With less than a month to go until Dendreon (Nasdaq: DNDN) releases pivotal clinical study results for its potential prostate cancer drug Provenge, lets go over a few points to consider when trying to determine the likelihood that its phase 3 IMPACT study will be a success.

1) Dendreon announced during its interim IMPACT study results release in October that there was a 20% reduction in the risk of death in the Provenge arm of the study relative to placebo” and that the final analysis will require Provenge to achieve a 22% reduction in the risk of death versus placebo to hit the study’s statistical significance target (some unknown figure south of 0.05).

Therefore the remaining Provenge patients in the study will have to show something better than a 22% reduction in the risk of death versus placebo for the study to be a success. This is a very important point that should not be lost on Dendreon investors.

Based on when drug developers like Onyx with Nexavar (sorafenib), Genentech with Avastin (bevacizumab) and Pfizer with Sutent (sunitinib) chose to do analyses in their pivotal studies as well as the survival data from Dendreon previous studies, our estimate would be that Dendreon undertook its interim analysis at some time between 65% to 80% of the 304 events in IMPACT.

Dendreon has made the point that in D9901 that the survival advantage seen with Provenge grew over time (i.e. the separation between the Provenge and placebo survival curves grew) so if you believe that Provenge truly does work then this is something to hang your hat on.

Survival curves from Dendreon's D9901 study comparing Provenge (in bold) to placebo (from 29-03-07 clinical briefing documents)

2) The next point that we wanted to bring up was something that some of the analysts on the interim IMPACT study results conference call in October seemed a little confused about. If the final analysis requires a 22% survival advantage to hit statistical significance then unless Dendreon went completely crazy and allocated the majority of its alpha (0.05) to the interim analysis, the p-value hurdle to overcome for the interim analysis had to have been harder than for the final analysis.

There are three key variables that affect the odds of the IMPACT study’s final analysis being successful:

• The p-value hurdle: This is normally 0.05 but since Dendreon allocated some portion of that alpha to the interim analysis then the final   p-value hurdle will be something south of 0.05

• The number of events in the analysis: The more data points there are, the more precise the data produced will be.

• The size of the treatment effect

The three variables above that will determine IMPACT’s success are why we are upset with Dendreon’s decision to move up the expected date of its final analysis from 2010 to 2009. Last year Dendreon decided to give up some of the alpha that was intended for the interim analysis and allocate it for the final analysis (a good decision) but then in order to get the final analysis sooner than 2010, Dendreon allocated fewer events (304 versus 360) to the final analysis, which strikes us as bad and not a very conservative thing to do when Dendreon could have just kept the 360 events for the final analysis and thereby improved the power of the study.

3) One thing that always raises the chances of something going wrong is when study enrollment criteria gets changed after a clinical trial has already begun.

Dendreon had valid reasons to change its enrollment criteria for IMPACT back in 2005 (two years after the start of the study) and the use of a Cox  model to analyze IMPACT’s data does provide an added measure of safety against possible enrollment imbalances. We’ll also note that Dendreon’s CRO appears to have done a good job of balancing D9901 and even D9902A despite those being much smaller studies.

Nevertheless, it has been our experience that the unexpected often seems to happen when data results day comes up or FDA review time rolls around after studies have been amended multiple times.

4) Shares of Dendreon rose 38% last Friday on news that the American Urological Association saved a spot in its annual meeting for Dendreon to present the final IMPACT results if they are ready in time for the meeting. This adds nothing new to the Dendreon story and the fact that a medical meeting saved a spot for the IMPACT data does not increase the odds that IMPACT will be a success, nor does it hint that IMPACT has already produced good results.

What the AUA placeholder does tell us is that if IMPACT hits its survival endpoint, it will be extremely significant news worthy of headlining a major medical meeting. This is something that everyone should know already though, considering that Sanofi-Aventis’ (NYSE: SNY) Taxotere (docetaxel) is the only FDA-approved therapy that has shown an overall survival benefit as a treatment for metastatic prostate cancer.

If Dendreon knew the IMPACT data already, they would have to release at least the top-line results to the investment world or face the investor class action lawsuits. That’s why the AUA placeholder does not alter the odds we ascribe to IMPACT’s success.

5) With most upcoming clinical trial results we have to worry about not just the efficacy data but the safety data as well. Due to the terminal nature of androgen independent prostate cancer and Provenge’s relatively benign safety record in previous studies, we are not too concerned about what sort of adverse event data Provenge might produce in IMPACT.

There was some cerebral vascular events seen with Provenge but since it is not being proposed for an early stage of prostate cancer yet, we can discount these safety issues being a potential impediment to Provenge’s marketing approval.

6) Our final point of concern is that so much about how prostate cancer is treated has changed since 2000 (when D9901 began enrolling patients). Prostate cancer survival times have been improving quite a bit (although data from the 21st century is not out yet) and with treatment and so much else surrounding prostate cancer changing in the time span between IMPACT results and D9901  study results, the odds of the unexpected happening are higher than if IMPACT data was coming out closer to when D9901 survival data was announced (back in 2004).

Now here’s the part where we put on our investor hats and ignore the fact that we are always rooting for new medical advances: we wouldn’t feel comfortable going long or short Dendreon’s shares at its current $6.30 share price (April 10). If Dendreon hadn’t changed the study enrollment criteria for IMPACT, if Dendreon had chosen to raise the final powering of IMPACT when it lowered the interim analysis powering, and if IMPACT’s results weren’t so far removed in time from Dendreon’s previous Provenge phase 3 studies then we would have more comfort in forming our expectations for IMPACT’s final results.

While we believe the odds favor IMPACT failing its primary overall survival endpoint, perhaps a 70% probability based on the above factors and the general stochastic nature of clinical trials*, the payoff to being right on IMPACT failing is not large enough (at Dendreon’s current share price) to compensate for the still very real possibility that IMPACT does hit its primary endpoint when top-line results are released later this month. In other words, we feel that Dendreon is in its fair valuation range at its current $630 million enterprise value (accounting for its convertible notes) ahead of IMPACT’s study results.

If Dendreon’s shares reach the $8.15 mark (an approximate $810 million enterprise value) we would feel comfortable enough to recommend a short bias (with hedges) towards Dendreon ahead of IMPACT. If its shares fall below $3.60 (an approximate $360 million enterprise value) , we would be comfortable saying that Dendreon’s upside potential outweighs its downside risks. Unless Dendreon reaches either of those two points ahead of IMPACT we’ll be happy to recommend following the IMPACT final results release from the sidelines. In our opinion the payoff for being right on IMPACT just doesn’t justify the risks at this point.

* Even a study powered 80% to show statistical significance on its primary endpoint (if a drug’s effect is real) will fail  an average of 1 out of 5 times and that also assumes the powering assumptions are accurate (which they may or may not be for IMPACT)

For further reading click below:

www.fda.gov/ohrms/dockets/ac/07/briefing/2007-4291B1_2a.pdf

www.biotechspeculators.com/archives/235