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.
True. Margin of error is large and risk large. But I comfort myself with the fact that the risk ratio around 0.52 has room to grow and still produce statistically significant difference - HR less than or equal to 0.75 is usually good for significance in trials having about 200 subjects per arm.
Agreed! And no sane big-pharma CEO will commit to an acquisition before he/she knows everything there is to know. Overpaying an acquisition is fast way of taking the shine off from one's reputation, almost as fast as buying a failing drug or two. See Allergan and Vitae: VTP-38543 and VTP-43742.
These things take time when done right.
DaubersUP, well, I am, in principle, with Infinity on this one. Apples to apples comparison are to preferred and Co's should try to provide them, if possible. They don't and I must go digging data - I rather sit on my patio bombarding deer eating my sweet young siberian spruces with empty beer cans.
Very true, G.
Notice AF's art in full flight: He makes alarming noises about enzyme elevation of 2 times Upper Normal Limit (UNL). In fact, 2 * UNL is typically the lower limit for grade 1 AE for liver enzymes - mild. I have to assume that after all these years AF knows that, otherwise he is a hopeless ingrate. No wonder that The Street turned comments off towards the end of AF's tenure there.
I can get my enzymes over 2 *UNL with three shots of rye whiskey night before blood test. Cheap fun, pisses off my doctor every time. One can get there with several 'benign' medications, including over the counter stuff.
Baseline: there was one severe, grade 3 or 4, AE and it was in 50 mg dose group. Conclusion: not dose correlated - overall liver enzyme effects mild. But when you want to disparage something and don't have to face any consequences ...
Well, Infinity, actually they did tell us something.
I just takes an effort to lift ones mental ass off from the whiny chair and do something called estimating. Lemme show you how:
We know that there were 6 subjects (0.462 * 13 = 6) in prurisol 200 mg group with IGA 3 that had 2 point drop. Numbers for 200 mg ITT group were 28 subjects and 18 of them were IGA 3. Let's be stingy and assume that the seventh 2 point drop in ITT did not happen in IGA 3 group. We get:
prurisol 200 mg ITT IGA 3 - 2 point change percentage: 100*6 /18 = 33.3 % (at least)
Placebo group had ITT of 30 out which 23 was IGA 3. This group had 4 subjects with 2 point IGA drop. Let's be generous and assume that all four happened in IGA 3 subjects. That means
placebo ITT IGA 3 - 2 point change percentage: 100*4/23 = 17.4 % (at most)
Risk ratio associated with 2 point drop in IGA 3 groups: 0.52 for prurisol (and probably better). A number that is known to imply clear statistical significance in 100+ subjects trials.
BTW:In Esteem 2 trial Otezla scored PASI 75 percentage of 28.8 % when calculated using LOCF. Prurisols results in p2a are probably better than that. Strange: It looks like Leo did not lie, but famous Adam F. might have. Who would have thunk THAT!
Nothing burger? Hardy.
IPIX presentation was yesterday 4 PM local time. After that everybody went for burgers and beer etc. If there is going to be any productive meetings they will be today and we will not hear any specifics unless/until a contract is signed.
Hi Sox thanks for the insights. And, of course, I am going to pipe in.
Some time ago I tried to find out if there is a way to map IGA scores to PASI scores. Turns out that there is until there isn't. Based on 'vast' sampling (is not like this is a hot topic is psoriasis research) of two studies covering some 1700 subject very lopsided, one had about 1550 subjects, the other had about 150. The bigger study included subjects being various lengths of time in treatment - useless for my purposes.The smaller study (typical, aargh!) had a scatter plot IGA vs PASI. Using that it turns out that at the extremes mapping IGA to PASI works - elsewhere not so much.
A rough mapping of IGA to PASI
IGA clear and almost clear (0/1) was always PASI 3 or less.
IGA mild (2) is between PASI 3 and 12
IGA moderate (3) varies from PASI 3 to 20
IGA severe (4) almost exclusively PASI > 20
Looks neat and clean except, of course IGA 3, which looks problematic.
From the scatter plot I could figure out that about 50 to 60 % IGA 3 values where in the region PASI > 12, which is the inclusion limit for the new IPIX study. PASI 12 is interesting number. If a subject has IGA 3 with PASI > 12 then a 75 % drop in PASI does not necessarily require 2 category drop in IGA (from 3 to 1 or 0), but 2 category drop in IGA will be at least PASI 75 drop! The opposite is true for IGA 3, PASI < 12, two category drop in IGA does not necessarily translate to PASI 75 or better.
What is all this weird IGA PASI, PASI IGA gibberish? Well, we can infer some pseudo-probabilities:
In my reference trial about 50 to 60 % IGA 3 values where in the region PASI > 12. Also, in the completed psoriasis trial IPIX observed that 46.2 % of IGA 3 subjects had 2 category drop. Assume that the PASI value distribution in the reference trial is typical. Then, based on the distribution of IGA 3 in terms of PASI it is slightly more probable than not ( 50 to 60 % vs 40 to 50 %) that in IPIX trial the percentage of IGA 3 subjects experiencing PASI 75 drop was at least the same as the number of subjects experiencing 2 point IGA drop.
Minor good news.
Beware!: All my reasoning is based on ONE small study.
Thanks for insights.
On lighter note: I have followed another well pummeled company (for altogether different reasons. Let's put it this way: people on that board would be crying out of sheer happiness, if by some miracle a person like Leo would materialize as CEO) namely CytRx. Between management shenanigans they have managed to make old doxorubicin much more safer and effective - the results is compound called aldoxorubicin (now licensed to NantCell)
Why not combine Kevetrin with something like aldoxorubicin and see what happens. There might be some panic attacks at pricy cancer medicine circles.
Thanks KMBJN, I was aware of COTI-2, but did not know about Critical Outcome's difficulties. Good, because if one trusts Critical Outcome's press, COTI-2 is the one to watch. But, isn't that always the case with press releases. I have yet to see one that would tell the drug is mediocre at best but our Co is going to push it thru regardless in hopes that some nincompoop or extra income friendly doctors will prescribe it.
PK11007 is totally new to mean. Thanks again!
True, There is a difference.
Aprea says APR-246 reactivates the mutant type p53 and is mum about the wild type. IPIX says Kevetrin degrades the mutant type and activates the wild type.
But, simple me just can't figure which one is more beneficial based on this difference. It looks like ARP-246 would be fine only in cancers with high expressions of the mutant types and Kevetrin might be more widely applicable but in certain cancers not necessarily as potent as APR-246 can be. Dunno, been wrong before, maybe again.
However, what is getting clearer to me is this: Whichever gets first on the market can hardy keep the other out. They are too different to be covered with FDA's exclusivity rules. So, whichever company will be the first, the other will applaud loudly. Money-promise for both :)
Understood. Thanks.
Clarification: ARP-246 is in P2 not in P3, not so far ahead of Kevetrin.
APR-246's success, if it is the first P53 targeting drug on the market, will also be Kevetrin success, provided that Kevetrin keeps showing good results.
Have you ever heard me-too drugs? A new compound gets approved for a lucrative market and immediately about half a dozen drug companies are working on how to circumvent the existing patent. The fastest and usually cheapest way to come up with marketable me-too drug is to buy it. Hence, if Aprea posts successful P3 trial its value will go up and so will IPIXs value, at least in the eyes of the companies wanting their share of P53 market. And it looks like (I am not 100 % certain about this) there is a bonus associated with Kevetrin: who ever might end up with ARP-246 rights in US will have hard time in blocking Kevetrin's entry based on exclusivity - APR-246 and Kevetrin are different chemical entities.
If all goes well with APR-246 it probably will be the first p53 targeting drug on the market. That, however does not guarantee market dominance.
TradingPro, funny that you used golf analogy - Kaplan Meier estimate with Log-Rank test is very much like analysing a golf game hole by hole. Kaplan Meier describes the events during the course and Log-Rank test sums up the score event by event and in this case came up with the clear winner - Brilacidin.
However, you are correct in pointing out that I can only guess what the censors stand for and or what happened in OM duration study - besides concurring with IPIX that the duration study results are non-informative.
But I can say with mathematical certainty that only one censor had some (near infinitesimal) effect on the outcome of OM incidence study, which is unassailable affirmation that censoring WAS NOT USED to manipulate the outcome.
And that should be good news to any long side investor - Log-Rank provided confirmed clean hazard ratio of 0.45 favoring Brilacidin over placebo.
Nice! Where did you get your numbers? In case I am, again, missing something obvious.
FDA usually allows low populations for cancer trials and most of the time a single trial seems to be sufficient. My guess is that OM trial would fall into that category. In than case around 200 subjects per arm should be enough.
I see the light...
I guess you think I am in some nefarious way trying to undermine IPIX OM trial. Nope! No Way! I am actually very pleased with the trial results (except some whining about Leo not telling who were excluded from the final PP population).
But, few days ago, I DID react too hastily to censoring numbers. Since then I have concluded that whatever the censors do present, all except 1 of them have no effect on Kaplan Meier estimates. But because I already had put out some faulty info, I felt it was my responsibility to show reasons why I changed my mind. Done that.
The meaningful portion of censoring is minuscule and hardly could be better balanced. Period.
In other words: all is fine on censoring front. Period.
Trial results are very good, as far as I am concerned. Period to that, also.
Correct. And does not change anything I said. You are better off in understanding the mechanics of Kaplan Meier estimator by looking at its formula.
K(i) = K(i-1)*{1-n(i)/[N(i-1) - n(i) - c(i)]}
and
N(i) = N(i-1) - n(i) - c(i)
where
K(i): Kaplan Meier estimate at then end of period i
N(i): number of subjects at risk at the end of i:th period
n(i): number of events during i:th period
c(i): number of censors during i:th period
the term multiplying K(i-1) stays at 1 as long as there are no events, n(i) = 0, during the period. No change in estimate (should be intuitively clear, KM estimates failures like deaths. Censor is NOT a failure, so why should it register as one!).
Subject at risk, N(i), counts down with every event and censors. So, you can censor as many subjects as you like and they will not affect the value of Kaplan meier estimate as long as there are no subsequent events (one will do).
Yes, it did. Excluding subjects with less that 55 Gy of cumulative radiation was fair and proper leveling of playing field. It has been done by others, also. But the exclusion after that.
What was THAT FOR? Being mum about it does not help Leo at all.
Some clarification,
This is my take on consensus practise:
SOM developing during radiation therapy is considered to be caused by radiation therapy as is SOM developing during the 4 weeks following the radiation therapy. In trials for SOM the limit for 'eligible' SOM is therefore set to be 11 to 12 weeks. IPIX used 11 weeks. For study purposes SOM developing after 11 weeks is considered not originated by radiation and is excluded from the study. Hence, what's the point of following subjects that have not developed SOM by the end of week 11. No point. These kind of subjects are usually censored at the next scheduled visit. This is common practise in trials with time limits. And usully called admistrative censoring.
Proper administrative censoring does not have any affect on Kaplan Meier estimate, because it is executed after the last event/failure, and censor has an effect to Kaplan Meier proportion only if it has an event/failure follwing it. As for possibel effects on other proportional measures - those are done against intent to treat or per protocol population and censoring does not remove subjects from those populations, so - no effect.
Administrative censoring was just a hypothetical throwaway thought. But then I started thinking (having late starts, lately. Lamentable!).
The trial was planned for 7 plus 4 weeks. 7 weeks for the radiation treatment period then 4 weeks for additional incidence observations. That is 11 weeks or 77 days total.
Placebo Kaplan Meier curve runs for 88 days, Brilacidin for 86 days. Why? The simplest explanation is (to me. Feel free to disagree) that severe OM incidents were followed until they cleared in order to get correct durations. That would mean last severe incidence in placebo cleared on day 88 and in Brilacidin on day 86.
Brilacidin has 5 censoring marks on or after day 77 which is presumably the last day to observe severe OM incidence, placebo has 3. These probably present subjects that did not have severe OM incident before day 77 and are therefore administratively censored as subjects that don't count anymore neither against incidents or durations.
Ofcourse, I can't say for sure. On the other hand, maybe they are subjects that got tired of swishing after radiation was done and just walked away from the trial. I don't know. Maybe you can come up with a more plausible explanation.
scottsmith, I agree totally.
I am an optimistic long on IPIX. Still, I agree that any conclusion based on current facts should be associated with the preamble: If these numbers hold in a future, larger study.
But, that is the case with any small trial, with any small biotech company. One should always try to somehow estimate the risk associated with the assumption that the results are repeatable (Or not repeatable, logic should apply to the short side, also. But looks like it seldom does). And then act accordingly.
In my case I have concluded that in the long term the risk being long is significantly less than being short. Others may, and are free, to disagree. A well argued short side position is worth hearing because there is always a good possibility that I have overlooked or misinterpreted something, hence I might be wrong. High risk of being wrong is in the root of investing in biotech and also the reason you seldom see me talking about buying or selling. It is not my business to make that decision for anybody else.
An Educational Science Story plus Optional Science Fiction Story
The Science Story
Fact number 1: Typical duration of radiation treatment in head and neck cancer is about 50 days.
Fact number 2: In OM trial 93 % of severe OM incidents in placebo arm occurred ON OR BEFORE day 40
Fact number 3: In OM trial 71 % of severe OM incidents in Brilacidin arm happened ON OR AFTER day 40
Fact number 4: For the typical radiation treatment period hazard ratio for Brilacidin over placebo is better than 0.45.
Fact number 5: A bulk of censors happened well after the radiation treatment period.
Fact number 6: The risk of developing a severe OM drops drastically in few weeks after radiation treatment.
Fact number 7. Brilacidin downregulates hTRPV1 ion channel associated with sensitivity to capsaicin and heat.
Fact number 8. Brilacidin upregulates hTRPA1 ion channel associated with sensitivity with cold.
The Science fiction story in case you still are fussing about censors and just for fun.
Patient Creamy: “So the hell weeks are over – no more radiation, just the monotony of swishing brilacidin and dripping chemo. Oh man, I am getting an urge for ice cream.But, the last time I ate it, it was hurting cold. Yeah, must be that swish stuff. To hell with swishing, I am going to have myself a quart of Haagen Dazs Mint.”
Patient Hotty: “I hear you Creamy - you go for it. Did I tell you my brother-in-law is one full sized sadistic moron. Can you believe this? He managed to fool me into eating chili peppers just few weeks ago - with all those freaking ulcers still in my mouth. The strange thing – it did not burn that much. Mr. Moron looked disappointed. Maybe it was that swishing stuff. Dunno, but as long as the family sadist keeps visiting I keep swishing.”
Karin, the more I think about the censoring in OM trial the less concerned I am. One thing the imbecil in me ignored for a long time was what were the times for censoring.
In both arms all but 1 censor mark in each arm are after week 10 - near the end of observation period and probably after radiation treatment was already done. If memory serves (too lazy to check) the typical duration of radiation treatment in head and neck cancer is about 50 days or 7 weeks to maximum 72 Gy cumulative dose.
At this point I am leaning towards pure coincidence or even administrative censoring ie. they may have been censoring subjects with longest time without severe OM incident in order to accelerate the study. When the bulk of censoring happened they had placebo arm median 40 days vs Brilacidin arm median at least 70 days, if forced with an event(s) after censoring. I would call that a clear cut case.
Whatever was the reason for censoring, censoring had a minimal effect on the outcome of OM instance study. Only thing the censoring did was make already statistically highly unreliable duration study unreadable. I like to leave it that.
Yup, KMBJN, it is confusing without seeing what actually is in their protocol.
You are correct, from 61 randomized to 46 is due to 55 Gy minimum cumulative radiation, which is OK. What is 46 to 39? Angry Canadians? No idea, and I don't like it. If it was in the protocol agreed with FDA there should not be any problem making it public (blame FDA for letting you to do it). If it was not, they should call it additional fudging in order to look good.
Anyway, these population reduction are usually called exclusions, dropouts are the same as censors ie. those little balls in Kaplan Meier curve. So, dropout rate = censoring rate.
For a such small trial: LOOKS GOOD
Just for fun I run Brilacidin severe OM incidence Kaplan Meier data (it is easy to reverse engineer) thru statistical analysis. In case you are wondering what this small trial tells. Here is
Brilacidin / placebo
Log-Rank p-value: 0.0772, just a tad shy from significance
Hazard Ratio: 0.45, 95 % confidence limits [0.18 - 1.11]
As I have said before with small trials it is better to look at hazard ratio, anything below 0.75 is usually good for significance. Assuming, of course, that it survives in bigger trial.
KMBJN, You have no idea how confused I occasionally am about censoring. As far I know there does not exist a good, comprehensive treatment of censoring. A lot of stuff covering some parts, yes.
I am not certain that I have answers to your questions, but the lack of knowledge has seldom stopped me from opening my mouth (ask my wife) - here goes.
Intent to Treat (ITT) population:
all subject that were randomized for the trial. Any reduction to it is considered per protocol - even excluding subjects randomized but not treated. In OM trial case: 61
When you talk to FDA you need to have data at least and always for this group, which means IPIX will present that analysis to FDA at some point of time.
Per Protocol (PP) population:
Treatment for OM and radiation for head and neck cancer are concurrent. Hence you can't put minimum cumulative radiation dose in inclusion criteria for ITT, which is always before any treatment. Solution:
Employ per protocol group: cumulative radiation at least 55 Gy.
1. That way you have analysis that definitely addresses also radiation effects, not only chemo effects.
2. Number 55 Gy specifically is probably to make result comparable with Dusquetide, which had the same per protocol group in phase 2.
3. I suspect that FDA had its finger on points 1 and 2.
In OM case per protocol analysis is probably more important to clinical practitioners (and hence to pharma cos) than ITT analysis because it is THE ANALYSIS for high radiation group.
Good Luck.
Hi KMBJN,
Figuring out censoring pattern is almost always a mess. Anyhow, here is a short treatment that applies to usual time-to-failure Kaplan-Meier.
I use word failure instead of the usual word 'event' to describe what is tracked. And I use word 'event' as the general name for the two possible outcomes: failure or censor. The meaning of failure is probably clear. One way of explaining a censor is: Censor denotes an event after which the subject cannot be objectively included as a failure (present or future) in the analysis - there are several causes for censoring. Administrative censoring is a special case of censoring and is usually done at the end of trial at time AFTER all analysis eligible failures.
Some rules:
1. A subject can have only one event - failure or censor, not both
1a. follows: censored subjects cannot fail
1b. follows: failed subject cannot be censored
If either one 1a or 1b is evident somebody is fudging the big way.
About Kaplan-Meier curves
1. A censor has an effect on curve and analysis only if there is at least one failure with later time.
2. One censor mark may be for multiple censors
3. One failure mark (usually a vertical tick) may be for multiple failures, but often one can figure out the number of failures by observing the associated length of the drop in the curve.
4. Typically, the last censor mark on the curve is for all administrative censoring done at that time.
5. Uncertainty in Kaplan Meier estimate increases with censoring. When censoring goes over about 45 % the estimate is nearly worthless.
We, without the underlying event table for KM curve, are reduced to statements like at least x censors, or at most x censors. Posted at risk numbers at marked times will help, but will not resolve uncertainties completely.
Applying the above on Brilacidin severe OM incidence curve.
Number of subject 19
Number of severe OM incidences 7
Number of non-administrative censors 7
No median reached
What does that imply for Kaplan Meier curve for durations of severe OM.
1. Time zero for the curve is time when severe OM was first observed
2. Failure time is the time after the last observed time for severe OM
3. Subject censored in instance Kaplan Meier curve are censored at time zero in duration curve, except administrative censoring at the end. --> at least 7 censors.
4. Only 7 subjects had severe OM --> Can't be more than 7 durations --> median can be reached only with additional censoring
5. Any additional censoring makes resulting median unstable and very inaccurate, the same is true for all points after censoring exceed 45 % level, possibly even before median --> statistics with additional censoring will be unreliable.
Sometimes you can, still, see median forced into existence with censoring. Say, Kaplan Meier curve loiters around 60 % and time is running short for the trial. Solution: staged 'administrative' censoring. At time A censor all subjects that have been followed at least the current Kaplan Meier time plus some fixed duration. This censoring will make any failure drop following it larger that it would have been otherwise. So, did the next event or two have big enough drops to force the curve into median. If so, done, else repeat treatment at some near future time. The only time when you might see FDA accepting this practise is when it is done with a curve that was statistically superior before any additional censoring ie. "we know the winner already, let's stop wasting time and money" situation.
I guess that is enough, for now.
Sox, I suspect the same as you do, especially when adverse effect I know of are from Polymedix ABSSSI trials and don't apply to oral rinse.
One far fetched though thou.
Do you have any idea how and when opioid painkillers are/were administered? I had this off the wall thought that getting opioids may be linked to stage of OM. Say, not anymore so dear doc keeps saying: "you need have stage 3 before I give you opioids". Brilacidin works just enough to keep you teetering at upper limit of stage 2. Plain is bad, but no opioids.
Well, I know I am weak - I would walk out and register to another trial if that would mean getting potent painkillers. But that's just me.
Good point! Placebo was water, we don't know how Brilacidin solution feels in one's mouth.
Have you ever rinsed your mouth with hydrogen peroxide solution? How about when there was a sore in your mouth? If B solution produced something even close to that feeling with those multiple sores from OM ... that added to pain from OM probably would make me to say F. T. and throw the stuff on to wall.
KMBJN, Trying to answer your questions (feebly),
The problem I have in trying to reason out causes for censoring is the fact that most of the info for side effects and adverse effects with Brilacidin is from Polymedix ABSSSI trials ie. for systemic Brilacidin not for topical. Totally different beast. So, it will be a guessing game.
Keep in mind that the censoring rates between the arms in OM trial ARE NOT STATISTICALLY DIFFERENT. It is very possible that the difference is just dumb coincidence. That's about as far I risk to go without seeing adverse effect tables.
An example of dumb coincidence: when Temsirolimus was on pivotal trial for treatment of kidney cancer there was about half a dozen deaths at early part of temsirolimus treatment, maybe 1 in control arm. FDA got very alarmed and investigated. If I remember correctly, no early death in T. arm was found to be treatment related - all were censored as unrelated deaths. So, it was just dumb, if deadly coincidence. These things do happen.
What test can be used when one arm does not reach median? Answer: usually usual Log-Rank statistics and hazard ratio will still work. But there is one big caveat - about which later.
Maybe I should try to explain once more Leo's situation.
Kaplan-Meier is a presentation of cumulative survival function. It is pointwise exact as long as there is no censoring. Technically every subject in Kaplan Meier estimate experiences either an event or is censored, neither can experience both. It is either - or. Heavy censoring happens at the end of trial when subjects still in follow-up are censored, because a censor has a statistical effect only if it is followed by an event, this end of trial censoring does not have an effect on statistics.
Here it gets complicated. With censoring Kaplan Meier becomes estimate, more censoring, wider the potential error. I have stimulated what happens when censoring rate approached that of events. At censoring to events ratio above 0.9 estimated median starts to increase - becomes unstable. At censoring to events ratio above 1 reported median is completely unreliable and useless if it is even reached.
There is a reason for this: When censoring goes up events further down in the future get included in estimated median. Kaplan Meier compensates for this by increasing the size of the drop each event causes when censoring is present. At censor to event ratios above 1 this compensation mechanism fails radically.
The commonly reported censoring rates are different from my censor/event ratio. Censoring rate is number of censored subject during the trial divided by the intent to treat population and is given as percentage. But, 50 % censoring rate means that censor to event ratio must be above 1 and the median based on Kaplan Meier is crap. I would say that it is often fruitless to go to FDA with a trial having censoring over 40 %. You will have hard time over there.
Back to Leo. there was 19 subject in Brilacidin arm, 7 evere OM instance events and at least 7 censors before the end of trial (I ignore the last censoring mark as administrative censoring at the end of trial). Because nitwits er.. statisticians... do insist of calculating the median based on ITT, Leo did not have median time for Brilacidin incidence neither for duration.
Worse, I have a guess how Brilacidin duration curve behaved based on times of occurrence of the censors. First Kaplan Meier estimate for duration was showing that severe OM duration for Brilacidin was shorter than for placebo - looked good. Then censoring in brilacidin arm kicked in and longer and longer durations became 'eligible' to be included in pre-median curve. Kaplan Meier estimate for Brilacidin duration shot out, maybe even catched placebo duration. What do you say? Exactly what Leo did:" It looks like ..."
There is additional misery: One known instance where Log-Rank statistics becomes unreliable is when instantaneous hazard functions do cross. Cox constant hazard ratio model is not a savoir, it does the same. Guess what happened with Kaplan Meier estimates in the case I just described? I am 100 % sure that this was pointed out to Leo. If it was not, he needs to get better statisticians.
Censor is Not AE and I made a mistake
Censor is anything that takes a subject prematurely off the study. There usually are some wacky ones.
I have seen at least one of each these reasons for censor
+ moved out of state in order to hide from his wife
+ got arrested
+ killed by neighbor
+ stung by a bee (severe allergic reaction)
+ developed an intense dislike of the treating doctor
+ took a vacation trip abroad, got rolled and is waiting diplomatic travel back home
+ did not have the cancer trial was intended to (some cancers are hard hard to distinguish)
All except the last two usually go under the plain Subject Withdrew Consent category. The last ones go to another wide net - The Protocol Violations category. Violations covers usually everything from crooked doctors to tardiness and esoteric home remedies.
A sad situation, actually. Study reports could be so much more riveting reading if every censor was listed as it happened.
None of the above can be considered as Treatment Emergent Adverse Event. But adverse events can give you clues why subject may have withdrawn their consent for. For instance, if you seen high percentage of grade 3 or worse febrile neutropenia expect some extra walkoffs. In general, high severe adverse event rate begets high walk-off rate.
But, when it comes to Brilacidin arm I don't have a clue what might be causing drop-outs. Maybe it is just a coincidence. And in any case, when you evaluate against original ITT numbers (placebo 20, B 19) censoring rates are still in the same ballpark:
placebo 5 /20 = 25.1 %
Brilacidin 8 /19 = 42.1 %
My first calculation used censor to event ratios, and may have been a bit misleading in this setting. Let's see if get some details, some time.
About censoring, and why I worry.
Brilacidin seems to reduce the incidence of OM. Also, and very hypothetically, some of brilacidin 'side-effects' like inhibiting certain ion channels might actually reduce sensation of pain. Logically, one can conclude that patients would stay on brilacidin treatment better than in placebo treatment.
The censoring numbers in today's pr do not support that assumption.
Eyeballing:
placebo: 12 events, 5 censor times. Ratio 5/12 = 0.42.
Brilacidin 7 events, 7 censor times. Ratio 7/7 = 1.0.
Drop-out rate in placebo arm is plenty less than in Brilacidin arm. Why? If the cure works one would assume subjects want to stick around. Can there be some side-effects that make patients want to drop off? Tingling and numbness are intolerable for some?
There are, at least, couple of things that can throw off my mini-analysis.
1. Nothing dictates that a censor marker stands for a single censor, it can be for multiple censors. My numbers can be skewed to any possible direction.
2. It is a small trial, small coincidences tend have big effects.
I would love to see something about adverse effects and censoring.
Thanks Sox, especially for the great link!
I am not glossing over duration,
I suspect that the difference between duration medians is not that great and I also suspect that due to heavy censoring just before the end of observation period, the medians, generated by that censoring, are not stable. So, not much point of talking about them in phase 2 trial.
Let's put a bit of irony in. If Leo had given some numbers for durations and the numbers would show nice benefit for Brilacidin how likely it would be that types like AF (or, possibly, you) would immediately use the above argument against him? - Leo is peddling nonsense statistics, as usual.
I think what he said is fine: it looks like there is a benefit also in durations, but can't be sure.
If you really want to worry about something, worry about this: Why it looks like there is heavier censoring in Brilacidin arm than in placebo arm? Shouldn't it be the opposite? That worries me.
Chill Off! Data looks good!
I have this strange habit that years of savage ridiculing on all sorts of boards have not been able to cure: I look at the results and I think about what I see, then I scream if warranted. Here the preference seems to be the opposite: scream first then think if warranted.
If you take a look at the Kaplan Meier curve tracing the time of incidence for severe OM provided in today's PR you should notice that due to censoring towards the end of observation period (after week 10 or day 70) THE CURVE WILL NEVER reach median in Brilacidin arm. The same will be true for the median duration of severe OM in Brilacidin arm.
And while the censoring may have a minor effect on incidence time median for Brilacidin (treating censors as indicidences would put median somewhere after 70 days, and THAT IS THE WORST CASE SCENARIO) they have more profound effect on severe OM duration.
Because Leo does not have stable medians for the durations at hand he can't say definitely what is the difference in durations of severe OM between B and placebo. Leo made an honest statement, if you ask me.
Otherwise the addition of time to severe OM info tells me this rather OBVIOUS THING:
Brilacidin has now beaten Validive (link to offcial, nicely formatted data) and Dusquetide (link to MS word doc) in ALL MEASURES WE HAVE COMPARATIVE INFO:
Incidence rate:
Brilacidin vs placebo : 36.8% - 42.9% vs 60 %
Validive vs placebo: 42.6 % - 47.6 % vs 60 % (excluding missing values)
Dusquetide vs placebo: 67 % - 79 % vs 74 %
Median time to incidence of severe OM in days:
Brilacidin vs placebo : Not reached! vs 36 days
Validive vs placebo: 42 - 45 days vs 36 days
Dusquetide vs placebo: 41 - 59 days vs 42 days
BTW1: In case some of you are planning to went about censoring and the time to onset. Don't bother. As I said: Treating censors as incidences would STILL put median for B beyond 70 days. QED
BTW2: Both Validive and Dusquetide are in P3. What should a big fat CEO of a big fat pharma co to do about Brilacidin?
BTW3: The median for duration of severe OM has never been reported for Validive. For Dusquetide it was reported to be 9 - 14 vs 18 days days (placebo), in a study where the duration of OM was driven by subjects treated with cisplatin every 3rd week - in placebo group median for them was 30 days, What it was for subjects treated with weekly cisplatin? Oh, not reported. Should we wonder why?
This may explain the weird conundrum bounce with lousier results vs no bounce: OPRX-106 has the advantage of being orally administered. This in spite of Brilacidin posting nearly double the remission rate of OPRX : OPRX 36 % at week 8, Brilacidin 66 % at week 7 (calculation for B done with the numbers from IPIX slide).
Quotable of The Day: People without severe UP do dislike enemas. People with severe UP probably less so.
Hi G., I do understand. The stress is on 'much higher' then ...
I was more like stating my belief that 150 mg BID will show statistically significant difference.
Correct,300 mg vs placebo should show significance.
That 28 subjects in 400 mg looks like FDA's touch. They have done this before. They make you include a small check group for your benefit/safety. Sometimes in between doses, sometimes at the smallest or the largest dose.
Irony - No. Fascinating - Yes.
There is nothing strange in what IPIX is doing. 505b2 does not explicitly require that active ingredient is the same in the application drug (prurisol) and the 505b2 bridged drug (Ziagen). As I understand it, in this case both drugs have the same final metabolite. FDA will want to know if that metabolite is safe. Instead of running specific safety tests IPIX can point to safety studies made with Ziagen. That is called 505b2 bridging. What IPIX can't do is to use Ziagen efficacy data. And, why would they?