This page separates reported trial results from statistical interpretation. Numerical results and trial characteristics are restricted to the ClinicalTrials.gov record. The registry provides the official trial record.
Trial at a Glance
PIONEER 6 was a randomized, double-masked, parallel phase 3 trial investigating the cardiovascular safety of oral semaglutide compared with placebo in subjects with type 2 diabetes.
| Feature | PIONEER 6 |
|---|---|
| Trial name | PIONEER 6 |
| Brief title | A Trial Investigating the Cardiovascular Safety of Oral Semaglutide in Subjects With Type 2 Diabetes |
| Phase | Phase 3 |
| Therapeutic area | Endocrinology |
| Condition | Diabetes; Diabetes Mellitus, Type 2 |
| Design | Randomized, parallel, double-masked |
| Primary purpose | Treatment |
| Enrollment | 3183 |
| Interventions | Semaglutide; placebo |
| Primary endpoint type | Binary registered endpoint; analyzed as a time-to-event endpoint |
| ClinicalTrials.gov | NCT02692716 |
| Lead sponsor | Novo Nordisk A/S |
| Sponsor type | Industry |
Clinical Question
The central statistical question was whether oral semaglutide could be shown to have cardiovascular safety relative to placebo with respect to the time from randomisation to the first occurrence of a major adverse cardiovascular event (MACE), while also allowing a superiority analysis of the same endpoint.
Population
Subjects with diabetes mellitus, type 2 diabetes, enrolled in the phase 3 PIONEER 6 trial.
Intervention
Oral semaglutide.
Comparator
Placebo.
Primary question
Is oral semaglutide non-inferior to placebo for the time to first MACE, using the prespecified hazard-ratio margin, and what does the same analysis indicate about superiority?
Trial Design
Oral semaglutide
- Intervention: semaglutide.
- Serious adverse events: 301/1591 affected participants/participants at risk.
Placebo
- Comparator: placebo.
- Serious adverse events: 358/1591 affected participants/participants at risk.
Endpoints
| Endpoint | Registered definition | Time frame |
|---|---|---|
| Primary endpoint | Time From Randomisation to First Occurrence of a Major Adverse Cardiovascular Event (MACE) Composite Endpoint Consisting of: Cardiovascular Death, Non-fatal Myocardial Infarction or Non-fatal Stroke | Maximum treatment duration is dependent on event rates and is estimated to be no longer than 19 months + 5 weeks of follow-up period. |
| Secondary endpoint | Time From Randomisation to First Occurrence of an Expanded Composite Cardiovascular Endpoint Consisting of: Cardiovascular Death, Non-fatal Myocardial Infarction, Non-fatal Stroke, UAP Requiring Hospitalisation or Hospitalisation for Heart Failure | Maximum treatment duration is dependent on event rates and is estimated to be no longer than 19 months + 5 weeks of follow-up period. |
| Secondary endpoint | Time From Randomisation to First Occurrence of Each of the Individual Components in the Expanded Composite Cardiovascular Endpoint | Maximum treatment duration is dependent on event rates and is estimated to be no longer than 19 months + 5 weeks of follow-up period. |
| Secondary endpoint | Time From Randomisation to First Occurrence of a Composite Endpoint Consisting of: All-cause Death, Non-fatal Myocardial Infarction or Nonfatal Stroke | Maximum treatment duration is dependent on event rates and is estimated to be no longer than 19 months + 5 weeks of follow-up period. |
| Secondary endpoint | Time From Randomisation to First Occurrence of Fatal or Non-fatal Myocardial Infarction | Maximum treatment duration is dependent on event rates and is estimated to be no longer than 19 months + 5 weeks of follow-up period. |
| Secondary endpoint | Time From Randomisation to First Occurrence of Fatal or Non-fatal Stroke | Maximum treatment duration is dependent on event rates and is estimated to be no longer than 19 months + 5 weeks of follow-up period. |
| Secondary endpoint | Time From Randomisation to All-cause Death | Maximum treatment duration is dependent on event rates and is expected to be no longer than 19 months + 5 weeks of follow-up period. |
| Secondary endpoint | Time to First AE Leading to Permanent Trial Product Discontinuation | Maximum treatment duration is dependent on event rates and is expected to be no longer than 19 months + 38 days of ascertainment window. |
The registry's primary endpoint definition specifies that the reported event count represents the first event of a MACE, defined as cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke. The posted analysis uses the in-trial observation period, beginning at randomisation and including the period after permanent trial product discontinuation, if any, and ending at the follow-up visit regardless of adherence.
Statistical Methodology
Full analysis set
The primary analysis used the full analysis set (FAS), which comprised all randomised participants. This is important because the primary treatment comparison remains anchored to randomisation rather than being restricted to participants who remained on treatment.
Cox proportional-hazards model
The primary MACE endpoint was analyzed using a Cox proportional-hazards model. Treatment was included as a categorical fixed factor, and the analysis was stratified by evidence of cardiovascular disease at screening.
The Cox model relates the instantaneous event rate to treatment while allowing the baseline hazard to remain unspecified. The exponentiated treatment coefficient is interpreted as a hazard ratio.
Hazard ratio
The primary effect measure was the hazard ratio (HR). An HR below 1 indicates a lower estimated instantaneous event rate in the oral semaglutide group relative to placebo under the fitted model; an HR above 1 indicates a higher estimated instantaneous event rate.
The HR is a relative time-to-event measure. It is not an absolute risk difference, a probability of experiencing an event, or a statement that every individual has the same proportional change in risk.
Stratified analysis
The primary MACE analysis was stratified by evidence of cardiovascular disease at screening. Stratification allows the analysis to account for this prespecified factor when comparing the treatment groups.
Censoring
For the primary endpoint, participants were censored at the end of their in-trial observation period. Censoring means that participants without an observed event contribute information up to their censoring time rather than being treated as if an event had occurred.
Non-inferiority framework
The principal hypothesis was non-inferiority. The registry states that non-inferiority of oral semaglutide versus placebo was considered confirmed if the upper limit of the two-sided 95% confidence interval for the HR was strictly below 1.8.
Multiplicity control
The registry states that the non-inferiority hypothesis was controlled for multiplicity. It also states that the superiority hypothesis was controlled for multiplicity. This matters because the interpretation of a confirmatory P-value depends on the prespecified testing strategy rather than on the numerical P-value alone.
Statistical Methods Explained
Why was a Cox model used?
The primary outcome was not simply whether a participant ever experienced MACE. It was the time from randomisation to first MACE. A survival-analysis model such as Cox regression uses both event occurrence and the amount of observed event-free time, while accommodating right censoring.
What does an HR of 0.79 mean?
An HR of 0.79 corresponds to a 21% lower estimated instantaneous event rate for oral semaglutide relative to placebo under the fitted model. This is a relative model-based interpretation; it does not mean that 21% of participants avoided MACE or that an individual participant's absolute probability was reduced by 21 percentage points.
Why is the non-inferiority margin more important than whether the HR is below 1?
Non-inferiority asks whether the treatment can be ruled out as being worse than the comparator by more than a prespecified clinically acceptable amount. The registry's boundary was an HR of 1.8. Therefore, the decisive comparison for non-inferiority is the upper confidence limit versus 1.8, not simply whether the point estimate is below 1.
Why can the same endpoint have both non-inferiority and superiority analyses?
These are different questions applied to the same estimated treatment effect. Non-inferiority asks whether the upper confidence boundary excludes an HR at or above the specified margin. Superiority asks whether the treatment effect differs from the null value of HR 1 in the favorable direction under the prespecified testing procedure.
What does the 95% confidence interval tell us?
The confidence interval describes statistical uncertainty around the estimated hazard ratio under the model and sampling framework. For the primary MACE analysis, the interval was 0.57 to 1.11. It therefore provides substantially more information about precision than the point estimate of 0.79 alone.
Why does the P-value not measure the size of the treatment effect?
A P-value addresses evidence against a specified null hypothesis within the statistical testing framework. It does not tell us how large the treatment effect is. Effect size is described by the hazard ratio, while the confidence interval describes uncertainty around that estimate.
Results: Primary MACE Endpoint
The primary endpoint was the time from randomisation to first occurrence of a MACE composite consisting of cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke. The primary analysis used the full analysis set and a stratified Cox proportional-hazards model.
Non-inferiority analysis
Two-sided 95% CI: 0.57–1.11 · P < 0.0001
Non-inferiority margin: upper HR boundary strictly below 1.8
The estimated hazard ratio of 0.79 is below 1, corresponding to a 21% lower estimated instantaneous MACE rate for oral semaglutide relative to placebo under the fitted Cox model.
For the non-inferiority question, however, the important comparison is different. The upper limit of the two-sided 95% confidence interval was 1.11, which is strictly below the prespecified non-inferiority margin of 1.8. On the registry's stated criterion, the non-inferiority hypothesis was therefore confirmed.
The HR of 0.79 does not mean that 21% of participants experienced a benefit, nor does it represent an absolute 21% reduction in the probability of MACE. The confidence interval of 0.57–1.11 also crosses the null value of 1, so the interval is compatible with both a lower and a higher instantaneous event rate under the usual superiority interpretation.
The very small non-inferiority P-value should not be interpreted as the magnitude of the treatment effect. It addresses the prespecified non-inferiority testing question. The analysis also depends on the Cox model and its proportional-hazards interpretation, as well as on the handling of censoring and the defined in-trial observation period.
Superiority analysis of the same primary endpoint
Superiority analysis
Two-sided 95% CI: 0.57–1.11 · P = 0.1749
Hypothesis: superiority; controlled for multiplicity
The superiority analysis uses the same estimated hazard ratio and confidence interval: HR 0.79, 95% CI 0.57–1.11. The point estimate is below 1, but the confidence interval includes 1.
The reported P = 0.1749 is therefore not evidence of superiority under the stated analysis. Importantly, this does not reverse the non-inferiority finding. Non-inferiority and superiority are different hypotheses with different statistical thresholds.
The result also illustrates why the hazard ratio and P-value should be reported together but interpreted separately. The HR describes the estimated relative treatment effect; the P-value describes the evidence for the particular hypothesis being tested. Neither quantity alone describes absolute cardiovascular risk.
Secondary Endpoint Results
The registry posts additional Cox-model analyses for several cardiovascular and treatment-discontinuation time-to-event endpoints. These are distinct from the primary MACE analysis and, where the registry states that the hypothesis was not controlled for multiplicity, their P-values should not be interpreted as though each were an independently confirmatory test.
Expanded Composite Cardiovascular Endpoint
Expanded cardiovascular composite
Two-sided 95% CI: 0.61–1.10 · P = 0.1827
The expanded composite consisted of cardiovascular death, non-fatal myocardial infarction, non-fatal stroke, UAP requiring hospitalisation, or hospitalisation for heart failure. It was analyzed in the FAS using a stratified Cox proportional-hazards model. The registry states that this hypothesis was not controlled for multiplicity.
Individual Cardiovascular Components
| Endpoint component | HR | 95% CI | P-value |
|---|---|---|---|
| Cardiovascular death, including undetermined cause of death | 0.49 | 0.27–0.92 | = 0.0261 |
| Non-fatal myocardial infarction | 1.18 | 0.73–1.90 | = 0.5044 |
| Non-fatal stroke | 0.74 | 0.35–1.57 | = 0.4350 |
| Unstable angina pectoris requiring hospitalisation | 1.56 | 0.60–4.01 | = 0.3605 |
| Hospitalisation for heart failure | 0.86 | 0.48–1.55 | = 0.6227 |
These component analyses demonstrate why composite endpoints require careful interpretation. A composite combines several clinically distinct event types, and its overall estimate does not imply that every component has the same treatment effect. The component estimates here have different point estimates and different levels of precision.
All-cause Death, Myocardial Infarction, and Stroke Composite
All-cause death / MI / stroke composite
Two-sided 95% CI: 0.56–1.05 · P = 0.0952
This endpoint consisted of all-cause death, non-fatal myocardial infarction, or nonfatal stroke. The registry reports a stratified Cox analysis using the FAS and states that the hypothesis was not controlled for multiplicity.
Fatal or Non-fatal Myocardial Infarction
Myocardial infarction
Two-sided 95% CI: 0.66–1.66 · P = 0.8583
This analysis used the on-treatment observation period, with time measured from the first dose of trial product to the first EAC-confirmed fatal or non-fatal myocardial infarction. The registry states that participants were censored at the end of their on-treatment observation period.
Fatal or Non-fatal Stroke
Stroke
Two-sided 95% CI: 0.37–1.56 · P = 0.4485
This analysis used the in-trial observation period and assessed time from randomisation to the first EAC-confirmed fatal or non-fatal stroke. The confidence interval is relatively broad compared with the point estimate, illustrating the uncertainty that can accompany analyses of individual event components.
All-cause Death
Time to all-cause death
Two-sided 95% CI: 0.31–0.84 · P = 0.0078
The registry reports this analysis as time from randomisation to all-cause death, using the in-trial observation period and a stratified Cox proportional-hazards model. The registry states that this hypothesis was not controlled for multiplicity.
An HR of 0.51 corresponds to a 49% lower estimated instantaneous rate of death under the fitted model. That relative interpretation should not be converted into an absolute survival difference because the ClinicalTrials.gov record does not provide the corresponding absolute survival estimates.
Adverse Event Leading to Permanent Trial Product Discontinuation
Time to first AE leading to permanent discontinuation
Two-sided 95% CI: 1.42–2.30 · P < 0.0001
This analysis measured time from the first dose of trial product to the first adverse event leading to permanent trial product discontinuation. The registry reports a stratified Cox proportional-hazards analysis and states that the hypothesis was not controlled for multiplicity.
The HR of 1.81 means that the estimated instantaneous rate of the analyzed discontinuation event was higher in the oral semaglutide group than in the placebo group under the fitted model. This endpoint concerns discontinuation because of an adverse event; it is not itself a measure of the incidence or severity of all adverse events.
Safety
The registry provides serious adverse-event counts by treatment arm. These are reported as affected participants over participants at risk.
| Safety measure | Oral semaglutide | Placebo |
|---|---|---|
| Serious adverse events | 301/1591 | 358/1591 |
The ClinicalTrials.gov record does not provide a formal statistical analysis for these serious adverse-event counts in the posted statistical analyses summarized here. Therefore, the counts should be kept distinct from the Cox-model results above.
The serious-adverse-event data show the number of affected participants and the corresponding number at risk in each arm. A count comparison by itself does not provide a confidence interval, a P-value, a time-to-event estimate, or an adjusted treatment effect.
For a binary safety outcome, a risk ratio, risk difference, odds ratio, or an appropriate time-to-event method could be considered depending on the precise event definition and follow-up structure. The ClinicalTrials.gov record does not post such a formal comparison for this serious-adverse-event measure, so no additional effect estimate is inferred here.
How to Read the Primary Confidence Interval
The most important feature of the primary interval is its relationship to the non-inferiority boundary. The entire 95% confidence interval, from 0.57 to 1.11, lies below 1.8. That is the statistical basis for the posted non-inferiority conclusion.
The same interval also crosses HR 1. That distinction explains why the non-inferiority result and the superiority result can coexist without contradiction: the former uses 1.8 as its critical margin, whereas superiority concerns whether the treatment effect is different from HR 1.
Time-to-Event Analysis and Censoring
Several posted endpoints use the same general survival-analysis framework, but the exact time origin and observation period differ. These distinctions are important because changing the time origin or censoring rule changes the estimand being analyzed.
| Endpoint family | Time origin / observation period |
|---|---|
| Primary MACE | Randomisation; in-trial observation period |
| Expanded cardiovascular endpoint | Randomisation; in-trial observation period |
| Individual cardiovascular components | Randomisation; in-trial observation period |
| All-cause death / MI / stroke composite | Randomisation; in-trial observation period |
| Fatal or non-fatal myocardial infarction | First dose of trial product; on-treatment observation period |
| Fatal or non-fatal stroke | Randomisation; in-trial observation period |
| All-cause death | Randomisation; in-trial observation period |
| AE leading to permanent discontinuation | First dose of trial product; end of treatment visit framework |
This is more than a technical detail. An analysis beginning at randomisation answers a different question from one beginning at first dose. Likewise, an in-trial observation period can include information after permanent trial-product discontinuation, whereas an on-treatment analysis focuses on a defined exposure period.
Composite Endpoints: Why the Definition Matters
The primary MACE endpoint combines cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke. Composite endpoints can increase the number of observed events and therefore improve statistical information, but the components may differ in frequency, clinical importance, and treatment effect.
One composite estimate
The primary HR of 0.79 describes the first occurrence of any qualifying MACE component, not a separate effect estimate for each component.
Component-specific estimates
The registry separately reports HRs for cardiovascular death, myocardial infarction, stroke, unstable angina requiring hospitalisation, and hospitalisation for heart failure.
First-event logic
The primary endpoint counts the first occurrence of a qualifying MACE rather than treating repeated events as independent primary events.
Interpretation
A composite result should not automatically be described as though its individual components all had identical effects.
Non-inferiority: The Statistical Logic
PIONEER 6 is particularly useful for teaching non-inferiority because the posted primary result illustrates a case where the confidence interval crosses the usual superiority null value of 1 but remains comfortably below the non-inferiority margin.
Observed upper confidence limit: 1.11. Prespecified margin: 1.8.
The logic can be expressed in three steps. First, estimate the relative treatment effect with the prespecified survival model. Second, quantify uncertainty using the two-sided 95% confidence interval. Third, compare the upper confidence limit with the prespecified non-inferiority margin.
This framework prevents an overly simplistic interpretation such as "HR below 1 means success." The point estimate being below 1 is not what establishes non-inferiority. The margin and confidence interval determine whether the observed uncertainty is compatible with an unacceptably large disadvantage.
Multiplicity
The registry explicitly states that the non-inferiority hypothesis for the primary endpoint was controlled for multiplicity and that the superiority hypothesis was also controlled for multiplicity. This is important because the same observed HR can be evaluated through different hypotheses while preserving the trial's overall error-control strategy.
| Analysis | Hypothesis type | Multiplicity status | Key criterion |
|---|---|---|---|
| Primary MACE | Non-inferiority | Controlled for multiplicity | Upper two-sided 95% CI strictly below 1.8 |
| Primary MACE | Superiority | Controlled for multiplicity | Reported P = 0.1749 |
| Expanded cardiovascular endpoint | Other / not stated | Not controlled for multiplicity | HR 0.82; 95% CI 0.61–1.10 |
| Individual cardiovascular components | Other / not stated | Not controlled for multiplicity | Component-specific Cox estimates |
| All-cause death | Other / not stated | Not controlled for multiplicity | HR 0.51; 95% CI 0.31–0.84 |
| AE leading to discontinuation | Other / not stated | Not controlled for multiplicity | HR 1.81; 95% CI 1.42–2.30 |
The distinction is essential when reading the secondary P-values. A nominal P-value such as 0.0078 for all-cause death is a result of the posted analysis, but the registry specifically states that this hypothesis was not controlled for multiplicity. It should therefore not be treated as if it were an independently multiplicity-adjusted confirmatory test.
Analysis Population and Interpretation
The primary and most secondary efficacy analyses used the full analysis set, comprising all randomised participants. This approach preserves the treatment assignment generated by randomization and avoids defining the primary comparison solely by treatment exposure or adherence.
The analysis population is particularly relevant in a safety-focused cardiovascular trial because treatment discontinuation and subsequent observation can create different analytical questions. The primary MACE endpoint explicitly uses the in-trial observation period, including the period after permanent trial product discontinuation, if any. By contrast, the myocardial-infarction analysis uses an on-treatment observation period beginning at first dose.
If participants discontinue treatment but continue to be followed for the primary endpoint, their subsequent events can remain part of the randomized comparison. That is consistent with an analysis anchored to randomisation. An on-treatment analysis answers a narrower question concerning events during the specified treatment-observation period.
Important Limitations
- Registry-level information: this analysis is restricted to the ClinicalTrials.gov record. The registry does not provide every possible component of a full statistical analysis plan or publication-level results.
- No absolute event probabilities reported: the posted statistical analyses provide hazard ratios and confidence intervals, but the ClinicalTrials.gov record does not provide Kaplan-Meier survival probabilities or median event times.
- Hazard-ratio interpretation: Cox HRs are model-based relative measures. A single HR is most straightforward when the proportional-hazards interpretation is appropriate over the analyzed period.
- Censoring: time-to-event analyses depend on the defined censoring rules and observation periods. The primary endpoint censors participants at the end of their in-trial observation period.
- Different observation periods: not all secondary endpoints use the same time origin or observation framework. The myocardial-infarction analysis uses an on-treatment observation period, while the primary MACE analysis uses the in-trial observation period.
- Multiplicity: several secondary hypotheses were explicitly not controlled for multiplicity, so their nominal P-values require more cautious interpretation than the prespecified primary testing framework.
- Composite endpoint: MACE combines cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke. A composite estimate should not be interpreted as though it were an effect estimate for every component separately.
- Safety comparisons: the registry-reported serious-adverse-event data are counts by arm rather than a posted formal comparative analysis for that measure.
Why This Trial Matters Statistically
PIONEER 6 provides a compact teaching example of how a cardiovascular safety trial can combine randomized treatment assignment, a time-to-event primary endpoint, a stratified Cox model, a hazard ratio, and a formal non-inferiority margin.
| Concept | How it appears in PIONEER 6 |
|---|---|
| Randomization | Randomized allocation to two parallel treatment arms. |
| Double masking | Participants and study personnel were enrolled under a double-masked design. |
| Time-to-event endpoint | The primary endpoint measures time from randomisation to first MACE. |
| Kaplan-Meier framework | Time-to-event endpoints of this type are naturally described using survival-analysis methods that account for censoring. |
| Cox model | Primary and posted secondary time-to-event analyses use Cox regression. |
| Hazard ratio | Relative treatment effect is expressed as an HR. |
| Stratified analysis | The primary MACE analysis is stratified by evidence of cardiovascular disease at screening. |
| Non-inferiority | The primary hypothesis uses an upper HR margin of 1.8. |
| Confidence interval | The two-sided 95% CI is used directly in the non-inferiority decision. |
| Multiplicity | The primary non-inferiority and superiority hypotheses were controlled for multiplicity. |
| Different estimands | Some secondary analyses use in-trial observation, while another uses on-treatment observation. |
| Composite endpoint | MACE combines cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke. |
Statistical Methods Explained in Practical Terms
Non-inferiority is not superiority
The primary result can satisfy the non-inferiority criterion even though the superiority confidence interval includes HR 1.
The margin is prespecified
The HR value of 1.8 is a design criterion, not an observed treatment effect.
FAS preserves randomization
Using all randomized participants in the FAS keeps the primary efficacy comparison anchored to treatment assignment.
Censoring is informative about follow-up
Participants without an observed event contribute information until their specified censoring time.
Related Tutorials
Learn more about the methods used in this trial:
Related Calculators
Sources
- ClinicalTrials.gov: PIONEER 6, NCT02692716.
- Linked publication: PubMed record — PMID 29263194.
- Linked publication: PubMed record — PMID 30284349.
- Linked publication: PubMed record — PMID 31185157.
- Linked publication: PubMed record — PMID 34277983.
- Linked publication: PubMed record — PMID 35484580.
Continue through Clinical Biostats
Explore the statistical concepts behind randomized trials, survival analysis, non-inferiority testing, hazard ratios, confidence intervals, and clinical-trial methodology.
Record Summary
PIONEER 6 provides a clear example of the distinction between non-inferiority and superiority in a randomized time-to-event trial. The primary MACE analysis used the full analysis set and a stratified Cox proportional-hazards model, producing an HR of 0.79 with a two-sided 95% CI of 0.57–1.11. The upper confidence limit of 1.11 was strictly below the prespecified non-inferiority margin of 1.8, while the separately reported superiority analysis had P = 0.1749.
The secondary analyses extend the same survival-analysis framework to expanded cardiovascular outcomes, individual cardiovascular components, all-cause death, and adverse-event-related discontinuation. Their interpretation requires attention to the exact endpoint definition, time origin, observation period, censoring rule, and multiplicity status. The ClinicalTrials.gov record separately report serious adverse events of 301/1591 for oral semaglutide and 358/1591 for placebo.