This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
PIONEER 12 was a randomized, parallel, phase 3 trial evaluating oral semaglutide and sitagliptin in people with type 2 diabetes. The registry describes four interventions, quadruple masking, a treatment purpose, and a total enrollment of 1441 participants.
| Feature | PIONEER 12 |
|---|---|
| Phase | Phase 3 |
| Condition | Diabetes Mellitus, Type 2 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 1441 |
| Primary endpoint | Change From Baseline to Week 26 in Glycated Haemoglobin (HbA1c) (%) |
| Primary endpoint type | Binary |
| Hypothesis type | Superiority |
| Statistical method | MMRM (mixed model for repeated measures) |
| Effect measure | Mean difference |
| Trial status | Completed |
| Start | 2019-07-29 |
| Primary completion | 2021-10-27 |
| Lead sponsor | Novo Nordisk A/S |
| Sponsor type | Industry |
2. Clinical Question
The primary statistical question was whether treatment groups differed in change from baseline to week 26 in glycated haemoglobin (HbA1c). The registry identifies the hypothesis type as superiority and reports three primary endpoint analyses using a mixed model for repeated measurements.
Population
People with Diabetes Mellitus, Type 2 enrolled in a randomized phase 3 treatment trial.
Intervention
Oral semaglutide, represented in the reported analyses by 3 mg, 7 mg, and 14 mg groups.
Comparator
Sitagliptin 100 mg in the reported primary comparisons where a comparator is specified.
Primary question
How does change in HbA1c from baseline to week 26 differ between the reported oral semaglutide and sitagliptin treatment comparisons?
3. Trial Design
Active semaglutide interventions
- Oral semaglutide 3 mg
- Oral semaglutide 7 mg
- Oral semaglutide 14 mg
Active comparator intervention
- Sitagliptin 100 mg
- Placebo (oral semaglutide)
- Placebo (sitagliptin)
The intervention list contains oral semaglutide, sitagliptin, placebo for oral semaglutide, and placebo for sitagliptin. The registry therefore describes a masked parallel-group structure in which matching placebo interventions were part of the trial design.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Change From Baseline to Week 26 in Glycated Haemoglobin (HbA1c) (%) | Change in HbA1c from baseline to week 26 in percentage (%) point of HbA1c is presented. The outcome data was evaluated based on the on-treatment without rescue medication observation period and in-trial observation period. | From baseline to week 26 |
Observation periods
The registry defines two observation frameworks for the endpoint. The on-treatment without rescue medication observation period runs from the date of first dose of trial product following randomization through the first date of any of the following: the last dose of trial product plus 3 days or initiation of rescue medication.
The in-trial observation period is the time period in which participants were considered to be in the trial, regardless of discontinuation of trial product or initiation of rescue medication.
These two frameworks answer somewhat different questions. The on-treatment framework focuses on observations while treatment exposure remains relevant and before rescue medication enters the specified observation window. The in-trial framework retains the broader trial-period perspective, including participants after discontinuation or initiation of rescue medication.
5. Statistical Methodology
Full analysis set
The reported analyses use the full analysis set (FAS), which the registry defines as including all randomized participants. The number analyzed for the endpoint is the number of participants with available data for the outcome measure and, for each specified category, the number with available data in that category.
This distinction is important. Randomization establishes the treatment comparison, while the actual number contributing an observed HbA1c measurement at a particular analysis time can be smaller because outcome data are not necessarily available for every randomized participant.
Mixed model for repeated measurements
The registry reports a mixed model for repeated measurements (MMRM) as the method for each of the three primary analyses. HbA1c is measured longitudinally, so an MMRM is naturally suited to an endpoint defined by change from baseline to a specified follow-up time while using repeated observations across the study period.
For the reported primary analyses, the effect measure is a mean difference in percentage points of HbA1c. A negative estimate indicates a lower estimated HbA1c change in the first-named treatment comparison relative to the comparator under the model.
Superiority hypothesis
The registry identifies the hypothesis type as superiority. Thus, the inferential question is whether the estimated treatment difference provides evidence of a difference in the prespecified direction rather than whether a treatment is merely close enough to a comparator to satisfy a non-inferiority margin.
Confidence intervals
All three posted primary analyses report two-sided 95% confidence intervals. A confidence interval describes statistical uncertainty around the estimated mean difference. It does not describe the range of individual patient responses and does not give the probability that the true effect lies inside the particular interval.
P-values
The reported p-values quantify the compatibility of the observed data with the null hypothesis under the specified statistical model. They do not measure the magnitude of the treatment effect. The estimated mean difference and its confidence interval are therefore essential companions to each p-value.
6. Primary Results: HbA1c at Week 26
The registry posts three statistical analyses for the primary endpoint, all using MMRM and the full analysis set. The analyses evaluate the change from baseline to week 26 in HbA1c.
Oral Semaglutide 3 mg vs Sitagliptin 100 mg
Mean difference in HbA1c change
95% CI: −0.3 to −0.1 · P = 0.0078
Outcome unit: percentage point of HbA1C
The estimated mean difference was −0.2 percentage point of HbA1c for oral semaglutide 3 mg versus sitagliptin 100 mg. In practical statistical terms, the model estimated a 0.2-percentage-point lower change in HbA1c for the first-named treatment comparison relative to sitagliptin.
The estimate does not mean that every participant experienced a 0.2-percentage-point difference, nor does it describe the response of an individual patient. It is a population-level model-based comparison of mean change.
The two-sided 95% CI of −0.3 to −0.1 describes the precision of the estimated mean difference under the analysis framework. The interval is relatively narrow around the point estimate, but it still represents uncertainty rather than a distribution of individual treatment effects.
The p-value of 0.0078 is evidence against the null hypothesis under the specified model. It does not indicate that the effect is "0.78%" in size, nor does it quantify clinical importance. The effect estimate and confidence interval are the appropriate quantities for understanding magnitude and precision.
Oral Semaglutide 7 mg
Mean difference in HbA1c change
95% CI: −0.8 to −0.6 · P < 0.0001
Outcome unit: percentage point of HbA1C
The registry reports an estimated mean difference of −0.7 percentage point of HbA1c for the oral semaglutide 7 mg analysis. The corresponding two-sided 95% CI is −0.8 to −0.6, indicating substantially more separation from zero than the 3 mg analysis on the reported scale.
The registry's analysis field lists Oral Semaglutide 7 mg under "Groups Compared" but does not state the comparator in that field. The broader ClinicalTrials.gov record identifies sitagliptin 100 mg as the active comparator. The reported analysis should therefore be read using the registry's stated comparison information rather than assigning an additional comparison not specified in the analysis record.
The p-value is reported as <0.0001. This supports the superiority hypothesis under the specified statistical analysis, but the p-value itself does not establish the clinical importance of a 0.7-percentage-point mean difference.
Oral Semaglutide 14 mg vs Sitagliptin 100 mg
Mean difference in HbA1c change
95% CI: −1.1 to −0.8 · P < 0.0001
Outcome unit: percentage point of HbA1C
The estimated mean difference was −1.0 percentage point of HbA1c for oral semaglutide 14 mg versus sitagliptin 100 mg. The negative sign indicates a lower estimated change in HbA1c for the first-named treatment comparison relative to the comparator.
The 95% CI of −1.1 to −0.8 provides the uncertainty range around the mean-difference estimate under the MMRM framework. It is important not to interpret this interval as the range of responses that individual participants could experience.
The p-value of <0.0001 provides evidence against the null hypothesis under the reported model. It does not mean that there is a less-than-one-in-ten-thousand probability that the treatment difference occurred by chance, nor does it provide a measure of effect size. Effect magnitude is described by the mean difference, while precision is described by its confidence interval.
| Primary analysis | Method | Mean difference | 95% CI | P-value |
|---|---|---|---|---|
| Oral semaglutide 3 mg vs sitagliptin 100 mg | MMRM | −0.2 | −0.3 to −0.1 | 0.0078 |
| Oral semaglutide 7 mg | MMRM | −0.7 | −0.8 to −0.6 | <0.0001 |
| Oral semaglutide 14 mg vs sitagliptin 100 mg | MMRM | −1.0 | −1.1 to −0.8 | < 0.0001 |
7. How to Interpret the Primary Endpoint
The primary endpoint is expressed as a change from baseline, not simply an HbA1c value measured at week 26. That distinction matters because the statistical comparison concerns how the outcome moved over the specified period.
A negative mean difference means that the first treatment in the comparison had a lower estimated change in HbA1c than the comparator. The magnitude is expressed in percentage points of HbA1c. For example, a mean difference of −1.0 represents a one-percentage-point difference on the HbA1c scale between the model-estimated treatment changes.
Effect size
The mean difference communicates the estimated magnitude of the treatment contrast on the HbA1c percentage-point scale.
Precision
The 95% confidence interval communicates uncertainty around the estimated mean difference.
Evidence against the null
The p-value quantifies statistical evidence under the specified model; it is not an effect-size measure.
Population
The analysis population is the full analysis set, defined by the registry as all randomized participants, with available-data counts determining the participants contributing to the endpoint analysis.
8. Safety Results
The registry reports serious adverse events by treatment arm. These counts are presented as affected participants over participants at risk.
| Arm | Serious adverse events | Affected / at risk |
|---|---|---|
| Oral Semaglutide 3 mg | Serious adverse events | 16/361 |
| Oral Semaglutide 7 mg | Serious adverse events | 11/358 |
| Oral Semaglutide 14 mg | Serious adverse events | 11/361 |
| Sitagliptin 100 mg | Serious adverse events | 15/358 |
These safety data are descriptive counts by randomized treatment arm. They should not be interpreted as a formal hypothesis test because the registry information presented here does not provide a statistical comparison, confidence interval, or p-value for serious adverse events.
Affected/at-risk counts provide the numerator and denominator needed to understand how many participants experienced a serious adverse event relative to the number at risk in each listed arm. The counts alone do not establish whether observed differences are statistically significant, clinically important, or attributable to treatment.
9. Statistical Methods Explained
Why was an MMRM used?
HbA1c is a longitudinal outcome, and the registered endpoint is defined as change from baseline to week 26. An MMRM is designed for repeated measurements and can use the available longitudinal outcome information within a model rather than reducing the entire follow-up history to a single observed value. The registry specifically identifies MMRM as the method used for all three posted primary analyses.
What does a mean difference of −1.0 mean?
A mean difference of −1.0 means that the model-estimated change in HbA1c for the first-named treatment comparison was one percentage point lower than the comparator's estimated change, on the HbA1c percentage-point scale. It is a difference between estimated group means, not a statement about every individual participant.
What does the 95% confidence interval tell us?
The 95% confidence interval describes the uncertainty around the estimated treatment difference under the statistical model and sampling framework. For the 14 mg analysis, the interval is −1.1 to −0.8. The interval is therefore useful for assessing both the direction and precision of the estimated treatment contrast.
Why does the p-value not measure effect size?
A p-value addresses statistical evidence against a null hypothesis. It is affected by the magnitude of the effect, the variability of the data, and the amount of information contributing to the analysis. Consequently, two studies can have similar effect estimates but different p-values, or statistically significant results with very different effect magnitudes.
What is the full analysis set?
The registry defines the FAS as including all randomized participants. The endpoint-specific number analyzed is based on participants with available data for that outcome measure. This preserves the randomized population as the foundation of the analysis while recognizing that outcome measurements may not be available for every participant at every analysis.
Why does the observation period matter?
The registry distinguishes an on-treatment without rescue medication observation period from an in-trial observation period. These definitions determine which observations contribute to each version of the endpoint and therefore affect the interpretation of the resulting estimated treatment difference.
What does superiority mean here?
The registry classifies the hypothesis as superiority. The statistical objective is therefore to determine whether the treatment comparison demonstrates a difference rather than to establish that one treatment is no worse than another within a prespecified non-inferiority margin.
10. Confidence Intervals and Statistical Precision
The three reported primary estimates have two-sided 95% confidence intervals:
| Comparison | Estimate | 95% CI width | Interpretive role |
|---|---|---|---|
| Oral semaglutide 3 mg vs sitagliptin 100 mg | −0.2 | −0.3 to −0.1 | Uncertainty around the estimated mean HbA1c difference |
| Oral semaglutide 7 mg | −0.7 | −0.8 to −0.6 | Uncertainty around the registry-listed analysis estimate |
| Oral semaglutide 14 mg vs sitagliptin 100 mg | −1.0 | −1.1 to −0.8 | Uncertainty around the estimated mean HbA1c difference |
The confidence intervals should be read together with the point estimates. The point estimate gives the center of the reported treatment contrast, while the interval gives information about its statistical precision.
11. The Role of Randomization and Masking
The registry identifies the allocation as randomized and the design as parallel. Randomization is important because it creates the basis for comparing treatment groups without intentionally assigning participants according to their expected outcomes.
The trial is also identified as quadruple masked. Masking can reduce the opportunity for knowledge of treatment assignment to influence participant behavior, investigator conduct, outcome assessment, or other trial processes, depending on which parties are masked.
12. What the Mean Difference Does — and Does Not — Mean
A mean difference of −0.2, −0.7, or −1.0 describes a difference between model-estimated mean changes in HbA1c. It does not mean that every participant had exactly the corresponding change, and it does not describe the probability that an individual participant will achieve a particular HbA1c response.
The corresponding 95% confidence interval describes uncertainty around the population-level mean difference. A narrower interval indicates greater statistical precision than a wider interval, all else equal.
The reported p-values address statistical evidence under the model. They do not measure treatment magnitude, patient-level benefit, or clinical importance.
13. Statistical Issues That Matter for Interpretation
Available outcome data
The registry defines the FAS as all randomized participants but specifies that the number analyzed for the endpoint consists of participants with available data. Thus, "all randomized" and "all participants with an observed endpoint value" are not necessarily the same number.
Repeated measurements
Because the analysis uses MMRM, the statistical model is designed around repeated outcome measurements rather than treating the week-26 value as an isolated observation. This is particularly relevant when interpreting a change-from-baseline endpoint.
Model-based estimates
The mean differences are not simple arithmetic differences between two reported group averages in the registry. They are effect estimates produced by the reported mixed model for repeated measurements.
Multiple primary analyses
The registry reports three primary endpoint analyses. Their estimates and p-values should therefore be understood as a set of prespecified primary analyses rather than as three unrelated statistical exercises. The information reported by the registry does not specify a multiplicity-adjustment procedure for these three posted analyses, so no additional adjustment should be attributed to the trial.
14. Limitations
- Endpoint-specific available data: the FAS includes all randomized participants, while the number analyzed for the outcome consists of participants with available data.
- Registry-level methodological detail: the registry identifies MMRM as the primary method but does not provide the full model specification in the information presented here.
- One primary comparison field: the analysis record for oral semaglutide 7 mg lists the treatment group but does not state the comparator in the "Groups Compared" field.
- Safety interpretation: serious adverse event counts are reported descriptively without a corresponding formal statistical comparison in the posted analyses summarized here.
- Endpoint classification: the registry identifies the primary endpoint type as binary even though the endpoint is expressed as change in HbA1c in percentage points and analyzed using MMRM.
- Clinical interpretation: statistical evidence and effect magnitude should be distinguished from broader clinical judgments about treatment value.
15. Why This Trial Matters Statistically
PIONEER 12 provides a useful teaching example because the primary endpoint combines a clinically familiar continuous biomarker with a longitudinal statistical model. The analysis illustrates how randomized treatment comparisons can be expressed through estimated mean differences, confidence intervals, and p-values while accounting for repeated measurements.
| Concept | How it appears in PIONEER 12 |
|---|---|
| Randomization | The trial is randomized. |
| Parallel design | The registered design model is parallel. |
| Masking | The trial is quadruple masked. |
| Full analysis set | The FAS includes all randomized participants. |
| Repeated measurements | Primary analyses use MMRM. |
| Mean difference | The primary effect measure is mean difference. |
| Confidence interval | Each posted primary analysis reports a two-sided 95% CI. |
| P-value | Each posted primary analysis includes a p-value. |
| Superiority testing | The registered hypothesis type is superiority. |
| Observation periods | The endpoint is evaluated using both on-treatment without rescue medication and in-trial observation periods. |
| Safety analysis | Serious adverse events are reported as affected/at-risk counts by arm. |
16. Related Tutorials
Learn more about the methods used in this trial:
17. Related Calculators
18. Sources
- ClinicalTrials.gov: PIONEER 12, NCT04017832.
- PubMed: PMID 38985161.
Continue through Clinical Biostats
Explore statistical tutorials, calculators, and additional clinical trial analyses.
19. Record Summary
PIONEER 12 is a randomized phase 3 parallel trial with quadruple masking and four registered interventions in people with type 2 diabetes. Its posted primary endpoint is change from baseline to week 26 in HbA1c, evaluated using a mixed model for repeated measurements in the full analysis set. Three primary analyses are reported, with mean differences of −0.2, −0.7, and −1.0, accompanied by two-sided 95% confidence intervals and p-values.
Statistically, the most important feature of the analysis is the combination of a longitudinal endpoint with an MMRM framework. The mean difference describes the estimated treatment contrast, the confidence interval describes its precision, and the p-value describes statistical evidence under the model. These quantities should be interpreted together rather than treating any one of them as a complete summary of the trial.