Plavix Patent Invalidated by the Distinction Between Demonstrated Utility and Sound Prediction

Apotex Inc v Sanofi-Aventis / clopidogrel 2011 FC 1486 Boivin J
            1,336,777 PLAVIX

Two unique features of Canadian law, the false promise doctrine, and the more recent rule that the basis for a sound prediction must be disclosed in the patent itself, mean that lack of utility is one of the most common reasons for invalidating pharmaceutical patents. This was the case in this clopidogrel decision, in which both these doctrines came together to result in a holding of invalidity based on lack of utility.

By way of context, clopidogrel is the enantiomer of the racemic compound referred to as PCR 4099, which is a member of the class of compounds called thienopyridines. An early member of this class, ticlopidine, was synthesized by Sanofi in the early 1970s [411]. Ticlopidine has been approved for use in humans (in 1978 in France [442]); Apotex, for example, holds an NOC for the sale of ticlopidine for humans. Unfortunately, ticlopidine has significant side-effects, so there was a motivation to find a better tolerated but equally or more active member of the class. Sanofi identified PCR 4099 as a lead compound, and carried out extensive studies on it in the 1980s, including roughly a dozen different studies on humans [444]. Based on these studies, “it was found that (i) PCR 4099 was the most potent thienopyridine compound synthesized to that point of time, and, (ii) it was significantly more effective and better tolerated than ticlopidine [446]. However, a long term animal study suggested that PCR 4066 could lead to convulsions. (It is not entirely clear from the discussion of the facts whether this was a side effect of ticlopidine, but it appears not.) A wide range of animal studies demonstrated that clopidogrel was at least as active as PCR 4099, but not toxic [467].

In summary, PCR 4099 was more efficacious than ticlopidine and clopidogrel was at least equally active. PCR 4099 was better tolerated than ticlopidine, except for convulsions, while clopidogrel was less toxic that PCR 4099. This means that at the relevant date it seemed that clopidogrel was clearly better than ticlopidine, and ticlopidine was actually approved for human use. Nonetheless, Boivin J held that Sanofi had neither established demonstrated utility or a sound prediction of utility. What explains this?

The first crucial doctrine is that in Canadian law utility is measured against the promise of the patent. This means that the invention may have the “scintilla” of utility adequate to support a patent, but if it promises a higher degree of utility that is not delivered, the patent will nonetheless be invalid. In this case Boivin J interpreted the promise of the patent to be “use in humans” [123], [174]. This was more than the “potential use in humans” argued by Sanofi, but less the “guarantee” of treatment in humans argued by Apotex [123].

This conclusion as to the promise of the patent was crucial, as it seems clear beyond any doubt that the utility that had been demonstrated prior to the relevant date was ample to establish potential use in humans, and this is enough for patentable utility (compare e.g. rosiglitazone (NOC) 2011 FC 239, (blogged here) Hughes J holding a primary murine screen sufficient to establish demonstrated utility based on a promise of potential use). Thus the validity of the patent turned on the construction of the promise. I have criticized this doctrine in several previous posts (here, and more generally here), on the basis that a perfectly good patentable invention may be held invalid because of the fine points of the interpretation of the disclosure, which is what happened here, even though the primary purpose of the disclosure is to explain the invention, not to define it, and in all other respects validity is measured against the claims. Another objection to this doctrine is that it is uniquely Canadian, and so serves as a trap when the disclosure is drafted in the context of a foreign filing. In this case Boivin J’s interpretation of the promise of the patent was reasonable enough; the problem is that validity should turn on the interpretation of the claims, not on the interpretation of description.

Even if we accept that utility is to be measured against usefulness in humans, a puzzle remains as to why that utility was not demonstrated, given that clopidogrel was apparently shown to have been better than ticlopidine, which itself was actually approved for human use. Boivin J’s discussion of demonstrated utility occupied only 11 paragraphs ([339] - [349]) of a 790 paragraph decision, and was addressed entirely to the reliability of a single human trial conducted on clopidogrel. He held the study was not reliable for a variety of reasons, the most compelling of which is that it was a double-blind study in which the blind was not broken until after the relevant date for assessing utility. If Sanofi did have evidence from the study at the relevant date, it would have been unreliable as the protocol would have been breached.

Even without this study, there was ample evidence, discussed above, of utility of prior compounds, in particular the racemate and ticlopidine, in humans, and ample evidence, including animal studies, that clopidogrel was equally effective and less toxic. All that was missing was a human study on clopidogrel itself. How much would that add to our knowledge of clopidogrel’s utility? It is true that the enantiomer may have different properties from the racemate in terms of efficacy and toxicity – that is the reason for separation – but it was established prior to the relevant date that clopidogrel was effective in animals, and it was also established that the racemate was effective in humans. There was no evidence to suggest that as a general matter an enantiomer that is effective in other animals is often ineffective in humans. Consequently, it seems to me not unreasonable to conclude that efficacy in humans was demonstrated as of the relevant date, even in the absence of a human study. Toxicity is more of an issue, as that may vary between species (baboons, according to the evidence, are prone to convulsions [476].) But presumably it is not necessary to demonstrate complete lack of toxicity in humans in order to prove utility in humans; many approved drugs, such as ticlopidine, have serious side-effects. As noted above, the evidence indicated that the racemate was better tolerated than ticlopidine, and from this it would seem to follow that clopidogrel would also be better tolerated than ticlopidine. (So far as I understand, an enantiomer will never be more toxic than the racemate, except to the extent that a given quantity of the racemate will have half the dosage of the same quantity of the enantiomer.) The fact that the side-effects of the racemate were sufficiently severe that Sanofi chose not to continue its development, given the compounds already on the market, does not mean that it is not useful for patent purposes. It seems not unreasonable to say that a skilled person would know that clopidogrel was safe enough to have patentable utility in humans, though perhaps not regulatory utility, even in the absence of the questionable human study.

Of course, this chain of reasoning above depends on some assumptions of fact on which expert evidence would be necessary before arriving at any conclusions, and Boivin J made no express findings as to these points, presumably because they were not argued. But my point is not that clopidogrel had demonstrated utility at the relevant date, but rather that it is not evident that human trials are needed to establish demonstrated utility in humans.

It might well be said that my arguments above, even accepting the factual premises, are pushing the boundaries of what we can say is “known,” rather than what it only “predicted” even though the prediction may be a very confident one. This is fair enough as a semantic point regarding the nature of “knowledge” rather than “prediction”, but surely this semantic point should not make any substantive difference. If our understanding is so close to knowledge that it is a difficult semantic question as to whether we “know” or “predict”, then as a matter of common sense it would seem to follow that the patent should be valid as a matter of sound prediction. Indeed, from the relatively cursory treatment of demonstrated utility in this case (11 paragraphs), as compared with sound prediction ([358] - [585] = 227 paragraphs), it seems to have been implicit that if sound prediction cannot be established, then a fortiori, neither can demonstrated utility.

Though logical, this is not the case, because of the second problematic aspect of Canadian patent law. This is the doctrine which originated in Eli Lilly / raloxifene (NOC) 2008 FC 142 aff’d 2009 FCA 97, that for sound prediction, though not for utility, the factual basis for the prediction must be disclosed in the patent itself. Thus the distinction between demonstrated utility and sound prediction, which is often razor fine, as in this case, is nonetheless crucial. On the issue of sound prediction, Boivin J concluded that Sanofi did have both a factual basis [488] and a sound line of reasoning [563] supporting the conclusion that clopidogrel was useful in humans. This is entirely consistent with the argument above, that the evidence may even have been sufficient to establish demonstrated utility. However, Boivin J held that sound prediction had not been established, solely on the basis that the disclosure of the factual basis for sound prediction was insufficient [584]. If utility had been based instead on demonstrated utility, then the patent would have been valid, as it is clear law that the basis for utility need not be disclosed in the patent itself: Consolboard v MacMillam Bloedel, [1981] 1 SCR 504 at 526; Pfizer / atorvastatin calcium (NOC), 2008 FCA 108 [57]-[62]; Pfizer / sildenafil (NOC) 2010 FCA 242 [82].

In a previous post I have argued that this doctrine is unsound. There is only one utility requirement in the Act, and there is no basis in the Act or the SCC case-law to support this dramatic distinction between demonstrated utility and sound prediction. The doctrine of sound prediction as set out by the SCC in Wellcome / AZT 2002 SCC 77 is a way of establishing utility, not an entirely new requirement. This doctrine is particularly problematic as it was established by the Federal Court only four years ago. Even apart from any concerns about integrating Canadian law into a worldwide patent system, the patents that are being litigated today were drafted when this doctrine did not exist. It is true that the law must change and develop, but the pharmaceutical industry depends on future patent protection to fund research and development that often takes years or decades between patenting and marketing. If the law changes in a way that renders invalid a patent that would have been valid under prior law, it is essential that there be some compelling policy reason for the change. In my view the strongest argument in favour of disclosure of the factual basis for utility, does not support the drastic consequences of this uniquely Canadian doctrine.