# Oncothyreon: Analyzing Trial Data and Concluding Likely Success

Oncothyreon (ONTY) is a controversial stock. There has been a great deal of debate regarding the likely outcome of the Phase III START trial for the cancer vaccine Stimuvax. While a number of items are in serious dispute, there is one conclusion on which both the bulls and bears can agree: the median overall survival for the entire trial population at the time of the first interim look was approximately 33 months.

Based on this point of agreement, this article will explore the likely outcome of the START trial by examining the record of past trials in this patient population and the implications of these trials on START’s null hypothesis, which is that there is no statistical difference in overall survival between the control and treatment arms of the trial. By so doing, we will show that the evidence overwhelmingly favors rejecting the null hypothesis and that the START trial is likely to be successful.

The START trial is randomized 2:1 between the treatment arm and the control arm. It is designed to determine with 90% power whether or not Stimuvax (the treatment) results in a 6-month or greater improvement in median overall survival relative to the control arm. With the aggregate population currently estimated to show a 33-month median overall survival, a 6-month improvement can be seen if the control population is living 29-months or less while the treatment population lives 35-months or more. Mathematically, this would be represented as (1/3) * 29 + (2/3) * 35 = 33. With this in mind, we can examine the probability that the control population is living at least 29 months.

Before examining any scientific literature, it is important to understand the enrollment criteria and timing for the START trial so that we can be sure to compare trials on a like basis. The most important enrollment criteria for the START trial are:

(1) stage IIIa or stage IIIb unresectable NSCLC

(2) stable disease or better after completion of chemoradiation

(3) ECOG score of 0 or 1

(4) no plural effusions

(5) patients can have either concurrent or sequential chemoradiation before trial enrollment

While there are a myriad of inclusion and exclusion criteria, the five listed above are the most important for determining comparable trials. A full list of the criteria can be seen here.

In addition to understanding the enrollment criteria, it is also important to know the timing for the START trial. Randomization into the START trial occurs between four and twelve weeks after the conclusion of chemoradiation (i.e. one to three months after chemoradiation). This means that patients in the START trial will have been getting treated for NSCLC for roughly four months by the time they are randomized into the trial: about two months of waiting after about two months of chemoradiation (the total chemoradiation time will depend on whether it is sequential or concurrent). Because there is a four-month delay between patients starting chemoradiation and being randomized in the START trial, we estimate that roughly 25% of all stage IIIa and IIIb patients who begin chemoradiation would ultimately be ineligible for the START trial.

Now that we understand what needs to be believed for the null hypothesis (29-months of survival), who gets enrolled in the START trial, and when they are enrolled, we can examine the eight Phase III and Phase II trials that we believe are most similar to the START trial and what their implied equivalent lifespan is to the START trial. From most recent to oldest, these trials are:

1) The Lu Trial: Lu was a phase III trial published in 2010. It enrolled Stage IIIa and Stage IIIb patients with an ECOG of 0-1. Malignant plural effusions were excluded though cytologically negative plural effusions were allowed. Median overall survival in the control arm was 15.6 months and the treatment arm was 14.4 months for about 15 months across the entire population. Median survival was measured from randomization, which took place before initial chemoradiation. This was a first-line trial so we need to adjust for the roughly 25% of patients that would be ineligible for START and dropped prior to START randomization and the four month delay that would be seen in the randomization date for START relative to the Lu randomization date. We also need to adjust because the method of staging may not have required imaging and because of the presence of cytologically negative plural effusions. In aggregate, we would expect this to equate to an 18 month overall survival in the START population.

2) The Bastos Trial: Bastos was a Phase II trial published in 2010. It enrolled Stage IIIa and stage IIIb patients with an ECOG of 0-1. Both malignant pleural and pericardial effusions were excluded. Median overall survival was 14.8 months. Median survival was measured from randomization, which took place before initial chemoradiation. This was a first-line trial so we need to adjust for the roughly 25% of patients that would be ineligible for START and dropped prior to START randomization and the four month delay that would be seen in the randomization date for START relative to the Bastos randomization date. Staging was done with CT or PET scans, so no other adjustments need to be made. In aggregate, we would expect this to equate to a 15 month overall survival in the START population.

3) The Movsas Trial: Movsas was a Phase II trial published in 2010. It enrolled Stage IIIa and stage IIIb patients with an ECOG of 0-1. Both malignant pleural and pericardial effusions were excluded. Median overall survival in the control was 16.1 months from the point of randomization. Randomization in this trial took place after chemoradiation and all patients were staged with advanced imaging (CT, MRI, or bone scan), so no additional adjustments need to be made. Note that enrollment took place before chemoradation and that 78% of the patients enrolled were fit to be randomized (i.e. met ECOG score of 0-1 and restaging in week 10 showed no evidence of recurrence or progression). This is consistent with our prior 75% estimate. In aggregate, we would expect this to equate to a 16 month overall survival in the START population.

4) The Bradley Trial: Bradley was a Phase II trial published in 2010. It enrolled unresectable NSCLC patients with stage I-IIIb but it stratified the group, so we can analyze just the IIIa and IIIb population. Patients had a Zubrod PS of 0-1 (like an ECOG 0-1) and were staged with advanced imaging (CT, MRI, or Bone Scan). Supraclavicular lymph node metastasis, pleural or pericardial effusions were all excluded so the patients were a bit healthier than in START. Median overall survival in the Phase IIIa and Phase IIIb subset was 21.6 months. Median survival was measured from randomization, which took place before initial chemoradiation. This was a first-line trial so we need to adjust for the roughly 25% of patients that would be ineligible for START and dropped prior to START randomization and the four month delay that would be seen in the randomization date for START relative to the Bradley randomization date. We also need to adjust because of the stricter exclusion of patients with supraclavicular lymph node metastasis. In aggregate, we would expect this to equate to a 21 month overall survival in the START population.

5) The Garrido Trial: Garrido was a Phase II trial published in 2009. It enrolled Stage IIIa and stage IIIb patients with an ECOG of 0-1. The two separate treatment arms had median overall survivals of 13.1 and 14.7 months. All plural effusions were excluded and patients were staged with CT or bone scans. Median survival was measured from randomization, which took place before initial chemoradiation. This was a first-line trial so we need to adjust for the roughly 25% of patients that would be ineligible for START and dropped prior to START randomization and the four month delay that would be seen in the randomization date for START relative to the Garrido randomization date. In aggregate, we would expect this to equate to a 14 month overall survival in the START population.

6) The Kelly Trial: Kelly was a phase III trial published in 2008. It enrolled Stage IIIa and stage IIIb patients with an ECOG of 0-1. Patients were staged with a CT scan and both pleural and pericardial effusions were excluded. The control arm had a median overall survival of 35 months and this trial is therefore the most frequently cited trial by bears on ONTY as to why we might expect the control arm to live in excess of 29 months. We covered this trial in detail in our original post, which can be found here. In short, the Kelly trial went through two separate prequalification chemoradiation stages whereas the START trial is only going through one stage of chemoradiation. As a result, the Kelly trial eliminated the 57% of the total patient population, whereas we expect START to eliminate only about 25%. This is backed up because the first round of chemoradiation in the Kelly trial only eliminated 25% of the participants. Normalizing the Kelly trial to the START trial results in an expected median overall survival in the comparable population of about 22 months.

7) The Hanna Trial: Hanna was a phase III trial published in 2008. It enrolled Stage IIIa and stage IIIb patients with an ECOG of 0-1. Patients were staged with a PET or CT scan and both malignant pleural and pericardial effusions were excluded. There were 203 patients initially enrolled of which 147 were randomized. The 72% randomization factor is equivalent to the 75% we assume for the START trial. The control arm, which was part of the 72% that were randomized, had a median overall survival of 23.2 months from the time of enrollment in the trial. Normalizing for the difference between the enrollment time in the Hanna trial relative to the enrollment time for the START trial equates to a 20 month overall survival in the START population.

8) The Nyman Trial: Nyman was a phase II trial published in 2008. It enrolled Stage IIIa and stage IIIb patients with a PS of 0-1. The method of patient staging and the treatment of pleural effusions were unclear. Median overall survival in the trial was 17.8 months. Median survival was measured from randomization, which took place before initial chemoradiation. This was a first-line trial so we need to adjust for the roughly 25% of patients that would be ineligible for START and dropped prior to START randomization and the four month delay that would be seen in the randomization date for START relative to the Nyman randomization date. We also need to adjust for the unclear treatment of pleural effusions and the possibility of staging mistakes. In aggregate, we would expect this to equate to a 20 month overall survival in the START population.

We chose these eight trials as the most representative of the patient population being treated in the START trial. While some of the trials require extensive adjustments to make them apples-to-apples, such as Kelly, others are nearly identical, such as Movsas. Importantly, none of the trials gave us any reason to expect a median survival that even approaches 29-months.

In contrast to our diligent approach of looking for trials that are comparable to the START trial, we believe that Oncothyreon’s vocal short seller Martin Shkreli has chosen trials that are not representative of the START population. In addition to the Kelly trial, he chose the following three trials:

1) The Franca Trial: Franca was not a trial. It was a review of patient medical charts focused on resected NSCLC patients whereas the START trial focuses on unresectable NSCLC. Resected means that a surgeon has gone in and cut out the cancer whereas unresectable patients cannot be operated on. As a result, reseted patients have always had and will always have a much higher median overall survival.

2) The Gadgeel Trial: Gadgeel was a phase II trial in 2011. It enrolled Stage IIIa and stage IIIb patients with a PS of 0-1. Patients were staged with PET or CT scans and malignant pleural effusions were excluded. It did not have a control arm but it did have a treatment arm that showed 34-months of overall survival. The treatment arm received Cisplatin and Alimta (pemetrexed) concurrently during initial chemoradiation. Docetaxel was given once initial chemoradiation was complete. This is different from START in two respects: (1) the use of Alimta is not allowed at any time during the START trial (though it may be used once a patient progresses and is removed from the trial) and (2) the use of Docetaxel is not allowed at any time during the START trial (though it may be used once a patient progresses and is removed from the trial). This was a first-line trial so we need to adjust for the roughly 25% of patients that would be ineligible for START and dropped prior to START randomization and the four month delay that would be seen in the randomization date for START relative to the Gadgeel randomization date. We are not sure how to normalize this to the overall survival in the START population.

3) The Chang Trial: Chang was a phase II trial in 2010. It enrolled Stage IIIa and stage IIIb patients with a Karnofsky scores of 70-100. Patients were staged with PET or CT scans and malignant pleural effusions were excluded. It did not have a control arm but it did have a treatment arm that showed 29.4 months of overall survival. The treatment arm received concurrent chemoradiation and the radiation therapy used proton beam therapy. This is different from START in two respects: (1) START patients may have either concurrent or sequential chemoradiation (concurrent has a roughly 3-month benefit to overall survival) and (2) START patients do not have proton beam therapy (there are only 3 sites in the US that give this to lung cancer patients). This was a first-line trial so we need to adjust for the roughly 25% of patients that would be ineligible for START and dropped prior to START randomization and the four month delay that would be seen in the randomization date for START relative to the Chang randomization date. In aggregate, adjusting only for the use of concurrent chemoradiation, we would expect this to equate to a 28 month overall survival in the START population.

In total, this article has reviewed eleven recent papers. Of the eleven papers, ten concerned trials and one concerned a chart review in a patient population largely irrelevant to the START trial. In the ten relevant trials, only two suggested that it was at all likely for the control group in the START trial to live in excess of 29-months. One of these two trials, Kelly, had a far different design than START. Normalizing for the design differences between the trials suggests that the Kelly population would have about 22 months of median overall survival when placed on a like basis with the START population. That leaves only one trial that showed a median overall survival of greater than 29 months, the Gadgeel trial. Gadgeel was a very small trial (28 patients), had no control arm, and the treatment arm received multiple treatments that are not available to START patients. Nonetheless, Gadgeel shows us that it is possible that the control group is living in excess of 29-months.

The trials that I have reviewed strongly suggest that the control group is living somewhere on the order of 20-months, which is in-line with the expectations of Merck KGaA when they designed the START trial. There is still an explanation that the bears will cling to: new drugs such as Alimta, Tarceva, and Avastin are extending lifespans. There are, however, problems with this contention:

(1) Avastin is only approved for use during initial chemoradition and is not approved as a second-line therapy. Patients are not eligible for START if they used Avastin in their initial chemoradiation therapy. Avastin showed a roughly 2-month improvement in median overall survival

(2) Tarceva is a second line therapy that patients are allowed to use only once they are removed from the START trial due to progression. Tarceva’s improvement in median overall survival is 3.3 months as a second line therapy, but it would probably be less if used as a third-line therapy for patients coming out of START

(3) Alimta is approved for use during initial chemoradition, as a maintenance therapy, and as a second-line therapy. Patients are not eligible for START if they used Alimta in their initial chemoradiation therapy. Alimta showed a roughly 0.4 month improvement in median overall survival as a second line therapy

Clearly, it is difficult to imagine any of the above treatments substantially extending the median overall survival of the control group.

In order to assess Oncothyreon as a stock, let us pose the only important question: Is it likely that the control group has a median overall survival of more than 29 months? If you believe that this group is living in excess of 29-months, then you would logically believe that the trial is likely to fail. In contrast, if you believe that the control group is living less than 29-months, then you would conclude that the trial will succeed. I submit that the historic literature from similar trials is overwhelmingly in favor of the control group living less than 29-months and that the START trial therefore has a 90+% probability of success.

Disclosure: I am long ONTY.