ASTRAZENECA CANADA INC. v. AKTIEBOLAGET HÄSSLE, 2017 FC 726

Date: 20170726 Dockets: T-1409-04 T-1890-11 T-2300-05 Citation: 2017 FC 726 Ottawa, Ontario, July 26, 2017 PRESENT: The Honourable Mr. Justice Barnes Docket: T-1409-04 BETWEEN: ASTRAZENECA CANADA INC. AND AKTIEBOLAGET HÄSSLE Plaintiffs (Defendants by Counterclaim) and APOTEX INC. Defendant (Plaintiff by Counterclaim) Docket: T-1890-11 AND BETWEEN: ASTRAZENECA AB AND AKTIEBOLAGET HÄSSLE Plaintiffs (Defendants by Counterclaim) and APOTEX INC. Defendant (Plaintiff by Counterclaim) Docket: T-2300-05 AND BETWEEN: APOTEX INC. Plaintiff and ASTRAZENECA CANADA INC.

Defendant PUBLIC JUDGMENT AND REASONS Table of Contents I. During the Period of Infringement of the 693 Patent, Did Apotex Have an Available Non-Infringing Alternative 4 A. Are the Proposed NIAs Bioequivalent to LOSEC? . 23 B. Could Apotex Have Conducted Human Clinical Trials on its Proposed NIAs to Prove Bioequivalency? 38 C. Are the Proposed NIAs Sufficiently Stable? . 41 D. Would the Proposed NIAs Have Received Regulatory Approval? . 59 E. Would Apotex Have Obtained and Used a Third-Party NIA? . 63 F. Conclusion on the Availability of a NIA .. 79 II. How Should the Court Reconcile the

Section 8 Judgment in Favour of Apotex in Court Docket T-2300-05 with the Infringement Judgment in Favour of AstraZeneca in Court Dockets T-1409-04 and T-1890-11 . 79 III. With Respect to Apotex’s Profits From the Infringement of the 693 Patent, What Allowance Should be Made for Profits-On- Profits . 92 A. Tax Effects on Profits-on-Profits . 94 IV. With Respect to the Infringement of the 693 Patent, What Allowance is Required Having Regard to the United States District Court Award For the Infringement of the United States 505 Patent And Apotex’s Satisfaction of That Award . 95

V. Disposition. 106 [1] In these bifurcated proceedings, AstraZeneca Canada Inc., Aktiebolaget Hässle and AstraZeneca AB [collectivelyAstraZeneca] seek an accounting of the profits earned by Apotex Inc. [Apotex] from the infringement of AstraZeneca’s Canadian LettersPatent No 1,292,693 [693 Patent]. In the liability phase of the infringement actions (Court dockets T-1409-04 and T-1890-11), the Court found in favour of AstraZeneca: see AstraZeneca v Apotex, 2015 FC 322, 134 CPR (4th) 1, aff’d in

part 2017 FCA 9, [2017] FCJ No 22(QL). The commercial product covered by the 693 Patent is an omeprazole formulation marketed by AstraZeneca in Canada under thetrade name LOSEC. The period of infringement by Apotex runs from September 5, 2003 to December 3, 2008. [2] In Court docket T-2300-05 Apotex, in turn, seeks an offset for its

section 8 damages for being held out of the Canadianmarket for its generic omeprazole formulation [Apo-Omeprazole] between January 3, 2002 and December 30, 2003 by reason ofAstraZeneca’s failed Patented Medicines (Notice of Compliance) [NOC] application concerning its 762 Patent. [3] All of these references were consolidated by a case management Order dated December 11, 2013 and were tried together atToronto. [4] To their considerable credit, the parties have resolved most of their quantification issues and have reduced their agreement towriting [see Exhibit AZ 24].

They have also agreed that their respective accounting experts will adjust their calculations as required bythe streamlining agreement and by any other issues resulting from the Court’s Judgment. Any remaining points of disagreement betweenthe accounting experts will be referred to the Court for final resolution. [5] The parties have left with the Court the following matters for determination: (

a) During the period of infringement of the 693 Patent, did Apotex have an available non-infringing alternative [NIA]; (

b) How should the Court reconcile the

section 8 Judgment in favour of Apotex in Court docket T-2300-05 with theinfringement Judgment in favour of AstraZeneca in Court dockets T-1409-04 and T-1890-11; (

c) With respect to Apotex’s profits from the infringement of the 693 Patent, what allowance should be made for profits-on-profits; and (

d) With respect to the infringement of the 693 Patent, what allowance is required having regard to the United States DistrictCourt award for the infringement of the United States Patent No 4,786,505 [505 Patent] and Apotex’s satisfaction of that award. [6] The matter of costs is to be left pending further submissions from the parties. I.

During the Period of Infringement of the 693Patent, Did Apotex Have an Available Non-InfringingAlternative [7] It is now well established in Canadian law that a NIA defence is available to a patent infringer to potentially reduce aninnovator’s claim to damages or to the recovery of the infringer’s profits. [8] The onus rests on Apotex to prove that a NIA was available and at what cost.

This point was made in Reading & BatesConstruction Co v Baker Energy Resources Corp (1992), 44 CPR (3d) 93 at pp 106-107, 56 FTR 22 (FCTD), aff’d (1994) (FCA), 58 CPR (3d) 359, 175 NR 225 (FCA), where Justice Barry Strayer held: I also agree with the learned referee’s conclusions of law that the onus is on the defendant to prove that an alternative non-infringingmethod existed and the costs of using that method.

Although the defendant cited several cases to the contrary, these were cases from theCircuit Courts of the United States, one of which was over 100 years old and none of which were less than about 50 years old. On theother hand, I believe that such Canadian jurisprudence as exists is consistent with the burden being on the defendant to prove thealternative and its cost. It has been held in this court, for example, that in accounting for profits the burden is on the defendant to provehis costs, and thus establish the net profits from his sales: [citations removed].

Consistently with this fundamental principle, it is equallyincumbent upon the defendant to prove his real net profits from using the infringing method by establishing on a balance of probabilitieswhat his costs would have been had he used the most likely non-infringing alternative method. Therefore, the learned referee was right inlaw in imposing that burden on the defendant in this case.

Also see Apotex Inc v Merck & Co, 2015 FCA 171 at para 74, 387 DLR (4th) 552 [Lovastatin FCA], and Pfizer v Teva, 2016 FCA 161 at paras 53-66, 400 DLR (4th) 723, where the Court said: “[m]ere possibilities short of probabilities do not suffice” [para 56]. [9] The NIA defence was initially received in Canada with some hesitation perhaps because of the qualified language used inMonsanto Canada Inc v Schmeiser, 2004 SCC 34, [2004] 1 SCR 902.

More recently, however, the Federal Court of Appeal has fullyendorsed the defence, at least in conceptual terms: see Lovastatin FCA and Apotex Inc v ADIR, 2017 FCA 23, [2017] FCJ No 110 (QL)[Perindopril FCA]. But, as with any legal principle, the real challenge lies in its application to the evidence. This case is no different. [10] Lovastatin FCA, above, contains a useful discussion of the theory behind the NIA defence and the method of applying it. At

its root is the need for a causal link between the infringement and the claimed recovery. Behind the application of the NIA idea is saidto lie “robust common sense” about what would and could have happened “but for” the infringement. The following passages from thedecision are particularly instructive: [48] The difficulty with the Judge’s approach is that if damages for lost profits are calculated never having regard to an availablenon-infringing alternative, the patentee will sometimes be better off than it would have been in the absence of infringement. This is so forthe following reason.

Where a defendant can make and sell a non-infringing alternative, the patent does not confer a complete monopolyon the patent holder. Instead, the patent confers a share of market power upon the patentee. In this circumstance, where, instead of usinga non-infringing alternative, a defendant infringes, it is a question of fact whether, “but for” the infringement, the defendant would nothave competed with it.

The defendant’s lawful competition in the “but for” world may have deprived the patentee of some sales. [49] Put another way, in cases where, in the “but for” world, the infringer could and would have made and sold a non-infringingalternative, these sales may well reduce the patent owner’s sales. Awarding the patentee full damages for lost profits in every case will,therefore, sometimes over-compensate the patentee. [50] Perfect compensation requires consideration of: (

i) what, if any, non-infringing product the defendant or any other competitorscould and would have sold “but for” the infringement; and, (ii) the extent lawful competition would have reduced the patentee’s sales. … [73] When considering the effect of legitimate competition from a defendant marketing a non-infringing alternative, a court isrequired to consider at least the following questions of fact:

i) Is the alleged non-infringing alternative a true substitute and thus a real alternative? ii) Is the alleged non-infringing alternative a true alternative in the sense of being economically viable? iii) At the time of infringement, does the infringer have a sufficient supply of the non-infringing alternative to replace the non-infringing sales?

Another way of framing this inquiry is could the infringer have sold the non-infringing alternative? iv) Would the infringer actually have sold the non-infringing alternative? [74] As a matter of principle, the burden lies on the defendant to establish the factual relevance of a non-infringing alternative on abalance of probabilities. Indeed, Apotex acknowledged in oral argument that it bears the persuasive burden, on a balance of probabilities,to prove that it would have used the non-infringing alternative. This is consistent with jurisprudence such as Rainbow Industrial CaterersLtd. v.

Canadian National Railway Co., (SCC), [1991] 3 S.C.R. 3, 84 D.L.R. (4th) 291. [Emphasis in original.] … [89] While this is dispositive of the appeal on this issue, I also find that Apotex failed to establish that it would have replaced itsinfringing sales.

I reach this conclusion on the following basis. [90] First, as Apotex conceded in oral argument: • The real world informs our construction of the “but for” world. • Conduct in the real world is “very important” to what would have happened in the “but for” world. • Findings of fact from the liability decision are relevant to constructing the “but for” world. • “Brazen” infringement in the real world makes it very difficult to prove that the defendant would have deployed the non-infringing alternative in the “but for” world. [91] In the liability phase, the Judge found, at paragraph 309 of her reasons (reported at 2010 FC 1265), that if Blue Treasure hadbeen using the non-infringing process to ferment lovastatin, it would have lost significant amounts of money for each kilogram ofproduct it shipped to AFI.

However, Apotex knew that once Blue Treasure began to use the allegedly non-infringing process it becameprofitable.

The inference to be drawn is that Apotex knew Blue Treasure was in fact using the infringing process; yet Apotex used thatbulk product to prepare and sell its lovastatin tablets. [92] In this circumstance it is relevant to note that from January 1, 1997 to January 1, 2001 Apotex believed Merck’s patent wasinvalid. [93] Apotex’ evidence falls far short of demonstrating that it would have sold the non-infringing product when one considers: thescale of Apotex’ infringement; its likely knowledge that Blue Treasure was supplying it with infringing lovastatin; its belief the Merckpatent was invalid; its failure to call a witness from AFI to support its contention that, had it known the product was infringing, it wouldhave resurrected operations at AFI in Winnipeg; and the fact the Judge found that the testimony of Apotex’ only fact witness was, albeitnot on this point, unsubstantiated and self-serving. [11] The NIA defence was more recently endorsed in Perindopril FCA, above.

There Apotex advanced the defence based on theasserted availability of the patent-protected product from certain foreign sources for sale into non-infringing markets. The Courtexpressly rejected the idea that a NIA could not take the exact form of the patented product. Such an approach, it said, wouldinappropriately extend the territorial reach of the Canadian patent into non-infringing jurisdictions. The Court was also unperturbed bythe fact that, at the beginning of the infringing period, none of the identified foreign third-party suppliers of Perindopril had the

compound at hand. The question posed was whether, in the hypothetical world, Apotex could and would have obtained sufficientquantities of non-infringing product and that it could and would have used that product [see para 41]. The Court discussed this point inthe following way: [42] As this Court later explained in Pfizer Canada Inc. v. Teva Canada Limited, 2016 FCA 161, 483 N.R. 275, (Effexor) at paragraph50, both the “could have” and “would have” requirements are important. To prove “could have”, the defendant must demonstrate that itwas possible for it to secure non-infringing product.

To prove “would have”, the defendant must demonstrate “that events wouldtranspire in such a way as to put them in that position” (Effexor, paragraph 50). The importance of the “would have” requirement is thatby requiring a defendant to show that it would have used a non-infringing alternative, the defendant shows that the value of the patentedinvention is not such that reliance on alternatives is unlikely or fanciful.

Put another way, notwithstanding the availability of a non-infringing alternative, the defendant must show that there are no impediments to its use. [12] AstraZeneca contends that the jurisprudence does not support a NIA that is not perceived by the infringer to be non-infringingat the point of the infringement. It also posits that a NIA must be “foreseeable” to the infringer at the relevant time.

Anything short ofthis is said to be speculative. [13] In support of the “knowledge” requirement, AstraZeneca relies on the trial decision in Wellcome Foundation Ltd v Apotex Inc(1998), (FC), 82 CPR (3d) 466 at paras 32-33, 151 FTR 250 (FCTD) [Wellcome FC], aff’d (FCA), [2001] 2 FCR 618, 11 CPR (4th) 218 (CA). AstraZeneca cites to Lovastatin FCA, above, at paras 93-95 on the issue offoreseeability. [14] I do not read these decisions as broadly as AstraZeneca suggests.

In Wellcome FC, above, Justice MacKay did focus onwhether Apotex had actual knowledge that its proposed NIA was non-infringing, but he also considered whether “it could have known”[para 33]. Knowing whether or not a proposed NIA would infringe is, of course, a factor in determining whether the infringer “wouldhave” employed it in place of the infringing product. But this falls well short of making prior knowledge of non-infringement anabsolute pre-requisite to the assertion of a NIA. [15] I also place little significance on the stray reference to “foreseeability” in Lovastatin FCA, above.

In the context of its use Itake that reference to mean only that the concept of a viable NIA would have been available to the infringer based on what was known inthe art at the time.

If foreseeability meant that the infringer must have the asserted NIA in mind at the time of the infringement, it couldpotentially punish those who had no idea their product was infringing while rewarding those who had an appreciation of the risk andcourted it, but nevertheless had a back-up, work-around solution available. [16] In its Closing Argument on NIA at para 56, AstraZeneca cites two United States authorities (Grain Processing Corp vAmerican Maize-Products Co, 185 F 3d 1341 (Fed Cir 1999) [Grain Processing], and Micro-Chemical Inc v Lextion Inc, 318 F 3d 1119(Fed Cir 2003) [Micro-Chemicals]) for the idea that a NIA requiring the infringer to “invent around the patented technology” is notconsidered to be “available” to the infringer.

I do not agree with this

interpretation and in oral argument counsel retreated somewhatfrom the above proposition. Neither Micro-Chemical, above, nor Grain Processing, above, stand for the idea that the availability of aNIA is necessarily contingent on the amount of inventive effort required to make it. The time and effort of coming up with a non-infringing solution is certainly relevant to whether the infringer would have pursued it, but they are not absolute barriers to the defence.

That this was all Judge Rader for the Court was saying in Micro-Chemical is clearly evident from his statement at p 1123 that high costsand the complexity of the exercise “to design or invent around the patented technology to develop an alleged substitute weighs [sic]against a finding of availability”. The Court in Grain Processing makes the same point. [17] The American authorities cited by the parties also do not, on my reading, support an argument for exclusion of a NIA that isnot “on the market” at the time of infringement.

In Grain Processing, above, the Court was only concerned with the hypotheticalavailability of a NIA “including but not limited to products on the market” [p 1349]. Where the substitute was not on the market at therelevant time, the Court observed that an inference of unavailability could be drawn but not that it must be drawn. The Court went on tosay at p 1353 that “the trial court must proceed with caution in assessing proof of the availability of substitutes not actually sold duringthe period of infringement”.

In that case, however, the trial court had found that the asserted substitute could have been made by aprocess that was known in the art. That finding was upheld on appeal. I can see nothing in the Micro-Chemical decision that detractsfrom the above view. [18] There is, of course, a difference between cases like Perindopril FCA and this one. In Perindopril FCA the NIA was known toexist at the time of infringement. The NIAs Apotex proposes in this case were unknown and never made by anyone before or during theinfringing period let alone approved for use in Canada, the United States or elsewhere.

Notwithstanding this distinction, I acceptApotex’s point that in the hypothetical, but for pharmaceutical world the infringer’s failure to produce a viable NIA formulation in thereal world is not a threshold bar to the use of the NIA defence.

In this context, the question is: Could the infringer have made theproduct had it attempted to do so at the relevant time and would the infringer have sold the product on some reasonable financial basis insubstitution for the infringing product? [19] I think this is the point being made by Justice Eleanor R Dawson for the Court in Perindropril FCA, above, when she said atpara 62 “the fact that an event does not take place in the real world does not necessarily mean that the event could not and would nothave taken place in the hypothetical world”.

Added to this is the recognition in Perindropril FCA that the availability of a NIA is not tobe foreclosed simply because it was not immediately available to the infringer, i.e. on the eve of first infringement. The Court is stillobliged “to consider whether at some later point in time a supplier would and could have provided” a replacement product [see para 67].

This lends support to Apotex’s view that a viable NIA need not exist at the exact time of infringement. [20] All of this is not to say that the post-infringement development of a NIA does not present problems of proof for the infringerasserting the defence. Indeed, as explained below, serious problems of proof are manifest in this case. [21] One of the difficulties with an ex post facto NIA solution was recently discussed in Bell Airbus Helicopters SAS v Bell Helicopter Texteron Canada Limitée, 2017 FC 170 at para 295, 144 CPR (4th) 281 [Airbus]. There Justice Luc Martineau explained that

the Court must be very wary of hindsight bias when it considers the claimed ease with which an after-the-fact NIA could be developed, tested, scaled-up and approved for use. In a case where the use of a product carries considerable infringement risk, one is left to wonder why the supposedly simple, non-infringing, equal cost version was never attempted. The “could have and would have” evidentiary concerns are also magnified when the proposed hypothetical NIA(

s) were never, at any time, submitted to the relevant regulator for assessment and approval. [ 22 ] I do not, however, think that Justice Martineau’s decision in Airbus , above, stands for the proposition that ex post facto NIAs of the sort proposed in this case must be excluded from consideration as a matter of law. Justice Martineau simply expressed reservations about the dangers of relying on a NIA that was either unknown during the period of infringement or had been previously discarded.

He was appropriately concerned about the reliability of this type of look-back evidence and the risk of hindsight bias [see para 295]. [ 23 ] I have similar concerns to those expressed by Justice Martineau about the NIA evidence presented by Apotex in this case concerning its recently developed in-house NIA formulations. [ 24 ] It is one thing to rely upon a NIA that is known and available for use during the period of infringing activity. It is quite another thing to propose a NIA made long after an infringement has taken place.

When a pharmaceutical NIA has been created and has obtained regulatory approval, one is not left to wonder whether it “could” have been available for use (assuming a capacity to obtain it in commercial amounts). In this case, however, Apotex’s self-created NIAs were made in non-commercial batches, without full stability, bioequivalency or clinical studies, and without obtaining the required regulatory approvals for commercial use. Indeed, Apotex had no intention of ever developing these formulations for commercial exploitation.

Many questions, therefore, remain about whether and, if so, when any of the formulations could have been used successfully during the period of infringement. [ 25 ] Apotex attempts to explain away the evident weaknesses in its testing evidence with the argument that AstraZeneca and its experts misconceived Apotex’s NIA burden.

Apotex puts the issue in the following way, at para 113 of its Closing Submissions: …The issue before the Court is whether one or more of the NIA formulations could meet regulatory requirements had Apotex manufactured them at a commercial scale and made the requisite regulatory filings, not whether the data generated is sufficient to meet regulatory standards. • Were Apotex to have done what Astra requires of it, millions of capsules would have had to be manufactured and studied over the period of a year.

Moreover, hundreds of humans would have needlessly been subjected to clinical studies. [ 26 ] The difficulty with the above idea is that, without ever acquiring the data necessary to satisfy regulatory requirements for its proposed NIAs, Apotex cannot directly establish that any of them would have obtained that approval. Incomplete or inconclusive data is weak data. The fact that Apotex began its stability testing too late to get it finished before trial and did not conduct clinical bioequivalency research at all does not make its case for NIA viability any stronger.

The same can be said of the experimental short-cuts and less-than-optimum testing protocols employed by Apotex in the generation of its stability data.

While these approaches may be entirely appropriate for the purpose of making in-house formulation choices to advance product development, they have diminished probative value where the question is whether a particular formulation would have been sufficiently viable to obtain regulatory approval on a balance of probabilities. [ 27 ] It is also of some significance that Apotex unsuccessfully asserted a NIA defence in the damages-assessment phase of the United States litigation.

Apotex argued there that it could have made adjustments to the infringing formulation, adopted an existing non- infringing formulation or used a microtablet formulation. These arguments were wholly rejected by the United States District Court for the Southern District of New York [District Court] in AstraZeneca AB v Apotex Corp, 985 F Supp 2d 452 (2013).

The Court characterized Apotex’s proposed formulation adjustments in the following way, at p 499: As for Apotex's proposals for tinkering with the ingredients in its pellets, it is pure speculation whether any of its various proposals would create a stable, bioequivalent product that was non-infringing. Apotex has never asked one of its many experts to try to create the revised formulation, much less to create and test it. See SynQor, Inc. v.

Artesyn Techs., Inc., 709 F.3d 1365 , 1382 (Fed.Cir.2013) (where an alleged substitute is not on the market, “the accused infringer has the burden to overcome the inference that the substitute was not ‘available’”) (citation omitted). There is a reason that Apotex chose the ingredients that it did for its pellets following six years of research and testing. Those ingredients created a successful product.

This is no easy task given the challenges of working with the omeprazole molecule and delivering it sufficiently intact to the part of the body in which it is most effective. [ 28 ] In this case, Apotex belatedly attempted to overcome the problem identified by the District Court by developing a set of alternative formulations. However, Apotex has not adequately explained why it waited until late 2015 to begin its stability testing when it knew or ought to have known as of 2007 from the United States litigation that Apo-Omeprazole infringed AstraZeneca’s formulation patents.

Inexplicably, Apotex mounted a purely theoretical NIA posture in the damages-assessment phase of the United States proceeding and by the end of that case, the Court observed at p 449 that it had “largely abandoned its argument that it could have altered the infringing formulation successfully” . The rejection by the District Court of Apotex’s NIA defence was based on a different and presumably weaker evidentiary record than the record before me. Nevertheless, I am left to wonder why Apotex failed to work-up its asserted alternative formulations in this case long before the end of 2015.

Its excuse that it thought its formulation was non-infringing is undermined by the 2007 District Court finding of infringement [see AstraZeneca AB v Mylan Labs Inc et al, 490 F Supp 2d 381 (2007)] which was subsequently upheld on appeal in 2008 in AstraZeneca AB v Apotex Corp, 536 F 3d 1361 (Fed Cir). Apotex’s stability testing thus commenced long after it knew or ought to have known that Apo-Omeprazole was infringing. [ 29 ] Apotex’s failure to complete the testing of its alternative formulations and to instead rely on extrapolations from its experts in this case is an unacceptable approach.

A recognition of this strategy would potentially reward Apotex for its delay by excluding from consideration finished stability test results – data that may well have established that the alternative formulations would not work. It cannot be to Apotex’s advantage that its delay in the initiation of obvious testing avoids the potential for failed results. What Apotex is

asking is that the Court predict a result that it could have but failed to establish. On the evidentiary record before me, I am not preparedto draw the inferences Apotex is seeking. [30] AstraZeneca relies heavily on the principle that in the assessment of the but for world of NIAs the Court must look at whattook place in the real world including the behaviour and state of mind of the infringer.

Apotex does not deny this as a point of principlebut argues for its reduced significance. [31] Initially I did have reservations about the idea that the availability of a NIA can be informed, in part, by the willfulness of theinfringement. But as I understand the decision of the Federal Court of Appeal in Lovastatin FCA, the idea is no more than this: wherean infringer brazenly infringes a valid patent, or substantially courts the risk of doing so, an inference may arise that no viable substitutewas available.

If it were otherwise the rational choice would always be to employ the NIA and not the infringing product. [32] It seems to me that what Apotex knew at the time and what it did in response to that knowledge in the real world are importantconsiderations in the assessment of the hypothetical availability of its after-the-fact NIAs.

The suggestion today that the developmentand commercial exploitation of the asserted NIAs would have been simple, cost-effective and speedy is substantially belied by historicalfact. [33] It is worth noting that it took Apotex many years to develop and obtain regulatory approval for Apo-Omeprazole – a productthat Dr. Bernard Sherman apparently thought at the time would not infringe the 693 Patent or the United States 505 Patent.

This factbelies the argument that any of the NIAs would have enjoyed an easier route to success if they were developed from scratch and withoutthe benefit of the development of Apo-Omeprazole. Indeed, as I found in the liability phase, omeprazole is not an easy molecule toformulate. [34] Dr. Sherman’s evidence that a work-around NIA solution was a straight forward task is also belied by the experience ofproducing the now-asserted NIA formulations. Initially Dr. Sherman thought the solution lay in the removal of the alkaline reactingcompound [ARC] from the infringing formulation. Indeed, that was Dr.

Sherman’s evidence in the United States litigation. Howeverwhen that approach was adopted for this proceeding, it failed [see Exhibit APO 130, Chow Report #1 at paras 84-85]. [35] It is also noteworthy that none of the first 14 NIA formulations produced by Apotex were pursued in this litigation. Thissupports an inference that each of them failed. Of those formulations that did go forward to further testing, a number clearly failed tomeet the necessary stability or bioequivalency requirements. Of those formulations that Apotex continues to assert, several weredeveloped later in the selection process.

All of this undermines Apotex’s argument that numerous viable NIA options would have beenimmediately obvious to a skilled formulator like Dr. Sherman. [36] Dr. Sherman’s excuse for not exploring his NIA options during the infringing period was that he had no reason to think Apo-Omeprazole was infringing. This evidence does not stand up to scrutiny. Indeed, as discussed above, it either was or should have beenincreasingly obvious to Dr. Sherman that Apo-Omeprazole was likely an infringing product. Notwithstanding what Apotex knew orought to have known, it persisted with its use of Apo-Omeprazole.

This continued infringing conduct was unreasonably stubborn ordogmatic, if not wilfully blind to the consequences, and it contradicts Dr. Sherman’s trial testimony that, had he known, Apotex wouldhave immediately searched for other options. [37] It is also of some significance that despite increasing evidence of infringement Apotex chose not to examine Apo-Omeprazoleto determine if it incorporated an infringing subcoat. All of this conduct undermines Dr. Sherman’s evidence that if he had only knownApo-Omeprazole was infringing, he could easily have developed or purchased a NIA.

The more sustainable inference is that Apotex wasprepared to run with Apo-Omeprazole whatever the likely consequences and it is doubtful it would ever have pursued a NIA option. That is particularly the case for pursuing a third-party NIA. Dr. Sherman made it very clear that such an approach would not have beenconsidered unless and until he had exhausted his in-house options. [38] In assessing Dr.

Sherman’s evidence about what Apotex would have done in the hypothetical world it is necessary to considerwhat he knew in the real world and what Apotex did or did not do with that knowledge. [39] At least as early as 2000, Apotex knew that AstraZeneca was asserting an infringement allegation based on an in situ formedsubcoat in connection with another generic omeprazole formulation.

In AB Hassle et al v Canada et al, (FC), 10 CPR (4th) 38, 102 ACWS (3d) 185 (FC), aff’d 2002 FCA 147, 18 CPR (4th) 558, Justice Daniele Tremblay-Lamer made a finding ofinfringement on that basis. [40] In 2000, AstraZeneca made the same allegation against Apotex and other generics in the infringement action in the UnitedStates. In the first wave of that litigation, concluded in 2002, the District Court found infringement on the part of a different defendantfor an in situ formed subcoat.

At the conclusion of the second wave of cases in 2007, discussed above, the same finding was madeagainst Apotex. [41] In 2003, the Federal Court of Appeal construed the 693 Patent claims to cover an in situ formed subcoat and rejected Apotex’s arguments to the contrary: see AB Hassle v Apotex Inc, 2003 FCA 409, 29 CPR (4th) 23. [42] In 2004, AstraZeneca commenced the first of these proceedings in Canada against Apotex for damages, alleging again thatApo-Omeprazole infringed the 693 Patent on the basis of an in situ formed subcoat. [43] Notwithstanding the above history, Apotex took no steps to pursue a NIA formulation or even to test whether Apo-Omeprazole capsules contained an infringing subcoat layer. [44] On March 16, 2015, I rendered a Judgment in these proceedings finding Apo-Omeprazole to be infringing of the 693 patentbecause it incorporated a sub-coat layer formed in situ. [45] Having regard to the above background and to the fact that no effort was made by Apotex until late 2015 to develop any NIA

formulations nor at any time or to pursue a third-party formulation, considerable caution is warranted. [ 46 ] Apotex had no readily available NIA options at any time during the infringing period and it had no back-up plan to develop or purchase one. Instead it ran with Apo-Omeprazole to the end. Even now Apotex produced late, incomplete and inconclusive stability and bioequivalency data suggesting that it did not, and to this day does not, have a viable in-house NIA option.

Notwithstanding these concerns, I will proceed with an assessment of Apotex’s evidence concerning its asserted NIAs to determine whether they were available and true non-infringing substitutes for Apo-Omeprazole. [ 47 ] In that regard I can readily dispose of two issues raised by AstraZeneca: (

a) whether Apotex had the capacity to commercialize one of its asserted NIA formulations (the could-have question); and (

b) whether Apotex has failed to prove that each of its asserted in-house NIAs is non-infringing. [ 48 ] While I accept that there would be manufacturing challenges for Apotex during scale-up to commercial NIA production, I believe that, with the exception of NIA formulation MR8620E1, these could be overcome by a successful and sophisticated producer like Apotex. Enteric coatings have been commercially used for many years and Apotex had considerable experience in successfully applying them to its formulations. Dr.

Davies identified a number of production obstacles that Apotex may not have fully resolved in its small- scale batches. However, I am left with the impression that Apotex could and would have resolved most of these issues without the inordinate expenditure of time or money and without compromising the dissolution profile of the enteric coatings used in the NIA formulations. [ 49 ] I exclude from this finding formulation MR8620E1. That formulation was designed to avoid an in situ subcoat by reducing the water content of the MACP enteric coating dispersion.

This change reduced the potential for a reaction at the enteric coating/core interface. [ 50 ] I am not satisfied that MR8620E1 could have been commercially developed because, as Dr. Davies explained, it failed to meet the MACP manufacturer’s specification for solids content and this repeatedly caused nozzle blockages [see Exhibit AZ 137 at paras 136- 37] . In the absence of persuasive evidence proving that this production problem could be overcome at commercial production levels, I am not convinced that it would have worked.

Indeed, if it was as obvious a work-around as Apotex now suggests, one is left to wonder why it was not attempted until well into Apotex’s NIA development and why larger scale enteric coated batches were either not attempted or were left undocumented. [ 51 ] I am also satisfied on the evidence provided by the Apotex witnesses that it had ample in-house capacity to produce the remaining NIA formulations, sufficient to match its infringing sales. [ 52 ] AstraZeneca contends that Apotex has failed to prove that its proposed NIAs would not infringe the 693 Patent.

Although AstraZeneca has stipulated that Apotex’s proposed NIAs produced at batch scale do not infringe [see Exhibit APO 69], it does not concede the same point for any of the NIAs if produced at a commercial scale. I do not accept this argument because it lacks direct evidentiary support. [ 53 ] If the NIAs are non-infringing at batch scale, one would expect them to remain non-infringing on commercial scale-up. That expectation might be rebuttable with cogent evidence that a production scale-up would be likely to give rise to an infringing characteristic (e.g. an in situ subcoat layer).

No evidence directly on point was before me and I find that the asserted NIA formulations at commercial scale would not infringe the 693 Patent. A. Are the Proposed NIAs Bioequivalent to LOSEC? [ 54 ] Dr. Mario González is an expert in pharmacokinetics, clinical pharmacology and biopharmaceutics [including the development and application of in vitro – in vivo correlations and relationship in predicting bioequivalence of formulations].

He provided expert opinion evidence on behalf of Apotex as to whether, in the absence of in vivo data, one could reasonably predict that any of the asserted NIAs would be likely to be bioequivalent to LOSEC and, if so, how the prediction could be made. [ 55 ] After advising Apotex that, in certain circumstances, such predictions could be made, Apotex gave Dr. González its pharmacokinetic/statistical clinical data comparing the bioequivalency of Apo-Omeprazole and LOSEC along with its in vitro dissolution data and testing protocol. From that information Dr.

González was asked to provide an opinion “as to whether any of the [NIAs] would be expected to be bioequivalent to Losec” . [ 56 ] Dr. González’s first report [see Exhibit APO 41] acknowledges that assessing bioequivalency for regulatory purposes between two pharmaceutical compounds is carried out with randomized, cross-over human clinical testing where blood plasma concentrations are measured over time and compared. Acceptable clinical studies would require at least 12 human subjects but more typically between 18 to 24 subjects “to gain meaningful data” .

The tested population needs to be large enough such that the data are not unduly thrown-off by intra- and inter-subject variability. Health Canada will accept two formulations as bioequivalent if the comparative data meet minimum statistical standards [see para 32]. [ 57 ] Dr.

González’s first report states, at para 33: where it is undesirable or impractical to conduct a comparative bioavailability study to determine whether two formulations are bioequivalent, it is, in certain circumstances, possible to use alternative methods, such as an in vitro/in vivo correlation (“IVIVC”) or an in vitro/in vivo relationship (“IVIVR”), to provide a reasonable prediction that two formulations will be bioequivalent. [ 58 ] In this case an IVIVC could not be carried out and Dr.

González was limited to using the less robust IVIVR method – a technique that he said “can be of great value during formulation development” [see para 38].

[ 59 ] Dr. González began his IVIVR bioequivalency work-up by plotting the bioequivalency data for Apo-Omeprazole and LOSEC. He observed them to be bioequivalent in the fasted state. According to Dr. González any NIA formulation with an in vitro dissolution profile that fell between those of Apo-Omeprazole and LOSEC “would be expected to be bioequivalent to these formulations” [see para 52]. Such a prediction could not be made with the same degree of confidence for a NIA with an out-of-range profile. Of the fifteen NIA formulations Dr.

González examined, eight fell within the Apo-Omeprazole and LOSEC dissolution profiles. The others were said to be “less likely to be bioequivalent to Losec® and Apo-Omeprazole based on dissolution data” [see para 60]. Dr. González then carried out a comparison using the f 2 metric for similarity for the remaining eight NIA formulations and observed them to be similar to either LOSEC or Apo-Omeprazole. He concluded that all eight “would be expected to be bioequivalent to Losec® and Apo-Omeprazole” [see para 65]. [ 60 ] AstraZeneca countered Dr. González’s evidence with the opinion of Dr. David Taft.

He was qualified as an expert in pharmaceutical sciences, including pharmacokinetics. [ 61 ] Dr. Taft was asked to advise if IVIVR was an accepted and reliable technique for predicting bioequivalency generally and, more specifically, for predicting the bioequivalency of the Apotex NIAs to LOSEC or Apo-Omeprazole based on the available data. [ 62 ] Dr. Taft defined bioequivalency and its regulatory significance in the following way in his responding report [Exhibit AZ 160 ]: 37.

Bioequivalence has been defined as, “the absence of a significant difference in the rate and extent to which the active ingredient or active moiety in pharmaceutical equivalents or pharmaceutical alternatives becomes available at the site of drug action when administered at the same molar dose under similar conditions in an appropriately designed study.” (emphasis added) 38. Government agencies that regulate marketed drugs, like the FDA and HC, set standards for how “absence of a significant difference in the rate and extent” are to be determined.

C max and AUC are the parameters used to measure the rate and extent of absorption. [Emphasis in original.] [Footnotes omitted.] [ 63 ] Dr. Taft expressed the opinion that the bioequivalency of LOSEC and Apo-Omeprazole was not established by the data relied upon by Dr. González. He was particularly concerned by the exclusion of data from one of the test subjects (OM75). Had those data been included, the results of the Apotex study would not have satisfied either the Health Canada or the United States Food and Drug Administration [FDA] requirements for bioequivalency. Notwithstanding Dr.

Taft’s concern, the fact remains that regulatory approval was obtained for Apo-Omeprazole and presumably the exclusion of OM75 was accepted by the regulators. On the evidence presented, including Dr. Taft’s evidence at Transcript p 4087 and Exhibit AZ 78 , I am not prepared to look behind that decision to reassess its scientific merit. [ 64 ] Dr. Taft was highly critical of Dr. González’s analysis for other reasons. He challenged Dr.

González’s use of mean dissolution and absorption data obtained from the Apotex Apo-Omeprazole biostudy based on the wide variability of the actual plasma concentrations for each of the test subjects as compared to the mean. Figure 6 of his report, at p 31, nicely depicts this variability. According to Dr. Taft this variability weakens the IVIVR model as a foundation for comparing LOSEC and Apo-Omeprazole to the NIAs. [ 65 ] According to Dr.

Taft, the problem of subject-to-subject variability in the Apotex biostudy data for Apo-Omeprazole is compounded by the variability in the in vitro dissolution data for the NIAs used by Dr. González. Because all of the dissolution data for the NIAs exceeded the accepted margin of 10% for the IVIVC method, the data, he thought, were “too variable to use for reliably predicting bioequivalence” using the less robust IVIVR method [see para 116]. [ 66 ] Dr. Taft was also critical of Dr. González’s use of IVIVR as a predictive tool in proof of bioequivalence.

He pointed out that IVIVR is only useful as a tool for guiding formulation development. According to Dr. Taft, there is nothing in the literature supporting its reliability for predicting bioequivalence or indicating that it can be used as a substitute for an IVIVC. Furthermore, Dr. González’s IVIVR method did not meet the minimum standards required of an IVIVC. Dr. Taft concluded this part of his report in the following way: 126. More fundamentally, Dr. González’s “IVIVR” technique is incompatible with the underlying principles for an IVIVC.

An IVIVC is based on a formulation consistently demonstrating a relationship between in vitro dissolution and in vivo absorption. Dr. González’s “IVIVR” is based on an entirely different proposition. In particular, Dr. González’s “IVIVR” technique contends that where different formulations with different in vitro dissolution profiles ( i.e., Apo-Omeprazole and LOSEC) have similar in vivo absorption profiles, any test formulation ( i.e., an AF) that has an in vitro dissolution profile lying between the other two will also have the same in vivo absorption profile (and hence be bioequivalent).

That is, as he seems to acknowledge, Dr. González’s “IVIVR” technique is not based on a correlation between in vitro dissolution and in vivo absorption; but rather on the assumption that such a correlation is irrelevant. However, if dissolution is irrelevant to absorption, then it is not predictive of absorption. 127. Based on the foregoing, in my view the scientific community has not accepted and would be unwilling to accept Dr. González’s “IVIVR” technique for establishing bioequivalence of the [NIAs] to Apo-Omeprazole or LOSEC. [Footnotes omitted.] [ 67 ] Dr. Taft had several other concerns about Dr.

González’s use of IVIVR to predict the bioequivalence of the NIAs to LOSEC and Apo-Omeprazole. These included his failure to estimate the error rate, the lack of validation as required by the FDA for IVIVCs and the lack of any data showing bioequivalence in the fed state. [ 68 ] Dr. González’s reply report generally addressed the points advanced by Dr. Taft [see Exhibit APO 43]. In particular, he picked up on Dr. Taft’s point that LOSEC and Apo-Omeprazole exhibited different dissolution rates for the first 40 minutes. When

Dr. González adjusted for the five minute time lag in dissolution between the two formulations, the profiles were similar. [ 69 ] Dr. González commented on Dr. Taft’s view that some in vivo absorption data for the NIAs was needed for a valid IVIVR. Dr. Taft’s point was weakly countered with the following statement, at para 45: “[h]owever, if there were in vivo data available for these formulations, this would remove the need for an IVIVR” . Dr.

González’s substantive response was limited to the observation that, as “minor” variations to Apo-Omeprazole, the NIAs would be assumed to have the same linear relationship between dissolution and absorption. [ 70 ] Dr. González accepted that “there may be some larger than expected variance” in the dissolution data for the NIA formulations. This he attributed to the fact that the NIAs came from pilot scale batches “as opposed to optimized formulations” but “this does not mean that the data are unsuitable for use in an IVIVR” [see para 57]. [ 71 ] Dr. González answered Dr.

Taft’s concern about the IVIVR technique not being peer-reviewed or generally accepted in the scientific community in the following way: 58. At paragraphs 118 to 127 of his report, Dr. Taft provides the opinion that the IVIVR technique that I used in my prior report was not peer-reviewed or generally accepted.

However, I know from my personal experience in the pharmaceutical industry that, while IVIVRs may not be the subject of large numbers of journal articles, IVIVRs are routinely used within the pharmaceutical industry, and have been used since the early 1980s during formulation development to direct the modification of formulations. An IVIVR allows a formulator to decide if a modified formulation is likely to be successful in a bioavailability study, and thus whether to test in a human population. 59.

During formulation development, it would be rare for an IVIVC to be available for an early formulation, and it is not practical to conduct an in vivo bioavailability study for each formulation prepared during the development of a product. Rather, in vivo data will only be obtained for a select number of formulations, and an IVIVR will be used to identify a bio-relevant dissolution test to identify additional formulations that would be expected to be bioequivalent to, or to have a better drug absorption profile, for example, a longer t max than, a reference formulation. [ 72 ] Dr.

González’s reply also took issue with Dr. Taft’s application to the IVIVR model of a 10% prediction error threshold used in IVIVCs. His less than compelling response was that the IVIVR model would be expected to throw off a higher error rate than an IVIVC such that “a prediction error of 10-20% [would] be good for an IVIVR” . He summed up the point in the following way: 73. For these reasons, it is not at all surprising that an estimate of prediction errors with an IVIVR would differ from those typically seen with a more rigorous and robust IVIVC.

The intent of an IVIVR is to provide assurance of the likely performance or bioequivalence of a test formulation when compared to a reference formulation without requiring the need for larger amounts of in vivo data, and also allowing for a greater range of differences between formulations. [ 73 ] Dr.

González responded to the absence of NIA bioequivalence data in the fed state by pointing out that because Apo- Omeprazole is bioequivalent to LOSEC in the fed state, and because the NIA formulations were similar to LOSEC and Apo-Omeprazole, the NIAs would also be expected to be bioequivalent in the fed state. [ 74 ] Under cross-examination Dr. González accepted that the best means of establishing bioequivalence is through a well-designed clinical study [Transcript p 1235]. Dr. González also stated that one does not “run a bio study with some little pilot formulation.

You want to make sure that you have a really good formulation before you get into a bio study” [Transcript p 1241]. [ 75 ] Dr. González was questioned closely on his view that, because the NIA formulations were similar to LOSEC and Apo- Omeprazole, they would be expected to behave in the same way. He conceded that he was not a drug formulator and he was clearly out of his depth with respect to this issue, as can be seen from the following exchange at Transcript p 1307: Q.

Moving away from the compression forces to the ingredients: In your view, these formulations are similar because they have the same ingredients? A. Yes. Omeprazole, mannitol, magnesium hydroxide and povidone, yes. Q. Let's look at the first experimental formulation, 15-1214B. That doesn't have magnesium hydroxide. It doesn't have povidone? A. That is true. Q. Let's look at the third one. A. Qualitatively, that one is different. Q. The third one, that is different too. Fifteen -- I think it should read "12"? A. Povidone is missing. Q. Next one, the ingredient is missing? A. "Magnesium hydroxide" is missing. Right.

Q. Next one, an ingredient is missing? A. That is true. Q. Next one, as well, 6108-288C. So many of these formulations do not have the same ingredients. Correct?

A. That is true, but I am not using any of these formulations to arrive at the IVIVR. I am using the dissolution profiles from these formulations to see how well they fall within the other two dissolution profiles. Q. You just said, Dr. González, that you were relying on the ingredients being the same to assume that they would have the same relationship. A. Yes, I did say that. Q. So you can't make that assumption? A. Not for all of these formulations, no. On the next page, the other four all do have the same ingredients and, in fact, even the same percentage, so I don't know how -- [ 76 ] Dr.

González was asked about the differences between IVIVCs and IVIVRs. He referred to the IVIVC as “a predictive mathematical model” and conceded that the IVIVR “doesn’t have such clout” [Transcript p 1313]. The IVIVR is less reliable in predicting bioequivalence [Transcript p 1315]. Unlike the quantitative information produced by an IVIVC, the IVIVR shows only a “qualitative” relationship [Transcript pp 1314, 1320-21]. Establishing an IVIVR would not be accepted by a regulator as evidence of bioequivalence because it does not produce the required quantitative data [Transcript p 1321].

Rather, IVIVRs are typically used as a directional screening tools to guide formulation development. They are not a replacement for biostudies [Transcript pp 1355-56]. [ 77 ] The predictive value of an IVIVR was further explained by Dr. González in the following exchange at Transcript p 1359: Q. But you didn't have a biostudy for a single one of the experimental batches? A. No. Q. Do you agree with Devane and Butler that in vivo data for the experimental batches would permit a reality check on your IVIVR model? A.

Well, I would agree that at some point in time, Apotex would eventually run a biostudy on one or two of those formulations. But that it needs to be done for them to select a formulation to pursue, I don't think you have to run a biostudy at that stage. [ 78 ] In another exchange Dr. González refused to accept a characterization by Devane and Butler that IVIVRs have “limited value” . He answered by saying that IVIVRs produce “some predictive value, that your dissolution shows a relationship to in vivo absorption” [Transcript p 1361]. Even in the absence of biostudy data for the NIA formulations, Dr.

González expressed the opinion that “there is a good chance that they are bioequivalent” [Transcript p 1370] and “[s]ome of them appear to be bioequivalent on the basis of IVIVR or should be bioequivalent on the basis – of the fact that their dissolution falls within the dissolution of the product that we are testing” [Transcript p 1371]. And further at Transcript pp 1372 and 1378, he stated: We have a set of two dissolution profiles. Now I have a bunch of formulations that have a similar mechanism of release or at least what I assume is a similar mechanism of release.

There is a good chance those are going to fall -- that the ones to fall within the two dissolution profiles will have a good chance for bioequivalence. That is it. … That is exactly what I did actually. I took two extreme batches that were bioequivalent, and now I am hoping that these other ones will fall in there, that they are going to be clinically relevant. [Emphasis added.] [ 79 ] Generally speaking Dr. Taft was a better witness than Dr. González and he provided more reliable evidence on the central issue of the validity of Dr. González’s IVIVR bioequivalency analysis.

I would add to this that many of the central aspects of Dr. Taft’s written opinion and his evidence in Chief were left unchallenged on cross-examination. Furthermore, some of the issues addressed in Dr. Taft’s responding report were not answered in Dr. González’s reply. The failure to fully engage with Dr. Taft’s opinions supports an inference that his unchallenged evidence was unimpeachable. [ 80 ] By way of example, Dr. Taft was concerned about the high variability in the Apotex biostudy data for Apo-Omeprazole (what he called a “wide dispersion” of data around the mean result). According to Dr.

Taft, data variability is an important factor in the establishment of a valid IVIVR. He addressed this point at Transcript p 4033: Q. Going back in your report, paragraph 106, given those features, what does that tell you about the reliability of using that data to try to generate an IVIVR? A. Again, as I have mentioned in paragraph 105, according to Cardot and Davit, in that situation where the mean curve does not reflect the individual behaviour, IVIVC is not recommended.

In that context and looking at what we just talked about of the relative data from the OMCP10 that seems to match what is being described by Cardot and Davit, it is my opinion that using an IVIVR based upon biostudy OMCP10 to predict bioequivalence is unreliable. JUSTICE BARNES: Does it matter that what Dr. González was looking at was not an IVIVC but rather than IVIVR? Does that change anything? And if so, how? THE WITNESS: In my opinion -- I believe this is the opinion the scientific community -- the only surrogate bioequivalence is a level A IVIVC. You have asked me the question. Essentially, what Dr.

González, in my opinion, is attempting to do is to use an IVIVR as if it was a level A IVIVC.

[ 81 ] Dr. Taft also spoke to the problem of variability of subject data and the corresponding potential for error in relying on the mean to support an IVIVR. According to Dr. Taft the variability of the data called into question the reliability of the mean as a measure of absorption or dissolution [see Transcript p 4036]. [ 82 ] On the issue of using IVIVR generally to predict bioequivalence, Dr. Taft testified as follows at Transcript p 4044: Q.

Turning to the next point, this is more broadly to the extent that that technique has been generally accepted in the scientific community for establishing bioequivalence. What are your views there? A. In the first case, for example, if it is not available in the peer-reviewed literature, you can certainly look towards regulatory agencies and their opinions and views through guidances and other things.

In the documents that I have reviewed not only in this case but in my day-to-day professional activities, I have never come across a guidance document that would support using an IVIVR to establish bioequivalence of alternate formulations. [ 83 ] It is of some significance that in answer to questions from me, Dr.

Taft attributed some value to the IVIVR technique but only as a rough screening tool [Transcript p 4045]: JUSTICE BARNES: Before you go there, to run this to ground a little bit, it strikes me from what I have heard so far that IVIVR is a "recognized technique," if I could put it that way, in the pharmacokinetic world. It has some value in some places. THE WITNESS: Yes, Your Honour, it does. JUSTICE BARNES: Yes, so it is recognized as providing some scientific value to some sort of an analysis. Where does it fit in the scheme of things then?

THE WITNESS: I believe that the Devane paper provided some background on the use of IVIVR in formulation development. Certainly formulators will attempt to rely upon in vitro data to establish or to move products forward in screening. It is not unlike what I do in my day-to-day activities, looking at in vitro pharmacokinetic parameters and relating them to in vivo. The issue really becomes, in the case of using an IVIVR, to establish bioequivalence.

That is, in my view, a totally different application that essentially what you are attempting to do is use the principals [sic] of a level A IVIVC to make that determination, and that is my point. [ 84 ] This evidence is not materially different from Dr. González’s ultimate trial testimony. [ 85 ] Dr. Taft went on to calculate the internal prediction error rate associated with Dr. González’s IVIVR techniques – a form of validation Dr. González failed to employ. Dr. Taft found a prediction error range between 18.7% and 36.6% – well beyond the validation threshold of 10 to 15%. This led Dr.

Taft to reasonably conclude that Dr. González’s IVIVR model, based on mean data, was not able to capture the predicted profiles for the test subjects [Transcript p 4052]. [ 86 ] Dr. Taft repeated his point that an IVIVR model cannot be used to predict bioequivalency in the fasted state let alone the fed state [Transcript pp 4052-53]. On this issue, I agree with Dr. Taft that a prediction of bioequivalency in the fed state cannot be extrapolated from data observed in the fasted state. Dr. González had no data to support his opinion.

That opinion rested only on an assumption that the NIA formulations were sufficiently similar to LOSEC and to Apo-Omeprazole that the ingestion of food would not make any difference to their bioequivalency. The firmness of that view wavered under cross-examination where Dr. González conceded the difficulty of predicting food-effects on formulation bio-availability. He ended with the equivocal statement that one “may be able to predict how things are going to work out” [Transcript p 1327]. [ 87 ] The far better evidence on this point came from Dr. Taft.

He described the problem of predicting bioequivalency in the absence of a biostudy in the fed state. The presence of food in the stomach, he said, can “influence the absorption of a drug or drug formulation, particularly for an enteric-coated product” [Transcript p 4000; see also Transcript pp 4052-53]. This evidence was not challenged under cross-examination. [ 88 ] Apotex’s failure or inability to conduct NIA biostudies in the fed state represents a large gap in its case for bioequivalency.

That is so because, if Health Canada required bioequivalency data for any of the NIAs, biostudies in the fed and fasted state would have been necessary just as they were for Apo-Omeprazole [see Exhibit AZ 158 at p 1 ]. [ 89 ] Dr. González failed to counter much of Dr. Taft’s evidence about the validity of Dr. González’s methods. I accept Dr. Taft’s concerns about the variability of the data employed by Dr. González and the corresponding potential for error. Dr. González should also have run a validity analysis of his own and I reject Apotex’s criticisms of Dr. Taft’s approach to that issue. I also agree with Dr.

Taft that the error range for Dr. González’s analysis exceeded acceptable levels. [ 90 ] The above-noted methodological problems are sufficient on their own to wholly undermine Dr. González’s prediction of NIA bioequivalency. But there is a more fundamental problem with Dr. González’s use of IVIVR as a predictive tool – a problem that Dr. González acknowledged up to a point. In his trial testimony Dr. González retreated somewhat from his firmer written opinion. [ 91 ] In his first report Dr.

González concluded that eight of the Apotex NIA formulations “would be expected to be bioequivalent to Losec® and Apo-Omeprazole” [Exhibit APO 42, para 65]. However under cross-examination, he conceded that IVIVR is a directional or screening methodology that only showed the selected NIAs to “have a good chance for bioequivalence” [Transcript p 1372]. Given the inherent limitations of IVIVR as described by Dr. Taft, I reject the suggestion that the IVIVR analysis carried out by Dr. González supports a balance of probabilities finding of bioequivalency.

Indeed, the technique in its present form has only a limited value in the area of pre-formulation selection or screening. It produces nothing of quantitative value and Apotex’s attempt to extend its reach is unjustified. In these circumstances Dr. González’s bioequivalency opinion does not rise above the level of speculation. It is

certainly a wholly inadequate proxy for the kind of data required for establishing NIA bioequivalency necessary for regulatory approval. Furthermore, it is an unreliable platform for drawing an inference of bioequivalency. In the result, I find that Apotex has failed to establish that the NIA formulations assessed by Dr. González are, or would be seen by a regulator to be, bioequivalent to either LOSEC or Apo-Omeprazole. II.

Could Apotex Have Conducted Human Clinical Trials on its Proposed NIAs to Prove Bioequivalency? [ 92 ] Apotex contends that it could not conduct human clinical trials for its proposed NIAs on ethical grounds. According to this view, because the NIAs were only developed for litigation and no potential health benefits would result, the risks to human subjects would always outweigh the parochial objectives of the work. AstraZeneca disagrees and argues that human clinical trials could have been conducted.

AstraZeneca also says that Apotex should, at a minimum, have sought permission from its research ethics board to conduct these studies. By failing to even ask, Apotex failed to produce definitive evidence on the point and should not be the beneficiary of any lingering doubt. Whatever the outcome, the issue, AstraZeneca says, would be conclusively resolved, rendering any ex post facto ethics opinion on the point effectively moot. [ 93 ] Each party called an eminent ethicist in support of its position. Apotex relied on the evidence of Dr. Michael McDonald and AstraZeneca presented evidence from Dr. Charles Weijer.

Dr. McDonald testified that a properly qualified research ethics board would not have authorized human clinical studies in these circumstances. Dr. Weijer came to the opposite conclusion. [ 94 ] There is some attractiveness to AstraZeneca’s argument that opinion evidence on this issue ought to be excluded in the absence of a research ethics board ruling. I am not convinced, however, that AstraZeneca’s suggested approach would necessarily have resolved the matter – at least if the answer provided by the Apotex research ethics board was in the negative.

In that event it would remain open to AstraZeneca to argue that the decision was self-serving or wrong. In short, a fact-based negative ruling was unlikely to be a complete and final answer to the question of the propriety of conducting human trials. That is not to say, however, that the approach asserted by AstraZeneca would not have been helpful in determining whether the administration of the Apotex NIA formulations to humans would be unethical and would never be authorized. An actual decision on-point would certainly have provided relevant and likely probative evidence. [ 95 ] I do not accept Dr.

McDonald’s point that Apotex would have been ethically constrained from even asking its research ethics board if a biostudy could be carried out. Dr. McDonald testified that the circumstances of this case were novel and he posed the question: “how does one deal with a novel case?” [Transcript p 2517]. In the absence of clear guidelines and where there are opposing views about the standards of ethical review to be applied, it cannot be the case that Apotex is entitled to make a pre-emptive and potentially self-serving decision not to approach its research ethics board for direction.

Apotex could and should have sought direction from its research ethics board in this case. [ 96 ] It does, however, seem doubtful that Apotex would have been authorized to conduct human biostudies in connection with its hypothetical NIAs, given that the sole purpose of the experiments would be to advance Apotex’s litigation interests. [ 97 ] For this, I accept Dr. McDonald’s views over those expressed by Dr. Weijer. In particular, I accept Dr.

McDonald’s evidence concerning the risk-benefit ratio and the ethical requirement that the importance of a research objective outweigh the risks faced by the research subjects. In the case of the Apotex NIAs the risks may have been slight but they were not unworthy of consideration and concern. Where the only benefit of the trials would be to further Apotex’s business interests (in contrast to a public good), it is doubtful that ethical approval could have been obtained. [ 98 ] I also disagree with Dr.

Weijer that a biostudy involving the human ingestion of an unapproved drug carrying anticipated risks and side-effects and done solely for private commercial or litigation purposes ought to the subject of less rigorous standards or some form of diminished risk-benefit analysis. His idea that there is an underlying public interest in the outcome of private litigation has no appeal. Courts of law decide cases on the basis of available evidence and, in many cases, the evidentiary record is incomplete.

It is inconceivable that any Court could or would ever order a litigant to conduct human testing to answer a question relevant to the outcome of a case. At most, an adverse inference can be drawn where a party fails to advance evidence that is potentially available to it. [ 99 ] The fact that Apotex probably could not ethically conduct human bioequivalency studies to prove the efficacy of its unapproved NIAs does not, however, assist it in advancing its substantive case. At most this barrier to human testing prevents the Court from drawing an adverse inference.

Human bioequivalency studies are an important means to prove the viability of a NIA and are often required for obtaining regulatory approval. The indisputable fact remains that those tests were not done, leaving a significant gap in the evidence as to whether any of Apotex’s NIAs could have been shown to be bioequivalent to LOSEC or Apo-Omeprazole and approved for sale in Canada or the United States. A.

Are the Proposed NIAs Sufficiently Stable? [ 100 ] Apotex contends that its testing data are sufficient to meet its burden of establishing that the asserted NIAs were commercially viable substitutes for Apo-Omeprazole. One of the key requirements for proving that viability concerns the stability of those formulations (i.e., did they have an acceptable shelf-life both from a regulatory and commercial standpoint?). [ 101 ] Apotex’s stability case was based on data it obtained from in-house stability testing commenced in late 2015, as interpreted by Dr. Kwok Chow. Dr.

Chow was accepted as an expert in pharmaceutical product development and drug delivery systems including the design, execution and management of formulation screening, including with respect to the physical and chemical properties of drug substances and excipients, including, specifically, stability thereof.

[ 102 ] Dr. Chow was asked by Apotex to review its in-house stability test data for fifteen NIA formulations to determine if they would be expected to have sufficient stability to be useful as pharmaceutical products. [ 103 ] Dr. Chow confirmed in his first report dated September 16, 2016 [Exhibit APO 130] that in-house stability testing of the sort conducted by Apotex is usually done under the research conditions recommended by the appropriate regulatory bodies and/or the International Council for Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use [ICH] guidelines.

Those guidelines set out the minimum data requirements for a new drug submission under long-term (12 months), intermediate (six months) and accelerated (six months) storage with varying temperature and humidity conditions. Throughout the assessment period the formulations are tested for assay values and impurity levels [see Chow Report #1, Exhibit 1]. [ 104 ] Dr. Chow's first report was based on stability data for three different formulation groups (A, B and C). Group A formulations (four formulations) had data covering at least eight weeks. Group B (nine formulations) had some four-week data.

Group C (two formulations) had only the initial assay data. [ 105 ] Dr. Chow reviewed Apotex's testing protocols and concluded they met ICH guidelines. [ 106 ] Notwithstanding the preliminary nature of the data, Dr. Chow professed in his first report to be able to predict the stability outcomes for some of the formulations. According to Dr. Chow two of the Group A formulations "are expected to meet the acceptance criteria at the end of the stability program" for total impurities [see para 86]. For one of those formulations, he predicted a target shelf- life of two years.

For the others, the data was insufficient to support a definitive two-year shelf-life. [ 107 ] For four of the Group B formulations, Dr. Chow expressed the view that they "may well meet acceptance criteria at the end of the stability program and achieve a target shelf-life of two years" [see para 115(a)]. For the other four Group B formulations, Dr. Chow expressed uncertainty related to the interim impurity data obtained. The last Group B formulation failed to meet the required dissolution criteria. [ 108 ] According to Dr. Chow both of the Group C formulations "meet the specification" [see para 124].

Because of the compositional similarities to Apo-Omeprazole, Dr. Chow expressed the view that Group C formulation MR8620E1 "would be expected to have sufficient stability since Apo-omeprazole is an approved product" [see para 125]. [ 109 ] Dr. Chow concluded his first report by stating that all of the NIA formulations would be readily scalable to commercial quantities. [ 110 ] Dr. Chow updated his stability opinion in his report of January 6, 2017 by reviewing the stability data developed since his first report [Exhibit APO 131]. [ 111 ] For the two remaining Group A formulations, Dr.

Chow dismissed an assay anomaly for one (despite not knowing the cause) and higher than expected impurity levels at 24 weeks under accelerated conditions for both. However, he did reduce his shelf-life expectation for both formulations to 18 months.

He concluded that there was "a high probability that both formulations will meet acceptance criteria for at least an 18-month shelf-life" [see para 30]. [ 112 ] Of the Group B formulations, two were predicted to meet acceptance criteria for a two-year shelf-life [see para 47(a)], three had higher impurity levels and a predicted shelf-life of 18 months [see para 47(b)] and three did not meet specification criteria under accelerated conditions; nevertheless, one of those was predicted by Dr.

Chow to have "a high probability" of an 18-month shelf-life while the other two were "less likely to meet specification criteria" [see para 47(c)]. [ 113 ] Dr. Chow reviewed the 16-week data for the Group C formulations. He dismissed an assay anomaly for one as being likely caused by poor sample preparation. He concluded that both were expected to continue to meet the requirements for drug release and "there is a high probability that both formulations will meet acceptance criteria for a 24-month shelf-life" [see para 60]. [ 114 ] AstraZeneca responded to Dr. Chow's opinions through Dr. Martin Davies. Dr.

Davies had testified in the liability phase of this case and he was qualified to testify as an expert in physical chemistry, pharmaceutical stability, drug delivery and biomedical surface chemistry and in the development, testing and analysis and characterization of pharmaceutical formulations. [ 115 ] Dr. Davies expressed the view that Dr. Chow's stability predictions were unsound because the supporting data were incomplete and, with one exception, the small-scale test batches were unsuitable to predict the stability of large scale commercially-produced products. Dr.

Davies identified a number of problems in the scale-up of production that might be barriers to success and he identified testing procedures that failed to meet the applicable stability guidelines. These included the failure to test more than one batch (to reduce the effect of batch variability) and the general failure to use pilot scale test batches (one-tenth commercial scale). Only one formulation met the pilot scale standard. [ 116 ] Dr. Davies pointed out that Dr.

Chow's first report was based on incomplete stability testing and, without completed testing, "there is insufficient data to reliably predict shelf-life" [Exhibit AZ 137, para 79] . He was particularly critical of Dr. Chow's extrapolation for Group C formulations for which only early data existed. He also criticized Dr. Chow for assuming linear degradation kinetics in the absence of empirical support and he challenged Dr. Chow's treatment of data anomalies. [ 117 ] Dr. Chow's reply report addressed Dr. Davies' comments about the sufficiency of the incomplete stability data in the following way: 62.

Many of the differences in the opinions of Dr. Davies and me appear to come down to the different perspectives from which we have looked at the stability studies conducted by Apotex. While I have looked at the stability studies to assess whether or not the formulations being studied would likely exhibit sufficient stability to be useful as a drug product, Dr. Davies appears to have focused on whether the studies, standing alone, could be presented to a regulatory agency, such as Health Canada, as part of a drug submission. I do not disagree with Dr.

Davies that the Apotex studies in their present form would not be complete for filing with a regulatory authority as

part of a drug submission. However, this does not mean that the studies fail to show whether the formulations are likely to have sufficient stability for use as a drug product. In general, by focusing on regulatory filings, Dr. Davies has failed to consider the realities of formulation development, and how stability studies are conducted and used in the pharmaceutical industry. 63. In light of Dr. Davies’ opinions, it should also be understood that, as part of the formulation development process, formulations are initially prepared on a laboratory or pilot scale.

Based on the test data obtained for these formulations, which will include abbreviated (as opposed to full-length) stability studies, formulators make informed predictions as to whether formulations are expected to possess the properties that are desired for the final dosage form. Scale-up of a formulation will only be conducted for formulations that are expected to be successful based on this initial testing of material prepared on a small-scale.

For formulations that show sufficient stability in the lab or pilot scale, the formulator’s expectation is that the formulation will also be stable when made on a larger scale using the same manufacturing steps. 64. Based on my experience in the pharmaceutical industry, I cannot recall an instance where the volume of information that has been collected for the Apotex formulations would not be considered to be suitable for deciding which formulation or formulations would be expected to be successfully scaled-up. [Footnotes omitted.] [ 118 ] Dr. Chow accordingly accepted Dr.

Davies’ point that the standards and methods employed by Apotex would not meet regulatory requirements. He said that, from a product development perspective, multiple batch testing was not required. He made the same point concerning the small lab tests batches. This, he said, was “a regulatory concern and does not relate to whether or not a formulation is stable, or expected to be stable” [see para 67]. [ 119 ] In response to Dr. Davies’ criticism that Apotex’s preliminary data were incomplete and insufficient to support a stability prediction, Dr.

Chow stated, at para 73: … At paragraph 60 of his Expert Report, Dr. Davies comments on the fact that, at the time of my September 6, 2016 Expert Report, the stability studies for the Group A, B and C formulations were not complete. As discussed above, this issue appears to be directed more to whether the studies could be submitted to a regulatory agency at that time, not whether the available information could be used to determine whether a formulation was likely to be stable throughout the duration of the stability program.

However, as noted above (see paragraph 63), it is standard practice to use preliminary data from stability studies in order to determine which batch or batches to scale up. … 74. In his Expert Report, Dr. Davies takes issue with my extrapolation of stability data, noting that the ICH Q1A stability guideline provides specific conditions for the extrapolation of stability. Again, Dr. Davies is confusing the

[…]