SAGE Therapeutics Inc (SAGE) – Debunking Asclepia Capital’s “Scientific Response” by Kerrisdale Capital
We are short shares of SAGE Therapeutics Inc. Please click here to read full disclosures.
At the end of October, the value investor Mohnish Pabrai gave a presentation and took part in a Q&A session at Boston College and Harvard Business School on the Uber Cannibal Investor Framework, which he has developed over the past decade. Uber Cannibals are the businesses “eating themselves by buying back their stock,” the value Read More
In the wake of our report on Sage Therapeutics, Asclepia Capital, a biotech-focused hedge fund, released a seven-page response purporting to “debunk” what it characterized as our “superficial analysis.” While we always appreciate substantive criticism, we disagree with all of Asclepia’s key points, which we address one by one below.
It’s also worth noting what Asclepia does not discuss. It offers no explanation for why SAGE-547 would outperform placebo in Phase 3 despite the weakness of the drug’s mechanism of action, and it doesn’t even mention, let alone refute, our argument that the SRSE market may be more than 6x smaller than SAGE Therapeutics suggests.
SAGE Therapeutics Inc (SAGE) – Debunking Asclepia Capital’s “Scientific Response”
Asclepia: SAGE-547 is more than just a “Band-Aid” as the reduction in excitatory brain activity is exactly what is needed for patients with epileptic disorders.
First, status epilepticus is not an epileptic disorder. Epileptic disorders are chronic diseases with the potential to cause seizures throughout patients’ lives. Status epilepticus, by contrast, is an acute symptom of a wide variety of root causes (from tumors to infections to strokes), giving rise to prolonged seizures in patients who often have no history of such problems. It thus stands to reason that patients with epileptic disorders would have different needs that patients with status epilepticus.
Second, the reason we refer to SAGE-547 as a Band-Aid is because its effects are temporary and do nothing to address the root cause of the underlying condition. SAGE-547 temporarily reduces brain activity by modulating GABA receptors, but by the time patients reach super-refractory status epilepticus, they are already on multiple other treatments that inhibit brain activity. The average patient in SAGE Therapeutics’s Phase 1/2 trial was already on 12.9 medications; why would one more with a mechanism already covered by other therapies make a difference?
Again, super-refractory status epilepticus is a symptom, not a disease, just as coughing is a symptom of the common cold, along with many other ailments. Treating a patient with a cough suppressant will temporarily relieve the cough, but, unless the underlying problem is resolved, once the cough suppressant wears off, the coughs will come back. Giving a patient multiple cough suppressants won’t confer a long-term health benefit after the patient stops taking them, and it probably won’t even provide any additional benefit relative to using a single cough suppressant.
SAGE Therapeutics’s drug is conceptually similar. It acts as a temporary suppressant and achieves, at best, a small incremental reduction in brain activity – like many other treatments that doctors already use with SRSE patients. Once the drug is removed, there is no reason to expect a meaningful difference in outcomes between patients on SAGE-547 and patients on placebo.
Asclepia: The use of SAGE-547 is NOT intended to address the underlying etiology of SRSE, but rather to help halt the seizure and wean the patient off of the anesthetic. It is just as imperative to halt the seizure and stop the administration of an anesthetic as it is to address the underlying etiology. The longer a patient is in SE/RSE/SRSE the greater the probability that he/she will experience permanent neurological damage or death.
Temporarily controlling seizures is not the problem; the problem is ensuring that seizure control doesn’t require anesthesia. We disagree that it is “just as imperative to halt the seizure and stop the administration of an anesthetic as it is to address the underlying etiology” because, for SRSE patients, general anesthesia has nearly always already halted the seizures, and, at any rate, there is no evidence that SAGE-547 can halt seizures that anesthetics can’t. (Besides, patients whose seizures continued despite anesthesia would likely be excluded from SAGE Therapeutics’s Phase 3 trial anyway.) The primary endpoint of the Phase 3 trial is weaning patients off of anesthetics and SAGE-547 or placebo, not halting seizures.
But once the seizures have stopped, there is no good reason to think that further suppression will help. Patients do not suddenly emerge from status epilepticus just because they’re switched from an anesthetic to SAGE-547; to the contrary, they need to be weaned off of SAGE-547 as well, at which point the etiology, if unresolved, will re-assert itself, the same as with anesthetics.
Asclepia: Although SAGE-547 is a positive modulator of the GABAA receptor, such as many of the other GABAergic agents, the mechanism of action and the binding sites are in fact NOT identical to that of standard drugs used for SE/RSE/SRSE. Additionally, the recent criticism states, “a body of scientific research shows that other anesthetics such as midazolam and propofol” also work on both synaptic and extra-synaptic GABAA receptors. While this may be correct, one needs to take into account the “flavor” of GABAA receptor present at each of these sites, the local concentration of each receptor type present, and their properties.
This goes back to our point that SAGE-547 offers patients no incremental value; its mechanism of action is already covered by other drugs that most patients are already on. Asclepia admits as much but focuses on minor, nuanced differences between the types of extrasynaptic receptors affected by various drugs. However, if these differences mattered to clinical outcomes, why do the electrographic data fail to reflect that? As discussed on p. 11 of our original report (the key graph is reproduced below), changes in patients’ suppression ratios have little to no correlation with the concentration of SAGE-547 in patients’ plasma, especially in the range of concentrations that SAGE Therapeutics is actually targeting, where the effect is close to zero. The proof is in the pudding; if SAGE-547’s impact on ?-subunit–containing receptors is truly so different from, say, propofol’s, then why doesn’t it show up clearly on an EEG?
Kerrisdale Re-Presentation of Sage’s EEG Data
Source: Belfort et al., “Adaptation of a Pre-Clinical Biomarker for SAGE-547 for Use in a Phase 1/2 Clinical Trial for Super Refractory Status Epilepticus,” presented at the 69th annual meeting of the American Epilepsy Society, December 4-8, 2015; Kerrisdale analysis
Asclepia: SAGE Therapeutics never purported that SAGE-547 could “rewire axons”; they only claim that it can help stop and prevent seizures from coming back during the period of weaning the patients off of anesthetics.
Asclepia is referring to the following passage in our original report:
No one believes that general anesthesia is itself a cure. SAGE-547 is no different – another way to temporarily suppress brain activity and give patients a chance to heal. It can’t rewire axons.
Our intention was to communicate that SAGE-547 has no permanent positive effect on patients. We made this statement because many investors seem to believe that it does have such an effect. But how can the drug prevent seizures from coming back after a patient has been completely weaned off of it? We’ve yet to hear a convincing or, frankly, even an un-convincing argument that bridges the gap between, on the one hand, SAGE-547’s modest effects at the level of modulating receptors and affecting brain activity and, on the other hand, SAGE-547’s purportedly very high response rate. In our view, the small magnitude of the first is totally out of proportion with the second, which makes us very skeptical.
Asclepia: Many drugs, not limited to but including benzodiazepines, have the potential of becoming tolerant; this feature is hardly unique to SAGE-547. Tolerance is dependent on the dosage taken and made bioavailable, as well as the duration of dosing. At the proper exposure levels and duration of exposure to SAGE-547, tolerance will be a nonissue.
This is likely in response to our noting that investigators showed in a 2001 rat study that exposure to allopregnanolone led to increased seizure rates, which we think is a strong knock against the possibility of SAGE Therapeutics ever using allopregnanolone for any chronic indications. (To be clear, this is an argument about the value of SAGE Therapeutics’s preclinical pipeline, not about the use of SAGE-547 in SRSE, where the severity of the condition makes concerns about tolerance and withdrawal less important.)
Is it really true, as Asclepia asserts, that tolerance is necessarily a nonissue at “proper exposure levels and duration of exposure”? What law of nature guarantees that, for any given drug (or is there somehow a special rule for SAGE Therapeutics alone?), an effective dose is also a safe dose?
In reality, there’s no justification for simply assuming that safe exposure levels will necessarily suffice for the chronic treatment of any condition. The 2001 study provides strong evidence that the proper chronic exposure level for exogenous allopregnanolone is zero unless SAGE Therapeutics comes up with an innovative way to chemically alter allopregnanolone and proves that this new chemical entity is safe and works better than the similar, already time-tested benzodiazepines. Given the risks posed by allopregnanolone, it’s fair to ascribe SAGE Therapeutics’s pipeline very little value.
Asclepia: The critics inappropriately concluded from an article published in a scientific journal that benzodiazepines and neurosteroids have the same pharmacological and side effect profiles. It requires considerable emphasis that these results are conclusions for mice treated a particular way to induce seizure that cannot translate over to humans in SE. A close read of the study and the methodology used unquestionably eliminates the study for use as a comparison between the safety and efficacy of midazolam and SAGE-547 in human SRSE. The study cited induces seizures in mice using picrotoxin, which binds to GABA receptors and behaves as a negative allosteric modulator. The experimental animal model of epilepsy that uses picrotoxin to induce seizure is classified as a “simple partial, acute” model of epilepsy. Generalized seizure animal models, which include “generalized tonic-clonic” and “generalized status epilepticus”, are extremely different. Again, the experimental model of seizures that compared the efficacy of allopregnanolone (SAGE-547) and midazolam differs greatly from seizures seen in humans in SRSE; thus, the data cited by the critic is inappropriate for comparison and not translatable to humans in SRSE.
First, all of this emphasis on a single reference (Cz?onkowska et al. 2001) is puzzling and misplaced. In addition to this particular study, we also cite, among other sources, a 17-page review paper from 2011 entitled “Tolerance to allopregnanolone with focus on the GABA-A receptor.” That paper in turn draws on dozens upon dozens of other papers involving data from both animals and humans collected over many years by many different researchers. It’s simply not the case, as Asclepia appears to imply, that our concerns about the risks of chronic allopregnanolone use stem from just one animal study.
Second, Asclepia seems to misinterpret our use of a quotation from the Cz?onkowska et al. 2001 paper. The main point of the quotation, which appears amid a discussion of the history of allopregnanolone in medical research (not an analysis of allopregnanolone in SRSE specifically), is that informed researchers harbored serious doubts about the usefulness of the compound many years ago. That doesn’t by itself rule out the possibility that allopregnanolone will end up being valuable in SRSE or other indications (which we disbelieve for many other reasons), but it’s certainly useful background information for anyone trying to estimate the odds of clinical success. Whether the picrotoxin model is a good match for SRSE in humans is irrelevant to the point we’re trying to make, and we never made a claim about it one way or another.
Asclepia: The criticism failed to draw attention to the positive properties of SAGE-547 such as its higher potency than many of the anesthetics used as a 3rd line treatment. SAGE-547 is active at nanomolar (nM) concentrations whereas pentobarbital and propofol and other commonly used drugs require micromolar (?M) to millimolar (mM) concentrations to reach similar efficacy thus making it much more potent.
There’s no point in discussing dosing if SAGE-547 doesn’t benefit patients. The fact that SAGE-547 might have an effect at nanomolar rather than micromolar concentrations makes no difference to the question of whether it will outperform placebo in the Phase 3 trial.
Asclepia: Any attempt to compare data from the SAGE-547 Phase 1/2 study with those of an ongoing prospective study with only preliminary results published will be met with serious difficulties due to a lack of consistency in protocol, definitions, and other important variables such as quality of health management, availability, quality, and condition of medical equipment used, and experience in dealing with SE/RSE/SRSE. For example, the protocol for initially treating SE upon diagnosis was not consistent: Only 33% of the cases used a 1st line drug treatment while 61% used 2nd line drugs and 6% used a 3rd line drug at the start of treatment. Also, only 56% of the cases used two or more 2nd line drugs while 479 patients were reported to receive an anesthetic. Since the treatment protocol regimen was not consistent, the 74% recovery rate cited cannot definitively conclude which patients (SE or RSE or SRSE) on what drug therapy were part of that statistic.
Furthermore, approximately 350 patients received a second anesthetic at unknown variable time points and kept on the anesthetic for an unknown duration of time (lack of consistency, reporting, and protocol); again, the status of these patients (SE or RSE or SRSE) is not clear. The study also classified 413 patients into 3 categories: (i) recovered from SE (74%) (ii) patient died (22%) (iii) patient withdrew from therapy (4%). The category referring to ‘recovered from SE’ is defined as seizure control only and does not include the time frame the patient remained on the anesthetic, or whether they were SE or RSE or SRSE, nor the neurological outcomes.
There is no fixed protocol for pre-SRSE treatments in SAGE Therapeutics’s trial either. If anything, the protocol and exclusion criteria in SAGE Therapeutics’s trial should have enhanced patient outcomes even if the drug were ineffective. As for the point about the status of the patients, we know that the average ICU stay was 18 days for patients in the study that Asclepia is referring to, with 69% of patients staying in the ICU for more than 7 days, so it is clear that most of the patients did have SRSE, as we discussed in our original report (p. 14).
Asclepia: As has been made evident, the analysis of clinical data from the referenced study used to compare results from the SAGE-547 phase 1/2 study lacks the proper scientific rigor to be conclusive and therefore should not be considered comparable, as agreed by the authors of the study when they made the following statement:
“As a registry, it would not be possible to reach any definitive conclusion about efficacy of treatments of refractory status epilepticus from these data.”
Moreover, the authors from the 2012 published study themselves acknowledged the pitfalls of any direct comparisons to the data in their study because of the uncertainty of the quality and accuracy of the data used to create the statistical profile of each drug. They write:
“These were, as mentioned above, largely anecdotal reports and each drug cannot be compared with the others. Furthermore, a number of significant potential biases exist.”
What the authors of these studies are cautioning against in the quoted passages is using their data to determine which third-line treatment works best. But we aren’t interested in comparing midazolam and propofol, so these warnings don’t apply; instead, we’re using the studies to estimate the rate of SE resolution among SRSE patients. These are the best available benchmarks for forming expectations about the placebo group in SAGE Therapeutics’s Phase 3 trial. “Proper scientific rigor” is a worthy aspiration, but investors can’t wait for perfect, “conclusive” data; we need to make the best judgments we can from the data that actually exist, even if they fall somewhat short of the ideal. Dismissing inconvenient information on the basis of boilerplate caveats of the sort found in almost every scientific paper is a good way to avoid ever changing one’s mind.
Asclepia: The salient misconceptions presented by the short report regarding SAGE-547 have herein been shown to be unfounded; 547 has not only clearly elucidated its unique mechanism compared to other GABAergic agents, but rather it has also thereby substantiated the phase 1/2 clinical results supporting its unique mechanistic effects.
Furthermore, the short report draws conclusions with no solid evidence to defend its claims as well as references studies that are incongruous for proper comparison, which translates to an improper assessment of the probability of success of the upcoming Phase 3 study called STATUS.
While all Phase 3 studies inherently involve a significant level of risk, assessing the likelihood of success of STATUS based on the superficial analysis from the short report poses a daring and unwise proposition. Rather, the scientific virtues of SAGE-547 described in this analysis that debunk the salient features of the short report should be considered to better weigh the risk-reward of its outcome when it reports topline data in the second half of the year.
Our report provided ample evidence to support our views, which remain the same: SAGE-547 has no unique mechanism of action and has no meaningful, lasting clinical benefits in SRSE, so there is no reason to think it will do any better than placebo at enabling SRSE patients to be weaned off of anesthetics.