Investing in the biotech sector is scary enough for most investors. Investing in biotech stocks with a lead drug that has only produced messy study results is even scarier. Dendreon (NASDAQ:DNDN) has lately been haunted by unflattering statistics that are likely to frighten more than a few investors.

Before the FDA issued an approvable letter in May to Dendreon for its potential prostate cancer treatment Provenge, Dendreon had completed two small 127-person and 98-person phase 3 clinical trials that were labeled the D9901 and D9902A studies.

In the D9901 study, Provenge failed its primary endpoint of slowing down prostate cancer, but was successful on improving 36-month survival, even though that was not a pre-specified endpoint of the study. Why does this matter? Non-statisticians may not see the importance of pre-specifying study endpoints, but one of my favorite quotes goes something like this: "An exact answer to a wrong question can always be made precise." And then there's always the "three kinds of lies" made famous by Mark Twain: "lies, damned lies, and statistics."

Due to its smaller size, Provenge fared even worse in the D9902A study; it failed on the same progression-free survival endpoint and also showed an inability to improve overall survival in a statistically significant manner. Some people like to point out that the 3.3 months improvement in median survival time that patients experienced in the D9902A study is meaningful, even though it wasn't statistically significant.

But let's shape the D9902A trial results in another way. If we use survival confidence interval data from a FDA briefing document, we see that if the study was run 100 times, then 95 of those times the placebo patients would have median survival times in the 12.8 month-to-25.4 month range. Unfortunately, 95 of those times the Provenge patients would have median survival times in the 13.6 month to 31.9 month range. So if randomness wasn't in Dendreon's favor in this study, placebo patients could easily have lived nearly twice as long as the Provenge patients (depending on the sample used). Yikes!

Even after correcting for study imbalances in the D9902A trial, using what statisticians call a Cox model, the FDA states "there is not enough evidence in support of (Dendreon's) finding of statistically significant difference between the (Provenge treated patients and placebo group) in overall survival."

Dendreon just completed patient enrollment this week in another phase 3 study that will serve to either put Provenge on the market or seal its fate among the "hoped for but failed" drug candidates like Telik's (NASDAQ:TELK) Telcyta.

The first interim results of this large, more than 500-person phase 3 study dubbed IMPACT are due out in the second half of 2008. If Provenge shows a strong enough improvement in overall survival when these interim results come out, then Dendreon will likely have FDA approval of Provenge sometime in 2009. The only problem with this scenario is that interim results have a much higher statistical significance hurdle to overcome than final efficacy results of a study.  

Dendreon is trading at valuations commonly assigned to companies that have produced positive phase 3 data, even though it has only phase 2-quality results. New investors in this company would be making more than one of those critical errors that investors sometimes make.

Those looking for an oncology-focused drugmaker with more impressive compounds in development should take a look at GPC Biotech (NASDAQ:GPCB) or Rule Breakers pick Exelixis (NASDAQ:EXEL) instead. It's time to go to Motley Fool CAPS and rate Dendreon underperform -- at least until the interim IMPACT data comes out.

Want to know what other companies give us the frights? You can view the rest of our hair-raising stocks here.