Safety and Efficacy of Empagliflozin (BI 10773) and Sitagliptin Versus Placebo Over 76 Weeks in Patients With Type 2 Diabetes: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Safety and Efficacy of Empagliflozin (BI 10773) and Sitagliptin Versus Placebo Over 76 Weeks in Patients With Type 2 Diabetes (NCT01289990), a non-randomized phase 3 study. The primary outcome is Changes From Baseline in Glycosylated Haemoglobin (HbA1c) (%) After 52 Weeks of Treatment.

2. Clinical Question

3. Trial Design

The registry classifies this as a phase III, double-blind, parallel-group study with non-randomized allocation. It reports an enrollment of 2705 and a completed status. The recorded start is 2011-02 and primary completion is 2013-05. Those dates and the enrollment describe the registry record; they do not establish how many participants contributed to any particular endpoint analysis.

The title describes a placebo-controlled extension involving empagliflozin and sitagliptin over 76 weeks. The official title specifies a minimum 76 weeks, including 24 weeks of a preceding trial. The primary HbA1c endpoint, however, is change from baseline after 52 weeks of treatment. These timeframes refer to different aspects of the study and should not be conflated. The registry also lists secondary HbA1c change after 76 weeks and blood pressure and body-size outcomes measured at 52 or 76 weeks.

The study included low- and high-dose BI 10773 groups, placebo groups, and a sitagliptin group, with several background-therapy settings. That structure means there is not one universal comparison across all participants: the available groups depend on whether participants were drug naive or using a background therapy. The listed arm descriptions identify matching placebo tablets for some treatment settings.

Double masking describes the stated masking, while non-randomized describes allocation. These are distinct design features. Masking does not remove the potential for differences between groups that arise when treatment assignment is not randomized. The registry does not report how allocation was determined, or whether any statistical adjustment was intended to address group differences.

The primary population description refers to “all randomised patients” who received study drug and had a baseline HbA1c assessment, while the overall allocation field says non-randomized. This discrepancy in the registry record cannot be resolved from the available information. Accordingly, interpretation should retain the recorded design caution rather than assume randomized allocation.

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.

Only the first 6 of 13 groups are shown.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Full Analysis Set (FAS) which contained all randomised patients who received at least 1 dose of study drug and had a baseline HbA1c assessment, irrespective of participation in the extension trial. (LOCF:Last observation carried forward)

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: Full Analysis Set (FAS): which contained all randomised patients who received at least 1 dose of study drug and had a baseline HbA1c assessment, irrespective of participation in the extension trial.

Only the first 6 of 14 groups are shown.

6. Primary Endpoint

The registry lists the primary outcome as Changes From Baseline in Glycosylated Haemoglobin (HbA1c) (%) After 52 Weeks of Treatment, measured over Baseline and 52 weeks.

The primary endpoint measures change from baseline in glycosylated haemoglobin, or HbA1c, after 52 weeks of treatment. HbA1c is a continuous measure used to assess average blood glucose over time. A change score expresses how the measured value at follow-up differs from the participant’s baseline value. The endpoint is reported as a least-squares mean, with standard error as the stated measure of dispersion.

In general, a continuous endpoint can preserve information about the size of change, rather than reducing outcomes to a yes-or-no category. A mean describes the average change in the analysis population, while a standard error describes the estimated precision of that mean. Neither measure by itself gives the full range of participant-level outcomes. A confidence interval, when reported, would convey a range of values compatible with the data under the model and its assumptions.

The registry specifies a Full Analysis Set comprising patients who received at least one dose and had a baseline HbA1c assessment, irrespective of participation in the extension trial. It also states that the analysis used last observation carried forward. The actual least-squares mean changes, standard errors, between-group contrasts, confidence intervals, and p-values for the primary endpoint are not reported in the registry record. Thus, the endpoint definition is available, but its estimated results and precision cannot be described from this record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a least-squares mean represent?

A least-squares mean is a model-based estimate of an average. In general, a statistical model can account for specified factors when estimating group means, so the result need not be the same as a simple arithmetic average. The registry identifies the primary endpoint statistic as a least-squares mean, but does not report the model formula, the factors included, or the estimation details. It is therefore not possible to describe which adjustments, if any, were used for the primary result.

The endpoint is a change from baseline in HbA1c. Change scores relate follow-up values to baseline values, while a model-based mean summarizes estimated average change for a group. A between-group contrast would compare such estimated changes; it is distinct from the within-group change in each group. The contrast and its uncertainty are not reported in the registry record. No direction or magnitude of treatment difference can be inferred from the endpoint definition alone.

What does the standard error tell the reader?

A standard error describes the estimated precision of a statistical estimate, such as a mean. It is not the same as the standard deviation, which describes variability among observations. In general, the standard error tends to be smaller when an estimate is based on more information and when the underlying variability is lower, though the exact relationship depends on the model and data. The registry names standard error as the dispersion measure for the primary endpoint but does not provide the values.

A confidence interval is a general way to express uncertainty around an estimate. Its width reflects precision, conditional on the statistical model and assumptions. A confidence interval for a treatment contrast would help show which effect sizes are compatible with the observed data. No such primary interval is reported in the registry record. A p-value, when available, addresses compatibility with a specified null hypothesis under model assumptions; it does not show the size or clinical importance of an effect. The primary p-values are also not reported.

Why does allocation affect interpretation?

Randomization, in general, assigns participants by chance and supports comparisons by reducing systematic differences between groups in expectation. Here the allocation field is NON RANDOMIZED. The record also states double masking, but masking and allocation answer different design questions. Masking concerns knowledge of assigned treatment; it cannot ensure that groups began with comparable characteristics. The registry does not report the allocation procedure or whether analysis methods adjusted for baseline imbalances. Group comparisons should therefore be read as potentially affected by confounding rather than automatically interpreted as causal treatment effects.

The record has a further wording inconsistency: the Full Analysis Set is described as consisting of “all randomised patients,” whereas the study allocation is marked non-randomized. The registry does not explain this discrepancy. It is not appropriate to select one description as definitive or infer that allocation was randomized from the population wording alone.

How should the analysis population and imputation be understood?

The primary population includes patients who received at least one dose and had a baseline HbA1c assessment, regardless of participation in the extension trial. This definition indicates that both treatment exposure and baseline measurement were required for inclusion in the described set. It does not tell the reader how many individuals contributed to each particular analysis, how missing follow-up differed between groups, or how other exclusions were handled.

The registry states that last observation carried forward was used. In general, this method substitutes a participant’s most recent recorded outcome for a later missing value. It can retain participants with incomplete follow-up in an analysis, but it relies on an assumption that the carried-forward value is a useful representation of the missing later measurement. If outcomes change over time, carrying an earlier value forward may fail to represent that change. This approach also does not fully reflect uncertainty about the unobserved value. The registry does not report the amount or pattern of missing HbA1c data, or sensitivity analyses using alternative assumptions.

What does multiplicity mean in this study?

Multiplicity arises when investigators examine several outcomes, timepoints, treatment groups, or contrasts. In general, if many hypotheses are tested, the probability of at least one small p-value under a collection of null effects can rise. A prespecified testing sequence or adjustment can address this issue, but the registry does not report whether one was used. It lists 13 outcome measures and 143 analyses; those counts indicate a broad analysis record, but do not identify which comparisons were prespecified or how their results should be prioritized.

The registry describes the primary HbA1c timepoint, which helps identify the principal endpoint, but it does not provide a statistical analysis plan or multiplicity strategy. Consequently, readers cannot determine from this record whether other analyses were confirmatory, exploratory, or adjusted for multiple testing.

What do sample-size planning calculations mean?

In general, a planning calculation estimates how many participants may be needed to detect a specified effect under stated assumptions. The registry’s planning assumptions describe a two-sided significance level of 0.05, equal allocation between two arms, a normal approximation, and an effect defined as a fraction of the observed effect. Such calculations are illustrations, not observed trial results and not observed power. They do not establish how the original study was designed. The registry does not report the protocol’s sample-size justification, and the planning assumptions do not establish that the study used them.

For continuous outcomes, variation in the measure and the size of the effect being targeted influence the required sample size in general. A planning scenario based on a fraction of an observed effect is conditional on that effect estimate and should not be mistaken for independent confirmation. Numerical planning results are not reported in the registry record addressed. No sample-size requirement or achieved power can be inferred from the enrollment figure alone.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("BI 10773 Low (Drug Naive)", "BI 10773 High (Drug Naive)", "Placebo (Drug Naive)", "Sitagliptin 100mg (Drug Naive)", "BI 10773 Low (Pioglitazone)", "BI 10773 High (Pioglitazone)")
n   <- c(224, 224, 228, 223, 165, 168)
m   <- c(-0.7, -0.82, 0.09, -0.58, -0.63, -0.71)
sp  <- c(0.05, 0.05, 0.05, 0.05, 0.07, 0.07)
sd  <- sp * sqrt(n)      # standard errors converted to SDs
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  BI 10773 Low (Drug Naive)  224  -0.700  0.748  -0.799  -0.601
CARM  BI 10773 High (Drug Naive)  224  -0.820  0.748  -0.919  -0.721
CARM  Placebo (Drug Naive)  228  0.090  0.755  -0.009  0.189
CARM  Sitagliptin 100mg (Drug Naive)  223  -0.580  0.747  -0.679  -0.481
CARM  BI 10773 Low (Pioglitazone)  165  -0.630  0.899  -0.768  -0.492
CARM  BI 10773 High (Pioglitazone)  168  -0.710  0.907  -0.848  -0.572

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 13 outcome measures and 143 analyses. It identifies change from baseline in HbA1c after 52 weeks as the primary endpoint, while also listing HbA1c at 76 weeks and outcomes involving blood pressure, body weight, and waist circumference. With many outcomes and analyses, multiplicity is a concern in general: repeated testing can increase the chance of a nominally notable result even when there is no corresponding effect. The registry does not report a multiplicity adjustment, a testing hierarchy, or which analyses were prespecified.

The primary analysis population is described as the Full Analysis Set: patients who received at least one dose and had a baseline HbA1c assessment, irrespective of participation in the extension trial. The description also calls these patients randomised, despite the allocation field stating non-randomized. The record does not resolve this difference in terminology, report the number included in each endpoint analysis, or explain how exclusions from the analysis set were handled.

Last observation carried forward is specified for the primary endpoint. In general, this method fills a missing later measurement using the participant’s most recent recorded value. Its validity depends on assumptions about how outcomes would have changed after that last observation. The registry does not report missing-data counts or patterns, reasons for missingness, or sensitivity analyses testing alternative approaches. The model specification and the statistical plan are also not reported in the registry record. As a result, the robustness of the primary estimate to missing-data assumptions and analytic choices cannot be assessed from this information.

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 serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: cellulitis (1.2%), coronary artery disease (1.0%), osteoarthritis (1.0%), nephrolithiasis (1.0%), gastroenteritis (0.9%).

The registry reports, for each listed group, the number of participants at risk and the number affected by serious adverse events, as well as corresponding counts for other adverse events. These are descriptive counts. For example, the drug-naive empagliflozin 10 mg group has 229 at risk and 23 with a serious adverse event; the corresponding placebo group has 223 at risk and 16 with a serious adverse event. These figures do not establish that a treatment caused an event or that group differences are statistically reliable.

Counts should be interpreted in the context of each group’s own at-risk figure and clinical setting. The listed groups include drug-naive participants and participants taking pioglitazone, metformin, or metformin plus sulfonylurea. They are not all interchangeable comparisons. Because allocation is marked non-randomized, differences in counts may also reflect differences between the groups. No comparative statistical analysis of these safety counts is reported in the registry record.

The most frequently listed serious-event terms are cellulitis at 1.2%, coronary artery disease at 1.0%, osteoarthritis at 1.0%, nephrolithiasis at 1.0%, gastroenteritis at 0.9%, and cerebral infarction at 0.9%. The record does not give an arm-specific breakdown for these terms. Deaths are not reported in the registry record’s group entries. These summaries are not a complete account of clinical context, event timing, or causal attribution.

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

The registry record establishes that this was a phase III, double-blind, parallel-group extension study of BI 10773 and sitagliptin with placebo groups. It identifies change from baseline in HbA1c after 52 weeks as the primary endpoint and specifies least-squares means with standard errors. It also describes a Full Analysis Set and use of last observation carried forward. These details explain what the primary analysis was intended to measure and how the registry says missing values were handled.

The primary endpoint estimates and comparisons are not reported in the registry record. It is therefore not possible to state the amount or direction of HbA1c change, how groups differed, or how precise any treatment contrast was. The stated endpoint and study title do not establish the results. Similarly, the adverse-event counts describe events reported by group but do not establish that any treatment caused or prevented them.

The study’s allocation is marked non-randomized. In general, this makes causal interpretation more difficult because participants in different groups may differ in ways that affect outcomes. The double-blind designation does not remove that concern. The record’s use of “randomised” in the Full Analysis Set description conflicts with the allocation field, and the discrepancy is not explained. No allocation procedure or adjustment strategy is reported.

Nor does the record establish the adequacy of its handling of missing outcomes or multiplicity. Last observation carried forward has assumptions that may not hold when outcomes change over time, and details of missingness or alternative analyses are not reported. The record identifies many outcomes and analyses but does not report a multiplicity strategy. Thus, the information supports a careful description of design, endpoint definitions, and adverse-event reporting, but not a conclusion about efficacy, comparative safety, or clinical importance.

14. Why This Trial Matters Statistically

This trial is statistically notable for the combination of a defined continuous primary endpoint and a complex set of treatment contexts. The primary measure is HbA1c change from baseline after 52 weeks, while the study title and official title describe an extension period of at least 76 weeks, including 24 weeks from a preceding trial. Keeping endpoint timing distinct from the overall study duration is essential when interpreting what each result would represent.

The groups span drug-naive participants and people receiving different background therapies. Consequently, a comparison in one setting should not automatically be generalized to another. The non-randomized allocation designation adds a further interpretive limitation: treatment groups may not be comparable at baseline, even though the study is described as double blind. The registry’s conflicting use of “randomised” in the analysis-population description makes transparent reporting of design details particularly important.

The record also illustrates why an endpoint definition is not the same as an endpoint result. It specifies a least-squares mean and standard error, but does not report the estimated HbA1c changes or comparisons. Without estimates and uncertainty, readers cannot assess the magnitude, direction, or precision of the primary finding. Reporting these quantities alongside a clear analysis model would support a more informative statistical interpretation.

Finally, the record lists 13 outcome measures and 143 analyses, as well as a last-observation-carried-forward approach. Multiplicity and missing-data assumptions can affect interpretation even when the outcome and analysis population are clearly named. The registry does not report how these issues were addressed beyond the stated imputation method. These are useful reminders that statistical conclusions depend not only on the endpoint, but also on allocation, analysis choices, missingness, and the treatment of multiple comparisons.

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.

The primary outcome is defined, but its results are not reported in the registry record. There are no stated least-squares mean changes, treatment contrasts, standard errors, confidence intervals, or p-values for the primary HbA1c endpoint. Accordingly, the magnitude and precision of the primary comparison cannot be assessed from the registry information.

The allocation field marks the study as non-randomized, while the analysis-population description refers to “all randomised patients.” The record does not explain this inconsistency. It also does not describe the allocation procedure, baseline balance, or any adjustment for group differences. The double-blind designation does not resolve these limitations for comparisons between groups.

The Full Analysis Set requires at least one dose and a baseline HbA1c assessment and includes patients irrespective of participation in the extension trial. The record does not state the number included in each analysis or the extent of follow-up. It specifies last observation carried forward but does not report missing-data patterns, reasons for missingness, or sensitivity analyses. The statistical model and its covariates are not reported either.

Although the registry lists 13 outcome measures and 143 analyses, it does not describe a multiplicity procedure or identify the confirmatory analysis hierarchy. The safety information is also limited: event counts are reported by group, but deaths are not reported in the registry record’s group entries, and the listed serious-event terms are not broken down by group. These constraints limit comparative interpretation of both efficacy and safety.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Safety and Efficacy of Empagliflozin (BI 10773) and Sitagliptin Versus Placebo Over 76 Weeks in Patients With Type 2 Diabetes — NCT01289990.
• Registry record last updated 2014-07-15; status last verified 2014-07.
• Results first posted 2014-07-15.
• 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

The registry describes a completed phase III, double-blind, parallel-group study with non-randomized allocation and enrollment of 2705. It concerns BI 10773, identified as empagliflozin in the title, with placebo groups and a sitagliptin group in specified treatment settings.

The primary endpoint is change from baseline in HbA1c after 52 weeks, summarized as a least-squares mean with standard error. The primary population is described as a Full Analysis Set, with last observation carried forward for missing data. The primary estimates and comparisons are not reported in the registry record.

The record lists 13 outcome measures and 143 analyses, along with serious and other adverse-event counts by group. It does not report a multiplicity strategy, detailed missing-data patterns, or deaths in the group entries. The allocation designation and analysis-population wording are inconsistent, which further limits interpretation of comparisons.