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.
I normally don't waste my time posting here anymore because of too many dunning-kruger 'experts' like flipper.
I always had you as a someone retired but expert in the field - Here just out of general interest - you're one of the few that genuinely know your shit when it comes to the stats. But my god your desperation shows who you really are (if you're even one person).
As you know but conveniently ignore you infringe on a patent if you use every component of a particular claim. Go through claim 1 of the family of dcvaxl patents. The minor differences of atl-dc do not deviate in the slightest from claim 1 of the dcvaxl patents. I.e dcvax-l claim1 describes atl-dc.
You guys are just trying to scare a few Newby investors!what a joke!
"This is what you were trying to limit the damage to in the Optune trial:
Quote:
....where i personally think novocure are being immoral however is quoting a totally unreliable 5-year survival % -Jimmy James
But then you turn around and state you agree with Dr. Liau and her colleagues (whom I agree with) were essentially correct when it was brought to your attention they said Optune does not work, and it was a poorly designed trial, but you think it just doesn’t work that well. "
What are you on about? the unreliable 5-year KM stat and whether optune actually works or not are totally different issues. There's no contradiction between having an unreliable 5-year stat and a therapy being statistically proven to work. You really show your lack of understanding of the stats when you come out with this rubbish. Same goes for the forecasting/extrapolation.
"And here on January 5 you defend the Optune censoring as essentially the same between arms and always fine as long as defined ahead of time.
Quote:
Then there's no issue so long as the censoring reasons are the same for both arms and the numbers are similar which they more or less seem to be. Also it doesn't matter so much that it's unblinded so long as those reasons for censoring are defined a priori.
"
You should put this back in its original context. Here i'll do it for you (post: 345029):
"Flipper: "The excessive censoring occurred to a higher percentage in the Optune arm, but Stupp’s SOC arm was pumped with excessive censoring as well."
Me: Then there's no issue so long as the censoring reasons are the same for both arms and the numbers are similar which they more or less seem to be. Also it doesn't matter so much that it's unblinded so long as those reasons for censoring are defined a priori. "
I was replying to your assertion that the SOC arm was also heavily censored. Maybe would have better if i had stated "if that were the case then there's no issue...". I had always thought that the optune arm was more heavily censored than the SOC arm. But actually looking at that data from page 18, table 4 of the SSED (https://www.accessdata.fda.gov/cdrh_docs/pdf10/P100034S013b.pdf) it doesn't look as bad as i originally thought. There's still clearly some bias in the trial:
1) there's slightly more disease progression and withdrawal with optune compared to SOC - that's the classic bias where heathier patients stay in the trial.
2) there are a lot of active follow up censors. The good thing is that it's a similar % on both arms. However there's no information on the breakdown of those numbers and for sure some bias can creep in there. Overall if the criteria for censoring is defined a priori this should be kept to a minimum.
But overall i think there would have to be a lot of bias to get those two arms overlapping. Hence, i think optune works but difficult to say exactly how much (for sure less than they claim) and at the same time i think that 5-year KM survival stat is meaningless. That's been my opinion the whole time only now i have slightly more faith that it works seeing a better break down of the censors.
i've had this opinion the entire time.
Rather than defining an arbitary timepoint beyond which they shouldn't be allowed to quote survival stats...i'd like to see a law where they are not allowed to quote beyond the time point where say the p-value goes over 0.05. That way a better powered trial with more patients and more events can quote longer in time than a lower powered trial even though they might have the same median follow-up.
"She said it Jimmy"
I never said she didn't
"So, a part of it is having the FDA think a little bit differently about what constitutes approval. On the converse, like the Optune device, that Novocure device, is FDA-approved. But I must say a lot of patients are like, “Well, I don’t really like it.” And I think on the physician side, we just don’t understand how it works and there are a lot of problems with the clinical trial design.
There was no true placebo arm, but they did design it in a way that met the FDA checkboxes. So, it was approved, but a lot of us don’t really believe that it works. I think it’s one of these things where if your ultimate goal is to cure GBM, it’s hard to reconcile all these different factors; the regulatory factors, the financial factors, as well as the science and the biology of the tumor itself, which is very complex."
She said it very matter of fact. I'd generally agree with everything she said. I however think it works just not as well as shown by the trial data. i'd have done exactly as the EMA/NICE did...EMA approved it..NICE refused to pay for it.
"No, but when you are conducting a seminal trial, a blinded trial prevents the dangerous benefits of excessive manipulative (real-time) censoring."
yes ofcourse...that's clinical trials 101..but it's not very ethical to make terminal brain cancer patients walk around with an unpowered helmet attached to thier head for 18 hours a day. Given the promosing early data of TTFs you would have had two choices:
1) Run an unblinded trial
2) Don't run any trial and just trash it in the garbage.
I'd go for 1 every day even though it's obviously not optimal.
You seem to be saying that because there was the potential for novocure to be biased that they were and not only that that they did it on a massive scale. Yet you have no hard evidence for that at all.
"You simply can’t handle words like unreasonably “extrapolate” or “forecast.”"
KM analysis is not a forecast either. It's simply a way of making the best of lossy data.
" Why do you think Dr. Liau would state that her, and many of her colleagues do not believe Optune works? "
When did linda liau say that?
you are basically saying that unblinded trials should never be allowed. If that were the case GBM patients wouldn't be living longer using optune now.
The key phrase is 'a priori'...that's why the SAPa nd protocol are prespecified...that's what prevents cherry picking even in an unblinded context. For sure some can creep in anyway but it would take an extreme amount to get those two KM curves overlapped. You'd need a lot of MDs in a lot of different centers that were complicit.
"unreasonably forecasting/padding long-term survival stats"
There was no issue with this. They were using standard KM analysis. They simply don't have long enough follow up to say anything meaningful about long term survival and it was immoral of them to quote that 5-year KM stat.
"The excessive censoring occurred to a higher percentage in the Optune arm, but Stupp’s SOC arm was pumped with excessive censoring as well."
Then there's no issue so long as the censoring reasons are the same for both arms and the numbers are similar which they more or less seem to be. Also it doesn't matter so much that it's unblinded so long as those reasons for censoring are defined a priori.
"Hell, they even censored if they did not have the exact date of death in an unblinded trial, yet they don’t document the number of censors there"
i presume those numbers belong to the lost to follow up category as they have to be in one of those categories and it's clearly not the others. So documented numbers at the time of approval for optune arm are =<1, =<1, =<3, =<12 and SOC arm are =<1, =<2, =<3, =<4 at 6, 12, 18 and 24 months respectively. You're not going to move the KM curves very much fiddling with those numbers.
"Are you going to put that in your letter of complaint?"
i wasn't planning on writing one.
is that what you're going to write in your letter of complaint?
check page 18, table 4. Here's a more detailed breakdown of the censors of the data used for approval. You've got fairly similar censoring % between the 2 arms. Slightly more on the optune arm for withdrawing and disease progression/AE but no where near enough to account for the differences on the KM curve between the two arms.
https://www.accessdata.fda.gov/cdrh_docs/pdf10/P100034S013b.pdf
still this doesn't change the fact that they don't have reliable data to say anything about 5-year survival and in my opinion it shouldn't be legal for them to advertise that % KM figure.
"Then we can send your graph to FDA saying don't waist your time just approve it. Case closed. Picture worth $1M"
do you think the receptionists at the FDA are so bored they need a chuckle? lol
On top of that I'd note that cancer on average kills younger people than covid19. Hence it costs a lot more life years.
I think part of the issue is effect size. If a cancer therapy came along tmrw and interim results of a p3 showed 95% 2-year survival of test patients versus 50% in the control arm (with sufficient events to be meaningful) then I could envisage an emergency use authorization. From a stats perspective it's a no brainer. Both the moderna and biontech vaccines have such big effect sizes that they are also no brainers so was easy to give EUA so long as there were no safety issues.
just out of interest sake, I guess not, but do you know if the FDA has ever given emergency use authorization for an anti-cancer drug?
"There are still those 17 missing SOC/placebo patients from this trial and their absence speaks loud and clear to me."
Don't you think they explained what happened to the missing 17 placebo patients when they announced that they'd changed the SAP and made this a single arm trial at least with regards to the primary endpoint?
"Only 16 out of 466 patients lived beyond five years."
by itself that info is meaningless as it depends on how many patients had the possibility of being alive beyond 5 years. Because of the short follow up of the trial that number was ofcourse low.
"What Optune did was extrapolate the survival of those 41 patients prematurely to assume they’d all survive until five years from time zero. That’s reckless (and wrong) extrapolation."
This is not 'extrapolation'. Do you want to know what the best word for what they have done is? it's called 'kaplan-meier analysis'...you seem to have a problem with it. People who use this analysis all the time know that the righthand side of the curve is almost always un reliable. Like i said many times even today is that this data is meaningless, the p-value for that 5-year survival would be very insignificant. It's a normal situation but the one thing that novocure have for sure done that is immoral is tout that meaningless 5-year KM survival %. That should be illegal but isn't.
"even more so because Novocure engaged in unblinded excessive censorship early on"
The company ofcourse had to run an unblinded trial, you can't make terminal brain cancer patients walk around with an unpowered helmet just for the sake of it. You really have no idea what lead to most of that censorship so stop throwing baseless accusations around. More than likely it was simply patients dropping out for not maintaining the threshold of compliance. That will most likely result in bias as it will more than likely be the unwell patients that don't comply leaving healthier patients in the optune arm. It's just the unfortunate nature of this type of trial that has to be dealt with and i trust the FDA/EMA to have analyzed the full data set in the appropriate way.
"What is the conclusion one can draw about four and five year survival in the Optune trial?"
As i've said a million times before none because the data is not reliable at that righthand end of the curve with so few events/at risk.
"They are idiotically extrapolating that those 41 patients still alive between 24 months and 60 months will somehow magically survive beyond five years and make five year survival 13%."
They are not extrapolating! They are doing exactly as i stated in post 342887. They are simply showing unreliable data...this is why we have a p-value...the p-value for that 5-year survival of 13% would show insignificance.
biosect. I'm not so familar with medical devices and the level of scrutiny that they go through. But the FDA/EMA definitely would have had the data required to explore the effect of disproportionate censoring of one arm over the other. They had the data to for 'testing' whether optune was efficacious.
Ofcourse none of this relates to the unreliable 5-year KM survival estimates that novocure are touting for which you'd need longer follow up data on many more patients.
"just because they sit behind a white coat does not mean that they are any more intelligent than their fellows"
No but in general they are more intelligent than their fellows - that's how they got the job in the first place. Plus it's not just intelligence but years of study and experience working in a specific field.
"It constantly amazes me that so many people put blind trust in all who don a white coat just because these people dedicated a few years to study."
most of these people studied at least several years and have dedicated their entire careers to their given professions. Not blind trust but i would certainly take their word over yours when they give their opinion on their respective fields.
ofcourse none of this means that people are not corruptible...but you have to deal with that everywhere.
If you're talking about follow-up data post approval then ofcourse the FDA/EMA didn't have this data when they decided to approve optune. At the time of approval they would have had all the necessary data to explore the effect of the disproptionate censoring between the arms relating to the originally defined endpoints.
and you are certainly not the person who can decide what is magical thinking...
but i'm sure the FDA/EMA reviewers have seen the full dataset...so question is... do you trust their intelligence and skillset?
"I was not using that definition"
even the more loose defintion describes it as a trend
"the action of estimating or concluding something by assuming that existing trends will continue or a current method will remain applicable."
even so you are talking about KM analysis - a mathmatical procedure - so use the more appropriate mathematical definition. Especially if you are doing to use it to take some moral highground to attack a company that's developed an approved therapy for glioblastoma whilst at the same timing assuming that all the reviewers at the FDA/EMA must be so stupid that they couldn't notice this issue.
"I'd like to think that the FDA/EMA took all this into account when they scrutinized the optune data - after all they are not stupid. ex mentioned that they'd have the full set of censor data and the reasons for each. So a reviewer would likely have reconstructed a set of new KM graphs for the optune arm via markov modelling with alternative censoring (At least that's what i would have done). I suspect that it still showed efficacy but nowhere near what was quoted in stupps et al., 2017."
Actually, i think they'd have to do this the other way around. Markov modelling censoring random samples of the control temozolomide patients that fit the censoring spectrum from the optune arm. This would bring the control arm up but again i suspect it would still show at least some efficacy. Impossible to tell without the full dataset.
Just to add to this, as i think i mentioned before it's quite likely that novocure are entirely innocent at least with the over censoring of the optune arm relative to the control arm. It would be likely that those optune patients who were not doing well were more likely to drop out of the trial through lack of compliance and hence become a censor (i can't remember the inclusion/exclusion criteria but i presume they had to use the device for a minimum % of time to be included). Hence they've probably ended up with an optune arm full of better prognosis patients to start with through no real bad intent.
It's similar to their finding that increased compliance with optune led to increased survival in GBM - it's inherently biased as those who are feeling well are more likely to use the device.
I'd like to think that the FDA/EMA took all this into account when they scrutinized the optune data - after all they are not stupid. ex mentioned that they'd have the full set of censor data and the reasons for each. So a reviewer would likely have reconstructed a set of new KM graphs for the optune arm via markov modelling with alternative censoring (At least that's what i would have done). I suspect that it still showed efficacy but nowhere near what was quoted in stupps et al., 2017.
where i personally think novocure are being immoral however is quoting a totally unreliable 5-year survival %
Better to use the mathematical definition for extrapolating if you're using it in a mathematical KM analysis context:
"the extension of a graph, curve, or range of values by inferring unknown values from trends in the known data."
The key word there is "trend"...extrapolation is about derivatives. What you are calling extrapolation does not involve derivatives, it is simply a translation moving the starting point of a curve in the y-axis (in the case of the optune arm translating the KM curve less down due to the censors)for a given range in the x-axis. Those multiple y-axis translations that are performed over the x-axis (for each additional and suspected unwarranted censor event - still no hard proof of that but i also suspect it) have the effect of making the optune arm look higher in the KM graph than it should have done without those additional censors...but there is no extrapolation.
" NWBO will, on the other hand have five year data on all patients from their respective dates of enrollment excepting a very small and reasonable number of LTFU."
it's still a KM estimate albeit a vastly more accurate one compared to stupps et al., 2017.
And there's no extrapolation in KM analysis....one day you will learn....maybe...
as we've been over a million times
1) Novocure simply don't have the data to say anything about 5 year survival either way and they shouldn't be allowed to say anything about that...that's false advertising.
2) There's nothing wrong with heavy censoring in a trial so long as it's done with the same criteria for both arms. Clearly it wasn't for stupps et al., 2017 as there are high % of censors on the optune arm relative to the control arm and it was open label.
3) Flip doesn't have any hard evidence for what he claims. Simply seeing more censoring on one arm relative to the other doesn't necessarily mean anything untoward happened ..though i agree it should raise alarm bells. Overall if i were the FDA/EMA i'd want to see the breakdown of the reasons for each individual censoring event for both arms - i'd want to confirm they are more or less consistent between the two.
I'd add to this that my personal suspicion is that the stupps 2017 result shows an exageration of the effectiveness of the therapy. More than anything because the arms diverge way too quickly on the KM curve...literally from day 1. That reeks of bias to me...
Overall personally i would have done exactly as for example the EMA and NICE did. EMA approved it and NICE refused to fund it.
BSB the location of the editing offices of the journal are irrelevant. All that matters is it's standing in the academic/medical world and the lancet and the NEJM are the top dogs for clinical research. The lancet will make their decisions totally independently of the MHRA. Anyway like Vubru was saying...not meeting those orginal endpoints doesn't make it impossible to publish there especially if they can show data to explain why those endpoints weren't met (in a journal they can be more 'exploratory' with the analysis unlike when formally asking for approval)...just that bit harder.
"It may may have taken them some time to identify and agree on the right comparison trials and construct these survival curves (I'm not sure sure whether they had easy access to the raw patient-level trial data from other trials)."
yes i think that raw patient-level trial data from other trials is exactly what they need to be taken seriously. Hopefully they can do something like i did with the estimation but substituting the DCVax-L blended data with the real test-arm data (and obviously more stringent handling of the historical comparison and no need to generate a markov-model):
http://jammyjames.com/analysis.php
but as you say i don't know how easy it will be for them to extract these specific details from other trials.
"As I have said in prior posts, if their original planned PFS and OS outcomes (DCVAX vs. the internal placebo group) are not significant and the only significant findings are the comparisons to historical controls, getting this accepted by NEJM or Lancet may be challenging"
yep and exactly the same likely goes for approval.
"Once the company has an acceptance letter in hand, they could in theory announce TLD in a PR saying the paper has been accepted by XX journal, and provide some additional context in the PR based on what is written in the article."
exactly what i suspect. Though I don't think they'll wait for a preprint.
sorry if that sounded harsh before. Didn't mean it to sound like that. Nothing personal... i just bang my head against the wall everytime i see someone compare this trial population to the general trial population...apples and oranges. We've been over it on this board a thousand times just in the last 3-4 years that i've been here and i imagine even more before.
True. My point is that they were all intent for gross resection. They'll have removed much more tumour tissue than a typical biopsy that's for sure. The point is the prognosis of the selected set of patients for the trial should be significantly better than the 15-17mOS for the general population. Been discussed a thousand times but people still keep saying the same thing.
Almost all the patients in the trial had gross resection. What's the mOS and 5 year survival rate for patients with nGBM with gross resection...it's not 15-17 months and 5% I can tell you.
No patient ever wants to be in the control arm in any trial. Why would they be in it in the first place?
Well I was considering being tax resident in Gibraltar for 6 months and 1 day so I can use the money for my own cancer research but maybe there are better options with more to see than just monkeys climbing a rock!
Isn't your capital gains just considered as normal income? And taxed via income tax? I always thought that's how it worked in Belgium and the Netherlands but maybe I got it wrong.
There's no substitute for mature data...go away!
not bad but not quite as good as the DCVax-L blended data at 18 months.
It's a fairly comparable patient population as they also have mainly gross resection since the exclusion criteria states:
"Presence of greater than 1 cm x 1 cm residual tumor enhancement on postoperative MRI"
https://clinicaltrials.gov/ct2/show/NCT03491683
ahh yes i'm just exploring the different types of continuation patent. I'm wondering why they did this from a strategic point of view. I wonder why didn't they include this as a claim to start with given it was already supported by data in the application (i.e. not a continuation in part by the look of it).
interesting...not sure how it works with expiry dates with a continuation patent grant. Do all the claims in the continuation patent have the new expiry date? or only the new claims?
It's duffy!
well i'm only expecting a reasonable sized benefit due to the inclusion/exclusion criteria as i've shown before:
http://jammyjames.com/analysis.php
that's before we start taking into account any dilution of signal due to cross-over. I'm hoping it comes out stat sig due to the length of the trial (and that's why they chose to go way past the original designated stop point) and hopefully adapted the testing criteria (again as i've mentioned previously i'm hoping the new SAP includes something like a fleming-harrington(0,1) test instead of a basic logrank test).
so no if that's the case the trial wouldn't have ended.
it's not the p3 data though that's for sure.
1) The P3 data would include all the authors from all of the different sites.
2) If the unblinded P3 data is as good as most people hope it'll be in NEJM or the Lancet...Nothing less.