`
`CENTER FOR DRUG EVALUATION AND
`RESEARCH
`
`
`APPLICATION NUMBER:
`206073Orig1s000
`
`STATISTICAL REVIEW(S)
`
`
`
`
`
`
`
`

`

`
`
`
`
`U.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 T A T I S T I C A L R E V I E W A N D E VA L U A T I O N
`CLINICAL STUDIES
`
`NDA/BLA #:
`Supplement #:
`Drug Name:
`Indication(s):
`Applicant:
`Date(s):
`
`NDA 206073
`Original-1
`Empagliflozin (BI 10773)/linagliptin (BI 1356)
`Type II Diabetes Mellitus
`Boehringer Ingelheim
`Stamp Date: January 30, 2014
`User Fee Goal Date: January 31, 2014
`
`Standard
`Review Priority:
`
`
`Division of Biometrics II
`Biometrics Division:
`Jennifer Clark, Ph.D.
`Statistical Reviewer:
`Concurring Reviewers: Mark Rothmann, Ph.D., Statistical Team Leader
`Thomas Permutt, Ph.D., Division Director
`
`Division of Metabolism and Endocrinology Products
`William Chong, M.D.
`Callie CapelLynch
`
`
`Medical Division:
`Clinical Team:
`Project Manager:
`
`Reference ID: 3644133
`
`

`

`2 
`
`3 
`
`Table of Contents
`1  EXECUTIVE SUMMARY ................................................................................................................................. 5 
`1.1 
`CONCLUSIONS AND RECOMMENDATIONS ....................................................................................................... 5 
`1.2 
`BRIEF OVERVIEW OF CLINICAL STUDIES ........................................................................................................ 5 
`1.3 
`STATISTICAL ISSUES AND CONCERNS ............................................................................................................. 8 
`INTRODUCTION ............................................................................................................................................... 8 
`2.1 
`OVERVIEW ...................................................................................................................................................... 8 
`2.1.1 
`History of Drug Development ................................................................................................................ 9 
`2.2 
`DATA SOURCES ............................................................................................................................................ 10 
`STATISTICAL EVALUATION ...................................................................................................................... 10 
`3.1 
`DATA AND ANALYSIS QUALITY ................................................................................................................... 10 
`3.2 
`EVALUATION OF EFFICACY .......................................................................................................................... 11 
`3.2.1 
`Study Design and Objectives ............................................................................................................... 11 
`3.2.2 
`Statistical Methodologies ..................................................................................................................... 15 
`3.2.3 
`Patient Disposition, Demographic and Baseline Characteristics........................................................ 16 
`3.2.4 
`Results and Conclusions ...................................................................................................................... 21 
`FINDINGS IN SPECIAL/SUBGROUP POPULATIONS ............................................................................. 30 
`4.1 
`GENDER, RACE, AGE, AND GEOGRAPHIC REGION ........................................................................................ 30 
`4.2 
`OTHER SPECIAL/SUBGROUP POPULATIONS .................................................................................................. 31 
`SUMMARY AND CONCLUSIONS ................................................................................................................ 33 
`5.1 
`STATISTICAL ISSUES ..................................................................................................................................... 33 
`5.2 
`COLLECTIVE EVIDENCE ................................................................................................................................ 33 
`5.3 
`CONCLUSIONS AND RECOMMENDATIONS ..................................................................................................... 34 
`
`4 
`
`5 
`
`
`
`
`
`Reference ID: 3644133
`
`2
`
`

`

`LIST OF TABLES
`
`Table 1: Efficacy Results in Proposed Label for Empagliflozin Monotherapy compared to Placebo .......................... 9 
`Table 2: Efficacy Results from the label for Linagliptin (Tradjenta) compared to Placebo ....................................... 10 
`Table 3: Applicant created table of Treatment Regimens/Study Intervals for 24 Week Analysis .............................. 12 
`Table 4: Descriptive Statistics for the Metformin Treated Population ....................................................................... 17 
`Table 5: Baseline Descriptive Statistics for the Treatment Naive Population ............................................................ 20 
`Table 6: Mean HbA1c Results for Metformin Treated Subjects ................................................................................ 22 
`Table 7: Primary Endpoint Adjusted Differences between Treatments for Metformin Subjects ............................... 22 
`Table 8: Mean FPG Results for Metformin Treated Subjects ..................................................................................... 23 
`Table 9: FPG Differences between Treatment Arms for Metformin Treated Subjects .............................................. 23 
`Table 10: Mean Body Weight Results for Metformin Treated Subjects ..................................................................... 24 
`Table 11: Change in Body Weight Differences between Treatment Arms, Metformin ............................................. 24 
`Table 12: Proportion of Subjects with HbA1c<7% in the Metformin Treated Subjects ............................................ 24 
`Table 13: Odds Ratios between Treatment Arms for HbA1c<7%, Metformin .......................................................... 25 
`Table 14: Mean HbA1c Results for Treatment Naive Subjects .................................................................................. 25 
`Table 15: Primary Endpoint Adjusted Differences between Treatments, Treatment Naive ....................................... 25 
`Table 16: Mean HbA1c at 52 Weeks .......................................................................................................................... 26 
`Table 17: HbA1c Adjusted Differences at 52 Weeks between Treatments, Treatment Naive ................................... 27 
`Table 18: Mean FPG for Treatment Naive Subjects ................................................................................................... 27 
`Table 19: FPG Differences between Treatment Arms, Treatment Naive ................................................................... 28 
`Table 20: Mean Body Weight for Treatment Naive Subjects ..................................................................................... 28 
`Table 21: Body Weight differences between Treatment Arms, Treatment Naive ...................................................... 28 
`Table 22: Treatment Naive Subjects with HbA1c<7% ............................................................................................... 29 
`Table 23: Adjusted ORs for HbA1c<7%, Treatment Naive ....................................................................................... 29 
`Table 24: Subgroup Statistics for Change in HbA1c in Metformin Treated Subjects ................................................ 30 
`Table 25: Subgroup Statistics for Change in HbA1c in Treatment Naive Subjects .................................................... 31 
`Table 26: Other Subgroup Stats for Change in HbA1c in Metformin Treated Subjects ............................................. 31 
`Table 27: Other Subgroup Stats for Change in HbA1c in Treatment Naive Subjects ................................................ 32 
`
`
`
`
`Reference ID: 3644133
`
`3
`
`

`

`LIST OF FIGURES
`
`Figure 1: Testing Hierarchy Results for Metformin Treated Subjects .......................................................................... 7 
`Figure 2: Testing Hierarchy Results for Treatment Naive Subjects ............................................................................. 7 
`Figure 3: Applicant created schematic for the Primary Endpoint Testing Hierarchy ................................................. 14 
`Figure 4: Proportion of missing data at each visit for the metformin treated population ........................................... 17 
`Figure 5: Proportion of Missing Data at each visit for the Treatment Naive Population ............................................ 19 
`
`
`
`
`
`Reference ID: 3644133
`
`4
`
`

`

` 1
`
` EXECUTIVE SUMMARY
`
`This is a statistical review for Boehringer Ingelheim’s submission of empagliflozin/linagliptin
`combination as treatment for type 2 diabetes mellitus (T2DM). The applicant is seeking
`approval based on a change from baseline in glycosylated hemoglobin (HbA1c) at week 24. A
`low fixed dose combination (FDC) of empagliflozin 10 mg + linagliptin 5 mg as well as a high
`FDC of empagliflozin 25 mg + linagliptin 5 mg were studied against their monotherapy
`counterparts. Analysis results from this study are presented in this review
`
`
`1.1 Conclusions and Recommendations
`
`The submitted study for FDC empagliflozin/linagliptin showed some efficacy benefits when
`compared with monotherapy. While the trial did not show significant improvements in the
`treatment naïve population, it did show statistically significant improvements in the metformin
`treated subjects.
`
`The lack of evidence showing an improvement in efficacy when comparing high FDC to low
`FDC does leave some reservations on the need for the higher dose. The findings in this review
`support approvability of FDC in the metformin population.
`
`
`1.2 Brief Overview of Clinical Studies
`
`There was one clinical study, BI Trial No. 1275.1, for empagliflozin (BI 10773) / linagliptin (BI
`1356) combination. This was a phase III randomized, multi-national, double-blind, parallel
`group study to evaluate the efficacy and safety of once daily oral administration of low and high
`fixed dose combination (FDC) tablets compared with the individual components (BI 10773 25
`mg, BI 107723 10 mg, and linagliptin 5 mg) for 52 weeks, with the primary endpoint taken at
`week 24. The study was stratified and analyzed separately for two different populations,
`treatment naïve and metformin treated patients with type 2 diabetes mellitus and having
`insufficient glycemic control. The results given in this review are based on separate analyses for
`each population. The trial lasted from 21 August 2011 to 10 September 2013. A hierarchical
`testing procedure was specified for each population. Due to the ordering of this procedure, all of
`the primary and secondary endpoints in the treatment naïve population were considered to be
`exploratory. The results found for this review, using methods similar to what was specified in
`the protocol, are given in Figure 1 and Figure 2. Results for endpoints shown in grey boxes
`should only be considered exploratory, even when p<0.05, due to the protocol specified testing
`hierarchy.
`
`The protocol specified a comparison of each FDC with respective monotherapies to show
`improved efficacy. There was, however, no specification for testing whether the higher FDC had
`any greater efficacy than the low FDC. Analyses done in this review showed a non-significant
`difference of -0.11 (-0.28, 0.06) in the metformin treated population and 0.15 (-0.05, 0.35) in the
`treatment naïve population. Sample size calculations done by the applicant during the IND stage
`
`5
`
`Reference ID: 3644133
`
`

`

`(IND 108388) were run based on an effect size of 0.5% between FDC and all monotherapies for
`change in HbA1c. Using this effect size for a post-hoc power calculation comparing low and
`high FDC, we see that this study appears to be adequately powered to detect such a difference
`between FDC in both populations. Since a significant effect of 0.5% remains outside both
`confidence intervals, we maintain reservations on whether the use of high FDC provides any
`meaningful difference when compared with low FDC.
`
`Subgroup analyses of this combination confirmed previous findings in the empagliflozin
`monotherapy trials of possible interactions of treatment with renal function (See the statistical
`review for NDA 204629 by Dr. Dongmei Liu, signed October 30, 2013). Since such interactions
`did not preclude the approval of empagliflozin as monotherapy for T2DM, it will not be further
`discussed here.
`
`Analysis results following the pre-specified hypothesis testing hierarchy for primary and key
`secondary endpoints, which are in line with what is in the study report, are shown in Figure 1 and
`Figure 2. Although the trial failed for the treatment naïve population when using the hierarchy,
`there does appear to be improvements in efficacy when compared with some monotherapies, but
`because of specifications within the protocol, these can only be considered “exploratory”.
`
`There is the possibility that the trial failed not because the FDC treatment was not any better than
`monotherapy in the treatment naïve population, but because there may be an efficacy “ceiling”.
`Treatment naïve subjects given empaliflozin 25 mg monotherapy seem to hit this ceiling, so
`adding linagliptin to this treatment would not result in added benefit. Results seen with
`empagliflozin 10 mg indicate there may still be room for additional benefit. Proving the
`existence of such a ceiling effect would be difficult and beyond the scope of this study.
`
`
`
`
`
`Reference ID: 3644133
`
`6
`
`

`

`Figure 1: Testing Hierarchy Results for Metformin Treated Subjects
`
`
`Figure 2: Testing Hierarchy Results for Treatment Naive Subjects
`
`
`
`
`
`7
`
`
`
`
`
`Reference ID: 3644133
`
`

`

`1.3 Statistical Issues and Concerns
`
`There were several concerns that emerged during the course of this review. One had to do with
`failure of part of the trial due to the specification of the hierarchical testing procedure. The other
`concern had to do with an absence of methods to show increased efficacy of high FDC over low
`FDC.
` The biggest statistical issue for this study has to do with the failure of the primary and all
`secondary endpoints in achieving statistical significance as was pre-specified for the
`treatment naïve population. While there did appear to be some evidence of improved
`efficacy in this population, these results must be considered “exploratory”, as deemed by
`the protocol,
`.
`
`
`
` A separate issue with this submission has to do with the lack of evidence of improved
`efficacy in both populations when comparing high FDC to low FDC. This was neither
`specified in the protocol nor provided in the submission. Post-hoc analyses to test if the
`two FDCs were equivalent failed to show evidence of improved efficacy in terms of the
`chosen endpoints for the high FDC.
`
` Trial results for each population were not supportive of each other as analysis results
`were not significant for one population.
`
`
`
`
` 2
`
` INTRODUCTION
`
`2.1 Overview
`
`BI 10773/linagliptin is a combination product of two oral antidiabetic agents for the treatment of
`type 2 diabetes mellitus (T2DM). Linagliptin as a monoproduct was approved on May 2, 2011
`under NDA 201280 and is currently available under the name Tradjenta. Empagliflozin was
`submitted for approval as a monotherapy on March 5, 2013 under NDA 204629 and was
`subsequently approved under will be marketed under the trade name Jardiance. The proposed
`name and indication for this fixed-dose combination (FDC) tablet is Glyxambi which is to be “an
`adjunct to diet and exercise to improve glycemic control in adults with type 2 diabetes mellitus
`when treatment with both empagliflozin and linagliptin is appropriate.” The current study
`examined safety and efficacy in both treatment naïve and metformin treated patients with type 2
`diabetes mellitus with insufficient glycemic control. This was a 52 week phase 3 randomized,
`double-blind, parallel group study with five treatment arms, Empagliflozin 25mg/Linagliptin 5
`mg (nmetformin =137, nnaïve=137), Empagliflozin 10mg/Linagliptin 5 mg (nmetformin =135,
`nnaïve=136), Empagliflozin 25mg (nmetformin=141, nnaïve=135), Empagliflozin 10mg (nmetformin=140,
`nnaïve=134), and Linagliptin 5 mg (nmetformin =132, nnaïve=135).
`
`
`
`
`
`Reference ID: 3644133
`
`8
`
`(b) (4)
`
`

`

`2.1.1 History of Drug Development
`Empagliflozin is an oral selective inhibitor of SGLT-2 which was reviewed and approved for
`treatment levels of 10 mg and 25 mg daily. Previous phase 3 studies compared empagliflozin as
`a monotherapy or add on to metformin or other therapies to placebo. Table 1, copied from the
`proposed label, shows study results for empagliflozin (Jardiance) as monotherapy compared with
`placebo. This study demonstrated an improvement in lowering HbA1c below 7%, fasting
`plasma glucose, and body weight with both low and high dose empagliflozin when compared to
`placebo.
`
`
`Table 1: Efficacy Results in Proposed Label for Empagliflozin Monotherapy compared to
`Placebo
`
`
`Linagliptin is a DPP-4 activity inhibitor, prolonging the half-life of GLP-1, which can be
`administered orally or intravenously. Clinical studies run by the applicant indicated that for this
`treatment there were no relevant interactions with metformin, pioglitazone, glyburide, or
`empagliflozin. Previous phase 3 studies compared 5 mg linagliptin (trade name Tradjenta) with
`placebo, metformin, and various combinations of linagliptin and metformin. Results from the
`product label for Tradjenta are given in Table 2. The applicant was able to show a statistically
`significant difference in the change from baseline in A1C (%) and FPG when compared with
`placebo. There does not seem to be a difference in the proportion of patients achieving A1C
`under 7% in this study, though.
`
`
`
`
`
`Reference ID: 3644133
`
`9
`
`(b) (4)
`
`

`

`Table 2: Efficacy Results from the label for Linagliptin (Tradjenta) compared to Placebo
`
`
`
`
`Since empagliflozin inhibits SGLT-2 and linagliptin inhibits DPP-4, the applicant rationalizes
`that the combination of the two “may lead to additional effects on glycemic control.” The
`convenience factor of having only one tablet instead of two is also offered as further justification
`for this combination.
`
`
`2.2 Data Sources
`
`The data and final study report were submitted electronically and archived under the network
`path location <\\CDSESUB1\evsprod\NDA206073\206073.enx>. The information needed for this
`review was contained in Module 1 FDA Regional information (cover letter, meetings, response
`to information requests) and Module 5 (clinical study reports). Independent coding for the
`analysis was run for this review. An information request was made for the applicant’s analysis
`code for verification.
`
`
` 
`
` STATISTICAL EVALUATION
`
`3.1 Data and Analysis Quality
`
`This submission is in the electronic common technical document (eCTD) form with an xml
`backbone. A statistical analysis plan was pre-specified in section 7 of the protocol and further
`detailed in a separate SAP. The methods described in this plan along with other sensitivity
`analyses were used in this review. Study datasets were provided as SAS XPORT transport files.
`
`The clinical study report mentions a monitoring visit on February 21, 2012 which turned up
`evidence of scientific and data misconduct in a site in the USA recording fraudulent data. This
`
`10
`
` 3
`
`Reference ID: 3644133
`
`

`

`site was closed and subjects were sent to a new site which was opened to replace the old one. It
`was also found that several patients were screened/randomized at multiple sites within the study
`which increased the patient numbers by 23. The multiple screenings/randomization of patients
`were in 13 sites, two of which only contained subjects with these violations. Data from the
`patients at the fraudulent site and those with multiple screenings/randomization were excluded
`from the primary analysis and clinical study report.
`
`Original datasets were submitted and used for the initial analysis for this review. Analysis
`datasets were inadvertently omitted from the original submission and later sent on May 7, 2014
`under the sequence number 0005.
`
`
`3.2 Evaluation of Efficacy
`3.2.1 Study Design and Objectives
`
`This was a phase III randomized, multi-national, double-blind, parallel group study to evaluate
`the efficacy and safety of once daily oral administration of BI10773 25 mg/linagliptin 5 mg and
`BI 10773 10 mg/linagliptin 5mg Fixed Dose Combination Tablets compared with the individual
`components (BI 10773 25 mg, BI 107723 10 mg, and linagliptin 5 mg) for 52 weeks in treatment
`naïve and metformin treated patients with type 2 diabetes mellitus with insufficient glycemic
`control. The study was powered to analyze two separate patient groups at week 24, metformin
`treated and treatment naïve patients. Statistical inference was carried out separately for each
`population.
`
`The study consisted of five periods, screening, placebo run-in, treatment phase in one of five
`treatment arms (Empagliflozin 25mg/Linagliptin 5mg, Empagliflozin 10mg/Linagliptin 5mg,
`Empagliflozin 25mg, Empagliflozin 10mg, and Linagliptin 5mg), post-treatment, and post-study.
`After a 2 week run-in period patients were randomized, stratified by patient population, to one of
`five treatment arms for the 52 week treatment period.
`
`
`
`Reference ID: 3644133
`
`11
`
`

`

`Table 3: Applicant created table of Treatment Regimens/Study Intervals for 24 Week
`Analysis
`
`
`
` X is 1, 3, or 7 days for pulse rate, safety laboratory, and AE respectively.
`
`
`
`There were a total of 2504 patients enrolled with 1363 entering the study. The population was
`stratified by either having a metformin background (n=686) or treatment naïve patients (n=677).
`These were recruited in 188 centers with screenings across 22 countries in Asia (10.8%), Europe
`(27.7%), Latin America (11.5%), and North America (50% of the screened population).
`
`For the metformin group, this was a multi-national study that included 1179 enrolled patients
`from 188 centers in 22 countries across Asia, Europe, Latin America, and North America. 686
`patients were randomized to each arm in a 1:1:1:1:1 ratio. Some subjects were excluded from
`the applicant’s analysis due to protocol violations described in section 3.1.
`
`
`
`
`
`12
`
`Reference ID: 3644133
`
`

`

`
`
`
`
`
`
`
`
`
`
`
`
`In the treatment naïve study population there were 1325 enrolled patients in the multi-national
`study from 201 centers across 22 countries in Asia, Europe, Latin America, and North America.
`There were 677 patients randomized in a 1:1:1:1:1 ratio.
`
`
`
`
`
`
`
`
`
`
`
`
`Primary Efficacy Objective and Endpoint
`The main objective of this study was to show superiority of the combination therapy of
`empagliflozin and linagliptin against the individual monotherapy components. With this study
`objective in mind, the applicant used a primary efficacy endpoint of change from baseline in
`HbA1c (%) at 24 weeks of treatment.
`
`Key Secondary Endpoints
`The key secondary endpoints listed in the protocol were to be tested hierarchically in the
`following order:
`1. Change from baseline in fasting plasma glucose (FPG) after 24 weeks of treatment
`2. Change from baseline in body weight after 24 weeks of treatment
`In an amendment to the original protocol an exploratory endpoint was taken to be the third key
`secondary endpoint,
`3. Occurrence of treat to target efficacy response: HbA1c of <7.0% (<53.0 mmol/mol)
`after 24 weeks of treatment
`
`
`Testing Hierarchy
`A separate testing procedure was followed for each of the patient populations with the high dose
`of the FDC being tested first against each respective monotherapy. Given that these were
`
`
`13
`
`Reference ID: 3644133
`
`

`

`significant then the low FDC was tested against respective monotherapies. The testing hierarchy
`used for both the primary endpoint and first secondary endpoint of FPG is given in Figure 3.
`
`
`Figure 3: Applicant created schematic for the Primary Endpoint Testing Hierarchy
`
`
`
`
`Both null hypotheses had to be rejected with a 2-sided hypothesis test with alpha=0.05 at each
`step before proceeding. If all primary endpoint hypotheses are significant, then the first key
`secondary endpoint of change in FPG could then be tested using the same hypothesis hierarchy.
`If all hypotheses were significant for FPG, then the testing hierarchy would continue with
`hypotheses for change in body weight. The following testing hierarchy was implemented for this
`secondary endpoint with a two-sided testing procedure setting alpha at 0.05.
`
`
`1. H0,1: Effect of BI 10773 25 mg/linagliptin 5 mg is equal or inferior to that of linagliptin
`5 mg;
`Against the alternative HA,1: Effect of BI 10773 25 mg/linagliptin 5 mg is superior to
`linagliptin 5 mg.
`
`2. H0,2: Effect of BI 10773 10 mg/linagliptin 5 mg is equal or inferior to that of linagliptin
`5 mg;
`Against alternative HA,2: Effect of BI 10773 10 mg/linagliptin 5 mg is superior to
`linagliptin 5 mg.
`
`3. H0,3: Effect of BI 10773 25 mg/linagliptin 5 mg is equal or inferior to that of BI 10773
`25 mg;
`
`
`
`Reference ID: 3644133
`
`14
`
`

`

`Against the alternative HA,3: Effect of linagliptin 5 mg/BI 10773 25 mg is superior to BI
`10773 25 mg.
`
`4. H0,4: Effect of BI 10773 10 mg/linagliptin 5 mg is equal or inferior to that of BI 10773
`10 mg;
`Against the alternative HA,4: Effect of BI 10773 10 mg/linagliptin 5 mg is superior to BI
`10773 10 mg.
`
`
`The third secondary endpoint of treat-to-target efficacy response HbA1c<7% was tested last if all
`preceding hypotheses were significant. The following order was specified here for testing the
`dosages of FDC versus different monotherapies for this endpoint:
`1. High dose FDC vs. linagliptin 5 mg
`2. Low dose FDC vs. linagliptin 5 mg
`3. High dose FDC vs. empagliflozin 25 mg
`4. Low dose FDC vs. empagliflozin 10 mg
`
`
`
`3.2.2 Statistical Methodologies
`
`Metformin background and treatment naïve patients were treated as two independent patient
`populations in this study. ANCOVA with last observation carried forward (LOCF) was used on
`the full analysis set (FAS) with randomized treatment for the primary endpoint. The model
`included fixed effects for treatment, region, and also the covariates baseline HbA1c, and
`screening eGFR to measure renal function.
`HbA1c change from baseline = overall mean + baseline HbA1c + treatment + renal
`function + geographical region + random error
`
`
`In a TSAP submitted in response to advice given by the Agency after unblinding, the main
`analysis was changed to be an MMRM. Due to the fact that both methods impute data based on
`data observed before dropout, meaning results will be similar, and ANCOVA was pre-specified,
`this review focuses on results from the ANCOVA. Renal function was measured through a
`bivariate eGFR score of <90 or at least 90.
`
`The MMRM analysis for treatment comparison of the adjusted mean change in HbA1c from
`baseline at Week 24 was originally specified as a REML based sensitivity analysis using the
`same population specified for ANCOVA (later changed to be the main analysis method in the
`TSAP). An unstructured covariance was to be used unless it failed to converge, in which case
`the following covariance structures were to be implemented with the best model used in the
`primary analysis as determined by AIC: unstructured, compound symmetry, variance
`components, and Toeplitz. The model for the MMRM analysis was,
`HbA1c change from baseline = overall mean + baseline HbA1c + treatment + screening
`eGFR + region + visit + visit by treatment interaction + random error.
`
`
`Other sensitivity analyses for the primary endpoint used the 24 week and 52 week FAS
`completers and PPS populations in order to assess the impact protocol violations and premature
`discontinuation had on the model used in the primary analysis. For my own analysis, I tested all
`
`15
`
`Reference ID: 3644133
`
`

`

`possible combinations of treatments (not just the ones specified by the sponsor) in order to gain a
`better understanding of where higher doses and combination therapy could be an improvement
`over low doses and monotherapies. Since the study was not designed for these hypotheses, only
`results that seemed relevant to better understanding the FDC are discussed.
`
`Once all the hypotheses in the primary endpoint hierarchy, described earlier, were found to be
`significant, the key secondary endpoint of fasting plasma glucose (FPG) was then specified to be
`tested using the same hierarchy. A similar ANCOVA model but with an additional baseline FPG
`covariate (baseline HbA1c also stayed in the model).
`
`If this set of FPG hypotheses were found to be significant, then the set of hypotheses specified
`for body weight could then be tested. Again, a similar ANCOVA model as specified for the
`primary endpoint was used also including an additional baseline weight covariate.
`
`The last secondary endpoint that was tested was the treat-to-target endpoint (HbA1c < 7%) for
`four different hypotheses testing the FDC therapies against their respective monotherapies.
`Testing was done using a logistic regression, imputing all missing data as failures. The model
`for this regression was similar to the ANCOVA model with factors for treatment, baseline renal
`function, geographical region, and a baseline HbA1c covariate. Reported odds ratios along with
`95% CIs and p-values were based on this model.
`
`
`3.2.3 Patient Disposition, Demographic and Baseline Characteristics
`
`
`Metformin Treated Subjects
`In the study report the applicant states that there were 1179 subjects enrolled with 747 entering in
`the placebo run-in period. Of these, 686 were randomized in a 1:1:1:1:1 ratio to the study arms
`with:
`
`137 on empagliflozing 25 mg/linagliptin 5 mg,
`136 on empagliflozin 10 mg/linagliptin 5 mg,
`141 on empagliflozing 25 mg,
`140 on empagliflozin 10 mg,
`132 on linagliptin 5 mg.
`
`They found 628/686 (92.5%) completed the 24-week treatment period for the primary analysis
`with 58 (8.5%) prematurely discontinuing trial medication. Of these, 17 subject (2.5%)
`discontinued due to adverse events, 15 (2.2%) were lost to follow-up. For my own analysis, the
`proportion of dropouts in the dataset over time are shown in Figure 4. Table 4 contains
`descriptive statistics for the baseline characteristics in the study. It should be noted that the
`sample size that I used for the descriptive analysis and what was stated in the study report differ.
`In staying with the ITT principle I used all randomized subjects in the original dataset for the
`purposes of the descriptive analysis. The proportions were ultimately the same and appear to be
`balanced between the five arms.
`
`
`
`
`Reference ID: 3644133
`
`16
`
`

`

`Figure 4: Proportion of missing data at each visit for the metformin treated population
`
`
`
`
`
`
`
`Table 4: Descriptive Statistics for the Metformin Treated Population
`
`Empa 25 mg
`
`Empa 10 mg
`
`Empa 25 mg
`
`Empa 10 mg
`
`Lina 5 mg
`
`Characteristic
`
`Category
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Sex
`
`Race
`
`Ethnicity
`
`Region
`
`
`
`Female
`Male
`
`
`
`White
`Black / African American
`Asian
`American Indian / Alaska Native
`
`
`Hispanic / Latino
`Not Hispanic / Latino
`
`
`
`North America
`Latin America
`
`Lina 5 mg
`(N=146)
`66 (45.21%)
`80 (54.79%)
`
`
`
`
`
`
`
`105 (71.92%)
`11 (7.53%)
`22 (15.07%)
`8 (5.48%)
`
`42 (28.77%)
`104 (71.23%)
`
`68 (46.58%)
`19 (13.01%)
`
`Reference ID: 3644133
`
`Lina 5 mg
`(N=141)
`55 (39.01%)
`86 (60.99%)
`
`
`107 (75.89%)
`13 (9.22%)
`18 (12.77%)
`3 (2.13%)
`
`
`(N=144)
`76 (52.78%)
`68 (47.22%)
`
`
`103 (71.53%)
`14 (9.72%)
`20 (13.89%)
`7 (4.86%)
`
`
`(N=146)
`62 (42.47%)
`84 (57.53%)
`
`
`110 (75.34%)
`11 (7.53%)
`19 (13.01%)
`6 (4.11%)
`
`
`
`
`
`37 (26.24%)
`104 (73.76%)
`
`
`69 (48.94%)
`18 (12.77%)
`
`41 (28.47%)
`103 (71.53%)
`
`
`69 (47.92%)
`20 (13.89%)
`
`
`
`
`
`52 (35.62%)
`94 (64.38%)
`
`
`(N=141)
`69 (48.94%)
`72 (51.06%)
`
`
`103 (73.05%)
`14 (9.93%)
`15 (10.64%)
`9 (6.38%)
`
`
`
`
`
`48 (34.04%)
`93 (65.96%)
`
`71 (48.63%)
`19 (13.01%)
`
`67 (47.52%)
`18 (12.77%)
`
`17
`
`

`

`
`
`
`
`
`
`
`