was highly statistically significant, with a p-value of 0.01. That meant there was about a one in 100 possibility that the finding was due to chance, and not necessarily the drug, which is good enough to provide confidence to even the most conservative biostatisticians.
After three years of follow-up, the drug appeared to be looking even better. About one out of every three Provenge patients were still alive then (34 percent), compared with one out of 10 in the placebo group (11 percent). Dendreon has hypothesized that the drug appears to look better over time because it has a “delayed effect” as it takes time to rev up the immune system against prostate cancer.
These findings were not from the primary goal of the study, which prompts some statisticians to dismiss them out of hand. And another identical study in 98 patients, called 9902a, failed to produce evidence that was statistically significant, meaning it’s possible the first study could have been a fluke.
Still, when those two trials are pooled, as they were originally designed to be, the findings again favored Provenge. The median survival benefit in a total of 225 patients was still 4.3 months, with a statistical p-value of 0.0006. (That means it’s extremely unlikely that the survival benefit seen was due to chance, or that if Dendreon ran the trial 1000 times, they’d still expect to see a survival benefit each time.)
Since we’re talking about Dendreon, there is likely to be some drama and debate about this data. The latest trial, called Impact, started in 2003 and took six years to produce final enroll all the patients and follow them for the survival result. That’s a lot of time to enroll patients and follow them, and Dendreon will surely have to scrub through each individual patient record to make sure meticuluous records were kept to satisfy the FDA.
The trial also was modified half-way through, in 2005, to alter eligibility criteria for who could get in the study. Early on, the trial had been restricted
