Quantification of obviousness
What is clear (to say the least) from this message board is that everybody was taken aback by the differences in the standard of obviousness required by the judge, the Patent office or the FDA.
Judge acted as if EPA was having mechanistic effects on two other parameters lipids and TG. She missed the fact that there are numerous unknown ones (genetics nutrition randomness etc.…) which could have caused the results and not EPA. The underlying hypothesis of Judge DU ruling is that the Mori and Kuribayashi would be systematically reproduced. And that the opinion from Amarin’s lawyers amounted to no more than hairsplitting.
Clearly, she took a mechanistic approach, which typically consider that components of any results are real, solid and visible. If it works once it’ll work over and over again.
The FDA has devised requirements to make sure trial results have a high likelihood of being reproduced. The task of Judge DU was not to be retrospectively more knowledgeable than biologist or physicians at that time. Her task should have been to assess if the necessary prerequisites for the mandatory standard of proof were met. This is the reason IMO her obviousness considerations should be appealed.
There is a benchmark for obviousness: the necessary but not even sufficient requirements for approvability set for by the Center for Drug Evaluation and Research (CDER) which ensures that safe and effective drugs are available for the people of the United States.
Hence the standard of proof for a drug should take into account the particular nature of biology and the lack of apparent logic because not all information is known.
I propose counting the unmet conditions for approvability as an approximate for lack of obviousness. It is a kind of quantification which can be applied to Judge Du ruling. The goal of the demonstration is to make judges or anyone not aware of the difficulty of proving obviousness understand there are a lot of necessary conditions to ensure a result is repeatable and not a chance finding.
Lawyers should have put their emphasis on the prerequisites more than on the details of the results. They should have been very pedagogical, used a lot of examples as to why these requirements are necessary. There is an incredible number of failed trials to illustrate the demonstration.
Required parameters not met:
Randomization
Not performed according to possible risk factors such as gender age smoking etc. How EPA is processed in the body could likewise depend on age, smoking. Nobody knows in advance. There are discrepancies in the percentage of risk factors in the two arms of the trial. For instance: 5 times as many smokers in the EPA group. Etc…
Why is it so important?
Interesting to illustrate with the debate on chloroquine. At day 5, no treated patient is positive as compared to 90% for the control group! p value at day 5 below 0,0001. The efficacy is obvious, isn’t it?
In the treated group there were mostly women and younger patients as compared to the control group. They are noticeably less prone to the acute phase of the disease …
Furthermore Covid 19 has a lot of unknown parameters. Why some old people are not affected much and why some young patients die from it without any known risk factor? Even if it had been randomized according to age and gender it could be for instance possible to have imbalances with more old patients less prone to the disease in the treatment arm just because of the small size of the group.
These kinds of Imbalances are much more likely when using a small population. P value only works if the two populations are totally similar, the only difference being the treatment effect. The only way to approximate similarity for unknown factors is to increase as much as possible the size of the studied population.
P value cannot be trusted if there are known or unknown imbalances
Conclusion: size of group matters and randomization according to known risk factors matters. Reviewing the necessary preconditions is not splitting hair because ex-ante, nothing is obvious as demonstrated by this debate on chloroquine.
Similarity of studied population
The studied population was only Japanese:
As for alcohol, there could be genetic differences in the way EPA is processed in an Asiatic population as compared to a US population.
They were women only.
It is not proven EPA is processed the same way regardless of the gender or the ethnicity of the studied population. (for instance, women are less prone to severe forms of Covid 19).
Triglycerides at baseline
It is impossible to know in advance if EPA has the same effect on TG whatever the level. EPA could be “overwhelmed” if the level are higher. Nobody knows in advance.
Population size
Not enough to take into account possible unknown differences (same way people of same age and gender react differently to coronavirus) making patients with different metabolism react differently for unknown reasons to EPA ingestion.
Not enough to take into account possible change of diet among the clinical trial participants with obvious impact on LDL and TG.
EPA dosing
Not the same:1.8 g EPA as compared to 4g in the Marine trial.
The bigger the dosing the better?
You would expect as obvious the success of a trial with an even greater amount of EPA. Splitting hair? Well the strength trial testing Epanova with an even greater amount of EPA than Jelis (2.4g as compared to 1.8g) failed miserably while Jelis did not. Nothing is guaranteed in biology. The more the best? It depends again on many unknown biological pathways.
Therefore, you cannot compare the effect of 1.8g of EPA in japan with 4g in the USA.
Molecule tested
Not exactly the same as Vascepa: indeed, EPA was associated with estradiol.
Possibility of synergetic or antagonistic effects which again can blur results. Drugs are notorious for interacting one with each other. For instance, possibility of estradiol effect on cytochromes or other unknown effects.
Independent review
The safety and data integrity have not been reviewed by an independent committee. EPA was not considered a drug but a supplement. Possibility of bias to enhance the results.
P value of the Kurabayashi trial:
« In light of the statistically-significant differential effects reported between the EPA and control groups, a person of ordinary skill in the art would have attributed the reduction in Apo B to EPA.”
As stated, p value is meaningful only if the two arms of the trial have no imbalance. Furthermore, you cannot draw any conclusion on EPA if you test a combination of drugs using a different dose!
The trial was too small et not randomized according to risk factors.
If the Kurabayashi trial p value amounted to enough evidence for not increasing LDL then why did the Strength trial failed to confirm the Jelis trial for cardiovascular events?
Hadn’t both Jelis and Kurabayashi a p value looking like a definite proof of efficacy?
Oddly the Jelis trial findings were not reproduced by the Strength trial because may be:
not the same dosing,
not the same association of drugs
not the same population
not the same diet
Or a combination of the above causes which exactly mirrors the reasons why Kurabayashi could not have been trusted. And Jells had 16000 patients enrolled as compared to 69 in the Kurabayashi experiment! It is an illustration there is no possible conclusion to draw from a limited trial when necessary requirements are not met since even huge trials cannot not be reproduced when only missing a few prerequisites. It demonstrates that “Common Sense” is no substitute for science.
Publication bias and pseudo "meta-analysis" by judge DU.
“Publication bias is a type of bias that occurs in published academic research. It occurs when the outcome of an experiment or research study influences the decision whether to publish or otherwise distribute it. Publishing only results that show a significant finding disturbs the balance of findings, and inserts bias in favor of positive results”
“Other proposed strategies to detect and control for publication bias[4] include p-curve analysis [7] disfavoring small and non-randomised studies because of their demonstrated high susceptibility to error and bias.[2]”
“We have confirmed the presence of publication bias in a cohort of clinical research studies. These findings suggest that conclusions based only on a review of published data should be interpreted cautiously,”
At the time of the Mori and Kurabayashi study there could have been unpublished studies showing no good results of EPA on LDL casting doubt on EPA not increasing cholesterol.
It was for a person of ordinary skill impossible at the time of the Kurabayashi experiment to conclude there was no unpublished study contradicting Kurabayashi and Mori findings. It was a realistic possibility since they were small sized trials without proper randomization.
This non-exhaustive list gives a count of no less than 9 missing requirements for reproducibility and with a more thorough investigation the number could be much higher.
A judge might find obviousness debatable when there are only a few minor errors but should be made conscious that the higher the number the lower the bar for obviousness.
Then a “person of ordinary skill” should have been aware of the great limitations of the Kurabayashi experiment and could not in any way find as obvious the repeatability of the findings as proven by the failure of the Strength trial.
Conclusions from Mori and Kuribayashi trials should not have been discussed: they were not valid because they failed to meet the necessary requirements for reproducibility. Judge DU got lost into the consideration of the conclusions but was not made really aware of the necessary prerequisites for obviousness and what it really means to a be a person of ordinary skill.
Even if a court wants anyway to discuss the studies’ conclusions, I should be shown that on top of not meeting the requirements for reproducibility they were not even interpreted correctly.
It should be underlined to the judges of the appellate court the dire consequences of having a double standard for obviousness: that two non-matching burdens of proof creates an “arbitrary zone” where a drug can be considered obvious by a judge but cannot be approved by the FDA with 2 major unpleasant consequences:
It alters the whole process of developing drugs from natural extracts (NE) if the sheer existence of one tiny trial prevents companies from patenting any NE. It goes contrary to the intent of the patent process which is to promote innovation. That should not be the purpose of the courts nor of the law. The goal of the law and justice system should be, whenever possible, to minimize as much as possible the extent of this “arbitrary zone”.
It removes any incentive to promote a drug as a standard of care at the start of its lifecycle. It takes a lot of time and efforts and money to get physicians aware of the existence of a new indication. No company would be enticed to promote the drug as it would benefit all the other players. The ensuing lack of marketing would prevent the expected decrease in the number of deaths and vascular events from happening. More death and misery than necessary. The discounted price of the generics wouldn’t even be a bonus for the insurers, since it would require more spending to cure patients, which might run contrary to the Judge Du’s intent.
Therefore, judges should be made aware that biological systems including humans are dependent upon a complex set of parameters where apparent causality is generally misleading without necessary requirements. Let’s hope the actual debates about Covid 19 potential treatments make judges conscious of the precarity of preliminary findings.
Judge Du ruling implicitly suggested that FDA requirements are non-necessary and frivolous but failed to demonstrate why (the burden of providing evidence they are unnecessary should rely on the defendant). She didn’t realize she was, without any theoretical or empirical justifications, upending all the rules set for demonstrating reproducibility and lack of chance finding. It makes this ruling looks capricious, grandstanding and obvious only in hindsight. A person of “ordinary skil” would have been wary of fast conclusions!
