SAGE-547: Fooling some of the people all of the time

9 min readApr 25, 2017

“Why don’t you slip into something more comfortable… like a coma” — Jerry Lawler

Sage Therapeutics (SAGE) is a clinical stage biopharmaceutical company developing drugs for neurological conditions. It currently trades at a market cap of $2.74 billion, based primarily on the perceived value of its phase 3 drug candidate, SAGE-547, in development for super-refractory status epilepticus (SRSE) and for post-partum depression.

Status epilepticus (SE) is said to occur when seizures are prolonged, or are recurrent and the patient does not recover between seizures. Notably, SE is a symptom of an underlying disorder of the brain, and not a disease itself. The causes of SE vary widely, and may occur in patients with tumors, infections, autoimmune disease, trauma, or strokes. The first line standard-of-care for SE consists of benzodiazepines, such as diazepam and lorazepam, which are positive allosteric modulators of inhibitory GABA-A receptors and thus reduce brain activity. Although benzodiazepines lead to resolution of SE in most cases, some patients progress to established status epilepticus. These patients are typically administered anti-epileptic drugs, including valproate and phenytoin. Patients for whom these treatments do not work are classified as refractory and progress to general anesthesia, such as propofol, midazolam, or pentobarbital. Physicians will attempt to wean refractory patients off of anesthesia after 24 hours. Most of these patients emerge seizure-free, but if the seizures return, the patients are classified as super-refractory. These patients then go back on general anesthesia for 5–7 day cycles followed by additional weaning attempts, which will typically lead to resolution. In essence, the paradigm of care for SRSE is to induce a coma to allow the patient’s underlying disorder time to heal itself, which determines whether the subsequent attempt to wean from general anesthesia will be successful.

SAGE-547 is an intravenous formulation of allopregnanolone, a naturally occurring neurosteroid produced from the sex hormone progesterone. Allopregnalone, like benzodiazepines, enhances the effects of neurotransmitters that activate GABA-A. Although benzodiazepines and SAGE-547 bind to different allosteric sites on the receptor and may have varying affinities for different isoforms of the receptor, the underlying mechanisms for sedation are very similar.

SAGE asserts that their drug is distinct from benzodiazepines in that it works at the extrasynaptic GABA-A receptor. They claim that this is important because resistance to benzodiazepines — which bind to synaptic GABA-A receptors — occurs because the synaptic GABA-A receptors leave the membrane surface during protracted seizures. They further claim that extrasynaptic GABA-A receptors remain on the surface even in the setting of long-term seizures, during which the modulation of extrasynaptic receptors will effectively calm the brain down[1].

The issue with this line of reasoning is that SAGE-547 is not unique in its ability to target extrasynaptic GABA-A receptors. Indeed, all of the general anesthetics commonly used in SRSE, including propofol[2], midazolam[3],[4], and barbiturates[5], are active at extrasynaptic GABA-A receptors[6]. If targeting extrasynaptic receptors is truly the key to resolving SRSE, it highly unlikely that SAGE-547 will have a unique effect given that these SRSE patients are already receiving general anesthetics.

In light of their drug’s undifferentiated MOA, why is the market ascribing so much value to SAGE? Bulls point to the results of their Phase 1/2 trial. In this open-label, single-arm study, 25 patients were administered both SAGE-547 and general anesthetic for a period of five days. The primary efficacy endpoint was to determine the proportion of patients that could be successfully weaned from general anesthetic and from SAGE-547 and emerge from coma seizure-free. Of 22 evaluable patients, 17 were weaned off of general anesthetic, and the same number were weaned off of both anesthetic and SAGE-547 without recurrence of SE in a 24-hour follow-up. SAGE bulls applauded this apparent 77% response rate — in spite of the fact that the trial did not include a placebo comparator — by declaring that the expected response rate from a placebo arm would have been around 35% based on the published literature.

It is noteworthy that this purported 77% response rate corresponds to a denominator of 22 evaluable patients, not the 25 total patients enrolled in the trial. In particular, SAGE’s definition of “evaluable” excludes patients for whom treatment was disrupted or no attempt was made to wean from general anesthesia[7]. In other words, this would comprise patients whose condition was so poor that physicians would not even risk taking them out of a coma. If the response rate is instead computed on an intent-to-treat basis (the gold standard for reporting clinical trial results), we arrive at a response rate of 68% (17/25). Even among the 17 responders, four experienced recurrence of SE in the four weeks following treatment, implying that SAGE-547 did not truly resolve the condition, so a more conservative definition of “response” would have further depressed this response rate.

Separately, a closer inspection of the 35% response rate parroted by management and sell-side analysts uncovers that it is in reference to a single study of 24 patients[8]. In this trial, refractory SE patients were administered propofol or barbiturates for 36–48 hours and subsequently weaned, with a primary endpoint of seizure control after the first course of drug. Although the overall rate of seizure control from the first wean was 35% (8/23), this dataset does not provide an apples-to-apples comparison of the results from SAGE’s Phase 1/2 trial. For instance, patients who did not achieve burst suppression on the pre-selected doses of anesthetic were counted as failures, whereas such patients were excluded from SAGE’s Phase 1/2 trial altogether[9]. Another patient was classified as a failure because propofol infusion syndrome led to discontinuation of drug; this patient would similarly have been excluded from SAGE’s trial. Removing these seven patients from the denominator leads to a response rate of 50% (8/16) from general anesthesia.

In the above trial, patients were also classified as failures if there was a need to re-titrate the study drug during the first seven days after initial wean (such patients would then be considered super-refractory). Physicians were then able to choose the subsequent course of treatment, for which seizure control was achieved in 4 out of 8 patients in the propofol group and 5 out of 7 patients in the barbiturate group. Thus, the overall response rate for SRSE in this trial was 60% (9/15). The difference between this 60% response rate and SAGE’s 68% response rate based on intent-to-treat chalks up to a mere 1–2 patients, which can reasonably be attributed to chance.

While citing this small study as the source for a 35% response rate is clearly off the mark, a more thorough analysis of the literature confirms a typical response rate for standard-of-care that is much higher than 35%. A comprehensive review and meta-analysis from 2012 determined that control of refractory and super-refractory status epilepticus was achieved with propofol, midazolam, or barbituates in 74% (678/920) of cases[10]. Although this includes both refractory and super-refractory SE cases, the duration of anesthesia suggests that SRSE patients comprised the great majority of those administered propofol or barbituates, for which the combined response rate was approximately 66%.

Another large, ongoing study — aiming to create a registry of refractory and super-refractory SE treatment cases — demonstrated that seizure control is achieved in 74% (304/413) of cases[11]. In this study, 69% of patients were in the ICU for over 7 days, implying that the majority of patients were, in fact, super-refractory. Thus, the totality of the evidence available points to a response rate in SRSE materially higher than 35%.

Clearly, a more calibrated view of the clinical data suggests that SAGE-547 provides minimal benefit relative to standard-of-care. The physiological data from the trial, where SAGE measured the change in suppression ratio before and after treatment as a biomarker for neurological hyperactivity, lends further support to this view[12]. Although SAGE presents the dose-response data as evidence of an increase in suppression ratio with increasing concentration of drug, this interpretation of the data is flawed. A significant number of data points indicate no change in suppression ratio, even at higher concentrations of drug. Several data points even show a negative change, which would entail an increase in brain activity. Furthermore, without the three outliers in the 400–450 nM range — well above the target dose range of 200–300 nM — the data would clearly show no correlation between drug concentration and suppression ratio. In light of this lack of dose-response, and with almost as many data points exhibiting zero or negative change in suppression ratio as positive, the biomarker results present no evidence that SAGE-547 works in these patients.

SAGE’s Phase 1/2 EEG biomarker data — a higher suppression ratio implies less brain activity as measured by EEG, so a positive change in suppression ratio over the course of the trial would suggest that SAGE-547 is working.

The weight of the evidence suggests that SAGE-547 is, at best, another lemming among the host of GABAA agonists already being used to treat SE. The notion that SAGE-547 is differentiated in its ability to treat or cure SRSE contradicts what we know of the SRSE treatment paradigm. General anesthesia does not actually repair the network dysfunctions that underlie seizures — which can have vastly different etiologies — but only stabilizes patients within a coma in an attempt to allow the dysfunction to resolve itself. Since there is no evidence SAGE-547 does anything different from general anesthesia to affect the underlying biology, it will likely fail to show any benefit in the larger, double-blind, placebo-controlled phase 3 trial that will read out in Q2 of this year.

In this trial, 140 SRSE patients were randomized 1:1 to received standard-of-care anesthetics in combination with either SAGE-547 or placebo over six days, with a primary endpoint of resolution of SE after weaning without resumption of seizures in the 24-hour follow-up. I expect that the placebo arm will exhibit a response rate that is in line with historical data (between 50–70%), and the SAGE-547 arm will come out about the same, missing the threshold for a statistically significant benefit.

Interestingly, the Phase 3 trial also includes an open-label protocol, whereby patients that fail blinded treatment are treated with a higher dose of SAGE-547. As outlined above, SE can resolve spontaneously with standard-of-care at a high frequency, especially as the duration of treatment increases (giving the underlying condition more time to improve). Even in SAGE’s Phase 1/2 trial, 2 of the 5 “non-responder” patients eventually recovered — an outcome that cannot in hindsight be attributed to SAGE-547 due to its short, one-hour half-life.

Continuing wean attempts from general anesthesia will lead to the eventual recovery of a substantial fraction of these refractory patients. A cynical interpretation of this open-label study design is that, by concurrently administering these patients SAGE-547, SAGE will aim to replicate the superficially positive Phase 1/2 results in the context of this pivotal trial. Thus, even if the trial misses its primary endpoint, SAGE can argue for the approvability of SAGE-547 based on the unblinded results and the severity of the unmet need in SRSE. I believe that the FDA would see through this argument and abide by the results of the randomized, controlled protocol.

Given the extent of downside potential in SAGE stock, and the likelihood of failure in the pivotal SRSE Phase 3, I have initiated a short position in SAGE of 200 shares. Although the outcome of SAGE’s ongoing Phase 3 trial of SAGE-547 in post-partum depression is less predictable, the outcome of the SRSE trial should be the company’s next meaningful catalyst and represents the lion’s share of sell-side valuations, where a near-certain probability of success appears to be priced in.

Disclaimer: All content in this blog post is my personal opinion. I am not receiving compensation for this post, and have no business relationship with any company whose stock is mentioned herein. All information in this post is obtained from sources I believe to be reliable and accurate, but I make no express or implied warranty of such. I have a short position in SAGE stock and stand to realize gains in the event of a decrease in the price of the stock.