Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
The DNDN LONGS ARE EVEN MORE SHORT SIGHTED WHEN THEY MAKE THREATS AGAINST PEOPLE WILLING TO LOOK AT THE DATA.
Biowatch - You are one of the few posters who usually offers a balanced and informative view unlike the biased posts of some well-known posters here. However, your blanket statement above about all "The DNDN LONGS" is not just uncalled for, it is likely incorrect. Let's not forget that the extreme views posted here and other boards, positive or negative, are at most from just a minority few who spend all their waking time on these investment boards. Some of these people are clearly in arrested development but sadly too vocal.
The FDA did NOT say NO, they said give us more data. Big difference.
This statement does not take into account the short time to live that some PC patients do not have to wait for new data. As the ABC news piece tonight showed, the true disappointment in the FDA decision is from many PC patients who have been deprived of the ability to try, apparently at little risk, a treatment with fairly strong evidence of efficacy, even if that still seemed inadequate to some.
<Does it ring true that the projected survival is higher in the trial that allows symptomatic patients (9902b) than in the trials that were restricted to asymptomatic patients?>
Good. Now you are saying something concrete that we can discuss. Your inference of wrongdoing was based on your inherent dislike of DNDN management and the two facts that you stated: (1) the slide showing the median survival time in 9902b is better than that in 9901 and (2) the extension of 9902b to include mild pain.
But is that all that is known about the data? Here are a few more for you to ponder: (1) about 100 early 9902b patients had GS<=7, (2) those were found to be somewhat healthier than D9901 per the Provost data, (3) about 60% of patients in 9901 and 9902a were GS<=7, and (4) extension to include mild pain does not exclude no pain. And, you really want to stretch it, think about the number of bone mets for somewhat healthier patients and the fact that the Halabi nomogram does not account for it.
Now factor those into your inference to see if it still holds? If you have not done so, then, IMO, your inference is lacking. But as IO pointed out, you don't really need to work through this mildly challenging brain teaser. DNDN does not need the legal risk as they surely will be liable if they were found lying to promote stock price (as you aptly pointed out) when the data come out at trial end. The point here is that there is enough objective data to work with that you don't need a faith-based reasoning such as trusting Novartis or distrusting Dendreon management.
The problem with inference is that you need some reliable ground facts. When you have made up your mind with questionable ground facts, your inference is suspected.
You seem to pull all debates down to winning and losing instead of exchanging information and mutual learning. There are two separate issues about Dendreon, debating about their data and their presentation of that and debating about it as an investment vehicle. On the former, your mind has made up in a clear direction so we won't need to explore that further. On the latter, I don't know what it has done to you to make you so obsessive in hacking it but Dendreon has been quite good for me.
Your shrillness is getting tiring so I'll stop this thread here. You may have the last word.
Dew - Did your post below make sense in answering my question?
<It’s trivial—all you’d have to do is lop off a few patients who produced the lowest projected survival according to the nomogram.>
In your own words, your statement above "created the impression (perhaps intentionally)" that DNDN was acting in a mindless and criminal manner, when, in fact, you had not a shred of evidence of it. Indeed, "factual posting is evidently not your style".
Now, about the Halabi computations, I know of databases that allow blind computations based only on schemas (prognostic factors in terms of these trials). I DO NOT KNOW FOR CERTAIN, but if this is how the D9902b database is managed, Dendreon do not need to have access to any actual data to run the calculations. The game that you insinuated would be hard to play in that case.
I replied to your post on this board before responding to a post on the iVillage board. Here, my post was simply dismissive of the absurdity of your post (read yours and my reply again to see). On the iVillage board, I gave more details as some people there are not familiar with blinded data. The point of my post was that if Dendreon did what you suggested in your earlier post, that would be unlawful.
<I see that you replied to me after all, but you took it over to iVillage to obtain a more hospitable venue for your misstatements. Extremely lame.>
Unlike you, I do not spend 24 hours on these boards reading and responding to posts and I make no distinction between the two boards. Btw, you seem a bit obsessive and your mindless bashing and the occasional bullying of posters here or there are, IMHO, juvenile and unworthy of you. Good day.
<their intent was always 9902b and that the whole 9901 BLA was just for show.>
So you are saying they believe in D9902b so much, they abandoned an SPA that virtually guaranteed success and changed the protocols to weaken it?
Then, continuing to wait for D9902b, they didn't do a secondary after the AC to lock in the funds to survive until the trial is done in a few years because?
Dew - You can do better than that. But don't bother.
Strictly speaking, there is a difference between excluding patients who did not have the data for an analysis algorithm and simply excluding patients. But that isn't relevant. This data was provided blind by the DSMB. Do you know of a way to game that?
If the data was up to date, then it was even better than I thought. Or are you saying that Dendreon faked that data?
I agree. There are also factors related only to trial conduct that affects the analysis. For example, about 10% of the 9902a treatment arm did not get fully treated. That got to have some confounding effect on the final results. Hopefully IMPACT will do better with that. In addition, with a much larger enrollment number, the imbalances will be less pronounced.
Sarcasm unwarranted. The Halabi data presented today were probably generated when the enrollment number was somewhere below 400. So that could be skewed due to early 9902b enrollees who were relatively healthier than D9901. Under certain reasoanble assumptions, the number of GS<=7 in IMPACT might be around 60-65%, comparable to or better than the 60% or so in D9901. So even if the presented Halabi data might not hold up as well at the end of enrollment, it probably will turn out not too bad comparing to D9901.
<If you are doing Cox Regression anyway then randomization strat isn't worth a lot - except that it minimizes the difference between plain Log Rank and the Cox Regression.>
Careful. There is an assumption being made in the above statement that might not always hold in how a CR is done relative to how the stratification is done. The scale matters.
<What factors led to the relatively poor results of 9902A? How much of the shortfall in log rank p-value and HR is attributable to smaller size? trial imbalances favoring placebo, etc>
If the trial was run well, the small size would have had relatively little effect on the results while the sicker population and imbalances such as bone mets counted for more. However, a large factor might have been that there were more patients in D9902a than in D9901 who did not get full treatment (around 15% vs 5%). Per the patient distribution decision tree in the stat review, there was even one patient on the treatment arm of D9902a who did not get treated! Because of that, the small trial size might have played a larger role due to a reverse cross-over situation where a significant number patients on the treatment arm were acting like placebos. That might have had some effect on the poor TTP result in D9902a too.
<Why should the results of 9902B look much more like 9901 than 9902A at the interim and final?>
The larger enrollment number will make the trial more balanced across prognostic factors, both the ones accounted for and the ones unknown (the latter is really what people refer to when they say that a large trial is more robust than a small trial - we don't always know everything, e.g., genetic variations, etc.). In addition, stratifying on bone mets and Gleason scores should help. Last, per the CD54 data, the early enrollees on the treatment arm were compliant and hopefully the recent ones would be too due to better published trial data. So there should be less of the reverse cross-over issue above.
<The p-values you cite for 9901 and 9902a are based on cherry-picked Cox analyses. In 9902b, the primary endpoint will be based on a Cox analysis that is fully pre-specified.>
"cherry-picked" is hard given that the FDA statisticians also tried a few other models that showed stat sig or trended toward stat sig. However, the large p-value improvement from log-rank to Cox might not repeat with the same magnitude in D9902b. The trial size is much larger, hence less chance for imbalances. Also, it sounded like they are stratifying for bone met count which was heavily imbalanced against Provenge so Cox was effective.
<if you have a low efficacy in a big trial (which you would have to have in order to get p near 0.05) then that lesser imbalance is still just as big as the efficacy.>
This is, of course true. But I am assuming that the efficacy effect is not small. Also Cox only corrects for imbalances that we know about, it cannot correct for factors that are not recorded. In theory at least, a reason for a large trial is to minimize the chance of missing some factor that might turn out to be significant and imbalanced.
In any case, how do you run MC with Cox without data on individual patients? Are you randomizing around the medians of the prognostic factors? What assumptions about the distributions in that case?
With 500 patients, 9902b will be much more balanced than the integrated 9901+02a dataset. Although Cox should give more precision than log-rank, the p-value improvement might not be much more. So for the purpose of simulation, it might be better to just do log-rank since we have no data on the prognostic factors per each data point. What do you see as the power of the trial with HR=1.20 or HR=1.40? Same question for the interim result if N=180 or N=240?
I've discussed on the iVillage board about the possibility that frozen Provenge is inferior to fresh Provenge, primarily because the three doses are essentially three first doses, each with 2/3 of normal volume.
<I think the recent politicking is not really about whether Provenge should be approved but rather who should decide.>
That's basically it. 4 years ago when the OO was created, almost everything related to cancers went to it except for the Tumor Vaccine Branch which remained in OCTGT. This fight is not about a drug but about Dr. Pazdur's ideology or just plain ego.
<So if the follow up to DN-101 eventually brings the log rank p value under 0.05, would you support approval?>
When the time comes, I will look at the entirety of the data first before making a decision if I am still interested. My post pointed out an inconsistency in Dr. Scher's reasoning, ie, a counter-argument. It does not imply anything further. You can read more into it but that is your prerogative.
<How do the FDA's alternate cox models (#2, 3, and 4) that gave p values above 0.05 support the survival benefit finding?
As usual, it is easy to spin things one way or the other.>
To your second question, the statement about the sensitivity analysis was made by the FDA stat reviewer, not me. Again, if you need to spin that as a spinning argument to make a point, it is your prerogative. And my prerogative is not to reply further to that sort of base insinuation. Bye.
<His letter seemed sincere, although some of the mistakes documented here and on IV raise questions about craftsmanship. I find it very hard to believe that he is being dishonest.>
See my post on iVillage 76998.
http://www1.investorvillage.com/smbd.asp?mb=971&mn=76998&pt=msg&mid=1902510
The confusion about the p value for DN-101 was probably genuine. However, it was a bit suspicious for him to point out the imbalances on soft tissue mets and GS while ignoring the imbalance on the number of bone mets which was highly predictive of mortality in both D9901 and D9902a and went against the treatment arm. In addition, he should have known about the 6% CVAs in the phase-2 trial of Tax and how that might have translated into the phase-3 trial even if they did not collect that data. For him to make a big deal on that about the Provenge trial and the comment on 20 years down the line on an old-man and fast terminal disease like AIPC was something else. I am rather surprised that somebody like him would write such a letter without vetting all the facts. Make one wonder what sort of pressure he has been in since the meeting, either internally or externally generated.
The best way to thank David is to get a subscription to BSR.
http://www.biotechstockresearch.com/subscribe.php
Different protocols but relevant to the GM-CSF discussion...
GM-CSF as a systemic adjuvant in a phase II prostate cancer vaccine trial.
Simmons SJ, Tjoa BA, Rogers M, Elgamal A, Kenny GM, Ragde H, Troychak MJ, Boynton AL, Murphy GP.
Cancer Research Division, Pacific Northwest Cancer Foundation, Northwest Hospital, Seattle, Washington 98125, USA.
BACKGROUND: Recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF; Leukine [sargramostim], Immunex Corp., Seattle, WA) was administered to a subgroup of 44 patients in a phase II clinical trial for prostate cancer using DC pulsed with HLA-A2-specific prostate-specific membrane antigen (PSMA) peptides. Our purpose was to determine if GM-CSF caused any enhancement of patients' immune responses, including enhancement of clinical response to the DC-peptide treatment. This report compares the clinical responses to DC-peptide infusions with and without systemic GM-CSF treatment. METHODS: GM-CSF was administered by subcutaneous injection at a dose of 75 microg/m2/day for 7 days with each of six infusion cycles. Prefilled syringes were supplied to the patients for self-administration. RESULTS: One complete and 8 partial responders were identified among 44 patients who received GM-CSF, as compared to 2 complete and 17 partial responders among 51 patients who did not receive GM-CSF. For patients who received GM-CSF and were tested by delayed-type hypersensitivity (DTH) skin test, 3 cases of improved immune response were identified, compared to 5 cases of improvement in patients who did not receive GM-CSF. The main GM-CSF side effects reported were local reactions at the site of injection, fatigue, pain, and fever. Most reported side effects were of mild severity, with some cases of moderate severity leading to discontinuation of GM-CSF. CONCLUSIONS: Our results suggest GM-CSF as employed in this trial did not detectably enhance clinical response to DC-peptide infusions, or significantly enhance the measured immune response.
Corpstrat & Dew,
Thanks for the link. Will check that out.
Panel review webcast replay?
Does anyone know if there is a replay somewhere? I was on the plane home from conference yesterday and missed the whole thing. It would be great to see the dynamics of the meeting.
biologically explain why a therapy is at a disadvantage to no therapy at all?
Sounds like your reasoning went something like this:
1. The DNDN pumpers claim that Provenge is at a disadvantage because TTP did not get stat sig due to some weird delay effect.
2. But it's being measured against placebo.
3. Ergo, they are claiming that it is disadvantaged against placebo.
A link in that chain of reasoning is missing.
How long the delay effect is is a different discussion. I was only addressing your confusion about Provenge being disadvantaged against placebo.
I don't know enough biology to say anything about whether the max immune response time should be the point for when the delay effect should end. You'll be the expert there - I hope. But eyeballing the KM curves for both 9901 and 9902a, you can see the steep drop-off where both curves more or less coincide lasted about 8-10 weeks.
<So how does that apply to the TTP misses in 01 and 02A? It doesn't. Provenge was tested against placebo, not a chemo. You would have to argue that Provenge is at a disadvantage versus placebo due to the immune system ramp up time!>
That would be silly. The delay effect simply says that for 10 weeks or so that it takes for the immune system to ramp up, both arms of the trial behave in the same way. That makes it hard to get stat sig. But non-stat-sig does not mean performing worse.
<Note that of course it is certainly possible that the treated arm had proportionately more non-cancer deaths (and thus will get a better p value when non-cancer deaths are censored) - just due to bad luck. But it shouldn't be "expected".>
Perhaps I am missing something, but why shouldn't that be expected? If a drug prolongs life, you would expect the probability of a patient taking it dying of something other than cancer to improve over not taking the drug. In the limiting case when the drug is a cure, that probability would be 1. So, IMO, it is quite possible that D9902a would have similar improvement in statistics as D9901 as far as cancer specific mortality is concerned.
Baseline bone lesion count > 10 was significantly imbalanced against the treatment arm in D9902a. That might have been the strongest factor contributing to the poor p value. So yes, adding a few more patients properly randomized would not have changed that imbalance enough to make a difference to the p value.
Also the same bone lesions counts were much worse in D9902a than in D9901 over the entire ITT. This is one of the strong indicators of advanced disease state since PC has a preference to metastasize to bones. This would imply a faster progression time, earlier death, and a more exaggerated delay effect when comparing the KM curves for TTP and survival. For TTP in D9902a, the delay effect essentially wiped out any difference between the curves.
Baseline PSA and bone lesion counts.
Ok, mid Summer - was posting from memory - my bad. As to the rest of your post, old ground. This PGS's conspiracy theory that you follow will be proven or not soon so not worth debating any more.
The full Cox model was described in Dr. Small's ASCO presentation in 2005 before the D9902a completed in late Summer same year.
Dr. Martino has been off ODAC for quite a while. Dr. Hussain's specialties include prostate cancer. She was one of the coauthors of the SWOG Taxotere paper. Other notable authors of that paper were Petrylak, Tangen, Raghavan, Small and Burch. The last two, of course, were major players in the Provenge trials. It's a small club of influential thought leaders and they all likely do know the Provenge story well.
Disease progression and survival are many years out for this patient population. These endpoints in P-11 may not come in for another few years after D9902b is done in 2008/2009. Here's one study from John Hopkins:
"The median actuarial time to metastases was 8 years from the time of PSA level elevation."
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?itool=abstractplus&db=pubmed&cmd=Retrieve&...
<But if you want to find something that killed patients I'd suggest that it is the infrequency of the meetings of the DSMB.>
This is truly unfortunate as it is really easy to build triggers into these databases so that dangerous imbalances would raise alarms and the DSMB could convene as appropriate.
Did you see the part about the 6-month transition period?
The facts are that (1) the agreement was signed in Oct 2001 and (2) clause 1.4 talked about a possible BLA approval in 2003 which would dovetail with the plan at the time to file on TTP for AA. So the stricken out date in 12.3 should refer to that tentative date whatever it was. Nobody in their right mind would think that that had anything to do with the current tentative BLA approval date. With the contract ending this year, DNDN decided not to renew and that was that.
But this is Dew's typical drive-by mouth shooting mode without thinking through. When pointed out wrong as by p3analyze in this case, he would mumble some weak excuses before slinking off with the usual "Take the last word if you want." You can be sure that if he could have the last word, he would. Maybe Dew will come back some time to explain why DNDN finished the factory or filed the BLA after his prediction that they wouldn't because they were just lying about the whole thing.
<4. Changing the 9902b protocol to make Cox the primary endpoint supports the contention in item #3 above and makes 9902b the real pivotal study, in my view. (I.e., DNDN learned ex post facto that Cox gave a good result in 9902a and hence they are testing this Cox endpoint prospectively in 9902b. This is how things are supposed to be done.)>
This is a truly neat insight. I didn't know that Dendreon does research in biostatistics and is trying to get the FDA approval to sell a that to cure prostate cancer. I always thought that the endpoint for 9902b was survival. This Cox endpoint is something else. I guess we learn something new every day. Thanks Dew.
<25.9 is provenge overall median,>
I thought that's what you meant but wanted to confirm. 25.9 is the Provenge median for 9901 only. The current data is from the integrated group and that's 23.2. If you do that, the result should be closer to what I calculated, around 20. Our calculations should be equivalent given the assumptions.
What is the M=25.9? Did you mean M=23.2?