Foxiga Korea Local Phase 4 Study: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Foxiga Korea Local Phase 4 Study (NCT02564926), a randomized phase 4 study. The primary outcome is Adjusted Mean Changes From Baseline in Total Body Fat Mass by DXA Scan.

2. Clinical Question

3. Trial Design

This was a randomized, open-label, parallel-group, multicentre phase IV study of dapagliflozin compared with glimepiride in people with type 2 diabetes inadequately controlled with metformin. Both study regimens included metformin 1000mg: dapagliflozin was given at 10mg, while glimepiride was given at 1-2mg. The registry describes the study as lasting 12 months and identifies Week 52 as the primary endpoint timeframe.

The trial is marked completed. The recorded study start was 2016-01-05, and primary completion was 2018-01-15. Enrollment was 125. Those are registry-level design and status details; they do not establish how many participants contributed to every analysis. The primary outcome reports n 56 in each arm, while the safety summaries report at-risk denominators of 60 and 61. The registry record does not explain the differences among these reported denominators or provide a participant flow account in the information available for this analysis.

Random allocation is intended to reduce systematic differences between groups at assignment, although it does not guarantee that the groups are identical. The open-label design means treatment masking was not used. Knowledge of assigned treatment can affect participant behavior, clinical management or assessment, depending on how outcomes are collected. DXA provides an instrumental measure of body composition, but the record does not state whether image assessment was blinded or give further details of measurement procedures. Those points are not reported in the registry record.

The record identifies a phase IV study and an active-drug comparison, rather than a comparison with no treatment. The resulting contrast is therefore about the assigned regimens in this trial context. Generalizability beyond the described Korean population, background metformin use and study follow-up cannot be established from the registry summary alone.

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.

Before assignment: 178 patients provided informed consent. 121 patients received study treatment and were included in the SAF population. 112 patients received study treatment and had at least one post-baseline primary efficacy assessment and were included in the FAS population (56 patients in each treatment group).

Overall Study

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.

6. Primary Endpoint

The registry lists the primary outcome as Adjusted Mean Changes From Baseline in Total Body Fat Mass by DXA Scan, measured over From baseline to Week 52.

The primary endpoint was adjusted mean change from baseline in total body fat mass measured by DXA, with the timeframe from baseline to Week 52. The registry reports the endpoint in grams and presents arm-level least-squares mean changes with 95% confidence intervals. It also reports a between-group LS mean difference in kg. These are descriptions of the same primary outcome as recorded in the registry; the record does not provide further detail on the adjustment model in the information available for this analysis.

Total body fat mass is a continuous measure of body composition. DXA is used to assess body composition, and measuring change allows the analysis to describe how the outcome moved relative to baseline over follow-up. An adjusted mean aims to account for specified variables in the analysis model, but which covariates were included is not reported in the registry record. The endpoint is not a direct measure of every aspect of diabetes control or of how participants feel or function.

In general, an interval estimate communicates uncertainty around an estimated average change or contrast. A narrower interval indicates greater precision under the model and assumptions used; an interval that excludes no difference is conventionally inconsistent with that null value at the stated confidence level. It does not establish the size of a clinically important effect. The registry does not report a threshold for clinical importance for this endpoint, so statistical evidence and clinical importance should not be treated as interchangeable.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: it is a measured quantity that can take a range of values, rather than a count of participants meeting a yes-or-no criterion. The registry reports adjusted mean changes, also called least-squares means. In general, a least-squares mean is a model-based average adjusted for variables included in the model. It is not necessarily the same as a raw arithmetic average. The registry does not identify the adjustment variables for the primary endpoint, so the exact interpretation of that adjustment cannot be reconstructed from the record.

What does the confidence interval describe?

A confidence interval expresses sampling uncertainty around an estimate, conditional on the model and assumptions used. The primary arm-level results are each accompanied by a 95% confidence interval, and the reported between-group LS mean difference also has a 95% confidence interval. The interval for the contrast is the direct uncertainty statement about the estimated difference between regimens. In general, an interval excluding zero for a difference is evidence against a zero-difference value at the corresponding confidence level, under the analysis assumptions. It does not quantify the probability that a particular true value lies inside the interval, nor does it resolve whether a difference matters clinically.

The arm-specific means and their intervals answer a different question from the between-group contrast. Each arm estimate describes change in that group; the contrast compares groups. A reader should not treat the arm-level intervals as substitutes for the interval on the comparison. The registry supplies the sponsor's contrast estimate, interval and p-value, which should be read as the trial's reported inferential analysis.

What does the p-value add?

A p-value describes how unusual data at least as incompatible with the null model would be if that model and its assumptions held. The reported primary p-value is p < 0.001. It is not the probability that the null hypothesis is true, and it does not measure the size or importance of the observed contrast. The magnitude and precision are better conveyed by the estimated difference and its confidence interval, considered alongside the endpoint's context.

How should the secondary binary outcome be read?

The HbA1c threshold outcome is binary: participants either met the registry's stated threshold or did not. The posted counts are 27 of 56 for dapagliflozin and 22 of 56 for glimepiride. A recalculated risk difference expresses the absolute difference in proportions on a probability scale, with its confidence interval. The estimate is -0.089 for glimepiride relative to dapagliflozin, with a 95% confidence interval from -0.272 to 0.094, and unadjusted p = 0.341. The interval includes zero, so this unadjusted calculation does not rule out no difference at the conventional level. The registry also reports a logistic-regression odds ratio, which uses a different scale and model. Odds ratios and risk differences are not interchangeable measures, and their numerical values should not be compared as though they were the same quantity.

The recalculated binary comparison is an unadjusted analysis from posted summary counts. It does not reproduce the sponsor's logistic regression and does not account for any adjustment variables. The posted counts and the sponsor's model-based result therefore answer related, but not identical, statistical questions.

What are the planning illustrations?

General sample-size planning uses assumptions about the target effect, variability or event frequency, significance level, allocation and desired power. The registry's planning illustrations are based on the observed effect, use a two-sided significance level of 0.05, equal allocation, and a normal approximation. They consider planning for 100%, 75% and 50% of the observed effect. These are hypothetical design calculations, not evidence about the trial's actual power and not a sample-size justification documented by the study team. Smaller target effects generally require more participants for a fixed power under otherwise similar assumptions; this is a general statistical relationship, not a finding about the trial's design.

Planning quantities must not be mistaken for observed group sizes or for a retrospective probability that the study detected an effect. The trial's recorded enrollment and the denominators appearing in endpoint and safety results describe different aspects of the record. The registry does not report the original sample-size justification or further details of its planning assumptions.

Why do model and reporting details matter?

Inference depends on the analysis model, endpoint definition, assumptions about observations and handling of incomplete data. For continuous outcomes, model-based estimates can depend on covariate choices and variance assumptions. For binary outcomes, a logistic model estimates an odds ratio, whereas a risk-difference calculation compares probabilities directly. In general, differences in estimands and models can produce different numerical summaries without indicating a transcription error.

The record identifies ANCOVA for the primary comparison and logistic regression for the reported HbA1c analysis. It does not provide the full model specifications, covariates, diagnostics, missing-data method, analysis-population definition or multiplicity procedure. Consequently, the displayed inferential results can be described, but the detailed analysis cannot be independently reproduced from this record alone.

10. Analyses Reported by the Sponsor

For the primary outcome, the registry reports the sponsor's ANCOVA result as an LS mean difference of -2.6 kg, with a 95% confidence interval from -3.346 to -1.819 and p < 0.001. The arm-level adjusted changes are reported in grams, while the comparison is reported in kg. The registry supplies no further detail that would justify independently reconstructing the model-based contrast from those summaries, so the sponsor's reported analysis is the appropriate source for that inferential result.

The recalculated result for the secondary HbA1c threshold is an unadjusted comparison from posted counts. It gives a risk difference of -0.089, with a 95% confidence interval from -0.272 to 0.094 and p = 0.341. The sponsor reports logistic regression for that outcome, with an odds ratio of 1.45, a 95% confidence interval from 0.673 to 3.113 and p = 0.343. The different effect measures and methods mean the estimates are not expected to match exactly. Their p-values are similar, but one is unadjusted and the other is model-based.

No discrepancy is flagged for the recalculation in the registry summary. That does not make the recalculation a replacement for the sponsor's prespecified analysis: it remains a simple check based on summary statistics and cannot reproduce adjustments or other model details that are not reported. For the primary endpoint, no corresponding recalculation is reported in this analysis; the sponsor's ANCOVA result remains the central comparative estimate.

11. Multiplicity, Missing Data and Analysis Populations

The record identifies a primary endpoint and 14 secondary outcomes, and it reports 16 outcome measures and 16 analyses. It does not report a multiplicity strategy, such as an adjustment across tests, a hierarchy for testing endpoints, or a distinction between confirmatory and exploratory secondary analyses. Without that information, the nominal p-values should be understood as specific to the analyses reported, not as evidence that the full set of outcomes was protected against false-positive findings.

In general, evaluating many endpoints creates a multiple-testing problem: when each test is considered separately, the chance of at least one apparently unusual result can rise across the family of tests. Methods such as prespecified hierarchies or adjusted testing can address this, but the registry record does not establish whether any such approach was used. It would be inappropriate to assume either that adjustment was performed or that it was omitted from the trial's full analysis solely because the registry summary does not describe it.

The registry also does not report how missing outcome data were handled, whether participants were analyzed according to randomized assignment, or how the analysis populations were defined. These details matter because incomplete follow-up or exclusions can affect the estimate and its uncertainty. The primary outcome lists n 56 in each arm, whereas the safety summaries list at-risk denominators of 60 and 61. The record does not explain the relationship among these denominators, and they should not be combined or treated as a participant-flow account.

The registry states enrollment of 125 but does not provide a detailed account of screening, randomization, withdrawals, endpoint availability or reasons for missing data in the information available for this analysis. Nor does it report sensitivity analyses that might show how conclusions change under alternative assumptions about missing outcomes. These omissions limit assessment of potential attrition bias and robustness. They do not demonstrate that missing data or multiplicity compromised the findings; they mean the relevant methods cannot be evaluated from the registry record.

12. Safety Results

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

Deaths

Other (non-serious) adverse events

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

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: rib fracture (1.7%), splenic injury (1.7%), subdural haemorrhage (1.7%), intervertebral disc protrusion (1.7%), transient ischaemic attack (1.7%).

The safety summaries report serious adverse events in 6 of 60 at-risk participants in the dapagliflozin group and 6 of 61 in the glimepiride group. Other adverse events are reported in 11 of 60 and 10 of 61, respectively. The registry reports zero deaths in each group. These are counts with the stated at-risk denominators; they should be read as descriptive safety summaries rather than as adjusted comparative estimates.

The registry lists rib fracture, splenic injury, subdural haemorrhage, intervertebral disc protrusion, transient ischaemic attack and conversion disorder among the serious adverse-event terms, each reported as 1.7%. The record does not provide enough detail to establish the timing, duration, relatedness to treatment or clinical circumstances of each event. A serious-event term does not, by itself, establish that the study drug caused the event.

Similar counts do not prove equal safety, and different counts do not establish a treatment-related difference. Safety comparisons require attention to exposure time, definitions, ascertainment, severity and uncertainty, none of which are fully described in the registry summary. The reported figures characterize events recorded during the trial; they are not advice about treatment and do not establish safety beyond the population and follow-up studied.

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

The primary result is evidence of a between-group difference in adjusted change in total body fat mass over the recorded follow-up. The arm-level estimates move in opposite directions, and the sponsor's reported ANCOVA contrast has a 95% confidence interval below zero with p < 0.001. Statistically, this is inconsistent with a zero contrast under the reported model and its assumptions.

That conclusion is limited to the endpoint and comparison studied. It does not show that every aspect of body composition changed, that the result applies to outcomes not measured, or that the difference is clinically important. The registry does not report a threshold for clinical importance for total body fat mass. Statistical significance alone cannot supply such a threshold.

The trial compared regimens that both included metformin, in Korean patients with type 2 diabetes inadequately controlled with metformin, and assessed outcomes through Week 52. It does not establish the same effects for other populations, background therapies, doses, treatment durations or clinical settings. The study was open-label, and the registry does not report whether DXA assessment was blinded.

The primary body-fat finding should also be kept separate from the secondary HbA1c threshold result. The latter has an unadjusted interval that includes no difference and a sponsor-reported logistic-regression interval that includes the null odds ratio. Different endpoints can yield different statistical evidence; one result neither validates nor negates the other.

Finally, the registry summary does not describe missing-data handling, analysis populations, multiplicity control or full model specifications. These gaps limit independent assessment of robustness and of the broader pattern across outcomes. The reported result is informative about the specific adjusted comparison, but it is not proof of a mechanism, a universal effect, or a participant-level response.

14. Why This Trial Matters Statistically

This trial is statistically notable because its primary endpoint is a continuous body-composition measure, while the posted comparison is presented both as adjusted arm-level changes and as a model-based between-group contrast. That makes the distinction between within-arm change and between-arm evidence especially important. The contrast, rather than separate changes viewed in isolation, addresses the randomized comparison directly.

The study also illustrates that different outcomes need different estimands and models. Total body fat mass is analyzed as a continuous measure, whereas the HbA1c threshold is binary. The primary ANCOVA result is a difference in adjusted means; the sponsor's logistic regression reports an odds ratio; the recalculated binary analysis reports a risk difference. Each summary has a distinct interpretation, and the scales cannot be substituted for one another.

Another statistical lesson is the importance of precision and context. The primary confidence interval describes uncertainty around the contrast, whereas a p-value alone does not convey the plausible range of its magnitude. Conversely, statistical evidence does not state whether that magnitude is clinically important. The registry does not provide a clinical-importance threshold for the primary endpoint.

Finally, this record makes visible how design and reporting details shape interpretation. The randomized, parallel-group design supports a comparative question, but open-label conduct, differences among reported denominators, and unavailable information on missing data, multiplicity and model specification constrain what can be independently assessed. Clear reporting of those elements would help readers judge the strength and scope of the evidence without confusing a registry result with a complete analysis report.

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 reporting gaps constrain interpretation. The registry describes randomization and an open-label parallel-group design, but it does not report the randomization procedure, allocation concealment or whether outcome assessors were masked. Since the primary endpoint was measured by DXA, details about acquisition, reading and quality control would help readers assess the measurement process; those details are not reported in the registry record.

The primary endpoint is described as adjusted, and the sponsor identifies ANCOVA, but the covariates, model specification and assumptions are not reported in the information available for this analysis. The registry also does not give a clinical-importance threshold for total body fat mass. Thus, the estimate and its statistical precision can be described, but clinical interpretation cannot be anchored to a stated threshold.

There is limited information about participant disposition and analytic populations. Enrollment is recorded as 125, while the primary endpoint and safety summaries have different denominators. The registry does not explain the difference or report the extent and reasons for missing endpoint data, withdrawals or exclusions. Without missing-data methods and sensitivity analyses, the effect of incomplete follow-up cannot be assessed.

Multiplicity handling is also not reported, despite the registry listing 14 secondary outcomes. This makes it difficult to distinguish confirmatory secondary evidence from exploratory findings or assess whether reported secondary p-values account for multiple testing. The registry's outcome list should therefore not be read as a complete interpretation of all results.

Finally, follow-up is described as 12 months, with the primary measure assessed through Week 52. The record does not establish longer-term outcomes. The open-label design, endpoint reporting gaps and lack of detailed analysis information all warrant caution, without proving that any particular bias occurred or invalidating the posted result.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Foxiga Korea Local Phase 4 Study — NCT02564926.
• Registry record last updated 2019-08-20; status last verified 2019-08.
• Results first posted 2019-08-20.
• 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

NCT02564926 is the registry identifier for the completed Foxiga Korea Local Phase 4 Study. The randomized, open-label, parallel-group, multicentre trial compared dapagliflozin 10mg plus metformin 1000mg with glimepiride 1-2mg plus metformin 1000mg in Korean patients with type 2 diabetes inadequately controlled with metformin. Enrollment was 125; the recorded study start was 2016-01-05 and primary completion was 2018-01-15.

The primary endpoint was adjusted mean change in total body fat mass by DXA from baseline to Week 52. The arm-level estimates were -1486 grams for dapagliflozin and 1096 grams for glimepiride. The reported ANCOVA contrast was -2.6 kg, with a 95% confidence interval from -3.346 to -1.819 and p < 0.001. The registry also lists secondary glycemic, body-composition and blood-pressure outcomes. Further details on multiplicity, missing data, analysis populations and full model specifications are not reported in the registry record.