Don't Get Burned by the Next Cell Genesys

I spent a good bit of time last March responding to readers incredulous that I could doubt the GVAX prostate cancer vaccine. In response to one such email, I wrote the following:

"If you remember one rule about biotech investing, or more specifically, one rule about investing in biotech oncology stocks, it's that you should never (never!) trust phase II data from uncontrolled studies.

Print that out. Place next to your computer. Memorize."

Well, it turns out that the stupendously positive phase II data on GVAX Prostate was completely worthless.

Prostate cancer patients taking GVAX in the phase II trial had a median survival of 26.2 months. In a subset of patients at the higher GVAX dose, median survival was 35 months.

But all patients in the phase II study received GVAX, so there was no control, and the study wasn't randomized so there was nothing to which GVAX patients' survival could be compared.

Of course, that didn't stop Cell Genesys or its believers to compare these GVAX patients to historical data. In this case, Taxotere, approved for prostate cancer patients, produced a median survival of about 19 months in its own pivotal study.

Since 26 or 35 months' survival from GVAX's phase II study is way better than what Taxotere produced in the past, Cell Genesys fans reckoned that the phase III study pitting GVAX directly against Taxotere was a sure bet to succeed.

Not so much, as we learned Thursday.

Here's how I closed my response to that Cell Genesys investor in my March column:

"The quality of treatment changes, patient characteristics are different, there is selection bias -- heck, everything is different from one study to another. When the phase III GVAX results are announced next year, patients on the vaccine will live far shorter than they did in the phase II study. Likewise, Taxotere patients in the study will live longer than expected. This is why the odds are strongly against GVAX Prostate and why I'm no fan of the stock."

The only thing I ended up getting wrong was the timing. We didn't need to wait for 2009 to find out that GVAX was a dud.