 patients each, with solanezumab administered twenty times over eighty weeks. Those trials both failed on their primary endpoints. However, Lilly seized upon a posthoc analysis of data pooled from both trials that they believed pointed to cognitive benefit in patients with mild AD. The finding that benefit was restricted to mild patients could be interpreted as meaning that, by the time Alzheimer’s has reached moderate stage, it is too late to intervene in the amyloid pathway, and Lilly chose to interpret it that way. They also noted that 26% of the patients enrolled on the basis of cognitive impairment had negative results on florbetapir scans, concluding that the lack of plaque meant that these individuals did not have Alzheimer’s, or at the very least, did not have sufficient amyloid involvement to make solanezumab responsivity likely. This overlooks the confound that florbetapir only identifies plaque, while solanezumab targets soluble beta-amyloid, not plaque.
Lilly’s final Phase III trial (EXPEDITION3), enrolled 2100 patients who met the criteria for mild AD, infusions given monthly for eighty weeks, and amyloid pathology had to be shown via imaging or CSF assay. In theory, this enhanced the homogeneity of the population, but constitutes a pathophysiological tightrope: Scanned patients had to be advanced enough to have significant plaque deposits that show on PET, but not so advanced that they had anything more than mild cognitive impairment. All of this was for naught: The Sola group showed a consistent pattern of slowed decline on the six measures, ranging from 7% on the FAQ (p=.14) to 15% on the ADCS-ADL (p=.009) and on the CDR-SB (p=.004). On the primary endpoint, the ADAS-cog14, the drug group declined 11% less than the placebo group (p=.095). Lilly acknowledged that the magnitude of observed therapeutic benefit, generally in the 11-15% range, was not clinically meaningful, and thus did not warrant continued development–a refreshing example of ethical clarity–some might have considered filing based on the overall preponderance of evidence, from all solanezumab trials, that the drug has a statistically ‘visible’ effect on disease-progression, even if the clinical impact was so small as to be invisible. This conundrum will be encountered again in the AD area, not everyone will make the ethical choice.
Solanezumab was also selected for two major prospective studies of intervention in nonsymptomatic individuals:
1) The A4 Alzheimer’s study of nonsymptomatic elderly individuals with evidence of amyloid pathology based on florbetapir scans. 1150 individuals are to be enrolled in this five-year treatment duration study, largely funded by the NIH, results coming in 2022.
2) Solanezumab is one of three treatments (along with gantenerumab and a BACEi) used in the DIAN trial, enrolling genetically vulnerable individuals, compared with placebo, plus family members without the genetic mutation. That study, assessing whether any of the three drugs impacts the progression of genetically vulnerable individuals to AD, should have its solanezumab/gantenerumab outcome data in 2019, the arm with JNJ’s BACEi will complete its biomarker (amyloid-PET) phase in 2023, but the drug’s impact on cognitive decline will be tracked longterm.
Aducanumab: Licensed by Biogen from Neurimmune in 2007, this mAB targets an undisclosed beta-amyloid epitope, identified as relevant in elderly individuals who had not shown any cognitive deterioration, the premise being that this reflects AD-‘resistance.’ It is not intuitively apparent why a ‘resistance’-linked target would still be pertinent in an active AD population that was non-‘resistant’. Aducanumab binds to plaque and fibrils, not monomers, which differentiates it from solanezumab.
Biogen ran a 166pt study (30-32 patients per dose-group) PhIIb. There were four dose levels, 1mg/kg, 3mg/kg, 6mg/kg, and 10mg/kg, all infused on a monthly basis. The 6mg/kg dose was added after the study was underway, due to vasogenic edema concerns and a high discontinuation rate with the 10mg dose. The 6mg dose did not separate from placebo at 26 weeks on the CDR-SB or MMSE, unlike the other doses, thus the dose-response relationship was not what was hoped for. The other major cognitive assessment tools used, the NTB and FCSRT, showed no effect from the drug at any dose.
The most positive interpretation of the data is that the drug, at best, produced an approximately six-month’s worth of impact on the pace of degeneration. However, the rate of vasogenic edema in the 6mg group was almost the same as seen with the highest dose (10mg) group. Biogen stated that the cases of vasogenic edema emerged early and eventually resolved, but that was within the context of a clinical trial setting wherein neuroimaging was routine. While 78% of ARIA cases were described as ‘mild to moderate’, 22% were more severe, some requiring hospitalization. Top-line results from the ongoing long-term extension study were just released, going out 36-48 months. The numbers of patients in each treatment group is dwindling, but the trendlines of amyloid reduction and (limited) impact on decline rates apparently remain the same, and the incidence of ARIA (vasogenic edema) is 25%, as detected by neuroimaging. Given what appears to be a limited magnitude of therapeutic impact, the cost-benefit analysis of whether the gain achieved is commensurate with the safety monitoring needed remains very much in doubt.
Because of the association of vasogenic edema risk with the ApoE4 genotype and dosing, the Phase III trials, each enrolling 1350 mild AD patients, are slowly titrating dosing. Originally limited to 6mg doses in the ApoE4+ groups, the EMA eventually gave Biogen permission to go as high as 10mg, the same as for ApoE4- patients. The treatment duration is eighteen months, and unlike the previous trial, these trials will allow concurrent ChEi use. One would hope that an interim futility analysis is planned, but otherwise these trials will not finish until 2020. Even if the hint of a signal proves out in the larger trials, they face the distinct risk that achieving even a borderline level of clinical impact will court the risk of vasogenic edema. It should be noted that there is a minority school of thought that holds that some vasogenic edema may be necessary in order to disrupt the blood-brain-barrier and thereby facilitate brain access for the mAb, but the aducanumab record thus far suggests that the degree of vasogenic edema ‘achieved’ too often goes beyond anything that could be construed as beneficial.
Gantenerumab: The results for aducanumab led Roche to reconsider its plan to shelve R1450/gantenerumab, a mAb (licensed from Morphosys) that attacks both plaque and fibrillar forms of AB. There have been reports of vasogenic edema, not surprising given its plaque-targeting and stimulation of microglial activity, even though less than 1% of the mAb reaches the brain. Originally, the drug entered Phase II, focused on prodromal AD, this was then expanded into a pivotal Phase II/III trial, more than doubling the planned enrollment, to 770, given monthly for two years, Roche conducted an interim analysis that they interpreted as providing a ‘go’ signal, the trial continuing into its Phase III stage, but it was eventually terminated in the wake of an interim futility analysis. The gantenerumab group did show significantly reduced CSF p-tau and total tau over a one year period, and PET-scan data indicating better brain amyloid clearance compared to placebo. Additionally, the interim analysis of the first 312 completers did not include 190 patients who had dropped out, and Roche subsequently concluded that the removal of patients who were rapidly progressing while on placebo prevented the identification of a signal, and that a nascent, dose-related post hoc signal of effect on progression rate could be discerned, and that higher dosing was needed. Interestingly, in the small group of completers who had florbetapir imaging, reductions in amyloid plaque load appeared associated with a reduction in phosphorylated tau in CSF, suggesting that impacting either of the major AD targets can impact the other.
The drug continues in another Phase III trial that began in 2014, aimed at enrolling 1000 mild AD patients, but it was cut back to 389 patients, data expected next year. In an attempt to focus higher dosing on a slightly earlier-stage population, Roche has recently initiated two 760pt, two-year duration Phase III studies in MCI/prodromal AD patients, which will report in 2022. Along with solanezumab and a BACEi, gantenerumab is offered in the DIAN trial (previously described) as a preventative/slowing option; enrollment is complete, data for the two mAbs is expected in 2019.
Crenezumab: This AB antibody program came from AC Immune, who outlicensed it to Genentech in 2006. It is based on an Ig backbone thought to be less pro-inflammatory than its peers in terms of stimulating microglial activity, it does show activity against both soluble and insoluble forms of beta-amyloid. It is likely to have a lower risk of vasogenic edema, and was chosen as the disease-modifier of choice for the API trial, as is described below.
Crenezumab failed in a 431pt Phase II trial that enrolled both mild and moderate AD patients, using both subQ and IV delivery. The IV was deemed equivalent to a ‘high’ dose, and in the mild subgroup of patients, a statistically significant impact (3.44 point mean difference) was seen on the ADAS-cog, reflecting a 52% reduction in the rate of decline on that endpoint, along with a 41.5% reduction in the rate of decline on the CDR-SB. Post hoc analyses of the high-dose (n=247) population by MMSE score at entry showed a relatively consistent trend towards cognitive preservation, on the ADAS-cog and some domains of the CDR-SB, in more mildly impaired patients, with an inverse correlation between baseline severity and improvement. Such posthoc population parsing is fraught with risk, but it was sufficient to convince Genentech to run another trial. A 750pt Phase III, using a two year treatment period and restricting enrollment to patients with mild AD (MMSE>22) is projected to finish in 2021.
Crenezumab also continues in the API study, a prospective trial wherein the mAb is used as prophylaxis in members of an extended Colombian family (approximately 5000 members) with a presenilin gene variant associated with very early onset AD. A variety of biomarkers are being tracked, and the trial will run until 2020, though interim results are expected before then. Such prospective studies should begin to answer the question of whether amyloid antibodies work when employed early enough, if one assumes that the API trial, this particular presenilin mutation, and the accelerated dementia that it produces, are sufficiently similar to sporadic AD to be predictive for the nonfamilial type. Which will cast a significant shadow of doubt across any trends that may emerge.
Roche has thus doubled down on the amyloid mAb concept; they control two Phase III mAB candidates that have each offered just enough posthoc promise–from their own data and from a competitor’s–to offer a rationale for continuation. Both crenezumab and gantenerumab are also in prophylaxis studies that are partly funded by other entities, and will not report data for years. NIR had previously commented that it was hard to imagine Roche funding two amyloid mAb programs in parallel, not when there are other targets worthy of exploration and investment–but they are indeed doing so, at a cost that is reported to be over $4 billion. We continue to doubt that they will end up getting ‘their money’s worth.’
BAN2401: BioArctic (subsequently partnered with Eisai) developed an antibody that attacks AB protofibrils. BAN2401 was subsequently repartnered by Eisai with Biogen, and was the subject of the most recent round of hyperbolic Alzheimer’s headlines. BAN2401’s Phase II data from a substantially scaled dose-finding trial (854 patients treated, four dose levels ranging from 2.5mg/kg bimonthly to 10mg/kg bimonthly) were billed as showing both a significant impact upon amyloid levels and a dose-related impact upon progression rates. The highest dose was reported to show a 30% slowing of decline on the ADCOM (a composite endpoint developed by Eisai) at eighteen months; a 46% slowing of decline on the ADAS-cog, and a 26% slowing of decline on the CDR-SB, all compared with the placebo group. Which would most likely have exceeded even Biogen/Eisai’s expectations, had only it been genuine. Once again, a major pharma company(s) had tripped itself up with an unrealistically parsed presentation of data from a high-profile Alzheimer’s amyloid-antibody trial. They neglected to mention a confounding factor that invalidated all of the comparisons between the highest-dose cohort and the placebo group: ApoE4 genotype. While the study was underway, regulators became concerned (due to aducanumab) that the risk of vasogenic edema (VE) was excessive when combining the highest dose with ApoE4+ genotypes, and stopped the enrollment of ApoE4+ patients in the high-dose group, while withdrawing ApoE4+ patients from the study who had not been in for at least six months. What this produced was a highly apparent discrepancy between the 71% of patients in the placebo group with an ApoE4+ genotype, versus the 30% in the high-dose group.
This matters, because the best evidence available indicates that ApoE4+ alleles are associated with more rapid clinical decline in Alzheimer’s. Following a spate of studies with conflicting findings, Holtzman et al reported in Nature last year data from the state-of-the-art ADNI study: Over a ten year period, patients homozygous for ApoE4 declined at a 23% more rapid rate than ApoE4 negatives, heterozygous patients declined 14% faster than ApoE4 negative patients. Which strongly suggests that the slower decline in the BAN2401 highest-dose group was partially due to their having a less-pathological ApoE4 genotype. Biogen/Eisai could have, but did not, report on what percentage of BAN2401-study patients were homozygous as opposed to heterozygous, but a simple eyeball assessment suggests, indeed compels, the conclusion that the reported rates of slowed decline on the aforementioned outcome measures were tainted by this confound, and should have been reported with an bold-type asterisk, if reported at all. Our unsophisticated assessment of the outcome curves suggests that when the ApoE4 imbalance is taken into account, that BAN2401’s impact on progression is in a range where clinical significance is dubious.
The second issue is safety, vis-a-vis ARIA/vasogenic edema. One out of every seven (14.6%) of ApoE4+ patients who did receive the highest dose-regimen of BAN2401 developed ARIA, though only three out of the 48 cases were considered ‘severe.’ While this appears to indicate a less problematic prevalence of ARIA than seen with aducanumab, if BAN2401 were deployed in the Alzheimer’s population as a clinical option, that might still be considered an unacceptable risk given the sheer numbers of patients involved.
As was the case for aducanumab, one can argue that there are hints of success to be found here; this antibody was at its essence designed to reduce amyloid levels, and the PET scan data shows that it does so, in a dose-related manner. But whether such beta-amyloid reductions lead to clinically meaningful impact on disease progression is still not proven. Both of the 10mg hint-of-a-signal groups also display a not insubstantial rate of ARIA/VE for both E4+ and E4- groups. In other words, efficacy and risk continue to be inextricably intertwined for the beta-amyloid mAbs, and whether there is enough of the former to justify the latter continues to be unclear.
A German group (Technical University Munich) reported that some amyloid mAbs appear to increase cortical neuron hyperexcitability in the presence of AB, which could have a negative effect on cognitive function. This has yet to be shown in humans. One hypothesis is that this might occur as a result of breaking down plaque into oligomers that directly trigger hyperexcitation. Thus, beyond the questions of brain access and target suitability traditionally raised, there is a possibility that amyloid mAbs could trigger deleterious consequences that offset some or all of the benefit that might have been expected from AB dissolution/clearance.
from the November-December 2017 Issue
Parkinson’s Review: The Gut-Brain Connection
This is a hot topic in the neurodegeneration field, nowhere more than in Parkinson’s. An Emory University group has reported that inflammatory cytokines (e.g. IL-8, IL-6) are elevated in colonic tissue and stool samples from PD patients. There is a highly speculative theory that categorizes PD as an age-related immunodeficiency syndrome, chronic infection and inflammation setting the stage for an increased vulnerability to PD: In this model, chronic inflammation does not cause PD, but it creates a permissive environment wherein it is more likely to manifest. Various researchers have produced contradictory profiles for the gut bacteria that they consider emblematic of prodromal PD, thus this is a narrative that is still early in development. It has also been speculated that the early development of hyposmia (impaired sense of smell) could be the manifestation of infection/inflammation via another vulnerable somatic interface, the nasal passages.
This is congruent with work published by Axial Biotherapeutics, a newco established around gut-brain pathology work done at Caltech. A PD model (published in Cell) showed that short-chain fatty acids produced by gut bacteria promoted CNS inflammation and motor symptoms. The group transplanted microbes from the gut of PD patients into germ-free, alpha-synuclein transgenic mice; compared with mice who received transplants from non-PD human sources, the PD-microbes were associated with greater motor dysfunction.
Suggesting something more than a ‘permissive’ environment is data from a Danish retrospective study (Annals of Neurology) of individuals who had their vagus nerves severed as a treatment for ulcers–some 15,000 patients were in the database. The data indicated that the risk of developing Parkinson’s was cut by half, but partial transections did not have this effect. The researchers hypothesized that PD begins in the GI tract, and travels to the brain via the vagus nerve.
This gut-brain connection is a conceptual parallel to a theory advanced by a UCSF researcher that suggests gut-based bacteria infiltrate the brain in the course of neurodegenerative disease, yielding an unorthodox neuroprotectant approach underlying Cortexyme, looking to commercialize that work. Their premise postulates the entry of an unspecified bacterial protease into the brain, interacting with misfolded proteins, leading to neurodegeneration. Cortexyme has an antibiotic that targets that protease, being readied for the clinic, with Alzheimer’s their primary indication.
This is semi-reminiscent of work previously done by Rudy Tanzi’s group on amyloid as an anti-infection agent, with misfolding/aggregation impairing that function. There is speculation that alpha-synuclein may have a similar role, thus Parkinson’s is a potential second target.
As was noted earlier, Takeda has partnered with Prana Biotechnology on PBT434, which Prana had left relatively neglected as it focused on Alzheimer’s and Huntington’s. Prana has some animal data indicating motor function restoration, but Takeda is specifically interested in the compound’s impact upon GI dysfunction, including the assessment of its impact in an alpha-synuclein mouse model. They hope to be in Phase I this year.
This is far more thought-provoking than validated at this point, but reflects the increasingly prevalent view that the CNS is best conceptualized and evaluated as part of an organismic whole, including the ‘inflammasome’ and immunomodulatory systems.
Pride Goeth Before the Fall: Axovant and Hubris
It is possible that intepirdine’s trial in Lewy Body Dementia may yet yield positive results. There is a significant cholinergic deficit in LBD, and with intepirdine being given as a monotherapy, it may have a better shot than as an adjunct to another cholinergic drug, as was the case in its AD Phase III. There may be a limit to how much one can drive a dying cholinergic system. As was the case in AD, there is little chance of anything robust, but even a mediocre treatment effect is something that has commercial potential.
But not like in Alzheimer’s, a market fifty times larger than is LBD. Axovant has not suffered a fatal blow by any means, but intepirdine’s prospects must be dialed down considerably. Which may have a salutory effect on Axovant’s primary defining characteristic to this point, hubris. Give the founding CEO Vivek Ramaswamy credit, he charmed some pretty astute investors (and some who are not so astute) into giving him over a half billion dollars, including an enormous IPO, at $315 million. All this primarily on the basis of Ramaswamy convincing them that Axovant was the proverbial ‘smartest kid in the room’, able to perceive and exploit the potential of a drug GSK had written off, and was willing to outlicense for just $5 million upfront. But NIR was never enamored of the RVT-101/intepirdine thesis in Alzheimer’s. As we said in NeuroPerspective’s Alzheimer’s review in September:
“…the premise is that an earlier, mediocre differentiation from the control group can be buttressed and enhanced via a higher-powered trial. Our view is that the higher-powered trial may provide intepirdine’s results with statistical significance, though the idalopirdine outcome is a cautionary note that trial size does not guarantee a better statistical profile. But even if it does accomplish this for intepirdine, GSK’s data points to a clinically mediocre, arguably meaningless effect.”
‘Discovering’ a second-in-line 5HT-6 antagonist as a potential second-in-line competitor for pimavanserin, and ‘discovering’ Lewy Body Dementia, are markers of a highly derivative and incremental road map, not a dramatic leap forward for CNS drug development. Ramaswamy once told NIR that “I want to cure Alzheimer’s;” it was in retrospect, perhaps unfortunate that we did not immediately respond with: “Well, you’re going about it in a really weird way.” Ramaswamy was quoted in an interview (with Forbes‘ Matthew Herper and Nathan Vardi) following intepirdine’s Phase III failure as saying that ‘“This is the single greatest failure of my career…I feel much more accountable and it hurts me to have disappointed others who took a bet on something I believed in…It’s a humbling day.” Good. Given his undeniable salesmanship skills, this could be a positive turning point, because it would be terrific if Roivant/Axovant now turns some of its talent and resources to projects that might make a real difference in clinical care.
from the September-October 2017 Issue
Alzheimer’s: The Plasma Fountain of Youth
Alkahest is working on a therapeutic program that is a next-generation elaboration upon something of a throwback neuroregenerative option; taking blood plasma from the young and infusing it into older adults in the hope of spurring neuroregeneration. Regeneration through the absorption of ‘young’ blood is a theme that has percolated through primitive rituals, vampire tales, and some ill-fated experiments of early medicine; but this program, spun out as Alkahest, incubated in a JNJ Jlabs facility, comes from work done at Stanford, which sets it distinctly apart from these predecessors. Animal experiments there showed that infusing plasma from young mice into old mice stimulated the generation of new brain cells, and improved cognitive functioning/learning, while injecting plasma from old mice into the young had an opposite and deleterious effect. Work published in Nature Medicine showed that plasma improved synaptic plasticity, operating via CREB, critical to learning/memory. In 2015, Grifols acquired 45% of Alkahest for $37.5 million, and made an additional $12.5 million payment, in return for worldwide rights.
While seeking to isolate which particular factors were upregulated via young plasma, the simplest tactic for establishing preliminary HOC was to infuse plasma itself. An 18 patient, double-blinded Phase II trial was initiated, using weekly plasma infusions from individuals under age thirty, in patients with mild to moderate (MMSE 12-24) Alzheimer’s, with a trial duration of nine weeks. Most standard cognitive tests were utilized as efficacy measures, nine weeks into the trial. The trial was finished earlier this year, preliminary results will be presented at CTAD. It is highly unlikely that any marked contrasts will be seen, since this is a tiny, short-duration trial, and it is not known whether (as is the case in other AD trials) these patients may be too advanced in their disease to benefit from the trophic effects of plasma, while it is also not known whether the donors themselves are too old to provide maximally ‘potent’ plasma in terms of the inherent level of regenerative factors. If there is a therapeutic product to be had here, this would raise all sorts of practical and ethical issues around plasma supply, and IP regarding the application of the technology. As NIR has commented tongue-in-cheek in the past, one can only imagine the dinner table conversations between middle-aged parents and their teenage offspring: “Yes, you can borrow the car, but there’s something you have to do for us in return…” This even made it onto HBO’s satirical comedy, Silicon Valley, where a narcissistic CEO was depicted lounging at home with his plasma donor.
From a longterm pragmatic perspective, identifying critical regenerative factors would be far more useful, and those Stanford researchers reported (in Nature) that they had identified a key protein in cord plasma that they believe accounts for the regenerative effects of umbilical cord plasma in mice: TIMP2. TIMP2 levels are reduced in the hippocampi of older adults, and infusing mice with the protein induced the same procognitive changes as did cord plasma. NIH Director Francis Collins rhapsodized about the finding in his blog: “...might the elusive elixir of youth that people have long pined for been inside us all along?” As we have noted previously, that is more than a tad hyperbolic, but the TIMP2 finding suggests both a mechanism and a commercial path forward for Alkahest, if TIMP2 can be developed as a therapeutic agent. Alkahest believes human testing is at least two years away (it does not fall under their Grifols agreement, which is limited to plasma products). The premise is not universally embraced: Science quoted a researcher from across the Bay (UC Berkeley‘s Conboy) questioning whether the magnitude of benefit was truly noteworthy compared to more mundane interventions, like exercise, while postulating that the contrast between cord and aged plasma is actually due to the presence of degeneration-associated proteins in aged plasma.
from the July-August 2017 Issue
Spinal Cord Injury: Inhibiting the Inhibitors
As was noted above, barriers to regeneration/rewiring are both physical and biochemical, and finding a way to interrupt growth-inhibiting signals has been perhaps the most heavily pursued strategy in SCI pharmacotherapy R&D. Many of these factors are components of myelin, easily found in the myelin ‘debris’ around the spinal injury site. Three of the inhibitory factors found in myelin are MAG (myelin-associated glycoprotein), OMGP (oligodendrocyte myelin glycoprotein), and the best known such factor, Nogo-A/Reticulon-4, These all bind to different areas of the Nogo receptor, which itself is part of a complex involving that receptor plus either p75 or the protein TROY (both related to the TNF family, and hence relevant to inflammatory processes), and/or the protein Lingo-1. When this receptor complex is activated via any or all of these binding sites, the next step along the pathway is Rho, which is discussed in more detail below. When Rho is activated, the growth cones that guide axonal extension collapse, and growth stops. Additionally, inhibiting the inhibitors at this stage can also mean inhibiting apoptosis, since p75 is involved in apoptosis. Classifying this pathway as purely regenerative is something of a misnomer used for the sake of brevity: Choosing the right target will not only allow regeneration to proceed, but will also protect surviving tracts from subsequent apoptotic cell death.
From a historical perspective, this crucial understanding of endogenous ‘anti-growth’ processes dates back to when Martin Schwab first identified Nogo, in 1988. Nogo has been recharacterized over time as a regulator of axonal sprouting, and multiple programs aimed at suppressing Nogo have been conceived–and demised. The first tactic licensed by Schwab to Regeneron involved the use of monoclonal antibodies (NI-35, NI-250), but those antibodies eventually proved nonviable. Novartis (with whom Schwab subsequently worked) eventually produced another antibody, IN-1. In a primate model, where 90% of tracts were destroyed, full function was eventually restored. IN-1 blocked the Nogo protein itself, and was the basis for ATI-355, a product that went through a Phase I trial (the FDA did not allow US sites due to safety concerns) that utilized intrathecal infusion, beginning within 4-14 days of the injury. The program eventually disappeared from Novartis’ pipeline; Schwab was reportedly going to run another trial in Switzerland, at his own University of Zurich laboratory, but there is no evidence that this ever started.
Biogen took a slightly different tack to Nogo, working with NEPI-40, a Nogo antagonist peptide, but that program was dropped. Sanofi had a program pursuing the Nogo target, but sidelined it. GSK ran a trial in ALS with their Nogo mAb, ozanezumab/1223249, rather than pursuing SCI, and after that trial showed no therapeutic benefit, they dropped the program. The initial allure of Nogo as a target has pretty much vaporized. While it plays a role in the pathological cascade, it does not appear to constitute a viable target.
AbbVie did some discovery work in this area, in regard to the use of a selective antibody against Nogo Receptor 1/Nogo-66, reporting in 2013 that mAb 50 did enhance functional motor recovery in a rat model of SCI. But they have shifted their SCI target to Repulsive Guidance Molecules (RGMs). ABT-555/elezanumab is an antibody to RGMa, whose presence on the cell surface inhibits axonal regrowth and remyelination. In 2015 they initiated a PET-study of ABT-555 in the UK, aimed at RRMS; that trial was suspended last October, and terminated this past February. But this was not due to a safety issue, AbbVie has begun enrolling a 30pt Phase II trial in RRMS, with a six-month treatment duration. They cite early 2019 as the expected completion timeframe, but they are also planning to initiate a Phase II trial in SCI, perhaps this year.
There are other candidates for the role of inhibitory mediator, through which growth-inhibiting signals from various sources must pass, that offers an attractive target for blocking these inhibitory processes, and in our view, Rho has been the leading possibility. As was noted above, activation of the two major Nogo complexes (combined with either p75 or TROY) eventually leads to Rho activation. Rho can also be activated by proteoglycans found in the scar tissue, the aforementioned neurochemical obstacle arising from the extracellular matrix components of scarring. Rho antagonism thus blocks not only the impact of Nogo, but also impacts pathways triggered by other growth-retarding factors, including proteoglycans and axonal repulsion factors. Rho proteins modulate signal transduction within the growth cone itself, controlling axon growth and cell proliferation. Blocking Rho promotes neuroprotection and axon growth, working upstream of nogo, and both in vitro and in vivo studies have shown axonal growth after Rho antagonism. One concern with the Rho target is the fact that Rho proteins are ubiquitous throughout the body, but no safety/tolerability issues have arisen in the clinical testing done thus far.
There were originally two companies pursuing the Rho target, BioAxone and Migragen. Both took a known Rho-antagonist with very poor absorption, the enzyme c3-transferase, and modified it. Migragen combined CT-3 with a component taken from botulinum toxin, which increased membrane permeability. But Migragen ran out of money and sold their IP to Schering, who did not have interest in it for therapeutic development.
In contrast, BioAxone has persevered (albeit in two corporate iterations) for more than fifteen years with the development of Cethrin, a recombinant version of c3-transferase that antagonizes Rho. It is neuroprotective (reducing apoptotic cell death 50% in one model), and reduces TNF-alpha, thus reducing inflammation and scar formation. It is effective in promoting growth, and no significant treatment related adverse events have been seen.
In a 48pt (mean time to treatment was 52 hrs post-injury) open-label PhI/II program using five doses of Cethrin, 43% of the patients showed functional gains of two ASIA grades or more, from a starting point at ASIA-A (complete loss of function below the level of injury). Some improved up to Level D, where at least half of the muscles innervated from below the injury regained significant functionality. In the 12 patients with cervical injuries (thoracic injury patients tend to show little benefit from interventions, and were included primarily to assess safety), the mean improvement over twelve months was 27.3 points for the 3mg group, 21.3 points for the 1mg group. Historical control data suggests only about 10% of ASIA A patients show this level of ASIA-category improvement, and it usually occurs fairly early in the post-injury period. Motor function and sensory improvement were noted, and no adverse event or tolerability problems were reported.
Historical control comparisons must be viewed with considerable skepticism, though these results are striking. The the lack of a control group in that dataset considerably complicated the process of finding partners, and the task of assembling and completing a placebo-controlled Phase II/III trial turned into a marathon, one not yet fully run. BioAxone originally partnered Cethrin with Alseres in 2006, but Alseres failed to carry out the promised PhIIb trial, and BioAxone eventually (and with no small amount of effort expended) regained those rights and reconstituted itself. In 2014, BioAxone found a highly credible partner in Vertex Pharmaceuticals, which paid $10 million upfront, the deal included an option to acquire BioAxone. A placebo-controlled Phase IIb for VX-210 began in February 2016, aimed at enrolling 150 patients with acute cervical SCI, using two dose-levels, surgery done within 72 hours of the injury. Improvement over six months in upper extremity strength is the primary endpoint, and that trial is projected to finish next June, a highly significant dataset for the SCI world.
LRP1(Low-Density Lipoprotein Receptor Protein-1) has been identified by Novoron Bioscience as a target for intervention, based on LRP1’s role as a convergence point for multiple inhibitory factors, with the downstream effect of hyperactivating Rho-A. Novoron is working with a biologic (a recombinant derivative of RAP–Receptor-Associated Protein–a chaperone protein) that prevents that hyperactivation. Working with grant funding thus far, Novoron is first pursuing remyelination in MS, but SCI would be the next indication-of-interest.
Yale‘s Martin Strittmatter developed a fusion protein which served as a decoy receptor, binding three anti-regenerative, myelin-based factors in combination: Nogo, MAG, and OMG. Strittmatter’s animal work indicated that inhibiting all three nogo receptor ligands provided additive benefit, exceeding what could be achieved via a single factor. The belief has been that working with chronic SCI might be easier, given that six months out, natural recovery has run its course, thus removing an element of placebo response, and permitting the use of smaller patient samples without the pressure of trying to enroll and treat within a few days of the injury. This program was a core component of Axerion Therapeutics, which has been reorganized as ReNetX, and at present they are not disclosing the status of any of their inhouse programs.
Much of the work in this area as it pertains to SCI has been continued by academic groups. A Harvard Medical School-affiliated group has worked on the OMGP pathway to Nogo modulation, while a Johns Hopkins group has reported animal work wherein recombinant neuraminidase blocked MAG and enhanced motor function.
A Genentech group identified leukocyte immunoglobin B2 (LILRB2, also known as PirB–generally thought to have a role in controlling inflammation, since it is a receptor for MHC antigens) as being bound by several inhibitory factors, including Nogo-A, OMG, and MAG. In mice, a PirB antibody significantly improves neurite growth, suggesting that LILRB2 antagonism in humans might be another broad-spectrum approach to restricting axonal growth inhibitors. However, Genentech‘s work on LILRB2 focused primarily on its potential role in Alzheimer’s, not SCI, and that program appear to have been culled.
Articulating the Lingo
Biogen has pursued utilizing the Lingo-1 component of the growth inhibitor complex as a target: After a promising trial in acute optic neuritis, wherein the mAb against Lingo-1(BIIB033) improved the speed of signal transmission, they had disappointing results in RRMS, which they attributed to anomalies in the dose-response profile (they intend to resume testing in RRMS). They have not articulated any plan to go into SCI, but if they do succeed in eventually spurring remyelination in RRMS, SCI would be a logical avenue for label-extension.
Whereas the growth-inhibitory factors described above, and the physical barriers posed by proteoglycans discussed later, constitute obstacles outside of the neuron that slow or prevent regrowth post-injury, there are also components within the neuron that pose an obstacle to regrowth. The highest-profile intrinsic factor is PTEN (phosphatase and tensin homolog), a protein that regulates cell growth. Due to its role in tumor suppression, it was long pursued by a number of companies in oncology, who have sought to upregulate its action. However, PTEN’s restrictive effect on cell growth also led to a flurry of research by a number of academic groups, particularly at Harvard and UC Irvine, looking at its inhibition in SCI, in the service of fostering axonal regeneration. This came to a head in 2014, when the UC Irvine group published work showing dramatic results when rats, injected with an AAV vector delivering PTEN inhibition (via RNAi), showed impressive functional recovery (up to 95%), but only when the PTEN RNA-silencing vector was accompanied by salmon fibrin, believed to provide a scaffold for the regenerating axons. In this model, the PTEN-silencing and fibrin injection were done before the induced injury, and while the lead investigator said that recovery is also seen with treatment concurrent with the injury, it has been unclear if this intervention, administered post-injury, would have the same effect. It is also not clear whether PTEN knockdown can be accomplished without the risk of tumor development: PTEN has been a target-of-interest in the treatment of glioblastoma, and given the relative youth of the SCI population, safety over the very long run would have to be established. BioAxone and the Harvard/Boston Children’s Hospital group did receive some NIH/NINDS funding for their work on developing RNAi to knock-down PTEN expression, BioAxone’s BA-434 is their PTEN RNAi payload, a ‘self-delivering’ RNAi factor that has good cell permeability even without utilizing a vector (like AAV). In rodents BA-434 knocks down PTEN with an extending duration of effect, and in rat retina, promotes neurite growth and axonal regeneration.
Bob Yant, who founded the SCI-supply company CureMedical, and who himself has quadriplegia following an SCI, has licensed some of Harvard‘s IP around PTEN, and is collaborating with venBio‘s Corey Goodman in forming a newco that will focus on PTEN in SCI; no details have been made available.
from the May-June 2017 Issue
Insomnia: Paradigm Shift-Orexin Antagonism
The orexin system, also targeted by the alertness drug Provigil, became the context for the next real paradigm shift in insomnia treatment, orexin receptor antagonism. The key concept here was that, in contrast to other insomnia drugs in clinical use, which upregulate sleep (and sedation) associated pathways, orexin antagonism downregulates excitatory/arousal pathways, or as it was put in the popular press, ‘turning down the wake switch instead of turning on the sleep switch.’ This, at least in theory, might provide a different risk-benefit profile.
Merck became the trail-blazer in this area, as a GSK/Actelion entrant faltered (see below) and they successfully developed the dual orexin subtype (OX1R and OX2R) antagonist MK-4305/suvorexant. In two 1000+ Phase III trials that followed patients for three months, suvorexant’s high dose (40mg, 30mg for elderly) hit all of its co-primary endpoints, reducing TSO (sleep onset delay) by 8.4 minutes and 12.2 minutes, and WASO by 22.9 and 19.4 minutes. In a 737pt 12-month study, suvorexant improved sleep latency, and was reported to have no impact on next-day driving, whereas zopiclone did negatively impact driving. Where the suvorexant story became particularly surprising was at the FDA. The FDA’s Advisory Committee voted 13-3 to support starting doses of 15 and 20mg (in elderly and non-elderly respectively), but 8-7 against the higher doses (30/40mg) that had produced Merck’s most clearly positive results. There was some discussion of whether the high dose was sufficiently tolerable in terms of residual effects, and whether Merck should be required to run an even lower-dose trial (10mg). But after receiving manufacturing data for the 5 and 10mg dose tablets, the FDA eventually approved the NDA for suvorexant at 10mg, a dose that even Merck did not see as efficacious (depending on endpoint, optimal dosing varied from 20 to 40mg), and Belsomra was commercialized. At time that this occurred, NIR commented that, while we were glad to see Merck’s work rewarded and Belsomra approved, that this was a potentially confusing outcome in terms of guidance for those who might follow Merck. It can also be argued that, while there was obviously some benefit to Merck in not having to run another Phase III, approval at a subtherapeutic dose is not without its own cost: Patients and prescribers can grow disenchanted by an initial lack of impact, and while some might then escalate dosing, some will not, and this is not the kind of ‘word of mouth’ experience that gives a new drug momentum. Merck has still not broken out the sales for Belsomra, which suggests that the numbers are weak: Would they have been better off had they further pursued a starting dose that matched their own efficacy data? We will never know, but this was a downside to the FDA’s flexibility wherein they prioritized the avoidance of side effects, even those encountered with any hypnotic, over efficacy.
In any event, Belsomra’s clinical case vis-a-vis existing drugs was not definitively made by its pivotal testing, and some payors are treating it as a higher-tier option to be utilized in cases where generic hypnotics have not been satisfactory. Kaiser Permanente, for example, requires failure on four generic drugs before they will cover Belsomra. The more euphoric analyst predictions for Belsomra’s sales growth have been ratcheted down quite a bit.
The efficacy questions did not end with the FDA: The Australian Department of Health initially rejected Merck’s application to sell Belsomra there due to what they viewed as moderate therapeutic benefit accompanied by the same kind of next-day drowsiness as plagues current drugs, along with ‘abnormal sleep behaviors and suicidal ideation’ (albeit without any particular evidence of the latter; rare instances of sleep paralysis, and about a 10.6% rate of somnolence at the 20mg dose). They eventually did approve the 15 and 20mg starting doses. Unlike the FDA, they did not approve the 10mg dose, nor did they approve Belsomra for longterm use (the low-dose trials only went out three months).
Merck did have a follow-on molecule, filorexant/MK-6096, also a dual receptor antagonist, but they dropped development after Phase II testing was done in 2014.
There is some debate in the field as to whether blocking both orexin receptor subtypes inevitably means that the pro-sleep effects of OX2R antagonism will be offset by sedation/dulling associated with OX1 antagonism. Rodent studies indicate that OX2R antagonism is more critical for improved sleep, but dual-receptor antagonists appear to induce sleep somewhat faster. Regardless of the specificity issue, the fact that suvorexant’s half-life is in the 9-13.5 hour range likely accounts for the next-day drowsiness risk.
Minerva/JNJ‘s MIN-202 could provide the first opportunity to test the specificity thesis. MIN-202 is an OX2R antagonist that was developed by Janssen/JNJ (as JNJ-42847922) then licensed by Minerva Neurosciences for co-development in 2014, when insomnia was deprioritized at Janssen. A 28pt Phase IIa showed MIN-202 did improve sleep parameters over a five day period, somnolence and abnormal dreams were reported. Minerva has also done testing parsing MIN-202’s impact on sleep from its benefit upon depression in pilot studies, but has not yet initiated a Phase IIb study in insomnia, perhaps hoping for either a partner or funding from licensing another of their assets.
Eisai has developed their own dual orexin-receptor antagonist, lemborexant/E2006. In a 291pt Phase II, five dose-levels all showed efficacy over a fifteen day period, only the highest dose appeared associated with next-day drowsiness (the half-life has not been disclosed). It should be noted that drowsiness was measured via self-report, not via any objective cognitive or functional testing. Purdue Pharma partnered with Eisai, in an agreement that includes cost-sharing plus co-promotion in the US. Purdue has the advantage of having brought in the Merck scientist who drove Belsomra’s development there, saving them from having to ‘reinvent the wheel’ on many points. A 950pt Phase III in individuals over 55 will finish 3Q:17, comparing lemborexant to zolpidem-CR. Another Phase III, enrolling 900 patients and following daily use for twelve months, is expected to finish late 2018. There is also a Phase II trial in Alzheimer’s patients with sleep disturbance.
There was a time that GlaxoSmithKline and Actelion were at the head of the orexin field with the dual receptor antagonist almorexant. Actelion stated that they achieved PhII benefit in both induction and sleep maintenance, without impacting sleep architecture, and claimed that no cognitive sequelae were seen, though we have heard that questioned by interested observers who had seen the data; almorexant’s 40hr half-life would seem to guarantee residual adverse effects. Phase III used zolpidem as the active comparator, and Actelion announced completion at the end of 2009, successful save for some “safety observations.” Those must have been significant, because eventually the partners dropped almorexant, shifted to a backup orexin antagonist, and then GSK exited completely.
Actelion has carried on: The backup ACT-541468 is in a 300pt Phase II trial expected to complete later this year, using zolpidem as an active comparator. A 50pt Phase II in elderly patients should finish around the same time. All of this will now be done under the auspices of the R&D newco (run by Actelion management) spun out of JNJ/Actelion (along with a $1 billion in funding) when JNJ acquired Actelion earlier this year. GSK itself had an inhouse orexin dual receptor antagonist, SB-649868, but they dropped this early on, due to an undisclosed toxicity issue. Novartis disclosed OX2R selective compounds back in 2013, which were 30X more selective for OX2 than OX1. It does not appear that Novartis did anything with these assets.
Heptares Therapeutics has done early-stage work on selective OX1R antagonists, applicable to compulsive behavior disorders and addiction, but has not discussed any interest in OX2R and insomnia.
The major unknown/undisclosed variable for MIN-202, lemborexant, and ACT-541468 is half-life, since extended bioavailability would seem likely to lead to the risk of next-day residual effects. But then again, zolpidem is not without the risk of residual effects, and its half-life is cited as being in the two-three hour range, so this is not such a clearcut criterion for viability.
from the January-February 2017 Issue
The Year of Living Dangerously–the Sequel, Coming to Everyone’s Neighborhood….
2016 passes into the history books with hardly a ripple, the progress of humankind towards a brighter tomorrow smoothly advancing without so much as a hiccup. 2017 promises to be modestly more dramatic, with pundits predicting outcomes ranging from Apocalypse to Armageddon. The stock market has been booming. Somewhere, we can hear Nero fiddling as the Empire binges on ayahuasca.
When it comes to the biopharma world in general, there was a sense of déja vu. Another year, another amyloid antibody yielded its bounty of disappointment. Some pharma observers believe that the sudden 180 degree turn in Presidential mindsets means that the heat on the industry will be off, the pressure on pricing diminished by an industry-friendly administration. It does not look that way to everybody. NIR quoted Allergan‘s Brent Saunders in November, now we cite this excerpt from a piece he wrote for the December 1 issue of Forbes:
“…our industry has a social contract. Patients understand that making new medicines requires significant investment. Companies doing the hard, long and risky work of bringing new medicines to market have generally understood that they have to price medicines in a way that makes them accessible to patients while providing sufficient profit to encourage future investment. Everybody wins when it works the way it should. But some have violated this social contract, and the whole industry is under attack. The election results don’t change that debate. The election may give the industry time to demonstrate that it can self-govern. Time to demonstrate that the social contract is alive and well. If not, our elected leaders may act without us–at the peril of innovations that could help millions of patients.”
Citing Tufts estimates for the cost of developing drugs, and whining about the costs of Innovation, have been burned out as arguments with the public audience. They simply do not believe us, and an administration with a limited tolerance for scientific truth, a minimal attention span, a low bar for ‘facts’, and a penchant for populism cannot be relied upon as a shield when we offer such a convenient target: It’s not as if we make something useful, like air conditioners. The bill is coming due for Pharma’s public relations, which have been a chronically abysmal failure. We have allowed callow sociopaths like Martin Shkreli to seize the limelight and frame all of us as bathed in a miasma of unmitigated greed. This is not accurate, and is not fair, but we have failed to make our case or live up to our part of the social contract. If we do not show some collective spine in developing self-discipline, it will be done for and to us.
from the January-February 2017 Issue
Everything You Always Wanted To Know About the Amyloid Hypothesis (But Were Afraid To Ask Because the Shareholders Might Not Like What They Hear)
“Maybe we’re all just stumbling from the right questions to the wrong answers, or maybe from the right answers to the wrong questions.” –Malek, R. Mr. Robot 2016
“This (the sola study) is confirmation of the amyloid hypothesis, our strongest confirmation to date.” –Aisen, P. 12/8/16 CTAD
Which just shows how low the amyloid bar is, confirmation-wise. But in this instance, the conversation ricocheted from the wrong question to the wrong answer.
The wrong question: ‘What do these results say about the amyloid hypothesis?’
This infers that there is a single, archetypal amyloid hypothesis (AH), but in fact there are several, because there are multiple iterations of the Amyloid Hypothesis which have evolved over time as the earlier versions began to show their flaws.
AH 1.0: The original hypothesis says that Alzheimer’s is caused by the accumulation of amyloid plaque within the brain. If AH 1.0 were correct, aducanumab’s ability to reduce plaque deposition by 25-30% should translate into tangible cognitive/functional benefit. But while the 10mg dose showed a tentative signal on the CDR-SB, the 6mg dose, in spite of its similar PET-scan credentials, was a flop at 110 weeks. That could be due to the variance born of small cohorts, only a larger trial will tell. It is worth noting that solanezumab, in spite of not binding to plaque at all, produced a consistent albeit clinically meaningless effect on cognition.
Any question about AH 1.0 is the wrong question. It’s like asking about the ‘link’ between vaccinations and autism.
AH 2.0 postulates that amyloid in some non-plaque form (e.g. fibril, oligomer, monomer) is the causal key to Alzheimer’s. One counter-argument to AH 2.0 is offered by solanezumab, which produced a dramatic effect upon plasma amyloid, increasing it 500-800 fold, but the weakness of the therapeutic effect in the face of that biomarker does not portend a robust relationship. Sola’s first two Phase III trials showed an increase in total CSF AB40 as well, again without any connection to a clearcut treatment effect. The sola results neither prove nor disprove AH 2.0: We do not know if the weak response promoted by solanezumab means that the antibody binds to a suboptimal form of AB; not enough antibody reached the brain to bind enough AB; the patients were too advanced to benefit from a post hoc reduction of AB; or the answer includes a combination of some or all of these. It is worth noting that a recent in vitro study published in PNAS (Sun et al) did not support the traditional contention that longer, more aggregation-prone variants of beta-amyloid (e.g. AB42) would be associated with more severe, earlier-onset forms of Alzheimer’s–no correlation was found. While this does not disprove the amyloid hypothesis, it adds to the evidence stacked against AH1.0 or AH2.0.
AH 3.0 proposes that amyloid is a secondary rather than primary pathophysiological feature; it is not the foremost pathological actor, but plays some facilitative or ancillary role. This is consistent with the recent hypothesis that amyloid sets the stage for tau to disperse and cause the bulk of the direct damage. AH 3.0 seems to be best current model for understanding the data up to this point, one that leaves open the possibility that targeting amyloid could have a clinically meaningful role in AD treatment–-but by no means guarantees it, and makes it unlikely to be the optimal target for a monotherapy.
But wait, there’s more. These are lower profile, partly because of the potential for terminal embarrassment for the leading lights of neuropharm if it turns out that either one is true.
AH 4.0 says that amyloid is benign. It could be a warning flag for ‘system errors’, cellular misprocessing of APP which exerts a toxic effect separate from amyloid-as-byproduct. It could reflect problems in APOE processing, which have been linked to effects on neuroinflammation and mitochondrial function that do not trace their roots back to amyloid.
AH 5.0 claims that amyloid has been framed, not only wrongly convicted, but in reality playing an erstwhile (albeit eventually futile) part in defending the brain. Rudy Tanzi has proposed that beta-amyloid plays a role as a defense against infection, encapsulating microbes in amyloid protofibrils. A University of Madrid group hypothesized that beta-amyloid is a defense against fungal infection, while Cortexyme is based on UCSF work indicting a bacterial protease as critical to Alzheimer’s in an infection motif, though in that scenario, beta-amyloid is an ancillary but damaging byproduct, more in line with AH 3.0.
None of these have been proven. AH 1.0 has been largely disproven, but the rest are all conceptual models, hypotheses to be tested.
Returning to CTAD, and the question posed to the sola panel: It was ‘wrong’ because it referred to ‘the amyloid hypothesis’, as opposed to ‘an amyloid hypothesis,’ without specifying which one. And with all due respect to Paul Aisen, who knows far more about Alzheimer’s than NIR ever will, he provided the ‘wrong answer’, regardless of which AH species he may have been referring to: Nothing was ‘confirmed’ by the sola results.
So much has been invested by so many in the amyloid hypotheses that any results that do not contradict it/them is seized upon like a drowning man grabbing hold of a flotation ring. There are outstanding scientists who have devoted their professional lives to the explication of one or more Amyloid Hypotheses, and it is human nature to seek corroboration, particularly when there is no viable reverse gear by which they can switch belief systems.
Having commented upon the confluence of ‘wrong questions’ and ‘wrong answers’ as pertains to the Amyloid Hypotheses, we have our own list of questions and answers about the state of the mAb art and what this means, some of which are guaranteed to be just as ‘wrong’ as those discussed above. Being wrong is an inevitable side effect of scientific inquiry, where it becomes problematic is when a belief system takes on a life of its own, despite evidence to the contrary.
At the CTAD presentation of the detailed results from the most recent sola trial, Lilly‘s Eric Siemers several times responded to questions with the honest and accurate “I really don’t know.” For most of the questions oft-asked, no one knows the answer, yet. But based on the hodgepodge of information gleaned from the several AB mAb trials that have reported results, this is how it looks.
1) Solanezumab reached statistical significance in its impact on four out of six outcome variables. What future does it have, if any?
The drug group showed a consistent pattern of slowed decline on the six measures, ranging from 7% on the FAQ (p=.14) to 15% on the ADCS-ADL (p=.009) and on the CDR-SB (p=.004). On the primary endpoint, the ADAS-cog14, the drug group declined 11% less than the placebo group (p=.095). Based on the selection of the ADAS-cog14 as primary, the trial failed, this would at best be considered a trend. Recalling that Lilly had originally cited co-primary endpoints, including a functional endpoint combining the ADCS-iADL and FAQ, eyeballing the p values for those two tests suggests that combining them would not have reached statistical significance, and would not have changed the verdict on the trial. It should also be kept in mind that the p values reported were not corrected for their multiplicity, they would have been higher had they been corrected.
But that really does not matter–Lilly acknowledged that the magnitude of observed therapeutic benefit, generally in the 11-15% range, is not clinically meaningful, and thus does not warrant continued development for sola. The fact that the primary endpoint missed hitting p=.05 saved Lilly from having to decide whether to file for a drug that they knew achieved statistical but not clinical significance in its effect, a debate that would have been roiled by commercial needs as much as clinical and ethical considerations. The same question might have been begged by the p=.004 value reported for the CDR-SB, a functional endpoint being utilized as the primary by Biogen in its aducanumab Phase III program. At the time that the EXPEDITION3 trial was being designed, the CDR-SB was not viewed as a primary endpoint candidate, but even if it had been, a 15% slowing of decline is not clinically meaningful. Lilly should be commended for its current clarity on this point. There are companies that would have considered filing based on the overall preponderance of evidence, from all solanezumab trials, that the drug does have a statistically ‘visible’ effect on disease-progression. But Lilly had expected, with the requirement of amyloid-positive biomarkers for inclusion, that this ‘purer’ patient population would reveal a treatment effect of around 30%, close to what they reported from the amyloid-positive dataset pooled from the first two trials. Falling so far short of that begs questions of what might have happened had a higher dose been used (the drug is very well tolerated), or if they had somehow identified an earlier-stage population (hard to do when requiring amyloid biomarkers for inclusion) for intervention, but also suggests that tweaking dose and population is unlikely to achieve clinically meaningful results. Lilly did hedge on whether they would consider going ahead with sola based on such tweaks. Lilly’s announcement the day after CTAD that they were partnering with AstraZeneca on another AB1-42 antibody program may be tangible proof that they have moved on from sola–although that particular choice seems like switching from Coke to Pepsi for health reasons, unlikely to make any difference at all.
2) What does the sola outcome say about Biogen and aducanumab?
Not much, though it may reassure Biogen that CDR-SB is a more sensitive endpoint and perhaps better for their Phase III program. The Phase IIb data for aducanumab did suggest a possible signal, but the small size of the study makes this tentative, as does the very shaky dose-response relationship. The 6mg dose group performed more poorly than the 3 or 10mg groups on the CDR-SB, and matched placebo on the MMSE. It is not reassuring that switching the placebo group to active treatment with the 3mg or 6mg dose at 12 months did not seem to impact the slope of deterioration at all, while the slope of the 10mg group was inconsistent. The aducanumab data also does not support AH 1.0: The two highest dose levels reduced plaque deposition rates by 25-30%, but while the 10mg dose showed the strongest association with cognitive benefit, the 6mg dose performed poorly, in spite of its PET scan results. Reducing plaque is meaningless save as a marker of specific target-engagement. Aducanumab shows a trend suggesting the highest dose has an effect on cognitive decline, but the dose-response relationship is nonlinear, and the inconsistency means that the small size of the dataset precludes taking anything as having been demonstrated efficacy-wise.
Aducanumab Sings an ARIA-again: These less-than-exemplary efficacy results came at the price of a substantial incidence of vasogenic edema/ARIA, a safety issue not encountered by solanezumab. At CTAD, Biogen reported results from a 23pt titration trial of aducanumab, wherein APOE4 patients, who are most at risk of ARIA, showed a reduction of ARIA frequency when the dose of aducanumab was slowly titrated up to 10mg: Titration reduced the incidence of ARIA by 36%, from the 55% seen in the fixed dose 10mg group, down to 35%. But all of those ARIA cases developed on the 3 or 6mg dose-levels; thus avoiding the 10mg dose in Phase III, where 6mg is the highest dose being given APOE4 patients, is unlikely to avoid ARIA. Biogen accentuated the fact that in the extended treatment phase of the regular PhIIb trial, no patients developed ARIA during their second year of drug treatment. But 25% of the group who switched from placebo to 3 or 6mg did develop ARIA. Thus while it is mildly reassuring that those who do not develop ARIA relatively early, during their first year of treatment, are unlikely to develop it later, this does not offset the significant incidence of ARIA, even with titration, even at lower doses.
In the original Phase IIb, 22% of ARIA cases were severe, some requiring hospitalization, which would justify a back-of the-envelope projection that about 6% of the APOE4 population (which is roughly two-thirds of the AD population overall) would develop severe ARIA. This is not an inconsequential safety problem. APOE4 genotyping can be easily done, but administering MRI scans on a routine and repeated basis to hundreds of thousands, even millions, of AD patients would constitute a major logistical and financial headache, but this scale of incidence cannot be ignored. Biogen does have experience with a safety issue with Tysabri, where PML is of much lower frequency, albeit much higher lethality, and identifying a predictor of PML vulnerability became a high priority. Biogen knows better than anyone that they have two tasks: Not only showing that there is a consistent benefit afforded by aducanumab that is clinically meaningful, but also to find some biomarker identifying at-risk patients who would need early and repeated imaging.
3) What do the sola/adu results say about the amyloid hypotheses?
Nothing definitive. AB mAbs are not a homogenous group, they target different epitopes on beta-amyloid and hence bind very different beta-amyloid species. Solanezumab only binds AB monomers; aducanumab does not bind monomers, but instead binds to aggregates (plaque, fibrils, oligomers). There are those who think that the monomer binding approach is too inefficient, sopping up large pools of AB monomers using an antibody that has very limited (0.1%) brain penetrance is unwieldy at best.
The consistent but therapeutically impotent response produced by solanezumab raises as many questions as answers. If the antibody binds to a suboptimal form of AB, what is optimal? Setting aside plaque as a failed candidate, one could consider targeting oligomers only (as Acumen does with their antibody) to see if that is a more focused and efficient target. If not enough antibody reached the brain to bind enough AB, that does not infer that Lilly should run a high-dose sola trial, the magnitude of benefit strikes us as simply to small to be correctable via dosing. But with just .1% of the antibody achieving brain penetrance, one could seek technologies that enhance brain availability, a theme Roche and Genentech have been enthusiastically pursuing. If the patients were too advanced to benefit from a post hoc reduction of AB, one would have to find a valid and reliable marker of pre-Alzheimer’s (and pre-plaque) MCI to zero in on during the testing. Since using PET and CSF levels did not produce the kind of differentiation hoped for, this throws into question the criteria via which many sponsors now seek to achieve a more homogenous patient population.
4) Next steps?
With all the resources allocated to bapineuzumab, solanezumab, gantenerumab, crenezumab, and aducanumab, we still do not know if the amyloid theory that spawned them offers any hint of a blueprint for clinical therapeutics–or not. Lilly’s effort to now be rational about sola’s prospects stands out for its rarity. Hopefully, the rationality with which Lilly stopped sola’s development will not become diluted and diverted into some exploratory study using higher doses in MCI patients.
Whether other antibody programs might thread the therapeutic efficacy needle more effectively than the first three remains to be seen. There are several still on deck in various stages of development: Roche/Genentech have gantenerumab and crenezumab; Lilly has the new collaboration with Astra Zeneca; Acumen has its oligomer-specific antibody; Biogen/BioArctic, Sanofi have others. It is worth noting that Roche/Genentech are running their trial of crenezumab dosing at 60mg/kg–at the very least, this represents an imposing manufacturing challenge, since a dose might average 4.2 grams of antibody. We propose that the Millennials now spawning craft breweries by the bushel be recruited to manufacture artisanal antibodies in small scale.
The eyes of the amyloid world will now turn to Merck‘s BACEi Phase III program; verubecestat will report its first results next year. BACEi in theory offers the prospect of turning off the source of beta-amyloid rather than retrospectively cleaning it up, and there are several other BACEi programs also out there (e.g. Biogen/Eisai, JNJ, Lilly/AZ, Novartis). Whether BACE is the optimal vehicle for doing so has not been proven, and Probiodrug, Lilly, and American Life Sciences Pharma have programs targeting alternative beta-amyloid pathways. Then there are off-the-beaten track strategies that tap into amyloid pathology in divergent ways; Proclara has its cross-motif inhibitor in the clinic, Cortexyme has a lead compound chosen for development, to name but two.
We are not going to review the various alternative mechanistic frameworks available to AD therapeutic programs–we utilized 60 pages to do so in the September/October 2016 issue, and will doubtlessly reach the same volumetric marker of verbosity in next fall’s edition. The field largely, albeit not unanimously, wedded amyloid more than twenty years ago, based on in vitro findings showing beta-amyloid’s effect on cultured neurons. The use of in vitro and animal models to predict the course of the human disease has been an unmitigated disaster, so we are not going to repeat that error with tau. Tau is very much on the ascendancy as an Alzheimer’s target, and there has been work that postulates a tau-centric model for AD that better approximates the human disease than do amyloid-centric models, but it still has to earn its stripes in clinical testing. The same can be said for the pursuit of neuroinflammation as a key pathophysiological factor.
5) The Bottom-Line
For a target that was the consensus choice from 1995-2010, and had tens of billions of dollars devoted to it, amyloid has thus far offered far more disappointment than utility, regardless of how much researchers say that they have learned from the process. The four mAbs with significant clinical data thus far in hand; Bapineuzumab, solanezumab, aducanumab, and gantenerumab, must have totalled over $5 billion in development costs. It seems fair to say that no one has gotten their money’s worth.