Recently some people have declared Provenge, a treatment for prostate cancer, "dead." This is because of an analysis made independently by Marie Huber [See Insight: New doubts about prostate-cancer vaccine Provenge (Reuters)]. Marie has a BA in Biochemistry from Cambridge and a Masters of Philosophy in Bioscience Enterprise from Cambridge and MIT. Her work history is not as a scientist but as a business analyst, and she claims to currently be an independent citizen-scientist.
Before getting to her analysis, I should say there is no point to the character assassination that apparently Ms. Huber has been subjected to. Her analysis either stands on its own weight or it does not. [Disclaimer: my own biostatistics credentials are challengeable too. I took one probability course in college, worked for 1 year at a low-level job at a biostatistics company, and otherwise I am mainly self-taught. I have done professional analysis of biotechnology companies for about 5 years now, invest in biotechs myself, and have a good, but not perfect, track record.]
If you want to see it for yourself, Marie Huber's analysis is available on the internet as her video Understanding the Provenge Trials as is the paper she co-authored, published in the the Journal of the National Cancer Institute (of Britain): Interdisciplinary Critique of Sipuleucel-T as Immunotherapy in Castration-Resistant Prostate Cancer.
In the video she specifically tries to convince oncologists and urologists to not prescribe Provenge for patients with metastatic prostate cancer. She says her analysis shows that Provenge "might" be harmful, rather than helpful to patients. Given the low rate of side effects for Provenge, far lower than typical cancer treatments including chemotherapy and physical interventions, that is an extraordinary claim.
Ms. Huber starts with a perfectly reasonable review of prostate cancer, the immune system, the Provenge mechanism, and the trial design and results. Her first factual critique (here I am assuming she is being factual about data Dendreon released) is that only 27% of treated patients showed an immune response to PAP, the target protein. For me, that in itself might account for the low response rate; but we already knew there was a low response rate, one acceptable to the FDA given the low risk of side effects. More troubling, those 27% who did respond (with longer time till death) to PAP, as with the 73% who responded to the PAP-catalyst combination meant to induce response, could not be shown to have a survival difference. So, if that holds up, the process is prolonging survival in some patients, but not for the reasons usually given.
She points out that Provenge showed no effect on time to progression. But that was known long before the final Phase III trial. This should not, in itself, be a problem, and it was not for the FDA, because there are drugs that do delay time to progression, but don't increase overall survival. Overall survival is the gold standard, not time to progression.
It is data released the after the FDA approval that really gets Ms. Huber going (available for you to see at http://www.fda.gov/BiologicsBloodVaccines/CellularGeneTherapyProducts/ApprovedProducts/ucm213554.htm).
What the data proves most obviously is that the FDA raked Provenge data over the coals before approving the therapy. But we already knew that.
Huber focuses on a particular table of age-related survival data. This shows that the benefits in the trial were far greater for patients over 65 years of age than for those under 65 years old. Within the patients receiving the placebo, older patients died about a year earlier than younger patients, which should not be the case if all prostate cancers are equal and are the cause of death. The table also shows, but does not seem to interest her, that African Americans do really, really well on Provenge. So much so that I am surprised no one else has pointed this out.
In most vaccines, per Huber, older patients show lesser responses to vaccines than younger patients. In the Provenge data, again using 65 as the age division point, while the therapy seems to work in older patients, the younger patients who get Provenge survive longer.
Huber believes the most likely explanation of this data is that older patients were hurt by the placebo, but for some reason younger patients were not. If so, what we had may have been double blind, but in fact we were testing against a poison, not a placebo. The claim then would be that the Provenge process, aside from the immune-protein creation part, has negative health consequences for patients.
Interesting, but it involves leaping to statistical conclusions on an arbitrary dividing point, the age 65. A Dendreon spokesperson has indicated that, dividing at age 71, the anomaly goes away.
Statistics work best on like things, like a perfectly balanced, perfectly spaced roulette wheel. Each human is different, genetically and in state of health, when they enter a trial. We don't force cancer patients to eat identical diets or live otherwise identical lives while in trials. We hope our statistics catch meaningful aggregations of data. Dendreon has never claimed that Provenge cures prostate cancer. On the whole, no matter how you divide up the patients, Provenge prolonged survival in a double blind study (actually, 3 studies). If, like other cancer drugs, it hurt some patients while helping others, that is just par for the course. Chemotherapy frequently kills elderly patients before their cancer can. From Huber's statistics, at worst Provenge causes slight harms to some subsets while showing greater benefit to other subsets.
At the extreme, you can take the ten percent of patients who live longest and the ten percent who die quickest, and claim Provenge (or any therapy) does not work because the second set does worse than average. The right question to ask is: did the average improve, with minimal down side for the worst-affected? By that criteria, Provenge passes, and that is basically the criteria used by the FDA for approval.
Huber speculates that the blood processing reduced the number of lymphocytes in the blood returned to placebo patients. This would be easy to test. Perhaps her team can get a grant for that. I would like to see the results. Of course, if you want to kill lymphocytes, that is easy to do. Dendreon's process was intended to preserve lymphocytes. Huber admits that her speculation is unproven.
If lower lymphocytes in placebo patients were a problem, then the question arises: what caused the early deaths? In the beginning Huber stated that one reason patients have prostate cancer in the first place is that their immune system is not attacking cancer cells. Removing lymphocytes should not alter that. So the deaths should be attributable to other causes, like infections. As far as I can tell, the data does not support more placebo patient deaths from infections. The deaths are from the cancer. There may be other explanations, but the obvious one is that the placebo patients were just that: placebo patients. Which means the treated patients who lived longer did so because they received Provenge. The placebo patients lost the lottery. Sad, but that is what happens in clinical cancer trials of effective therapies.
Marie Huber did some interesting work, and I have no problem with her publishing it in a science journal. Whether or not her conclusions are correct, continuing to collect and analyze data is appropriate. It should soon be possible to generate statistics on patients treated with Provenge since its approval. In addition, Provenge should certainly be studied, in clinical trials, at earlier stages in the cancer progression process.
My problem is: suppose Ms. Huber is wrong. Suppose doctors don't recommend Provenge because they are busy and Huber's work reinforces a prejudice, or a salesman for a rival therapy overstates Huber's case, or whatever. Then lives will be shortened because someone rushed to put out data without checking it by actually conducting a study.
The FDA can get it wrong, but it is the agency responsible to see that drugs are safe and effective. If the FDA thinks Huber & crew are right, they should take Provenge off the market, even if it means they look bad for not being careful enough the first time around. Even if it means I lose some money. But to me the story that comes through Ms. Huber is that the FDA did a very thorough job, while Ms. Huber admits she has not proved her point. Right or wrong, it is speculation.
Disclaimer: I am long Dendreon. I won't trade the stock for at least 3 days after this article is posted.
Before making a decision be sure to read my Q4 2011 Dendreon call summary and check out my main Dendreon page, which has loads of background information.