GW Pharmaceuticals — Update 24 February 2016

Company description: Cannabinoid medicines specialist

Founded in 1998, GW Pharmaceuticals is a UK-based biopharmaceutical company that discovers, develops and commercializes proprietary cannabinoid (cannabis-derived) medicines for a broad range of diseases. GW’s cannabinoid platform generated the world’s first plant-derived cannabinoid therapeutic, Sativex, for the treatment of spasticity due to multiple sclerosis (MS), sold by multiple global partners. GW is also developing a broad pipeline of cannabinoid medicines targeting diabetes and central nervous system (CNS) disorders including epilepsy, schizophrenia, autism and neonatal hypoxic-ischemic encephalopathy (NHIE). The company currently has four Phase III trials ongoing in Dravet syndrome and LGS, particularly severe forms of epilepsy. GW listed on AIM in 2001 and on the NASDAQ exchange in 2013.

Our DCF-based valuation is now £1.35bn or 517p/share (vs £1.21bn, 510p/share) after increasing the probability of success for Epidiolex in Dravet and LGS to 70% from 50% following positive data, increasing its probability of success in schizophrenia to 25% from 20% after proof-of-concept data and adding TSC and IV GWP42003 for NHIE into our model. This was mitigated by the removal of estimates for Sativex for cancer pain and GWP42003 for ulcerative colitis due to the failure of those programmes and a higher share count. We also note that our valuation represents fair value for the stock today, ahead of Phase III results with Epidiolex. For illustration, increasing the probability of success to 90% on positive outcomes from these trials would raise our overall valuation to approximately £1.73bn or 663p/share. Epidiolex also holds wider potential in other refractory forms of childhood epilepsy, which we do not currently capture in our model.

GW Pharma is subject to the sensitivities common to most biopharmaceutical companies, such as potential clinical or regulatory failure or delay, commercialization risks (launch, uptake, pricing, reimbursement, competition) and reliance on partners. With investor focus on the four Phase III trials in Dravet syndrome and LGS, any hiccups in relation to these trials could have a significant impact on the valuation of the company (especially given they all revolve around the same product, which represents the bulk of GW’s value). Based on a sensitivity analysis, it appears that investors are currently pricing in a ~30% chance of success for those trials, so there would be downside if they fail. However, given the size of the trials, the expanded access data and the historic placebo responses in other trials, we believe they are more likely to succeed than fail.

The company announced that it finished Q116 with £219.3m ($324.1m) in cash. GW-funded research and development was £21.9m for the quarter, up £0.5m sequentially, but up £13.2 compared to the fourth calendar quarter of 2014. We have increased our R&D expense due to larger trial sizes and removed Otsuka-related R&D fees and expenses from future estimates (although the NPV impact is zero on the Otsuka changes). GW also announced a slight delay in receiving the results of its second Dravet trial to H216 from mid-2016, although the NDA filing is still expected to occur in Q416. While this change is minor, we had originally expected all results to be released in H116, so we do not now expect Epidiolex to launch in 2017 and have removed those revenues from our model. We currently estimate profitability in 2018 with a total cash burn of around £160m from 2016-17, providing GW with ample cash flow to achieve its goals without the need for additional capital. However, this profitability is dependent on successful Epidiolex trials in Dravet and LGS, as well as approval.

GW has developed a broad pipeline, which is focused on areas of continued unmet medical need despite a number of approved therapies. Epilepsy patients have quite a number of choices for treatment, but 36% have pharmacoresistant epilepsy1 that cannot be controlled by even three or more drugs taken concurrently. GW is focused on some of the most difficult-to-treat subpopulations in Dravet and LGS. It has recently expanded this to Tuberous Sclerosis Complex (TSC), another subtype of epilepsy, and is expanding to other CNS areas such as schizophrenia, where it recently had proof-of-concept data, and NHIE, a severe issue for newborns that often leads to disease or lifelong brain damage.

The development of Epidiolex, a liquid formulation of highly purified CBD extract, has been a rapid and clear example of the inherent potential in GW’s cannabinoid technology platform. This has resulted in the investment case advancing well beyond Sativex, for many years the only significant valuation driver for the stock. In 2014, GW received Orphan Drug and Fast Track designations from the FDA for Epidiolex in the treatment of Dravet, as well as Orphan Designation from the EMA in Europe. Also in 2014, GW received Orphan Drug designation from the FDA for Epidiolex to treat LGS. These designations, coupled with highly encouraging data from an expanded access programme in the US (313 children treated across 16 clinical sites), have led to the design of an extensive pivotal clinical study programme for Epidiolex in DS and LGS, with data expected from four Phase III trials in the next six months.

Dravet is an extremely malignant form of childhood epilepsy that typically presents itself within the first year of life with prolonged febrile and afebrile, generalized clonic or hemiclonic epileptic seizures in otherwise normally developing children. The incidence of Dravet ranges from 1:20,000 to 1:40,000 births, which suggests an overall disease prevalence of 5,500 patients in the US and 6,700 European patients.2 The primary genetic cause is a mutation of the SCN1A gene which, while helping to increase the rate of diagnosis, has not done much for the outcomes of these patients. 10-14% of Dravet patients end up dying, typically around the age of six or seven (see Exhibit 2).

Besides the risk of death, by the time the children are teenagers they exhibit either severe or profound learning disabilities (see Exhibit 3). In one study of 31 typical and borderline Dravet patients (14 were typical Dravet, 17 were borderline) who were followed until adulthood, 22.6% could speak no words at all, 29% cold speak several words, 29% could make primitive conversation and 16.1% could make simple conversation and read to some extent. Only one (3.2%) with borderline Dravet could lead an independent life, although he developed psychosis.3

Thanks to data from the company’s expanded access programme, which provided the drug to 313 patients at 16 different sites, including 40 with Dravet, we have evidence that the drug is efficacious in this subtype of epilepsy (see Exhibit 4). Patients were 12 years old on average and, importantly, were already on three other anti-epilepsy drugs on average. At week 12, Dravet patients saw a median decrease of 63% in total seizures, with 13% being seizure free, a very impressive result in this high refractory population. In terms of safety, somnolence, diarrhea, fatigue and decreased appetite were the most common, but only 4% discontinued therapy due to an adverse event. Importantly, as the sample size for Dravet patients has increased from the original 23 patients reported back in April 2015, the data have remained consistent.

There is an ancient Chinese saying that “he who lives by the crystal ball ends up eating broken glass”. This is often especially true for drug trials, as it seems a thousand things need to go right for a successful trial, while it only takes one thing to go wrong for the trial to be doomed. The design of the first Phase III trial is pretty straightforward (see Exhibit 5), with a four-week baseline observation period to understand the seizure rate in the patients before therapy. This is followed by a two-week titration phase up to 20mg/kg of Epidiolex or placebo, with each arm having 60 people (120 total).

This design is relatively similar to the expanded access study, which produced exceptional results in Dravet patients, but with one difference, which is that dosing was not a flat 20mg/kg. However, the average dose was 19mg/kg in the Dravet patients and there was no clear increase in response rate for those dosed above 20mg/kg. In the second Phase III, which will test two doses of Epidiolex (10mg/kg and 20mg/kg) versus placebo in around 150 patients, the design is almost identical except for the inclusion of the lower 10mg/kg dose.

The percentage change from baseline in convulsive seizure frequency during the treatment period (0-14 weeks) compared to placebo is the primary endpoint of the trial, so it is important to understand what sort of placebo effect we can expect. There is very limited data from previous placebo controlled trials in Dravet patients. In a 41-patient trial comparing stiripentol to placebo in patients currently on valproate and clobazam, placebo patients actually saw a 7% increase in seizures and only one patient (5%) was deemed a responder, which was defined as a 50% reduction in the frequency of clonic seizures. In a smaller 21-patient trial with stiripentol, the placebo response rate was a 12.7% decrease in seizures with, again, only one responder (9%). While these trials are small, given that Dravet is such an incredibly difficult subtype of epilepsy to treat, any placebo effect will likely be minimal, if not non-existent. According to the company, the trial has higher than 90% power to detect a 30% difference between the treatment arm and placebo. That means that even if the expanded access programme was overstating efficacy by 20-30% at week 12 (assuming little/no placebo efficacy), the trial would likely still be successful.

As an illustration of how the powering/statistics may work out, it is helpful to look at the clobazam trial in LGS patients (see Exhibit 6) since the individual treatment arms and placebo arm are roughly the same size as the arms in both the GW Dravet trials (although the clobazam trial was larger as a whole due to more dosage arms), and the trial design was generally similar. In the low-dose arm with 53 patients, clobazam demonstrated a 29.1% placebo-adjusted reduction in seizure rate, which was good enough to achieve a p-value of p=0.012-0.017, depending on the statistical tests. Higher placebo-adjusted reductions in seizure rates quickly added additional zeros to the p-value, signifying a highly statistically significant result. We expect data from the first Phase III in Dravet in March and the second in H216.

With an NDA filing expected in Q416, we expect a launch in early 2018, allowing for the FDA review, as well as DEA rescheduling (which by law should take 90 days). As patients and physicians are comfortable using a combination of drugs to treat these disorders, we would expect the addition of Epidiolex, with its novel mechanism of action, to be widely adopted (assuming the data support its use). In our model we assume 50% penetration in the US and 25% penetration in the EU. We project worldwide peak sales of around $200m in Dravet patients.

LGS, like Dravet, is another rare form of epilepsy, although it typically starts later in life, at between two and eight years of age vs six months for Dravet. As with Dravet, outcomes are extremely poor for these patients, with 90% becoming mentally handicapped with a progressive reduction in IQ. The mortality rate is high, although the exact percentage varies based on the study and ranges between 3% and 25%.4 Incidence estimates for LGS vary, but it accounts for approximately 2-5% of all childhood epilepsies. This suggests 16,000 patients with LGS in the US and 24,000 in Europe.

While the sample size from the expanded access data is more limited than with Dravet, it is still impressive and the median seizure reduction has been consistently high, with the most recent data point suggesting a 71.1% median reduction in atonic seizures (which is also the primary endpoint of the two LGS trials).

The first Phase III trial results for LGS, which will have results in Q216, has a design similar to the Dravet trial with two arms, although there will be 171 patients (approximately 85 patients each). This increases the power of the trial significantly and allows for both a lower than expected treatment effect and a higher than expected placebo effect. Historically, however, LGS placebo response rates have not been much more than 10% across trials (see Exhibit 8), so the trial should be powered for statistical significance even with only a 20% placebo-adjusted difference.

The second trial is expected to have three arms of more than 70 patients each (placebo, 10mg/kg and 20mg/kg), but the powering should still be enough to detect a treatment effect. Note that in the lamotrigine trial in LGS patients (treatment arm had 78 patients, while 89 were in placebo), a 23% placebo-adjusted difference was able to achieve a p-value of 0.002.5

As with Dravet, combination therapy is very much the norm and Epidiolex has a novel mechanism of action compared to other therapies, so we expect that it will be widely used. In our model, we assume 50% penetration in the US and 25% penetration in the EU. We project worldwide peak sales of around $600m in LGS patients.

TSC is a multisystem, autosomal dominant genetic disorder resulting from a mutation in one of two tumour suppressor genes, TSC1 (encoding hamartin) or TSC2 (tuberin). TSC is characterized by benign tumours, known as hamartomas, in various organs, most commonly the skin, brain, kidneys, heart and lungs. A hamartoma is composed of an overgrowth of mature cells and tissues, which normally occurs in the affected tissue. TSC affects both sexes and all ethnic groups, affecting as many as 25,000-40,000 individuals in the US and about one to two million individuals worldwide, with an estimated prevalence of one in 6,000 newborns (tuberous sclerosis fact sheet).

The signs and symptoms of TSC vary depending on the organs involved. The most common symptom is epilepsy, which occurs in around 80-90% of patients and is a significant cause of morbidity and mortality.6 Seizures of all types may occur, the most common being infantile spasms, partial motor seizures and generalized tonic-clonic seizures. Seizure onset occurs in the first year of life in almost two-thirds of patients, and within the first three years in around 80% of patients.7 The seizures are often severe, and up to two-thirds of TSC patients are refractory to available medical and surgical therapies.2 Developmental delays occur in around 50-60% of TSC patients, ranging from mild learning difficulties to severe mental retardation, and about one-third of children with TSC meet the criteria for autism spectrum disorder (tuberous sclerosis fact sheet). Behavioral problems are common and can be difficult to manage. The prevalence of cognitive impairment and neuropsychiatric and developmental disorders has also been found to be higher in those with refractory epilepsy.

Under an expanded access IND from the FDA, Epidiolex has been made available in licensed clinics as an adjunct treatment in children and young adults with drug-resistant epilepsy. GW presented updated data on the trial at the American Epilepsy Society’s annual meeting, showing encouraging outcomes in 10 TSC patients treated with Epidiolex under this programme. Four of these TSC patients also had cognitive impairment. Six of the patients responded to adjunctive treatment with Epidiolex, with response defined as a 50% decrease in seizure frequency at 16 weeks of treatment compared to a four-week baseline period.

Importantly, patient 9, who suffered from five separate seizure types, was on three concomitant antiepilepsy drugs and had a total of seven in the past, became seizure free after two months of treatment and has remained seizure free for 10 months. Diarrhea was the only side effect attributed directly to CBD. Drowsiness and irritability were due to drug-drug interactions in five patients, which were alleviated with antiepileptic drug dose adjustments while maintaining seizure control.

On the basis of these results, GW will start a Phase III trial in TSC in Q116. Once data are available for a larger number of TSC patients, we will see whether the seizure response and cognitive and behavioral effects are repeated. If the results suggest wider cognitive/behavioral benefits, this could provide the basis for an expanded use of Epidiolex beyond epilepsy.

In September, GW Pharma announced positive proof-of-concept data from a Phase IIa study of CBD in 88 patients suffering from schizophrenia who had failed at least one previous therapy. There was no primary endpoint, just a series of exploratory endpoints. CBD showed a statistically significant impact in the PANSS positive subscale, CGI-S and CGI-I, while showing trends in both cognition and the PANSS negative subscale (see Exhibit 9). The two most recent FDA approvals of schizophrenia medications (Vraylar and Aristada) both had PANSS total score as the primary endpoint, with CGI-S being the secondary endpoint in one and CGI-I being the secondary endpoint in the other. Given that CBD has demonstrated statistically significant results in some very key endpoints, we have increased our probability of success for the programme from 20% to 25%. We await further clarity on the complete dataset from the clinical trial, as well as the plan for the programme, before increasing the probability of success further.

NHIE is a condition that results from an interruption of blood flow and oxygen delivery to the brain due to a variety of reasons including placental insufficiency, cord compression or foetal hemorrhage (see Exhibit 10). According to Medscape, incidence is currently estimated at 2.5 per 1,000 live births, so as there are four million live births per year in the US; according to the CDC, there are around 10,000 cases of NHIE per year. Without treatment, 23-27% with moderate to severe NHIE die before leaving the hospital, with the overall death rate at 30-38% at the 18- to 22-month time point.8,9 Those that survive often have cerebral palsy (36%), epilepsy (16%), hearing (10%) and visual impairment (13%).10

GW Pharma is developing an IV formulation of GWP42003, where CBD is the primary cannabinoid but also contains other cannabinoid and non-cannabinoid components, for use in neonates. There are no clinical data yet, as the company first needs to conduct a juvenile toxicology study followed by a ~100 newborn proof-of-concept study. The preclinical data seen so far have been quite remarkable, with brain activity and necrotic cell counts close to normal in newborn piglets.11 CBD appears to be anti-inflammatory and modulates cerebral hemodynamic impairment and brain metabolic derangement, while also preventing the appearance of brain edema and seizures.

There are high costs associated with these infants, so if GW can demonstrate a large enough benefit it would be able to receive a premium price for its product. For the purposes of our valuation, we are currently assuming a price of $50,000 per treatment (which will last at least 72 hours and may last as long as a week) and a ~40% peak penetration, which results in ~$450m in worldwide peak sales.

Sativex is an oromucosal spray consisting of a formulated extract of the cannabis sativa plant, which contains the principal cannabinoids delta-9-tetrahydrocannabinol (THC) and CBD. Multiple sclerosis (MS) affects approximately 1.3 million people worldwide, of which up to 80% suffer from spasticity, a symptom of MS characterized by muscle stiffness and uncontrollable spasms. There is no cure for spasticity and Sativex provides an alternative for patients who fail to respond to conventional oral therapies (Baclofen/Zanaflex).

Sativex is approved as a treatment for MS spasticity in 28 countries (outside the US). The product is licensed to a number of partners across global territories, including Almirall (EU ex-UK/Mexico), Otsuka (US), Bayer (UK/Canada), Novartis (Australia/New Zealand/Asia/Middle East/Africa) and Ipsen (Latin America ex-Mexico). GW receives upfront fees, milestones and royalties from these collaborations. In-market 10ml vial sales volumes grew 22% in FY15 and GW recorded £4.2m in Sativex commercial-related revenues in FY15, down £0.2m from the previous year. While it is a small product providing minimal value to GW currently, it did serve as proof of concept for cannabinoids as a medicinal product.

Our DCF valuation has increased to £1.35bn or 517p/share (vs £1.21bn, 510p/share) by increasing the probability of success for Epidiolex in Dravet and Lennox-Gastaut syndromes to 70% from 50% following the positive expanded access data presentation in April and the update in December, increasing its probability of success in schizophrenia to 25% from 20% after positive proof-of-concept data and by adding TSC and IV GWP42003 for NHIE into our model (Exhibit 11). We estimate that the TSC programme could reach the market in 2018 and achieve peak global sales of $255m. For IV GWP42003 in NHIE, projected market launch is 2021, with peak sales of $448m.

We have also rebased our model to 2016, and adjusted our revenue estimates for Sativex. We include cash at end December of 2015 of approximately £219m.

Offsetting the increase is the removal of estimates for Sativex for cancer pain and GWP42003 for ulcerative colitis due to trial failures. We have also pushed back our timeline for a launch in Dravet and Lennox-Gastaut from 2017 to 2018 as the company expects to file in Q416 and will also have to wait for the Drug Enforcement Agency (DEA) to reschedule the drug.

As a reminder, GW has the most extensive portfolio of cannabinoid-based therapeutics in the industry. It is at the forefront of developing cannabinoid-based medicines that hold potential across a wide spectrum of therapeutic areas. We note that the company announced that it is currently targeting autism spectrum disorders, Duchenne Muscular Dystrophy, oncology and cachexia in its preclinical research, although we do not ascribe any tangible value to those programmes yet.

The company announced that it finished Q116 with £219.3m ($324.1m) in cash. GW-funded research and development was £21.9m for the quarter, up £0.5m sequentially, but up £13.2 compared to the fourth calendar quarter of 2014. While we do expect R&D expenses to moderate as the four Phase III programmes conclude, we have increased our R&D expense assumptions somewhat to take into account the continued high spending rate. We have also removed all Otsuka-related R&D fees and expenses (~£7.5m from each line), although this has no NPV impact as every dollar of Otsuka-related revenues was matched by a dollar of Otsuka-related expenses. GW also announced a slight delay in receiving the results of its second Dravet trial to H216 from mid-2016. The NDA filing is still expected to occur in Q416 and the company will request a pre-NDA meeting with the FDA as soon as the first trial results are released. While this change is minor, we had originally expected all results to be released in H116, so we do not now expect Epidiolex to launch in 2017 (although it was likely only to be on the market for three months or so in our previous expectation) and have removed those revenues from our model. We currently estimate profitability in 2018 with a total cash burn of around £160m from 2016-17, providing GW with ample cash flow to achieve its goals without the need for additional capital.