Must the Factual Basis for Sound Prediction Be Disclosed in the Patent?

Eli Lilly Canada Inc. v. Apotex Inc. / raloxifene (NOC), 2008 FC 142, 63 CPR (4th) 406, aff'd 2009 FCA 97, 78 CPR(4th) 388.

In Eli Lilly / raloxifene (NOC), 2008 FC 142 Hughes J held that the data supporting a sound prediction of utility must be disclosed in the patent itself [163-64]. This was the primary basis for holding the allegation of invalidity to be justified [183]. The holding that the disclosure must be in the patent was crucial, as Hughes J held that a good factual basis for the sound prediction did exist, and if it had been disclosed in the patent, that ground of attack would have failed [156 - 58]. Hughes J’s holding on this point was specifically affirmed by the Court of Appeal in 2009 FCA 97 [15] Noël JA: Desjardins, Trudel JJA. In this post I will argue that despite this clear holding, the doctrine that the factual basis for sound prediction must be disclosed in the patent itself is conceptually unsound and inconsistent with the leading cases, including those of the Supreme Court of Canada. Recent jurisprudence from the FCA has provided a clarification that will allow the application of the doctrine to be avoided in many, perhaps most cases, but we nonetheless are left with an unjustifiable and technical doctrine that is uniquely Canadian, and so may serve as a trap for unwary patent drafters.

Taking the conceptual difficulty first, it is well established that the data supporting utility need not be disclosed in the patent itself. Indeed, the patentee need not even explain how the invention is useful: 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]. Consequently, the raloxifene rule implies a sharp distinction between a patent where utility has been established at the relevant date, and one where utility is based on sound prediction. In the former case the patent need not disclose any data supporting utility, but in the latter it must: see 2010 FC 1065 [92]. The difficulty with this distinction is that it is not consistent with a purposive interpretation of the Patent Act. On the text of the Act, there is only one utility requirement, namely that which defines an “invention” to be “new and useful.” There is therefore no textual basis for treating demonstrated utility and sound prediction differently. The purpose of that utility requirement is to ensure that a patent is not granted for “mere speculation”: Wellcome / AZT 2002 SCC 77 [69]. A line must be therefore drawn somewhere in the development of an innovation from conception to commercialization. That line is drawn at the point of sound prediction. Thus there is no purposive basis for the distinction between demonstration of actual utility and sound prediction. An innovation that has demonstrated utility is closer to practical application than one where there is only a sound prediction of utility, and an innovation that has actually been delivered to consumers is closer again; but none are based on mere speculation, and therefore all are equally “useful” under the Act. Note that the Supreme Court in Wellcome / AZT consistently treated demonstrated utility and sound prediction on the same terms: [46], [52], [56].


Hughes J’s policy reasoning was as follows. (The Court of Appeal did not offer an independent rationale.) The prior art did not adequately establish the usefulness of the drug in question [155]. A human trial undertaken by Lilly did adequately establish the utility of the drug, as a matter of sound prediction [156]. The patent did not disclose the human trial. The patent therefore did not disclose any more information than the prior art. Since the disclosure of information is the quid pro quo for which a patent is granted, it follows that without disclosure of new information, there could be no valid patent: “It is clear that the '356 patent does not disclose the [human trial]. The patent does not disclose any more than [the prior art] did. The person skilled in the art was given, by way of disclosure, no more than such person already had. No "hard coinage" had been paid for the claimed monopoly. Thus, for lack of disclosure, there was no sound prediction” [163].

This is a reasonable argument on its face. However, exactly the same could be said if the human trial had been sufficient to demonstrate present utility, rather than sound prediction. In that case it is clear law, as noted above, that no disclosure of the study would have been required. In Consolboard, which established this proposition, the invention consisted of a new type of wafer board, with wafers of specified dimensions. The Federal Court of Appeal was of the view that in order to disclose the invention, it was necessary to specify why the invention was useful, and not simply how to make it. The implicit point is that otherwise the public would not be given the new technical information that was the quid pro quo for the grant of the patent. The Supreme Court disagreed, saying at 527 that “the Federal Court of Appeal has confused the requirement of s. 2 of the Patent Act defining an invention as new and "useful", with the requirement of s. 36(1) of the Patent Act that the specification disclose the "use" to which the inventor conceived the invention could be put. The first is a condition precedent to an invention, and the second is a disclosure requirement, independent of the first.” That is, so long as the invention is in fact useful, it is not necessary to explain the basis for that utility in the patent. It is the fact of the usefulness, and not the evidence for that utility, that is the “hard coinage” of information with warrants the grant of the patent.

The response to the concern that no new information has been disclosed is that the patent itself, by claiming the invention, implicitly discloses that it is in fact useful. On the facts of raloxifene, prior to the issuance of the patent, the prior art gave no reason to believe the drug was useful, but the grant of the patent itself tells the public what it did not previously know, which is that the drug in fact has utility. This may seem like a formal point, but consider that the patentee does not have an interest in applying for a patent if it does not have a factual basis establishing utility, because if it applies without that data, the patent will be invalid if even it is granted. Thus the fact that the patentee applies for the patent is a signal that it has data which is a factual basis for either sound prediction or demonstrated utility, even if that data is not disclosed in the specification itself. It must also be kept in mind that the Patent Office is entitled to refuse the patent for lack of utility if it is not satisfied that there is a factual basis for sound prediction. The Patent Office may require more evidence than is disclosed in the patent itself. Once the Patent Office is satisfied that the invention is in fact useful (and that the other requirements of the Act are satisfied), the Patent Office is required to issue the patent. The patent is presumed to be valid, and implicitly it is presumed that there is data supporting the assertion of validity, even if not disclosed, because the patent has been examined prior to grant.

Now consider the authorities. Both Hughes J and the Court of Appeal relied on paragraph 70 of the Supreme Court’s decision in Wellcome / AZT. Hughes J stated at [164] that “[a] considered reading of paragraph 70 of the AZT decision leads to the conclusion that the disclosure must be in the patent, not elsewhere. The public should not be left to scour the world's publications in the hope of finding something more to supplement or complete a patent disclosure. As the Supreme Court said at paragraph 70, the quid pro quo offered in exchange for the monopoly is disclosure. It must be in the patent.” The FCA decision at [18] was to the same effect.

Here is the relevant passage from the Supreme Court [70], with emphasis added:

“Thirdly, there must be proper disclosure. Normally, it is sufficient if the specification provides a full, clear and exact description of the nature of the invention and the manner in which it can be practised. It is generally not necessary for an inventor to provide a theory of why the invention works. Practical readers merely want to know that it does work and how to work it. In this sort of case, however, the sound prediction is to some extent the quid pro quo the applicant offers in exchange for the patent monopoly. Precise disclosure requirements in this regard do not arise for decision in this case because both the underlying facts (the test data) and the line of reasoning (the chain terminator effect) were in fact disclosed, and disclosure in this respect did not become an issue between the parties. I therefore say no more about it.”

With respect, it is impossible to read this as a requirement that the disclosure of the factual basis for the sound prediction be in the patent itself. Discussion of the disclosure requirements was expressly reserved.

Though the Supreme Court did not discuss the point explicitly, on their facts the leading cases are all inconsistent with any requirement that the factual basis for sound prediction be disclosed in the patent. The seminal decision is that of Graham J in Olin Mathieson Chemical Corp. v. Biorex Laboratories Ltd., [1970] RPC 157 (Ch D). While the patent in question, GB857547, was upheld on the basis of sound prediction, there was no disclosure whatsoever of the factual basis for that prediction in the patent itself. The disclosure stated the utility of the compounds (at p 1-2), but provided no evidence at all to support that assertion. Indeed, there was not even any mention of tests or other evidence. Olin Mathieson was accepted into Canadian law in respect of sound prediction by the Supreme Court in Monsanto Co. v. Commissioner of Patents, [1979] 2 SCR 1108, in which Pigeon J stated “I should say immediately that I am in full agreement with the decision of Graham J. in Olin Mathieson.” In Monsanto, the claims were for 126 compounds for use in preventing premature vulcanization of rubber. The specification disclosed details for the preparation of only three compounds, and data supporting the utility was provided for only 2 of those three (at p 16-17 of the disclosure). The Federal Court of Appeal upheld the examiner, who had rejected the application on the ground that “In order to sustain claims to a broad group of compounds, the specification must illustrate with reasonable certainty that all members of the group are capable of being prepared by the disclosed process of preparation and have the same utility (inhibiting premature vulcanization) upon which their patentability is based.” The Supreme Court reversed, saying “in order to succeed, such attack will have to be supported by evidence of lack of utility. At present there is no such evidence and there is no evidence that the prediction of utility for every compound named is not sound and reasonable.” This is an express statement that the patentee need not provide evidence of utility in the specification.

In Wellcome / AZT itself, the Supreme Court said in paragraph 70 that “both the underlying facts (the test data) and the line of reasoning (the chain terminator effect) were in fact disclosed,” but this summary paragraph must be read in context. At [73]-[75] the Supreme Court set out and specifically approved the factual basis relied on by Wetston J at trial. Some of the bases specified were set out in the disclosure, but others were not. In particular, neither the toxicity tests [73(i)], nor the evidence that AZT would be absorbed by oral administration [73(ii)], were disclosed in the patent itself. Moreover, Wetston J at [185], stated that his conclusion was based on evidence of human cell line testing that was disclosed in the patent, “considered cumulatively, in conjunction with all of the evidence adduced and considered in this trial.”  This passage was specifically cited by the Supreme Court at [73(iv)] as the appropriate factual basis for the trial judge’s conclusion.

Further, in Ciba-Geigy AG v. Commissioner of Patents (1982), 65 CPR(2d) 73 (FCA), which was cited in Wellcome / AZT [63] as an application of the doctrine of sound prediction, the Federal Court of Appeal held that sound prediction had been established on the basis of evidence “that was not part of the original disclosure” (78). This is clearly contrary to the raloxifene holding. Indeed, the Court held that it was permissible for the evidence to be based on “something done after the filing of the application” (ibid). This can no longer be considered good law since Wellcome / AZT, but it emphasizes that the courts have never considered it necessary for the basis for sound prediction to be disclosed in the patent itself.

In Pfizer / sildenafil (NOC) 2010 FCA 242 the Federal Court of Appeal revisited its decision in Eli Lilly / raloxifene. The Court clarified that there is no need to demonstrate sound prediction of utility in the patent itself; it is enough that the factual basis of the sound prediction is disclosed. This serves only highlight the artificiality of the doctrine. On the facts, the disclosure that was found to be sufficient was the statement that “In man, certain especially preferred compounds have been tested orally in both single does and multiple dose volunteer studies. Moreover, patient studies conducted thus far have confirmed that one of the especially preferred compounds induces penile erection in impotent males” (2009 FC 638 [81]).  Neither the details of the study nor the results were disclosed. The quoted passage amounts to no more than a statement that a study was done that supports the asserted utility. In other words, on Pfizer / sildenafil, it is enough advert to the existence of the studies which provide a factual basis for the prediction; the study need not be identified, and neither methodology nor results need be provided. This negates the original justification for imposing the more stringent disclosure requirement: it does not satisfy Hughes J’s concern in Eli Lilly / raloxifene [164] that “[t]he public should not be left to scour the world's publications in the hope of finding something more to supplement or complete a patent disclosure.” The statement in the patent that a study exists does not provide any more actual information to the public than the grant of the patent by itself. By granting the patent the Patent Office in effect warrants that it is satisfied that there is a factual basis for a sound prediction of utility; to insist that the patentee say that, but no more, in the patent itself, is the kind of technical objection to the specification which the Supreme Court cautioned against in Consolboard (at 520).

With that said, Pfizer / sildenafil (NOC) suggests that the FCA is beginning to resile from the doctrine that the factual basis must be disclosed in the patent itself. It is perhaps too much to have hoped that in 2010 it would have directly repudiated its 2009 decision in Eli Lilly / raloxifene. But even so, I submit that notwithstanding Eli Lilly / raloxifene, it would be open to a trial judge in the Federal Court to follow Wetston J in Wellcome / AZT in considering any results disclosed in the patent “in conjunction with all of the evidence adduced and considered in this trial” in deciding whether sound prediction is established.

---

Addendum: I now realize that I misinterpreted Pfizer / sildenafil in this post. It was held both at trial [88] and at the Court of Appeal [90, 102] that utility was based on demonstrated utility, not sound prediction. The decision does not, therefore, undermine the raloxifene doctrine itself. The fact that utility was upheld with only a passing reference to key study in the disclosure itself therefore serves only to reinforce the sharp distinction between utility based on sound prediction, for which the Federal Courts requires disclosure of the factual basis for the prediction in the disclosure, and utility based on demonstrated utility, in which no such disclosure is required.

It is my policy not to substantively revise my posts once written (though I will correct typographical errors etc. if I come across them), which is why I am posting this correction as an addendum.