← Clinical Trials
Type 2 Diabetes Phase 3 Cardiovascular Outcomes NCT01179048

LEADER: Complete Statistical Analysis of Liraglutide in Type 2 Diabetes

A detailed statistical review of the randomized phase 3 LEADER trial evaluating liraglutide versus placebo in patients with diabetes and type 2 diabetes, with emphasis on the prespecified cardiovascular time-to-event endpoint, non-inferiority and superiority testing, Cox proportional-hazards modeling, secondary cardiovascular and microvascular outcomes, and safety.

Trial start: 2010-08-31  ·  Primary completion: 2015-12-17  ·  Enrollment: 9341
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results are restricted to the ClinicalTrials.gov record for NCT01179048. ClinicalTrials.gov provides the official trial registry record.

Registry note: 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

LEADER was a randomized, double-blind, parallel phase 3 trial evaluating liraglutide versus placebo in diabetes and type 2 diabetes. The registry reports 9341 enrolled subjects, two arms, a binary registered primary endpoint, and 20 posted statistical analyses.

9341
Enrolled
Randomized trial
2
Arms
Liraglutide vs placebo
0.868
Primary HR
95% CI 0.778–0.968
<0.001
NI P-value
Two-sided CI; one-sided testing framework
FeatureLEADER
PhasePhase 3
Therapeutic areaEndocrinology
ConditionsDiabetes; Diabetes Mellitus, Type 2
DesignRandomized, parallel, double-blind
AllocationRandomized
Primary purposeTreatment
Enrollment9341
InterventionsLiraglutide; placebo
Primary endpoint typeBinary; analyzed as a time-to-event outcome using a Cox proportional-hazards model
Results postedYes
Outcome measures posted6
Statistical analyses posted20
Lead sponsorNovo Nordisk A/S
Sponsor typeIndustry
ClinicalTrials.govNCT01179048

2. Clinical Question

The principal question was whether liraglutide, compared with placebo, affected the time from randomisation to the first occurrence of cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke—the registered composite cardiovascular primary outcome.

Population

The registry identifies the study conditions as diabetes and Diabetes Mellitus, Type 2. The analysis population for the posted primary endpoint was all randomised subjects.

Intervention

Liraglutide was the study drug intervention.

Comparator

Placebo was the comparator drug.

Primary question

Does liraglutide meet the prespecified non-inferiority criterion for the composite cardiovascular outcome, and, if so, is superiority subsequently established?

3. Trial Design

01
Randomize9341 enrolled
02
Parallel armsLiraglutide vs placebo
03
Double-blindMasked randomized comparison
04
Follow-upTime-to-event outcomes
05
AnalysisCox regression
INTERVENTION

Liraglutide

  • Drug intervention
  • Compared with placebo
  • Included in all randomised-subject efficacy analyses reported in the registry
COMPARATOR

Placebo

  • Drug comparator
  • Compared with liraglutide
  • Included in all randomised-subject efficacy analyses reported in the registry

The registry describes the allocation as randomized, the design model as parallel, and masking as double. These features establish the basic structure of the treatment comparison: subjects were assigned to one of two concurrent groups and the study was conducted under double masking.

4. Endpoints

EndpointRegistry definitionTime frameAnalysis
Primary cardiovascular composite Time from randomisation to first occurrence of cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke (a composite cardiovascular outcome). The percentage of subjects experiencing a first event of cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke is presented. From randomisation (visit 3; month 0) to last contact (visit 16; up to month 60+30 days) Cox proportional-hazards model
Expanded composite cardiovascular outcome Time from randomisation to first occurrence of either cardiovascular death, non-fatal myocardial infarction, non-fatal stroke, revascularisation, hospitalisation for unstable angina or for heart failure. From randomisation (visit 3; month 0) to last contact (visit 16; up to month 60+30 days) Cox proportional-hazards model
All-cause death Time from randomisation to all cause death. From randomisation (visit 3; month 0) to last contact (visit 16; up to month 60+30 days) Cox proportional-hazards model
Microvascular outcome Time from randomisation to first occurrence of a composite microvascular outcome. From randomisation (visit 3; month 0) to last contact (visit 16; up to month 60+30 days) Cox proportional-hazards model

The registry also posts analyses of each individual component of the expanded cardiovascular outcome and individual components of the composite microvascular outcome, as well as retinopathy and nephropathy composite outcomes separately.

5. Statistical Methodology

Cox proportional-hazards model

The central statistical method was the Cox proportional-hazards model. The primary analysis estimated the hazard ratio comparing liraglutide with placebo in all randomised subjects, with treatment group included as a factor. Secondary time-to-event analyses similarly used Cox regression with treatment as a fixed factor.

Primary effect measure
HR = estimated hazard in liraglutide / estimated hazard in placebo

An HR below 1 indicates a lower estimated instantaneous event rate in the liraglutide group relative to placebo under the fitted model. It is a relative time-to-event measure, not a direct statement about absolute risk or individual patient outcomes.

Analysis population

The primary endpoint was analyzed in all randomised subjects. The same population is specified for the posted secondary cardiovascular and microvascular analyses. Anchoring efficacy analysis to randomized assignment preserves the treatment comparison created by randomization.

Two-sided confidence interval with one-sided non-inferiority testing

The primary endpoint was evaluated with a two-sided 95% confidence interval for the hazard ratio. Non-inferiority was then judged against a prespecified hazard-ratio margin of 1.3, using either the upper confidence-limit criterion or the equivalent one-sided hypothesis test.

Prespecified non-inferiority logic
H0: HR ≥ 1.3   versus   Ha: HR < 1.3

The registry states that non-inferiority was considered confirmed if the upper limit of the two-sided 95% CI was below 1.3, or if the p-value for the one-sided test was less than 2.5% (or equivalent to 5% in a two-sided test).

Sequential non-inferiority and superiority testing

The primary analysis had two distinct inferential questions. First, could liraglutide be shown to be non-inferior to placebo with respect to the prespecified cardiovascular margin? If non-inferiority was established, superiority was then tested against a hazard ratio of 1.0.

Superiority logic
H0: HR ≥ 1.0   versus   Ha: HR < 1.0

The registry states that superiority was considered confirmed if the upper limit of the two-sided 95% CI was below 1.0, or if the corresponding one-sided p-value was less than 2.5% (or equivalent to 5% in a two-sided test).

Kaplan-Meier interpretation

The posted endpoints are time-to-event outcomes. For this type of endpoint, Kaplan-Meier estimation is the standard descriptive framework for representing the probability of remaining event-free over time while accounting for right censoring. The registry's posted inferential method is Cox regression; the ClinicalTrials.gov record does not provide Kaplan-Meier estimates or curves to reproduce here.

What censoring means statistically

For a time-to-event endpoint, some subjects may reach the last contact without experiencing the event of interest. Such observations are censored rather than treated as if the event never occurred. The analysis therefore uses both event information and the amount of follow-up available for each subject.

6. Primary Results: Composite Cardiovascular Outcome

The registered primary endpoint was the time from randomisation to first occurrence of cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke. The analysis population was all randomised subjects, and the groups compared were liraglutide versus placebo.

Non-inferiority analysis

Primary cardiovascular outcome

HR 0.868

Two-sided 95% CI: 0.778–0.968   ·   P < 0.001

Non-inferiority hypothesis: H0: HR ≥ 1.3 versus Ha: HR < 1.3

The upper limit of the reported two-sided 95% confidence interval was 0.968, which is below the prespecified non-inferiority margin of 1.3. The registry also reports a p-value of <0.001 for the primary non-inferiority analysis.

Clinical Biostats interpretation

What the estimate means: An HR of 0.868 means that the estimated instantaneous rate of experiencing the first component of the composite cardiovascular outcome was about 13.2% lower with liraglutide than with placebo under the fitted Cox model.

What it does not mean: It does not mean that 13.2% of subjects avoided an event, that each individual subject had exactly a 13.2% reduction in risk, or that the absolute probability of an event was reduced by 13.2 percentage points.

Confidence interval: The two-sided 95% CI of 0.778–0.968 describes uncertainty around the estimated hazard ratio under the statistical model and sampling framework. It does not describe the range of effects among individual patients.

Why the p-value is not an effect size: The p-value addresses the prespecified hypothesis test. It does not tell us how large the treatment effect is. The HR and its confidence interval provide the magnitude and precision information.

Non-inferiority caution: The relevant non-inferiority comparison is against 1.3, not simply against 1.0. A confidence interval entirely below 1.3 establishes the registry's stated non-inferiority criterion even though the estimated HR is also below 1.0.

Model caution: Cox regression relies on the proportional-hazards framework. The single HR summarizes the relative hazard under that model and should not automatically be interpreted as a constant relative risk at every time point.

Superiority analysis

Superiority after non-inferiority

HR 0.868

Two-sided 95% CI: 0.778–0.968   ·   P = 0.005

Superiority hypothesis: H0: HR ≥ 1.0 versus Ha: HR < 1.0

After non-inferiority was established, the registry states that superiority was tested. The upper limit of the two-sided 95% CI was 0.968, below 1.0, and the reported superiority p-value was 0.005.

Clinical Biostats interpretation

What the estimate means: The same HR of 0.868 represents the estimated relative treatment effect for the superiority comparison. Under the fitted Cox model, the estimated instantaneous event rate was lower in the liraglutide group than in the placebo group.

What it does not mean: The HR is not an absolute risk reduction and does not specify how many subjects must receive liraglutide to prevent one event.

Confidence interval: The 95% CI of 0.778–0.968 quantifies uncertainty around the HR estimate. Its upper bound is below both the non-inferiority margin of 1.3 and the superiority reference value of 1.0.

P-value: The p-value of 0.005 measures the evidence against the specified superiority null hypothesis. It is not a measure of clinical importance or effect magnitude.

Sequential testing matters: Superiority was tested only after the registry's non-inferiority criterion had been established. The two inferential stages therefore should not be treated as unrelated independent tests.

7. Secondary Cardiovascular Results

The registry contains additional Cox regression analyses for cardiovascular and mortality outcomes. These analyses use all randomised subjects and compare liraglutide with placebo. The ClinicalTrials.gov record does not provide p-values for these secondary analyses, so interpretation here focuses on the hazard ratio and its 95% confidence interval rather than assigning significance based on an unreported test.

Secondary outcomeHR95% CIRegistry analysis
Expanded composite cardiovascular outcome0.8810.807–0.962Cox regression; superiority
All-cause death0.8470.739–0.971Cox regression; superiority
Cardiovascular death0.7830.656–0.934Cox regression; superiority
Non-fatal stroke0.8940.721–1.107Cox regression; superiority
Non-fatal myocardial infarction0.8780.747–1.031Cox regression; superiority
Hospitalisation for unstable angina pectoris0.9800.763–1.258Cox regression; superiority
Coronary revascularisation0.9120.797–1.044Cox regression; superiority
Hospitalisation for heart failure0.8720.727–1.046Cox regression; superiority

Several of these confidence intervals extend across 1.0, while others do not. That distinction describes the precision of the individual estimated treatment effects relative to 1.0; it should not be converted into a ranking of the individual components. The component analyses also have a different inferential role from the prespecified primary composite.

Reading component endpoints correctly

A composite endpoint can provide more statistical events than any single component, which can improve efficiency for detecting an overall treatment effect. Once the composite is separated into its individual components, event counts and statistical precision can differ substantially. A component HR therefore answers a narrower question than the primary composite HR.

8. Secondary Microvascular Results

The registry also reports time-to-event analyses for a composite microvascular outcome and its components. All analyses use all randomised subjects and Cox regression with treatment as a fixed factor.

Microvascular outcomeHR95% CIRegistry analysis
First composite microvascular outcome0.8410.730–0.969Cox regression; superiority
Composite nephropathy event0.7820.666–0.918Cox regression; superiority
New onset of persistent macroalbuminaria0.7380.602–0.905Cox regression; superiority
Persistent doubling of serum creatinine0.8900.667–1.189Cox regression; superiority
Need for continuous renal-replacement therapy0.8690.607–1.244Cox regression; superiority
Death due to renal disease1.5930.521–4.869Cox regression; superiority
Composite retinopathy1.1490.869–1.519Cox regression; superiority
Treatment with photocoagulation or intravitreal agents1.1590.869–1.546Cox regression; superiority
Development of diabetes-related blindness0.3350.004–30.847Cox regression; superiority
Vitreous haemorrhage1.4540.845–2.502Cox regression; superiority
Precision is especially important for rare outcomes. The confidence interval for development of diabetes-related blindness extends from 0.004 to 30.847. That very wide interval indicates substantial statistical uncertainty around the point estimate of 0.335. A point estimate by itself would therefore provide an incomplete description of this analysis.

The same principle applies to the renal-death estimate of 1.593 with a 95% CI of 0.521–4.869. The estimate is above 1.0, but the confidence interval is wide and includes values below and above 1.0. Without an accompanying p-value in the ClinicalTrials.gov record, no additional formal significance statement should be inferred.

9. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected subjects divided by subjects at risk.

Safety measureLiraglutidePlacebo
Serious adverse events2320 / 46682354 / 4672

These counts provide the number of subjects affected and the corresponding number at risk in each arm. They should be distinguished from a time-to-event efficacy endpoint: the ClinicalTrials.gov record does not provide a hazard ratio, confidence interval, or p-value for serious adverse events.

Liraglutide

2320 subjects with serious adverse events among 4668 subjects at risk.

Placebo

2354 subjects with serious adverse events among 4672 subjects at risk.

10. Non-Inferiority: Why the Margin Matters

The most instructive statistical feature of LEADER is that the primary endpoint was evaluated through a non-inferiority framework followed by a superiority assessment. This is different from simply asking whether a conventional null hypothesis of HR = 1 was rejected.

StageNull hypothesisCriterion reported in the registryObserved result
Non-inferiorityH0: HR ≥ 1.3Upper two-sided 95% CI below 1.3, or one-sided p-value less than 2.5%Upper CI = 0.968; P < 0.001
SuperiorityH0: HR ≥ 1.0Upper two-sided 95% CI below 1.0, or one-sided p-value less than 2.5%Upper CI = 0.968; P = 0.005

The margin of 1.3 has a fundamentally different interpretation from 1.0. The non-inferiority question asks whether the data exclude a hazard ratio as large as or larger than 1.3. The superiority question asks whether the data support a hazard ratio below 1.0.

Clinical Biostats interpretation

A common mistake is to interpret non-inferiority as merely "the p-value was significant." That is not the logic of a non-inferiority trial. The clinically relevant reference point is the prespecified margin. Here, the reported upper confidence limit of 0.968 lies below 1.3, directly satisfying the registry's stated non-inferiority criterion.

Once non-inferiority is established, the registry describes a separate superiority test against 1.0. Because the reported upper confidence limit is also below 1.0 and the superiority p-value is 0.005, the same estimated HR supports the second inferential stage as reported by the registry.

11. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

The primary endpoint is defined by time from randomisation to first occurrence of a cardiovascular event. Time-to-event methods are appropriate because subjects can have different follow-up times and can be censored before experiencing the event. The Cox model summarizes the relative event hazard between randomized groups while incorporating the timing of events and censoring.

What does an HR of 0.868 mean?

An HR of 0.868 is a relative measure of the instantaneous event rate estimated by the Cox model. In simple terms, 0.868 corresponds to an estimated hazard about 13.2% lower in the liraglutide group relative to placebo. It does not mean that the probability of an event was 13.2 percentage points lower, nor that every patient had the same relative reduction.

Why is the non-inferiority margin 1.3 rather than 1.0?

Non-inferiority and superiority answer different questions. A value of 1.0 represents equal hazards, whereas the non-inferiority margin of 1.3 represents the largest hazard ratio that the prespecified analysis would still regard as compatible with non-inferiority. The trial therefore needed to rule out the possibility that the true hazard ratio was 1.3 or higher.

Why use a two-sided 95% confidence interval for a one-sided non-inferiority question?

The registry explicitly describes the equivalence between the upper limit of a two-sided 95% confidence interval and the corresponding one-sided 2.5% test. This gives a convenient interval-based way to assess the one-sided non-inferiority hypothesis without treating the confidence interval as if it were a different inferential question.

Why was superiority tested only after non-inferiority?

The registry explicitly states that if non-inferiority was established for the primary outcome, a test for superiority was performed. This sequential structure distinguishes the initial safety/efficacy threshold from the stronger claim that the treatment has a hazard below the equality value of 1.0.

Why can a secondary HR above 1.0 still have a wide confidence interval?

The point estimate and its confidence interval convey different information. A secondary estimate such as 1.593 is the fitted point estimate, while its 95% CI of 0.521–4.869 describes substantial uncertainty around that estimate. When the interval is wide, the point estimate should not be treated as if it were precise.

12. Confidence Intervals and Statistical Precision

Confidence intervals are particularly useful in this trial because the posted analyses include both relatively precise estimates and very wide intervals. The interval is not simply a yes-or-no significance device; it shows how much uncertainty remains around the estimated hazard ratio.

Estimate95% CIWhat the interval illustrates
Primary cardiovascular outcome: HR 0.8680.778–0.968Relatively narrow interval around the primary estimate
All-cause death: HR 0.8470.739–0.971Interval remains below 1.0
Non-fatal stroke: HR 0.8940.721–1.107Interval crosses 1.0
Death due to renal disease: HR 1.5930.521–4.869Very broad uncertainty
Diabetes-related blindness: HR 0.3350.004–30.847Extremely broad uncertainty around a rare outcome estimate

The width of a confidence interval depends on the amount of statistical information available for the endpoint. Composite outcomes can accumulate more events than individual components, while rare individual outcomes can yield much less precise estimates.

13. Composite Endpoints

The primary endpoint combines three events: cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke. A composite endpoint allows the trial to evaluate a clinically defined collection of related events as a single time-to-event outcome.

What the composite answers

It measures time to the first occurrence of any of the specified cardiovascular components.

What it does not answer

It does not imply that the treatment effect is identical for cardiovascular death, myocardial infarction, and stroke individually.

Why components matter

The registry separately analyzes individual components, allowing their estimated treatment effects and precision to be examined.

Interpretive caution

The clinical importance and frequency of individual components can differ, so the composite should remain the primary unit of interpretation for the primary endpoint.

14. Analysis Population and Randomization

The posted primary and secondary efficacy analyses specify all randomised subjects. This is an important statistical feature because the treatment comparison is defined by randomized assignment rather than by treatment received after randomization.

Randomization principle
Randomized assignment → comparable treatment groups in expectation → outcome comparison

Randomization creates the basis for a causal comparison by making treatment assignment independent of measured and unmeasured baseline factors in expectation. The analysis then compares subsequent event histories according to that randomized assignment.

The registry does not provide baseline characteristics, stratification factors, crossover information, or missing-data/imputation methods in the ClinicalTrials.gov record. Those topics are therefore not reconstructed here.

15. Secondary Analysis and Multiplicity

The registry reports one primary endpoint and multiple secondary statistical analyses. The ClinicalTrials.gov record identifies the secondary analyses as superiority analyses but do not provide a complete multiplicity-adjustment scheme or adjusted p-values for the secondary endpoints.

Multiplicity caution: The secondary hazard ratios should not be treated as though each were an isolated confirmatory test. With multiple endpoints and component analyses, the probability of observing apparently notable findings by chance can increase unless the testing strategy explicitly controls the relevant familywise error rate or another prespecified error criterion.

This is especially relevant when a composite endpoint is followed by analyses of each component and additional microvascular outcomes. The appropriate interpretation is to distinguish the prespecified primary analysis from the broader set of secondary estimates.

16. Time-to-Event Analysis: What Is Actually Being Compared?

A time-to-event analysis does more than compare the percentage of subjects who eventually experience an event. It uses the timing of events and the follow-up contributed by subjects who remain event-free at the end of observation.

Conceptual survival quantity
S(t) = probability of remaining event-free through time t

Kaplan-Meier estimation describes the event-free survival function, while the Cox model estimates a relative hazard between treatment groups.

The registry's primary endpoint is therefore naturally expressed as a hazard ratio rather than as a simple difference in percentages. The ClinicalTrials.gov record does not include the underlying event times, censoring records, or Kaplan-Meier coordinates needed to reconstruct a survival curve.

17. Primary Endpoint vs Secondary Endpoints

RoleOutcomeStatistical perspective
PrimaryComposite cardiovascular outcomeFormal non-inferiority analysis followed by superiority testing
SecondaryExpanded composite cardiovascular outcomeSuperiority; Cox regression
SecondaryAll-cause deathSuperiority; Cox regression
SecondaryIndividual cardiovascular componentsSuperiority; Cox regression
SecondaryComposite microvascular outcomeSuperiority; Cox regression
SecondaryNephropathy and retinopathy outcomesSuperiority; Cox regression
SafetySerious adverse eventsAffected subjects / subjects at risk by arm

18. Important Limitations and Interpretation Issues

19. Why This Trial Matters Statistically

LEADER is a useful teaching case because it illustrates how a modern randomized cardiovascular-outcomes trial can combine randomization, masking, time-to-event analysis, non-inferiority testing, and a subsequent superiority assessment within a single primary endpoint framework.

ConceptHow it appears in LEADER
RandomizationThe trial is randomized with two parallel treatment arms.
BlindingThe registry identifies the study as double-masked.
Time-to-event endpointThe primary endpoint measures time from randomisation to the first cardiovascular event in a composite.
Cox regressionThe primary and secondary time-to-event analyses use Cox regression.
Hazard ratioThe primary treatment effect is expressed as an HR of 0.868.
Confidence intervalThe primary two-sided 95% CI is 0.778–0.968.
Non-inferiorityThe primary analysis uses a hazard-ratio margin of 1.3.
One-sided testingThe registry specifies one-sided tests for both non-inferiority and superiority.
SuperiorityAfter non-inferiority was established, superiority was tested against HR 1.0.
Composite endpointCardiovascular death, non-fatal myocardial infarction, and non-fatal stroke form the primary composite.
Secondary endpointsExpanded cardiovascular, mortality, microvascular, nephropathy, and retinopathy outcomes were analyzed.
PrecisionConfidence intervals range from relatively narrow primary estimates to very wide rare-event estimates.

20. Clinical Biostats Interpretation of the Primary Finding

Putting the primary analysis together

The primary HR was 0.868, with a two-sided 95% CI of 0.778–0.968. The registry's non-inferiority criterion required the upper confidence limit to be below 1.3, and the observed upper limit was 0.968. The registry therefore reports confirmation of non-inferiority.

The same confidence interval also has an upper limit below 1.0. The registry consequently proceeded to the prespecified superiority assessment, which reported P = 0.005.

Statistically, this sequence is important: the conclusion is not based solely on the fact that the HR is below 1.0. The primary analysis was designed around a non-inferiority margin and then, conditional on establishing non-inferiority, a superiority question.

Relative effect versus absolute effect

The ClinicalTrials.gov record provides the hazard ratio and its confidence interval, but not absolute event rates or median event times. Consequently, the most defensible interpretation from this dataset is relative: the estimated hazard under the Cox model was lower with liraglutide than with placebo.

An HR alone cannot tell a reader how many additional or fewer subjects experienced an event, how large an absolute risk difference was, or what the number needed to treat would be. Those quantities require absolute event information that is not included in the ClinicalTrials.gov record.

What the secondary results add

The secondary analyses broaden the statistical picture. The expanded cardiovascular composite and all-cause death have HRs below 1.0 with confidence intervals that remain below 1.0, while several individual cardiovascular and microvascular components have intervals that cross 1.0. Other rare outcomes have extremely wide intervals. This pattern illustrates why a clinical-trial analysis should consider both point estimates and precision rather than treating every estimate as equally informative.

21. Related Tutorials

Learn more about the methods used in this trial:

22. Related Statistical Calculators

23. Sources

Continue with the Clinical Biostats statistical pathway

Explore the statistical methods behind randomized trials, survival analysis, hazard ratios, confidence intervals, and non-inferiority testing.

24. Record Summary

LEADER provides a detailed example of time-to-event analysis within a randomized, double-blind phase 3 trial. Its primary cardiovascular endpoint was analyzed with a Cox proportional-hazards model in all randomised subjects, producing an HR of 0.868 with a two-sided 95% CI of 0.778–0.968. The registry specifies a non-inferiority margin of 1.3; the upper confidence limit was below that margin, and the reported non-inferiority p-value was <0.001. After non-inferiority was established, the registry reports a superiority p-value of 0.005 against HR 1.0.

The secondary analyses extend the statistical assessment to an expanded cardiovascular composite, all-cause death, individual cardiovascular components, and multiple microvascular outcomes. Their confidence intervals demonstrate an important lesson in statistical interpretation: estimates with intervals close to their reference value can be more informative than point estimates alone, while rare outcomes can generate very wide uncertainty intervals. The serious-adverse-event data provide an additional safety perspective but are reported here only as affected subjects and subjects at risk because no comparative effect estimate was reported.

Clinical Biostats methodology: A trial-results page should not merely repeat an outcome estimate. The goal is to explain what the estimand means, how the analysis addresses the endpoint structure, how the confidence interval and hypothesis test should be interpreted, and which conclusions are supported by the registry-reported evidence.