This page separates reported trial results from statistical interpretation. Numerical results and trial characteristics are limited to the ClinicalTrials.gov data posted on ClinicalTrials.gov for NCT01147250 and the associated analysis records identified there.
1. Trial at a Glance
ELIXA was a randomized, parallel, triple-masked phase 3 trial evaluating lixisenatide (AVE0010) versus placebo in patients with type 2 diabetes after acute coronary syndrome. The registered primary endpoint was a time-to-event composite of cardiovascular death, non-fatal myocardial infarction, non-fatal stroke, or hospitalization for unstable angina.
| Feature | ELIXA |
|---|---|
| Trial name | ELIXA |
| NCT ID | NCT01147250 |
| Therapeutic area | Endocrinology |
| Condition | Acute Coronary Syndrome |
| Phase | Phase 3 |
| Status | Completed |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 6068 |
| Interventions | Lixisenatide (AVE0010); Placebo |
| Lead sponsor | Sanofi |
| Sponsor type | Industry |
| Study period | 2010-06 to 2015-02 |
2. Clinical Question
The central statistical question was whether treatment with lixisenatide, compared with placebo, changed the time to first occurrence of the registered primary cardiovascular event in patients represented by the ELIXA trial population.
Population
Patients represented by the trial's condition and eligibility framework: type 2 diabetes in the setting of acute coronary syndrome.
Intervention
Lixisenatide (AVE0010).
Comparator
Placebo.
Primary question
How does lixisenatide compare with placebo for the time to first occurrence of the primary cardiovascular event?
3. Trial Design
Lixisenatide
- Intervention: Lixisenatide (AVE0010)
- Trial role: randomized treatment group
- Serious adverse events: 625/3031 affected/at risk
Placebo
- Intervention: Placebo
- Trial role: randomized comparator group
- Serious adverse events: 669/3032 affected/at risk
The registry characterizes the allocation as randomized, the design model as parallel, and the masking as triple. These design features establish the basic structure of the comparison: participants were assigned to one of two concurrent treatment groups, with masking intended to reduce the influence of treatment knowledge on trial conduct and assessment.
4. Trial Timeline
Study start
The ELIXA study began in June 2010 according to the registry profile.
Randomized comparative study
The study used randomized allocation, a parallel design, and triple masking with lixisenatide and placebo as the two interventions.
Primary time-to-event assessment
The registered primary endpoint was followed from randomization through the end of study, with a median follow-up of 25 months.
Primary completion
The registry profile identifies February 2015 as the primary completion date.
5. Primary Endpoint
| Endpoint | Definition / time frame | Analysis |
|---|---|---|
| Time to First Occurence of Primary CV Event | CV Death, Non-Fatal MI, Non-Fatal Stroke or Hospitalization for Unstable Angina; from randomization up to the end of study (median follow-up of 25 months) | Cox proportional-hazards model and log-rank test |
The registry definition specifies that Kaplan-Meier plots of the cumulative incidence rate by treatment group were used to depict the onset of the primary cardiovascular endpoint over time. The number of observed participants with endpoint events was reported in the registry results framework, and a cardiovascular event adjudication committee reviewed and adjudicated all potential events in a blinded fashion.
6. Analysis Population and Covariate Adjustment
The primary efficacy analysis used the intention-to-treat (ITT) population, defined in the registry analysis as all randomized participants analyzed according to the treatment group allocated at randomization.
| Analysis feature | ELIXA specification |
|---|---|
| Analysis population | Intent-to-treat (ITT) |
| ITT definition | All randomized participants analyzed according to the treatment group allocated at randomization |
| Groups compared | Placebo vs Lixisenatide |
| Cox model covariates | Treatment groups and region |
| Regions used as covariate categories | North America; South and Central America; Western Europe; Eastern Europe; Africa/Near East; Asia/Pacific |
| Confidence interval | Two-sided 95% |
Including region as a covariate is a form of covariate adjustment. The model therefore estimates the treatment hazard ratio while accounting for the regional categories specified in the analysis record. This is distinct from changing the randomized treatment assignment: the treatment comparison remains anchored to the original randomized groups.
7. Statistical Methodology
Kaplan-Meier estimation
The registered endpoint is a time-to-event outcome, and the registry states that Kaplan-Meier plots were used to depict the cumulative incidence rate by treatment group over time. Kaplan-Meier estimation is designed for situations in which participants can have different lengths of observed follow-up and some participants have not experienced the event when observation ends.
Here, di represents events at an event time and ni represents participants at risk immediately before that time.
Cox proportional-hazards model
The primary registry analysis used a Cox proportional-hazards model. This model relates the instantaneous event rate to treatment and specified covariates and expresses the treatment contrast through a hazard ratio.
The treatment hazard ratio is obtained from the corresponding model coefficient. A hazard ratio below 1 indicates a lower estimated instantaneous event rate in the first group relative to the reference group, while a value above 1 indicates a higher estimated instantaneous event rate.
Log-rank test
The registry also reports a log-rank analysis for the primary endpoint. The log-rank test compares the observed and expected numbers of events between treatment groups across event times. It is therefore naturally suited to randomized time-to-event comparisons.
Intention-to-treat analysis
Analyzing randomized participants according to their assigned treatment preserves the treatment contrast created by randomization. The ITT principle is particularly important in a time-to-event trial because treatment discontinuation or other post-randomization events should not simply cause participants to be reassigned to a different randomized group for the primary efficacy comparison.
Covariate adjustment
The Cox analysis incorporated region as a covariate in addition to treatment. The six registry-specified regional categories were North America, South and Central America, Western Europe, Eastern Europe, Africa/Near East, and Asia/Pacific.
8. Primary Endpoint Results
The ClinicalTrials.gov record contains two formal analyses of the same registered primary endpoint. One is the Cox proportional-hazards analysis used for the non-inferiority comparison; the other is the log-rank analysis reported for the superiority hypothesis.
Non-inferiority analysis
Hazard ratio: Lixisenatide vs Placebo
95% CI: 0.886–1.168 · Two-sided 95% CI
Non-inferiority criterion: upper bound of the 2-sided 95% CI < 1.3
The Cox proportional-hazards analysis was performed using treatment group and region as covariates. The analysis population was the ITT population, and the effect measure was the hazard ratio.
The estimated hazard ratio of 1.017 is very close to 1. Under the fitted Cox model, the estimated instantaneous rate of the first primary cardiovascular event was therefore close to the corresponding rate in the comparator group.
The hazard ratio does not mean that 1.7% of participants experienced an event, nor does it mean that the probability of an event was exactly 1.7% higher. A hazard ratio is a relative model-based time-to-event measure, not an absolute risk difference or risk ratio.
The two-sided 95% confidence interval of 0.886 to 1.168 describes the statistical uncertainty around the estimated hazard ratio under the specified analysis framework. It includes 1, so the interval is compatible with a range of relative effects on either side of no hazard-ratio difference.
For the registered non-inferiority framework, however, the relevant comparison is not simply whether the interval includes 1. The registry criterion states that non-inferiority of lixisenatide versus placebo was to be claimed if the upper bound of the two-sided 95% confidence interval was less than 1.3. The reported upper bound, 1.168, is below that prespecified threshold.
The confidence interval therefore provides the key information for the non-inferiority interpretation. The p-value from a conventional superiority test is not a measure of effect size and is not the rule specified by the registry-reported non-inferiority criterion.
Superiority analysis
Hazard ratio and log-rank test
95% CI: 0.886–1.168 · P = 0.8542
Superiority criterion: upper bound of the 2-sided 95% CI < 1.0
The registry's second primary analysis used the log-rank test and reported the same hazard ratio and two-sided 95% confidence interval. The analysis notes specify that superiority of lixisenatide versus placebo was to be claimed if the upper bound of the two-sided 95% confidence interval was less than 1.0.
The estimated hazard ratio of 1.017 does not indicate a large relative separation between the randomized groups. Because the estimate is close to 1, the estimated instantaneous event rates are close under the fitted model.
The 95% confidence interval, 0.886–1.168, crosses 1.0. More importantly for the registry's stated superiority rule, its upper bound of 1.168 is not below 1.0. The registry-reported analysis therefore does not satisfy the stated confidence-interval criterion for superiority.
The reported P = 0.8542 is a result of the superiority-oriented hypothesis test. It does not quantify the size or clinical importance of the treatment effect. A p-value is a measure of compatibility with the specified null hypothesis and statistical model; it is not the probability that the null hypothesis is true and it is not a percentage reduction or increase in risk.
Because this is a Cox/log-rank time-to-event analysis, interpretation also depends on the assumptions underlying those methods, including the proportional-hazards framework for the Cox hazard ratio and appropriate handling of censoring. The ClinicalTrials.gov record does not provide a separate diagnostic assessment of those assumptions.
9. Non-Inferiority: Why the Margin Changes the Question
ELIXA's primary analysis included a non-inferiority hypothesis. This is statistically different from asking whether the treatment is superior to placebo.
The registry analysis states that non-inferiority of lixisenatide versus placebo was to be claimed when the upper confidence-limit for the hazard ratio was below 1.3.
| Question | Relevant quantity | ELIXA result |
|---|---|---|
| Estimated treatment effect | Hazard ratio | 1.017 |
| Statistical uncertainty | Two-sided 95% CI | 0.886–1.168 |
| Non-inferiority threshold | Upper CI bound < 1.3 | 1.168 < 1.3 |
| Superiority threshold | Upper CI bound < 1.0 | 1.168 is not < 1.0 |
This illustrates why a non-inferiority trial cannot be interpreted simply by asking whether a conventional p-value is below a chosen threshold. The margin defines the largest unfavorable relative effect that the analysis is designed to exclude. Here, the registry criterion uses 1.3 as that boundary for the hazard ratio.
10. Serious Adverse Events
the ClinicalTrials.gov record reports serious adverse events by randomized arm using affected participants and participants at risk.
| Treatment group | Affected | At risk |
|---|---|---|
| Placebo | 669 | 3032 |
| Lixisenatide | 625 | 3031 |
The affected counts should be interpreted together with the corresponding at-risk denominators rather than as raw counts alone. The ClinicalTrials.gov record identifies 669/3032 for placebo and 625/3031 for lixisenatide. The ClinicalTrials.gov record does not include a formal statistical analysis of this safety measure, so no comparative p-value or confidence interval is added here.
11. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
The primary endpoint is a time-to-first-event outcome. Participants can experience the event at different times, and some observations may be censored before an event occurs. A Cox model is designed for this structure and produces a hazard ratio that summarizes the relative event rate while allowing the model to incorporate specified covariates such as region.
What does a hazard ratio of 1.017 mean?
A hazard ratio of 1.017 means that the fitted model estimated the instantaneous event rate in the lixisenatide group to be approximately 1.017 times that of the placebo group, after incorporating the treatment and regional covariates specified in the analysis. It does not mean that the cumulative probability of an event was exactly 1.7% higher.
Why is the non-inferiority margin 1.3 rather than 1.0?
A non-inferiority question permits a prespecified amount of possible disadvantage while still considering the treatment acceptably close to the comparator for the stated purpose. In the registry-reported ELIXA analysis, the upper limit of the two-sided 95% confidence interval had to be below 1.3. A limit below 1.0 would instead represent a stricter criterion associated with superiority in the stated analysis.
Why does the confidence interval matter more than the point estimate for non-inferiority?
The point estimate describes the best single estimate under the model, but non-inferiority concerns whether the plausible range of effects remains within the prespecified margin. The reported interval of 0.886–1.168 is therefore more informative for the non-inferiority decision than the point estimate of 1.017 alone.
What does the P-value of 0.8542 tell us?
The reported 0.8542 is associated with the superiority analysis. It does not measure the magnitude of the treatment effect. It also does not establish the probability that the null hypothesis is true. Its interpretation depends on the particular hypothesis, test statistic, model, and assumptions used in the analysis.
Why use the intention-to-treat population?
The ITT definition reported by the registry includes all randomized participants and analyzes them according to their randomized treatment assignment. This preserves the comparison established by randomization and avoids changing the primary treatment contrast because of events occurring after randomization.
12. Understanding Censoring in a Time-to-Event Trial
A time-to-event endpoint contains more information than a simple yes/no event indicator because the timing of the event matters. A participant who has not experienced the primary cardiovascular event by the end of observation contributes information about being event-free up to that point.
Event
The participant experiences the first qualifying component of the registered composite endpoint.
Censoring
The participant reaches the end of their usable observation without a qualifying event being observed during that period.
Risk set
At each event time, the survival analysis considers participants who remain under observation and at risk immediately before that time.
Time-to-event comparison
The Cox model and log-rank test use the timing of events rather than reducing the entire follow-up period to a single binary outcome.
The registry definition also states that potential cardiovascular events were reviewed and adjudicated by a cardiovascular event adjudication committee in a blinded fashion. That adjudication process is relevant because a composite cardiovascular endpoint depends on consistent classification of the qualifying event components.
13. Reading the Composite Endpoint
The ELIXA primary endpoint combines four clinically defined event components:
| Component | Role in endpoint |
|---|---|
| Cardiovascular death | Qualifying component of the first primary cardiovascular event |
| Non-fatal myocardial infarction | Qualifying component of the first primary cardiovascular event |
| Non-fatal stroke | Qualifying component of the first primary cardiovascular event |
| Hospitalization for unstable angina | Qualifying component of the first primary cardiovascular event |
Because the registered endpoint is defined as the time to first occurrence, the analysis focuses on the first qualifying component experienced by each participant. The hazard ratio therefore summarizes the treatment comparison for the composite time-to-first-event outcome rather than providing a separate hazard ratio for every component.
14. Primary Analysis: Putting the Numbers Together
| Statistical quantity | Reported result | How to interpret it |
|---|---|---|
| Hazard ratio | 1.017 | Estimated relative instantaneous event rate for lixisenatide versus placebo under the Cox model |
| 95% CI | 0.886–1.168 | Uncertainty interval around the hazard-ratio estimate |
| Non-inferiority margin | 1.3 | Prespecified upper confidence-limit threshold in the registry-reported analysis |
| Superiority criterion | Upper CI bound < 1.0 | Prespecified confidence-interval condition for superiority |
| Superiority P-value | 0.8542 | Reported test result for the superiority analysis; not an effect-size measure |
The statistical story is therefore multidimensional. The point estimate is close to the null value of 1.0, the confidence interval extends below and above 1.0 but remains below the non-inferiority margin of 1.3 at its upper bound, and the reported superiority analysis has a P-value of 0.8542 with an upper confidence limit above 1.0.
15. What the Hazard Ratio Does — and Does Not — Mean
A hazard ratio of 1.017 means that, under the fitted Cox model, the estimated instantaneous rate of the first primary cardiovascular event in the lixisenatide group was approximately 1.7% higher relative to the placebo group on the hazard-ratio scale.
It does not mean that 1.7% more participants had cardiovascular events, that the probability of an event was 1.7% higher, or that every participant experienced the same relative change. Those would require absolute risk information or another specifically defined effect measure.
The 95% confidence interval of 0.886–1.168 shows that the point estimate is not the entire statistical story. The interval spans values below and above 1.0, while its upper endpoint remains below the registered non-inferiority boundary of 1.3.
The superiority P-value of 0.8542 describes the result of the specified hypothesis test. It is not a percentage, does not quantify the size of the hazard ratio, and should not be converted into a statement such as "there is an 85.42% chance that the treatment has no effect."
16. Non-Inferiority and Superiority Are Different Questions
Non-inferiority question
Is the upper bound of the two-sided 95% confidence interval below the prespecified hazard-ratio margin of 1.3?
Superiority question
Does the upper bound of the two-sided 95% confidence interval fall below 1.0, according to the registry-reported analysis criterion?
The same estimated hazard ratio and confidence interval can therefore be examined against two different thresholds. This is an important statistical distinction: a result can be compatible with the stated non-inferiority criterion without simultaneously establishing superiority.
The point estimate is 1.017, the lower confidence limit is 0.886, and the upper confidence limit is 1.168. The upper limit is below 1.3 but above 1.0.
17. Safety Interpretation
The serious-adverse-event data posted on ClinicalTrials.gov for ELIXA provide affected counts and corresponding at-risk denominators for both randomized groups.
| Measure | Placebo | Lixisenatide |
|---|---|---|
| Serious adverse events, affected | 669 | 625 |
| Participants at risk | 3032 | 3031 |
| Reported format | 669/3032 | 625/3031 |
The denominators are nearly the same size, which makes the affected counts directly interpretable as arm-specific safety summaries when read with their denominators. However, the registry profile does not provide a formal statistical analysis of this safety measure, so a hazard ratio, risk ratio, risk difference, or p-value should not be manufactured from the counts.
18. Results Scope
The ClinicalTrials.gov record reports 4 outcome measures and 2 statistical analyses. Both registry-reported statistical analyses concern the registered primary time-to-event endpoint. The ClinicalTrials.gov record does not contain formal statistical-analysis estimates for secondary endpoints.
19. Important Limitations and Interpretation Issues
- Composite endpoint: the primary outcome combines cardiovascular death, non-fatal myocardial infarction, non-fatal stroke, and hospitalization for unstable angina. The ClinicalTrials.gov record does not provide component-specific statistical estimates.
- Hazard-ratio interpretation: the Cox hazard ratio is a model-based relative time-to-event measure, not an absolute risk difference or risk ratio.
- Proportional-hazards assumption: the Cox model relies on a proportional-hazards framework for its standard hazard-ratio interpretation. The ClinicalTrials.gov record does not report a separate assessment of this assumption.
- Censoring: time-to-event analyses depend on appropriate handling and interpretation of censored observations. The ClinicalTrials.gov record does not provide detailed censoring diagnostics.
- Non-inferiority logic: the relevant criterion is the prespecified margin of 1.3, not simply whether the confidence interval includes 1.0.
- Superiority versus non-inferiority: satisfying a non-inferiority criterion does not by itself establish superiority. The registry-reported superiority criterion requires the upper confidence bound to be below 1.0.
- Analysis population: the primary efficacy analysis uses the ITT population, defined as all randomized participants analyzed according to their allocated treatment group.
- Covariate adjustment: the Cox analysis adjusts for the specified regional categories. The interpretation is therefore tied to the stated model specification.
- Safety analysis: the ClinicalTrials.gov record is descriptive affected/at-risk counts; no formal comparative safety statistic is provided in the ClinicalTrials.gov recordset.
- Secondary outcomes: although 4 outcome measures are identified as posted, the ClinicalTrials.gov record does not provide estimates for secondary endpoints.
20. Why This Trial Matters Statistically
ELIXA is a useful teaching case because the registry analysis brings together several core concepts in clinical-trial statistics: randomized allocation, masking, a composite time-to-event endpoint, Kaplan-Meier estimation, Cox regression, log-rank testing, ITT analysis, covariate adjustment, confidence intervals, and a prespecified non-inferiority margin.
| Concept | How it appears in ELIXA |
|---|---|
| Randomization | Randomized allocation to lixisenatide or placebo |
| Parallel design | Two concurrent randomized intervention groups |
| Triple masking | Registry design characteristic |
| Time-to-event endpoint | Time to first occurrence of the primary cardiovascular event |
| Kaplan-Meier estimation | Used to depict cumulative incidence over time by treatment group |
| Hazard ratio | Primary effect measure, estimated as 1.017 |
| Confidence interval | Two-sided 95% CI of 0.886–1.168 |
| Cox proportional-hazards model | Primary model with treatment and region as covariates |
| Log-rank test | Reported formal comparison for the primary endpoint |
| Intention-to-treat | All randomized participants analyzed according to assigned treatment |
| Non-inferiority | Upper two-sided 95% CI required to be below 1.3 |
| Superiority | Upper two-sided 95% CI required to be below 1.0 |
21. A Practical Reading Sequence for ELIXA
When reading the primary result, a useful statistical sequence is to move from the endpoint to the estimand, then to uncertainty, and finally to the hypothesis framework.
This sequence prevents a common statistical mistake: beginning with the p-value and attempting to infer everything else from it. The endpoint definition, analysis population, effect estimate, confidence interval, and prespecified hypothesis framework each answer a different question.
22. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
23. Related Statistical Calculators
24. Sources
- ClinicalTrials.gov: ELIXA — NCT01147250.
- Linked publication: PubMed PMID 26630143.
- Linked publication: PubMed PMID 39963952.
- Linked publication: PubMed PMID 36546588.
- Linked publication: PubMed PMID 35817031.
- Linked publication: PubMed PMID 33046070.
Continue through the Clinical Biostats statistical pathway
Connect the ELIXA analysis to deeper tutorials on survival analysis, confidence intervals, non-inferiority design, and other clinical-trial methods.
25. Record Summary
ELIXA provides a clear example of how a randomized clinical trial can combine a composite time-to-event endpoint with both non-inferiority and superiority hypotheses. The registry-reported analysis used the ITT population, a Cox proportional-hazards model with treatment and region as covariates, a two-sided 95% confidence interval, and a log-rank test. The reported hazard ratio was 1.017 with a 95% CI of 0.886–1.168. For the registered non-inferiority framework, the upper confidence limit was compared with the prespecified margin of 1.3; for the superiority analysis, the stated criterion required the upper confidence limit to be below 1.0, with a reported P = 0.8542.
The most useful statistical reading therefore does not reduce the trial to a single p-value. It considers the endpoint definition, randomized analysis population, hazard-ratio estimate, confidence interval, non-inferiority margin, superiority criterion, model covariates, and the limitations inherent in interpreting a composite time-to-event outcome.