CENTER FOR DRUG EVALUATION AND RESEARCH
`
`APPROVAL PACKAGE FOR:
`
`APPLICATION NUMBER
`
`NDA 21-574
`
`Statistical Review(s)
`
`

`

`DEPARTMENT OF HEALTH AND HUMAN SERVICES
`FOOD AND DRUG ADMINISTRATION
`CENTER FOR DRUG EVALUATION AND RESEARCH
`OFFICE OF BIOSTATISTICS
`
`Statistical Review and Evaluation
`
`CLINICAL STUDIES
`
`NDA: 21-574
`
`Name Of drug: Fortamet extended release (metformin XT)
`
`Applicant: Andrx Labs
`
`Indication: Type 2 diabetes
`
`Documents and datasets
`’eV‘ewed:
`
`\\Cdses_ub1\n21574\N 000\2003—01-17\clinstat
`\\Cdsesub1\n21574\N 000\2002-12—17\crt\datasets
`
`Dates: Received 12/19/02; user fee (1 0 months) 10/19/03
`
`HFD—510
`
`Project manager:
`
`Jena Weber, 8.8.
`
`Clinical reviewer: Robert Misbin, M.D.
`
`HFD—715
`
`,Team leader and J. Todd Sahlroot, Ph.D.
`statistical reviewer:
`
`Biometrics division 8. Edward Nevius, PhD.
`
`director and secondary
`‘
`reviewer:
`
`-
`
`Keywords: NDA review, clinical studies, dropouts
`
`

`

`TABLE OF CONTENTS
`
`1 Summary and conclusions
`2 introduction
`
`3 Design
`4 Baseline and demographic variables
`5 Disposition
`6 Statistical methods
`7 Results
`
`7.1 Dosing of test drug
`7.2 Efficacy
`7.2.1 HbAlc
`
`7.2.2 Fasting plasma glucose
`7.3 Comedications
`
`7.4 Compliance
`7.5 Assay sensitivity
`8 Suggestions for labeling
`
`LDCDCDmeQQ-P-AOJ
`
`AAA—LA
`
`OSGDQJNCD
`
`AppeG‘S ““5 WC”
`On Ollgl‘na‘
`
`

`

`1 Summary and conclusions
`
`The sponsor submitted data from two Phase 3, controlled trials of Fortamet
`(metformin XT or “XT”), an extended release oral anti-diabetic medication given
`once-a—day.
`
`Study 301 was a multi-center, randomized, double-blind (double-dummy), active-
`controlled clinical trial in 680 patients with type 2 diabetes. The trial comparedXT
`and Glucophage, an immediate-release oral anti-diabetic medication given twice—
`a-day.
`
`Study 302 was a multi-center, randomized, double-blind (double-dummy),
`controlled trial in 115 Type 2 patients. The primary objective was to compare the
`tolerability and safety of 2000 mg and 2500 mg of XT given once a day and the
`same dose of Glucophage given twice a day. The rationale for the study was to
`provide sufficient safety data for XT at the 2 highest doses to give 100 patients at
`each dose for both studies combined. HbA1c data were collected but the
`
`protocol stated that “efficacy will not be evaluated" because this was not the
`stated objective of the trial. Trial 302 was not reviewed for efficacy.
`
`In trial 301, XT was non-inferior to Glucophage on the primary efficacy variable,
`HbA1c change from baseline, using the pre—defined non-inferiority margin of
`0.40. Mean changes from baseline for XT and Glucphage were 0.40 and 0.14,
`respectively. The least square mean treatment difference was 0.25 (2-sided 95%
`CI = 0.14, 0.37). XT was also statistically inferior to Glucophage since the lower
`bound of the Cl excluded zero (p<.0001).
`
`One hundred twenty-five (125, 18%) randomized patients did not complete the
`trial. The ratio of these dropouts in the XT : Glucophage groups was 3 to 2.
`Although XT was shown to be non—inferior to Glucophage for the trial as a whole,
`- dropouts appeared to represent a significant subgroup of patients who were
`unable to establish diabetic control with XT. The 61 XT dropouts with on-
`treatment data had a mean HbA1c of 8.10, an increase of 0.73 over baseline.
`
`The 38 Glucophage dropouts with on—treatment data had a mean HbA1c of 7.38,
`an increase of 0.19 over baseline. The treatment difference was 0.54 for
`
`dropouts.
`
`Eighteen (18, 5%) XT patients and 8 (2%) Glucophage patients dropped out due
`to a stated lack of efficacy (p=.047). However, the poor XT response for
`dropouts was not confined to patients that dropped due to lack of efficacy but
`was also seen for patients who dropped for other reasons as well.
`
`The groups were similar with respect to study drug dosing, concomitant insulin
`and oral anti-diabetic use, and compliance. Therefore, the statistical difference
`between the groups on the primary endpoint could not be attributed to any
`imbalances between the groups in these variables.
`
`

`

`2 Introduction
`
`The sponsor submitted data from two Phase 3, active—controlled trials of
`Fortamet (metformin XT or “XT"), an extended release oral anti-diabetic
`medication given once-a-day.
`
`Study 301 was a multi-center, randomized, double—blind (double-dummy),
`controlled trial in 680 patients with Type 2 diabetes. The trial compared XT and
`Glucophage, an immediate-release oral anti-diabetic medication given twice-a-
`day. The objective of the trial was to evaluate the non—inferiority of XT compared
`to Glucophage at therapeutic doses over a 6-month period on the change from
`baseline in HbA1c. The pre—defined non-inferiority margin was 0.4 (%).
`
`Study 302 was a multi-center, randomized, double-blind (double—dummy),
`controlled trial in '115 Type 2 patients. The primary objective was to compare the
`tolerability and safety of 2000 mg and 2500 mg of XT given once a day and the
`same dose of Glucophage given twice a day. The rationale for the study was to
`provide sufficient safety data for XT at the 2 highest doses to give 100 patients at
`each dose for both studies combined. HbA1c data were collected; however, the
`
`protocol stated that “efficacy will not be evaluated” because it was not the stated
`objective of the trial. No power calculations were performed. (Thisreviewer
`calculated the power of the study to be approximately 21% to test for non—
`inferiority, given the number of patients studied and assuming the same
`parameter estimates from Study 301.) For these reasons, Trial 302 was not
`reviewed.
`
`3 Design
`
`Table 1 shows major design characteristics of trial 301.
`
`Table 1. Stud characteristics
`
`# randomized
`Design
`Duration
`Patients
`Primary endpoint
`of double
`
`
`blind period
`
`
`155-301
`M and F
`Metformin' XT QD Randomized
`6 weeks
`n=339
`ages 30—70
`titration
`
`
`
`with NlDDM 1 Glucophage BID
`47 US
`followed by
`centers
`receiving
`n=341
`20 weeks
`
`
`Glucophage.
`'
`Change from
`maintenance
`
`
`
`HbA1c 59%
`7/00 — 6/01
`baseline in HbA1c
`
`at Visit 1
`
`1 NlDDM = non insulin dependent diabetes mellitus (Type 2 diabetes)
`
`active-controlled
`double-blind
`
`

`

`Table 2 shows study visits and corresponding weeks on study.
`
`Table2. Stud visitschedule
`
`Treatment
`
`
`Patients were randomized at Visit 3 (Week 0)
`
`Patients entered the trial on Glucophage and were randomized at Visit 3 to XT or
`Glucophage. The dose of study drug (1000, 1500, 200 or 2500 mg/day) was
`determined from the prior Glucophage dose and Visit 2 fasting plasma glucose
`(FPG) level. During the first 6 weeks, the dose of XT or Glucophage was titrated
`up or down in 500 mg increments based on FPG . Patients remained on a fixed
`doSe of study drug for the last 20 weeks of the study.
`
`There were 4 types of study medication dose categories. Baseline, starting and
`final doses were 1000, 1500, 2000 or 2500 mg:
`
`.
`
`Prior dose — Glucophage dose prior to entering the trial
`
`. Baseline dose — Dose assigned at randomization based on prior
`Glucophage dose
`
`. Starting dose — Actual dose taken by the patient at randomization and
`based on Visit 2 FPG level. Equal to the baseline dose, baseline dose +
`500mg, or baseline dose — 500mg depending on FPG.
`
`.
`
`Final dose — Dose at the end of titration
`
`The primary objective of the trial was to evaluate the efficacy of XT in comparison
`to Glucophage over a 6-month treatment period.
`
`The primary efficacy variable was change from baseline in HbA1c at endpoint.
`
`Secondary efficacy variables were (1) changes from baseline (mean of Weeks —2
`and O) in HbA1c at Weeks 9, 13, 17 and 21; (2) changes from baseline (mean of
`Weeks -1 and 0) in fasting plasma glucose at Weeks 9, 13, 17, 21, 26 and
`endpoint; (3) changes from baseline (Week 0) in fructosamine and insulin at
`Weeks 9, 17, 26 and endpoint; and (4) change in dose from baseline dose to
`final dose.
`
`For all efficacy variables, endpoint was defined as the last fasting value up to 3
`days after the last dose.
`
`

`

`The sponsor calculated that 240 patients per group would give 90% power
`assuming the true difference in HbA1c change from baseline was zero, non-
`inferiority margin = 0.4, SD = 1.3% and alpha (1-sided) = 2.5%. Randomization
`was increased to 600 total patients to account for an estimated 15-20% dropout
`rate. The sample size was later amended to 700 patients (see protocol
`amendments below for more details).
`
`Protocol amendments
`
`The protocol was amended twice.
`
`Amendment 1 (dated December 18, 2000, approximately midway through the
`trial) increased the number 'of randomized patients from 600 to 700 “in order to
`achieve a sufficient number of evaluable patients". The actual number of desired
`evaluable patients (240 / group) did not change in the amendment. The sponsor
`did not state a rationale for the amendment, or what type of patient was
`considered as evaluable.
`In the statistical analysis, the sponsor designated 542
`patients as per-protocol.
`
`Amendment 2 (dated August 24, 2001) clarified the definitions of baseline and
`endpoint for all efficacy variables. The primary endpoint definition was changed
`from the Week 26 value to the last observation on treatment. Baseline HbA1c
`was defined as the mean of Visits 1 and 3 in the protocol and as the mean of
`Visits 1 and 2 in the statistical analysis section. The definition was clarified as
`the mean of Visits 1 and 3 and extended in the event that one or more of the Visit
`
`1 or 3 data were missing.
`
`4 Baseline and demographic variables
`
`Race, age, weight, height and BMI patient characteristics were similar between
`groups for all randomized patients. The mean age was 57 years. The mean
`weight was 94 kg and mean BMI was 31 kg/mz. 75% of patients were
`Caucasian, 14% Hispanic. There was a nominal imbalance (p=.029) in gender
`(XT, 63% males; Glucophage, 55% males) which did not translate into between-
`treatment differences in weight, height or BMl.
`
`5 Disposition
`
`Table 3 shows the number of patients on study at various time points during the
`study. On-study time is defined by the time of the last HbA1c measurement.
`82% of patients completed the trial. About 50% more patients discontinued in
`the XT group (n=76) compared to the Glucophage group (n=49).
`
`

`

`
`Table 3. Number of oatients b weeks on stud
`
`
`Last week on stud
`Metformin XT
`Gluco-ohae
`
`341 100%)
`680 (100%)
`
`332 97%
`659 97%
`
`
`
`
`
`321 (94%)
`632 (93%)
`316 93%
`614 90%
`
`339 100%
`327 96%
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Reviewer's ITT
`327 96%
`
`
`
`
`On-study time is defined by time of last HbA1c measurement.
`2 Sponsor's designation
`3 Same as sponsor‘s sensitivity population
`
`
`
`
`
`
`659 (97%)
`
`Table 4 shows numbers of dropouts and reasons for dropout. More XT patients
`discontinued for lack of efficacy than did Glucophage patients (n=18 vs n=8,
`p=.047). XT Patients withdrew consent at almost double the rate as Glucophage
`patients (7% vs 4%). Patients were not asked by the sponsor to state why they
`withdrew consent.
`
`Withdrawals due to AE’s were similar in the groups.
`
`
`Table 4. Numbers of dropouts and reason for droout
`
`Reason
`
`Metformin XT
`
`Glucophage _
`
`15 (4%)
`13 (4%)
`4 (1%)
`1 (<1%)
`3 (1 %)
`8 (2%)
`
`32 (5%)
`35 (5%)
`10 (2%)
`3 (<1%)
`7 (1 %)
`26 (4%)
`
` Medication non-compliance
`
`AE
`Consent w/d
`Lost to F/U
`
`'
`
`17 (5%)
`22 (7%)
`6 (2%)
`2 (1 %)
`4 (1 %)
`Protocol violation
`18 (5%)
`Lack of efficacy
`
`2%
`Other
`
`76 (22%)
`49 (14%L
`
`

`

`6 Statistical methods
`
`The sponsor compared treatment groups on HbA1c change from baseline using
`ANCOVA with treatment group and center as main effects and baseline HbA1c
`as a covariate. Non—inferiority was assessed. using a 97.5% one-sided Cl (upper
`bound only) for the difference in least square means (XT change from baseline
`minus Glucophage change from baseline) and a non-inferiority margin of 0.4.
`
`The sponsor employed 3 analysis populations: intent-to-treat (HT), per-protocol
`and sensitivity. The ITT population consisted of patients with on-treatment data
`and such that the last value was no later than 3 days of the last dose. HbA1c
`values had to be fasting as well. The sensitivity population consisted of all
`patients with any on-treatment data, consistent with the commonly applied
`definition of ITT.
`'
`
`This reviewer constructed 2-sided 95% Cl’s for the treatment difference for
`
`HbA1c to assess non—inferiority. The 2—sided 95% Cl yields the same upper
`bound as the 97.5% one-sided Cl and also provides a lower bound (“best case")
`for the treatment effect. The sensitivity population and completers were
`examined.
`'
`
`7 Results
`
`7.1 Dosing of test drug
`
`Final mean daily doses were similar for the 2 groups, 2112mg for XT and
`2122mg for Glucophage. Table 5 shows cross-classifications for the start and
`final doses of XT and Glucophage by treatment group. The shaded numbers in
`the Table are the numbers of patients whose final dose increased from their
`starting dose. Only a few patients in each group had their dose titrated down.
`
`A slightly greater % of Glucophage patients had dose increases during titration
`(137/341; 40%) compared to XT patients (120/339; 35%). For patients with dose
`increases, the mean increase was 688mg for XT and 653mg for Glucophage.
`Overall, mean dose changes from start to final dose were similar between groups
`(XT, +256mg; Glucophage, +235mg).
`
`None of the dose differences was statistically significant.
`
`

`

`
`Table 5. Number of oatients b startin- dose and final dose
`.
`
`
`
`
`Start dose
`Final dose ma/ da
`
`1500
`1000
`(mg / day)
`2000
`
`
`__———
`,
`
`
`
`
`
`
`
`
`_500_n .. _
`
`—-_
`
`
`
`_—All doses
`
`
`_—1000
`
` 1500
`-_
`
`2000
`
`2500
`_
`
`All doses
`
`
`
`—-
`
`Shaded cells show patients whose final dose was increased over their starting dose
`
`7.2 Efficacy
`
`7.2.1 HbA1c
`
`Table 6 shows the analysis results for this reviewer’s ITT analysis at Week 26,
`equivalent to the sponsor’s sensitivity analysis. Mean changes from baseline for
`XT and Glucphage were 0.40 and 0.14, respectively. The least square mean
`treatment difference was 0.25 (95% CI = 0.14, 0.37). The upper bound of the 2-
`sided 95% confidence interval (CI) for the mean was 0.37, smaller than the pre-
`defined non—inferiority margin of 0.40. XT was therefore non-inferior to
`Glucophage according. to the pre-defined criterion to establish efficacy.
`
`_ All 3 analysis populations gave roughly similar results on the primary endpoint.
`For the completers population, the LS mean treatment difference was 0.18
`(upper bound of CI = 0.30).
`
`The FDA disqualified the investigator. at Center E
`unchanged when data from the 9 patients at this site 1',
`removed from the analysis.
`
`:1 Results were
`_
`- 3, were
`
`

`

`Table 6. HbA1c (%) results
`Reviewer’slTl' co ulation LOCF 1
`
`Metformin XT
`‘ Glucophage
`n=332
`
`7.04 (0.88)
`L
`
`7.07 (0.76)
`
`1
`
`7.21 (0.97)
`7.44 (1.09)
`
`7.29 £0339)
`7.19 (0.98)_
`
` Baseline
`
`Endpoint
`mean (SD)
`Comoleters 2 mean SD
`Change from baseline
`Mean (SE)
`Min
`Max
`Adjusted mean 3 (SE)
`Adj. treatment difference
`Mean (SE)
`95% Cl
`4
`Same as sponsor's sensitivity analysis
`2 Sponsor's designation
`3 Adjusted for center and baseline HbAlc
`4 Sponsor calculated one-sided 97.5% CI which gives equivalent upper bound
`
`0.40 (0.04)
`L
`E
`0.41 (0.05)
`
`'
`
`'
`
`0.14 (0.04)
`’
`1
`
`J
`
`0.25 (0.06)
`0.14, 0.37
`
`'
`
`Because the lower bound of the CI for the mean difference excluded zero, XT
`
`was also statistically inferior to Glucophage (p<.0001) in addition to being
`clinically non-inferior. The statistical difference is illustrated graphically in Figure
`1 which shows the cumulative distribution functions (CDF) of the primary
`endpoint for each treatment group. The graph shows a clear separation in the
`CDF's between groups.
`
`Appears This Way
`On Original
`
`10
`
`

`

`Figure1
`
`Cumulative Dis ribution {unctions
`HbAlc change at Week 26
`(Metformin XT n=327; Clucophaqe n=332)
`
`1
`
`
`
`cumulativepercentofpatients
`
`888883888
`10
`
`ll
`
`

`

`V Figure 2 shows individual data for HbA1c change from baseline. Regression
`lines are drawn for each group.
`
`i
`
`HbA
`
`ic
`change from b
`i
`'
`=
`(Mei ormln Xi
`H
`
`t 2':-
`
`aseiine HbAic
`age n=352)
`
`change
`
`'HbA1c
`
`treatnent g'oq)
`
`5‘35 Guoophage
`
`'
`
`'
`
`' MatfornhnXl'
`
`12
`
`

`

`Figure 3 shows mean HbA1c change from baseline over time (weeks) for
`completers. The time courses in the two groups from Weeks 9-26 are similar.
`Up to Week 9, however, XT patients experienced an increase in HbA1c whereas
`Giucophage patients experienced a decrease. This pattern up to Week 9 is not .
`explained by the dosing data showing the 2 groups had comparable titrations.
`
`The graph does not show time~course data for dropouts. Dropouts will be
`analyzed in more detail in the next section.
`
`Figure 3
`Meian HbAic
`for completers
`Malformin XT n=263; Giucophage n=292
`
`'
`
`Mean
`
`l—le1cQ.07.03.03.07NNNNNN010)\1oo«3oANwA01
`
`—— 012345678911 1 11 11 11 12222222
`21
`01234567890123456
`
`Vkiéek
`
`treatnent group
`
`BTB‘BGuoophage
`I
`I MatfornhnXl'
`
`l3
`
`

`

`Dropouts
`
`Table 7 shows mean the HbA1c change from baseline by dropout cohorts.
`Dropout cohorts are mutually exclusive groups of patients defined by the time of
`last HbA1c on study. The combined dropout cohorts on XT experienced a mean
`0.73 increase in HbA1c from baseline compared with a 0.19 increase on
`Glucophage (shaded row in the Table). The mean treatment difference was
`0.54.
`
`Table 7. Mean HbA1c change from baseline by dropout cohort 1
`Reviewer’s ITT . o - ulation
`
`
`
`
`Dropout cohort
`Metformin XT
`Glucophage
`
`
`Difference SE
`(n=327)
`
`
`
`
`0.44 (0.46)
`Week 3
`0.46
`
`
`
`
`(n=ZL
`
`
`
`.‘041 (0 56)
`Week 6
`0.21
`
`
`
`‘
`(n=9)
`
`
`
`
`
`
`£513)
`
`
`
`0. 42 (0.45)
`0.46
`
`
`(n=10_)
`
`
`
`0.76 (0.58)
`0.83
`
`
`
`(n=8)
`
`
`
`
`
`
`
`(n=14)
`
`Weeks 3 to 21
`0.73
`
`
`
`
`
`combined
`(n:61)
`
`
`0 32
`
`
`(n2=66L
`(n=294)
`
`
`Dropout cohorts are mutually exclusive groups of patients defined by the'time of last
`HbA1c on study.
`2 Cohort contains data for all completers plus 5 non-completers with Week 26 data.
`Two completers in the cohort did not have Week 26 data. Their last data on study was
`Week 21.
`
`0. 18 (.0 06)
`
`This reviewer compared age, sex, race and baseline HbA1c characteristics
`between dropouts and completers (Table 7, last 2 rows) to see whether there
`were factors that might have contributed to the observed difference in responses.
`Only baseline HbA1c was statistically different between completers (mean 7.01)
`and dropouts (mean 7.30) (p=.001).
`
`Interestingly, patients
`Table 8 further explores HbA1c responses for all dropouts.
`who dropped due to AE’s had a mean treatment difference of 0.65, higher than
`the mean for all dropouts (not shown in Table). The poor XT response for
`
`14
`
`

`

`dropouts was therefore not confined to patients that dropped due to a stated lack
`of efficacy.
`
`'
`
`
`
`
`
`
`
`Table 8. HbA1c for Dropouts with data — Reviewer’s ITT LOCF
`
`Metformin XT
`Glucophage
`Treatment
`
`n=38
`Difference
`
`
`Baseline
`
` 7.37 (0.94)
`
`
`7.45
`
`Endpoint
`
`
`
`8.10 (1.28)
`Mean (SD)
`
`
`7.80
`Median
`
`
`Change from Baseline
`
`
`
`0.73 (0.95)
`Mean (SD)
`
`
`0.50
`Median
`
`7.19(0.91)
`7.13
`
`7.38 (0.90)
`7.10
`
`0.18
`0.32
`
`0.72
`0.70
`
`Mean (SD)
`Median
`
`
`
`0.19 (0.91)
`-0.03
`
`.
`
`0.54
`0.53
`
`Longitudinal analysis
`
`To further investigate the impact of dropouts, I conducted a post—hoc analysis of
`the repeated measures data over time (Longitudinal data) using a mixed model
`with patient as a random effect. The purported strengths of longitudinal models
`are that they use multiple observations on the same patient, provide a method for
`handling missing data, and allow the examination of treatment effects over time.
`
`The model was:
`
`HbA1c change from baseline = baseline HbA1c + group + week + center +
`group*week + baseline HbA1c*week
`
`Twenty-seven (2,7, 4%) patients in the sensitivity population had Week 3 or Week
`6 HbA1c data. The data at these early, sparse time points were deleted so that
`the model could converge. The analysis involved a total of n=2906 observations
`at Weeks 9, 13, 17, 21 and 26.
`
`The treatment contrast at Week 26 (0.26) was almost identical to the LOCF result
`(95% CI = 0.14, 0.38). The mean treatment effect did not vary significantly over
`the time period examined (p=.27).
`'
`
`15
`
`

`

`Response rate
`
`This reviewer performed post—hoc analyses of response rates using a criterion for
`response suggested by Robert Misbin, the reviewing Medical Officer. A positive
`response was defined as a change from baseline in HbA1c of less than +0.7, a
`negative response as a change of +0.7 or greater. The 0.7 figure, while
`somewhat arbitrary, was chosen because it corresponded to the smallest effect
`size seen in placebo controlled trials for Type 2 diabetes drugs approved to date
`by the Medical Division and also represents an approximate 10% increase in
`HbA1c over baseline for patients in the trial. Table 9 shows response rates for
`each group.
`
`Response rates
`
`
`
`
`
`Table 9. HbA1c resonse rates at Week 26 ITT - LOCF ‘
`
`
`Metformin XT
`Glucophage
`246/327
`276/332
`75%
`83%
`
`
`
`-value Fisher‘s exact test)_
`
`A positive response for a patient was defined as a HbA1c Week 26 change from
`baseline < +0.7
`'
`
`_
`
`
`
`An analysis using a responder cutoff of 0.5 gave similar results.
`
`7.2.2 Fasting plasma glucose
`
`Table 10 shows this reviewer’s FPG results for the sponsor’s l'lT population (all
`values fasting; last value within 3 daysof last dose of study drug). The 95% CI
`for the difference in least square mean change from baseline was (0.6, 12.3)
`measured in mg/dL units. These results agreed with the sponsor's results.
`
`Appears This Way
`On Original
`
`16
`
`

`

`146.7 (31.8)
`@251.5_)_
`
`
`
`Table 10. Fasting Plasma glucose (mg/dL results
`
`
`Sponsor’s ITT_Eo ulation LOCF
`
`Metformin XT
`Glucophage
`
`(n=333L
`E329)
`
`
`
`145.6 (29.4)
`(72, 236)
`
`Endpoint
`
`
`149.8 (41.2)
`mean (SD)
`156.8 (48.5)
`
`
`
`148.7 (£9)
`Comjleters ‘ mean ED)
`148.8 (37.3)
`
`Change from baseline
`
`
`4.2 (2.0)
`Mean (SE)
`10.1 (2.2)
`
`
`4.5 (a)
`Adjusted mean 2 (g)
`11.0 (2.4)
`
`Adj. treatment difference
`
`6.5 (3.0)
`Mean (SE)
`.
`
`(0.6, 12.3) 95% CI
`
`
`fLvalue
`0.03
`Sponsor's designation
`2 Adjusted for center and baseline HbA1c
`
`7.3 Comedications
`
`Table 11 shows insulin use during the trial by WHO drug preferred term
`(specific).
`
`
`
`Table 11. Number ofpatients taking insulins and insulin analogues
`Metformin XT
`Glucgphage
`
`26
`23
`Insulin and analogues total
`7
`3
`Insulin
`
`
`Insulin human
`2
`3
`
`Insulin human injection, isophane
`11
`14
`
`
`-
`Insulin human semisynthetic
`0
`1
`'
`
`
`Insulin human zinc suspension
`6
`11
`
`Insulin-injection, biphasic
`0
`1
`
`Insulin injection, isophane '
`6
`3
`Insulin isophane human'semfiynthetic
`0
`1
`Insulin Iispro
`1
`2
`
`Insulin zinc suspension 2 0
`
`
`
`
`The same approximate numbers of patients in both groups were taking insulin or
`analogues; however, 16 of 23 ( 70%) Glucophage patients taking insulins took
`multiple insulins vs just 6 of 26 (23%) XT patients. Results for the primary
`
`l7
`
`

`

`endpoint were essentially unchanged (95% CI for treatment difference: 0.14,
`0.38) when patients taking one or more insulins were removed from the analysis
`
`The analysis above was repeated for patients taking concomitant glibenclamide
`or glipizide (XT, n= 144; Glucophage, n= 157), the most frequently prescribed
`Type 2 oral anti-diabetic medications. Statistical results were similar for patients
`taking either medication and patients taking neither medication.
`
`7.4 Compliance
`
`Compliance rates were similar in the groups, 96% for XT and 97% for
`Glucophage. The compliance rate was presumably calculated as
`
`Compliance = 100% — (# tablets returned /# tablets dispensed)
`
`7.5 Assay sensitiVity
`
`Although GlucophageIS an established, standard treatment for Type 2 diabetes
`one could ask whether the drug was effective in the trial since patients receiving
`Glucophage experienced a mean increase in HbA1c over baseline This result
`might call the assay sensitivity of the trial into question.
`
`Assay sensitivity can be directly inferred in placebo controlled trials whenever a
`drug is shown to be superior to placebo.
`In active control trials, where the goal is
`to show non—inferiority, evidence of assay sensitivity usually cannot be inferred in
`the same way since there is typically not a placebo group present. Evidence of
`assay sensitivity typically comes from historical data outside the trial.
`In the
`current trial withOut a placebo control, however, assay sensitivity can be inferred
`from the trial itself since one of the treatments (Glucophage) was shown to be
`statistically superior to the other (XT).
`
`8 Suggestions for labeling
`
`, 1. Study 302 was a tolerability and safety study and not powered for efficacy.
`Efficacy data should not be labeled. Safety data can be put in the
`Adverse Reactions section of the label.
`
`2. Table 5 -— present 2-sided Cl's for HbA1 0 treatment difference
`
`3. Figure 2 — show results for completers and SE bars instead of SD.
`
`18
`
`

`

`4. The text describing results for LDL—C, HDL—C, TC and TG can be deleted
`since Tables 6 and 7 have the same information. Table 6 may be
`unnecessary.
`
`5. Although Study 301 showed XT to be non-inferior to Glucophage on
`HbA1c, dropouts fared poorly on XT. A significant percentage of these
`dropouts were due to a stated lack of efficacy. The label should
`incorporate these data in some appropriate fashion.
`
`Appears This Way
`On Original
`
`19
`
`

`

`This is a representation of an electronic record that was signed electronically and
`this page is the manifestation of the electronic signature.
`
`Todd Sahlroot
`
`10/20/03 11:07:28 AM
`BIOMETRICS
`
`S. Edward Nevius
`
`10/29/03 04:13:34 PM
`BIOMETRICS
`Concur with review.
`
`