`
`CENTER FOR DRUG EVALUATION AND
`RESEARCH
`
`
`APPLICATION NUMBER:
`204063Orig1s000
`
`STATISTICAL REVIEW(S)
`
`
`
`
`
`
`
`
`
`Department of Health and Human Services
`Food and Drug Administration
`Center for Drug Evaluation and Research
`Office of Translational Sciences
`Office of Biostatistics
`
`S TAT I S T I C A L R E V I E W A N D E VA L U AT I O N
`CARCINOGENICITY STUDIES- ADDENDUM
`
`NDA:
`Drug Name:
`Indications:
`Applicants:
`
`204063
`BG00012 (Dimethyl Fumarate) delayed action capsule
`Multiple Sclerosis
`Sponsor: Biogen Idec, Inc.
` 14 Cambridge Center, Cambridge, MA 02142
`
`To Reviewer: 16 January 2013
`Date:
`Standard
`Review Priority:
`Division 6
`Biometrics Division:
`Steve Thomson
`Statistical Reviewer:
`Concurring Reviewers: Karl Lin, Ph.D.
`Medical Division:
`Neurology Products
`Toxicologist Team:
`Melissa Banks-Muckenfuss, Ph.D.
`Project Manager:
`Nicole Bradley, Pharm.D.
`Keywords: Carcinogenicity, Cochran-Armitage test, Poly-k test, Trend test
`
`
`
`
`
`
`
`
`
`
`
`
`
`Reference ID: 3264435
`
`(b) (4)
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Addendum Biogen Idec, Inc.
`
`
` 4
` 2
` 4
` 3
` 3
` 4
` 1
`
` 0
` 0
` 0
` 2
` 0
` 0
` 0
`
`High
`
` 0
` 2
` 1
` 3
` 3
` 4
` 1
`
`High-
`Med
` 1
` 0
` 2
` 5
` 4
` 2
` 0
`
`The original submissions for this carcinogenicity study involved two standard two year
`studies, in rats and mice.
`
`. The study report and data for the rat study were originally submitted with
`IND 73061 from the same Sponsor. The results of the FDA analysis of the rat study are
`summarized in the statistical carcinogenicity review dated 22 April 2008. The data for the mice
`were submitted later, and the data for both species were reanalyzed in a review dated 28
`September 2012.
`
`After completion and posting of these reviews the Sponsor submitted new data for the
`renal data in both rats and mice. The table below indicates the changes in reported tumor
`incidence:
`
`Table Add 1. Kidney Tumor Incidence (Adenomas and Carcinomas Only)
`
`Original Incidence
`Revised Incidence
`Animal-Tumor
`High-
`Veh-
`Low Med-
`High Veh-
`Low Med-
`Med
`icle
`ium
`icle
`ium
`Male Rats – Adenomas
` 0
` 1
` 0
` 1 1
` 1
`Female Rats– Adenomas
` 1
` 0
` 1
` 0
` 0
` Carcinomas
` 0
` 0
` 0
` 0
` 2
`Male Mice – Adenomas
` 1
` 0
` 0
` 5
` Carcinomas
` 0
` 2
` 0
` 2
` 4
`Female Mice–Adenomas
` 0
` 0
` 0
` 0
` 2
` Carcinomas
` 0
` 0
` 0
` 0
` 0
`
`All these changes in incidence tend to reduce any apparent dose related indication of
`carcinogenicity in kidneys. This caused some concern to this reviewer, but the toxicology
`reviewer noted that: “An expert in rat renal histopathology conducted a re-evaluation of the
`original renal findings in the carcinogenicity assays. His re-evaluation was conducted to identify
`the nature of the reported lesions, taking into account histopathological events that would yield
`information about the mode of renal tumor formation. In his analysis of the rat renal data, he
`discovered that three of the identified renal tumors were of non-renal origin and that a number of
`others were of a particular morphology now known to be of spontaneous origin in rats. These
`alterations in the interpretation of the findings produced a substantially different renal tumor
`incidence in rats.” (personal communication)
`
`For incidence only data, the typical analysis is based on a so-called Cochran-Armitage
`test of trend, which basically does a regression type analysis of the incidence of the event under
`study, in this case the development of a specific organ tumor combination, regressed on dose.
`Each animal at each dose level is treated as being equally likely to develop the tumor. But in
`practice some animals die early and it may not be appropriate to consider those animals as
`having the same chance of developing the tumor as those animals in the same dose group that
`survive to the end of the study. The usual FDA analysis for carcinogenicity uses the so-called
`poly-k modification of the Cochran-Armitage test of trend where the risk set for the specific
`organ tumor combination is reduced by animals that die early in the study without the tumor..
`
`2
`
` 0
` 0
` 0
` 2
` 0
` 0
` 0
`
`Reference ID: 3264435
`
`(b) (4)
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Addendum Biogen Idec, Inc.
`
`Inspecting a large number of studies, Bailer and Portier (1988) noted that survival time seemed
`to fit a Weibull distribution, generally with a shape parameter of between 1 and 5, with 3 a
`typical value. With tmax denoting the maximal time to terminal sacrifice and tobs the time to
`death of the animal, they proposed weighting the animal by (tobs/tmax)k, so that an animal that
`survives for say 52 weeks in 104 week study without the tumor being analyzed is counted as
`(1/2)k of an animal when computing the size of the risk set for that tumor. For k = 3, that means
`that particular animal would count as 1/8 of an animal in the Cochran-Armitage analysis of that
`tumor. Further, the k = 3 specification seems to represent tumor incidence where some animals
`are perhaps more sensitive and respond earlier to the insult than the remaining animals. Under
`this structure, time to incidence would tend to follow a cubic expression. Thus an animal with
`the specific tumor being studied or who survives to terminal sacrifice without the tumor will be
`given a weight of 1 when counting the number of animals at risk. However, animals that die
`early without the tumor are down weighted when counting the number of animals in the risk set
`for that specific tumor. With differential mortality, this can mean a substantial reduction in the
`size of that risk set. This seems to be an appropriate adjustment whenever there is differential
`mortality across dose groups. The report of the Society of Toxicological Pathology “town hall”
`meeting in June 2001 recommended the use of this poly-k modification of the so-called Cochran-
`Armitage tests of trend.
`
`As transmitted to this reviewer the data consisted only of revised tumor incidence without
`the attendant mortality data. Under some circumstances, this could present a problem, and
`requesting original data was considered. However, it seems clear that in this case imputing a few
`new values will have no particular effect upon conclusions. In particular, data values were
`recoded as discussed below:
`
`1. In the original data, the high dose group in male rats included four rats identified with
`adenomas in the kidneys. The Sponsor’s reanalysis deleted all four of these adenomas. Note
`that this change is unique in the sense that we know exactly which animals should have the
`tumor incidence changed, and hence the computed p-values would be remain the same if the
`Sponsor had provided the corrected data.
`
`2. In female rats, one adenoma was added to control group totals, the tumors assigned to the
`original two identified carcinomas in the high medium dose group were deleted (animals E660
`and E663) as in 1. above. Further, three of the four carcinomas in the high dose group were
`deleted. Since animals that die early are downweighted, and thus reduce the size of the risk set,
`choosing a control group animal that survived to the end of the study will have the smallest
`effect. This was done for the single imputed adenoma in the control. Among the four animals
`originally identified with carcinomas in the high dose group, retention of the highest surviving
`animal will be least favorable for the Sponsor and thus is used here. Any other choice would
`increase the size of the risk set, thus decreasing statistical significance.
`
`3. In male mice, one adenoma was added to the control group. An animal with the least effect
`on the overall risk set was selected (animal 43 with survival to day 735). Again, this choice
`would be least favorable to the Sponsor.
`
`
`3
`
`Reference ID: 3264435
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Addendum Biogen Idec, Inc.
`
`
`4. In female mice, the tumor reanalysis had no effect on tumor incidence and hence results are
`unchanged.
`
`
`Using these imputed values allows one to use the poly-k methodology and adjust for
`differences in mortality. Note that in cases 2. and 3. above, these imputations will increase
`uncertainty and thus actual variance in results. But it was felt that this increase in “noise” would
`be difficult to assess, should be conservative, and did not warrant actual adjustment of results.
`
`
`In the following tables, for each species by gender the number of animals analyzed and
`used in the statistical tests is presented first. The entry for each tumor is preceded by the
`adjusted number of animals at risk for that tumor. It seems clear that an animal that dies early
`without a tumor reduces the size of the risk set for that getting that particular umor. The poly-k
`test down weights such animals, and the sum of these poly-k weights seems to be a better
`estimate of the number of animals at risk of getting that tumor. This sum is given in the row
`labeled “Adjusted # at risk”. Tumor incidence is presented next, with the significance levels of
`the tests of trend, and the results of pairwise tests between the high, high-medium, low-medium,
`and low dose groups versus vehicle. For this analysis, incidence in the vehicle group is used to
`assess background tumor incidence, and thus whether a tumor is considered to be rare
`(background incidence <1%) or common. Finally, the significance level of the original
`Cochran-Armitage test of trend over all five dose groups, with no poly-k adjustment for
`mortality, is presented under the heading for trend, below the corresponding poly-k trend result.
`Please note that the poly-k results are strongly recommended, but the Cochran-Armitage test
`does not depend upon the imputed incidence cited in 2. and 3. above.
`
`To adjust for the multiplicity of tests the so-called Haseman-Lin-Rahman (HLR) rules are
`often applied. That is, when testing for trend over dose groups and the difference between the
`highest dose group with a control group, to control the overall Type I error rate to roughly 10%
`for a standard two species, two sex study, one compares the unadjusted significance level of the
`trend test to 0.005 for common tumors and 0.025 for rare tumors, and the pairwise test to 0.01 for
`common tumors and 0.05 for rare tumors. Using these adjustments for other tests, like testing
`the comparisons between the low, medium, and medium-high dose groups versus vehicle can be
`expected to increase the overall type I error rate to some value above the nominal rough 10%
`level, possibly considerably higher than the nominal 10% rate.
`
`Table Add. 2. Kidney Tumors in Rats
`Gender/Tumor Incidence Significance Levels
` Med- Hi vs MedHi Medvs Low
` Veh Low ium MedHi High Trend Veh vsVeh Veh vsVeh
`Male Rats
` # Evaluated 75 75 75 75 75
` Adjusted # at risk 49.6 48.6 40.7 25.9 22.2
` RENAL TUBULE- ADENOMA 0 0 1 1 0 .3026 . .3378 .4494 .
` Cochran-Armitage test
`
`
` .4802
`
`
`
`
`4
`
`Reference ID: 3264435
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Addendum Biogen Idec, Inc.
`
`Table Add. 2. (cont.) Kidney Tumors in Rats
`Gender/Tumor Incidence Significance Levels
` Med- Hi vs MedHi Medvs Low
` Veh Low ium MedHi High Trend Veh vsVeh Veh vsVeh
`Female Rats
` # Evaluated 75 75 75 75 75
` Adjusted # at risk 54.9 53.2 48.7 48.1 52.0
` RENAL TUBULE- ADENOMA 1 0 1 0 1 .5055 .7429 1 .7222 1
` Cochran-Armitage test .5200
`
` Adjusted # at risk 54.9 53.2 48.7 48.1 52.0
` Overall Adenoma 1 0 1 0 2 .2330 .4857 1 .7222 1
` Cochran-Armitage test .2422
` Adjusted # at risk 54.8 53.2 48.7 48.1 51.5
` RENAL TUBULE-CARCINOMA 0 0 0 0 1 .2008 .4857 . . .
` Cochran-Armitage test
` .2000
` Adjusted # at risk 54.9 53.2 48.7 48.1 52.0
` Overall Adenoma/Carc. 1 0 1 0 3 .0923 .2947 1 .7222 1
` Cochran-Armitage test
` .0969
`
`
`In either gender in rats, none of these tests in kidneys achieved the nominal 0.05 level of
`statistical significance, let alone any of the the multiplicity adjusted HLR levels. One female rat
`in the high dose group was identified with a “Pelvis- Transitional Epithelium Adenoma” in the
`kidney. It is added to the overall adenomas, but the latter still shows no statistically significant
`tests of trend or pairwise differences.
`
`Table Add. 3. Kidney Tumors in Mice
`Gender/Tumor Incidence Significance Levels
` Med- Hi vs MedHi Med vs Low
` Veh Low ium MedHi High Trend Veh vs Veh Veh vs Veh
`Male Mice
` # Evaluated 75 75 75 75 75
` Adjusted # at risk 47.5 52.1 50.6 48.7 22.1
` Adenoma 2 2 0 5 3 .0163 .1811 .2264 1 .7286
` Cochran-Armitage test
` .1373
` Adjusted # at risk 47.5 52.1 50.7 47.7 22.0
` Carcinoma 0 0 2 4 3 .0015 .0265 .0585 .2631 .
` Cochran-Armitage test
`
` .0280
` Adjusted # at risk 47.5 52.1 50.7 49.3 22.8
` Adenoma/Carcinoma 2 2 2 8 5 .0008 .0298 .0525 .7145 .7286
` Cochran-Armitage test
`
` .0407
`Female Mice
` # Evaluated 75 75 75 75 75
` Adjusted # at risk 56.6 52.9 52.1 51.8 29.1
` Adenoma 0 0 0 2 4 .0002 .0117 .2248 . .
` Cochran-Armitage test
`
` .0024
` Adjusted # at risk 56.6 52.9 52.1 51.8 27.6
` Carcinoma 0 0 0 0 1 .1134 .3253 . . .
` Cochran-Armitage test
`
` .2000
` Adjusted # at risk 56.6 52.9 52.1 51.8 29.1
` Adenoma/Carcinoma 0 0 0 2 4 .0002 .0117 .2248 . .
` Cochran-Armitage test .0024
`
`
`
`5
`
`Reference ID: 3264435
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Addendum Biogen Idec, Inc.
`
`
`In male mice, using the HLR rules, the tests of trend in carcinoma and pooled adenoma
`and carcinoma were statistically significant (p = 0.0015 < 0.025 and p = 0.0008 < 0.005,
`respectively). Continuing in male mice, the pairwise test between the high dose and vehicle in
`carcinomas would be statistically significant (p = 0.0265 < 0.05) while the pairwise test between
`the high dose and vehicle pooled adenoma and carcinoma would be somewhat close to
`statistically significant (p = 0.0298 > 0.01). In female mice, the tests of trend in carcinoma and
`pooled carcinoma and adenoma of the kidneys were identical (since the only animal with a
`carcinoma of the kidneys also was identified with an adenoma). Both tests were statistically
`significant (both p = 0.0002 < 0.025 and p = 0.0117 < 0.025, respectively).
`
`
`
`
`6
`
`Reference ID: 3264435
`
`
`
`---------------------------------------------------------------------------------------------------------
`This is a representation of an electronic record that was signed
`electronically and this page is the manifestation of the electronic
`signature.
`---------------------------------------------------------------------------------------------------------
`/s/
`----------------------------------------------------
`
`STEVEN F THOMSON
`02/20/2013
`Addendum to Statistical Carcinogenicity Review
`
`KARL K LIN
`02/21/2013
`Concur with review
`
`Reference ID: 3264435
`
`
`
`Department of Health and Human Services
`Food and Drug Administration
`Center for Drug Evaluation and Research
`Office of Translational Sciences
`Office of Biostatistics
`
`S TAT I S T I C A L R E V I E W A N D E VA L U AT I O N
`CARCINOGENICITY STUDIES
`
`NDA:
`Drug Name:
`Indications:
`Applicants:
`
`204063
`BG00012 (Dimethyl Fumarate) delayed action capsule
`Multiple Sclerosis
`Sponsor: Biogen Idec, Inc.
` 14 Cambridge Center, Cambridge, MA 02142
`
`Date(s):
`
`Submitted: 27 February 2012
`To Reviewer: 30 July 2012
`Standard
`Review Priority:
`Division 6
`Biometrics Division:
`Steve Thomson
`Statistical Reviewer:
`Concurring Reviewers: Karl Lin, Ph.D.
`Medical Division:
`Neurology Products
`Toxicologist Team:
`Melissa Banks-Muckenfuss, Ph.D.
`Project Manager:
`Nicole Bradley, Pharm.D.
`Keywords: Bayesian analysis, Carcinogenicity, Cox regresson, Kaplan-Meier product limit,
`Survival analysis, Trend test
`
`
`
`
`
`
`
`
`
`Reference ID: 3207500
`
`(b) (4)
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Biogen Idec, Inc.
`
`
`Table of Contents
`1. EXECUTIVE SUMMARY ....................................................................................................................................3
`1.1. CONCLUSIONS AND RECOMMENDATIONS .......................................................................................................3
`1.2. BRIEF OVERVIEW OF THE STUDIES ....................................................................................................................11
`1.3. STATISTICAL ISSUES AND FINDINGS ..................................................................................................................11
`1.3.1. Statistical Issues.........................................................................................................................................11
`1.3.2. Statistical Findings ....................................................................................................................................17
`2. INTRODUCTION .................................................................................................................................................17
`2.1. OVERVIEW.........................................................................................................................................................17
`2.2. DATA SOURCES .................................................................................................................................................17
`3. STATISTICAL EVALUATION ..........................................................................................................................17
`3.1. EVALUATION OF EFFICACY................................................................................................................................17
`3.2. EVALUATION OF SAFETY ...................................................................................................................................17
`4. FINDINGS IN SPECIAL/SUBGROUP POPULATIONS ................................................................................36
`5. SUMMARY AND CONCLUSIONS ...................................................................................................................36
`5.1. STATISTICAL ISSUES AND COLLECTIVE EVIDENCE ............................................................................................36
`5.2. CONCLUSIONS AND RECOMMENDATIONS ..........................................................................................................36
`APPENDICES............................................................................................................................................................37
`APPENDIX 1. FDA SURVIVAL ANALYSIS..................................................................................................................37
`APPENDIX 2. FDA POLY-K TUMORIGENICITY ANALYSIS .........................................................................................42
`APPENDIX 3. BAYESIAN TUMORIGENICITY ANALYSIS IN RATS ................................................................................66
`APPENDIX 4. REFERENCES........................................................................................................................................70
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Reference ID: 3207500
`
`2
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Biogen Idec, Inc.
`
`1. EXECUTIVE SUMMARY
`
`Reports and data from two studies, in rats and mice, were provided.
` The study report and data
`for the rat study were originally submitted with IND 73061 from the same Sponsor. The results
`of the FDA analysis of the rat study are summarized in the statistical carcinogenicity review
`dated 22 April 2008. Although the current analysis of the rat study is slightly different, only the
`new mouse study contains completely new results.
`
`According to the mouse report provided by the Sponsors: “The purpose of this study was
`to evaluate the potential carcinogenicity of BG00012 following once daily oral gavage to CD-1
`mice for at least 104 weeks and toxicokinetics following once daily oral gavage for 180 days.”
`(page 18 of report) The objective of the rat report is expressed similarly. Each study included
`five treatment groups, as described below.
`
`1.1. Conclusions and Recommendations
`The Sponsor describes the drug vehicle as hydroxypropylmethylcellulose (HPMC) or
`hypromellose (3,500-5,600 cps), 0.8% w/v in reverse osmosis deionized water. For each study,
`in each gender, there are five treatment groups. Animals were dosed once daily by oral gavage.
`Gross aspects of the study designs for the main study animals are summarized in Tables 1 and 2
`below:
`Table 1. Design of Rat Study (75 animals per main study group/gender)
`Dosing
`Dose
`BG00012
`Treatment
`Vehicle
`Concent
`Volume
`Dosage
` Group
`or Test
`rationa
`(mL/kg)
`(mg/kg/day)
`Article
`HPMCa
` 0
` 10
` 0
`1. Vehicle
`BG00012
` 25
` 10
` 2.5
`2. Low
`BG00012
` 50
` 10
` 5
`3. Medium
` 100
`4. Med-High b BG00012
` 10
` 10
`5. High b
`BG00012
` 150
` 10
` 15
`a Hydroxypropylmethylcellulose or Hypromellose (3,500-5,600 cps), 0.8% w/v in reverse osmosis deionized water.
`b Due to mortality, the High dose group (150 mg/kg/day, Group 5) males were terminated during Week 86 and the
`Medium-High (100 mg/kg/day,Group 4) males were terminated during Week 88.
`
`Table 2. Design of Mouse Study (75 animals per main study group/gender)
`Treatment
`Vehicle
`BG00012
`Dose
`Dosing
` Group
`or Test
`Dosage
`Volume
`Concent
`rationa
`Article
`(mg/kg/day)
`(mL/kg)
`1. Vehicle a HPMC
` 0
` 10
` 0
`2. Low
`BG00012
` 25
` 10
` 2.5
`3. Medium
`BG00012
` 75
` 10
` 7.5
` 200
`4. Med-High BG00012
` 10
` 20
`5. High
`BG00012
` 600/400 b
` 10
` 60/40 b
`a Hydroxypropylmethylcellulose or Hypromellose (3,500-5,600 cps), 0.8% w/v in reverse osmosis deionized water.
`b Due to mortality observed in the high-dose group, beginning on Day 9,.the dose level for Group 5 was decreased
`from 600 mg/kg/day to 400 mg/kg/day (40 mg/mL) . For tests it is treated as dosage at 400 mg/kg/day.
`
`
`
`3
`
`Reference ID: 3207500
`
`(b) (4)
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Biogen Idec, Inc.
`
`
`More detailed descriptions of the studies are provided in Section 3.2.1 and 3.2.2 below.
`In this report the vehicle group is sometimes referred to as the “HPMC” or “control group” while
`the other dose groups are referred to as “actual dose groups”, and, purposes of assessing trend,
`the Vehicle, Low, Medium, Medium-High, and High dose groups (i.e., Groups 1-5) as “treated
`groups.” Simple summary life tables in mortality are presented in the report in these sections of
`this report. Also, because early very high mortality in the mice study, on day 9 dosage was
`reduced from 600 mg/kg/day to 400 mg/kg/day. On all tests used in the FDA analysis it is
`treated as the 400 mg/kg/day.
`
`In Appendix 1, Figures A.1.1 and A.1.2 for rats display Kaplan-Meier estimated survival
`curves for each study group for each gender. Two sets of plots are displayed for mice. Figures
`A.1.3 and A.1.4 for mice display the Kaplan-Meier estimated survival curves for gender using
`the original data while Figures A.1.5 and A.1.6 display the corresponding plots deleting those
`animals that died before day 9, because of initial high mortality. Results of tests on survival in
`rats and mice are summarized below:
`
`Table 3. Statistical Significances of Tests of Homogeneity and Trend in Survival in the Rat
`Study
`Hypothesis Tested
`
`Males a Females
`Log rank Wilcoxon Log rank Wilcoxon
` < 0.0001
` < 0.0001
` 0.4421
` 0.2737
`Rat Homogeneity over Groups 1-5
` < 0.0001
` < 0.0001
` 0.2904
` 0.1602
` Homogeneity over Groups 1-4
` 0.0292
` 0.0153
` 0.2090
` 0.1594
` Homogeneity over Groups 1-3
` < 0.0001
` < 0.0001
` 0.3427
` 0.2033
` No trend over Groups 1-5
` < 0.0001
` < 0.0001
` 0.1345
` 0.0389
` No trend over Groups 1-4
` 0.0169
` 0.0119
` 0.1386
` 0.0734
` No trend over Groups 1-3
` < 0.0001
` < 0.0001
` 0.4943
` 0.3229
` No Difference Between Groups 1 vs 5
` < 0.0001
` < 0.0001
` 0.2503
` 0.0931
` No Difference Between Groups 1 vs 4
` 0.0180
` 0.0142
` 0.1205
` 0.0613
` No Difference Between Groups 1 vs 3
`a Due to mortality, the High dose group (150 mg/kg/day, Group 5) males were terminated during Week 86 and the
`Medium-High (100 mg/kg/day,Group 4) males were terminated during Week 88.
`
`From Figure A.1.1, in male rats, the HPMC vehicle and Low dose group are largely
`intertwined, with the highest survival, while the Medium dose group has the next highest
`survival, and the Medium-High and High dose groups having the lowest but quite close survival
`over the study period. Note these groups were sacrificed early due to low survival, but in the
`analysis such deaths are considered as censored times, not deaths. This is sufficient to cause
`significant tests of lack of homogeneity, no trend, and no difference between the highest doses
`and vehicle in groups 1-4 and groups 1-5 (all 12 p < 0.0001, usually much less than the 0.0001
`level). Even the seperation of the Medium dose group from the Vehicle and Low dose is
`sufficient to result in consistently statistically significant results (all six p ≤ 0.0292). The
`situation in female rats is quite different. From Figure A.1.2, in female rats, although the HPMC
`vehicle seems to have slightly highest survival, the survival curves of the other four dose groups
`are generally quite close and are largely intertwined. While this is not strong evidence of no
`
`
`
`4
`
`Reference ID: 3207500
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Biogen Idec, Inc.
`
`differences, it is evidence of no strong differences when comparing all five groups plus the High
`dose to HPMC (all six p ≥ 0.2033). Results of the tests using treatment subgroups 1-4 and 1-3
`are primarily included to match those for male rats. Note that the Wilcoxon test is more sensitive
`to early differences in survival and the corresponding test of lack of trend is barely statistically
`significant at the usual level (i.e., Wilcoxon p = 0.0389), but this conclusion is not supported by
`the Logrank test ( p = 0.1386).
`
`Table 4. Statistical Significances of Tests of Homogeneity and Trend in Survival in the
`Mouse Study
`Hypothesis Tested
`
`Males Females
`Log rank Wilcoxon Log rank Wilcoxon
`Mouse Homogeneity over Groups 1-5 a
` < 0.0001 < 0.0001
` < 0.0001
` < 0.0001
` 0.1115
` 0.3129
` 0.4603
` 0.5208
` Homogeneity over Groups 1-4
` No trend over Groups 1-5 a
` < 0.0001 < 0.0001
` < 0.0001
` < 0.0001
` 0.1119
` 0.3181
` 0.1847
` 0.2348
` No trend over Groups 1-4
` No Difference Between Groups 1 vs 5a
` < 0.0001 < 0.0001
` < 0.0001
` < 0.0001
` 0.4876
` 0.7711
` 0.1099
` 0.1322
` No Difference Between Groups 1 vs 4
`a Printed P-value bounds are identical whether one conditions on survival to day 9 or not.
`
`Figures A.1.3 and A.1.4, in Appendix 1, display the gender specific survival curves over
`the five dose groups in mice. Note that in both genders the High dose group is clearly separated
`from the remaining dose groups. This is sufficient to result in the highly statistically significant
`tests of homogeneity, lack of trend, and no difference between the High dose and the vehicle
`dose group (for each gender all six p < 0.0001). One might speculate that the separation of the
`High dose from the other dose groups may be due solely to the early deaths in the High dose
`group, reflected in the initial drop in the Kaplan-meier curve apparent in the figures. Figures
`A.1.5 and A.1.6 in the appendix display the survival curves over the five dose groups
`conditional upon animals surviving to at least until day 9. Even with this criterion, the High dose
`remains seperated from the remaining dose groups. While the actual values of the significance
`levels for the tests of homogeneity and trend over groups 1-5, and the pairwise comparison
`between the High dose and vehicle are larger (and thus less statistically significant), the printed
`values remain the same (i.e., all tests still have all six p < 0.0001).
`
`In both sets of plots, for both genders, the survival curves for the dose groups 1-4 are
`more or less closely intertwined, but with much higher survival than that in the High dose group.
`This is consistent with the tests of survival completely deleting the High dose group, i.e., none of
`the tests of homogeneity, lack of trend, and no difference between the next highest dose and the
`Vehicle dose group were particularly statistically significant, at least at the usual 0.05 level
`(Males: all six p ≥ 0.1115, Females all six p ≥ 0.1009). This suggests that differences in survival
`are largely due to differences in the High dose from the remaining dose groups.
`
`
`Of course in a carcinogenicity study, primary interest is on the occurrence of cancers.
`The statistical analysis of tumors compares tumor incidence over dose groups. Complete tumor
`
`
`
`5
`
`Reference ID: 3207500
`
`
`
`NDA 2034063 BG00012 (Dimethyl Fumarate) Biogen Idec, Inc.
`
`incidence tables for each organ listed by the Sponsor in the submitted data sets and those
`combined by this reviewer are provided in Tables A.2.3 through A.2.6 in Appendix 3. For each
`species by gender by organ the number of animals analyzed and used in the statistical tests is
`presented first. The tumor incidence for each organ is presented next, with the significance
`levels of the tests of trend, and the results of pairwise tests between the high, medium, and low
`dose groups. These statistical tests are supposed to be conditioned on the animals actually
`evaluated, ignoring those not analyzed. In other words, animals not analyzed are treated as being
`not at risk.
`
`To adjust for the multiplicity of tests the so-called Haseman-Lin-Rahman rules discussed
`in Section 1.3.1.5, below, are often applied. That is, when testing for lack of trend in
`carcinogencity over dose and no difference between the highest dose group with a control group
`using Peto or poly-k tests, to control the overall Type I error rate to roughly 10% for a standard
`two species, two sex study, one compares the unadjusted significance level of the trend test to
`0.005 for common tumors (incidence > 1%) and 0.025 for rare tumors, and the pairwise test to
`0.01 for common tumors and 0.05 for rare tumors. As also discussed in section 1.3.1.5,
``employing these adjustments for other than the overall tren