- Zafgen’s beloranib was tested on extremely obese and weakened individuals suffering from PWS, or Prader-Willi Syndrome.
- Two patients in the PWS trial did die from thrombosis (one in the OLE, or open label extension portion).
- We present significant evidence to show that thrombosis may be a significant *pre-existing* factor in deaths of PWS patients — not a primary cause of the drug.
- Even if beloranib moderately increases thrombotic event risks, there may be a viable risk mitigation strategy. We will write a follow-up article when more data is available.
- As a counterbalance to potential safety issues, we highlight extremely powerful efficacy evidence presented this Sunday, April 3, 2016.
In its January article, Kerrisdale Capital publicly announced its Zafgen (NASDAQ:ZFGN) short position with a sweeping analysis of the current data. While holding Zafgen is still a risky proposition as we cannot rule out thrombosis risk as a side-effect of the mechanism of action, Kerrisdale made several misleading statements in its article, at a minimum. We dissect some of these statements here, while highlighting the highly positive anti-obesity effects of Zafgen’s beloranib.
Let us first briefly go over what PE and DVT are (and some possible risk factors for it), the main questions in the Zafgen safety saga.
As the Mayo Clinic states,
Thrombophlebitis (throm-boe-fluh-BY-tis) occurs when a blood clot blocks one or more of your veins, typically in your legs. Rarely, thrombophlebitis (sometimes called phlebitis) can affect veins in your arms or neck.
The affected vein may be near the surface of your skin, causing superficial thrombophlebitis, or deep within a muscle, causing deep vein thrombosis (DVT). Thrombophlebitis can be caused by trauma, surgery or prolonged inactivity. Superficial thrombophlebitis may occur in people with varicose veins.
Superficial thrombophlebitis usually self-resolves, but it is often recommended to take blood thinners, just in case. A DVT should also be treated (usually with blood thinners) lest it turn into a PE.
A DVT can turn into a PE, or pulmonary embolism. A PE is a life-threatening condition where the blood clot travels into the lungs and starts blocking arteries. Shortness of breath, chest pain, and cough can be symptoms of a PE. A PE is often treated by blood thinners to break up the clot.
The risk of a DVT in the general population is .1% a year. For obese individuals this risk is significantly increased, as we shall detail further.
1) In one study, it was found that deep vein thrombosis resolution is not accelerated with increased neovascularization.
Neovascularization is the formation of new blood vessels, similar to angiogenesis, which is the expansion of blood vessels from pre-existing blood vessels. Scientists found that in rats, neovascularization did not help resolve deep-vein thrombosis. This is a strong argument against the claim that beloranib increases likelihood of thrombosis due to its action of angiogenesis inhibition.
2) Contraceptives — something multiple patients who had reported a thrombosis-related event were using — use can increase risk of DVT. In fact,the safety information for a commonly used contraceptive implant warns about increased chance of blood clots and deep vein thrombosis:
“The use of NEXPLANON may also increase your chance of serious blood clots, especially if you have other risk factors, such as smoking. If you smoke and want to use NEXPLANON, you should quit. Some examples of blood clots are deep vein thrombosis (legs), pulmonary embolism (lungs), retinal thrombosis (NASDAQ:EYES), stroke (brain), and heart attack (heart). It is possible to die from a problem caused by a blood clot, such as a heart attack or stroke.”
Further evidence: see guidelines by Association of Reproductive Health Professionals, warning against possible venous thromboembolism and deep vein thrombosis.
Three of the patients with adverse effects in the Beloranib arms used implanted contraceptives. Two of the patients were in the ZAF-201 study on 2.4mg beloranib, and one patient was in the ZAF-203 study on 1.2mg Beloranib – see slide 21 on the January P3 top-line results here.
3) Of the two deaths reported in the PWS trial, both are at highly increased risk for pulmonary embolism and deep vein thrombosis due to pre-existing medical conditions. They were randomly assigned to the beloranib arms, but potentially would have suffered the same fate if they were assigned to placebo arm.
One patient was 55 BMI with multiple co-morbidities, on 1.8mg Beloranib.
The other patient had ongoing history of thrombophlebitis.
4) The placebo arm in ZAF-311 had an n=34 (34 patients), 1.8mg beloranib an n=36, and 2.4mg beloranib an n=37. There were 34 patients on placebo compared to 73 on beloranib; in other words, there were 115% more patients on beloranib than placebo.
5) In the ZAF-201 study, 35 people dropped out before full completion. Most of these drop-outs were in the 2.4mg arm due to insomnia (losing weight means sleep patterns would have to be re-learned and not based on food coma). However, those who dropped out in the placebo arm may have had unrecorded adverse events, such as thrombotic events.
6) A person with PE event in the 201 trial had Factor V Leiden, which causes abnormal blood clotting and multiplies risk by 5x – 7x (or potentially 25-50x depending on the form): see this and this for more information.
7) Another person had varicose veins, which multiplies risk by 4x to 8x.
8) A ZAF-311 2.4mg patient had an Androgel 1% transdermal patch, which is a hormone treatment currently often-provided for PWS patients to attempt to treat some conditions, but which according to some literature may cause sudden death – this may be due to an unnoticed pulmonary embolism.
9) A gout attack and extended immobilization, as reported in one patient in ZAF-201, could of course cause a blood clot to develop.
10) In the diabetes trial (ZAF-203), one patient had an SAE of PE, with previous conditions of an implanted contraceptive, heart failure, and systemic pulmonary inflammatory disease.
11) In obese patients (especially women), the thrombosis risk increases substantially: a 5.2x multiple in patients under 40 years for PE and DVT, and in females, a 6.1x risk in DVT:
“Obesity had the greatest impact on patients aged less than 40 years, in whom the relative risk for pulmonary embolism in obese patients was 5.19 (95% CI = 5.11-5.28) and the relative risk for deep venous thrombosis was 5.20 (95% CI = 5.15-5.25) (Table 1). In females aged less than 40 years, the relative risk for deep venous thrombosis comparing obese with non-obese patients was 6.10 (95% CI = 6.04-6.17). In males aged less than 40 years, the relative risk for deep venous thrombosis was 3.71 (95% CI = 3.64-3.79).”
12) According to multiple previous statements by Zafgen in its 10-K,
“There were no deaths or any SAEs deemed to be possibly, probably, or definitely related to beloranib, although there were two serious thrombotic adverse events which, while not attributed to beloranib treatment, may point to the utility of assessment of prior history of thrombotic events in patients enrolled in subsequent trials and added vigilance for AEs related to blood clotting during future clinical trials”.
This statement has since been amended after the deaths in the PWS study, but it shows that trial investigators did not believe the previous events were caused by the drug. Except for two events in the 203 study, in every other event noted in the 201, 203, and 311 study, there are significant pre-existing risk factors. There were two events in the diabetes 203 study designated as possibly related to beloranib four and nineteen weeks after the last dose was taken — we believe this designation was due to a swing to extreme caution.
Weight Loss and Hyperphagia Reduction Efficacy
“Recent excitement about beloranib’s efficacy is misplaced given the terrible risk/reward balance: the drug leads to 4-5% weight loss but quadruples the risk of death.”
Kerrisdale substantially misrepresents the bestPWS Phase 3 (ZAF-311) data in a hard-to-treat PWS (Prader-Wili Syndrome) population. In the two drug arms,weight loss was linear in most weeks, suggesting continued treatment cancontinue to reduce weight. In the placebo arm, already significantly obese patients gained an average of 4.2%. The difference between the 1.8mg arm versus the placebo arm was 8.2% +- 1.3%, and the difference in the 2.4mg arm versus the placebo arm was 9.5% +- 1.3% (note the low variability). A 8.2% to 9.5% difference between drug and placebo arms is twice as big as the factually misleading number helpfully provided by Kerrisdale. Furthermore, the drug was only taken approximately half a year on average, and with the near-linear reduction in weight loss seen in this trial, we could surmise an 16-19% weight loss reduction for a whole year of treatment in the PWS population. This does not account for the possibility of exercise, something that is made substantially easier with an enormous drug-based weight loss reduction and a reduction in time spent eating and searching for food due to the enormous decrease in hyperphagia (i.e.: constant, unceasing hunger):
Total Score at Baseline
|Unit Change in
adjusted Change in
|2.4 mg beloranib (n=37)||18.3||-7.4||-7.0||0.0001|
|1.8 mg beloranib (n=36)||17.4||-6.7||-6.3||0.0003|
1Endpoint results shown are Least Squares mean values.
The HQ-CT is a PWS-specific study instrument that provides an assessment by caregivers of the food-seeking behaviors exhibited by patients. The scale provides a composite value from nine questions, each rated on a scale of zero to four units (total range of score of zero to 36). Patients in the ZAF-311 trial were enrolled only if their baseline HQ-CT total score was greater than 12 units, representing moderate-to-severe hyperphagia related behaviors at baseline. While hyperphagia-related behaviors were stable over six months of treatment in the placebo arm, both the 2.4 mg and 1.8 mg beloranib arms showed highly statistically significant reductions in HQ-CT total score, indicative of reduced hunger-associated behaviors.
Kerrisdale writes, “Moreover, instances of hyperphagia were actually morecommon among beloranib patients than placebo patients”. Yes, they were — at baseline!
Further, Kerrisdale does not consider any previous study in its deeply flawed analysis. In a 12-week 2012-2013 ZAF-201 trial of 147 obese patients, a weight loss of up to 22% (an average of average 16%) on the 2.4mg arm was achieved — in just 12 weeks of dosing.
Risk of Death
Kerrisdale inaccurately portrays the risk of death at “quadrupling”. In the ZAF-311 study, average patient age was 20.9 (+-7.8) on placebo, 19.2 (+-5.2) on 1.8mg beloranib, and 19.5 (+-5.8) on 2.4mg. The death rate in this trial and subsequent OLE (open-label extension) in the drug arms was 2 out of 73, or approximately 5.5% on a yearly basis. For all patients, the death rate was 2 out of 107, or approximately 3.7% yearly. Literature puts the historical death rate at 3%. Given high rates of variability with an n of 107 and two deaths, we can safely say that the risk of death is not quadrupled, as Kerrisdale claims.
(While a recent Loker study has reported a mean age of PWS sufferers of 34 at the time of death, and other historical data suggests 32 [Zafgen Q3 update call, page 7], we are further assessing the possibility that the risk of death substantially rises in the age range of 15-25 years, especially in extremely obese individuals. We may discuss this in a subsequent article once we have analyzed more data.)
In its article, Kerrisdale cites three studies to support its claim of an actual .7% 5-year yearly death rate, for ages 20-25: Grugni 2008 (n = 425), Lionti 2012 (n = 163), Einfeld 2006 (n = 37). It is worth noting that the average age in ZAF-311 was closer to 20, not 25 as the Kerrisdale analysis seems to assume, roughly ranging between 28.7 and 13.1.
In the Lionti study, 26.4% of trial participants never developed obesity, and41.9% were not genetically diagnosed as having PWS. To compare this trial to the Zafgen 311 study of genetically confirmed morbidly obese PWS patients seems to indicate, at best, lack of statistical competency. The Lionti study suggests that obesity-related death is the primary cause of death. Furthermore, out of 10 known causes of death, 2 are pulmonary embolisms, at 15 (63.2 BMI) and 29 (48.1 BMI), which is in the patient range for Zafgen’s trial. Both participants, and indeed all deaths recorded in this study where the genetic mechanism was known, had a chromosome 15 deletion, as did 71% of ZAF-311 participants.
The Grugni study cited by Kerrisdale studies 425 subjects, using data through a form filled out by parents of PWS children identified by 25 medical centers in Italy, with genetically confirmed diagnosis of PWS: “Two hundred thirty-eight patients had del15, 104 had UPD15, 4 demonstrated a translocation affecting chromosome 15 and 79 showed a positive methylation test.” In the Grugni study, which is not a long-term study but an assessment of a form filled out by parents, there were only 18 deaths reported, which suggests a lower-than-expected death rate. However, there are some problems in comparing this study with Zafgen’s trial:
1) Only 62.6% of those subjects with recorded BMIs were obese. While we acknowledge that obesity often begins in the teen years, and that this study states the vast majority of adults (those over 17 years of age in this study) had developed obesity, this still does not compare well with the patients in the Zafgen trial where both adults and younger teens were extremely obese, ranging from a BMI of 29.4 to a BMI of 53, according to slide 10 of the January 20 PWS Zafgen-provided trial results.
2) Sixty-two (14.6%) of subjects were over the age of 29, 17 (4%) over 36, and 4 (.9%) were over 42. Given the low IQ of the average PWS sufferer and the need to take care of them long after a normal age in the regular population, we believe that even older PWS sufferers would be under the care or supervision of their parents. (i.e.: under a group home) If the mortality data is as low as Kerrisdale suggests, where are the older children? The age characteristics of the population, with just 4% over the age of 36 and 0 older than 47 indicates a potential bias in the data towards respondents with living children, potentially greatly under-counting the number of deaths in the population. We again note that in the Loker study of a US population (as in the Zafgen trial), the average age at death was 34.
Einfeld’s 2006 study, which tries to understand the contribution of intellectual disability to risk of death, does not cite any figures for its population regarding prevalence or severity of obesity, so we cannot compare it well to ZAF-311 — except to show that every known cause of death was caused by obesity-related factors, with 3 out of 6 deaths occurring at the ages of 20, 21, and 23:
|Gender||PWS status||Age at death||Cause of death|
|Male||Maternal uniparental disomy||20||Cardiorespiratory failure 4 weeks, morbid obesity|
|Female||Paternal deletion||21||Cardiomegaly, morbid obesity|
|Male||Paternal deletion||23||No death certificate available|
|Female||Proven PWS, type unknown||32||No death certificate available|
|Male||Maternal uniparental disomy||37||Adult respiratory distress syndrome, possible sepsis, possible aspiration, right ventricular failure|
|Female||Maternal uniparental disomy||43||Pneumonia, cerebro vascular accident, diabetes mellitus|
How many Deaths Are Caused By PE?
Finally, Kerrisdale cites Jim Loker’s as-yet-unpublished study showing the risk of death from PE (pulmonary embolism)-related causes at 6%. Well, first of all, it’s actually 6.6% — 19 out of 287* deaths were directly caused by PE. (In another 4 or 1.4% of deaths, there was evidence of prior PE, but patients died from a different cause.) Then, Kerrisdale take their 6% and their unrealistic .7% death rate and come up with a probability of 3,520 to 1 of deaths occurring from PE as natural. And yet, in the Lionti study, 2 out of 10 (20%) died from PE.
* Unpublished data — email correspondence. 287 patients died from complications due to obesity.
How can we possibly reject a 3,520-to-1 odds? First, some back-story:
1) According to the Johns Hopkins Medicine Health Library,
“PE is often difficult to diagnose because the signs and symptoms of PE are a lot like those of many other conditions and diseases. Imaging tests and blood tests are used to look for a PE.”
2) According to this study,
“The results validate previous studies that this diagnosis [pulmonary embolism] is often missed… Even when the diagnosis is considered, however, thrombolytics are not routinely given, even without contraindications… It also underscores the continued difficulty in the diagnosis of this disease.”
3) According to this 2011 study, all sources of PE are often not been considered on autopsy, causing an autopsy to sometimes miss a determination:
“Periprostatic or paravaginal venous thromboses are rarely considered clinically as sites of clot origin in patients with pulmonary thromboembolism. The majority of emboli have been demonstrated to originate in the veins of the legs. This report raises awareness of pelvic vein thrombosis as a potential source of pulmonary embolism that is rarely considered or detected clinically, and which usually requires postmortem examination for recognition. It also reviews the possible routes emboli may take to reach the lungs.”
4) This 2015 study paints a similar picture, where, while PE was diagnosed, no source of PE was identified in 22.6% of patients: “Chronic obstructive pulmonary disease (AOR .173, 95% CI .046 to .646, P = .009) was predictive of not identifying a PE source.” Recall that in the Leonti study, one patient had pulmonary heart disease as cause of death, and one patient had chronic pulmonary hypertension. Including these two, this would raise the identifiable sources of death caused to PE to 40% in the Leonti study.
Notably, in the Loker study, 170 patients (at least 35%) did not have a cause of death listed out of over 480.* There is a significant body of evidence to suggest that a substantial amount of unlisted causes of death were caused by a pulmonary embolism, as detailed above. Given that we know up to 19 out of 480+ patients died due to confirmed PE, and potentially another 4 died with evidence of PE, what happens to our chances when we add, say, 160 patients into the PE category? The chances then become 38% per death. In the ZAF-311 trial, the death rate was approximately 3.74%, counting both (n=34) placebo and the (n=73) drug arm. With a 3.74% “natural” death rate, and an actual observed 3.74% death rate, the chance that two patients will die due to PE is 14.4% (.38 * .38), or 1-in-7. While this does assume some optimistic projections and is subject to further study, it is a far cry from 1-in-3,520 proposed by Kerrisdale.
* Unpublished data — email correspondence.
In PWS sufferers, it is commonly reported and indeed common knowledge that pain tolerance is significantly higher than normal. This helps partly explain why pulmonary embolisms are often not diagnosed – some may have not felt the pain that comes with PE right away, and thus no doctor performed the necessary procedure in time to diagnose the pain as PE before death. Given that there seems to be significant evidence of under-diagnosis of PE, we can potentially chalk up these deaths as bad luck, not a dangerous drug.
April 3 Presentations
Even if beloranib moderately raises the risk of PE/thrombosis, we believe that this risk may be mitigated with blood thinners. Further, the known benefit profile has been greatly enhanced with the disclosure of additional benefits as reported by Zafgen in the week and day of April 3. These benefits track those reported in the ZAF-201 trial. Data from the ZAF-201 trial is provided below:
Interestingly, although this seems counter-intuitive, systolic blood pressure was very significantly lowered in the ZAF-201 beloranib arms compared to placebo, despite the direct mechanism of action that shrunk the amount of blood vessels. This was likely a result of the significant weight loss in the beloranib group of up to 22% in 12 weeks. Similar to the ZAF-201 trial, patients in the ZAF-311 trial enjoyed an enormously lowered LDL (the “bad cholesterol”), as well as enormously lowered total cholesterol, leptin, and CRP, while seeing increased adiponectin (these are all hugely positive developments):
|Mean at Baseline||101||174||6.9||4.6||7.6 μg/mL|
|2.4 mg beloranib (n=37)||-17.8**||-18.0*||-24.0**||+1.9**||-53%**|
|1.8 mg beloranib (n= 36)||-16.6**||-17.2*||-20.7**||+1.7**||-56%**|
For the moment, likely the most significant usage of beloranib is in PWS. If thrombotic issues are fully resolved, beloranib’s value as a general obesity medication would skyrocket. However, beloranib may have orphan indications for obesity treatment beyond just Prader-Willi Syndrome patients. Obesity is also a major feature in Bardet-Biedl syndrome, Cohen syndrome, Ayazi syndrome, and MOMO syndrome.
Written in collaboration with Agamemnus.
Disclosure: I am/we are long ZFGN.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.