`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`_____________________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`_____________________________
`
`
`MYLAN PHARMACEUTICALS INC., TEVA PHARMACEUTICALS USA,
`INC. and AKORN INC.,1
`Petitioners,
`
`v.
`
`ALLERGAN, INC.
`Patent Owner.
`
`_____________________________
`
`Case IPR2016-01127 (US 8,685,930 B2)
`Case IPR2016-01128 (US 8,629,111 B2)
`Case IPR2016-01129 (US 8,642,556 B2)
`Case IPR2016-01130 (US 8,633,162 B2)
`Case IPR2016-01131 (US 8,648,048 B2)
`Case IPR2016-01132 (US 9,248,191 B2)
`_____________________________
`
`DECLARATION OF DANIEL A. BLOCH, PH.D
`
`
`1 Cases IPR2017-00576 and IPR2017-00594, IPR2017-00578 and IPR2017 -
`00596, IPR2017 -00579 and IPR2017-00598, IPR2017 -00583 and IPR2017-
`00599, IPR2017-00585 and IPR2017-00600, and IPR2017 -00586 and IPR2017-
`00601, have respectively been joined with the captioned proceedings. The word-
`for-word identical paper is filed in each proceeding identified in the caption
`pursuant to the Board’s Scheduling Order (Paper 10).
`
`MYLAN - EXHIBIT 1040
`Mylan Pharmaceuticals Inc. et al. v. Allergan, Inc.
`IPR2016-01127, -01128, -01129, -01130, -01131, & -01132
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`
`
`TABLE OF CONTENTS
`
`I.
`
`QUALIFICATIONS ........................................................................................... 1
`
`II.
`
`SCOPE OF WORK ............................................................................................ 4
`
`III.
`
`PHASE 2 CLINICAL TRIAL ............................................................................... 5
`
`IV. PHASE 3 CLINICAL TRIAL ............................................................................. 11
`
`V.
`
`PHARMACOKINETIC STUDIES ........................................................................ 22
`
`VI. CONCLUDING STATEMENTS .......................................................................... 35
`
`VII. APPENDIX - LIST OF EXHIBITS ...................................................................... 36
`
`
`-i-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`I.
`
`I, Daniel A. Bloch, Ph.D., declare as follows:
`
`QUALIFICATIONS
`
`1.
`
`I am an Emeritus Research Professor in the Department of Health
`
`Research and Policy, Division of Biostatistics at Stanford University. I obtained a
`
`Ph.D. in mathematical statistics in 1967 from Johns Hopkins University. I began
`
`my career in biostatistics as a Research Fellow in the department of
`
`biomathematics at Cornell University Medical School in 1967 and as a Research
`
`Associate at the Sloan-Kettering Institute for Cancer Research in New York. I
`
`became an Assistant Professor of biomathematics at the Cornell University
`
`Medical School in 1969 and became an Assistant Professor of Mathematics at
`
`California’s Sonoma State University in 1970.
`
`2.
`
`In 1984 I became the Head Biostatistician for the Arthritis,
`
`Rheumatism, and Aging Medical Information System (ARAMIS) and for the
`
`Stanford Arthritis Center (SAC) at Stanford University. I was appointed Associate
`
`Professor in 1993 and have been a full professor of Health Research and Policy,
`
`Division of Biostatistics since 2001. I became Emeritus in 2007.
`
`3. My primary responsibilities include application of mathematical
`
`statistics to scientific studies and to advance biostatistical research methodology.
`
`Before becoming Emeritus, my salary support at Stanford was fully funded
`
`through research grants in various fields of medicine, with grants covering a broad
`
`1
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`range of topics, including rheumatic diseases, substance abuse, sleep apnea, basic
`
`immunology, juvenile arthritis, vascular structure and stem cell transplantation as
`
`immunotherapy.
`
`4.
`
`I have published more than 200 original articles in peer-reviewed
`
`journals and have additional articles either in press or recently submitted for
`
`publication. Approximately 170 of these articles have appeared in medical
`
`journals, mostly co-authored with medical faculty at Stanford and at other
`
`universities in the United States, Canada and Europe. Often these collaborations
`
`have afforded me the opportunity to explore the application of recently developed
`
`statistical methods by illustrating their use on real data, and to explore the
`
`development of new statistical methods.
`
`5.
`
`Among the several dozen articles I have published devoted to
`
`statistical methods, topics include efficiency and unbiasedness of estimators,
`
`sample size estimation, kappa statistics, non-parametrics, methods of assessing
`
`multi-parameter endpoints, and, most recently, designing Phase II and Phase III
`
`trials to estimate rare events. I believe that recognition of my expertise as a
`
`statistical researcher and at-large applicator to diverse fields of medicine were
`
`behind a Nobel Prize committee’s invitation to me to provide nominations for the
`
`Nobel Prize in Medicine.
`
`-2-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`6.
`
`I have extensive teaching experience, having taught courses offered
`
`by both the Mathematical Statistics and the Health Research & Policy departments
`
`at Stanford. At Stanford I have supervised the statistical efforts of numerous post-
`
`doctoral fellows and medical students. I have also been invited to give dozens of
`
`biostatistical presentations, in such diverse settings as Harvard, Johns Hopkins
`
`University, the Mayo Clinic, the FDA Center for Biologics Evaluation and
`
`Research, the College of Problems for the Drug Dependence annual meetings and
`
`at Institute of Mathematical Statistics meetings.
`
`7.
`
`Since 1987, I have been a consultant on an ad hoc basis to
`
`pharmaceutical and biotechnical firms, including both start-up and established
`
`companies. Herein I participate in all aspects of applying statistics to implement
`
`investigational plans, e.g., for protocol development, design of trials, data base
`
`design, analysis of data, and reporting results. I have served on numerous data
`
`safety monitoring boards and have been a member of the FDA Statistical Advisors
`
`Panel.
`
`8.
`
`A significant focus of mine has been the use of statistics to plan,
`
`conduct and evaluate experimental research, particularly medical research. I have
`
`extensive experience in reviewing published papers. At the request of an editor, my
`
`review deals with evaluating all aspects of how the author planned and applied
`
`statistics to meet study objectives. Peer-reviewed journals for whom I have
`
`-3-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`reviewed manuscripts include but are not limited to: The New England Journal of
`
`Medicine, The Journal of the American Medical Association, Lancet, The Archives
`
`of Internal Medicine, The American Journal of Medicine, Arthritis and
`
`Rheumatism, The Journal of Rheumatology, The Journal of Clinical Epidemiology,
`
`Medical Care, Cancer, Radiology, Osteoarthritis and Cartilage, Addiction, The
`
`Journal of Vascular Surgery, Statistics in Medicine, Biometrics, and The Journal
`
`of the American Statistical Association.
`
`9. My education, academic and professional appointments, a listing of
`
`my prior deposition and trial testimony over the past four years, and publications
`
`are set forth in my curriculum vitae. EX1043.
`
`II.
`
`SCOPE OF WORK
`
`10.
`
`I have been retained by the Petitioner to provide statistical analyses
`
`for certain data reported in Stevenson, Sall Figures 1-2 and to provide statistical
`
`analyses for Allergan’s animal PK studies testing cyclosporine ophthalmic
`
`emulsions as used by Dr. Attar in her Exhibit B to her Declaration presented to the
`
`USPTO. In forming my opinions, I relied on my knowledge, education, skills,
`
`experience, and training, in addition to the documents and materials cited in this
`
`Report. A full list of materials I considered in preparation of this Report is
`
`provided in the Appendix in Section VIII.
`
`-4-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`11.
`
`I am being compensated at the rate of $500 per hour. I am also being
`
`reimbursed for any necessary and reasonable travel expenses and other expenses
`
`incurred as a result of my involvement in this proceeding. I have no financial
`
`interest in the outcome of this matter.
`
`III. BIOSTATISTICS BACKGROUND
`
`12.
`
`Statistics refers to a body of methods for making valid scientific
`
`inferences from data about a “population” of interest. The data are but a sample of
`
`evidence from this population. Scientists use the techniques of statistics, among
`
`other reasons, to estimate the confidence with which one may draw conclusions
`
`based on the sample of evidence. Statistical inference may be thought of as the art
`
`of using the sample to confidently draw conclusions about this population. Often a
`
`conclusion is based on a data summary; for example, the difference in two sample
`
`averages (e.g., one sample from a group treated with a formulation containing
`
`0.05% cyclosporin, the other sample from a group treated with a formulation
`
`containing 0.1% cyclosporin). With a proper choice of statistical method, the
`
`scientist can estimate the chance that an observed result, for example a difference
`
`in averages, is not the result of random chance.
`
`13. Clearly a person can be more confident that an observed effect is real
`
`if the study yields a sample which is both representative of the population of
`
`interest and is large enough to capture the variation of data values in the population
`
`-5-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`of interest. If the set of observations from a group cluster closely, the sample
`
`average is a good indicator of individual observations; if, on the other hand, the
`
`observations were widely spread, the sample average would be a poor indicator of
`
`most individual observations. So the representativeness of the average depends
`
`upon the variability among sample values.
`
`14. The most widely used measure of variability is called the standard
`
`deviation. It is defined as the square root of the quantity called the variance. The
`
`standard deviation directly quantifies the degree to which individual measurements
`
`cluster near the mean. Brown, B.W. and Hollander M. “Statistics: A Biomedical
`
`Introduction.” John Wiley & Sons, N.Y. 1977 (“Brown, et al.”).
`
`15.
`
`If the number of individuals sampled is n, then the variance of a
`
`sample average is 1/n times the variance. The square root of the variance of the
`
`sample average is called the standard error; the smaller the standard error, the
`
`more confident one can be that the sample average is close to the population mean.
`
`
`
`A.
`
`Statistical Significance
`
`16. The statistical significance of the study result, for example the
`
`observed difference in two group averages, is obtained by applying a statistical
`
`test. A very important mathematical statistical principle is that, if the sample size is
`
`“large,” then the distribution of averages of like samples is nearly a normal
`
`-6-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`distribution (the so-called “bell-shaped” curve), centered at the “true” population
`
`average with standard deviation equal to the standard error.
`
`17. This is called the central limit theorem of statistics. When two
`
`averages are compared, the numerical value of T= (difference in averages) /
`
`(standard error of the difference) is the quantity that determines statistical
`
`significance. Large values of this ratio give statistical evidence that the difference
`
`between the two groups to be compared is real. This will be large if the difference
`
`in means is large compared to the standard error (the “signal to noise” ratio). If the
`
`value of “T” is not large, then all we can conclude is “maybe the result is real,” but
`
`not with the confidence desired. Brown, et al.
`
`18. The “confidence desired” is obtained when the result is “statistically
`
`significant.” If the desired confidence is 95%, by far the most common choice,
`
`then the result is statistically significant only if the observed difference in averages
`
`could possibly be the result of random chance at most 5% of the time. Through
`
`applying the statistical test (e.g., the “T” test described above) and making
`
`reference to an appropriate statistical distribution, one quantifies the achieved level
`
`of significance derived from the data (the “P-value”). If the P-value is very small,
`
`for example less than 1%, then we conclude the difference is very significant. And
`
`if P=5.0%, then the result is on the broader line of statistical significance. Brown,
`
`et al.
`
`-7-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`19. The Two Sample T-Test is used to determine whether the means of
`
`two groups are equal to each other (e.g., compare averages of two study groups
`
`receiving different treatments in the context of an experiment). The results from
`
`applying the two-sample t-test is valid if one assumes that (a) the data from each
`
`treatment group is normally distributed, or by assuming (b) the sample sizes are
`
`large enough so one can assume the averages that are being compared are,
`
`themselves, “close to” being normally distributed. Brown, et al.
`
`20.
`
`If the sample size is small and the data values themselves are not
`
`normally distributed, use of a t-test will give inaccurate P-values. In those
`
`circumstances, instead of using the two-sample t-test, the Wilcoxon 2-Sample Rank
`
`Test is usually applied. The Wilcoxon test gives exact p-values and is “non-
`
`parametric;” that is, it does not rely on normal theory.
`
` Comparability of Treatments
`B.
`
`21. The compared samples must also be properly comparable with respect
`
`to the variable being studied. This is best achieved by ensuring that two groups are
`
`similar except for the variable being studied. In that way, scientists may properly
`
`draw conclusions that the variable being studied is responsible for the difference
`
`between the compared groups.
`
`-8-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`22. For example, let’s say we want to compare the pharmacokinetic (PK)
`
`performance of two formulations over a 12 hour period having differing % castor
`
`oil, with the following compositions and study characteristics:
`
`Formulation I
`0.05% cyclosporin
`0.625% castor oil
`0.5% polysorbate 80
`Administered as single dose
`
`28.5 µL dose
`Assessed in Study I
`
`Formulation II
`0.05% cyclosporin
`1.25% castor oil
`1.0% polysorbate 80
`Administered twice daily for
`9.5 days
`50 µL dose
`Assessed in Study II
`
`Note that Formulation I and Formulation II differ with respect to amount of castor
`
`oil, amount of polysorbate 80, dosing regimen, and dose size.
`
`23.
`
`If a statistical difference is observed in a measure evaluating these
`
`formulations, it would be impossible to determine whether the difference was due
`
`to % cyclosporin, % castor oil, % polysorbate 80, the combination of cyclosporin
`
`and castor oil, the combination of the castor oil and polysorbate 80, or due to the
`
`different dosing regimens, dose size or the difference in scale of the outcome
`
`measure. Alternatively, let’s say the study involved identical formulations, except
`
`for % castor oil as in the following formulations:
`
`Formulation I
`0.05% cyclosporin
`0.625% castor oil
`1.0 % polysorbate 80
`Administered as single dose
`
`Formulation II
`0.05% cyclosporin
`1.25% castor oil
`1.0% polysorbate 80
`Administered as single dose
`
`-9-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`28.5 µL dose
`
`28.5 µL dose
`
`24.
`
`If “Study 1” and “Study 2” had identical investigational protocols, or
`
`if the comparison was made with a single study, then if one concluded that those
`
`two formulations resulted in a statistically significantly different performance with
`
`respect to outcome, it would be scientifically reasonable to conclude that the
`
`difference was due to percent castor oil, as that was the only difference between
`
`the two formulations (assuming, of course, that the study was properly designed
`
`and implemented).
`
`25. By comparing two groups that are identical, except for the variable
`
`that is being studied, scientists can attribute any statistically significant differences
`
`between the two groups as being associated with the single difference between the
`
`two groups. However, if there are multiple variables that differ between the two
`
`groups to be compared, scientists cannot easily attribute any statistically significant
`
`differences between the two groups.
`
`IV. PHASE 2 CLINICAL TRIAL
`
`26.
`
`I have been asked to evaluate the results presented in Figures 1-5 of
`
`Stevenson. I note that Stevenson does not provide data tables of the raw values
`
`graphed in each of Figures 1-5, however this information can be gleaned through
`
`precise measurements of the y-axis and bars within each bar graph. Stevenson also
`
`discloses the number of patients in each treatment group. EX1015 (“Stevenson”) at
`
`-10-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`970, Table 1; see also id. at 969 (identifying “the efficacy analysis presented here
`
`is confined to the evaluation of th[e] moderate-to-severe subgroup.”). However,
`
`Stevenson does not provide, either in table or graphical form, the error associated
`
`with each formulation’s reported “mean change.”
`
`27. As noted by Allergan’s expert Dr. Sheppard, “[a] statistical analysis
`
`depends on a lot more than arithmetic. It depends on the numbers that are fed in
`
`and all of these parameters, you know, P values and ranges and significance and all
`
`these numbers that statisticians apply to numbers.” EX1037 at 181:16-22 (and
`
`continuing “we clinicians, who are not statisticians, which is the vast, vast majority
`
`of us, have to depend on a statistical analysis.” Id.at 181:23-182:2). Due to the lack
`
`of error bars, a person of ordinary skill in the art could not independently verify
`
`any of the p-values presented by Stevenson or determine whether one formulation
`
`outperformed the other. Thus, the only statistical analysis available from the
`
`Stevenson reference is that reported by Stevenson. Importantly, Stevenson does
`
`not disclose any statistically significantly different result between the 0.05% and
`
`0.10% CsA formulations in any of the objective efficacy measures during the
`
`treatment period. EX1015.
`
`V.
`
`PHASE 3 CLINICAL TRIAL
`
`28.
`
`I have been asked to evaluate the results presented in Figures 1-2 of
`
`Sall (EX1007, “Sall”). Allergan’s experts Drs. Thorsteinn Loftsson and John D.
`
`-11-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`Sheppard have made conclusions that the performance of the 0.05% CsA
`
`formulation in Figure 2 of Sall was superior to that of the 0.10% CsA formulation.
`
`EX1036 (Loftsson), 200:5-201:22; EX1037 (Sheppard), 196:21-24 (claiming
`
`“substantially superior performance by the 0.05 percent concentration” in “Figure
`
`2 of Sall”), 246:2-20 (claiming “very different activity between the two
`
`compositions” appears “clearly from the Sall article in Figure 2”).
`
`29.
`
` Both Drs. Loftsson and Sheppard admitted to not performing
`
`statistical analyses on the Sall data, as well as noted that they are not statisticians.
`
`EX1036 at 175:22; 176:18-19; 215:5-11; EX1037, 183:5-8 (“Q. You did not
`
`perform, yourself, any statistical analyses of the data reported in the Sall paper;
`
`correct? A. Correct.”). Indeed, when asked if a statistically significant finding
`
`between one formulation and the vehicle, and an absence of a similar finding for a
`
`second formulation is the same thing as a finding of statistically significant
`
`differences between the two formulations, Dr. Loftsson responded that this type of
`
`analysis was “outside the scope of my knowledge.” EX1036 at 218:16-219:4; see
`
`also, id. at 176:15-20 (“Q. And at a high level, what is the standard deviation? A. It
`
`indicates the variation, but I’m not a statistician. If you are testing me about that,
`
`asking me about that, I can’t answer that.”); 178:7-10 (does not “have an
`
`understanding as to the relationship between a standard deviation and an evaluation
`
`-12-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`of statistical significance.”); 241:19-242:9 (unable to define what a “median value”
`
`is).
`
`30. The conclusions of Drs. Loftsson and Sheppard that Sall Figure 2
`
`demonstrated a difference between the two CsA formulations is incorrect. Unlike
`
`Drs. Loftsson and Sheppard, I am skilled in the art of applying statistics to science
`
`and I have performed a statistical analysis of the data in Figures 1-2 of Sall, which
`
`Sall indicates reports results of two objective measures of dry eye disease: corneal
`
`staining and categorized Schirmer Tear Test with anesthesia.
`
`31.
`
`I am informed that Drs. Sheppard and Schiffman believed that the Sall
`
`study was not adequately statistically powered to make a pairwise comparison
`
`between the 0.05% and 0.1% CsA treatment groups. See EX1037 at 200:10-14
`
`(wherein Dr. Sheppard testified that Sall teaches the formulations comprising
`
`cyclosporin can be compared “to baseline, and that is improved. You can compare
`
`to vehicle. You are not comparing between the two groups. So it’s not statistically
`
`powered to do that.”); EX1035 at 44:2-6 (“You’re often not powering these studies
`
`to have enough power to demonstrate with a high degree of statistical significance
`
`that there are – that these differences are highly likely to be due to the drug and not
`
`chance alone.”). This is incorrect. Phase 3 consisted of two identical multicenter,
`
`randomized, double-masked, parallel-group clinical trials. EX1007 at 632. Each
`
`of the Phase 3 trials was powered separately to make statistical pairwise
`
`-13-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`comparisons between the 0.05%, 0.1%, and vehicle treatment groups. Id. at 634.
`
`Pooling the two studies’ data, as was done in Sall, only increased the power. The
`
`fact that Sall did not report any statistically significant differences between the
`
`0.05% and 0.1% formulations (as will be discussed in detail below) reflects the
`
`absence of such a finding and not, as Dr. Sheppard implicitly suggested, a lack of
`
`statistical power for this comparison.
`
`32. Unlike in Stevenson, however, the data presented in Sall Figures 1-2
`
`include error bars. Thus the monthly averages and standard errors can be
`
`measured. The sample sizes for each treatment group are contained in Sall Table 1.
`
`With this information, the two sample t-test (described in the BioStatistics
`
`Background Section) can be employed for evaluation of statistical significance,
`
`and a biostatistician would have been able to perform his or her own analysis to
`
`confirm Sall’s statistical findings, as I have done below.
`
`33. Figure 1 of Sall depicts the mean change from baseline in corneal
`
`staining.
`
`-14-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`
`
`34. Regarding this data, Sall reports that there was a statistically
`
`significant improvement from baseline in corneal staining “within all treatment
`
`groups at all follow-up visits (P < 0.001).” EX1007 at 635. Sall also notes both the
`
`0.05% and 0.1% formulations showed statistically significant improvements
`
`compared to vehicle at month 4. Id. At month 6, the 0.05% formulation is taught to
`
`give a statistically significant improvement from baseline compared to vehicle. Id.
`
`Sall also notes the 0.1% formulation showed “a trend (P=0.062) toward a
`
`significantly greater improvement” than vehicle. Id.
`
`-15-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`35. Sall does not report a statistically significant difference in
`
`performance between the 0.05% and 0.1% formulations in Figure 2. Based on my
`
`own calculations2, summarized in Table 1, below, I too have found no statistically
`
`significant differences between these two formulations at any time point.
`
`Month
`
`Value
`Description
`
`P-Value
`
`B vs A C vs A B vs C
`
`Change from Baseline
`A
`B
`C
`Vehicle
`0.05/1.25
`0.1/1.25
`.239
`.159
`-0.53
`-0.64
`-0.63
`Mean Diff
` 1
`
`
`0.06
`0.05
`0.06
`Std Error
`
`.104
`.060
`-0.48
`-0.64
`-0.63
`Mean Diff
` 3
`
`
`0.06
`0.06
`0.07
`Std Error
`
` ≤0.044 ≤0.044
`N/A
`-0.67
`-0.73
`Mean Diff
` 4
`
`
`N/A
`0.06
`0.06
`Std Error
`
` 0.008
` 0.062
`N/A
`-0.89
`-0.85
`Mean Diff
` 6
`
`
`N/A
`0.06
`0.07
`Std Error
`
`Table 1: Mean Change From Baseline in Corneal Staining – Phase 3
`
`.898
`
`.914
`
` 0.480
`
` 0.665
`
`
`36. As can be seen above, by measuring the data presented in Figure 1,
`
`the mean change from baseline and the standard error can be determined for the
`
`vehicle (A), as well as for the 0.05% (B) and 0.1% (C) formulations. The three
`
`right-most columns of the table include pairwise p-values for the 0.05%
`
`
`2 The p-values in Table 1 for “B vs A” and “C vs A” at months 4 and 6 are the
`
`values reported in Sall. Hence I did not need to measure the vehicle mean
`
`difference and standard error at these months. In the Table “N/A” means not
`
`applicable. I calculated the p-values for the remaining comparisons that were not
`
`reported in Sall.
`
`-16-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`formulation versus vehicle (B vs A), the 0.1% formulation versus vehicle (C vs A),
`
`as well as the 0.05% formulation versus the 0.1% formulation (B vs C). As
`
`evidenced by Sall, the P-values in the comparisons of cyclosporin formulations to
`
`vehicle (B vs A, P≤ 0.044; C vs A, P ≤ 0.044), there is a statistically significant
`
`improvement from baseline compared to vehicle at month 4.
`
`37. Similarly, as evidenced by Sall, the 0.05% formulation showed a
`
`statistically significant improvement from baseline compared to vehicle at month 6
`
`(B vs A, P = 0.008); however, this statistically significant improvement was not
`
`observed in comparing the 0.1% formulation to vehicle (C vs A, P=0.062),
`
`although Sall reports this value as a “trend” toward significance. EX1007 at 635.
`
`38. Finally, a comparison of the two cyclosporin formulations against one
`
`another (B vs C) showed no statistically significant differences between the 0.05%
`
`and 0.1% formulations at any time point. In fact, the P-values are never close to
`
`approaching statistical significance.
`
`39. Figure 2 of Sall depicts the mean change from baseline in a second
`
`objective measure (categorized Schirmer Scores (with anesthesia)).
`
`-17-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`
`
`40. Drs. Loftsson and Sheppard base many of their conclusions on the
`
`above figure; however, both confirmed that they had not performed a statistical
`
`analysis of the data therein. EX1036 at 215:5-11; EX1037 at 183 (“Q. And with
`
`respect to Figure 2 of Sall, you have not evaluated whether there is a statistically
`
`significant difference in the Schirmer tear test with anesthesia values at month
`
`three between the 0.05 and 0.1 percent cyclosporin A treatment groups; correct? A.
`
`…Again, I have not performed any statistical analyses of this data myself.”); id. at
`
`184 (“I did not perform any statistical analysis of this data, nor do I have the data
`
`to do so.”). Again, when asked if he had an opinion as to “whether the studies
`
`-18-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`reported in Sall Figure 2” were “adequately statistically powered to detect a
`
`statistical difference between the 0.05 and the 0.10 percent CsA formulations,” Dr.
`
`Loftsson responded, “Again, I have to tell you that I am not qualified to answer
`
`this question. I am not a specialist in statistics.” EX1036 at 222:16-25.
`
`41.
`
`It is important to note what values are actually being reported in Sall
`
`Figure 2. Figure 2 represents the Categorized Schirmer values measured with
`
`anesthesia. Categorized Schirmer values differ from raw Schirmer values. Raw
`
`Schirmer values are measured in mm/5minutes. EX1007 at 632. Categorized
`
`Schirmer values are obtained by converting raw Schirmer values using a
`
`Categorized scale as follows:
`
`Raw Schirmer
`Score
`
`< 3 mm / 5
`min
`
`3-6 mm / 5
`min
`
`7-10 mm / 5
`min
`
`11-14 mm /
`5 min
`
`> 14 mm / 5
`min
`
`Categorized
`Schirmer Value
`
`1
`
`
`
`2
`
`3
`
`4
`
`5
`
`Id. at 633. Sall does not report raw Schirmer values, only categorized values. Any
`
`conclusions based on these categorical values are one step removed from
`
`conclusions that can be reached about raw values. The uncertainty of conclusions
`
`reached about a comparison between two categorical values is compounded when
`
`the average difference is only a fraction of one categorized scale unit.
`
`42.
`
`I note that Sall does not report a statistically significant pairwise
`
`comparison between 0.05% and 0.1% at month 3. Rather, Sall teaches that the
`
`-19-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`0.05% formulation showed statistically significantly greater improvement
`
`compared to vehicle (P = 0.009). Id. at 635. As evidenced by my analysis
`
`performed for Figure 1, it would be incorrect to make conclusions about the
`
`relative efficacy of 0.05% and 0.1% simply because 0.05% performed statistically
`
`better than vehicle and 0.1% did not at month 3.
`
`43.
`
`I performed the pairwise comparisons3 in Table 2, below, to confirm
`
`whether there is any statistically significant difference between the 0.05% and
`
`0.1% formulations at month 3. I have done the same analysis for the data provided
`
`for month 6, for which Sall notes that “there was a statistically significant
`
`improvement from baseline within both CsA groups. Moreover, the changes in
`
`Schirmer values in each of the CsA groups was significantly better than in the
`
`vehicle group (P < -0.007).” Id. at 635.
`
`Month
`
`Value
`Description
`
` 3
`
`Mean Diff
`
`Change from Baseline
`A
`B
`C
`Vehicle
`0.05/1.25
`0.1/1.25
`-0.24
`0.09
`-0.10
`
`P-Value
`
`B vs A C vs A B vs C
`
`0.009 0.246
`
`0.115
`
`
`3 The p-values in Table 2 for “B vs A” at months 3 and 6 and “C vs A” at
`
`month 6 are the values reported in Sall. Hence I did not need to measure the
`
`vehicle mean difference and standard error at month 6. In the Table “N/A” means
`
`not applicable. I calculated the p-values for the remaining comparisons that were
`
`not reported in Sall.
`
`-20-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`
` 6
`
`Std Error
`Mean Diff
`
`0.08
`N/A
`
`0.08
`0.39
`
`0.09
`0.26
`
`
`
`
`<0.007 <0.007 0.251
`
`
`
`0.08
`0.08
`N/A
`Std Error
`
`Table 2: Mean Change From Baseline in Categorized Schirmer Values
`Measured With Anesthesia – Phase 3
`
`
`
`44. As can be seen above, by measuring the data presented in Figure 2,
`
`the mean change from baseline, as well as the standard error can be determined for
`
`the vehicle (A), as well as for the 0.05% (B) and 0.1% (C) formulations. The three
`
`right-most columns of the table include pairwise p-values for the 0.05%
`
`formulation versus vehicle (B vs A), the 0.1% formulation versus vehicle (C vs A),
`
`as well as the 0.05% formulation versus the 0.1% formulation (B vs C). As
`
`evidenced by Sall, by the P-value in the comparison of the 0.05% formulation to
`
`vehicle (B vs A, P = 0.009), the 0.05% formulation showed a statistically
`
`significant improvement from baseline compared to vehicle at month 3.
`
`45. Similarly, as evidenced by Sall, both the 0.05% and 0.1%
`
`formulations showed statistically significant improvements from baseline
`
`compared to vehicle at month 6 (B vs A, P < 0.007; C vs A, P <0.007).
`
`46. Finally, a comparison of the two cyclosporin formulations against one
`
`another (B vs C) showed no statistically significant differences at either time point
`
`(B vs C, month 3, P = 0.115; B vs C, month 6, P = 0.251). The findings in Table 2
`
`are consistent with the fact that Sall does not report a statistically significant
`
`difference between the two formulations.
`
`-21-
`
`PROTECTIVE ORDER MATERIAL
`
`
`
`
`
`47. Dr. Sheppard testified that Sall teaches “no statistically significant
`
`differences among the groups,” at baseline, explaining that “that tells us something
`
`very important, and that is, that based upon this analysis, the three groups were the
`
`same when they started, and that’s pretty important to a clinical trial.” EX1037 at
`
`178: 19-25; See also id. at 179 (“So that range is statistically insignificant. So
`
`they’re the same.”). By the same token, the lack of statistically significant
`
`difference between the cyclosporin formulation groups in Sall Figures 1 and 2
`
`again tells us something very important – the two formulations performed the
`
`same.
`
`VI. PHARMACOKINETIC STUDIES
`
`48.
`
`I attended the deposition of Dr. Mayssa Attar. I understand from
`
`attending her deposition that she used data from two specific pharmacokinetic
`
`studies to generate Exhibit B of the Declaration she submitted to the patent office
`
`during patent prosecution of the patents at issue in these proceedings: PK-98-074
`
`(EX2027) and PK-00-163 (EX1027