How to Spot a Short Biotech Opportunity, Part II
What, exactly, defines pivotal drug trial data as “positive”?
On the surface it would seem that any drug that meets its trial’s primary endpoint would have produced data that could be defined reasonably as “positive.” But even that definition should be accepted with caution, as the primary endpoints that are submitted to the FDA 1) are hand-picked by the company that is developing the drug and sponsoring the trial; and 2) may have little to do with the standard clinical practice of the disease in question. Meeting a primary endpoint is relevant only if the endpoint itself is relevant.
Issues such as this call to mind the curious case of AP Pharma (APPA), which nearly two years ago released the “positive” data from its phase 3 trial of sustained-release granisetron (for chemotherapy-induced nausea and vomiting, CINV) – and, on the heels of this “good news,” subsequently saw its stock price drop about 33% in a day. The drug had met three of its primary endpoints, the company announced, but not the fourth (for treatment of delayed-onset CINV in highly emetogenic chemotherapy) – which was the only one that had any relevance to the investigational drug’s potential use in clinical practice.
I have included below the report I had prepared for a client three months before APPA released its data, as another case study in identifying a potential short opportunity in the biotechnology sphere. Similar to the case discussed in the first installment of this article (in which SNTA’s melanoma drug, elesclomol, was analyzed), the case of APPA and its investigational drug, APF530, also hinged on the binary event of phase 3 data release. But in the case of APPA, a further question lurked in the background – namely, even if APF530 were to achieve positive phase 3 results, and even receive distinction from the FDA as approvable, would it matter? Would the drug, even if it one day reached the marketplace, be prescribed? And, if prescribed, would anybody pay for it?
I argued “no” at the time, to all of the above questions, and though APPA never got past the first hurdle – the pivotal trial – these questions might represent the second wave of attack on a long position in any biotech firm whose fortunes rest on a single investigational product. A few other companies, too, might qualify as having “approvable” (or even approved) drug candidates that nevertheless have difficulty addressing the first two questions – Will anybody want to use it? Will anybody pay for it? – as well as a third (Over time, will it still prove to be safe?). AMAG, with its questionably beneficial drug for anemia in chronic renal disease, immediately comes to mind.
Before reading the 2008 report on APPA, I have included the five basic priniciples of the APF530 case that might lead any investor toward a short bias on any drug. They are:
1) The investigational drug in question is simply an off-patent drug, packaged in a new delivery formulation, and then marketed for a new indication. Changing a drug from an immediate release formulation to a sustained release formulation can be valuable, provided the new formulation is used for the same indication and increases patient compliance. But when an old drug has its delivery mechanism changed so that it can market itself for a whole new indication (in this case, putting granisetron in a sustained-release formulation and now claiming it works for delayed-onset CINV, instead of only early-onset CINV) the results are rarely successful – and, if successful, easily knocked off by potential generics competitors.
2) The control arms do not conform to academy standards, and so even positive results would not change clinical practice standards. Sometimes, new drugs for diseases that have established therapies are compared to placebo – this would be the most obvious example of a trial that, even with positive clinical data, might not change anything: the new drug may be better than nothing, but we already have more than nothing to treat the disease. In the case of APPA, the control group did not receive academy-endorsed treatment for delayed-onsent CINV in highly emetogenic chemotherapy.
3) The proposed therapy, even if successful, already has unfavorable cost-efficacy data. It is worthwhile to see if healthcare economists have already taken a look at the drug in question, particularly when the current presidential administration touts the importance of “comparative effectiveness.”
4) Results from early stage studies (phase 1 or 2) have not been published in peer-reviewed journals. The immediate inference here would be that articles derived from earlier studies were either rejected by such journals, or so flawed that exposure to public scrutiny would injure public relations.
5) Earlier stage studies that have been “successful” had undertreated patients in their control groups.
Remarkably, APPA still remains an actionable stock: the company is still seeking approval from the FDA for APF530 in the three indications in which it proved to be “non-inferior” to control, and its stock price continues to hover around $1. At its current price, it is hard to call APPA a short, but the full report I prepared in 2008 is provided below:
APF530 (biochronomer-treated granisetron) is a novel formulation of granisetron, a 5-hydroxytryptamine type-3 receptor antagonist (5-HT3RA) used for the prevention of the nausea and vomiting that often follows administration of chemotherapy. The active drug in APF530 is granisetron, which has a long, effective and safe history in the treatment of chemotherapy-induced nausea and vomiting (CINV). What distinguishes APF530 from typical formulations of granisetron is its method of delivery: APF530 is delivered in a bio-eroding polymer, administered by subcutaneous injection, that can sustain a controlled release of granisetron over a period of five days. AP Pharma (APPA), the manufacturer of APF530, contends that this prolonged release of granisetron will allow for control of CINV in both its acute setting (defined by clinical oncologists as occurring within 24 hours of chemotherapy administration) as well as in its delayed setting (greater than 24 hours after chemotherapy, with a generally recognized limit of 120 hours). To date, 5-HT3RAs as a class of drugs have been very effective at controlling acute CINV, but have been far less effective in controlling delayed CINV. For the control of delayed CINV, the steroidal anti-inflammatory dexamethasone has proven to be a successful clinical tool, though its mechanism of action remains unknown. Aprepitant, a neurokinin-1 (NK-1) receptor antagonist that has penetrated the market in the last 10 years, has had great success in the treatment of delayed CINV – probably because delayed CINV is mediated principally by substance P, a neurotransmitter that binds NK-1 – and has become the standard of care for delayed CINV for highly emetogenic chemotherapy (HEC), usually in conjunction with dexamethasone; it is also the standard of care in some moderately emetogenic chemotherapy (MEC) regimens. Olanzipine, an antipsychotic drug, has also had promising results in the treatment of delayed CINV (probably through its antagonism of dopamine and, to a lesser degree, serotonin receptors), but its use has been limited to second line and treatment failures because of its narrow therapeutic window. Of all the 5-HT3RAs to the present date, only palonosetron – which has a very long half-life and an unusually strong binding affinity to the receptor – has demonstrated any consistent success for the treatment of delayed CINV, despite the fact that the entire class of drugs is extremely successful in the treatment of acute CINV. This is probably because acute CINV is mediated by serotonin through its binding of the 5-HT3 receptor, and this mechanism of acute CINV rapidly attenuates after 16 hours, with a substance P-mediated mechanism gaining predominance after the first 24 hours. APPA is challenging the notion that granisetron is less effective than its clinical alternatives in the treatment of delayed CINV, and to that end has designed a pivotal phase III trial in the treatment of both acute and delayed CINV in HEC and MEC. APF530 has been tested in a three-arm trial: one arm gives a single injection of 5 mg APF530 on day 1 of chemotherapy; a second arm gives a single injection of 10 mg APF530 on day 1 of chemotherapy; and a third, active comparator arm gives a single infusion of 0.25mg palonosetron on day 1 of chemotherapy. All arms will also receive a dexamethasone infusion on day 1, and HEC patients will also receive oral dexamethasone on days 2-4. All arms will also receive either an IV or SC administration of placebo, so that all patients in the study will have received two IV infusions and one SC injection. Roughly half of the 1400 patients received HEC, and the other half received MEC. The trial is designed by APPA to demonstrate non-inferiority of APF530 to palonosetron for acute CINV in MEC and HEC, and to delayed CINV in MEC; the trial is also designed to demonstrate superiority of APF530 to palonosetron for delayed CINV in HEC. Results of the trial, already completed, are due 2H08.
Arguments Against the Generation of Positive Phase III Data and Adoption by Clinical Oncologists
- The phase III APF530 trial is not relevant to current American Society of Clinical Oncology (OTC:ASCO) guidelines for antiemetics in oncology, and so even positive results will not have a significant effect on clinical practice. The most obvious problem with the phase III APF530 trial design is that none of the arms represent the current standard of care for the prevention of CINV. The ASCO guidelines clearly state that all HEC patients should receive aprepitant on days 1, 2 and 3 of their chemotherapy course; both the National Comprehensive Cancer Network (NCCN) and the Multinational Association of Supportive Care in Cancer (MASCC) agree with these guidelines. However, aprepitant is not part of any regimen in this study, and so no matter the results that this pivotal trial produces, it will not be able to establish itself as either equivalent or superior to current standards of care (at least not for HEC patients). Likewise, ASCO guidelines also state that for MEC patients, dexamethasone must be continued through day 3 of the chemotherapy course if aprepitant is not being used; both NCCN and MASCC agree with these guidelines. However, the phase III APF530 trial is only giving dexamethasone on day 1 to the MEC patients, and so the treatment of MEC patients, too, has essentially omitted any comparison to current standards of care. It should also be noted that ASCO, NCCN and MASCC all agree that MEC patients receiving a combination of an anthracycline and cyclophosphamide should also receive aprepitant on days 1, 2 and 3 – as already stated, however, aprepitant is not included in this trial, and so this recommendation has not been recognized by APPA in its trial design. In defense of APPA, this may not all be the result of poor planning, but may have been the result of bad luck: APPA initiated its trial in 2006, only months before ASCO updated its antiemetic guidelines. Nevertheless, the phase III trial design does not reflect current standards of clinical care, and this will likely cause dismissal – or, in the very least, skepticism – of whatever positive results the trial produces.
- Published clinical evidence has established that the addition of a 5-HT3RA to dexamethasone does not significantly prevent delayed emesis, and that the use of a 5-HT3RA beyond 24 hours after chemotherapy does little to prevent delayed emesis. Dexamethasone – an old, safe, familiar and generic steroidal anti-inflammatory drug – is so effective at preventing delayed emesis in MEC that it does not benefit from the use of additional antiemetic therapy, and does not receive added benefit in HEC beyond the addition of aprepitant. This, in fact, is the official policy of ASCO (as well as NCCN and MASCC), and is substantiated by the clinical evidence that delayed emesis is not mediated by serotonin (which binds the 5-HT3R), but by the substance P-mediated binding of the NK-1 receptor. Trials that have looked specifically at the use of granisetron and dexamethasone for delayed emesis (6) have demonstrated that the administration of granisetron after 24 hours conferred no benefit over dexamethasone alone. Considering that the APF530 trial will compare not just granisetron and dexamethasone vs. dexamethasone alone – but against dexamethasone AND the extremely long half-life 5-HT3RA palonosetron – it is unlikely that this study will be able to detect even a small incremental effect of a 5-HT3RA when added to dexamethasone, let alone an effect that is uniquely attributable to granisetron. In trials that have looked at granisetron on day 1 with ondansetron on days 2-4 in combination with dexamethasone (9), none of the efficacy variables related to control of delayed emesis differed significantly between this combination and dexamethasone alone. And Paul Hesketh, M.D. – the Harvard Medical College oncology professor who is the author of the Hesketh algorithm for prophylactic management of CINV – has demonstrated that serotonin mediates the early CINV that occurs within 8-12 hours of chemotherapy, after which time substance P acting at NK-1 receptors becomes the dominant mediator of CINV (11). In sum, the entire premise of the APF520 trial – that longer acting 5-HT3RAs can be as effective at controlling delayed CINV as they are at controlling acute CINV – is flawed.
- Cost-benefit analyses of 5-HT3RA use for the delayed emesis (>24 hour) period have demonstrated that their minimal improvements do not justify their costs. Neither clinical evidence nor considerations of cost effectiveness justify using 5-HT3RAs beyond 24 hours after chemotherapy for prevention of delayed emesis (5). A meta-analysis acknowledged by ASCO determined that when 5-HT3RAs are administered in combination with dexamethasone, there was no added benefit, and so it is impossible (and irrelevant) to demonstrate cost-efficacy of a futile therapy. But even when 5-HT3RAs were administered beyond the 24 hour period and compared against no therapy at all, it was determined that 423 doses of ondansetron would have to be administered to prevent a single episode of vomiting. At an average price of $30.45 per ondansetron tablet, the drug acquisition costs to prevent a single episode of CINV would be $12,880. Now, if this $12,880 cost of a futile therapy were compared to the cost of one that actually works – e.g., dexamethasone, which as an IV formulation costs about $0.62 per dose (and less when given orally), or aprepitant, which has a full course cost of about $590 (but a probable drug acquisition cost of under $2000 when numbers needed to treat are considered) – the balance still weighs strongly against the use of a 5-HT3RA. When it is further considered that, even if demonstrated to show a modest clinical benefit, APF530 is likely to cost approximately $370 per dose (the cost of palonosetron, and roughly twice the cost of a four day ondansetron regimen), then it becomes even clearer that it will be virtually impossible to justify the cost of this drug, even in the unlikely scenario that it gains FDA approval. No managed care organization will bear the extreme costs of marginally effective therapy, let alone therapy that might add nothing at all to current standards.
- The cumulative dose of granisetron in the APF530 trial will be either 5 mg or 10 mg (over a 5 day period). There are published clinical data, however, that demonstrate 5 day cumulative doses of as much as 13 mg granisetron (given in multiple dose, in standard formulations) do nothing to prevent delayed emesis. In many ways, this contention is simply a re-statement of the fact that 5-HT3RAs do not work beyond the first 24 hours of therapy. Nevertheless, there is precedent that giving granisetron specifically over a five day period does not prevent delayed emesis (6). In a 1996 trial acknowledged by ASCO, granisetron was given as 13 mg over a five day period (and 17 mg over a period of seven days) in conjunction with dexamethasone, and it offered no benefit over dexamethasone alone. Now, the current APF530 trial is essentially the administration of granisetron with multiple doses over multiple days – the active drug (granisetron) is the same, but the biochronomer delivery system allows for a gradual absorption of the drug into the bloodstream over a five day period. Since the cumulative dose of granisetron in the APF530 trial is actually smaller (5 or 10 mg) than the cumulative dose in previous trials that have failed, it is even more unlikely that APF530 will be able to demonstrate any benefit at all.
- For acute emesis (<24 hours), all 5-HT3RAs are equivalent, and so APF530 is unlikely to demonstrate superiority in acute emesis in its head-to-head comparison with palonosetron, or in comparison to historical controls. This has been the conclusion of virtually every meta-analysis that has compared 5-HT3RAs – whether they have been sponsored by ASCO, NCCN, MASCC, the Cancer Care Ontario Practice Guidelines Initiative, or any other group: the 5-HT3RAs are all virtually the same. The only possible exception to the contention that all 5-HT3RAs should be considered equivalent – which, in fact, is the official policy statement of ASCO, NCCN and MASCC – is palonosetron. In both acute and delayed CINV, palonosetron has been able to demonstrate superiority to other 5-HT3RAs in head-to-head comparisons (12) – in the acute setting, probably because of palonosetron’s superior binding affinity to the receptor; in the delayed setting, palonosetron’s superiority to other 5-HT3RAs is probably due to its significantly longer half-life, and a blockade of the minor elements of serotonin-mediated CINV that exist beyond the 24 hour period of chemotherapy. With palonosetron as its active comparator, APF530 has quite a hill to climb: 1) in the acute phase, it is being compared to a drug that, in the best-case scenario, it can only hope to achieve non-inferiority; and 2) in the delayed phase, it is hoping to surpass the most effective drug in a class that does not have a clear indication for this clinical use. In general, there are two schools of thought within the oncology community: One is that all 5-HT3RAs are interchangeable – if true, APF530 is unlikely to succeed in its trial, because it will not demonstrate superiority. The other school contends that palonosetron is a best-in-class molecule – if this is true, then APF530 is unlikely to succeed in its trial, for it will not demonstrate non-inferiority.
- The pharmocokinetics of granisetron, even in a sustained-release formulation, compares unfavorably to palonosetron. Though the biochronomer formulation of granisetron will sustain release of the drug, it will do nothing to affect its half-life or, more importantly, its binding affinity. Palonosetron’s binding affinity (pK) to the 5-HT3R is nearly two logs (or 100 times) as strong as granisetron’s. The biochronomer delivery system that signifies APF530 as unique is merely an eroding polymer that sustains release of the drug – there is no change in the molecule that makes it bind its receptor target with greater affinity. So in this respect, at least in the first 48 hours of therapy, APF530 will have an extremely difficult time competing with palonosetron. Similarly, palonosetron’s half-life is nearly five times as long as granisetron’s – 40 hours, as compared to eight. Again, the polymer delivery system of APF530 does nothing to prolong half-life – the granisetron molecule is unchanged. But it does sustain a continued release of the granisetron over approximately five days. Given that the half-life of palonosetron is about 40 hours, this sustained release mechanism is unlikely to confer any clinical benefit to APF530 in the first 48-72 hours of therapy. Perhaps beyond that point – days four and five of the chemotherapy course – APF530 will still be active, while the palonosetron dose has dissipated. But at this point, the contribution of serotonin-mediated effects on CINV are so small that it is unlikely that a significant benefit over palonosetron can be measured.
- As experience with aprepitant accumulates and doctors grow more comfortable with its use, it is likely that it will expand its market into the moderately emetogenic therapy space and further squeeze out 5-HT3RAs for delayed emesis. Aprepitant has already gained a foothold in the delayed CINV market space – it has academy recommendation from ASCO, NCCN, and MASCC for HEC. It also has recommendations from those same academies for MEC that utilizes an anthracycline drug in combination with cyclophosphamide. In addition, aprepitant has the endorsement of the NCCN for its use in all MEC, including those regimens that do not employ anthracyclines – as clinical use with aprepitant becomes more widespread, and oncologists grow more comfortable with its use in specific MEC regimens, it is likely that the other two academies (ASCO, MASCC) will follow suit and recommend aprepitant more broadly. Even if successful in its trial, APF530 is likely to see a further competitive infiltration of its potential market space (specifically, MEC) by aprepitant.
- Three of the four goals of the APF530 trial, even if met, will do nothing to create a market for the drug. As stated publicly by APPA, the APF530 trial is looking to achieve four primary endpoints: 1) non-inferiority to palonosetron for acute CINV in MEC; 2) non-inferiority to palonosetron for acute CINV in HEC; 3) non-inferiority to palonosetron for delayed CINV in MEC; and 4) superiority to palonosetron for delayed CINV in HEC. However, the entire success of APF530 hinges on the fourth endpoint, and the fourth endpoint alone. Non-inferiority to cheaper and/or long-established treatment alternatives will not lead to adoption of APF530. To grab any market foothold at all, APF530 must demonstrate superiority to palonosetron for delayed CINV in HEC. Even then, APF530 will meet some very serious marketing challenges: as mentioned, aprepitant in combination with dexamethasone is an established and effective regimen for delayed CINV in HEC, and this trial will do nothing in the way of addressing whether APF530 can be just as effective as this standardized combination.
- Previous trials of APF530 that have demonstrated superiority to trials of palonosetron for delayed CINV in HEC have not standardized dexamethasone use, nor were they of equal statistical rigor. APPA has designed its phase III trial based on its phase I and II results with APF530, and historical standards of palonosetron trials that were submitted to the FDA. According to investor presentations by APPA, APF530 is non-inferior to palonosetron for acute and delayed CINV in MEC; it is insignificantly superior to palonosetron for acute CINV in MEC, and significantly superior to delayed CINV in HEC. There are numerous problems with these comparisons, however. For one, none of these comparisons were head-to-head – they compare a recent trial of APF530 to a historical trial of palonosetron. Two, the APF530 trials have been, to this point, open label, whereas the palonosetron trials to which it was compared were blinded and controlled. Three, the APF530 trials had approximately 20 patients in them; the palonosetron trials to which they have been compared had over 500 patients. Four, in the midstage trials of APF530 to date, dexamethasone use was not standardized: in most patients, dexamethasone was used, but not in all. In addition, HEC patients in previous APF530 trials only received dexamethasone on day 1, and not beyond (current standards call for its use through day 4, which is how it has been used in the phase III trial). John Barr, the Senior Vice President of Research & Development of APPA, notes that dexamethasone use was not standardized in any palonosetron trial, either, but this explanation is insufficient: the pivotal palonosetron trials took place prior to the publication of current academy guidelines. In addition, the poor design of a competitor’s trial does not justify the subsequent use of the same poor trial design.
- APPA has not published the results of its phase I or phase II APF530 trials in any peer-reviewed journal. It is troubling when supposedly promising results are not shared with the peer-reviewed academy. There are numerous reasons why a promising paper would not have been published: Studies with only 20 patients in them that use sub-standard protocols and, perhaps, do not achieve statistical significance seems like a likely explanation for why a study was never printed in an academic journal.
Disclosure: No positions