Bioequivalence Study in Healthy Volunteers for Empagliflozin/Metformin Fixed Dose Combination Compared to Separate Tablets: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Bioequivalence Study in Healthy Volunteers for Empagliflozin/Metformin Fixed Dose Combination Compared to Separate Tablets (NCT02028767), a randomized phase 1 study. The primary outcome is AUC (0-tz) (Area Under the Concentration-time Curve of Metformin in Plasma Over the Time Interval From 0 to the Last Quantifiable Data Point).

2. Clinical Question

The central question is whether metformin exposure after taking the empagliflozin/metformin fixed-dose combination is comparable with exposure after taking the empagliflozin and metformin tablets together. The study assessed this in healthy male and female volunteers under fed conditions, following a single dose in a randomized, two-period, two-sequence crossover design.

This is a pharmacokinetic comparison, not a study of clinical outcomes or treatment in people with a condition. Its stated purpose was to establish bioequivalence. The registry identifies the primary measure as metformin AUC from time 0 to the last quantifiable data point, reported as a geometric mean with geometric coefficient of variation. The posted analyses give two geometric mean ratio estimates and 90% confidence intervals. The record does not explain how the two entries correspond to the listed outcomes, or give the equivalence bounds used for a formal decision.

3. Trial Design

The registry classifies NCT02028767 as a Phase 1, randomized, open-label crossover study. The official title describes a single-dose, two-period, two-sequence study conducted under fed conditions. Participants received the fixed-dose combination and the separate tablets in different sequences, with each participant intended to contribute observations under both treatment conditions as part of the crossover design. The registry does not report the sequence allocation details or the interval between periods.

The enrollment is listed as 32. The primary pharmacokinetic results include 32 observations for the fixed-dose combination and 30 for the separate-tablet group. Those group counts are not the same as the enrollment field, and the record does not explain why the endpoint counts differ. The primary analysis population is called the pharmacokinetic set. It includes subjects in the trial set who had at least one observation for at least one primary pharmacokinetic endpoint and had no important protocol violations relevant to the statistical evaluation of pharmacokinetic endpoints.

The study started in 2014-01, had primary completion in 2014-02, and is listed as completed. Boehringer Ingelheim is named as the sponsor. The registry lists drug interventions including empagliflozin 2.5 mg, empagliflozin 10 mg, metformin 500 mg, and empagliflozin/metformin fixed-dose combination. The arm descriptions specify the fixed-dose combination as 12.5 mg empagliflozin / 500mg metformin and describe the comparator as empagliflozin and metformin tablets. The record does not give additional details about the dose presentation in the separate-tablet arm beyond that description.

Randomization concerns the order in which treatments are assigned; open label means participants and study personnel were not masked to treatment. In a crossover study, treatment comparisons can use within-participant information, but interpretation depends on period and sequence effects and on how those effects are handled analytically. The registry identifies ANOVA as the method for the reported geometric mean ratio analyses but does not give the model specification or describe handling of period, sequence, or carryover effects.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Treatment Period 1

Treatment Period 2

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Pharmacokinetic set (PKS): includes all subjects of the TS who provided at least 1 observation for at least 1 primary pharmacokinetic endpoint, and had no important protocol violations with respect to the statistical evaluation of pharmacokinetic endpoints.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: Treated set (TS): includes all subjects who were dispensed study medication and were documented to have received at least 1 dose of study medication.

6. Primary Endpoint

The registry lists the primary outcome as AUC (0-tz) (Area Under the Concentration-time Curve of Metformin in Plasma Over the Time Interval From 0 to the Last Quantifiable Data Point), measured over 1 hour (h) before drug administration and 20 minutes (min), 40min, 1h, 1h 30min, 2h, 2h 30min, 3h, 3h 30min, 4h, 5h, 6h, 8h, 10h, 12h, 24h, 34h, 48h and 72h after drug administration.

The primary endpoint is metformin AUC from time 0 to the last quantifiable data point, abbreviated in the registry as AUC (0-tz). AUC is the area under a concentration-time curve: it represents cumulative drug concentration in plasma across the measured interval, rather than concentration at a single blood draw. Here the measure runs from the time of dosing to the last data point at which metformin concentration could be quantified.

Blood sampling began 1 hour before administration and continued at 20 minutes, 40min, 1h, 1h 30min, 2h, 2h 30min, 3h, 3h 30min, 4h, 5h, 6h, 8h, 10h, 12h, 24h, 34h, 48h and 72h after administration. Those measurements provide the concentration-time observations from which the endpoint is derived. The registry reports the endpoint in ng*h/mL and summarizes it using a geometric mean and geometric coefficient of variation.

For a bioequivalence question, AUC is useful because it captures overall systemic exposure over the observed interval. Comparing exposure after the fixed-dose combination with exposure after the tablets administered together can assess whether the formulations yield similar metformin exposure in this study setting. The endpoint is pharmacokinetic: it does not itself measure symptoms, disease outcomes, or clinical benefit. A secondary outcome listed in the record is metformin AUC from time 0 extrapolated to infinity; its results are not described in the primary endpoint definition.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry identifies ANOVA as the method for both reported geometric mean ratio analyses and gives the geometric mean ratio as the analysis parameter. Further details of the model, transformations, hypotheses, and equivalence margins are not reported in the registry record. The following statistical explanations are general; they describe how these methods and quantities are commonly interpreted, not unreported features of this particular analysis.

What does a geometric mean represent?

Pharmacokinetic measures such as AUC are often analyzed on a logarithmic scale because concentrations and exposure measures are positive and can have right-skewed distributions. On that scale, multiplicative differences become additive, which can make model assumptions more reasonable. Converting an estimated mean back to the original scale yields a geometric mean. It is a measure of central tendency based on multiplicative rather than arithmetic averaging.

The posted geometric means are 6260.0 ng*h/mL for the fixed-dose combination and 6360.0 ng*h/mL for the separate tablets. A geometric mean ratio expresses one treatment’s geometric mean relative to the other. A ratio of 100 on the displayed scale represents equality. The reported estimates, 98.57 and 98.71, are near that equality reference. The registry does not state the direction used for the numerator and denominator in text, although it lists the treatment groups for each analysis.

What does ANOVA contribute?

Analysis of variance (ANOVA) is a modeling framework for comparing means while accounting for design structure and specified sources of variation. In a crossover study, a suitable model may account for repeated treatment observations from the same participant and for design features such as period and sequence. That structure matters because observations from one participant are generally not independent in the same way as observations from different participants. The registry reports ANOVA but does not specify its terms, covariance assumptions, or the details of how the crossover structure entered the model. Those particulars should not be inferred from the method label alone.

For a pharmacokinetic comparison, analysis is commonly performed after a logarithmic transformation, with the treatment contrast then converted into a geometric mean ratio. That is a general description of a common approach; the registry record does not explicitly describe the transformation used in this trial. The reported geometric mean ratio and confidence interval are the appropriate posted quantities to interpret without assuming further model details.

How should the confidence intervals be read?

A confidence interval gives a range of values compatible with the data and the statistical model at a stated confidence level. The registry reports 90% confidence intervals for both analyses: 94.662 to 102.639 and 94.783 to 102.796. In repeated use of the same procedure under its assumptions, 90% of intervals constructed in that way would cover the fixed underlying parameter. It is not a statement that there is a 90% probability that this particular interval contains the parameter.

For a ratio comparison, 100 is the equality reference. Both posted intervals include 100, so the estimates are not separated from equality by these intervals. An interval also conveys precision: narrower intervals indicate greater precision, while wider intervals indicate less. The bounds of these intervals alone do not determine whether equivalence has been demonstrated. That judgment requires prespecified equivalence bounds and the relevant analysis plan. Neither is reported in the registry record.

Why does the geometric coefficient of variation matter?

The geometric coefficient of variation describes relative spread on a multiplicative scale. It is not a confidence interval for a treatment ratio and should not be read as the uncertainty around the group’s geometric mean. The posted values are 20.4 for the fixed-dose combination and 21.8 for separate tablets. Greater variability generally makes it harder to estimate a treatment contrast precisely, all else being equal. In crossover studies, within-participant variability is especially relevant, but the record does not provide a specific within-participant variance estimate.

Over-dispersion is a general concern when observed variability exceeds what a statistical model assumes. If the model does not represent the variability adequately, standard errors and confidence intervals may be too small or otherwise miscalibrated. Whether over-dispersion was assessed or how variability components were modeled is not reported in the registry record. The group geometric coefficients of variation are descriptive and cannot by themselves establish that the model handled all sources of variation appropriately.

What does the p-value add?

A p-value quantifies how unusual data at least as discordant with a specified null hypothesis would be if that hypothesis and the model assumptions were true. It does not give the probability that the null hypothesis is true, nor does it measure the size or practical importance of a difference. Its interpretation depends on knowing the tested hypothesis. For an equivalence analysis, the relevant hypotheses concern whether the treatment contrast lies outside or within prespecified bounds; the p-value is not meaningful as a stand-alone test of bioequivalence without that context.

The registry reports p < 0.001 for one entry and p < 0.001 for the other. A formatted < 0.001 should not be read as an exact zero. The exact unrounded value, test hypothesis, and p-value calculation details are not reported. Consequently, the confidence intervals and p-values should be read as posted analysis outputs, with limits on what can be inferred about the formal equivalence procedure.

What can sample-size planning illustrations say?

Sample-size calculations based on an observed effect are planning illustrations. They can explore the number of participants per arm that might be needed under stated assumptions, such as a two-sided significance level of 0.05, equal allocation, a normal approximation, and an effect defined as a fraction of the observed effect. Such calculations are not a retrospective measure of the probability that this completed study detected its effect. They are not observed power.

The registry describes those planning assumptions but gives no numerical planning results in the record summarized for this analysis. Therefore no required participants-per-arm figure can be stated. The trial’s own sample-size justification is not reported in the registry record. A planning illustration based on an observed effect also inherits uncertainty in that observed effect, so it should not be treated as a definitive design recommendation.

10. Analyses Reported by the Sponsor

The registry reports two sponsor-associated ANOVA analyses of the geometric mean ratio: 98.57 with a 90% confidence interval from 94.662 to 102.639 and p < 0.001, and 98.71 with a 90% confidence interval from 94.783 to 102.796 and p < 0.001. These are the posted analyses to use when describing the sponsor’s reported results. The record does not explain why both entries are listed or specify the hypothesis associated with either p-value.

No separate recalculated estimate, interval, or p-value is reported in the registry record. Thus there is no numerical recalculation to compare with the sponsor’s analysis. In general, a recalculated comparison based only on posted summary statistics would be an unadjusted calculation and could omit the structure of a crossover analysis. It would not supersede a sponsor analysis that uses participant-level data and the prespecified model. Here, the registry does not identify a discrepancy between the sponsor’s analysis and a recalculation.

The sponsor’s reported estimates and intervals can be described directly, but the formal bioequivalence decision cannot be reconstructed without the equivalence margins and a fuller account of the analysis. The statistical interpretation rests on the posted confidence intervals and the limits of what the registry reports, not on an independently established decision rule.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists three outcome measures and six analyses, and it identifies one secondary outcome: metformin AUC from time 0 extrapolated to infinity. The record does not map each of the six analyses to a particular outcome measure or explain the relationship between the two reported geometric mean ratio entries. It also does not state whether the analyses were ordered as primary and secondary tests, or whether any multiplicity adjustment was used.

Multiplicity arises when several outcomes, comparisons, time points, or statistical tests are examined. The more analyses that contribute to a family of claims, the greater the chance of at least one apparently unusual result if no adjustment or hierarchy is used. This is a general statistical concern, not evidence that the reported analyses are invalid. The registry record does not report an adjustment method, testing hierarchy, or rationale for treating the analyses as separate or related.

The record identifies the pharmacokinetic set as the primary analysis population. It includes subjects in the trial set who had at least one observation for at least one primary pharmacokinetic endpoint and no important protocol violations relevant to statistical evaluation of pharmacokinetic endpoints. The endpoint arm counts are 32 for the fixed-dose combination and 30 for separate tablets. The reason for the difference between those counts is not reported in the registry record.

Beyond the pharmacokinetic-set definition, missing-data procedures are not reported. The record does not describe how missing concentration measurements, values below quantification limits, incomplete profiles, or missing endpoint values were handled. It also does not report sensitivity analyses for missing data or the number of participants excluded from the pharmacokinetic set. No assumptions about imputation or exclusion should be inferred from the endpoint counts alone.

The protocol’s statistical analysis plan, sample-size justification, interim analyses, and treatment of multiplicity are not reported in the registry record. Their absence limits assessment of whether the posted analyses follow prespecified decisions, but it does not establish that a particular procedure was or was not carried out.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent non-serious adverse event terms

The adverse-event figures list serious adverse events affecting 0 of 32 participants in the fixed-dose combination group and 0 of 30 in the separate-tablet group. These are the registry’s posted counts for the groups at risk. The registry lists no serious adverse event terms. Death counts are not reported in the registry record.

Other adverse events are listed for 11 of 32 participants in the fixed-dose combination group and 5 of 30 in the separate-tablet group. The record does not provide event terms, severity beyond the serious/non-serious grouping, timing, duration, or whether events were considered related to treatment. It also does not describe the rules used to collect or classify events. These counts should therefore be read as a brief registry summary, not as a complete account of safety findings.

Zero recorded serious events does not establish that a treatment has no risk. The study was conducted in healthy volunteers and focused on single-dose pharmacokinetics; its safety observations apply to the study conditions and recorded follow-up. The registry record does not give additional detail that would support a broader safety assessment or comparison of event rates beyond the listed counts.

13. What the Result Does, and Does Not, Mean

The result addresses a focused question about metformin exposure after a fixed-dose combination compared with empagliflozin and metformin tablets administered together. The primary measure is the area under the plasma concentration-time curve from dosing to the last quantifiable data point. In this healthy-volunteer, fed, single-dose crossover study, the posted geometric mean estimates were 6260.0 ng*h/mL and 6360.0 ng*h/mL, and the two reported ratio estimates were near 100 with 90% confidence intervals that included 100.

Those observations describe the trial’s pharmacokinetic comparison; they do not, on their own, settle whether formal bioequivalence criteria were met. That conclusion depends on the prespecified equivalence bounds and the analysis procedure. The registry labels the analyses as non-inferiority or equivalence analyses, but the bounds and the detailed hypotheses are not reported. The reported p-values cannot substitute for those details.

The result also does not establish how exposure would compare in people with a medical condition, under other meal conditions, after repeated dosing, or for outcomes other than the measured pharmacokinetic endpoints. The registry describes healthy male and female volunteers under fed conditions and a single-dose design. It does not report clinical efficacy outcomes. AUC is a drug-exposure measure, not a direct measure of symptom relief, disease control, or other clinical benefit.

The study’s design offers a within-participant comparison, which can reduce the influence of stable differences between participants in a crossover setting. At the same time, the interpretation depends on the sequence and period structure and on the analytic handling of those features. The registry names ANOVA but does not report its full specification, or details about the interval between periods or carryover assessment.

Accordingly, the most defensible reading is descriptive and bounded: the registry reports similar-looking geometric mean exposure summaries and ratio estimates near the equality reference, with confidence intervals that include it. How that evidence translates into a formal equivalence decision cannot be determined from the registry information alone. It should not be extended into a claim about clinical effectiveness or safety beyond what the registry reports.

14. Why This Trial Matters Statistically

This trial is statistically useful as an example of a formulation comparison where the endpoint is exposure rather than a clinical outcome. The distinction matters: the analysis asks whether metformin exposure differs between two ways of administering the drugs, not whether participants experience a change in health. AUC provides a cumulative exposure measure and the geometric mean ratio expresses the comparison on a multiplicative scale.

The crossover design is also instructive. Each treatment condition is part of a two-period, two-sequence design, so participant-level pairing can be important to the analysis. An analysis that ignores the design may not represent the uncertainty correctly. The registry names ANOVA but does not report the model terms, demonstrating why a method label alone may be insufficient to reproduce or fully assess a trial analysis.

The posted intervals highlight a second key point: evidence for equivalence is not the same as a failure to detect a difference. Formal equivalence relies on specified bounds and an interval-based decision. Here, the intervals include the equality reference, but the registry does not report the bounds needed to assess whether the intervals lie within acceptable limits. The result illustrates why confidence intervals, hypotheses, and decision margins need to be interpreted together.

Finally, the record distinguishes the trial enrollment from the counts appearing in the endpoint summaries and defines a pharmacokinetic analysis population. That information is relevant to understanding which participants contributed to the endpoint, while the reason for the different group counts and the missing-data procedures remain unreported. These features make this a useful case for reading registry results carefully: the estimates are informative, but their precision, analysis population, and decision context must all be considered.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

Several trial-specific details needed for a full statistical assessment are not reported in the registry record. The ANOVA model specification is not given, including how the crossover structure, period, and sequence were represented. The interval between treatment periods and any assessment or handling of carryover are also not reported. These omissions prevent a detailed evaluation of whether the analysis fully reflects the design.

The record does not report the equivalence margins, the specific null hypotheses tested, or the mapping between the two reported analyses and the listed outcome measures. The displayed p-values therefore lack enough context to establish their role in a formal equivalence decision. The record also does not identify the reason for the primary endpoint counts of 32 and 30, despite the enrollment field listing 32.

Missing-data handling is not described beyond the pharmacokinetic-set inclusion criteria. The record does not explain how incomplete concentration profiles or values that could not be quantified were handled, nor does it report sensitivity analyses or exclusions by reason. A sample-size justification, statistical analysis plan, and interim-analysis information are not reported in the registry record.

Generalizability is limited by the described study conditions: healthy male and female volunteers, fed conditions, and a single dose. The registry does not report clinical outcome measures. Safety reporting is also brief: serious-event counts and other-event counts are listed, but event terms and additional characterization are not. These limitations do not negate the posted pharmacokinetic summaries; they define what can and cannot be concluded from the registry information.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Bioequivalence Study in Healthy Volunteers for Empagliflozin/Metformin Fixed Dose Combination Compared to Separate Tablets — NCT02028767.
• Registry record last updated 2015-07-24; status last verified 2015-06.
• Results first posted 2015-07-24.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

NCT02028767 is a completed Phase 1 randomized, open-label, two-period, two-sequence crossover study sponsored by Boehringer Ingelheim. It compared a 12.5 mg empagliflozin / 500mg metformin fixed-dose combination with empagliflozin and metformin tablets administered together in healthy male and female volunteers under fed conditions after a single dose.

The primary endpoint was metformin AUC from time 0 to the last quantifiable data point, analyzed in the pharmacokinetic set. The registry reports geometric means of 6260.0 ng*h/mL for the fixed-dose combination and 6360.0 ng*h/mL for separate tablets, and two ANOVA geometric mean ratio analyses with 90% confidence intervals. Both reported intervals include 100. The registry does not report equivalence bounds or enough detail to reconstruct the formal decision procedure.

The registry lists 32 enrolled participants, endpoint group counts of 32 and 30, one secondary outcome, and six analyses. It reports no serious adverse events in either group, while other adverse events are listed for 11 and 5 participants. Additional detail on missing data, analysis specifications, event types, and the rationale for the endpoint count difference is not reported in the registry record.