Statistical analysis of the inadvertently released enrollment curve for Oncothyreon's ("ONTY") lung cancer vaccine, Stimuvax, shows there is a 50%+ probability Stimuvax is working.  At $8.84/sh, ONTY's EV is $350mm, but if Stimuvax works, the company is worth ~$1.8b, resulting in a stock price of ~$39/sh, ~340% above today's close. 

Stimuvax is currently in the Phase 3 Stimulating Targeted Antigenic Responses to NSCLC ("START") trial that we estimate shows a patient survival benefit of 18 months.  For Stage III, unresectable, non-small cell lung cancer ("NSCLC") patients, the two major advances of the past decade, concurrent chemo-radiation and Tarceva as maintenance therapy, have shown patient survival benefits of 2 months and 5 months, respectively (note: the 5 month advantage of Tarceva has not been shown in a Phase 3 statistically significant trial).  An 18 month survival benefit would make Stimuvax the standard of care resulting in a few billion in annual sales. 

The enrollment curve for the START trial was inadvertently disclosed by scientists working for EMD Serono, a division of Merck KGaA ("Merck"), at an American Society of Clinical Oncology ("ASCO") conference workshop.  If you're not a healthcare investor, an enrollment curve tells you when patients were admitted to a trial.  Additional publically available data about the trial, such as the number of deaths that occurred at the first interim look allows you to make a fairly accurate estimate of the median survival for the entire study.  Knowing how long patients with Stage III, unresectable, NSCLC generally live (i.e., the placebo group) allows you to estimate the median survival for patients receiving Stimuvax. 

Our work and the analysis of a biostatistics professor we employed indicate it is likely that Stimuvax will be shown to work at the second interim look, which will occur in the next 6 months.  As it is uncommon for efficacy to be shown at an interim look, there is an asymmetric risk-reward as the typical biotech downside (i.e., the compound doesn't work) is statistically unlikely, but if the trial is stopped because Stimuvax works, ONTY will skyrocket. 

While I've spoken with oncologists who believe the scientific rationale behind Stimuvax is sound, our investment is reached not by studying the drug, but through analysis of the enrollment curve for the START trial.  This isn't about understanding the science behind a new biologic vaccine - it's a math problem.

Company description:

ONTY is a clinical-stage biotechnology company focused on the development of therapeutic products for the treatment of cancer.  The company has 3 assets:

(i)                  Stimuvax, a cancer vaccine currently in two Phase 3 clinical trials, designed to target MUC1 expressing cancer cells.  Stimuvax's Phase 3 trials are paid for by Merck, which has a worldwide license for the development, manufacture, and commercialization of Stimuvax.  The Phase 3 START trial is the focus of this write-up as it is the crux of our investment thesis.  Merck is also running the Phase 3 Stimuvax Trial in Asian NSCLC Patients: Stimulating Immune Response ("INSPIRE"), which aims to enroll 420 patients in Asian countries.  INSPIRE was initiated in December 2009 and will be concluded in ~2014.  If Stimuvax is successful, ONTY earns an effective royalty that averages ~10.5% worldwide on Merck's sales.  Before beginning the START trial, Merck entered into a Special Protocol Assessment ("SPA") with the FDA, which means that the FDA has already approved the trial's design, clinical endpoints, and statistical analyses.  The START trial was designed to detect a 6-month survival improvement over placebo with a p-value of 0.025.  The FDA has also granted Fast Track status to the investigation of Stimuvax in the treatment of NSCLC.  While the SPA and Fast Track status do not guarantee approval, the FDA has approved & reviewed the START trial and recognizes the importance of expediting the review of a treatment for NSCLC.  Lastly, given MUC1's expression in various cancer cells, success in NSCLC could translate to future applications in breast, colorectal and other cancers.

(ii)                PX-866, an inhibitor of phosphoinositide 3-kinase, which may be able to suppress tumor growth.  PX-866 is currently in 4 Phase 2 tests with initial data expected in mid-2012.  Whereas Merck pays for all the costs of Stimuvax trials, ONTY funds PX-866's 4 trials.  Our valuation ascribes no value to PX-866, but 3 prior transactions of phosphoinositide 3-kinase compounds indicate PX-866 may be worth ~$200mm (Gilead Sciences' 2011 acquisition of Calistoga Pharmaceuticals for $375mm; Roche's 2008 acquisition of Piramed Limited for $160mm; Sanofi-Aventis' 2009 partnership with Exelixis for an upfront payment of $140mm). 

(iii)               ONT-10, an improved version of Stimuvax.  In our diligence, it seems reasonable to conclude that if Stimuvax works, ONT-10 also works.  The management team has confirmed this conditional thinking.  While ONTY receives a royalty on Stimuvax, ONTY has full ownership of ONT-10 and plans to begin Phase 1 testing in ~6 months.

Pro-forma for the early May secondary offering, ONTY's capital structure:
Sh price: $8.84 (7/19/11)
FD S/O: 47.3
Mkt cap: $418mm
Net cash: $68mm
EV: $350mm

ONTY has publically stated that the May secondary offering provides sufficient liquidity until 2013 and that they will not sell additional stock until the START trial is concluded.

Stimuvax background:

Stimuvax is part of a new class of drug known as cancer immunotherapies (i.e., cancer vaccines), which utilize the body's immune system to fight against cancer cells.  Most of us know of vaccines as drugs used to help prevent infections.  These vaccines use weakened or killed viruses, bacteria, or other germs to start an immune response in the body.  Much like regular vaccines, cancer vaccines are used to train the body's immune system to recognize cancer cells as foreign, to destroy them upon contact, and to remember these cells as future targets.  Unlike regular vaccines, which prevent disease, cancer vaccines are used to fight a disease that already exists in the body and are considered active immunotherapies.  The potential for adoption of cancer immunotherapies is high given that they are less toxic and have less harmful side effects than traditional chemotherapy drugs and therefore can also be used in conjunction with other treatments to maximize a patient's survival.  While cancer vaccines have been studied for decades, only two vaccines have received FDA approval - Provenge for the treatment of prostate cancer and Yervoy for melanoma.  Prior to these drugs, at least a dozen biotech companies failed in attempts to create cancer vaccines, such as Genitope, Aphton, and LipoNova. 

In the United States, lung cancer has been the leading cause of cancer death for many years for both men and women (29% of all cancer deaths in 2007 - 31% in men and 26% in women).  According to the National Cancer Institute, there will be an estimated 221,130 new cases of lung cancer in 2011 and 156,940 deaths.  Worldwide, there are 1.6 million new cases of lung cancer each year and 1.3 million deaths, with the highest rates of death in Europe and North America.  Among the various types of lung cancer, NSCLC is by far the most prevalent, representing 85% of cases according to the American Lung Association.  Current survival rates for NSCLC are low as early detection is difficult because a persistent cough is the most common symptom, but the majority of NSCLC patients are long-term smokers who frequently have persistent coughs.  The majority of diagnosed NSCLC patients are either Stage III (~30%) or Stage IV (~35%); there is no Stage V as beyond Stage IV is death.  According to M.D. Anderson Cancer Center, "treatment outcomes in advanced or metastatic non-small-cell lung cancer remain unsatisfactory, with low long-term survival rates"; the five year survival rate for all NSCLC patients is 15%.  Given the number of NSCLC cases and the need for more effective NSCLC treatment, the market opportunity for Stimuvax is substantial (more in "Valuation").

The current Phase 3 START trial is the result of a prior Phase 2b and a Phase 1/2 comparability trial, which both showed promising results.  In late 2002, enrollment was completed for a 171 patient Phase 2b trial of Stimuvax in Stages IIIB and IV, unresectable, NSCLC patients.  The median survival for patients that received Stimuvax was 4.2 months longer than for patients that received placebo.  The 3-year survival rate was 31% for patients that received Stimuvax and 17% for patients that received placebo (p=0.035).  Moreover, in the pre-stratified subset of locoregional Stage IIIB patients, the median survival of Stimuvax patients was 17.3 months longer (30.6 months vs. 13.3 months for patients that received placebo).  The 3-year survival rate of locoregional Stage IIIB patients that received Stimuvax was 49% for patients that received Stimuvax and 27% for patients that received placebo (p=0.070).  Additional detail from a report released earlier this month can be found at: http://www.scribd.com/doc/60222639?secret_password=8ddsekl334njw7axta3.

In Q2 2005, a small Phase 1/2 comparability trial was run because manufacturing changes were incorporated into the production of Stimuvax.  The results of this trial indicated that the revised formulation is equivalent to the formulation used in the Phase 2b trial.  While there was no placebo group, in mid-2008 Merck reported that the two-year survival for patients in the Phase 1/2 comparability trial was 64%.  Furthermore, while it is a small group of just 22 patients, 11 (50%) of the patients are alive today, which is approximately 6 years later.

Extrapolating from the Phase 2b and Phase 1/2 comparability trials that Stimuvax will have a successful Phase 3 trial is unwise.   There have been many Phase 2 trials with exciting results that were not able to show a statistically significant benefit in a Phase 3 trial.  Our investment case gives no credit to the Phase 2b results.  This information is presented so that readers are aware that there is a logical basis for the Phase 3 trial.

Enrollment curve:

The enrollment curve for the START trial can be found at:
http://university.asco.org/dgtfiles/ClinTrials/Symposium/3GlobalGaumond.pdf
EMD Serono is running the Stimuvax Phase 3 trial and gave a presentation explaining how to enroll large numbers of cancer patients in a drug study.  On page 19 (http://university.asco.org/dgtfiles/ClinTrials/Track3.pdf) it is noted that "only 3% of adults with cancer participate in oncology clinical trials."  I have a family member whose oncologist has recommend that she take part in numerous studies, but she never has despite the studies providing her best standard of care + a new drug.  In other words, she would lose nothing by participating, but like many cancer patients she's afraid to take a chance on a new treatment. 

When I asked ONTY CEO Bob Kirkman if he was aware that the START trial enrollment curve was public, he noted that while their intentions were good (i.e., educating other scientists on recruitment and retention techniques), EMD Serono's scientists published the enrollment curve without approval from ONTY or Merck's corporate office.  So the scientists inadvertently disclosed this information because they didn't realize investors could use the enrollment curve to estimate the survival benefit of Stimuvax.  And they certainly didn't realize that it would show that Stimuvax is likely working.  In fact, until an investor pointed this out to CEO Kirkman, he was unaware that the enrollment curve was public.  When he called Merck, they were also unaware. 

If a new drug study has the end point of death (vs. an end point such as improved quality of life), it often has at least one interim look given ethical concerns.  The ethical concern is two-fold: (i) if it can be statistically proven that the drug provides no benefit, the trial is stopped as it's unethical to give patients false hope (i.e., trial stopped for futility); and (ii) more importantly, if it can be statistically proven that the drug provides a survival benefit, the trial is stopped as it's unethical to keep patients on the placebo (i.e., trial stopped for efficacy).  Generally, the announcement of an interim look is not valuable as the company simply states, "we hit the interim look and the independent data monitoring committee determined that the trial should continue." (see ONTY's one sentence 8-K for their first interim look: http://files.shareholder.com/downloads/BIOM/1262956881x6311582xS950123-10-115837/1412067/filing.pdf).  For an investor, nothing can be determined because the only data point known is that a certain number of deaths have occurred, but because it takes years to fully enroll patients (e.g., the START trial's first patient enrolled in February 2007), you have no idea of the time between patients receiving the drug/placebo and their death.  However, knowing the enrollment curve, the timing of the first interim look, the number of deaths that triggered the first interim look, and having an estimate for the median survival of the placebo group allows us to estimate the median survival for patients receiving Stimuvax. 

We hired a professor of biostatistics and her report can be read at http://www.scribd.com/doc/60264073/Projection-of-START-Trial-Outcome-July-15-2011?secret_password=15dm852c2xizql7gpx96.  Her work is more sophisticated than this explanation, but of the 1,320 patients enrolled in this trial, 440 (one-third) were randomly assigned to the placebo group and 880 (two-thirds) to the Stimuvax group.  The trial had a first interim look at 353 deaths and because we know the amount of time between patients enrolling in the trial and the first interim look, we can estimate the hazard rate for all patients in the trial.  The hazard rate gives an average probability of a patient dying during any month.  As 2 patients received Stimuvax for each 1 patient that received placebo, knowing the hazard rate for all patients allows us to solve for the hazard rate for Stimuvax patients based on a given hazard rate for placebo patients.  We can then look for factors that create uncertainty and sensitize around these factors to determine the robustness of the analysis. 

Reading the entire report is a worthwhile use of time, particularly as it's written in a manner that makes it easy to understand, but the conclusions of the professor of biostatistics are:

(i)                  "My educated guess is that the true hazard rates lie in the range of" a median survival advantage of 16.3-20.8 months (~18 months);

(ii)                "The p-value obtained at the first interim look strongly suggested efficacy but the committee was cautious and wanted to see more data before stopping the trial"; and

(iii)               "Based on my work presented here (and additional background work, analyses and simulations that I have not presented here), there appears to be a substantial probability that the trial will be stopped for efficacy at the second interim look."

In short, the professor's analysis shows that Stimuvax provides a median survival advantage far greater than the 6 months required by the SPA and that this will be statistically proven at the second interim look.  As the agreed upon clinical endpoints will be satisfied, the trial will be halted and Merck/ONTY will request that the FDA approve Stimuvax for NSCLC patients.

The professor of biostatistics has been a member of many independent data monitoring committees for cancer studies.  In conversations to discuss her analysis, she stated that she strongly believes there is a 50%+ probability that efficacy will be reached at the second interim look.  I asked if she was willing to defer her compensation until the second interim look and multiply her bill (nearly $10,000) by the change in the stock price from the day of her report (July 15, 2011) to the day after the second interim look.  She accepted. 

This professor's confidence doesn't prove that Stimuvax works, but it does show that someone with significant similar experience, after more than 50 hours analyzing the START trial believes that the probability that efficacy will be reached at the second interim look is 50%+.  Prior to agreeing to pay her by the hour, I asked for an estimate of the number of hours likely to be billed and she estimated 5 and predicted that this would be "an exercise in explaining that nothing is yet known."  Upon concluding there is a high probability that Stimuvax works, she spent dozens of hours trying to poke holes in her work.  

According to ONTY CEO Bob Kirkman, on May 31^st, Merck informed him that they felt Stimuvax had a good chance of showing efficacy at the second interim look and decided to ramp-up Chemistry, Manufacturing and Controls ("CMC").  For non-healthcare investors, CMC ramp-up involves chemists/biologists working to increase production of a compound from the small amounts needed in a clinical trial to the large amounts needed following FDA approval.  This generally involves trying to reduce the amount of time and number of steps required to produce the compound.  While there are healthcare companies that ramped their CMC before trial conclusion only to see their compound miss its endpoint, it seems likely that Merck has completed statistical analysis similar to ours (note: Merck is in their blackout period until July 27^th so I haven't been able to ask them). 

Valuation:

Valuing any company on a hypothetical market size is an imprecise exercise, but we believe there are multiple ways to triangulate ONTY's fair value assuming the START trial is successful.  The most simplistic method is to use Merck's estimated worldwide market size for Stimuvax, which ONTY CEO Kirkman told us was $3b-$4b.  At ONTY's ~10.5% average worldwide royalty rate and the mid-point of Merck's market estimate of $3.5b, ONTY's annual Stimuvax royalty is $366mm.  A 5x multiple, in-line with other growing and long-lasting royalty payments, values ONTY at ~$1.8b, or ~$39/sh.  This ascribes zero value to ONTY's other assets.  Our valuation and a build-up to the $3-4b market potential: http://www.scribd.com/doc/60377061?secret_password=1ioyzaa6zqbubx6ywby8

Another method to arrive at a fair value for ONTY is to calculate the implied value of OSIP's lung cancer drug Tarceva.  OSIP was acquired for $4.0b by Astellas Pharma in June 2010 and OSIP management estimated that the value for OSIP's non-Tarceva assets was $1.3b, implying that Tarceva was worth $2.7b, or ~7x OSIP's annual Tarceva royalty.  Assuming that Stimuvax is also worth $2.7b, but using a 5x multiple to be conservative as 7x is a takeout valuation, Stimuvax is worth ~$1.9b, or ~$40/sh.  Our relative valuation:   http://www.scribd.com/doc/60377851?secret_password=czygighm8p577lc9wk5

DNDN and BMY are the only two companies with successful cancer vaccines.  As BMY's market value includes many other compounds, it's only possible to extrapolate the value of a successful cancer vaccine from DNDN.  At the current time, the market ascribes a value of $5.7b to DNDN's prostate cancer vaccine, Provenge.  While a pie-in-the-sky scenario, we do think it's possible the market ascribes a similar value to the combination of Stimuvax and ONT-10.  While Stimuvax is ~8 years further along in development than ONT-10, ONTY is entitled to only a royalty on Stimuvax sales.  However, similar to DNDN's Provenge, ONTY has full economics for ONT-10.  And while it's certainly not conservative to assume that a pre-Phase I drug will be successful, because ONT-10 is an improved version of Stimuvax (both are vaccines designed to target MUC1), Stimuvax success would give a strong indication that ONT-10 is also likely to be successful.  Both DNDN's and ONTY's vaccines target mid-to-late stage cancers, provide a meaningful survival benefit with good tolerability, and the overall market for NSCLC is approximately equal to prostate cancer (we estimate 75k worldwide annual patients for Stimuvax; Goldman Sachs' 2015 expectation for Provenge is 88k worldwide annual patients).  While it's pie-in-the-sky, if ONTY is valued similarly to DNDN, the stock is worth ~$120/sh.

Our estimated market potential for Stimuvax is above sellside expectations.  While it's uncommon for the sellside to understate a market potential or a price target, it doesn't appear that any of the analysts have carefully forecast the market potential.  For example, the most recent initiation is from Cowen, which forecasts Stimuvax's market penetration at 10-50% following approval.  This is illogical as approval for Stage III, unresectable, NSCLC, requires a 6+ month survival benefit, which is significantly better than any alternative therapy.  Combined with the mild side effects, Stimuvax would immediately become the standard of care for nearly 100% of patients.  Additionally, sellside price estimates for both DNDN and OSIP were wildly inaccurate.  According to Bloomberg, 7 analysts had an average price target of $1.00/sh for DNDN in early April 2009 before the trial announcement and within two weeks the average price target moved to $34.90/sh.  Separately, 10 analysts had a price target of $39.50/sh on OSIP at the beginning of April 2004 and by the end of the month, subsequent to the announcement of a successful trial, the same 10 analysts had an average price target of $96.85/sh.

We ascribe no value to ONT-10, which can be thought of as "Stimuvax v2.0," but it is quite likely that if Stimuvax works, ONT-10 works and unlike Stimuvax, ONTY fully owns ONT-10.  In discussing this dynamic with ONTY CEO Kirkman, he stated that if Stimuvax works, it would be very unlikely for Merck to leave ONTY a standalone company because Stimuvax will be a multi-billion dollar drug and ONT-10 will be a huge threat to it.  Additionally, we are ascribing zero value for the potential for Stimuvax to work in other solid-tumor cancers.  Assuming success in NSCLC, Merck will run trials for Stimuvax in multiple other cancers and ONTY is entitled to royalties on all indications.

Our probability-weighted fair value is $21.20.  This assumes ONTY is worth $39/sh if Stimuvax is successful and a 50% probability that efficacy is hit at the second interim look of the START trial.  While the professor's report shows that the chance of concluding that Stimuvax is futile at the second interim look is quite remote, we assume a 10% probability to be conservative and that ONTY would trade at $1/sh.  This implies a 40% chance that the second interim look makes no conclusion about efficacy or futility and the trial is continued to final data.  While we believe market expectations for efficacy at the second interim look are low, we conservatively assume that ONTY would decline to the price of its recent secondary offering, $4.00/sh.  This math:  http://www.scribd.com/doc/60379244?secret_password=152dac7r3gcl7s8qn2t8

Risks: 

As our investment is based on analysis of the enrollment curve, an incorrect assumption would cause an incorrect conclusion.  Therefore, the risks of this investment are: (i) the enrollment curve is mislabeled and does not accurately represent the START trial; (ii) our extrapolation of patient enrollment from the curve is incorrect; (iii) the biostatistics professor's analysis is flawed; and (iv) the placebo median survival is longer than 30 months. 

ONTY CEO Bob Kirkman confirmed to us that the START enrollment curve from ASCO is accurate.  So this isn't causing an error in our analysis. 

The ASCO poster gives the enrollment curve, not the exact date of each patient's enrollment.  Therefore, we must extrapolate the date of each patient's enrollment from the curve and while it's possible to zoom in to see a high level of detail, this does require making estimates.  We and the professor tried to be conservative in our interpretation of the enrollment curve - e.g., the professor assumed the first patient enrolled in May 2007, but ONTY issued a press release in Feb 2007 after the first patient enrolled.  Therefore, it is highly unlikely that our extrapolation of patient enrollment from the curve is causing an incorrect conclusion, particularly as it is Approach 2 that the professor considers most reasonable and in Approach 2 she assumes the average time for each patient in the trial is 2 months less than her best guess extrapolation of the enrollment curve.

While the scope of the biostatistics professor's work was not limited to only analyzing the enrollment curve, for this task alone she needed to make a number of calculations, apply statistical methods, and properly program simulations into software.  While her background as a member of many independent data monitoring committees seems ideally suited to analyzing the enrollment curve, everyone makes mistakes.  Therefore, she shared her work with other professors to ensure that she hadn't made an error before giving us her final report.  And we shared her work with a statistician to confirm that her methods and conclusions were logical.  With the exception of minor assumptions he would have made differently, which he noted would not change the output, the statistician found no mistakes.  Also, we shared this investment thesis with an oncologist who spent two years in a lab unsuccessfully attempting to develop anti-cancer vaccines and is quite bearish towards them.  We didn't provide him the professor's report.  Instead, he created his own statistical analysis.  His conclusion is that the assumptions required to believe that Stimuvax is no better than placebo are far more unreasonable than the assumptions required to believe it is working (i.e., there is a greater than 50% chance that Stimuvax is working).

Finally, in an attempt to learn from the past, we looked for prior instances of a public enrollment curve that investors incorrectly interpreted.  We were unable to find a situation that included the inadvertent release of the enrollment curve and an interim look that allowed investors to run statistical analysis to determine the likelihood the compound works.  A similar, but not comparable situation is NVLT's Phase 3 trial of NOV-002 for Stage IIIb and IV NSCLC patients (http://www.scribd.com/doc/60304651?secret_password=1o8jp6e3jdc3mj3r6in4).  This Phase 3 trial was designed to detect a 2.5-month survival improvement of NOV-002 + chemotherapy (expected median survival of 12.5 months) over standalone chemotherapy (expected median survival of 10.0 months).  Patients were enrolled from November 2006 - March 2008 (full enrollment curve was not known by investors) and the trial design did not include an interim look, only a final analysis upon reaching 725 patient deaths, 86.3% of the planned accrual of 840 patients.  This trial design is quite different from the START trial, which has a first interim look at 353 deaths (26.7% of the planned accrual of 1,320 patients), a second interim look at 529 deaths (40.1% of the planned accrual), and final analysis at 705 deaths (53.4% of the planned accrual).  For NOV-002's study, the median survival was hit well before an announcement of the total number of deaths.  There is almost no predictive value in knowing that the readout has or has not occurred because you are waiting for the "tail" deaths and not "median" deaths.  Having the enrollment curve for START and knowing the timing of the first interim look allows us to estimate the median survival for Stimuvax patients. 

Given our inability to find flaws in the professor's work, it doesn't seem plausible that her statistical analysis is incorrect. 

The "Conclusions" section of the professor's report states "Approach 2 entails the most reasonable assumptions regarding hazard rates...my educated guess is that the true hazard rates lie in the range of Approach 2 Scenarios 8-10."  The midpoint of her educated guess is Scenario 9 with a median survival for the placebo group of 24.8 months and median survival for the Stimuvax group of 43.3 months, resulting in an 18.5 month survival benefit.  But what if her educated guess is wrong?  What assumptions do we need to make about the median survival for the placebo group for Stimuvax to no longer show a 6 month survival benefit?  As shown in Approach 2 Scenario 4, the median survival for the placebo group is 30.1 months and the median survival for the Stimuvax group is 37.5 months, resulting in a 7.4 month survival benefit.  But below Scenario 4, the survival benefit is less than 6 months.  Therefore, what is the likelihood that the placebo median survival is longer than 30 months? 

While it's unfortunate that NSCLC is a very common cancer - it kills 100k+/yr - it allows for a fairly robust review, which included conversations with 10 oncologists, a review of dozens of prior trials that yielded 7 comparable studies, and consideration of the factors impacting median survival since trial design.  Detail is presented below that indicates it is not likely that the placebo median survival is longer than 30 months.

We surveyed 10 oncologists to determine if any are seeing Stage III, unresectable, NSCLC median survival north of 30 months.  No one is: http://www.scribd.com/doc/60380721?secret_password=26xzakdlrkhkknzlm8jo

The 2010 median survival of 18 months is lower than the 20 months estimated in the START trial design because patients in the START trial must respond positively to chemo-radiation.  Approximately 15-20% of patients do not respond to chemo-radiation and are excluded from the START trial.  These are the sickest patients and likely to die in ~6 months.  If we assume that 20% of the oncologists' patients have a median survival of 6 months, this implies a median survival of 21 months for the 80% of the population that is comparable to the START trial. 

9 of the 10 doctors surveyed work at hospitals in the U.S., but ONTY has stated that only 1/3 of the patients in the START trial are from North America.  The median survival at a U.S. hospital is likely far higher than a hospital in Eastern Europe - the best cancer center in Poland isn't a very good hospital by U.S. standards.  Therefore, the skew in our survey towards U.S. hospitals most likely over-states the expected placebo group median survival.

Below is a summary of 7 Phase 2 and Phase 3 clinical trials with Stage IIIA/IIIB NSCLC patients (sorted in descending order from the date the study began) that were used to estimate an overall expected median survival rate of 22.2 months for patients in the START placebo group.  Our detailed study analysis: http://www.scribd.com/doc/60320507?secret_password=2a3e06zek7m73paaa1fo

(i)                  Phase III Erlotinib (i.e., Tarceva) - May 2005:  243 Stage IIIA/IIIB patients were randomized with 122 patients receiving the drug and 121 patients receiving the placebo.  The median survival rates for the drug and placebo groups were 30.4 months and 25.1 months, respectively.  This trial failed to meet its enrollment goals and could not be completed.  Aside from the trial not fully enrolling, it was unusual as patients were randomized before receiving initial chemo-radiation therapy resulting in median survival rates approximately 4 months longer than we would expect in the START trial as patients are randomized before receiving initial Stimuvax/placebo therapy.  To be conservative we used the median survival for the drug group less 4 months, or 26.4 months.  Supporting timeline:

(http://www.itocd.com/fileadmin/presentations2010/Mornex,%20Francoise.pdf - slide 9).

(ii)                Phase II Sahlgrenska University Hospital Study - June 2002: 151 Stage IIIA/IIIB patients were randomized with no placebo group (often phase II studies do not have a placebo group).  The median survival rate was 17.8 months.

(iii)               Phase III Docetaxel (HOG LUN 1-24) - February 2002: 147 Stage IIIA/IIIB patients were randomized with 73 patients receiving the drug and 74 patients receiving the placebo.  The median survival rates for the drug and placebo groups were 21.2 months and 23.2 months, respectively.  This study failed and was closed early due to a slower than expected accrual rate and a higher than expected dropout rate.  Given the relatively small sample size, we use an average of the median survival rates of the two groups in our study analysis, or 22.2 months.

(iv)              Phase III Iressa (SWOG S0023) - June 2001: 243 Stage IIIA/IIIB patients were randomized with 118 patients receiving the drug and 125 patients receiving the placebo.  The median survival rates were 23.0 and 35.0 months for the drug and placebo groups, respectively.  This trial failed and is the only study we've seen in which the placebo arm lived so much longer than the drug arm that it was statistically significant.  While the 35.0 month median survival for the placebo arm is alarming, it is not comparable to the START trial.  This trial required patients to respond favorably to chemo-radiation and subsequent consolidation chemo before being randomized (for reference, most other trials including START, only require patients respond favorably to chemo-radiation before starting the drug/placebo).  Interestingly, it is possible to mathematically solve for the median survival rate of the group comparable to the START trial, which is 23.3 months and included in our study analysis.  Our math:

http://www.scribd.com/doc/60312102?secret_password=2b5wupabh8967d90dyp6

(v)                Phase II ASCO - "A California Cancer Consortium Phase II Trial" (article published in January 2001): 34 Stage IIIA/IIIB patients were randomized with no placebo group.  The median survival was 17.0 months.

(vi)              Phase II SWOG S9504 - October 1996:  83 Stage IIIB patients were randomized with no placebo group.  The median survival was 26.0 months in this trial, but the subsequent SWOG S0023 Phase III trial did not confirm this surprisingly high median survival.  We have included the 26.0 months median survival in our study analysis to be conservative.

(vii)             Phase III RTOG 9410 - July 1994: 595 Stage II/IIIA/IIIB patients were randomized with 201 patients receiving the drug and 201 and 193 patients receiving the placebo (two different placebo arms).  The median survival rates for the drug and the two placebo groups were 17.0 months, 14.6 months, and 15.2 months, respectively.  To be conservative we use the drug arm median survival of 17.0 months in our study analysis.

The START trial was designed in late 2006 with the assumption of a 20 month placebo group median survival.  According to ONTY, in late 2006, it was estimated that placebo group median survival was 17 months, but because of estimated improvements in median survival, an additional 3 months were added.  Since the mid-1980s, when placebo group median survival was 10-12 months, (http://www.scribd.com/doc/60311526?secret_password=2is7dh2sxk3szqcxabwy - see page 2, middle slide) Stage III NSCLC survival rates have increased ~0.5 months/yr.  The addition of an estimated 3 month survival improvement over the course of the ~6 years of the START trial seems appropriate.  But, have there been advances in care resulting in a survival benefit of more than 3 months?  Conversations with oncologists reveal 3 factors that have increased median survival for Stage III, unresectable, NSCLC patients:

(i)                  Improved diagnostic tests, specifically PET scans, that allow for better staging.  Most frequently, the result is previously classified Stage III patients are up-staged to Stage IV due to detection of distant metastasis.  Not including sicker, Stage IV patients results in an overall improvement in Stage III median survival relative to prior studies.  However, in 2005, the year before START was designed, 1.1mm PET studies were performed at 1,725 sites in the U.S.A. (http://www.scribd.com/doc/60313435?secret_password=1o916oadckd15dofwt1y).  To put this in perspective, in 2005 there were 1,402 Gap stores in North America.  With PET sites more ubiquitous than Gap stores, it's hard to imagine that PET studies result in more than a 1-2 month improvement in median survival since trial design;

(ii)                Studies show that concurrent chemo-radiation yields better outcomes than sequential chemo-radiation.  Offsetting this potential benefit, these studies were concluded before the trial was designed and should be mostly captured in the 3 month survival benefit assumed in the START trial, which makes it difficult to believe that concurrent chemo-radiation results in more than a 1-2 month improvement in median survival since trial design; and

(iii)               Tarceva, which was approved by the FDA in late 2004 for Stage IV NSCLC patients, and despite failing a trial for Stage III NSCLC patients and skepticism in the oncologist community about the benefits for Stage III patients, seems widely used for Stage III patients.  Therefore, it's possible that the 5 month median survival benefit noted above in the Phase III Erlotinib trial is incremental to the 3 month survival benefit assumed in trial design.  However, it's important to note that Tarceva works well for patients with EGFR mutation-positive NSCLC while doing nothing for EGFR mutation-negative NSCLC.  As sex and ethnicity have a significant impact, it's difficult to say with precision the percentage of EGFR mutation-positive NSCLC patients, but it's a relatively small minority.  And conversations with oncologists reveal that for this small minority, they would be very unlikely to enroll a patient in the START trial as it's not in the patient's best interest to enroll in a Phase III study with uncertain benefit as Tarceva works well for EGFR mutation-positive NSCLC patients.  Therefore, it seems unlikely that the full 5 month benefit is being realized in the START placebo group.

Of the 10 oncologists we spoke with, the highest estimate of the combined benefit of these factors was 4 months.  If we add a 4 month improvement to the estimated 20 month median survival, our estimate is that the START trial placebo group median survival is 24 months. 

Discussions with 10 oncologists, a review of 7 comparable studies, and consideration of factors impacting median survival since study design averages to a placebo group median survival of 22.4 months.  If a cushion of 6 months is added resulting in an assumed placebo group median survival of 28.4 months, the professor's analysis indicates Stimuvax patients will still show a survival benefit of 10 months relative to the placebo group (see Approach 2 Scenario 5).  While it's not a perfect comparison because the trial design assumed a placebo median survival of 20 months, START's p-value of 0.025, means that if Stimuvax is shown to have a survival benefit of 6 months, there is a 2.5% probability this is because of chance and a 97.5% probability it is because of treatment effect.  Because there are 440 patients in the placebo group, which is towards the higher end of most Phase 3 oncology clinical trials, we think the probability of the placebo group living 6 months longer than expected is low.  Therefore, our assumption above of 28.4 months, which includes a 6 month cushion, is conservative and still shows a Stimuvax treatment effect of greater than 6 months. 

Therefore, the probability the placebo group has a median survival longer than 30 months rendering Stimuvax ineffective is low, which supports our thesis that the probability that Stimuvax is working is much higher than implied by the current share price.

Why didn't Stimuvax show efficacy at the first interim look?:

If the survival advantage is close to 18 months, efficacy was likely hit at the first interim look.  Therefore, it's important to address why the START trial wasn't halted at the first interim look.  First, as noted at the bottom of page 6 of the professor's report, the independent data monitoring committee may have decided to continue the trial at the first interim look because they are waiting for the data to mature.  Or in the case of the START trial, because Stimuvax is working better than anticipated, the explanatory value of the Stimuvax arm was low at the first interim look and it was not possible to determine the difference in median survival.  While an unrealistic example, if no one in the Stimuvax group died prior to the first interim look, it would not have been possible to determine the median survival of the Stimuvax group.  For example, Approach 2 Scenario 9 estimates a 97% probability that the p-value was below the efficacy threshold, but only a 19% probability that a 6 month median survival difference could be determined at the first interim look.  Additionally, because of a clinical hold from March-June 2010, 156 patients that didn't receive proper dosing were included in the first interim look which may have weakened the survival advantage for Stimuvax at the first interim look.  These 156 patients will be excluded at the second interim look (the FDA agreed to this following the first interim look), which may allow Stimuvax to show a larger survival advantage.

What do others know?:

CEO Kirkman told me that "2 or 3 other funds have done the same analysis and come to the same conclusion" as presented in this write-up, which explains the YTD move in the stock (ONTY also completed a secondary offering, which was an overhang on the stock).  While it would have been great to be the first to figure this out, if we're correct that the stock is worth ~$39/sh, it's okay to be a bit late to the party. 

I've asked sellside analysts for an explanation of the move in the share price and they're unaware of the enrollment curve (see Wedbush note from June 22, 2011).  I dropped strong hints to one that the enrollment curve is public and he didn't grasp that it could be used to estimate the median survival for patients receiving Stimuvax.  Also, it doesn't appear that journalists are aware as a July 11, 2011 article from biotech journalist Adam Feuerstein titled "Oncothyreon: Deep Dive Into Stimuvax" (http://www.thestreet.com/story/11179526/1/oncothyreon-deep-dive-into-stimuvax.html) includes no reference to the enrollment curve. 

Short interest has moved up significantly as the share price has increased:
5/13/11 - 1.7mm shares sold short; ONTY: $5.91
5/31/11 - 2.3mm shares sold short (+36.2%); ONTY: $6.47 (+9.5%)
6/15/11 - 3.3mm shares sold short (+39.9%); ONTY: $7.38 (+14.1%)
6/30/11 - 6.7mm shares sold short (+104.7%); ONTY: $9.19 (+24.5%)

This seems indicative of investors unaware the enrollment curve is public, and the statistical analysis implying a 50%+ chance that efficacy is hit at the second interim look. 

Summary:

In most Phase 3 biologic trials, the chance of success is less than 10% (probably ~5%).  While two immunotherapy drugs, DNDN's Provenge and BMY's Yervoy, have received FDA approval, a quick search reveals more than one dozen that have failed.  Given this miserable track record, it's not surprising that the market prices in a fairly low probability that Stimuvax works.  But our statistical analysis shows the chance of success is 50%+ for Stimuvax, which creates a very asymmetric risk-reward.  Additionally, a 50%+ chance of success implies that the current price should be $21.20/sh.

It would be wonderful to have a dozen similar risk-rewards and create a portfolio of these investments, but we're not aware of other biotech investment opportunities with similar investment profiles.  So we have to play the one hand we're dealt.  Therefore, being highly certain that our analysis is complete & correct before the second interim result is critical.  So we've posted this to VIC as we'd like members to try to poke holes in our thesis by asking questions in the message section.  If questions are asked that we've already diligenced, but not discussed in this write-up, we will happily share our work in the message section.  Ideally, questions will be posed that we haven't answered so that we can dig-in to make our analysis more complete - if this is the case, please give us a few days to respond. 

Additional information:

ONTY May 2011 investor presentation: http://www.scribd.com/doc/60223374?secret_password=y7yejagobjuxpyonnm5
CEO & Pres Bob Kirkman, M.D.: [email protected], (206) 801-2110
CFO Julie Eastland: [email protected], (206) 801-2112

Ideally, no near-term catalyst as time is on your side - the longer it takes to announce the second interim look, the longer Stimuvax patients are living and the higher the probability that Stimuvax works.