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.
Ah, gotcha. I see the point you are trying to make.
Although, frankly I don't get legal TSM in the slightest. If it isn't statistically significant, it doesn't teach anything. That's why I find Du's point about 1 person having TG of 500 to be completely 100% non-sensical. You have to look at the average effect on a population. Real world data is stochastic. If one doesn't account for that, they are lost.
>Ok, according to Mori, EPA raised LDL.
That's not what Mori claims. In fact, Mori was unable to reject the null hypothesis that EPA treatment has the same average effect as placebo.
The main point, however, is that, upon reanalysis of Mori, the null hypothesis that DHA treatment has the same average effect as EPA also cannot be rejected.
Finally, Mori did show that the nulll hypothesis that DHA does raise LDL more than placebo can be rejected.
So Mori while Mori et al cannot claim a differential effect between DHA treatment and EPA treatment, nor can they claim that EPA raises LDL relative to placebo, they can and do claim DHA treatment raises LDL more than placebo.
Just to be clear, in the Mori data, the average LDL levels after EPA treatment ARE raised. However the size of the effect is not significantly different from placebo.
iwish -- great summary of the major issues with Mori (and at an early date!)
>Given the post trial finding reported by Megc, can the above be considered in the appeal.
I posted my opinion in the response to Megc. But really this is a question better answered by a lawyer, which I am not. My common sense tells me that since Mori is at the heart of Du's ruling, any argument about the validity of Mori should be entertained in a fair trial. But after reading a number of well-argued posts from lawyers on the board, I think the issue is more subtle than that.
Hi Megc,
The first issue is related, but not exactly the same point as raised by Bhatt and me. Related enough so that the Bhatt paper can be considered in the appeal? Well, that is being debated by lawyers who have different opinions. My opinion is that yes, it is related enough, but I am not a lawyer.
To be more clear:
Statement 1 from post trial finding: "the person of ordinary skill in the art would have attributed the absence of a statistically significant increase in the EPA arm to (1) the study’s small sample size (19 patients in the EPA group and 17 patients in the DHA group) ..."
How is this statement related to the result found by Bhatt and myself?
A POSA would have looked at the study and said to himself, "hmm, Mori did not directly compare EPA to DHA. If I want to claim DHA raises LDL levels more than EPA (a fact I would want to know if I am going to bother to purify EPA) then I better do the direct comparison". If the POSA did the direct comparison, and, hypothetically, found that DHA does raise LDL more than EPA then he would say, "OK, I can make the claim, sample size be damned, because the statistical test accounts for sample size". So in this instance sample size is irrelevant.
On the other hand, if he did the direct test and found there was not a statistically significant difference between the two group (which is what both Bhatt and I did and found this was indeed the case), he might still wonder why, if there is no statistical difference between the EPA and DHA groups, why did the DHA group show a statistically significant difference from placebo, whereas EPA group did not? Well to explain that fact you would invoke sample size -- either the direct comparison was not sufficiently powered or the EPA to placebo comparison was not sufficiently powered.
Now the second statement is a different issue entirely:
>(2) the difference in the baseline TG levels of the two groups, with the EPA group having a mean TG level 11% lower than the DHA group (2.01 mmol/L or 178 mg/dL for the EPA group versus 2.25 mmol/L or 199 mg/dL for the DHA group).
It is known that the expected change in LDL levels upon DHA treatment is correlated with the baseline TG levels in the patient (higher TGs = bigger increase in LDL upon DHA treatment). This has two implications, both of which were pointed out by Amarin's lawyers. First, the populations analyzed in Mori have much lower triglyceride levels on average than what is specified in the MARINE indication. So they argue Mori is irrelevant on those grounds. I tend to agree. The second implication is what is stated above -- that because the DHA and EPA groups have different mean TGs, any between group difference in change in LDL could be accounted for that. (one caveat is that the linear regression model that Mori used was not clearly specified, so perhaps this was accounted for in their model?).
I think Amarin's lawyers argued -- "We concede Mori shows something but it should be ignored based on small sample size and the TG level issue". I agree with this argument. But a stronger argument that can be made now is: "With respect to the question of whether EPA and DHA treatments have different effects on LDL levels, Mori teaches nothing". The TG argument is now not needed to disqualify what Mori teaches, because Mori teaches nothing w.r.t. that issue.
HDG --
I misspoke. Mori tested EPA and DHA groups each against the olive oil control group, not against baseline, and not against each other. (though baseline measurements were reported, the statistical test was (correctly) not done between treatment and baseline.
Zip, I think that what you've written is accurate.
I went back and looked at Mori et al. Given their introduction, one might have *expected* them to have done the direct comparison between DHA and EPA treatment (e.g. they write "it is important to determine the extent of any differential effects of EPA and DHA."). However, they did not in fact do a direct comparison. BUT, they also did not make a claim about a differential effect between DHA and EPA anywhere that I can find in the paper, so it's hard to lay the error at their feet. Then events proceeded as you have written.
Zip,
Thanks for the comment. My post was poorly worded.
I don't mean to suggest Du did anything maliciously. She did appear to essentially copy sections of the defense's brief for her opinion. This appears to be commonly done and was not meant to be evidence of any malintent on her part. Although I do not agree with her ruling, I do believe that she made an honest effort to do the right thing.
WRT the defense. Again -- worded poorly. I am not sure their cropping was intentional, but the error itself was egregious because what they claim is not supported by the data in the paper -- no further analysis needed. They either did not read the paper carefully, were very careless in interpreting it, or something more marginal. I tend to think it was the one of the first two thing, but it was an egregious, if unintentional, error.
The Kura issue is a whole nother can of worms (to my knowledge first uncovered by poster mc1988) and much more egregious than the Mori issue. That was a blatant misrepresentation of the paper by Generics lawyers that was then cut and pasted by Du (refuted by Covington? IDK?)
HK- great summary.
Kiwi,
Well it's actually all fairly subtle. I wouldn't say that Mori says nothing about the effect of EPA on LDL levels because the results were NS.
One way to look at it is as follows (this is an intuitive, non-rigorous argument)
One could argue that Mori puts a bound on the (average) effect size of EPA. So you can't really say that Mori "says nothing" because you've learned some real information. You've bounded the effect size. In other words Mori tells you that EPA probably doesn't have a HUGE positive or HUGE negative effect in the population tested. Because if it did, then we would see it, and the statistical test would have been significant.
Importantly, this estimate of the effect size of EPA in the population has some error associated with it, as does the estimate of the effect size of DHA. It happens to be that the error in the effect size of EPA treatment was large enough that you cannot confidently distinguish it from "no treatment" (i.e. no effect). Whereas for DHA the effect size could be distinguished from "no treatment". But when you compare the estimated effect size of EPA and DHA, because of the errors, you cannot say that the effect size of DHA on LDL is bigger or smaller than the effect size of EPA on LDL.
Now, all of this completely ignores the fact that Mori was performed in a population with characteristics that were essentially irrelevant to the target population for the MARINE indication, which was Covington's main argument.
Let's say you have two treatments A1 and A2, and you compare them to baseline.
Then you find that A1 has a statistically significant increase of some parameter (LDL) relative to baseline, but A2 does not.
You CANNOT say A1 treatment causes a bigger increase in LDL levels relative to A2 from this test alone. You HAVE to do a statistical test between A1 and A2 to make this claim.
Now look at figure 3 in Mori. The test was between DHA treatment and baseline and NOT between DHA and EPA. So you CANNOT conclude that DHA causes a bigger change in LDL levels than EPA.
In short, there is NO EVIDENCE in Mori that DHA causes a bigger change in LDL levels than EPA. It may be that when you do the appropriate statistical test, there is a difference, but it's hard to say without the raw data. By eye it doesn't look significant in Fig 3
Scientifically there is no way to justify her claim solely from the p-values in the paper. Therefore, the claim that DHA raises LDL more than EPA CANNOT be obvious to one skilled in the art (from Mori).
I think this is fundamental -- no need to rely on secondary considerations to rebut obviousness. The logic used to rebut the primary considerations was flawed.
Rafun,
I am not a lawyer, but your overall analysis resonates with what I have heard from other lawyers.
Regarding your comment:
>It it was: Amarin argued X, Generics argued Y, Du concluded Y and Amarin did not refute Y during the trial, that would be troubling.
I hate to say this, but I can certainly see how this viewpoint can be argued. If I recall correctly, much of Du's opinion was "cut and pasted" from the defense brief. That doesn't necessarily support the idea that Du concluded Z, but instead that she concluded Y.
>rmitra, I have to ask how long did it take you to do the math?
I dunno. I think I put together the document in like an hour, but I saw the mistake in Du's opinion as soon as I read Mori. That only happened because someone was kind enough to post Mori on the board. I probably wouldn't have bothered to dig up Mori had it not been posted. Perhaps that's what happened with whoever Covington hired (didn't dig up Mori).
Hi Rafa,
>Rather, our observation constitutes a re-interpretation of evidence that was already cited multiple times in the judge’s opinion.
Ah, I got you. I interpret that as referring to the fact that Du cited Mori multiple times in her opinion (which I did read!), rather than that Covington brought up this precise interpretation during the trial.
Since Mori was mentioned extensively in Du's opinion, they argue re-interpretation of Mori should be fair game for the appeal, which is a viewpoint that I tend to agree with.
alm2
>But Amarin had the right to expect a judge will apply her own sense of what is correct and incorrect in reaching her decision
I don't think Amarin could have reasonably expected Du to pick up the error in Mori. What I think they did expect was that she would be convinced by the argument that Mori was not performed in the correct patient population (i.e. very high triglycerides) and that this population has very unique characteristics w.r.t. LDL changes upon DHA treatment.
Amarin essentially bet the farm on that argument. I find that argument compelling with ample support from the literature, but ultimately it did not persuade Du.
>Are you certain that it was unrefuted by Covington? How do you know that they didn't bring it up during their testimony?
I am not. I have not read through the transcripts in detail.
I assume unrefuted because it is not a subjective issue. If it were brought up, then I cannot see how Du could rely so heavily upon that argument, because again it is not subjective.
So while you are 100% correct that I cannot make that claim with certainty, I would happily bet a beer on it :)
>wonder how they ( Covington ) explain that ? Weren't aware of it or were , but decided not to argue it ...simply expecting Judge Du to recognize the EPA/ LDL data was NS and what that meant .
My bet: unaware of it (but they, or their consultants, should have been). It's a clear statistical error, but one that I've seen postdocs make (though certainly not ones with BSs or PhDs in statistics). It would be crazy if Covington et al were aware of the error and expected a judge to deduce that the analysis was bogus.
It's a common error. And not one that is in any way subject to interpretation. But also way too subtle to expect a layman to recognize.
Hi zip,
>Mori conflated the two studies and erroneously concluded that a POSA would understand that EPA would significantly lower LDL
I think you mean Du. Mori did not do a direct comparison between the EPA and DHA groups, nor would Mori care one way or the other about a POSA.
>This study was published in a non peer reviewed journal...If it had been published in a peer reviewed journal, the error would been picked up.
Mori et al was published in a peer reviewed journal. From Am J Clinical Nutrition website "The most highly-rated peer reviewed, primary research journal in nutrition..."
> the error would been picked up.
Mori did not make an error. Mori did not claim a significant difference between EPA and DHA groups w.r.t. LDL. Mori only tested each versus baseline. Hikma and Reddy's lawyers made the error in interpreting Mori, which was unrefuted by Covington and then accepted by Du.
Interesting article re: coronavirus with implications for Vascepa
https://elemental.medium.com/coronavirus-may-be-a-blood-vessel-disease-which-explains-everything-2c4032481ab2
There has already been some discussion of the potential of Vascepa as a treatment for COVID due to the Das paper (https://pubmed.ncbi.nlm.nih.gov/32229155/), due to the clear effect of Vascepa on inflammation and the role of this same process in COVID , and due to the initiated clinical trial by Bhatt and colleagues.
This article (and the underlying paper in Lancet) add another rationale as to why Vascepa *may* be effective for the treatment of COVID. Vascepa is known to enhance endothelial cell function, and the Lancet article argues that the lethality of COVID is largely due to SARS-CoV-2's ability to infect endothelial cells and impair their function. As mentioned in the article, statins have been shown to have a protective effect against COVID, so that is also suggestive that Vascepa may be efficacious.
anfla, exactly.
Dukesking, TKE,
Thanks for the kind words. Having some trouble with the board -- LOL. Now back as rmitra.
mc -- excellent. Thanks, that is very helpful
alm2 -- ok, gotcha, thanks. Sent.
R
>Rmitra, do you know whether anything akin to your rationale is reflected in the record before the District Court. Review by the CAFC is on the record in that lower court.
I don't believe the exact calculation of the probabilities for EPA versus placebo or EPA versus DHA were reflected. That said, obviously there was a great deal of discussion about very related statistical issues.
>I suggest you email or text your analysis to Covington with a CC to Stephanie. Covington will consider whether a remand to the district court is appropriate or whether it would further settlement discussions.
I've forward a good deal of my earlier analyses to IR. I haven't forwarded any of the stuff that I've posted today. (if I forward every time I post, stephanie will think I'm a first rate narcissist hahaha).
gg-
>1)~12.6% chance that EPA does NOT raise LDL levels given the data in Mori
>2) <5% chance that DHA does NOT raise LDL levels given the data in Mori?
Yes I think that is a reasonable "normal speak" summary.
>So a POSA should not expect EPA to be LDL neutral or lowering by just looking at Mori UNLESS he was only allowed to pick either DHA or EPA, and not any other potential molecules, to treat TG>500.
I'm not sure I understand the claim. What I would argue is that if I were a POSA who had just read Mori, it would not be obvious to me that it would be worth it to go through all of the trouble to purify EPA, when the evidence in Mori that EPA raises LDL levels LESS than DHA is not statistically significant, and it's quite possible, given the data in Mori that EPA raises LDL levels relative to placebo (12.6% chance). Then, on top of all that, I would worry about the relevance of Mori due to TG levels.
Basically I would think that EPA probably raises LDL levels just like DHA does, but that Mori, being a small study, was just underpowered to detect it at an alpha of 0.05. And then, I would say to myself, "and it's probably even worse that what I think, because that population didn't even really have people with HIGH TGs, which makes the problem worse"
Currently, yes, that is my opinion. But I am only passingly familiar with the literature outside of Kura and Mori. Furthermore, it is tricky to say what studies that are performed in populations with the "wrong" distributions of TF levels actually teach. That is somewhat subjective issue, and as a result, I find it harder to separate my intrinsic biases from what a POSA might think. Also see my most recent post about secondary considerations.
>So is it obvious? Well trying to be unbiased, w.r.t. LDL, I can maaaybee see an argument, though I tend not to agree. But with regards to ApoB and Kura, I do not see it. Combined, I do not see it.
The one caveat to this statement is this -- as someone in the biomedical field, I tend to heavily weight what the courts are calling "secondary considerations". It's hard to overstate what a major paradigm shift the REDUCE-IT trial was. Very few in the field believe it would effect outcomes, and the results were spectacular. As a result, my opinions are certainly colored by the fact that I am acutely aware of what the scientific community believed before and after REDUCE-IT. But in this trial, REDUCE-IT is only tangentially relevant.
Kura and Mori, Part II.
1. Mori does not teach that DHA raises LDL levels MORE than EPA.
2. Mori does teach that DHA raises LDL level relative to a placebo group with greater than 95% probability. (or more precisely, you can reject the null with >95% percent probability, which is a slightly different, but more accurate statement).
Does Mori teach that EPA DOES NOT raise LDL levels relative to a placebo group? No. It teaches there is not 95% confidence that EPA raises LDL levels relative to placebo.
What is the probability, given the DATA in Mori that EPA raises LDL levels? One can easily compute this since they report the mean and SEM.
The answer is 87.38%
Or conversely there is a ~12.6% chance that EPA does NOT raise LDL levels given the data in Mori.
These probabilities can be changed by looking at data outside of Mori, of which there is some. But not in the right Tg populations. Nor is Mori the right triglyceride population. So is it obvious? Well trying to be unbiased, w.r.t. LDL, I can maaaybee see an argument, though I tend not to agree. But with regards to ApoB and Kura, I do not see it. Combined, I do not see it.
Kura and Mori --
I do think the problem with Du's logic w.r.t Kura is significantly stronger than the problem with Du's logic w.r.t. Mori.
Mori found
1. EPA treatment does not statistically significantly increase LDL (compared to placebo, olive oil). BUT there was an increase observed.
2. DHA does statistically significantly increase LDL compared to placebo.
Now if you want to claim DHA raises LDL more than EPA, you have to do a direct comparison. When this is done, there is no statistically significant increase.
So IMO, Mori ALONE does not teach, in an obvious way that EPA raises LDL levels less than DHA. But why not? Well the direct comparison fails the significance test, so you might assume that perhaps EPA treatment actually does increase LDL levels, but because of the small sample size, it failed to show up as significant. It is also POSSIBLE from Mori that it was the EPA versus DHA comparison that failed due to a lack of power, which would imply the EPA treatment raises LDL levels less than DHA treatment. Either option is possible, neither is "taught".
How could one decide between these options? Well, you could look at other studies, and if you found additional evidence that EPA does not raise LDL levels in a *similar* population, (this is another thorny issue -- good arguments can be made that the Mori population is NOT relevant to the MARINE indication) then you might reasonably conclude that DHA raises LDL while LDL does not. Is that obvious? No, I don't think so, certainly not from Mori alone. However, you can make the case that Mori DOES teach that DHA raises LDL levels relative to placebo, just not relative to EPA. There is some hint of the "right" path, that could then be gleaned by referring to additional literature.
Kura teaches there is no stat. significant difference between EPA and placebo. There is no other way to interpret those particular data that is not erroneous.
anfla,
Well said. This is exactly right.
Lem --
>but can it be a point on appeal.
I am not a lawyer. However, I suggest that it should be a point on appeal, because the prosecution's argument, the defense's argument, and the Judge's decision all cite Kura. Secondly there is some discussion in the decision specifically with regards to Kura about interaction between E3 and EPA that suggests that the Judge WAS aware of the non-significant p-value between groups, so this particular issue should also be fair game.
The Kura issue
I've seen a few posts talk about Dr. Heinecke and his claim that the ApoB result for Kura is significant, and some discussion about how this might then become a situation where two experts are pitted against one another in the appeal.
I don't believe this is the case. This particular issue, statistical testing between baseline and a randomized group as opposed to between two randomized groups, is very very common, and the statistical community has litigated this thoroughly. The answer is, it's a big no-no.
Board member Piyao posted a couple of links that discuss this (Hat tip to Piyao). I am reposting one of them here, and pasting a few quotes from this article, quoting past articles from the statistical literature. The article is not behind a paywall and highly readable, so I encourage people to take a look:
https://www.nature.com/articles/s41393-018-0203-y
Quoted from the article above:
-----------------------------------
Nonetheless, the practice persists. Here are some comments that should dampen enthusiasm for using p values in this way:
“….performing a significance test to compare baseline variables is to assess the probability of something having occurred by chance when we know that it did occur by chance. Such a procedure is clearly absurd.” p. 126 [2].
“P-values for baseline differences do not serve a useful purpose, since they are not testing a useful scientific hypothesis.” p. 2928 [3].
“With few exceptions, the statistical literature is uniform in its agreement on the inappropriateness of using hypothesis testing to compare the distribution of baseline covariates between treated and untreated subjects in RCTs.” p. 142 [4].
“Indeed the practice can accord neither with the logic of significance tests nor with that of hypothesis tests….I suspect that the practice has originated through confused and false analogies with significance and hypothesis tests in general.” p. 1716 [5].
Even if we ignore the criticisms of statistical testing for baseline differences, there is the added problem that an insignificant p value may merely reflect a small sample size. That is, there may be large differences that statistical testing fails to detect. And what if there is a significant difference on one baseline variable? It would be rather surprising if there was not given the high number of variables that are typically measured at baseline. A single p value less than 0.05 among many baseline statistical tests may just reflect a spurious finding.
------------------------------------
In summary, the error with Kura is not a topic that will be resolved by a debate between experts. It is an issue that has already been thoroughly examined and resolved by the community.
It may be that removing Kura as evidence for obviousness is not enough to tip the appeal Amarin's way. However, I find it very difficult to believe that Kura will not be removed as evidence for obviousness, as long as the legal team is aware of the mistake and the literature.
From a scientific point of view I think there is not much you can say about a small number of individuals with tg>500 in that population. Legally I think the defense was trying to make the point that people with Tg>500 had been given pure EPA in the past (which is the Marine indication).
From a scientific point of view I think
Hi Mc1988,
This is interesting. I haven't had a chance to read the paper yet, but here are some initial thoughts.
Assuming a normal distribution with a mean of 300 and stdev of 233, the Judge is correct, it is almost certain that at least one individual has triglycerides > 500. I could calculate this formally, but it is essentially a certainty.
Looking at the graphs below, it's curious that none of the 25 individuals have tg > 500 and they seem to be much more tightly distributed than the reported StDv, and the mean seems to be slightly less than 300 in this population. Of course, as you note, there were 28 individuals in the population. My interpretation would be that there were three outliers in the population that had VERY high triglycerides, which shifted the mean to the right and led to the higher st. dev. I would say that is likely the explanation, and supports the conclusion that there is at least one individual with >500 tgs.
mc1988-- NP. You made a great find. and thx for posting the full table.
Lem --> It was significant compared to "Baseline" which is not the appropriate comparison. (that was E3 + EPA)
It was not significant compared to the appropriate control (E3 only)
See for example:
https://www.nature.com/articles/s41393-018-0203-y
(this link was posted earlier by I belive Piyao, and I bookmarked it, but as the link states, this is a well known issue)
Hey man, I get it. It's a message board. People are going to do what they are going to do. My disclaimer was included so they don't use my words to do it :)
crap- misfire again.
I understand your point of view. Here is where I am coming from -- many people on the board are emotionally raw due to financial loss as a result of Judge Du's decision. Completely understandable. But, I don't think it is useful to vilify her, and I can easily see scenarios where it would be harmful. Better to rebut the facts that she used and the logic she applied to those facts, since that is the only way to productively move forward in any case.
edit: misfire