Stare Decisis and Statutory Interpretation

Janssen Inc v Canada (Attorney General) Mactavish JA: Stratas, Rennie JJA 2021 FCA 137 affg 2020 FC 904 Zinn J

esketamine hydrochloride / SPRAVATO

The FCA decision in Janssen is a straightforward application of the doctrine of stare decisis, with a loose end that raises some interesting questions for the application of stare decisis in the context of statutory interpretation.

Data protection is available for an “innovative drug” pursuant to section C.08.004.1 of the Food and Drug Regulations. “Innovative drug” is defined to mean:

a drug that contains a medicinal ingredient not previously approved in a drug by the Minister and that is not a variation of a previously approved medicinal ingredient such as a salt, ester, enantiomer, solvate or polymorph.

In a split decision in Takeda 2013 FCA 13, the FCA held that this definition should be interpreted to mean that the enumerated variations—a salt, ester, enantiomer, solvate or polymorph—were per se ineligible for data protection, regardless of the effort necessary to develop the data: see here. Stratas JA, dissenting, would have held that data protection is available for an enantiomer or other enumerated variation if regulatory approval for the drug required the submission of confidential data generated by considerable effort and the medicinal ingredient is materially different in terms of safety and efficacy: [22], Takeda [92]–[93]. I described the interpretive issue as a “hard case” where the purposive considerations relied on by Stratas JA pointed in one direction, while the textual and contextual considerations relied on by the majority pulled the other way.

In this case, Janssen applied for data protection for SPRAVATO, in which the medicinal ingredient is esketamine hydrochloride, an enantiomer of ketamine hydrochloride. Ketamine hydrochloride is the medicinal ingredient in the previously approved drug KETALAR. The Minister refused Janssen’s application. In the decision under appeal, Zinn J dismissed Janssen’s application for judicial review of that decision: see here. Zinn J accepted that there were factual differences between the cases. In particular, in Takeda both drugs were developed by the same company and used in the same manner for the same indication [FC 25], while in this case the enantiomer is indicated for an entirely different use: KETALAR is a narcotic, while SPRAVATO is an antidepressant. Thus, as a matter of policy, there is a stronger case for granting data protection in this case than in Takeda. (Note that in Takeda Stratas JA did not hold that data protection should be available for the enantiomer at issue on the facts, but only that the matter should be remitted for redetermination: Takeda [104]–[107].) However, as Zinn J pointed out, these facts were not material to the majority reasoning in Takeda, which held that the listed variations were per se ineligible, regardless of the amount of effort involved in developing the data: [41], [FC 26], Takeda [141].

In this appeal, Janssen acknowledged that the decisions below were consistent with Takeda. Its primary argument was that Takeda was wrongly decided and that Stratas JA’s interpretation of the provisions was correct [42], [71]. The FCA rejected this argument on the grounds of stare decisis, which permits departure from prior decisions only in “exceptional circumstances”: [72]–[79]. The FCA provided a useful brief summary of the relevant principles, with the most basic being that “one panel of this Court ought not to come to a different conclusion from a different panel, merely because it is of the view that the first decision was wrongly decided” [73], citing Miller 2002 FCA 370. The fact that Stratas JA was a member of both panels makes no difference: [80]–[83].

I agree with this result on the facts of this case, but I’m not comfortable with the Miller test. The law of stare decisis involves a trade-off between legal predictability and accuracy. In the easy cases, there is no conflict; if the first decision to address a point of law is clearly correct, then we get both predictability and accuracy by following it. But what if the first decision is wrong? When should we sacrifice accuracy, by following a wrong decision, for the sake of predictability? The Miller criteria set a very high bar for departing from a prior decision, giving very little weight to accuracy. The prior decision may be overruled only when it is “is manifestly wrong, in the sense that the Court overlooked a relevant statutory provision, or a case that ought to have been followed”: [75] quoting Miller [10]. “Manifestly wrong” has a “narrow” meaning, which is tantamount to saying the decision was made per incuriam: Olanzapine 2013 FCA 282 [8]. (It is a bit odd that the test for whether the FCA must follow its own precedent is much the same as the test for whether a lower court must follow a higher court.) Miller does implicitly acknowledge the trade-off between predictability and accuracy, but it adopts a threshold test for accuracy; where accuracy is unimportant outside the narrow threshold of manifestly wrong. This means that with Miller there is significant room for a lock-in effect from the first decision, where the initial decision may be generally considered to be wrongly decided, but will be reinforced by subsequent cases because it does not rise to the point of being manifestly wrong.

An alternative might be some kind of balancing test, which would more explicitly consider how likely it is that the prior decision was wrongly decided, how important the consequences are, and how much actual loss of predictability there might be. It might be said that the very use of a balancing approach to precedent would undermine predictability, because it would always be uncertain how a later panel would assess that balance. But I’m not sure that’s so different from where we are now, where problematic decisions are ignored, or gradually distinguished into oblivion. (Though that is easier when the point is factually suffused, rather than a narrow point of law.) How much worse would it be to explicitly recognize that some precedent is better than others, rather than doing it sotto voce? Another possibility would be greater use of expanded panels or en banc review, or some other approach entirely. I don’t have any answers here, but I do see a problem.

The problem of wrong precedent being locked in is particularly important in the context of statutory interpretation. As Professor Nina Varsava puts it “Does the judiciary have legitimate authority to knowingly misinterpret and misapply statutory law?”: How to Realize the Value of Stare Decisis 30 Yale J L & Human 62 (2018) 118. If the first decision interpreting a statutory provision is “manifestly wrong,” it can be revisited, but I am not sure it is sound to stick with an interpretation that is generally considered to be wrong, even if it is not “manifestly” wrong.

I should emphasize that I don’t think the main source of potential inaccuracy is poor legal reasoning on the part of the first panel to decide the issue. On that score, I do think the Miller criteria, amounting to whether the decision was made per incuriam, is appropriate. That is what makes this case an easy one. As the Court noted, the point was fully argued in Takeda and “[b]oth the majority and the minority decisions are thorough and carefully reasoned” [76]. Moreover, the case was a close call—my own view is that the majority decision was correct—with good arguments on both sides. There is much to be lost in terms of predictability if different panels flip-flop simply because they weigh close arguments slightly differently, and little to be gained in terms of accuracy of interpretation, given that either decision is about equally likely to represent the notional true intent of the legislature.

There are two other potential sources of inaccuracy that are much more important. One is that the full implications of a decision may become apparent only over time, as they are applied to different factual scenarios. A decision may be entirely well-reasoned and “right” given the information available at the time it was decided, and yet turned out to be wrong with the benefit of hindsight. That is largely what happened when, in Hospira 2016 FCA 215, the FCA departed from Aqua-Gem 1993 CanLII 2939 (FCA) on the standard of review for discretionary decisions of prothonotaries. The first reason the FCA gave for departing from Aqua-Gem was that experience had shown that the Aqua-Gem standard was difficult to apply and had generated confusion that detracted from the effective review of discretionary orders made by prothonotaries (Hospira [46]–[50]). A second reason was the persuasiveness of the reasoning on the same question in the Ontario courts (Hospira [51]–[55]). As I argued in my comment on Hospira, the FCA did not follow Miller in deciding to overrule Aqua-Gem: neither of these factors is relevant under the Miller test. Aqua-Gem was not poorly reasoned at the time it was decided (Hospira [49]) and it certainly was not manifestly wrong. This illustrates that the Miller criteria are poorly suited to a situation where the hindsight that comes with time and experience shows a prior decision to have been “wrongly” decided, even though it was entirely reasonable on the information available to the court at the time. None of these concerns are relevant to this case, as Takeda was recently decided and, so far as I can tell, nothing in our subsequent experience would cause us to weigh any of the arguments any differently.

Another source of error is when the case is poorly argued. This is perhaps particularly a problem for trial decisions, because the sheer number of factual and legal issues to be addressed can mean that a point that turns out to be important is not fully argued. My sense is that comity is not as powerful in the Federal Court for that reason. If a novel issue of law has been decided in a single prior FC decision, a subsequent FC judge will generally make their own assessment of the arguments before them, with the prior decision considered as one factor with considerable persuasive effect. (I’m afraid I can’t put my hands on particular examples right now.) This strikes me as a good approach. There is not much loss of certainty in the law, because a holding of law by the Federal Court is always liable to be overruled by the FCA in any event, and once a novel point of law has been addressed once, it is likely to be more thoroughly argued the next time. The same kind of problem can arise at the FCA level as well, particularly in cases (unlike Takeda), where there are multiple legal and factual issues being appealed.

The problem of a poorly argued case is particularly troublesome in the context of statutory interpretation. It’s disquieting to think that the law might depart from the intent of the legislature simply because a private litigant had argued its case poorly in the first decision on the issue.

This brings me to the loose end in this case. Janssen argued that the FCA should admit new evidence on appeal with respect to the effect of the repeal of NAFTA and the coming into force of CUSMA, and the impact that CUSMA should have on the interpretation of the data protection provisions [44]–[45]. In particular, “Janssen points out that under CUSMA, Canada agreed to provide protection to “new pharmaceutical products”, which are specifically defined as products that do not contain a chemical entity that has previously been approved” [46]: see CUSMA Art 20.48–20.49. Further, “Janssen notes that OSIP explicitly found that the medicinal ingredient in SPRAVATO had not been previously approved in a drug in Canada, and that this would qualify SPRAVATO for data protection under CUSMA” [46]. The FCA refused to admit the evidence on procedural grounds. First, Janssen could and should have made the argument in the Federal Court: [49]–[52]. The FCA noted that CUSMA was finalized in December 2019 and came into force on July 1, 2020, just before Janssen served and filed its application record in the Federal Court on July 17, 2020, and before its application was heard on August 31, 2020. Janssen therefore had the opportunity to make substantive submissions that CUSMA had changed the data protection regime, but failed to do so. Second, the question facing the Federal Court “was not to decide whether esketamine hydrochloride should be entitled to data protection. Its task was to determine whether OSIP’s decision to refuse data protection for SPRAVATO was reasonable, based on the record before it. OSIP [the Office of Submissions and Intellectual Property] rendered its decision on April 25, 2019, more than a year before the coming into force of CUSMA and the exchange of correspondence between the parties. Consequently, the agreement and the documents could have had no bearing on OSIP’s decision with respect to SPRAVATO, and they were not relevant to the task that the Federal Court had to perform” [54]–[55]. Both of these points strike me as compelling so far as procedural fairness to Janssen is concerned.

But what happens next time the same issue arises? Substantively, Janssen’s argument strikes me as quite reasonable. It would certainly have been taken into account if the interpretative question has been argued for the first time after CUSMA had come into force and it is at least possible that it would have tipped the balance. But the interpretive question is no longer a matter of first impression. As a matter of stare decisis, is the coming into force of CUSMA an “exceptional circumstance” that would warrant departing from Takeda, even though the legislation itself has not changed? This question raises an interesting point of principle. Suppose that the CUSMA argument was a knock-down argument that made it clear that Stratas JA’s position was correct. (I should say that this is purely for the sake of argument; I really have no opinion how the CUSMA argument would play out if it were added to the interpretive mix.) If the enactment of CUSMA was not an “exceptional circumstance” as a matter of the law of stare decisis, the result would be that a clearly wrong interpretation of the legislation would prevail simply because a private party had failed to make the argument at its first opportunity. I find this to be a troubling possibility; though perhaps it is simply a reflection of the fact that the interpretation of legislation through private litigation inevitably means that the interpretation will be shaped in part by the litigation strategy of the parties. Professor Varsava goes to far as to suggest that we should abandon stare decisis in the context of legislative interpretation, in part for that reason: see Varsava, How to Realize the Value of Stare Decisis 30 Yale J L & Human 62 (2018) 117–18. I would not go that far, but the point is an interesting one.

Because the FCA decided on procedural grounds, it was not necessary to address this question, and the Court—wisely, in my view—did not discuss the substantive point even tangentially. The door therefore remains open for this argument to be made. I expect we will find out the answer in due course, after the CUSMA argument is properly made to OSIP and it winds its way back up to the FCA on a more complete record.

What Did Whirlpool Decide?

Apotex Inc v Eli Lilly Canada Inc 2016 FCA 267 Pelletier JA: Boivin, Rennie JJA aff’g 2015 FC 875 Gleason J
            2,226,784 / tadalafil / CIALIS / NOC

This brief decision of FCA dealing with obviousness-type double patenting is important for its discussion of the FCA's view of precedent, both at the SCC level and its own. The decision first addresses the question of when the SCC should be taken to have settled a question of law. This topic is important not just for the issue at hand, but more broadly as the Federal Courts interprets and applies new Supreme Court precedent. The FCA also suggested that counsel might want to be more creative in thinking about how to approach obviousness-type double patenting. On both points, the FCA’s attitude bodes well for the future development of the law.

The 784 patent was attacked as being invalid for double patenting over patent 2,181,377 in two separate NOC proceedings, one involving Mylan and the other involving Apotex. The key issue in all the proceedings was the appropriate date for assessing obviousness-type double patenting: is it the claim date of the earlier patent, the claim date of the later patent, or the publication date of the later patent?

Mylan FC 2015 FC 17 was decided in Lilly’s favour in January 2015 (see here). The decision under appeal in this case, Apotex FC 2015 FC 875 was decided, also in Lilly’s favour, in July 2015 (see here). Mylan FC was affirmed by Mylan FCA 2016 FCA 119, which was issued in April of 2016. As discussed here, in Mylan FCA, the FCA rejected the latest date, namely the publication date of the later patent, and held that it did not need to decide between the two other possible dates, because the FC had held that on the facts, the 784 patent was not obvious over the 377 patent on either of the claim dates. Only a few days after Mylan FCA was issued, the appeal from Apotex FC was heard [8]. Pelletier JA remarked that “In the normal course, given that the double patenting issue is identical in both cases, this Court would have simply followed Mylan FCA and rejected Apotex’s double patenting argument on the basis that it had already decided this issue” [8]. But Apotex argued that Mylan FCA should not be followed because it was inconsistent with the SCC decision in Whirlpool 2000 SCC 67. The question therefore, was “What did Whirlpool Decide?” [28ff].

I won’t rehash the details of the Whirlpool decision, which was summarized by the FCA at [28]-[33]. Suffice it to say that Apotex made a plausible argument that the SCC in Whirlpool would have assessed double patenting as of the publication date of the later patent, if the issue had been before it [36-[37]. However, Whirlpool didn’t explicitly decide, or even address, the question of the appropriate date. The issue is whether the appropriate date should be considered settled by implication through exegesis of the SCC’s statements, or whether the question should be considered to have been left open.

Despite the appeal of Apotex’s argument, Pelletier JA rejected it (my emphasis):

[37] It is, in my view, improbable that Binnie J. would settle a significant point in the law of double patenting by implication and without addressing it directly. The care taken to justify the publication date of a patent as the date at which it is to be construed suggests that if Binnie J. intended to decide the issue of the comparison date of the claims of the patents in an obviousness double patenting suit, he would do so explicitly.

[38] Furthermore, the issue of the comparison date in double patenting cases had not arisen previously in the Supreme Court jurisprudence, nor in the works of the text book writers. It would be surprising, to say the least, if Binnie J. purported to deal with a novel question by implication. This is all the more so when one considers that the issue was there to be addressed if the Court chose to do so.

He therefore held that the question of the appropriate date remained open after Whirlpool [39].

In my view, Pelletier JA’s reasoning is both sound as a matter of policy and strongly supported by the SCC’s discussion in R v Henry 2005 SCC 76 [57]. In Henry the SCC reconsidered a number of its prior decisions, overruling some of them. The issue of the binding effect of SCC statements arose because R v Noël, 2002 SCC 67 was affirmed on its facts (Henry [49]) and the Attorney General of Ontario worried that some of the obiter would be seen as binding on trial courts. Binnie J, for the Court, that “ I do not think this ‘concern’ is plausible. The comment was neither part of the legal analysis nor a direction to trial courts. It was simply an observation by an experienced judge” [52]. Binnie J reviewed the law relating to the binding effect of its decisions, and summarized as follows (my emphasis):

[57] The issue in each case, to return to the Halsbury question, is what did the case decide? Beyond the ratio decidendi which, as the Earl of Halsbury L.C. pointed out, is generally rooted in the facts, the legal point decided by this Court may be as narrow as the jury instruction at issue in Sellars or as broad as the Oakes test. All obiter do not have, and are not intended to have, the same weight. The weight decreases as one moves from the dispositive ratio decidendi to a wider circle of analysis which is obviously intended for guidance and which should be accepted as authoritative. Beyond that, there will be commentary, examples or exposition that are intended to be helpful and may be found to be persuasive, but are certainly not “binding” in the sense the Sellars principle in its most exaggerated form would have it. The objective of the exercise is to promote certainty in the law, not to stifle its growth and creativity. The notion that each phrase in a judgment of this Court should be treated as if enacted in a statute is not supported by the cases and is inconsistent with the basic fundamental principle that the common law develops by experience.

This statement emphasizes that even direct statements by the SCC expressing a view on a point of law are not necessarily binding; it follows that inferences as to what the Court might have been thinking can never be binding. 

In his discussion, Binnie J acknowledged that in the 1970s the Court’s mandate shifted from error correction to development of the jurisprudence, and consequently, the Court’s work became more oriented to “the development of a general analytical framework which necessarily went beyond what was essential for the disposition of the particular case” [53]. While Binnie J noted that this is particularly true in the Charter context, there is no doubt that we have seen a similar shift in patent law as well. Binnie J acknowledged that this shift meant that some statements that are strictly dicta should nonetheless be considered to be effectively binding: “It would be a foolhardy advocate who dismissed Dickson C.J.’s classic formulation of proportionality in Oakes as mere obiter” [53]. Nonetheless, it is still true that “the common law develops by experience,” and much of that experience is gained at the trial and appellate levels. Binnie J cautioned that the lower courts should not be too ready to overly broaden what the SCC had decided, because:

the effect would be to deprive the legal system of much creative thought on the part of counsel and judges in other courts in continuing to examine the operation of legal principles in different and perhaps novel contexts, and to inhibit or skew the growth of the common law. This would be a consequence totally unforeseen and unintended by the Court that decided Sellars.[56]

The Supreme Court of Canada is the highest court in our system, and we like to think that it has some of the best judges, but very able judges are found at every level of court, and the Supreme Court justices are not omniscient oracles. A judge in the Federal Court or the Federal Court of Appeal, will have more insight into the proper disposition of a legal question that is immediately raised on the facts before her and fully argued by counsel, than can be found in the inferred intent of a Supreme Court judge. Binnie J’s assertion that it is not “plausible” that obiter statements by the SCC could be seen as effectively binding on lower courts, is, with due respect, clearly wrong, as we know well in patent law, but the remainder of his discussion is entirely compelling.

In summary, not only was the FCA in this case right to hold that the remarks in Whirlpool were not binding authority on the issue of the correct date for assessing obviousness-type double patenting, to have held otherwise would have been inconsistent with the SCC’s own guidance in R v Henry on the binding effect of SCC decisions.

Finally, on a point of considerable practical importance, Pelletier JA concluded with the following observation:

[40] I might add that the analysis in the Earlier Cases, in Mylan FCA, and in this case was driven by the fact that the parties chose to frame the issue in terms of the date for comparison of the claims of the patents in issue. Having conducted the analysis on the basis chosen by the parties, we should not be taken as having decided that this framework for analysis is the correct one. The fact that this issue has not arisen in this form in the past may be an indication that there may be other ways to approach it. Perhaps, the Court, having construed the claims of each of the patents with the assistance of the persons skilled in the art, simply compares the claims and decides whether the later claims are patentably distinct from the earlier claims on the basis of the insights which it has gained in the course of the construction of the patents. This appears to be what was done in [Comm’r of Patents v Hoechst [1964] SCR 49], and more recently in [Sanofi 2008 SCC 61]. This is not to argue that this approach is any more correct than the comparison date approach but rather that, going forward, parties should not feel that they are locked into the framework chosen by the parties in these cases.

This observation is of a piece with Pelletier JA’s discussion of Whirlpool. Adherence to precedent is important to the predictability which allows parties to plan in reliance on the law, but sound precedent cannot grow in a straightjacket imposed by the idiosyncracies of how particular cases were argued: to repeat Binnie J’s words, “The objective of the exercise is to promote certainty in the law, not to stifle its growth and creativity."

The Complexities of Comity and Stare Decisis

Bayer Inc v Apotex Inc 2016 FC 1013 Fothergill J
            2,382,426 / micronized drospirenone / YAZ YASMIN

As noted in yesterday’s post Bayer’s ‘426 patent covers an oral contraceptive comprising drospirenone and ethinylestradiol, where the drospirenone is in the form of fast dissolving particles. The claims of the ‘426 had previously been construed by Hughes J and the FCA in the prior NOC proceedings: Bayer v Cobalt 2013 FC 1061 J aff’d 2015 FCA 116 (blogged here). This raised the question of the extent to which Fothergill J was bound to follow the claim construction arrived at by Hughes J (on the basis of comity) and the FCA (on the basis of stare decisis) [45].

Bayer argued that Fothergill J should adhere to the FCA’s construction “unless the evidence demonstrates the prior construction was wrong, or if different evidence compels a different result” [46], relying on Allergan 2012 FCA 308, [44] (here). This position is in some ways surprisingly modest. Allergan did state that test, but in the context of comity, which governs the relationship between courts of the same level. As noted in Pfizer 2014 FCA 250 [59] (here) (also relied on by Bayer), claim construction is a matter of law, and stare decisis implies that a subsequent lower court should be strictly bound by a holding of law by a higher court. On the other hand, Cobalt and Apotex pointed out that because of their summary nature, even holdings of law in an NOC proceedings are arguably not strictly binding at all in a subsequent infringement action: Lilly v Novopharm / nizatidine [1998] 2 SCR 129 [95]-[96]; Lilly v Novopharm /olanzapine 2007 FCA 359 [41].

The Limits of Stare Decisis

Hospira Healthcare Corporation v The Kennedy Institute of Rheumatology, 2016 FCA 215 Nadon JA: Pelletier, Rennie, de Montigny, Gleason JJA
            2,261,630 / infliximab / INFLECTRA

As discussed in yesterday’s post, in Hospira the FCA held that the standard of review for discretionary decisions of prothonotaries set out in Aqua-Gem should be abandoned and replaced by the standard of review set out in Housen 2002 SCC 33. As well as the issue of whether Aqua-Gem should be abandoned, there was a question as to whether Aqua-Gem could be abandoned, because the Aqua-Gem standard had been approved by the SCC, and it was established FCA law. Hospira therefore raises the question of when the FCA can depart from SCC authority, and when it should overrule its own prior decisions.

How High a Bar for Overruling Prior FCA Decisions?

Eli Lilly Canada Inc. v. Apotex Inc. / olanzapine (NOC) No.2 2013 FCA 282 Evans JA: Stratas, Webb JJA aff’g 2010 FC 952 Gauthier J

 Under the “no reach back” rule established in Apotex v Syntex / naproxen (NOC) 2010 FCA 155 aff’g 2009 FC 494 and re-affirmed in Pfizer Canada Inc. v Ratiopharm / amlodipine besylate (NOC) 2011 FCA 215 (blogged here) the Court will not reach back and retroactively set aside an order of prohibition obtained in an NOC proceeding solely because the patent has been held invalid in a subsequent action. The consequence is that the generic will not be entitled to damages under s 8 of the NOC Regulations, which is triggered by success in the NOC proceeding itself, even though it will have been held off the market by an invalid patent. (See here for my posts on related cases).

In this proceeding, Apotex asked the FCA to reconsider this rule. The FCA declined, on the basis of stare decisis:

[8] Because this Court is normally bound by its own decisions, Apotex can only succeed in this appeal if it satisfies us that Syntex and Ratiopharm should not be followed because they are "manifestly wrong" within the narrow meaning of Miller v. Canada (Attorney General), 2002 FCA 370 at paras. 8, 10, and 22 (Miller). They were not per incuriam, nor subsequently overruled or seriously attenuated by decisions of the Supreme Court of Canada.

It is interesting to contrast the “manifestly wrong” standard applied by the FCA in overruling its own decisions, with the standard set out by the SCC in Canada v. Craig, 2012 SCC 43 [24]-[28] for overruling the SCC’s own decisions. The SCC does not provide a rule, but rather (unsurprisingly, from the SCC) engages in “a balancing exercise between the two important values of correctness and certainty,” [27], giving “careful and respectful consideration” [26] to the earlier decision. This appears to be a significantly lower standard than “manifestly wrong.” That impression is confirmed by the result in Craig, which the SCC overruled a prior decision on the basis that it was wrong (though there was no suggestion that it was manifestly wrong), and there had been significant judicial and academic criticism of the prior decision.