`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`__________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`__________
`
`MYLAN PHARMACEUTICALS INC.,
`Petitioner
`
`v.
`
`BIOGEN MA INC.,
`Patent Owner
`
`______________________________________________________
`
`Case IPR2018-01403
`Patent 8,399,514 B2
`______________________________________________________
`
`DECLARATION OF RONALD A. THISTED, PH.D.
`
`1
`
`
`
`Biogen Exhibit 2060
`Mylan v. Biogen
`IPR 2018-01403
`
`Page 1 of 77
`
`

`

`I.
`II.
`
`TABLE OF CONTENTS
`Introduction and Qualifications ....................................................................... 1
`Background ...................................................................................................... 5
`A. Documents Considered ......................................................................... 5
`B.
`Legal Standards ..................................................................................... 5
`C.
`Person of Ordinary Skill in the Art ....................................................... 5
`D.
`The ’514 Patent ..................................................................................... 6
`E.
`Background of MS ................................................................................ 7
`III. Based on Kappos 2006 and the Kappos Presentation, the results
`with 480 mg/day would be expected to be closer to those with 360
`mg/day than with 720 mg/day. ........................................................................ 8
`A.
`Review of the Phase II results reported in Kappos 2006 and
`the Kappos Presentation ........................................................................ 8
`There was no established effective dose range for DMF as a
`treatment for MS that would suggest 480 mg/day would be
`efficacious ........................................................................................... 11
`IV. The Biogen Phase II trial results do not reveal any flaw that could
`be adjusted post hoc as suggested by Drs. Corboy, Benet, and
`McKeague ...................................................................................................... 15
`A.
`Post hoc analyses can be unreliable .................................................... 15
`B.
`The post hoc analyses proposed by Drs. Corboy, Benet, and
`McKeague are inappropriate ............................................................... 17
`1.
`Dr. Benet’s subtraction analysis is unsound and
`unreliable ................................................................................... 20
`Dr. Benet’s division analysis is unsound and
`unreliable ................................................................................... 24
`
`B.
`
`2.
`
`i
`
`
`
`Page 2 of 77
`
`

`

`
`
`V.
`
`B.
`
`3.
`
`C.
`
`D.
`
`The prior art also did not provide sufficient
`information to perform Dr. Benet’s subtraction and
`division calculations accurately ................................................ 26
`A person of ordinary skill would not have conducted the
`analysis in Fox 2011 before February 2007 ........................................ 31
`The post hoc sensitivity analyses in EMA 2013
`demonstrated the robustness of Biogen’s Phase II results .................. 35
`Biogen’s Phase III trials demonstrated unexpected results ........................... 41
`A.
`The DEFINE and CONFIRM Phase III studies had the
`unexpected result that the effectiveness of 480 mg/day was
`similar to that of 720 mg/day. ............................................................. 41
`The 480 mg/day and 720 mg/day dosing arms may properly
`be compared, as confirmed by multiple regulatory agencies .............. 49
`VI. Conclusion ..................................................................................................... 55
`
`
`Page 3 of 77
`
`

`

`
`
`I, Ronald A. Thisted, Ph.D., hereby declare the following:
`
`I.
`
`Introduction and Qualifications
`I have been retained by Finnegan, Henderson, Farabow, Garrett &
`1.
`
`Dunner, LLP as an expert consultant for this inter partes review proceeding. I
`
`understand that the patent at issue is U.S. Patent No. 8,399,514 (“the ’514 patent”;
`
`Ex. 1001) and that it is owned by Biogen MA Inc. I am being compensated for my
`
`time at my standard hourly rate. My compensation is not contingent upon my
`
`opinions or the outcome of this or any other proceeding.
`
`2.
`
`I am Professor Emeritus in the Departments of Statistics and Public
`
`Health Sciences at the University of Chicago. At the time of my retirement in
`
`2018, I also held faculty appointments in the Department of Anesthesia & Critical
`
`Care and the Committee on Clinical Pharmacology and Pharmacogenomics. As a
`
`Professor in this Committee, I participated as faculty in the clinical pharmacology
`
`course for fourth-year medical students, lectured on pharmacoepidemiology, and
`
`mentored and advised advanced medical residents’ research as part of their
`
`advanced training fellowships in clinical pharmacology.
`
`3.
`
`I received a bachelor’s degree (B.A.) in mathematics and philosophy
`
`in 1972 from Pomona College in Claremont, California. In 1973 and 1977,
`
`Page 1 of 62
`
`- 1 -
`
`Page 4 of 77
`
`

`

`
`
`respectively, I completed a master’s degree (M.S.) and a doctorate of philosophy
`
`(Ph.D.) in statistics at Stanford University in Palo Alto, California.
`
`4.
`
`I have more than forty years of research, academic, and practical
`
`experience in the area of biostatistics. My research has focused on biostatistics and
`
`epidemiology, statistical computation, and the effectiveness of medical
`
`intervention from a statistical perspective.
`
`5.
`
`I have held positions on the faculty of the University of Chicago since
`
`1976. I was Co-Director of the Clinical Research Training Program (1999–2012),
`
`Chairman of the Department of Health Studies [now Public Health Sciences]
`
`(1999–2012), Director of Population Sciences for the Institute for Translational
`
`Medicine (2007–2014), and Scientific Director for the Biostatistics Core Facility at
`
`the University of Chicago Cancer Research Center (1999–2014). From 2014 until
`
`my retirement, I was Vice Provost for Academic Affairs at the University.
`
`6.
`
`I have taught courses on statistics and biostatistics for over forty
`
`years. I have also taught courses on statistical methods, computation,
`
`epidemiology, and clinical research methods. I have been awarded the Llewellyn
`
`John and Harriet Manchester Quantrell Award for Excellence in Undergraduate
`
`Teaching.
`
`7.
`
`I am an Elected Fellow of the American Association for the
`
`Advancement of Science (1992) and of the American Statistical Association
`
`- 2 -
`
`Page 5 of 77
`
`

`

`
`
`(1988). I have also been a member of several professional societies related to
`
`statistics and computation, including the Association for Computing Machinery,
`
`the International Biometric Society, the Institute of Mathematical Statistics, the
`
`Royal Statistical Society, and the Society for Industrial and Applied Mathematics.
`
`8.
`
`I have authored over one hundred publications in the area of
`
`biostatistics and epidemiology in peer-reviewed journals including the New
`
`England Journal of Medicine, the Journal of the American Medical Association,
`
`and The Lancet. I also served as Associate Editor of the Journal of the American
`
`Statistical Association (1979–1985, 1987–1988) and of ACM Transactions on
`
`Mathematical Software (1990–1992).
`
`9.
`
`I have served as Database Editor (1994), Managing Editor (1995), and
`
`Editor (1996–1998) for Current Index to Statistics. I sat on the editorial board of
`
`SIAM Journal of Scientific and Statistical Computing from 1983 to 1985. I have
`
`served as a referee for several journals related to biostatistics, including Annals of
`
`Statistics, PLoS One, and Statistics in Medicine. As a journal referee, I reviewed
`
`submitted articles for scientific quality. I have also acted as a referee for the
`
`National Institutes of Health (NIH) and the National Science Foundation (NSF).
`
`As a referee for NIH and NSF, I reviewed grant proposals for potential funding.
`
`10. Since the late 1970s, I have consulted for the pharmaceutical and
`
`medical device industries on the design of clinical trials and statistical analysis of
`
`- 3 -
`
`Page 6 of 77
`
`

`

`
`
`clinical trial results. I have consulted regarding the design of Phase I, Phase II, and
`
`Phase III clinical trials; designed data collection methods; planned and overseen
`
`statistical analysis of results; prepared reports for use by the FDA; and presented
`
`statistical aspects of clinical studies to the FDA.
`
`11. As part of my consulting practice, I have served as the principal
`
`biostatistician for studies involving MS patients. In particular, I was the principal
`
`biostatistician on two randomized clinical trials for treating specific symptoms in
`
`MS patients.
`
`12.
`
`I have extensive experience consulting in the pharmaceutical industry
`
`on the design and analysis of clinical trials, including dose-ranging trials. Dose-
`
`ranging trials are typically conducted as Phase II trials, and they seek to provide
`
`information on the dose-response curve for a drug in patients with a particular
`
`disease. Biogen’s Phase II trial is an example of such a dose-ranging trial.
`
`13. My teaching and consulting practices have also included design and
`
`analysis of pharmacokinetic and pharmacodynamic studies. As part of my
`
`consulting practice I regularly analyze pharmacokinetic and pharmacodynamic
`
`data, conduct pharmacokinetic modeling and simulations, and analyze dose-
`
`response relationships.
`
`14. A copy of my current curriculum vitae is attached as Appendix A.
`
`- 4 -
`
`Page 7 of 77
`
`

`

`
`
`II. Background
`A. Documents Considered
`15. My opinions are based on my knowledge and experience. In forming
`
`my opinions, I have also considered the documents listed in Appendix B.
`
`B.
`16.
`
`17.
`
`Legal Standards
`I am not an attorney and do not purport to offer legal opinions.
`
`I am informed by counsel and understand that a claim is unpatentable
`
`if it would have been obvious to a person of ordinary skill in the art before
`
`Biogen’s earliest filing date of February 8, 2007. I am informed by counsel and
`
`understand that an obviousness analysis involves considering the following factors:
`
`(1) the scope and content of the prior art; (2) the differences between the prior art
`
`and the claims; (3) the level of ordinary skill in the art; and (4) objective indicia of
`
`nonobviousness. I understand that one indicator of nonobviousness is if the
`
`claimed invention is unexpectedly different from the prior art.
`
`C.
`18.
`
`Person of Ordinary Skill in the Art
`I have been informed by counsel and understand that certain legal
`
`issues, such as claim construction and obviousness, are determined from the
`
`viewpoint of a person of ordinary skill in the art. I understand that a person of
`
`ordinary skill in the art is a hypothetical person who is presumed to have known
`
`the relevant art at the time of the invention. Further, I have been informed by
`
`counsel and understand that factors that may be considered in determining the level
`
`- 5 -
`
`Page 8 of 77
`
`

`

`
`
`of ordinary skill in the art include (1) the types of problems encountered in the art;
`
`(2) prior art solutions to those problems; (3) rapidity with which innovations are
`
`made; (4) sophistication of the technology; and (5) the educational level of active
`
`workers in the field.
`
`19. Dr. Benet states:
`
`[A] person of ordinary skill in the art here would have
`had (1) several years’ experience in designing clinical
`studies to meet regulatory expectations and/or analyzing
`data from such studies; (2) an advanced degree (PhD,
`MD, PharmD) and training in clinical pharmacology or
`experience treating MS; and (3) experience with the
`administration or formulation of therapeutic agents, their
`dosing, and the literature concerning drug development
`study and design.
`
`(Ex. 1003 ¶ 20.) For purposes of my report, I formulated my opinions with both
`
`this level of skill and Biogen’s in mind. I understand that Biogen has proposed
`
`that such a person would have had at least a medical degree with at least three
`
`years of training in neurology and at least three years of clinical experience treating
`
`MS. My opinions do not depend on which definition is adopted.
`
`D. The ’514 Patent
`20. The ’514 patent provides methods for using neuroprotective
`
`compounds in therapy for neurological diseases, including multiple sclerosis
`
`(“MS”). (Ex. 1001 at 1:12-14.) The claimed inventions include methods of
`
`treating subjects needing treatment for MS by (Ex. 1001 at claims 1, 11, 15, 20)
`
`- 6 -
`
`Page 9 of 77
`
`

`

`
`
`orally administering a pharmaceutical composition consisting essentially of a
`
`therapeutically effective amount of dimethyl fumarate (“DMF”), monomethyl
`
`fumarate (“MMF”), or a combination of the two, and one or more
`
`pharmaceutically acceptable excipients, in which the therapeutically effective
`
`amount of DMF, MMF, or the combination is about 480 mg/day (Ex. 1001 at
`
`claim 1); orally administering about 480 mg/day of DMF, MMF, or a combination
`
`of the two (Ex. 1001 at claim 11); orally administering a pharmaceutical
`
`composition consisting essentially of a therapeutically effective amount of DMF,
`
`and one or more pharmaceutically acceptable excipients, in which the
`
`therapeutically effective amount of DMF is about 480 mg/day (Ex. 1001 at
`
`claim 15); and treating the subject with a therapeutically effective amount of DMF,
`
`MMF, or a combination thereof, where the therapeutically effective amount of
`
`DMF, MMF, or the combination is about 480 mg/day (Ex. 1001 at claim 20).
`
`(Ex. 1001 at. 27:58-30:28.)
`
`E.
`21.
`
`Background of MS
`I have a general understanding regarding the aspects of MS measured
`
`in clinical trials as well as MS clinical trial populations. For ease of reference and
`
`as a basis for my opinions in my declaration, I have incorporated the following
`
`information in this section from Dr. Duddy’s declaration (Ex. 2058).
`
`22. Gadolinium enhancing (Gd+) lesions indicate active and recent
`
`- 7 -
`
`Page 10 of 77
`
`

`

`
`
`inflammation. (Ex. 2058 at ¶ 18.) Gd+ lesions disappear after about 6 weeks,
`
`because the blood brain barrier, which opens at the time of new lesion formation,
`
`closes, excluding the gadolinium contrast agent from the brain tissue. (Ex. 2058 at
`
`¶ 18.)
`
`III. Based on Kappos 2006 and the Kappos Presentation, the results with
`480 mg/day would be expected to be closer to those with 360 mg/day
`than with 720 mg/day.
`A. Review of the Phase II results reported in Kappos 2006 and the
`Kappos Presentation
`I have considered Kappos et al., 16th Meeting of the European
`
`23.
`
`Neurological Society (abstract to presentation May 30, 2006) (“Kappos 2006”)
`
`(Ex. 1007). Kappos 2006 concerned a 24-week randomized, double-blind,
`
`placebo-controlled clinical trial of four doses (one of which was placebo) of
`
`BG00012 in patients with relapsing-remitting MS “to determine the efficacy of
`
`three dose levels of BG00012 […] on brain lesion activity as measured by
`
`magnetic resonance imaging (MRI) […].” (Ex. 1007 at 27.) Although BG00012
`
`was described only as “a novel oral single-agent fumarate” and “a novel oral
`
`fumarate preparation” in Kappos 2006, I understand that BG00012 referred to a
`
`DMF pharmaceutical preparation. The doses studied were 0 mg/day (that is,
`
`placebo), 120 mg/day, 360 mg/day, and 720 mg/day. (Ex. 1007 at 27.) The
`
`prespecified primary treatment effectiveness outcome assessed was the total
`
`number of new Gd-enhancing (Gd+) brain lesions seen in four MRI brain scans at
`
`- 8 -
`
`Page 11 of 77
`
`

`

`
`
`12, 16, 20, and 24 weeks into the study against placebo. (Ex. 1007 at 27.)
`
`Prespecified secondary endpoints included the cumulative number of new Gd+
`
`lesions from week 4 to week 24 and the number of new or enlarging T2-
`
`hyperintense lesions at week 24, both calculated from MRI brain scan images,
`
`against placebo. (Ex. 1007 at 27.) Additional prespecified endpoints, including
`
`the number of new T1-hypointense lesions at week 24 against placebo, were also
`
`studied. (Ex. 1007 at 27.)
`
`24. Kappos 2006 reported that BG00012 at a dose of 720 mg/day
`
`“significantly reduced the mean number of new Gd+ lesions (the primary end
`
`point) compared with placebo.” (Ex. 1007 at 27.) It also reported that BG00012 at
`
`a dose of 720 mg/day “reduced the cumulative number of new Gd+ lesions, the
`
`number of new/enlarging T2-hyperintense lesions, and the number of new T1-
`
`hypointense lesions compared with placebo.” (Ex. 1007 at 27.) A press release
`
`dated May 30, 2006 confirmed that “[t]he results of the 120 mg and 360 mg BG-12
`
`treated groups were not statistically significant versus placebo.” (Ex. 1016 at 1.)
`
`25.
`
`I have also considered Kappos et al., 16th Meeting of the European
`
`Neurological Society (presentation May 30, 2006) (“Kappos Presentation”)
`
`(Ex. 1046 at 8-29), which appears to be the slide presentation related to the Kappos
`
`2006 abstract. As summarized in the table below, the Kappos Presentation
`
`indicated that the 720 mg/day dose of BG00012 showed a statistically significant
`
`- 9 -
`
`Page 12 of 77
`
`

`

`
`
`reduction in brain lesions compared to placebo in each of the four MRI outcome
`
`measures described above. (Ex. 1046 at 19-22.) In contrast, neither of the other
`
`two doses of BG00012 (120 mg/day and 360 mg/day) were statistically
`
`significantly different from placebo in any of the outcome measures. (Ex. 1046
`
`at 19-22.) Indeed, the Kappos presentation provided p values indicating statistical
`
`significance only for the 720 mg/day dose.
`
`Total daily dose 120 mg/day
`
`360 mg/day
`
`720 mg/day
`
`Dosing regimen 120 mg QD
`
`120 mg TID
`
`240 mg TID
`
`New Gd+ lesions, weeks 12-24
`
`New Gd+ lesions, weeks 4-24
`
`New T2-hyperintense lesions
`
`New T1-hypointense lesions
`
`NS
`
`NS
`
`NS
`
`NS
`
`NS
`
`NS
`
`NS
`
`NS
`
`p<0.001
`
`p=0.002
`
`p<0.001
`
`p=0.014
`
`
`Table 1. Comparison of MRI endpoints for each total daily dose compared to
`placebo. NS=not statistically significant (p>0.05). (See Ex. 1046.)
`
`
`
`26. Kappos 2006 and the Kappos Presentation demonstrated that
`
`BG00012 at a dose of 720 mg/day effectively reduced brain lesion activity
`
`measured by MRI in patients with relapsing-remitting MS, although it was not
`
`statistically significant for annualized relapse rate. (Ex. 1007 at 27; Ex. 1046
`
`at 19-22.) Specifically, the number of new Gd+ lesions seen in patients treated
`
`with 720 mg/day (by each of the four measures) was lower than that seen in
`
`patients taking placebo by an amount that was statistically significant (that is, by
`
`- 10 -
`
`Page 13 of 77
`
`

`

`
`
`an amount that was too large to be plausibly attributable to chance). (Ex. 1007 at
`
`27; Ex. 1046 at 19-22.) In contrast, Kappos 2006 and the Kappos Presentation
`
`revealed that the number of new Gd+ lesions seen in patients treated with 120
`
`mg/day and 360 mg/day (by each of the four measures) was not statistically
`
`significantly different compared to that seen in patients taking placebo. (Ex. 1007
`
`at 27; Ex. 1046 at 19-22.)
`
`27. Thus, in summary, Kappos 2006 and the Kappos Presentation
`
`reported that 720 mg/day was a statistically significant effective dose for reducing
`
`brain lesions in MS patients, while the lower doses of 360 mg/day and 120 mg/day
`
`were not.
`
`B.
`
`28.
`
`There was no established effective dose range for DMF as a
`treatment for MS that would suggest 480 mg/day would be
`efficacious
`I understand that the Board has presented a question of whether the
`
`discovery of 480 mg/day of DMF for treating MS was the result of optimizing the
`
`dose within an established effective range. The answer is no, because there was no
`
`established effective dose range for DMF as a treatment for MS at the time the
`
`’514 patent was filed. Dr. Corboy’s assertions that one could optimize within a
`
`“range” of effective doses are therefore incorrect. (Ex. 1002 ¶¶ 11, 126-127, 132,
`
`144-154, 173-181, 190-198.)
`
`29. The information in Kappos 2006 and the Kappos Presentation does
`
`- 11 -
`
`Page 14 of 77
`
`

`

`
`
`not indicate any known effective dose range for BG00012 in MS. All that Kappos
`
`2006 and the Kappos Presentation teach is that 120 and 360 mg/day each were not
`
`statistically significantly different from placebo, and that only a single dose (720
`
`mg/day) reached statistical significance in reducing brain lesions compared to
`
`placebo. (Ex. 1007 at 27; Ex. 1046 at 19-22.) Thus, because only one dose was
`
`known to reduce brain lesions, no established effective dose range can be derived
`
`from these results.
`
`30. These data from the Phase II trial also contradict Dr. Benet’s
`
`testimony. The lack of effect shown by the 120 and 360 mg/day doses reflects a
`
`flat, not statistically significant response at lower doses. These data demonstrate
`
`that Dr. Benet’s exemplary dose/response curve, which shows a steep increase in
`
`response at lower doses, does not accurately describe DMF in MS. (Ex. 1003
`
`¶ 167.)
`
`31. Moreover, in addition to showing a lack of any established effective
`
`dose range, the data are consistent with the 720 mg/day dose not having reached a
`
`maximum effect. Because the 720 mg/day dose resulted in a 69% reduction in new
`
`Gd+ lesions at weeks 12-24 compared to placebo, rather than 100%, the study
`
`results show that the 720 mg/day dose was not fully effective. (Ex. 1046 at 19.) In
`
`addition, the 720 mg/day dose was not statistically significant compared to placebo
`
`in reducing annualized relapse rate. (Ex. 1046 at 23, 27.) Thus, the data from the
`
`- 12 -
`
`Page 15 of 77
`
`

`

`
`
`Kappos Presentation indicate that it might be possible to achieve greater efficacy
`
`with higher doses than 720 mg/day.
`
`32. The results of the secondary endpoints in Biogen’s Phase II trial are
`
`further consistent with the absence of any known effective dose range for DMF in
`
`MS. (Ex. 1007 at 27; Ex. 1046 at 19-22.) Just as the 120 and 360 mg/day
`
`BG00012 doses were not statistically significant compared to placebo for the
`
`primary endpoint, they were also not statistically significant for the secondary
`
`endpoints of new Gd+ lesions over weeks 4 to 24 (Ex. 1046 at 20), new/newly
`
`enlarging T2 lesions at week 24 (Ex. 1046 at 21), or new T1 lesions at week 24
`
`(Ex. 1046 at 22). Indeed, for each of these three secondary endpoints, the outcome
`
`for 360 mg/day was worse than for 120 mg/day and was about the same as for
`
`placebo. By contrast, the 720 mg/day dose did have a statistically significant effect
`
`on these secondary endpoints (44% to 53% better than placebo), just as it did on
`
`the primary endpoint (which showed 69% improvement over placebo). (Ex. 1007
`
`at 27; Ex. 1046 at 19-22.) In my opinion, the consistency of the results when
`
`viewed as a whole, including the primary and secondary endpoints, further
`
`reinforces their reliability.
`
`33. My opinion that Kappos 2006 and the Kappos Presentation do not
`
`demonstrate any effective dose range is also consistent with the reporting in
`
`Kappos 2006. In the conclusion of Kappos 2006, the authors did not claim that
`
`- 13 -
`
`Page 16 of 77
`
`

`

`
`
`any effective dose range was known. Instead, they stated that BG00012
`
`significantly reduced brain lesion activity “in a dose-dependent manner.” (Ex.
`
`1007 at 27.) This “dose-dependent” language is consistent with the actual data
`
`reported, as a reduction in brain lesion activity in Biogen’s Phase II trial indeed
`
`depended on the dose, and a statistically significant response was evident only at
`
`the highest dose tested: the 720 mg/day dose was effective for reducing brain
`
`lesions while neither of the 120 mg/day or 360 mg/day dose results were
`
`statistically significantly different from placebo. (Ex. 1046 at 19-22; Ex. 1016 at
`
`1; see Ex. 1007 at 27.)
`
`34. The Kappos Presentation reported that the effects on brain lesions for
`
`the 120 mg/day dose group were essentially the same as those for the placebo
`
`group and that the differences between 120 mg/day and placebo were not
`
`statistically significant. (See Ex. 1046 at 19-22; Ex. 1016 at 1.) It further reported
`
`data indicating that tripling this dose (a 200% increase to 360 mg/day) also failed
`
`to produce a statistically significant reduction in brain lesions relative to placebo.
`
`(See Ex. 1046 at 19-22.) Consequently, based on the Kappos Presentation, a
`
`person of ordinary skill in the art would not have expected that increasing the dose
`
`by an additional 33% (to 480 mg/day) would produce statistically significant
`
`- 14 -
`
`Page 17 of 77
`
`

`

`
`
`efficacy when the lower dose (360 mg/day) did not.1 At most, the results would
`
`suggest only that the effects on brain lesions with a 480 mg/day dose would be
`
`closer to those with the 360 mg/day dose than those with a 720 mg/day dose.
`
`IV. The Biogen Phase II trial results do not reveal any flaw that could be
`adjusted post hoc as suggested by Drs. Corboy, Benet, and McKeague
`Post hoc analyses can be unreliable
`A.
`35. Performing additional analyses post hoc based on a review of
`
`unblinded data is inherently unreliable and must be viewed cautiously because
`
`such analyses necessarily involve hindsight and may therefore introduce significant
`
`bias. Conducting analyses that are motivated by the data (i.e., viewing the data and
`
`using the same data both to decide which after-the-fact analyses might produce
`
`favorable results and to carry out those analyses), rather than tested by the original
`
`study design and resulting data, is analogous to an archer redrawing the target after
`
`the arrow has landed.
`
`36. With pharmaceutical clinical trials, one purpose of typical post hoc
`
`analyses is not to determine the efficacy of a drug but rather to determine the
`
`extent to which the basic efficacy findings about a drug based on the
`
`
`1 I do not believe that Kappos 2006 and the Kappos Presentation provide any
`
`reason to even consider a dose of 480 mg/day. I merely provide an opinion
`
`regarding 480 mg/day because that is the dose at issue in this proceeding.
`
`- 15 -
`
`Page 18 of 77
`
`

`

`
`
`predetermined protocol are robust to an array of possible assumptions or
`
`alternative statistical models. This type of post hoc analysis is known as a
`
`sensitivity analysis, as it evaluates how sensitive the results are to a changing of
`
`assumptions and statistical models. The relevance of any particular result from a
`
`collection of sensitivity analyses depends heavily on the specific assumptions
`
`made that underlie the analysis. Another type of post hoc analysis is an
`
`exploratory investigation, which analyzes the data with hindsight to follow up on
`
`certain features in the data set that investigators may have noticed. Such features
`
`could derive from chance phenomena or from real effects, but without further
`
`empirical testing, it is generally impossible to determine which of these
`
`possibilities is present. Such results, whether from sensitivity analyses or
`
`exploratory analyses, can only be regarded as hypothesis-generating rather than
`
`hypothesis-testing. That is, they can generate questions but not provide new
`
`answers. For the same reasons, post hoc analyses suggested by the data are often
`
`spurious and without empirical replication they are unreliable. (See Wang et al.,
`
`Statistics in Medicine—Reporting of Subgroup Analyses in Clinical Trials, 357 N.
`
`ENGL. J. MED. 2189-94 at 2190 (2007) (Ex. 2071) (“Post hoc analyses refer to
`
`those in which the hypotheses being tested are not specified before any
`
`examination of the data. Such analyses are of particular concern because it is often
`
`unclear how many were undertaken and whether some were motivated by
`
`- 16 -
`
`Page 19 of 77
`
`

`

`
`
`inspection of the data.”).) Dr. McKeague agreed at his deposition, stating that post
`
`hoc analyses can only generate hypotheses, as “you’re actually cherry-picking” in
`
`a post hoc analysis. (Ex. 2064 at 121:8-15.)
`
`37. Drs. Corboy, Benet, and McKeague did not conduct any testing to
`
`analyze any prespecified hypothesis. They instead formed their opinions and
`
`conducted their analyses after reviewing the unblinded data in the Kappos
`
`Presentation with hindsight knowledge of the results of the 480 mg/day dose in the
`
`later Phase III trial. It is precisely within this context that a person of ordinary skill
`
`would have understood that a post hoc analysis is unreliable for forming any
`
`predictions or conclusions. (Rothwell, Subgroup Analysis in Randomized
`
`Controlled Trials: Importance, Indications, and Interpretation, 365 LANCET 176-
`
`86 at 184 (2005) (Ex. 2070) (“Post-hoc observations should be treated with
`
`scepticism [sic] . . . .”).)
`
`B.
`
`38.
`
`The post hoc analyses proposed by Drs. Corboy, Benet, and
`McKeague are inappropriate
`I have reviewed the declarations of Drs. Corboy, Benet, and
`
`McKeague, who contend that one of ordinary skill would have used mean baseline
`
`Gd+ lesion to adjust the results of Biogen’s Phase II data on the pre-specified
`
`primary endpoint. (Ex. 1003 at ¶¶ 169-178; Ex. 1004 at ¶¶ 24-30.) In particular,
`
`Dr. Benet states that a person of ordinary skill would be skeptical of the result
`
`reported for the 360 mg/day dose and that one could “normalize” the data to adjust
`
`- 17 -
`
`Page 20 of 77
`
`

`

`
`
`for the baseline Gd+ lesions. (Ex. 1003 at ¶¶ 169-178.) Dr. McKeague states that
`
`one of ordinary skill would adjust for baseline Gd+ lesions in a post hoc analysis
`
`because Gd+ lesions “can be measured accurately, and [are] clearly related to the
`
`endpoint investigated in the study.” (Ex. 1004 at ¶¶ 24-30.) Dr. Corboy offers a
`
`similar position, proposing a “correction” for baseline mean Gd+ lesions. (Ex.
`
`1002 at ¶¶ 203-211.) For the reasons discussed below, I disagree that the
`
`calculations proposed by Dr. Benet, as well as the positions taken by Dr. Corboy
`
`and Dr. McKeague,2 provide any reasonable interpretation of the data presented in
`
`the Kappos Presentation.
`
`39. First, the Kappos Presentation does not indicate any discrepancy in
`
`outcomes associated with the reported mean number of baseline lesions across
`
`study arms. Moreover, as discussed above, the Phase II results were coherent and
`
`consistent across all endpoints, with the 720 mg/day dose consistently
`
`demonstrating statistically significant superiority to placebo, while 360 mg/day
`
`failed to reach statistical significance on any endpoint. (Ex. 1046 at 19-22.) Given
`
`
`2 I note that only Dr. Benet proposes particular calculations that he states one of
`
`ordinary skill “could” make. (Ex. 1003 at ¶¶ 169-179.) Dr. McKeague and Dr.
`
`Corboy merely discuss Fox 2011 and EMA 2013, both of which I address later in
`
`my declaration. (Ex. 1002 at ¶¶ 203-211; Ex. 1004 at ¶¶ 24-30.)
`
`- 18 -
`
`Page 21 of 77
`
`

`

`
`
`this consistency of the reported results in Biogen’s Phase II trial, there is no
`
`suggestion in the presentation that any type of reanalysis would be warranted or
`
`necessary.
`
`40. Second, even if the presentation and data in the Kappos Presentation
`
`had suggested some type of recalculation was warranted, the “rough” calculations
`
`in Dr. Benet’s declaration are unsound and unreliable for many reasons. (Ex. 1003
`
`at ¶¶ 169-179.) I therefore disagree that a person of ordinary skill in the art would
`
`have relied on such calculations in drawing any conclusions as to what doses to
`
`test in future studies, let alone what doses would be efficacious.
`
`41. Dr. Benet presents two calculations for recomputing the data reported
`
`in the Kappos Presentation. In his first calculation (“the subtraction analysis”), he
`
`subtracts the mean number of Gd+ lesions at baseline from the mean total number
`
`of new Gd+ lesions counted during weeks 12-24 (i.e., the primary endpoint). (Ex.
`
`1003 at ¶¶ 172-174.) In his second calculation, he averages the total mean number
`
`of new Gd+ lesions measured during weeks 12-24 (by four scans at week 12, 16,
`
`20, and 24, respectively) of Biogen’s Phase II trial to a single-scan average by
`
`dividing by four. (Ex. 1003 at ¶¶ 175-177.) He then divides this single-scan
`
`average for each dosing group by the mean number of baseline Gd+ lesions for
`
`each respective dosing group. (Ex. 1003 at ¶¶ 175-177.) In my opinion, neither
`
`approach is a reliable method of recalculating the study results, even assuming a
`
`- 19 -
`
`Page 22 of 77
`
`

`

`
`
`“correction” for baseline lesions is ap