This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
EXAMINE was a randomized, parallel, quadruple-masked phase 3 study evaluating alogliptin versus placebo in participants with type 2 diabetes and acute coronary syndrome. The registered primary endpoint was the percentage of participants with primary major adverse cardiac events (MACE), analyzed as a time-to-event outcome using Cox proportional-hazards methodology.
| Feature | EXAMINE |
|---|---|
| Trial name | EXAMINE |
| NCT identifier | NCT00968708 |
| Phase | Phase 3 |
| Therapeutic area | Endocrinology |
| Conditions | Diabetes Mellitus, Type 2; Acute Coronary Syndrome |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 5380 |
| Interventions | Alogliptin; Placebo |
| Primary endpoints registered | 1 |
| Results posted | Yes |
| Outcome measures posted | 2 |
| Statistical analyses posted | 3 |
| Lead sponsor | Takeda |
| Sponsor type | Industry |
| Status | Completed |
2. Clinical Question
The central statistical question was whether alogliptin, compared with placebo, could be evaluated with respect to the occurrence of the registered primary MACE composite from randomization through the adjudication cut-off date of May 31 2013, with the primary analysis framed as a non-inferiority or equivalence comparison.
Population
Participants with type 2 diabetes and acute coronary syndrome enrolled in the EXAMINE phase 3 study.
Intervention
Alogliptin.
Comparator
Placebo.
Primary question
How does the time-to-event hazard for primary MACE compare between the alogliptin and placebo groups under the prespecified non-inferiority or equivalence framework?
3. Trial Design
Alogliptin
- Intervention: alogliptin
- Drug intervention
- Compared with placebo
Placebo
- Comparator: placebo
- Drug comparator
- Compared with alogliptin
The design is important statistically because randomization establishes the primary comparison between treatment assignments, while masking reduces opportunities for knowledge of assigned treatment to influence trial conduct or outcome assessment. The registry identifies the study as randomized, parallel, and quadruple-masked; the ClinicalTrials.gov record does not specify a randomization ratio.
4. Endpoints
The registered primary endpoint was a binary endpoint in the registry's primary-endpoint listing, while the posted statistical analysis treated the endpoint as a time-to-event outcome. That distinction matters: the analysis uses not only whether an event occurred, but also information about when the event occurred and censoring during follow-up.
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Percentage of Participants With Primary Major Adverse Cardiac Events (MACE) | Primary Major Adverse Cardiac Events were defined as a composite of cardiovascular death, nonfatal myocardial infarction and nonfatal stroke; these events were adjudicated by an independent cardiovascular endpoints committee. | From randomization until the adjudication cut-off date of May 31 2013 (maximum time on study was 41 months). |
| Percentage of Participants With Secondary Major Adverse Cardiac Events (MACE) | The registry-reported statistical analysis identifies a secondary MACE composite. | From randomization until the adjudication cut-of date of May 31 2013 (maximum time on study was 41 months). |
5. Statistical Methodology
Primary analysis population
The primary analyses were based on the full analysis set, defined as all randomized participants. This is closely aligned with the intention-to-treat principle because participants remain classified according to their randomized treatment assignment rather than being reclassified according to subsequent exposure.
Cox proportional-hazards model
The posted primary analyses used a Cox proportional-hazards model. This is a survival-analysis model that estimates a relative hazard between treatment groups while accounting for differing follow-up times and censoring.
An HR below 1 indicates a lower estimated instantaneous event hazard in the alogliptin group relative to placebo under the fitted model. An HR of 1 would represent equal estimated hazards.
Stratified analysis
One of the two posted primary analyses specifies stratification by endpoint renal function and geographic region. Stratification allows the Cox comparison to account for these prespecified factors rather than treating the trial population as completely homogeneous with respect to them.
Intention-to-treat analysis
The primary analysis population was the full analysis set of all randomized participants, and the registry-reported analysis text identifies intention-to-treat analysis as a relevant concept. The key statistical advantage is preservation of the randomized comparison. Treatment assignment remains the basis of the efficacy comparison even if participants do not follow an identical treatment course after randomization.
Repeated confidence intervals and group-sequential monitoring
The registry-reported primary analysis notes that the statistical analyses used sequences of one-sided repeated confidence intervals to assess non-inferiority or statistical superiority with respect to the null hypotheses. Each sequence was constructed using critical values from a one-sided stopping boundary for a group-sequential design, with the stated purpose of preserving an overall false-rejection rate.
Interim analysis
The analysis text explicitly identifies interim analysis and alpha spending as part of the statistical framework.
One-sided inference
The reported confidence intervals are 97.5% one-sided intervals, rather than ordinary two-sided 95% intervals.
6. Primary Results: MACE
The registry contains two posted primary statistical analyses for the same primary MACE endpoint. Both use a Cox proportional-hazards framework and both compare placebo with alogliptin in the full analysis set. They differ in the posted analysis specification: one identifies interim-analysis/alpha-spending concepts, while the other specifies stratification by endpoint renal function and geographic region.
Primary analysis 1
Hazard ratio for primary MACE
97.5% one-sided upper CI: 1.160 · P = 0.315
Analysis population: full analysis set (all randomized participants)
| Characteristic | Posted primary analysis 1 |
|---|---|
| Outcome | Percentage of Participants With Primary Major Adverse Cardiac Events (MACE) |
| Endpoint type | Time-to-event |
| Groups compared | Placebo vs Alogliptin |
| Method | Cox proportional hazards |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.962 |
| Confidence interval | 97.5% one-sided; upper bound 1.160 |
| P-value | 0.315 |
| Hypothesis type | Non-inferiority or equivalence |
The estimated HR of 0.962 means that the fitted model estimated the instantaneous hazard of the primary MACE composite in the alogliptin group at approximately 96.2% of the corresponding estimated hazard in the placebo group. Expressed as a relative difference, 1 − 0.962 = 0.038, so the point estimate corresponds to approximately a 3.8% lower estimated hazard.
The HR does not mean that 3.8% fewer participants experienced MACE, nor does it describe an absolute reduction in event probability. It is a relative, model-based time-to-event measure.
The 97.5% one-sided upper confidence bound of 1.160 indicates that the uncertainty in the estimate extends above 1.0. In this non-inferiority framework, that upper bound is particularly important because the registry-reported analysis description focuses on whether the repeated confidence bound satisfies the prespecified inferential boundary.
The P-value of 0.315 is evidence against a specified null hypothesis under the statistical testing framework; it is not a measure of the size of the treatment effect. A P-value cannot tell us that an effect is "31.5%" large or small.
The analysis also requires attention to the Cox model assumptions, censoring, the analysis population, and the group-sequential procedure. Because the registry extract does not provide a complete non-inferiority margin, that margin should not be inferred from the HR, confidence interval, or P-value.
Primary analysis 2: stratified analysis
Stratified hazard ratio for primary MACE
97.5% one-sided upper CI: 1.169 · P = 0.332
Stratified by endpoint renal function and geographic region
| Characteristic | Posted primary analysis 2 |
|---|---|
| Outcome | Percentage of Participants With Primary Major Adverse Cardiac Events (MACE) |
| Endpoint type | Time-to-event |
| Groups compared | Placebo vs Alogliptin |
| Method | Cox proportional hazards |
| Stratification | Endpoint renal function and geographic region |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.965 |
| Confidence interval | 97.5% one-sided; upper bound 1.169 |
| P-value | 0.332 |
| Hypothesis type | Non-inferiority or equivalence |
The stratified HR of 0.965 corresponds to an estimated hazard approximately 96.5% as large in the alogliptin group as in the placebo group under the stratified Cox model. The point estimate therefore corresponds to approximately a 3.5% lower estimated hazard.
Again, this does not mean that 3.5% fewer participants experienced MACE. It is a model-based relative hazard estimate over the analyzed time-to-event follow-up.
The 97.5% one-sided upper confidence bound is 1.169. The interval therefore includes values above 1.0, and its interpretation must be made within the trial's stated non-inferiority or equivalence framework rather than by treating the HR alone as a definitive conclusion.
The P-value of 0.332 should likewise not be interpreted as an effect-size measure. Its meaning depends on the null hypothesis and the prespecified group-sequential testing framework.
7. Non-Inferiority Logic
The primary analyses are explicitly classified in the registry as non-inferiority or equivalence hypotheses. The registry-reported analysis description states that, after accrual of 550 participants with MACE events, the upper bound of a one-sided repeated confidence interval for the hazard ratio was used in the inferential framework. It further states that an upper bound below 1.0 would support superiority of alogliptin to placebo for the primary MACE composite.
The registry extract establishes the confidence-bound framework but does not provide a complete numerical non-inferiority margin in the quoted analysis text. Therefore, no numerical margin is reported on this page.
This distinction is fundamental. A non-inferiority analysis does not ask only whether the observed HR is below 1.0. Instead, it asks whether the confidence interval excludes treatment effects worse than a prespecified amount of clinically acceptable difference. The registry-reported extract does not provide that complete margin, so a numerical non-inferiority conclusion cannot be reconstructed from the available data alone.
8. Secondary Endpoint Results
The ClinicalTrials.gov record contains one posted secondary statistical analysis for the secondary MACE endpoint. The registry does not report the statistical method for this secondary analysis, although it reports the hazard ratio as the effect measure and identifies intention-to-treat analysis as a relevant concept.
Secondary MACE hazard ratio
97.5% one-sided upper CI: 1.135
Analysis population: full analysis set
| Characteristic | Secondary analysis |
|---|---|
| Outcome | Percentage of Participants With Secondary Major Adverse Cardiac Events (MACE) |
| Time frame | From randomization until the adjudication cut-of date of May 31 2013 (maximum time on study was 41 months). |
| Endpoint type | Time-to-event |
| Groups compared | Placebo vs Alogliptin |
| Analysis population | Full analysis set |
| Method | Not reported in the registry analysis |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.952 |
| Confidence interval | 97.5% one-sided; upper bound 1.135 |
| Hypothesis type | Non-inferiority or equivalence |
The secondary MACE HR of 0.952 corresponds to an estimated hazard approximately 95.2% as large in the alogliptin group as in the placebo group, or approximately a 4.8% lower estimated hazard based on the point estimate.
The 97.5% one-sided upper confidence bound is 1.135. As with the primary endpoint, the confidence bound—not the point estimate alone—is central to interpreting a non-inferiority framework.
The registry extract does not report a formal statistical method for this secondary analysis beyond the reported hazard-ratio effect measure. Accordingly, the analysis should not be described as having used a particular survival model unless that method is independently documented in the ClinicalTrials.gov record.
9. Safety
The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm as affected participants divided by participants at risk. These figures provide an arm-specific count and denominator and should be kept separate from the efficacy analysis.
| Safety measure | Placebo | Alogliptin |
|---|---|---|
| Serious adverse events | 952/2679 | 907/2701 |
Placebo
952/2679 participants were reported as affected by serious adverse events.
Alogliptin
907/2701 participants were reported as affected by serious adverse events.
These are descriptive safety counts from the ClinicalTrials.gov record. They should not be converted into an inferred comparative effect measure without a prespecified safety analysis definition, appropriate population definition, and corresponding statistical analysis.
10. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
The primary MACE endpoint is fundamentally a time-to-event outcome in the posted statistical analysis. Participants can experience the event at different times, while others may remain event-free through their available follow-up. The Cox model uses this timing information and accommodates right censoring rather than reducing the analysis to a simple event/no-event proportion.
What does an HR of 0.962 mean?
An HR of 0.962 means that the fitted model estimates the instantaneous event hazard in the alogliptin group to be 0.962 times that in the placebo group. As a point estimate, that corresponds to approximately a 3.8% lower estimated hazard. It does not mean a 3.8 percentage-point reduction in the probability of MACE.
Why is the confidence interval especially important in a non-inferiority trial?
Non-inferiority is fundamentally a confidence-interval question. The trial asks whether the uncertainty around the estimated treatment effect remains within an acceptable prespecified range. A point estimate close to 1.0 cannot by itself establish non-inferiority. The registry-reported EXAMINE extract identifies a one-sided repeated confidence-interval framework but does not provide a complete numerical non-inferiority margin.
Why are the confidence intervals one-sided?
The primary analyses report 97.5% one-sided confidence intervals. One-sided inference is appropriate to a directional hypothesis such as non-inferiority when the protocol defines the inferential question in terms of whether the treatment effect crosses a particular unfavorable boundary. The confidence bound must therefore be interpreted in the direction specified by the trial design.
Why does interim analysis change the interpretation?
The registry-reported analysis text identifies interim analysis, repeated confidence intervals, and a group-sequential stopping boundary. Looking at accumulating data more than once can increase the false-positive rate if no adjustment is made. A group-sequential framework and alpha-spending approach are designed to control that error rate while allowing prespecified interim evaluation.
Why was stratification used in one primary analysis?
The second primary analysis was stratified by endpoint renal function and geographic region. Stratification permits the Cox analysis to account for these factors when estimating the treatment comparison. The resulting HR of 0.965 is therefore a stratified model estimate rather than a simple unadjusted ratio of event proportions.
11. Interpreting the Hazard Ratio Carefully
The primary HR estimates of 0.962 and 0.965 are both below 1.0. Under the fitted models, this corresponds to lower estimated instantaneous MACE hazards with alogliptin than with placebo. The point estimates themselves do not establish the clinical importance or statistical conclusion of the trial.
The first primary analysis has a 97.5% one-sided upper bound of 1.160, while the stratified primary analysis has an upper bound of 1.169. Because both bounds exceed 1.0, the reported uncertainty includes values corresponding to a higher estimated hazard under alogliptin. In a non-inferiority framework, whether this is acceptable depends on the prespecified non-inferiority margin, which is not numerically reported in the provided extract.
The reported primary P-values are 0.315 and 0.332. A P-value is calculated relative to a null hypothesis and the specified testing procedure. It does not quantify the magnitude of the HR, the probability that the null hypothesis is true, or the probability that a treatment effect is clinically meaningful.
The Cox proportional-hazards model carries assumptions about the relationship between treatment and the hazard over time. A single HR is most straightforward to interpret when the proportional-hazards assumption is reasonable. The ClinicalTrials.gov record does not provide a formal assessment of that assumption, so no such assessment is inferred here.
12. Randomization and Masking
Randomization is the central design mechanism supporting the causal comparison between alogliptin and placebo. By randomly assigning participants to treatment groups, known and unknown prognostic characteristics are expected to be distributed between groups in a way that supports an unbiased treatment comparison on average.
The registry classifies the study as quadruple-masked. Masking is distinct from randomization: randomization determines treatment assignment, while masking limits knowledge of treatment assignment during trial conduct and assessment. The ClinicalTrials.gov record does not specify the individual roles covered by the quadruple-masking designation, so those roles are not expanded here.
13. Interim Analysis and Alpha Spending
The primary analysis description explicitly references a group-sequential design and repeated confidence intervals. It states that critical values from a one-sided stopping boundary were used and that the repeated-confidence-interval sequence was constructed to preserve an overall false-rejection rate.
Why repeated looks require adjustment
If a trial repeatedly examines accumulating results without accounting for those looks, the chance of incorrectly rejecting a null hypothesis can become larger than the nominal level.
What the EXAMINE analysis did
The registry text describes repeated one-sided confidence intervals constructed using a group-sequential stopping boundary.
The ClinicalTrials.gov record does not provide the complete alpha-spending schedule, all information times, or the full numerical stopping-boundary specification. Those details therefore are not reconstructed on this page.
14. Analysis Population and the Intention-to-Treat Principle
The primary analyses used the full analysis set (all randomized participants). This is important because the treatment comparison remains anchored to randomization rather than to whether a participant subsequently adhered to assigned treatment.
| Population / concept | What the ClinicalTrials.gov record say |
|---|---|
| Full analysis set | All randomized participants; used for both posted primary analyses. |
| Intention-to-treat | Identified as a concept in the analysis text and relevant to the primary comparison. |
| Secondary analysis population | Full analysis set. |
Using all randomized participants protects the interpretation of the randomized treatment comparison. It also means that the estimated treatment effect should not be interpreted simply as the effect among participants who successfully completed or adhered to treatment.
15. What the Primary Endpoint Does — and Does Not — Tell Us
The primary endpoint is a composite of three clinically distinct outcomes: cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke. The composite therefore answers a combined question: how does treatment assignment relate to the time until the first occurrence of an event included in the MACE definition?
What it tells us
The reported HR summarizes the relative time-to-event hazard for the defined primary MACE composite under the specified statistical model.
What it does not tell us
The composite HR does not separately quantify the treatment effect for cardiovascular death, myocardial infarction, and stroke from the ClinicalTrials.gov record.
Composite endpoints are useful when clinically meaningful events can be combined into a prespecified outcome, but interpretation requires attention to the exact definition. A statistically similar composite hazard does not imply that every individual component has the same hazard ratio.
16. Limitations
- Incomplete non-inferiority margin: the registry analysis identifies a non-inferiority or equivalence framework and describes the confidence-bound logic, but the complete numerical non-inferiority margin is not provided in the registry-reported extract.
- Two primary analyses: the registry contains two formal primary analyses for the same MACE endpoint. They produce closely related but not identical estimates and P-values.
- Secondary-method reporting: the registry extract does not identify the statistical method for the secondary MACE analysis.
- Composite endpoint: the primary MACE result combines cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke. The ClinicalTrials.gov record does not provide separate component-specific estimates.
- Cox model: interpretation of the hazard ratio depends on the appropriateness of the proportional-hazards model. The ClinicalTrials.gov record does not report a formal proportional-hazards diagnostic.
- Interim analysis: the group-sequential framework means that ordinary unadjusted interpretations of the reported P-values would ignore an important feature of the design.
- Missing-data detail: the ClinicalTrials.gov record does not describe a specific missing-data or imputation strategy, so none is inferred.
- Safety interpretation: serious adverse-event counts are provided by arm, but the ClinicalTrials.gov record does not provide a formal comparative safety analysis.
- External validity: the ClinicalTrials.gov record identifies the study population through the stated conditions but does not provide detailed baseline characteristics or eligibility criteria in the registry-reported extract. Generalizability therefore cannot be assessed beyond the available information.
17. Why This Trial Matters Statistically
EXAMINE is a useful teaching case because it brings together several important concepts in modern cardiovascular outcomes-trial analysis: randomized treatment comparison, masking, a composite time-to-event endpoint, Cox proportional-hazards modeling, one-sided confidence intervals, non-inferiority logic, stratification, and group-sequential monitoring.
| Concept | How it appears in EXAMINE |
|---|---|
| Randomization | The study is classified as randomized. |
| Parallel-group design | The design model is parallel with 2 arms. |
| Quadruple masking | The registry classifies the trial as quadruple-masked. |
| Time-to-event endpoint | The posted primary analyses treat primary MACE as a time-to-event outcome. |
| Cox proportional-hazards model | Used for both posted primary analyses. |
| Hazard ratio | Primary effect measure for the treatment comparison. |
| One-sided confidence interval | Primary analyses report 97.5% one-sided confidence intervals. |
| Non-inferiority / equivalence | The primary analyses are classified under this hypothesis type. |
| Stratified analysis | One primary analysis is stratified by endpoint renal function and geographic region. |
| Interim analysis | The analysis description references interim analysis and repeated confidence intervals. |
| Group-sequential design | Critical values from a one-sided stopping boundary were used. |
| Intention-to-treat principle | The primary analysis population is the full analysis set of all randomized participants. |
18. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The posted primary Cox analyses produced HR estimates of 0.962 and 0.965, with 97.5% one-sided upper confidence bounds of 1.160 and 1.169. The analyses were framed as non-inferiority or equivalence comparisons and incorporated a group-sequential framework.
Clinical interpretation
The primary endpoint represents a composite of cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke. The ClinicalTrials.gov record permits interpretation of the reported composite hazard estimates but do not provide component-specific treatment estimates or a complete numerical non-inferiority margin.
Keeping these two levels separate is important. Statistical evidence describes what the prespecified analysis estimates and how uncertain that estimate is. Clinical interpretation asks what that evidence means for the defined clinical endpoint and population. Neither should be expanded beyond the information actually reported.
19. A Practical Reading of the EXAMINE Results
The endpoint is not simply "whether MACE occurred." The posted analysis uses time-to-event information through the adjudication cut-off date.
The HR summarizes the relative instantaneous event hazard under the Cox model. It is not an absolute risk difference.
The reported HR must be interpreted together with the one-sided confidence bound and the prespecified trial-design framework.
The efficacy comparison remains tied to randomized assignment.
The ClinicalTrials.gov record is a model-based summary of the composite time-to-event endpoint.
20. Related Tutorials
Learn more about the methods used in this trial:
21. Related Calculators
22. Sources
- ClinicalTrials.gov: EXAMINE, NCT00968708.
- Linked publication: PubMed PMID 34872085.
- Linked publication: PubMed PMID 32493335.
- Linked publication: PubMed PMID 31902327.
- Linked publication: PubMed PMID 30371278.
- Linked publication: PubMed PMID 29848776.
Continue through Clinical Biostats
Explore statistical tutorials, clinical-trial methods, and calculators related to survival analysis, non-inferiority design, and time-to-event inference.
23. Record Summary
EXAMINE provides a useful example of how a cardiovascular outcomes trial can combine randomized allocation, quadruple masking, a composite time-to-event endpoint, Cox proportional-hazards modeling, one-sided confidence intervals, non-inferiority or equivalence hypotheses, stratified analysis, and group-sequential monitoring. The posted primary analyses report hazard ratios of 0.962 and 0.965 for primary MACE, with 97.5% one-sided upper confidence bounds of 1.160 and 1.169, respectively. The secondary MACE analysis reports an HR of 0.952 with a 97.5% one-sided upper confidence bound of 1.135.
The most important statistical lesson is that these numbers cannot be interpreted in isolation. The HR describes a relative time-to-event comparison; the confidence bound describes uncertainty; the P-value is tied to a specified hypothesis; and the non-inferiority framework depends on a prespecified margin. The registry extract does not provide that complete numerical margin, so the appropriate analysis is to report the observed estimates and their confidence bounds without reconstructing an unsupported threshold.