NorthWest Biotherapeutics Inc (NWBO)

[Pyrrhonian]

NWBO Contrarian Analysis

Very recently, DCVax-Direct data were divulged at the ASCO "Industry Expert Theater." This was not an official ASCO presentation, which are selected by the ASCO committee for a trial's importance and data excellence. NWBO had likely submitted a standard abstract for admittance of DCVax-Direct data to the ASCO poster presentations but were rejected. The Industry Expert Theater is a first-come-first-served venue whereby those who have paid for and reserved booth space may also pay an additional $45,000 for a time slot to present there. Data presented are not peer reviewed.

In ASCO guidelines it states,

Once an abstract has been publicly released by ASCO and the embargo has lifted, you may widely distribute a press release containing the full data, including any additional data that will be presented at the meeting even if not included in the abstract itself.

The notion that NWBO did not go through the main channels of presentation at ASCO so that they would be able to present up to date information is incorrect. The above clearly states otherwise. They either did not submit an abstract (less likely) or did and it was rejected (more likely). The only way they could present was to pay their way in. There is nothing wrong with this of course, but it does create a distinction between the data they presented and that of others. This is also why a disclaimer was necessary in their press release stating it was not an official ASCO presentation.


DCVax-Direct Overall Survival

Slide 7 here: http://www.nwbio.com/NWBT_ASCO_2015_5-30-15.pdf

Shows a compelling-looking survival graph of the 39 evaluable Ph I patients. What is missing from any of these slides are a comparator group. The only way to know how well DCVax-Direct is or isn't working, in the absence of obvious extensive tumor shrinkage, would be to enroll a concurrent control group with identical inclusion/ exclusion criteria, give them placebo or no treatment at all and compare survival times between groups. That is usually beyond the scope of a Ph I study, however, which is largely information-seeking and has safety rather than efficacy as its primary endpoint. Though another way to have at least some idea how well each patient performs would be to provide a detailed analysis of each patient's life expectancy. Also the rate at which their cancer had previously been progressing.

Kris Carr is one moderately famous stage IV cancer patient with multiple lung and liver metastases that has been alive for over 10 years since her stage IV diagnosis. Her cancer is very slow-progressing. With all of the interest in their Ph I DCVax-Direct trial, it would not be difficult for the NWBO sponsored study researchers to find and enroll select patients with the highest life expectancy and slowest progressing tumors they could find. Not for nefarious reasons either, but to provide their experimental tx (in this case DCVax-Direct) the best opportunity to show effect by having the least serious obstacles they could afford it. Otherwise the patient may be too far gone for any therapy, regardless of how effective it may actually be. In a small Ph I study researchers can screen deeper and be highly selective.

Ph I studies are also always adaptive in design, meaning they can adjust criteria as they go. It appears they did just this in the DCVax-Direct trial. When it first appeared, the criteria called for only > 3 months life expectancy, and no checking of lymphocyte counts:

https://clinicaltrials.gov/ct2/show/NCT01882946?term=dcvax&rank=5.

This was changed to > 6 months so that patients could be expected to "live long enough to receive at least 4/6 doses," and also required an absolute lymphocyte count of >/= 500 mm3:

http://utm-ext01a.mdacc.tmc.edu/Dept/PROT/clinicaltrialswp.nsf/Index/2013-0160?OpenDocument&ExpandSection=5%2C4%2C3#_Section5

And the trial had always screened out heavily pretreated subjects. If you read down the list of criteria you can see how selective they were in choosing patients. They obviously chose patients with the greatest life expectancy and highest total lymphocyte counts they could find, having access to a great number of patient choices (vastly populated indications).

If they wanted to give a very honest reflection of how DCVax-Direct may be impacting survival, they would have included pertinent information--specifically, the life expectancy vs current survival time of each patient and the rate of progression before and after therapy. They did not do either. Also, a 24% growth of the sum of all measurable lesions is still considered stable disease (SD). Therefore a patient in this trial could be progressing at the same slow rate as before the study some 6-9 months in and be labeled SD.

Without knowing the prognosis of each patient, we cannot know how well DCVax-Direct is working or not. For sure, however, it is not producing adequate enough shrinkage to label any patient as even a partial response (PR). Future approval will thus be dependent upon survival endpoints, even surrogate ones (such as PFS). That would require more time/expense and enrolling a control group, and possibly developing a placebo.

The reason why so many Ph III studies fail even after compelling-looking data from earlier phase studies is due to a lack of powering and not using adequate controls in those earlier trials.

http://www.skepticalraptor.com/blog/wp-content/uploads/2012/08/Why-Pivotal-Clinical-Trials-Fail-Part-1_v12L_a.pdf

The DCVax-Direct study has no control, and researchers are providing no details on each patient. Therefore, considering how selective the criteria were, and the evidence they changed it to become even more selective mid-study, there is no reason to think the current overall median of about 10-12 months OS (trending) is exceptional.


Pancreatic Cancer

This is true in particular when considering the study's pancreatic cancer patients. Seven patients with metastatic pancreatic cancer were enrolled. The median overall survival (mOS) for these is only 3.9 months. One of the seven cases was pancreatic NET, a rare form of pancreatic cancer with a much improved prognosis: http://www.cancer.gov/types/pancreatic/hp/pnet-treatment-pdq

3.9 mOS is hardly promising, which is probably why they will avoid this very aggressive indication going forward. The two pancreatic cancer patients that have survived the longest from the trial are the panc. NET patient (expected) and Allan Butler, whose case was featured in "Stand Up to Cancer" last year. His case was likely a very slow progressing cancer. Being the first enrolled in a first-in-man study, researchers would want to find the healthiest specimen they could that still fell under the broad heading of stage IV unresectable cancer FDA required: high life expectancy with very slow progressing or even stable mets.


Many Ways to Interpret a Chart

One attempt made by NWBO to demonstrate effect of therapy was done in slide 9. Those who received 4 or more injections lived the longest. The claim being made is of course more injects correlates with improved survival. The simpler and much more obvious correlation to be made however is that those with a better prognosis just lived longer and so were still alive to receive them.

This is an example of confirmation bias. Interested parties, regardless of scientific background, looking for positive ways to interpret data will always find them, no matter how erroneous the premise.

The same could be said of slide 11. Here they show that SD at week 8 (meaning 24% tumor growth or less) correlates with an improved prognosis (will live longer). This is somehow then related to DCVax-Direct being the cause of SD at week 8. However, again it's clear that healthier patients will have slower progression or even no progression for longer periods of time. Being that NWBO was able to selectively pick the patients for this study, and have no control group, and have very high inclusion criteria standards, means that de facto a majority should be SD at week 8. It does not correlate cleanly with effect of therapy.

Without the ability of DCVax-Direct to generate significant tumor shrinkage, which is clear at this juncture, they will have to rely on larger studies with a concurrent control group with a survival endpoint for any kind of approval. Even if the Ph II studies generate compelling-looking survival rates it will be chalked up to patient-selection bias, just as with the information arm for DCVax-L.


DCVax-L Modeling

A German protocol surfaced which calls into question modeling that had been done for the Ph III trial. Apparently by late 2012 only 49 patients were randomized in the Ph III:

The study 020221 is an ongoing Phase III study. By August 15, 2012 patients were enrolling in 41 clinics in the United States in the study and 84 GBM patients were included. Of these 84 patients in the Phase III study, 49 were randomized (2:1, active treatment against placebo), and 35 patients were enrolled in an open label arm of the study, specifically for patients with tumor progression that is faster.

http://portal.dimdi.de/amispb/doc/pei/Web/2613044-palde-20140601.pdf

It seems they dropped data from the 33 that were enrolled before. There are also data from when they had no placebo and were enrolling patients into a non-placebo control group. These dropped out, but those enrolled into the treatment cohort continued to be treated. Apparently there is also some 10 years of compassionate use data for DCVax-L in GBM that has never been brought up. The only compassionate use data revealed to the public to date is the information arm data. Of course, these data are confounded by the intermixing of an indeterminate number of pseudo progressive patients, who have a relatively high mOS of around 27 months. Where are all the missing data? Have these data not been brought to light because the numbers are not flattering?

Back to the German protocol--the Company never again mentioned the 33 initial patients after early 2011, and with respect to the German protocol stating only 49 by later 2012, it seems most likely that the 33 were dropped from the analyses, and that they randomized 49 patients from May 2011 - Aug 2012. The leaked protocol also confirms this by way of its estimated accrual time and follow up. No mention of these 33 are found in the leaked protocol or the German protocol, the latter of which going into detail on all patients treated in the Ph III and their adverse events to date.

The ratio is also very consistent. You would expect to find 35 potential ePD from the same size group 49 normal PD patients were randomized from. If 33 were from 2008 when there was no information arm (enrolled only 2011-2012), then the numbers would be closer to "35 information arm and 82 randomized."

Therefore the most logical way to model is to drop that group of 33 from 2008 altogether.

Also, regarding the potential mPFS of placebo group, it was found once data had matured that the ICT-107 placebo group performed very well, at 10.1 months mPFS (about 12-13 months from surgery). This was without removing early progressive disease patents (ePD) during or immediately following radiotherapy. Also it was without the high bar of requiring 1,000 mm3 absolute lymphocyte count, which correlates strongly with improved PFS/OS rates. It also correlates with high CD4+ cell counts. The DCVax-L Ph III included both in its criteria, which further removed worse prognosis patients.

It isn't unreasonable therefore to expect the DCVax-L control group to perform even better--perhaps 12-15 months mPFS.

When we consider both factors, and have an enrollment ramp of:

-0 by May 1, 2011
-25 by Feb 1, 2012
-40 by June 15, 2012
-49 by Aug 15, 2012
-80 by Dec 15, 2012
-170 by Dec 15, 2013
-192 by March 1, 2014

And:

-66 PFS events by Dec 1, 2013

And assume:

-Control mPFS of 12 months

And so get:

Initial assumption (conservative): 50% of control group will experience PD at 12 months, 75% at 24 months, 95% at 36 months, and 99% at 48 months (12 months median PFS); tx group yet to be determined.

We arrive at an estimated enrollment ramp, conservatively figured of:

-Q1 2011 -- 0tx:0placebo
-Q2 2011 -- 2tx:1pl
-Q3 2011 -- 3tx:2pl
-Q4 2011 -- 4tx:2pl

-Q1 2012 -- 9tx:4pl
-Q2 2012 -- 9tx:4pl
-Q3 2012 -- 12tx:6pl
-Q4 2012 -- 14tx:7pl

-Q1 2013 -- 14tx:7pl
-Q2 2013 -- 15tx:7pl
-Q3 2013 -- 15tx:8pl
-Q4 2013 -- 16tx:8pl

And estimated PFS events occurred and projected of:

-From '11 by 12/'13 -- ?tx:4pl
-From '12 by 12/'13 -- ?tx:14pl
-Over '13 by 12/'13 -- ?tx:8pl

Therefore 26 pl group enrolled since '11 contributed to the 66 events of the former first interim trigger, the rest (40) from tx group. Here is the likely breakdown:

-From '11 by 12/'13 -- 8tx:4pl
-From '12 by 12/'13 -- 23tx:14pl
-Over '13 by 12/'13 -- 9tx:8pl

-66 Total PFS events triggered 12/'13

Deduced tx group assumption: about 16 months mPFS.

Trending about 4 months difference between groups. If placebo group is performing better than 12 months mPFS, the difference could fail to be stat sig.

There are a number of interesting small n academic studies that show higher mPFS and mOS than historical controls, but often these trials had younger, healthier participants selected for them. As usual, without a concurrent control group these data are suspect.


Moving Forward

There is enough activity to suggest some effect of therapy is being seen with DCVax-Direct, but there is also another explanation for nearly everything. The necrosis sometimes seen is apparently the result of a localized cytokine storm, and those lesions not injected that experience some affect are nearby. Inducing a cytokine storm has so far shown mixed therapeutic results in clinical practice and in trials. Note that PFS rates and even SD rates were not divulged from this Ph I DCVax-Direct study. The trial participants could be over 80% progressed (only 20% PFS) and we wouldn't know from these slides. Only survival is mentioned. This is certainly not painting a clear picture of how the patients are doing on this therapy.

The emergence of t-cells in necrotic regions is logical, as they would be induced to localize anywhere host-cells have undergone necrosis (i.e. frost bite). If the emergence of t-cells was noted in non-necrotic masses, that would possibly be indicative of a targeted effect, especially if in regions of the tumor where no injection had occurred, as all injections will produce some necrosis. Instead, t-cells have only been shown to congregate where injections and necrosis have occurred, but this is to be expected. There is no evidence to suggest DCVax-Direct induces a targeted t-cell response against tumor cells, causing the latter to differentiate their antigen profile from self-antigen presenting cells.

At this point, the future potential success of both DCVax-Direct and DCVax-L are entirely unknown. There are no studies testing autologous whole tumor lysate loaded DC vaccines showing impressive survival between a tx group and placebo group. DCVax-L for GBM is the first of its kind. There is no precedent, therefore the risk of failure is high.

As an investor in a speculative biotech you are taking on a large amount of risk, and even more if you are also investing in options. NWBO management will never tell you this. Trials can be delayed or extended for any number of reasons while your time value on those options diminishes to nothing. The trials could fail to show stat sig data and the stock plummet 80% (i.e. CLDN). You should be widely diversified and invest only what you are absolutely comfortable losing on NWBO. This is far from a de-risked investment.

The other potential problems lie in manufacturing the product on a mass scale. So far the only precedent we have for this attempt is DNDN, whose debt in getting up to scale swallowed up all of its equity until it had to file chapter 11, even with a $93K price tag for 3-months of treatment. If DCVax is cheaper to manufacture, is it sufficiently so to overcome the debt NWBO will have to take in to scale up appropriately? This is unknown.

In fact everything regarding the future success of NWBO is unknown. And that is the main reason why Woodford is willing only to risk 1%-2% of his investor's AUM on it. If you are impressed with Woodford and want to model his investment practices, you may find yourself far more overweight NWBO than you should be.

55% of all Ph III studies fail, and over 80% that were not preceded by a randomized Ph II. Statistically, and based on the variable glimpses of data management has selectively chosen to present to investors, DCVax-L and Direct have an 80% chance of failure.

As one frequent poster suggests: "Respect Risk, Manage Your Assets Wisely, Diversify."

Are you taking this sage advice?