Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 1 of 20 PageID #: 1487
`
`IN THE UNITED STATES DISTRICT COURT
`FOR THE DISTRICT OF DELAWARE
`
`
`
`Civil Action No. 15-697-RGA
`
`
`HOSPIRA, INC.,
`
`
`
`
`
`AMNEAL PHARMACEUTICALS LLC,
`
`
`
`
`
`
`v.
`
`
`
`
`
`
`
`Plaintiff,
`
`Defendant.
`
`
`
`
`AMNEAL’S RESPONSIVE POST-TRIAL BRIEF ON NON-INFRINGEMENT
`
`
`
`
`
`
`
`Of Counsel:
`
`Steven A. Maddox (pro hac vice)
`Jeremy J. Edwards (pro hac vice)
`Matthew C. Ruedy (pro hac vice)
`Kaveh V. Saba (pro hac vice)
`Maddox Edwards PLLC
`1900 K Street N.W., Suite 725
`Washington, D.C. 20006
`(202) 830-0707
`smaddox@meiplaw.com
`jedwards@meiplaw.com
`mruedy@meiplaw.com
`ksaba@meiplaw.com
`
`
`
`
`
`
`Frederick L. Cottrell, III (#2555)
`Kelly E. Farnan (#4395)
`Christine D. Haynes (#4697)
`Richards, Layton & Finger, P.A.
`920 North King Street
`Wilmington, DE 19801
`(302) 651-7700
`cottrell@rlf.com
`farnan@rlf.com
`haynes@rlf.com
`
`Attorneys for Defendant-Counterclaim Plaintiff
`Amneal Pharmaceuticals LLC
`
`
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 2 of 20 PageID #: 1488
`
`TABLE OF CONTENTS
`
`I. 
`
`II. 
`
`Procedural Note ................................................................................................................................... 1 
`
`Introduction .......................................................................................................................................... 1 
`
`III. 
`
`Argument ............................................................................................................................................... 2 
`
`A. 
`
`B. 
`
`C. 
`
`Sunovion Does Not Apply Because Amneal’s ANDA Specification Does Not
`Directly Address the Claim Limitation. ............................................................................... 3 
`
`Hospira’s Attempts to Dismiss Amneal’s Biostatistician Expert Dr. Bloch
`Because He Is Not a POSA Should Be Rejected as a Matter of Law. ............................ 6 
`
`The Stability Data Analysis and Opinions Offered by Dr. Linhardt Did Not
`Prove Infringement, as Confirmed by Dr. Bloch. ............................................................. 7 
`
`1.
`
`2.
`
`3.
`
`4.
`
`As Explained by Dr. Bloch, Dr. Linhardt’s Analysis Did Not
`Account for Any Variability of the Data, Even Though He Testified
`That Such Variability Typically Exceeds 2%. ....................................................... 8
`
`Dr. Linhardt’s Point Estimates Were Premised on Unsupported and
`Illogical Statistical Models. ....................................................................................... 9
`
`None of the Five-Month Concentration Decrease Predictions for
`Amneal’s ANDA Products Was More Likely Than Not to Be Less
`Than the Claimed About 2%. ............................................................................... 11
`
`Hospira’s Remaining Arguments About Statements of Drug Stability
`to FDA, and a Year-Belated Discovery Dispute, Are Irrelevant to the
`Issue of Infringement. ............................................................................................ 14
`
`IV. 
`
`Conclusion ........................................................................................................................................... 15 
`
`
`
`
`
`
`
`ii
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 3 of 20 PageID #: 1489
`
`TABLE OF AUTHORITIES
`
`Abbott Labs. v. Sandoz, Inc.,
`566 F.3d 1282 (Fed. Cir. 2009) ................................................................................................... 15
`
`Page(s)
`
`Abbott Labs., v. Torpharm, Inc.,
`300 F.3d 1367 (Fed. Cir. 2002) ............................................................................................. 2, 3, 5
`
`Allure Energy, Inc. v. Nest Labs., Inc.,
`No. 9-13-CV-102, 2015 WL 11110634 (E.D. Tex. Apr. 13, 2015) ................................................ 6
`
`Broadcom Corp. v. Emulex Corp.,
`732 F.3d 1325 (Fed. Cir. 2013) ................................................................................................... 14
`
`Calhoun v. Yamaha Moto Corp., U.S.A.,
`350 F.3d 316 (3d Cir. 2003) ........................................................................................................ 11
`
`Ferring B.V. v. Watson Labs., Inc.-Fla.,
`764 F.3d 1382 (Fed. Cir. 2014) ................................................................................................. 3, 4
`
`Fresenius Kabi USA, LLC v. Dr. Reddy’s Labs., Ltd.,
`65 F. Supp. 3d 379 (D. Del. 2014) (Andrews, J.) ......................................................................... 10
`
`Gemtron Corp. v. Saint–Gobain Corp.,
`572 F.3d 1371 (Fed. Cir. 2009) ............................................................................................... 5, 12
`
`Glaxo, Inc. v. Novopharm, Ltd.,
`110 F.3d 1562 (Fed. Cir. 1997) ................................................................................................. 3, 4
`
`Johnston v. IVAC Corp.,
`885 F.2d 1574 (Fed. Cir. 1989) ..................................................................................................... 5
`
`Novartis AG v. Actavis Elizabeth LLC,
`Nos. 14–1487, 15–150, 15–151, 15–975, 2017 WL 1398347 (D. Del. Apr. 17, 2017) .................... 7
`
`SEB S.A. v. Montgomery Ward & Co., Inc.,
`594 F.3d 1360 (Fed. Cir. 2010) ..................................................................................................... 6
`
`Sundance, Inc. v. DeMonte Fabricating Ltd.,
`550 F.3d 1356 (Fed. Cir. 2008) ..................................................................................................... 6
`
`Sunovion Pharm., Inc. v. Teva Pharm. USA, Inc.,
`731 F.3d 1271 (Fed. Cir. 2013) ......................................................................................... 2, 3, 4, 5
`
`The Medicines Co. v. Mylan, Inc.,
`853 F.3d 1296 (Fed. Cir. 2017) ..................................................................................................... 5
`
`iii
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 4 of 20 PageID #: 1490
`
`UltimatePointer, L.L.C. v. Nintendo Co.,
`816 F.3d 816 (Fed. Cir. 2016) ..................................................................................................... 14
`
`
`
`iv
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 5 of 20 PageID #: 1491
`
`I.
`
`Procedural Note
`
`Amneal submits this brief in the event that: (a) the Court adopts Hospira’s “long-term
`
`stability” construction of “no more than about 2% decrease” of dexmedetomidine (“dex”) for at
`
`least five months, over the intrinsic evidence of indefiniteness; and (b) the Court finds that “no more
`
`than about 2% decrease” under long-term conditions is not an inherent property of the claimed
`
`formulation in the sealed glass container, contrary to Hospira’s admission and all of the long-term
`
`stability data presented at trial. If the Court makes both of these rulings, then the question of
`
`infringement of the ’106 patent turns on whether Hospira’s Dr. Linhardt provided credible, or even
`
`admissible, expert evidence sufficient to meet Hospira’s burden—when considered against the
`
`unrebutted expert testimony of Amneal’s biostatistician Dr. Bloch that Dr. Linhardt’s statistical
`
`models lacked any statistical significance.
`
`II.
`
`Introduction
`
`Hospira relegates Dr. Linhardt’s trial evidence to the back of its brief for good reason. Dr.
`
`Linhardt arbitrarily selected two statistical models, and presented inaccurate and statistically
`
`meaningless results. He even admitted that he did not choose those models based on any scientific
`
`authority or literature. Instead, he chose them based on speculation and simplicity. In a post hoc
`
`effort to justify that choice, Hospira’s counsel provided him with unrelated articles from The Journal
`
`of Water Research and The Journal of Hazardous Materials. However, Dr. Linhardt was forced to admit
`
`that he did not know or use either of these journals—and that the articles did not concern dex or
`
`even pharmaceuticals in general.
`
`Dr. Daniel Bloch, the only biostatistician in the case, provided unrebutted evidence that Dr.
`
`Linhardt’s results were statistically meaningless, and so failed to establish even a 50% likelihood that
`
`Amneal’s product met the “no more than about 2% decrease” limitation. Dr. Linhardt dared not
`
`attempt to rebut Dr. Bloch’s evidence of statistical insignificance. Nor did Dr. Linhardt offer a
`
`1
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 6 of 20 PageID #: 1492
`
`substantive defense for his failure to account for what he admitted was a 2–5% variability in the data
`
`he used—that is, variability up to more than twice the 2% claim limitation. Whenever pressed on the
`
`substantive defects of his analysis, Dr. Linhardt fell back on accusing Amneal of not having
`
`provided enough data—the implication being that Hospira’s failure to pursue the data Dr. Linhardt
`
`requested somehow excuses the scientific inadequacy of his work.
`
`Hospira attempts to overcome Dr. Linhardt’s failure of proof by legal argument. However,
`
`Hospira’s Sunovion argument is contrary to the Federal Circuit’s own refusal to apply Sunovion where,
`
`as here, the ANDA does not address the specific limitation at issue. The text of Amneal’s ANDA
`
`does not directly address long-term stability testing at five months. Accordingly, in another post hoc
`
`rescue bid, Hospira offers its own mathematical calculations for the first time in post-trial briefing.
`
`Such untimely lawyer-provided expert opinion evidence is inadmissible—and in any event, it cannot
`
`change what is written in the ANDA.
`
`Hospira’s attempt to brush aside the unrebutted expert testimony of Dr. Bloch likewise lacks
`
`any basis in law. Courts uniformly have rejected Hospira’s argument that an expert who is not a
`
`POSA cannot offer evidence regarding infringement. Indeed, this Court already rejected Hospira’s
`
`argument in overruling the objection to Dr. Bloch at trial. Infringement is an objective factual issue,
`
`not one which must, or even may, be assessed through the subjective eyes of a POSA. Dr. Bloch
`
`opined on precisely his area of technical expertise as to the scientific propriety and power of
`
`Hospira’s proffered statistical analysis.
`
`Accordingly, if the Court makes the rulings described in the Procedure Note above, Amneal
`
`is entitled to judgment of non-infringement of claim 6 of the ’106 patent.
`
`III. Argument
`
` “[T]he determination of infringement is a question of fact…” Abbott Labs., v. Torpharm, Inc.,
`
`300 F.3d 1367, 1373 (Fed. Cir. 2002). “An infringement inquiry provoked by an ANDA filing under
`
`2
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 7 of 20 PageID #: 1493
`
`35 U.S.C. § 271(e)(2)(A) is focused on the product that is likely to be sold following FDA approval.”
`
`Id. The patentee must “prove that the product that [the accused infringer] ultimately would put on
`
`the market would likely infringe the [asserted] patent.” Glaxo, Inc. v. Novopharm, Ltd., 110 F.3d 1562,
`
`157 (Fed. Cir. 1997). In addition, “[o]ne who does not infringe an independent claim cannot
`
`infringe a claim dependent on (and thus containing all the limitations of) that claim.” Wahpeton
`
`Canvas Co. v. Frontier, Inc., 870 F.2d 1546, 1552 n.9 (Fed. Cir. 1989).
`
`Here, Hospira failed to prove infringement of the asserted claim 6 of the ’106 patent at trial,
`
`and should not now be permitted to circumvent that failure of proof by contending that it is entitled
`
`to infringement as a matter of law.
`
`A. Sunovion Does Not Apply Because Amneal’s ANDA Specification Does Not
`Directly Address the Claim Limitation.
`
`Hospira asks the Court to ignore the entire trial record, and instead find that infringement is
`
`mandated by Sunovion Pharmaceuticals, Inc. v. Teva Pharmaceuticals USA, Inc., 731 F.3d 1271 (Fed. Cir.
`
`2013). (See D.I. 101 at 4–7.) Hospira argues that because Amneal’s specification provides for an
`
`assay of 90%–110% of the label claim over two years, that means that there can be no more than
`
`10% loss over two years, and therefore theoretically could include products that exhibit less than 2%
`
`concentration decrease over five months. (See id.) In other words, Hospira argues that because the
`
`specification permits infringing products, a finding of infringement is required. This is exactly the
`
`type of argument that the Federal Circuit has squarely and repeatedly rejected.
`
`The Federal Circuit rejected this infringement theory in similar circumstances less than one
`
`year after Sunovion, in Ferring B.V. v. Watson Labs., Inc.-Fla., 764 F.3d 1382 (Fed. Cir. 2014), which was
`
`decided by a panel that included two of the same judges as Sunovion. In Ferring, the disputed claim
`
`limitation required “less than about 70% by weight [of drug] released at about 45 minutes,” and the
`
`ANDA specification recited that “at least 80 percent by weight of the [drug] would dissolve in 60
`
`minutes.” Id. at 1385–86. The district court determined that the generic defendant was infringing
`
`3
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 8 of 20 PageID #: 1494
`
`because the dissolution specification at 60 minutes included products that met the claimed
`
`dissolution at 45 minutes. Id. at 1387. The Federal Circuit disagreed. It concluded that whereas in
`
`Sunovion the “ANDA specified an infringing product,” the ANDA in Ferring did not “‘clearly describe[]
`
`a product that meets the limitations of the asserted claims.’” Id. (citing Sunovion, 731 F.3d at 1280)
`
`(emphasis added). The Ferring Court deemed the ANDA specification “silent with respect to the
`
`claim limitations of the patents-in-suit, which do not specify dissolved dissolution rate at 60 minutes.”
`
`Id. at 1387 (emphasis added).
`
`Indeed, it is well-settled that under the circumstances of Ferring and this case, the patentee
`
`cannot rely solely on the ANDA specification and instead must prove infringement by relying “on
`
`evidence that the ANDA applicant ‘would likely sell an infringing composition pursuant to an
`
`approved ANDA.’” Id. at 1388 (quoting Glaxo, 110 F.3d at 1570); see also Glaxo, Inc. v. Novopharm,
`
`Ltd., 110 F.3d 1562, 1567 (Fed. Cir. 1997) (rejecting the argument that “the alleged infringer must
`
`disprove infringement if the ANDA permits sale of a composition that may include an infringing
`
`product”).
`
`The analogous facts in Ferring compel the conclusion that Sunovion does not apply here. In
`
`Ferring, even though the 60-minute dissolution ANDA specification included products that met the
`
`claimed 45-minute dissolution, that was not sufficient for infringement under Sunovion, as a matter of
`
`law. Id. at 1387. Likewise, here, even though Amneal’s FDA-mandated specification of 90–110%
`
`over a shelf life of two years could theoretically include products that exhibit less than 2% decrease
`
`in concentration at five months, that is not sufficient for infringement, as a matter of law. Amneal’s
`
`ANDA specification does not directly address or specify any concentration decrease at five months.
`
`(JTX-76.4.) Because the ANDA is silent as to the claimed 2% limitation, Hospira must actually
`
`prove infringement as a matter of fact. Ferring, 764 F.3d at 1387.
`
`4
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 9 of 20 PageID #: 1495
`
`What is more, given the undisputed variability of the concentration measurement at issue
`
`here (“[a]ssay of dexmedetomidine by HPLC”), the ANDA specification cannot possibly be
`
`interpreted to directly address the claimed 2% concentration decrease over five months, as
`
`confirmed by Dr. Linhardt himself. (JTX-76.4.) Amneal’s ANDA specification recites only that the
`
`concentration measurements for at least a two-year period must be “[b]etween 90.0% and 110.0% of
`
`label claim.” (Id.) And as Dr. Linhardt testified, it is well-known that these individual concentration
`
`values are subject to inherent variability (e.g., 3–5%), such that the values alone are not conclusive of
`
`whether there is greater than or less than about 2% decrease from one measurement to the next.
`
`(Tr. 289:16–291:11, 353:3–354:8.) Accordingly, this case in particular is not one where the ANDA
`
`specification clearly defines the product “in a manner that directly addresses the issue of
`
`infringement.” Abbott Labs., v. Torpharm, Inc., 300 F.3d 1367, 1373 (Fed. Cir. 2002).
`
`As a last resort, Hospira seems to argue from whole cloth that even though Amneal’s
`
`ANDA specification does not specifically address the claimed 2% at five months limitation, it does
`
`match the limitation after some arithmetic in its briefing. (See D.I. 101 at 6–7.) Hospira had its
`
`chance in expert reports and at trial to offer expert evidence as to any interpretation or implications
`
`of what is written in Amneal’s ANDA. Hospira failed to do so. Hospira cannot cure that failure
`
`through argument in its brief. Gemtron Corp. v. Saint–Gobain Corp., 572 F.3d 1371, 1380 (Fed. Cir.
`
`2009); Johnston v. IVAC Corp., 885 F.2d 1574, 1581 (Fed. Cir. 1989).
`
`Regardless of how Hospira tries to stretch the ANDA specification or the law, the fact
`
`remains that because Amneal’s specification on its face does not directly address the infringement
`
`issue, it does not invoke infringement under Sunovion, under the controlling Federal Circuit law. See
`
`The Medicines Co. v. Mylan, Inc., 853 F.3d 1296, 1310 (Fed. Cir. 2017) (ruling Sunovion inapplicable
`
`because “Mylan’s ANDA does not on its face establish that Mylan’s compounding process” satisfies
`
`the disputed claim limitation).
`
`5
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 10 of 20 PageID #: 1496
`
`B.
`
`Hospira’s Attempts to Dismiss Amneal’s Biostatistician Expert Dr. Bloch
`Because He Is Not a POSA Should Be Rejected as a Matter of Law.
`
`The second way Hospira tries to salvage its failure of proof is to suggest that Amneal’s
`
`expert Dr. Bloch’s opinions should be disregarded because he admittedly is not a POSA, but rather
`
`an expert biostatistician. (D.I. 101 at 1–2, 12.) The Court already rejected this argument at trial, and
`
`should reject it again now. (Tr. 435:9–21 (overruling Hospira’s objection because Dr. Bloch’s
`
`“testimony is very relevant to evaluating Dr. Linhardt’s testimony”).)
`
`Hospira’s criticism of Dr. Bloch as a “non-POSA statistician” relies on the Federal Circuit’s
`
`decision in Sundance, Inc. v. DeMonte Fabricating Ltd., 550 F.3d 1356 (Fed. Cir. 2008). But Sundance
`
`involved the admissibility of testimony from a patent lawyer on technical issues, where the Federal
`
`Circuit explained that “[u]nless a patent lawyer is also a qualified technical expert, his testimony on
`
`these kinds of technical issues is improper and therefore inadmissible.” Id. at 1362. Far from
`
`holding that testimony on the factual issue of infringement must come from a POSA, the Sundance
`
`Court explicitly noted: “Those of ordinary skill (or those of extraordinary skill) in the particular art
`
`are not, of course, the only witnesses who may testify as experts in a patent trial.” Id. at 1363 n.5.
`
`Hospira’s reliance on Sundance for its dismissal of Dr. Bloch—who is unquestionably an
`
`expert in the relevant technical field of biostatistics—is entirely misplaced. As the Federal Circuit
`
`stated in SEB S.A. v. Montgomery Ward & Co., Inc., 594 F.3d 1360, 1373 (Fed. Cir. 2010), “[t]his case
`
`comes nowhere close to the unusual situation in [Sundance],” where the testimony offered was that
`
`“of a patent law expert ‘[d]espite the absence of any suggestion of relevant technical expertise.’” See
`
`also Allure Energy, Inc. v. Nest Labs., Inc., No. 9-13-CV-102, 2015 WL 11110634, at *2–3 (E.D. Tex.
`
`Apr. 13, 2015) (“Defendants conflate a [POSA] with the requirements for expert testimony…
`
`Defendants’ reliance on Sundance is misplaced, as the expert in that case possessed no relevant
`
`technical background, which the Federal Circuit repeatedly referenced.”).
`
`6
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 11 of 20 PageID #: 1497
`
`Indeed, a Court in this district recently rejected a similar argument that the testimony of a
`
`defendant’s biostatistician expert would not be relevant because he was not a POSA. In Novartis AG
`
`v. Actavis Elizabeth LLC, Nos. 14–1487, 15–150, 15–151, 15–975, 2017 WL 1398347 (D. Del. Apr.
`
`17, 2017), the Court explained: “That [defendant’s biostatistician expert] does not meet either
`
`parties’ proposed definition of a person of ordinary skill in the art does not mean that he lacks
`
`relevant expertise. Rather, [the biostatistician’s] knowledge, skill, experience, training, and education
`
`are likely to assist the trier of fact.” Id. at *1. The same is true here with respect to the non-
`
`infringement testimony of Amneal’s biostatistician expert Dr. Bloch.
`
`Dr. Bloch’s qualifications to testify on whether the data and Dr. Linhardt’s analysis proved
`
`infringement as to the quantitative 2% limitation are numerous, and undisputed. Indeed, Dr. Bloch
`
`has been a biostatistician with Stanford University for over 30 years and has published hundreds of
`
`papers in renowned statistical and medical journals, among many other notable credentials. (Tr.
`
`424:16–431:2; DTX-516.)
`
`C.
`
`The Stability Data Analysis and Opinions Offered by Dr. Linhardt Did Not
`Prove Infringement, as Confirmed by Dr. Bloch.
`
`There is a reason why Hospira argues that infringement of the 2% limitation should turn on
`
`something other than its expert’s evidence at trial. Dr. Linhardt’s evidence was manifestly
`
`insufficient, even if his proposed construction of the indefinite term was not erroneous as a matter
`
`of law. (See D.I. 100 at 20–28.)1
`
`Dr. Linhardt relied on Amneal’s stability testing done at the industry standard up to six
`
`months under accelerated conditions, and up to eighteen months under long-term conditions. (D.I.
`
`
`1 Hospira also mischaracterizes the testimony of Amneal’s expert Dr. Alpaslan Yaman, in suggesting
`he testified that the claimed stability of the ’106 patent refers to long-term (room temperature)
`testing. (D.I. 101 at 4 (citing Tr. 691:4–692:10).) The cited testimony reveals that Dr. Yaman said
`no such thing. Indeed, there can be no mistake as to Dr. Yaman’s opinion at trial that the 2%
`limitation is indefinite because of the conflicting intrinsic evidence on whether long-term or
`accelerated conditions at five months should be used to assess infringement. (Tr. 572:16–573:22.)
`
`7
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 12 of 20 PageID #: 1498
`
`101 at 7–9; JTX-56.) There were no measurements taken at the claimed five months in any study.
`
`(JTX-56.) And neither Hospira nor Dr. Linhardt claimed that the individual data points or raw data
`
`in the studies alone could prove infringement. Instead, Dr. Linhardt purported to “appropriately
`
`analyze the data” with “regression analysis.” (D.I. 101 at 9.) In short, he tried to fit the data to a
`
`statistical model to determine the five-month concentration decrease under long-term conditions.
`
`What followed, however, was a series of unsupported speculations, statistical failures, and
`
`miscalculations that fell far short of what is required of reliable, credible evidence of infringement.
`
`1.
`
`As Explained by Dr. Bloch, Dr. Linhardt’s Analysis Did Not Account
`for Any Variability of the Data, Even Though He Testified That Such
`Variability Typically Exceeds 2%.
`
`The first deficiency in Dr. Linhardt’s analysis that should result in a finding of non-
`
`infringement comes from his own testimony about the variability built into Amneal’s stability data.
`
`At the very outset of his regression analysis, Dr. Linhardt testified on direct examination that the
`
`individual data points in such studies are known to have a typical variability of “three percent, five
`
`percent”—which would swallow the 2% limitation at issue here. (Tr. 289:16–291:3.) On cross-
`
`examination, Dr. Linhardt reiterated the known variability of “two or three percent,” that “can be as
`
`high as five percent.” (Tr. 353:3–354:8.)
`
`As a hypothetical illustration of the impact of Dr. Linhardt’s variability, this means that an
`
`initial measurement of 97% could actually be 99% (or higher) concentration, and an ensuing
`
`measurement of 97% could actually be 95% (or lower) concentration. So, what would appear to be
`
`a 0% loss would in reality be a 4% (or greater) loss—solely because of the known variability. Yet
`
`Dr. Linhardt did not even attempt to explain how his estimates for infringement were not skewed by
`
`this variability (which is significant at least as it relates to the 2% limitation). (Tr. 446:11–447:16.)
`
`Tellingly, when given a chance on cross-examination to explain how he took the variability
`
`into account, Dr. Linhardt could only respond that he asked Hospira’s attorneys for more data, but
`
`8
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 13 of 20 PageID #: 1499
`
`never got any. (Tr. 357:18–358:3.) When confronted about the implications of this two-to-five
`
`percent variability for infringement, Dr. Linhardt’s simply responded, “I don’t present any error in
`
`my analysis.” (Tr. 357:11–17.) On redirect, Dr. Linhardt only exacerbated the variability problem.
`
`He testified that POSAs—and even graduate students—know how to deal with these variabilities in
`
`the data, by determining the “coefficient of variability or CV,” as he called it. (Tr. 420:1–421:2.) But
`
`Dr. Linhardt never performed or presented any such “CV” analysis in this case.
`
`Dr. Linhardt’s lack of any accounting for the greater than 2% variability alone undermines
`
`the rest of his analysis. It also refutes Hospira argument that the “actual individual stability data
`
`points” could serve as any sort of “final check” on infringement. (See D.I. 101 at 11.) This latter
`
`argument should be further rejected as contradicted by Dr. Linhardt’s own testimony about how
`
`concentration decrease should be determined—testimony that Hospira relied on just two pages
`
`earlier in its opening brief. (See D.I. 101 at 9 (citing Tr. 289:16–290:10, 343:24–344:7, 344:16–345:1);
`
`Tr. 343:18–345:1.)
`
`2.
`
`Dr. Linhardt’s Point Estimates Were Premised on Unsupported and
`Illogical Statistical Models.
`
`The next part of Dr. Linhardt’s analysis—the selection of two alternative models, zero-order
`
`(linear) and first-order (exponential)—was no less problematic. (See D.I. 101 at 10.) As Amneal’s
`
`biostatistician expert Dr. Bloch explained, to create a statistical model for a dataset, one must first
`
`determine the correct rate relationship (e.g., linear, exponential, mixed, among many others)
`
`according to the science. (Tr. 434:23–435:6, 437:10–15.) If the model is incorrect or
`
`unsubstantiated, the same will be true of any conclusions from the model. (Tr. 437:10–438:5.) That
`
`is precisely what happened in this case.
`
`Here Dr. Linhardt picked two alternative models without citing any scientific or literature
`
`support for either one. For example, Dr. Linhardt justified his choice of an exponential model by
`
`speculating that “with other types of losses, like oxidative loss or chemical decomposition, you might
`
`9
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 14 of 20 PageID #: 1500
`
`want to go to a higher order of loss, like a first order of loss…” (Tr. 295:22–296:18 (emphasis
`
`added).) Likewise, while he alluded to some sort of literature support for adsorption corresponding
`
`to zero order (linear) or “pseudo first order or zero order” loss, that literature never surfaced. (Id.)
`
`Dr. Linhardt chose these two models merely because he “would consider both types of mechanisms
`
`possible,” (Tr. 295:22–296:18 (emphasis added)), but admitted the true rate relationship “could be
`
`some other loss that we don’t really know about,” (Tr. 294:7–15). This type of speculative, wavering
`
`testimony is not a proper basis for infringement. E.g., Fresenius Kabi USA, LLC v. Dr. Reddy’s Labs.,
`
`Ltd., 65 F. Supp. 3d 379, 386 (D. Del. 2014) (Andrews, J.).
`
`Indeed, Dr. Linhardt admitted on cross-examination that his approach of just “looking at the
`
`two simplest rate equations possible” was not based on literature or other objective support—but
`
`instead on his own say-so and experience with students. (Tr. 388:12–389:16.) The only two journal
`
`articles that Dr. Linhardt did ever rely on did not concern dex, nor indeed any kind of
`
`pharmaceutical whatsoever. (Tr. 380:9–381:5, 382:14–383:13, 384:18–385:23; JTX-73 (from the
`
`“Journal of Hazardous Materials”); JTX-74 (from the “Journal of Water Research”).) Dr. Linhardt
`
`admitted that he had never used either journal in his own work. (Id.) Notably, Dr. Linhardt did not
`
`even mention these references during his direct examination. They had been supplied to him by
`
`counsel after Dr. Bloch had noted the absence of any authority in Dr. Linhardt’s opening report.
`
`(Tr. 378:17–379:11, 386:9–19.)
`
`Beyond the lack of scientific justification, Dr. Bloch explained that Dr. Linhardt’s modeling
`
`was also illogical—because if both linear and exponential loss were truly at play, the correct model
`
`may be a mix of the two, not separate alternatives. (Tr. 447:17–448:19.) Hospira now argues that
`
`the result of a mixed model would be no different. (D.I. 101 at 13–14 (citing Tr. 481:20–482:1).)
`
`But Hospira’s post hoc argument cannot cure Dr. Linhardt’s flawed technical analysis at trial. And the
`
`testimony of Dr. Bloch that Hospira cites plainly does not support this proposition. Instead, Dr.
`
`10
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 15 of 20 PageID #: 1501
`
`Bloch just reiterated what he said all along—that Dr. Linhardt’s alternative models made no sense.
`
`(Tr. 480:21–482:1.)
`
`In sum, the statistical models on which Dr. Linhardt’s opinions were built lacked the
`
`required scientific reliability to be credible, or even admissible, evidence of infringement on the
`
`claimed 2% limitation. Calhoun v. Yamaha Moto Corp., U.S.A., 350 F.3d 316, 322 (3d Cir. 2003)
`
`(affirming exclusion of expert testimony because “[t]here was no literature confirming [the expert’s]
`
`theory, nor demonstrable tests,” so “his testimony was speculative and unreliable.”).
`
`3.
`
`None of the Five-Month Concentration Decrease Predictions for
`Amneal’s ANDA Products Was More Likely Than Not to Be Less
`Than the Claimed About 2%.
`
`Ultimately, where Dr. Linhardt went most wrong was in his failure to provide even a single
`
`accurate prediction from his statistical modeling that was more likely than not to fall within the claimed
`
`less than about 2% decrease at five months, as required to prove infringement. As Dr. Bloch
`
`explained, for a statistical model to determine whether a predicted value (e.g., concentration
`
`decrease) at a given time point (e.g., five months) will fall within a defined parameter (e.g., no more
`
`than about two percent), one must calculate the predicted range of the true value within a desired
`
`confidence level—which for infringement is something above 50% or even odds. (Tr. 441:9–
`
`442:23.) Without providing this range of where the true value is likely to fall, Dr. Bloch explained
`
`that “you don’t know whether or not you are actually on or about a certain amount, which is at issue
`
`here.” (Tr. 442:24–443:5.) Thus, while Hospira offered Dr. Linhardt’s various point estimates at
`
`five months (or two months) as proof of infringement, these figures alone are facially insufficient to
`
`meet the more-likely-than-not legal standard. (See D.I. 101 at 10–11.)2
`
`
`2 Dr. Linhardt’s estimates under accelerated conditions are erroneous for the same reasons as his
`long-term estimates, but also because Dr. Linhardt’s proposal that the limitation can be satisfied by
`something less than five months (e.g., two months) under accelerated conditions contradicts the
`intrinsic evidence, and thus cannot be correct. (See JTX-4 at 13:61–14:20; D.I. 100 at 20–28.)
`
`11
`
`

`

`Case 1:15-cv-00697-RGA Document 105 Filed 10/10/17 Page 16 of 20 PageID #: 1502
`
`In fact, it is undisputed that Dr. Linhardt failed to show that any of his point estimates at
`
`five months (or at two months under the accelerated measure) was statistically more likely than not
`
`to be the true concentration decrease. Even Dr. Linhardt himself admitted in no uncertain terms
`
`that he pro