The Threshold for Demonstrated Utility

Apotex Inc v Janssen Inc 2021 FCA 45 Locke JA: Webb, Boivin JJA affg Janssen Inc v Apotex Inc 2019 FC 1355 Phelan J

            2,661,422 / abiraterone acetate & prednisone / ZYTIGA / old NOC

The invention at issue in this case relates to the use of the combination of abiraterone acetate (AA) & prednisone for the treatment of prostate cancer. Today’s post will discuss the utility aspects of this decision, which touches on three different issues: (1) the nature of the utility requirement for a combination, (2) the requirement to disclose the factual basis for a sound prediction; and, (3) most importantly, the threshold for demonstrated utility namely that it is sufficient if the results are “strongly suggestive” of utility, and no other explanation is likely. (See Monday’s post on obviousness.)

Yesterday’s post sets out the facts in more detail; for this post it is enough to say that androgen produced in the adrenal gland promotes the growth of certain types of prostate cancer and AA was a known adrenal androgen inhibitor, albeit a relatively new one. Most adrenal androgen inhibitors also suppressed other adrenal hormones, such as glucocorticoids; this caused side effects which were treated by co-administration of a glucocorticoid, such as prednisone. AA had an unusual mechanism of action, and at the filing date it was thought that AA would not require co-administration of a glucocorticoid to reduce side-effects; instead, the inventors though that prednisone would have a synergistic effect with AA in treating the cancer itself. As it turns out, that theory was mostly wrong, but not entirely wrong: AA does cause side effects, and prednisone is co-administered for that reason, not for its synergistic effects in cancer treatment, but Phelan J did find that there is also a small synergistic effect.

Utility requirement for a combination

As discussed here, for a combination invention there is a question as to whether the utility requirement is satisfied if the combination has a scintilla of usefulness in treating the disease, or if there has to be a scintilla of a synergistic effect, such that the combination is at least marginally better than the sum of the individual effects. The combination at issue clearly had utility on the simple usefulness test ([218]–[219], but Phelan J proceeded on the view that a scintilla of synergy was required: see eg [219]-[221]. Phelan J found that there was a scintilla of utility even with a synergy requirement, though it was a much closer call.

Locke JA’s decision on appeal proceeded on the same assumption as Phelan J, ie that synergy was required, but without specifically affirming that this was the correct approach as a matter of law: see [43], [45]. My impression is that the point wasn’t argued at either the FC or FCA level; for example, the reference to the synergy requirement at [43] was literally parenthetical. I hope this means that the point shouldn’t be considered settled; but I recall how the promise doctrine started with a few statements that weren’t really challenged in complex litigation where there was a lot else going on, until by the time it was directly addressed, it was considered settled.

Enhanced Disclosure Requirement for Sound Prediction

The evidence relied on to establish utility was two trials, referred to as “Study 001" (AA and a different glucocosticoid, namely dexamethasone, in human subjects) and “Study 004” (AA and prednisone, also in human subjects), collectively the “Cougar trials.” The hypothesis that the combination of AA and a glucocosticoid such as prednisone would have a synergistic anti-cancer effect was referred to as “de Bono’s hypothesis” after one of the inventors [40]; this was the line of reasoning that would connect the trial data to the utility.

At first instance Phelan J held that Janssen could not rely on these trials or the de Bono hypothesis to establish a sound prediction of utility because “no part of de Bono’s hypothesis or the Cougar studies were disclosed in the 422 Patent and therefore Janssen cannot rely on sound prediction” [223]; and see [212]. Without the enhanced disclosure requirement, a sound prediction of utility would “likely” have been made out [223]. Thus this was a case in which the enhanced disclosure requirement played a real role. I had somehow missed this when blogging on Phelan J’s decision, so I’ll make a few comments on Phelan J’s decision at the end of this post. For now I’ll say that while there is certainly FCA caselaw directly endorsing the enhanced disclosure requirement, as Phelan J noted [223], in Sanofi-Aventis v Apotex / clopidogrel 2013 FCA 186 [132] Gauthier JA raised serious concerns about the doctrine. In this case, there was no challenge on appeal to Phelan J’s holding on sound prediction [38], but towards the end of the decision Locke JA quoted from my article “Must the Factual Basis for Sound Prediction be Disclosed in the Patent?” (2012) 28 CIPR 38, and remarked “For the purposes of the present appeal, it is not necessary to address the question posed in the title of Prof. Siebrasse’s article” [50]. Maybe it’s wishful thinking, but I’ll take this as a hint that there may be room to revisit this issue at the FCA level; the only other FCA decision to touch on the point after Gauthier JA’s comments in Clopidogrel, namely Eurocopter 2013 FCA 219 [153], waffled a bit by saying that disclosure in the specification “may” be required to support a sound prediction of utility.

Demonstrated Utility

While Janssen could not rely on the Cougar trials and the de Bono hypothesis to establish sound prediction, Phelan J held that the same data, in conjunction with previous studies on the anti-cancer effects of prednisone, was sufficient to demonstrate utility [FC 221].

The FCA affirmed. Locke JA’s reasons rested on three points of law or principle, combined with a deferential standard of review.

An initial point of law is that neither study was sufficient on its own to demonstrate utility, but Phelan J held that the combination did. Locke JA affirmed that “testing should be considered cumulatively when assessing demonstrated utility” [41], citing with approval Teva v Novartis / imatinib 2013 FC 141 [215]–[216].

Secondly, while the 004 Study tested AA in combination with prednisone [FC 43], the endpoint was not survival rate, but rather prostate specific antigen (PSA) response, which is “a surrogate measurement for the effectiveness of prostate cancer treatments” [FC 19]. The claims were to the combination “for the treatment of a prostate cancer in a human” [FC 38], and Apotex argued that a study using a surrogate could not demonstrate the claimed utility, but would merely provide the factual basis for a prediction of utility [47].

In response, Locke JA first noted that “the Federal Court found, as a fact, that PSA indicated the response to prostate cancer in 2007: Reasons, para. 19. This suggests that a reduced PSA level can be a demonstration of utility in fighting cancer” [48, original emphasis]. The general point, as I understand it, is that even when we consider a fact to be demonstrated, we don’t necessarily directly observe it directly. The Higgs boson is said to have been observed, but what was directly perceived is a peak in the output of a particular detector, which, combined with a suitable statistical analysis, allowed the CERN scientists to say, with a very high degree of confidence, that the existence of the Higgs boson has been established. If we say it has been shown that an antibody binds to an antigen, what is directly observed will likely be a colour change, which, combined with a theory about the conditions under which the colour will be observed, allows us to infer the binding. Even in an RCT in humans, the strongest way of demonstrating utility of a drug, we don’t directly observe the drug curing the disease; what is observed is the disease symptoms in the control and treatment groups, and this, combined with suitable statistical analysis, may allow us to infer that the drug treats the disease. There is no magic in “statistical significance” with a p-value of 0.05, as the replication crisis in social psychology shows; physicists require 5-sigma significance for important discoveries.

Thus Apotex’s objection is correct in a way; the use of the PSA really only allows us to predict—or more accurately, infer—that the combination is effective in treating prostate cancer. But the same would be true even if survival had been used as an endpoint; any conclusion that the combination worked to treat prostate cancer would be an inference based on a statistical analysis of the observed outcomes, which would allow for the role of chance, placebo effect and so on. Depending on the strength of the correlation between PSA levels and prostate cancer, it might well be that a high-powered study using PSA levels would give a greater degree of certainty than a low-powered study measuring survival rate. While the established doctrinal test for sound prediction requires a factual basis and a sound line of reasoning, the same is true for demonstrated utility, even though that point is not expressed as a formal test. There is no sharp difference in kind between a sound prediction of utility and a demonstration thereof; the difference is only the confidence with which we are prepared to draw the inference of causation. So, the fact that the 004 Study used PSA as a criterion rather than studying the survival rate itself does not mean that it cannot demonstrate utility, in the same way that the fact that CERN used statistical analysis of detector energy spectrum rather than direct visual perception does not mean that the Higgs boson has not been observed. The question as to whether utility has been demonstrated rather than soundly predicted does not turn on the nature of the tests that were carried out.

So, it makes sense in principle that the claimed utility need not be directly measured in order to be demonstrated. The question then is to decide how strong the inference must be in order to demonstrate utility, as opposed to merely predicting it. Locke JA held that utility need not be established with certainty in order to be demonstrated [49]:

It is sufficient that the test results are strongly suggestive of utility, and that there is no other logical explanation for the test results is likely.

Finally, Locke JA stated that “More broadly. . . it must be acknowledged that it is difficult to draw a clear line between a demonstration of utility and a prediction of utility” [50]. He went on to quote a passage from my article “Must the Factual Basis for Sound Prediction be Disclosed in the Patent?” (2012) 28:1 CIPR 38, in which I argued that there is no basis for distinguishing between the two. He then concluded that he would not intervene with Phelan J’s decision, “[b]earing in mind the vagueness of the line between prediction and demonstration,” combined with the evidence relied on by Phelan J, and in light of the applicable (deferential) standard of review [51].

Locke JA did not explain exactly why the vagueness of the distinction between demonstration and sound prediction was relevant in upholding Phelan J’s finding that utility had been demonstrated. I made the point in my article to argue that because there is no sharp distinction in reality, there should be no sharp distinction in doctrine; either disclosure of the factual basis should be required for both demonstrated utility and sound prediction, or for neither. That is evidently not how Locke JA is invoking the argument, given that his decision proceeded on the assumption that there is an enhanced disclosure requirement for sound prediction and not for demonstrated utility. I think that what Locke JA was saying is that given that current doctrine establishes a line that is vague, as Locke JA put it—or not grounded in reality, as I might prefer to say—the law should not be too alert to enforce that boundary strictly, so as to avoid arbitrary results. Put another way, the threshold for demonstrated utility should be applied flexibly, given the drastic legal consequences of a vague and arguably arbitrary distinction. This is consistent with GSK/rosiglitazone 2011 FC 239 [98], for example, in which the utility of a claim to the compound rosiglitazone was held to be demonstrated on the basis of potential use for treatment of hypoglycaemia.

While this all makes sense, it does illustrate that the law faces a struggle in adapting to the enhanced disclosure requirement. Ultimately, the FCA will have to deal with the issue head-on. When that time comes, this decision will be an important part of the context.

Phelan J’s discussion of the enhanced disclosure requirement

As noted above, Phelan J proceeded on the view that there is an enhanced disclosure requirement, applicable to sound prediction but not demonstrated utility, such that the factual basis and line of reasoning to support a sound prediction must be disclosed, or at least referenced, in the patent itself. There is certainly FCA caselaw expressly so holding: see Lilly v Apotex / raloxifene 2009 FCA 97 [15]; Lilly v Teva / atomoxetine 2011 FCA 220 [46]–[47]. Subsequently, however, in Sanofi-Aventis v Apotex / clopidogrel 2013 FCA 186 [132] Gauthier JA raised serious concerns on this point in her concurring reasons, though that of course does not in itself overrule the prior FCA holdings; at most, it suggests that there might be some appetite to reconsider the issue in the future. So Phelan J was on firm ground in taking this position. With that said, I have a couple of concerns about his analysis. He didn’t refer specifically to either the Raloxifene or Atomoxetine decisions, but they were evidently the premise of his analysis. His discussion of the relevant case law was as follows:

[222] The Supreme Court of Canada stated in Apotex Inc v Wellcome Foundation Ltd, 2002 SCC 77 at para 70, [2002] 4 SCR 153, that sound prediction requires a factual basis, line of reasoning, and disclosure of the prediction in the patent.

In my view, it is not correct to say that the SCC in Wellcome / AZT held that disclosure of the factual basis is required in the patent. The passage was cryptic and was expressly obiter. That doctrine is really the product of the FC / FCA interpretation of that passage, particularly in Raloxifene. I also have to point out that while the SCC said in Wellcome / AZT that “both the underlying facts (the test data) and the line of reasoning (the chain terminator effect) were in fact disclosed,” that statement is incorrect.

Phelan J went on to say:

Although some commentary has suggested the Supreme Court of Canada may have overturned this statement, the Federal Court of Appeal in Bell Helicopter Textron Canada Limitée v Eurocopter, société par actions simplifiée, 2013 FCA 219 at paras 152-155, 120 CPR (4th) 394, confirmed that where the factual basis and line of reasoning are based on data that is not part of the common general knowledge, then disclosure is likely required to support sound prediction.

I believe the first sentence refers to some commentary suggesting the SCC in AstraZeneca 2017 SCC 36 overturned the enhanced disclosure requirement. I recall seeing that commentary, but I can’t find it or a case referring to it (if any reader can point me to it in a comment, I’d appreciate it.) In any event, I agree with Phelan J that the AstraZeneca decision does not affect the issue one way or the other.

My own view isn’t that the SCC overruled itself in AstraZeneca, it’s that Wellcome / AZT never established an enhanced disclosure requirement in the first place. The SCC’s statements in Teva 2012 SCC 60 are more to the point. The SCC stated at [37] that

The lack of certainty that comes from predicting rather than demonstrating an invention’s utility has led some courts to conclude that there is a “heightened” or “enhanced” disclosure requirement in cases in which a claim of utility is based on sound prediction: see, e.g., Eli Lilly Canada Inc. v. Apotex Inc., 2009 FCA 97, 78 C.P.R. (4th) 388, at paras. 14-15.

Notably, the SCC did not cite its own decision in Wellcome / AZT as suggesting an enhanced disclosure requirement. The SCC in Teva concluded its discussion of this point by saying “Since sound prediction is not an issue, the question whether there is an ‘enhanced’ or ‘heightened’ disclosure requirement with respect to sound predictions does not arise in this case and need not be addressed” [43]. The SCC did not say “We need not consider whether we should overrule our holding in Wellcome / AZT.” In my view, it is clear that the SCC does not consider that it established an enhanced disclosure requirement in Wellcome / AZT; I am quite certain that if the matter ever did go to the SCC, the Court would not consider that the question was whether it should overrule its own precedent, but rather whether it should overturn a rule established by the Federal Courts.