How to Spot a Short Biotech Opportunity, Part I
In 2008, Synta Pharmaceuticals (SNTA) concluded the analysis of its phase 3 trial for esleclomol, a drug with the potential indication for treatment of advanced melanoma. As investors awaited the release of data from its pivotal trial, elesclomol seemed to have all the makings of an approvable drug with a successful commercial future.
For one, elesclomol’s potential indication was not just for cancer, but for one with a particularly grim prognosis – a context which seems to promote the most sympathetic ears among regulatory panels. In addition, it had been decades since a new cytotoxic (“cell-killing”) drug had been approved for melanoma, and long positions seemed to rest on the grounds that it was high time the FDA approve something, provided it was reasonably safe and effective. Also, it did not hurt that SNTA had significant support from one of the biggest names in the pharmaceutical industry (NYSE:GSK).
And then, several months after the anticipated phase 3 data on elesclomol was expected to be released, the news finally broke: the drug did not work. The stock, which had been trading in the $8 range, fell to about $2. The partnership with GSK was severed.
I had prepared for a client an analysis of SNTA – or, more specifically, elesclomol – about nine months before the release of the pivotal trial data, in which I anticipated disappointing results. I have included that report here, as a sort of case study in identifying short opportunities in the biotechnology sector – particularly when a biotech stock’s price hinges on the binary event of phase 3 data release.
The report, included below, employs the same standard of evidence that the FDA would demand in evaluating any drug: evidence-based, academically oriented, peer-reviewed data, complete with references and citations. In other words, I had to go back to the library and read what the scientists had to say in the journals and in the quarterlies.
This approach lays in contrast to the more typical biotech analysis, which relies upon the impressions of “thought leaders,” or “key opinion leaders” in the field. Whereas it can never hurt to seek the guidance of an accomplished physician-researcher with expertise in a specific field, it should be remembered that the FDA does not accept expert opinion as a grounds for approval or dismissal: It is all about high-quality data, which can never be replaced by expert intuition.
Before reading the 2008 report on SNTA, I have included the five basic principles of the elesclomol case that might lead any investor toward a short bias on any drug. They are:
1) The mechanism of action (MoA) is theoretical, but not established, and there is scant pre-clinical data that attempts to define the MoA. I suppose it can be argued that, as long as a drug is reasonably safe and effective, we do not need to know exactly how it works. But in an era in which most molecules’ MoAs are well-defined, no molecule can be felt to be in expert hands if not even the expert knows how it works. And the absence of published pre-clinical data makes one wonder if the company running the trial possesses data that they do not want you to see. A more recent example of a drug that had promising early stage trials but scant published data on its MoA is dimebolin, a drug developed by Medivation (NASDAQ:MDVN) for Alzheimer’s disease – this drug, of course, failed its phase 3 trial.
2) There were significant baseline imbalances in preceding non-pivotal trials. Earlier stage studies might have significant differences in the baseline characteristics of its study group when treatments are non-randomized, or entire studies are open-label. This is fine when a study has the characteristics of a pilot investigation – but positive data can hardly be accepted as valid and ready for confirmation when earlier stage studies do not meet anyone’s standard of statistical rigor.
3) There were manipulations of the “intention to treat” principle. The most common manipulation of the intention to treat principle is to ignore it altogether – for example, not including in the final analysis any of the patients who dropped out of a study (a technique recently observed, but then corrected, by a company with an investigational weight loss drug), and thus skewing the data toward a favorable result. In the SNTA example, however, it held on to the intention to treat principle too strongly, attributing adverse events to control patients in early stage trials, even after they had crossed over to the experimental group.
4) Control groups are not given an active comparator, or are given a sub-therapeutic dose of an active comparator. This one is simple: If a disease has an accepted standard of therapy with an accepted dosing regimen, then that standardized treatment should be given to the control group – particularly when you are dealing with a disease that has a high mortality rate. Only when there is no therapeutic standard available are placebos (or, in the case of SNTA, real drugs given at dummy doses) acceptable.
5) The control group does not receive an academy-accepted standard of care. Even if data is positive, the drug will eventually need to receive FDA approval, and the FDA usually considers the value of the drug over therapies that are already available.
The full report that I prepared on elesclomol in 2008 is provided below:
Elements of the Phase I and Phase II Trials that Argue Against the Generation of Positive Phase III Data
- Not only did the paclitaxel (control) group have more patients with advanced (M1c) disease than the elesclomol + paclitaxel (experimental) group, but the control patients with M1c disease had more distant and/or multiple metastases than their experimental counterparts. The American Joint Committee on Cancer Staging melanoma guidelines define the M1c stage as any visceral (organ) metastasis other than the lung, or any distant metastasis (for example, a lymph node) with an elevated serum lactate dehydrogenase (LDH). To this end, it is already known that the rate of M1c disease (the most advanced stage) was higher in the phase II trial’s control group – 75% vs. 53% -- and so it would be expected that the control group, with its higher disease burden, would have a poorer outcome. However, SNTA reports that the rate of liver metastasis (a common site of melanoma spread) was equal in both the experimental and control groups, with a rate of 32% in each. As defined by the Joint Committee, however, liver metastasis would qualify as M1c disease – so if both the control and experimental groups had the same rate of liver involvement, but the control group had a higher rate of overall M1c disease, then we know that the control group had a higher rate of multiple or more distant metastases. Consider this: if the experimental group had a liver metastasis rate of 32%, and its incidence of M1c is 53%, then its rate of metastasis beyond the liver was 21%. Compare this to the control group, which also had a liver metastasis rate of 32%, but an overall M1c rate of 75%: that means that 43% (more than twice as many as in the experimental group) had distant metastases beyond the liver. In appreciation of this, it should be noted that the phase II trial not only had more advanced patients in the cohort, but even those patients in the most advanced disease cohort had, on average, more distant spread. This is an important point to recognize, since SNTA management has advanced the argument that, even correcting for M stage, the experimental group fared better. This may be so, but the baseline imbalance of the two cohorts is such that one cannot discount the possibility that this may have more to do with patient selection than with therapeutic effect. In sum, SNTA appears to have spun a positive result out of one of the quirks of cancer staging: quantum distinctions (e.g., M1a, M1b, M1c) are given to continuous processes, and so an appearance of equivalence can be assigned to an entity that has progressed further along the continuum (e.g., distant metastases are not distinguished from isolated liver metastases).
- The elesclomol + paclitaxel cohort also had a younger average age than the control group (57 vs. 61), and a lower rate of prior chemotherapy (56% vs. 68%). The latter point – that the control group had a 12% higher rate of prior chemotherapy than the experimental group – is probably the more significant, in that it suggests incomplete randomization, more advanced disease, disease more resistant to therapy, and/or a patient population already exposed to the deleterious effects of cytotoxic agents. Prior chemotherapy certainly suggests more advanced disease – and, in the case of melanoma previously exposed to cytotoxic agents, it is a disease which is well-described as being more resistant to other chemotherapeutic agents like paclitaxel. In fact, the investigators of the phase II study even drew this conclusion – patients previously treated with chemotherapy fared so poorly that the phase III trial has subsequently been designed to exclude them. SNTA maintains that, no matter the prior chemotherapy status, the experimental group still fared better. However, SNTA did not break down its data in a way that stratifies how many prior chemotherapy patients were in which M stage, and so prior chemotherapy status alone does not assure clinical similarity. The first point – that the experimental patients were slightly younger than the control patients – is self-explanatory: as SNTA acknowledged at its analyst day, patients over 60 (particularly male patients, which comprises 67% of the study) fare worse than those under 60. Again, the median control age was 61; the median experimental age was 57.
- There is no evidence to support oxidative stress as elesclomol’s mechanism of action. In addition, the pharmacokinetic data that comes out of the phase I trial suggests elesclomol’s mechanism of action may actually be its inhibition of paclitaxel metabolism and a subsequent prolongation of paclitaxel’s duration of action. There is no pharmacodynamic data that demonstrates whether elesclomol does anything at all – even the data which SNTA presented at its analyst day demonstrated that elesclomol has virtually no activity on melanoma when administered as a single agent. The only data, in fact, that is even suggestive of elesclomol’s mechanism of action is the pharmacokinetic data derived from the phase I study. That study demonstrated that elesclomol had a rapid elimination from plasma (the biologic half life was about an hour) and a low volume of distribution – not surprising pharmacokinetic characteristics in a drug that has not been shown to be effective as a single agent. At the same time, the co-administration of elesclomol with paclitaxel resulted in a significant decrease in total body clearance of paclitaxel, and a slower elimination of the paclitaxel – a finding that was noted to increase when the dose of elesclomol was escalated and the dose of paclitaxel was fixed. This measured pharmacokinetic data was supported by the clinical data of the phase I trial, in which co-administration of elesclomol with paclitaxel resulted in an increased incidence of paclitaxel-related toxicities (neutropenia, mucositis, myalgia, etc.) – another suggestion that elesclomol’s contribution to this regimen was its increase of the effective circulating dose of paclitaxel. In other words, it appears that elesclomol does not so much synergize with paclitaxel, or sensitize cancer cells to paclitaxel (both are claims made by SNTA) – it rather appears to be inhibiting the metabolism of paclitaxel and driving up its effective dose, thus allowing paclitaxel a greater opportunity for cytotoxicity. This would explain both greater therapeutic action, as well as greater adverse reactions – a trend that was indeed found in the subsequent phase II trial: that study demonstrated an improvement in clinical response with elesclomol, but it also demonstrated an increase in paclitaxel-related adverse events, despite the fact that the paclitaxel dose was fixed. This, of course, is not at all consistent with SNTA’s proposed mechanism of action, in which elesclomol pushes cancer cells over their oxidative threshold, but leaves healthy cells alone – if that were indeed the case, then you would not see an increase in paclitaxel-related toxicity in the experimental group. But, in fact, there was an extremely strong signal for neutropenia (a well-known toxicity of paclitaxel) that was completely absent in the control group – again, a suggestion that elesclomol may be acting in no other way then simply ramping up the levels of circulating paclitaxel and increasing that agent’s cytotoxicity. Though it will be discussed in greater detail at a later point in this report, it is important to recognize that the dose of paclitaxel used in both the phase II and phase III trials (80 mg/m2) has been established as subtherapeutic in melanoma and is well below the dose of paclitaxel used in other cancer therapies. In fact, the phase II and phase III dose is less than half the dose used in the phase I trial (175 mg/m2), which is a standard dose of the drug and one in which you would expect to see some clinical activity. That dose was dropped for phase II and phase III, of course, because the SNTA investigators knew that the effective half-life of paclitaxel would increase in the presence of elesclomol, and an unfavorable toxicity profile would emerge – so they more than halved the dose. What you are left with for phase II and III is, in essence, a control group on an established subtherapeutic dose (80 mg/m2) being compared to an experimental group that has an agent (elesclomol) that pushes its pharmacokinetic distribution into the therapeutic range – a drug that can’t work versus a drug that might work. What you do not have is a drug that sensitizes melanoma to paclitaxel.
- The phase II trial probably underestimated the incidence of adverse events in the experimental group, as the study was an intention-to-treat analysis with a remarkably high rate of crossover. The phase II study allowed for crossover of the control group when their treatment failed, but their data was included and analyzed by their original group assignment – a standard intention-to-treat analysis. However, there was a remarkably high rate of crossover in this study: 68% of the control group crossed over into the experimental group. As a result, any of the toxicities experienced by those control group patients after crossing over to the experimental group were assigned to control. This could represent a potential obstacle in the phase III study, where there will be stronger statistical powering (an estimated 630 patients will enroll, which is almost eight times the size of the phase II study) and no crossover – and thus a potentially larger, and more likely significant, difference between adverse event rates of the two groups that will not reflect favorably on the experimental group.
- Despite claims by SNTA, the on-going phase III trial is not confirmatory: it is exploratory, and would likely be held to the same scrutiny, and possible standard of confirmation, as any exploratory phase III trial. The press conferences and releases that have flowed from SNTA management have gone to great efforts to promote their phase III trial as a confirmatory trial, and not an exploratory trial more typical of a first phase III study. Management’s rationale for this position is that, unlike most phase II trials, the phase II elesclomol study was randomized, controlled, and blinded. But having explored the role of elesclomol for melanoma in the phase II trial, the investigators concluded that it was best indicated for patients who had never received chemotherapy, and so the phase III trial was designed to include chemotherapy-naïve patients only. Consider that 56% of the phase II experimental group and 68% of the control group had received prior chemotherapy: if you exclude these patients from the analysis (as they do not apply to the phase III study), then there were only 9 chemotherapy-naïve patients in the control group, and only 23 in the experimental. A grand total of 32 patients, despite randomization and blinding, is more consistent with a pilot study, and hardly qualifies as even a hearty phase II trial. It certainly is not strong enough to qualify its subsequent study as confirmatory – in fact, the phase III trial is exploring the hypothesis that elesclomol and paclitaxel work better together in chemotherapy-naïve patients. Even if this phase III study works, and I believe this report will make my doubts on that clear, it is likely that either the FDA or the oncology academy will demand confirmation. In addition, as stated before, the phase III study will not be a crossover study – this, too, will make the evaluation of the adverse events data exploratory, and not confirmatory.
- The dose of paclitaxel (80 mg/m2) used in both the phase II and on-going phase III trials is not a standard dose in either combination or monotherapy, and by all parameters would be considered subtherapeutic. Zimpfer-Rechner et al. (15) have demonstrated that paclitaxel monotherapy for melanoma at 100 mg/m2, or in combination with carboplatin at 80 mg/m2, does not improve survival. Walker et al. (14) has demonstrated that a dose of 80 mg/m2 for melanoma has virtually no anti-tumor activity. Bedikian et al. (12) concluded that a dose of 90 mg/m2 for melanoma is “marginally active.” Meanwhile, the standard paclitaxel dose for lung cancer ranges anywhere from 120 mg/m2 to 1000 mg/m2, and the standard paclitaxel dose for ovarian and breast cancer ranges anywhere from 120 mg/m2 to 200 mg/m2. And yet, despite all this antecedent data, SNTA chose as its control paclitaxel 80 mg/m2 – in other words, the control medication at the chosen dose is a drug that more or less has no anti-tumor effect in melanoma, effectively acting as a placebo with toxicity. Paclitaxel 80 mg/m2 is a poorly chosen control, and one has to wonder whether the FDA and/or the oncology academy would even accept positive results from a phase III trial as being relevant – if elesclomol + paclitaxel is better than subtherapeutic paclitaxel alone, does that mean anything? Referring back to an earlier point in this analysis, it seems that elesclomol’s mechanism of action may be a simple increase in the circulating dose of paclitaxel, which would mean that both the phase II and phase III trials essentially pit a possibly therapeutic dose of paclitaxel (paclitaxel + elesclomol, with elesclomol pushing paclitaxel into a therapeutic range) versus a known subtherapeutic dose – hardly an impressive head-to-head study. In fact, the adverse event data from the phase II study suggests that virtually inactive paclitaxel (control) was put up against a minimally active paclitaxel: Though the PDR lists neutropenia as occurring in 90% of patients on paclitaxel, it occurred in only 6% of the phase II experimental group, and 0% of the control. Though neuropathy occurs, as a standard, in 60% of paclitaxel patients, it only occurred in 31% of the experimental and 21% of control. Anemia is listed as occurring in 78% of paclitaxel-treated patients, but in only 24% in the phase II elesclomol study. This trend would similarly apply to alopecia, vomiting and diarrhea, but the main point is this: the dose of paclitaxel chosen as control for both the phase II and phase III study is too low for any clinical effect to occur – and so low that even toxic effects hardly occur.
- Single-agent paclitaxel is not an accepted standard of cytotoxic therapy for advanced melanoma, and so the degree of significance in demonstrating superiority of an experimental regimen over a sub-standard of care is highly questionable. This point more or less reiterates the point made immediately above: is anybody going to listen to even positive data that demonstrates improvement over a clinically irrelevant protocol? Dacarbazine, though hardly a miraculous drug in the treatment of advanced melanoma, is the standard of care, with progression free survival statistics that are similar to those found in the elesclomol phase II trial. It would have been far more valid to have had dacarbazine as the control medication, or to have designed a trial that had three arms (dacarbazine, paclitaxel, and elesclomol + paclitaxel) instead of two. It also would have helped to have designed a trial where the control medication is given at a standard, therapeutic dose.
Source Material and Correspondence
- Safi Bahcall, Ph.D., Director, President and CEO, Synta Pharmaceuticals, James Barsoum, Ph.D., Senior Vice President, Research, Synta Pharmaceuticals, et al., Synta Pharmaceuticals Analyst Day, March 25, 2008.
- Tsao H. et al.: Management of cutaneous melanoma. NEJM 2004;351:998-1012.
- Miller AJ et al.: Melanoma. NEJM 2006;355:51-65.
- Berkenblit A et al.: Phase I clinical trial of STA-4783 in combination with paclitaxel in patients with refractory solid tumors. Clin Cancer Res 2007;13:584-90.
- Ramaswamy MD: Rational design of cancer-drug combinations. NEJM 2007;357:299-300.
- De Giorgi V et al.: Management of cutaneous melanoma [letters]. NEJM 2004;351:2770-71.
- Diaz-Cano SJ et al.: Molecular mechanisms in melanoma [letters]. NEJM 2006;355:11395-96.
- Murphy PM: Chemokines and the molecular basis of cancer metastasis. NEJM 2001;345:833-34..
- Gogas H et al.: Prognostic significance of autoimmunity during treatment of melanoma with interferon. NEJM 2006;354:709-18.
- Gehrmann M: Drug evaluation: STA-4783 – enhancing taxane efficacy by induction of Hsp70. Curr Opin Investig Drug 2006;7:574-80.
- Korn EL et al.: Meta-analysis of phase II cooperative group trials in metastatic stage IV melanoma to determine progression-free and overall survival benchmarks for future phase II trials. J Clin Oncol 2008;26:527-34.
- Bedikian AY et al.: Phase II evaluation of paclitaxel by short intravenous infusion in metastatic melanoma. Mel Res 2004;14:63-66..
- Rao RD et al.: Combination of paclitaxel and carboplatin as second-line therapy for patients with metastatic melanoma. Cancer 2006;106:375-82.
- Walker L et al.: Phase II trial of weekly paclitaxel in patients with advanced melanoma. Mel Res 2005;15:453-59.
- Zimpfer-Rechner C et al.: Randomized phase II study of weekly paclitaxel versus paclitaxel and carboplatin as second-line therapy in disseminated melanoma: a multicentre trial of the Dermatologic Co-operative Research Group (DeCOG). Mel Res 2003;13:531-36.
- Balch CM et al.: Final version of the American Joint Committee on cancer staging system for cutaneous melanoma. J Clin Oncol 2001;19:3635-48.
Disclosure: No positions