This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical results on this page are restricted to the ClinicalTrials.gov record. The ClinicalTrials.gov record contains posted outcome measures but no formal statistical-analysis records, so no treatment-effect estimate, confidence interval, or p-value is asserted here.
1. Trial at a Glance
PLATO was a completed randomized phase 3 trial comparing ticagrelor with clopidogrel in patients with acute coronary syndrome. The registry describes a two-arm, parallel-group, quadruple-masked treatment trial with 18,624 enrolled participants and two registered primary binary endpoints.
| Feature | PLATO |
|---|---|
| Trial name | PLATO |
| Brief title | A Comparison of Ticagrelor (AZD6140) and Clopidogrel in Patients With Acute Coronary Syndrome |
| Phase | Phase 3 |
| Condition | Acute Coronary Syndrome |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 18,624 |
| Interventions | Ticagrelor; Clopidogrel |
| Primary endpoints | 2 registered binary endpoints |
| Trial status | Completed |
| Start | 2006-10 |
| Primary completion | 2009-03 |
| Lead sponsor | AstraZeneca |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT00391872 |
2. Clinical Question
The registered clinical question can be expressed as a randomized comparison of ticagrelor versus clopidogrel in patients with acute coronary syndrome, with treatment effects evaluated using two registered primary endpoints through the specified follow-up periods.
Population
Patients with acute coronary syndrome enrolled in the PLATO trial. The ClinicalTrials.gov record reports 18,624 enrolled participants.
Intervention
Ticagrelor, classified in the registry as a drug intervention.
Comparator
Clopidogrel, classified in the registry as a drug intervention.
Primary questions
How do the randomized groups compare with respect to the composite of vascular death, MI, and stroke, and with respect to major bleeding during the registered follow-up periods?
3. Trial Design
Random allocation is the central design mechanism for creating comparable treatment groups in expectation and supports an intention-to-treat treatment comparison.
Participants were assigned to one of two treatment arms rather than being repeatedly exposed to both interventions as part of a crossover design.
The registry identifies the trial as quadruple masked. The ClinicalTrials.gov record does not specify the four masked roles, so they are not inferred here.
The registered primary purpose was treatment, consistent with a randomized comparison intended to evaluate therapeutic interventions.
4. Trial Timeline and Registry Status
PLATO begins
The ClinicalTrials.gov record identifies October 2006 as the trial start.
Primary completion
The ClinicalTrials.gov record identifies March 2009 as the primary completion date.
Trial completed
The current the ClinicalTrials.gov record classifies PLATO as completed and reports that results have been posted.
The timeline is useful for distinguishing trial conduct from statistical reporting. A posted result does not necessarily mean that every underlying statistical comparison is reproduced in the registry's structured statistical-analysis fields. In the ClinicalTrials.gov record, the statistical-analysis array is empty even though outcome measures are posted.
5. Primary Endpoints
| Endpoint | Time frame | Registered definition | Type |
|---|---|---|---|
| Participants With Any Event From the Composite of Death From Vascular Causes, Myocardial Infarction (MI), and Stroke | Randomization up to 12 months | Participants with death from vascular causes, MI, or stroke. If no event, censoring occurs at the earliest of patient withdrawal consent or date of scheduled withdrawal from therapy. Intention To Treat (ITT) analysis of whole population. Events were adjudicated by an endpoint committee. | Binary |
| Participants With Any Major Bleeding Event | First dosing up to 12 months | Participants with major (fatal/life-threatening or other) bleed by a study protocol scale based on need for treatment, number of transfusions, hemoglobin decrease, and other factors. Events were adjudicated by an endpoint committee. | Binary |
These two endpoints have an important statistical distinction. The first is a composite efficacy endpoint in which a participant is counted if any one of several specified clinical events occurs. The second is a binary safety endpoint based on the occurrence of a protocol-defined major bleeding event. Both ultimately reduce each participant's outcome to an event/no-event classification for the registered analysis window.
6. Endpoint Construction and Event Adjudication
The composite endpoint
The first primary endpoint combines three clinically distinct events: death from vascular causes, myocardial infarction, and stroke. A participant enters the binary event category if any of those components occurs during the registered period. This is statistically efficient because a single composite can accumulate events from multiple clinically relevant pathways, but the interpretation is necessarily a statement about the composite rather than about every individual component separately.
The registry also states that events were adjudicated by an endpoint committee. That feature is important because the outcome classification is not simply an unreviewed report of what happened during follow-up. An adjudication process is intended to apply prespecified clinical definitions consistently when determining whether an event qualifies for the endpoint.
The major bleeding endpoint
The second primary endpoint is also binary, but its event definition is substantially more structured than a simple report of any bleeding. The registry describes major bleeding as fatal/life-threatening or other major bleeding according to a study protocol scale incorporating factors such as treatment requirements, number of transfusions, hemoglobin decrease, and other criteria.
For statistical analysis, the important consequence is that the protocol's clinical classification determines whether each participant is counted as having experienced the primary safety event. The statistical model should therefore be applied to the protocol-defined endpoint, not to an independently reconstructed definition of bleeding.
7. Planned Analysis
The ClinicalTrials.gov record indicates that results have been posted for both primary endpoints, but no formal statistical analyses were posted. Consequently, the ClinicalTrials.gov record does not contain a formal treatment-effect estimate, confidence interval, or p-value for either primary endpoint.
| Primary endpoint | What is posted | What a formal binary analysis would normally estimate |
|---|---|---|
| Composite of vascular death, MI, and stroke | Endpoint definition, time frame, ITT population statement, and adjudication statement | Event proportions by randomized treatment group, together with an appropriate between-group effect measure and confidence interval |
| Major bleeding | Endpoint definition, time frame, protocol-scale description, and adjudication statement | Major-bleeding proportions by treatment group, together with an appropriate between-group effect measure and confidence interval |
For a binary endpoint, several effect measures can be appropriate depending on the prespecified statistical analysis plan. A treatment comparison can be expressed using a risk difference, risk ratio, or odds ratio. A hypothesis test can accompany the effect estimate, but the exact test and modeling framework should come from the prespecified statistical plan rather than being retroactively selected from the observed data.
A risk difference describes the absolute difference in the probability of the event between treatment and comparator groups over the specified analysis period.
A risk ratio describes the event probability in the treatment group relative to the comparator group.
An odds ratio compares odds rather than probabilities. It should not automatically be described as a risk ratio, particularly when events are not rare.
8. Statistical Methodology
Why the primary endpoints are binary
Both registered primary endpoints classify participants according to whether an event occurred within a defined observation window. This is different from a continuous endpoint such as blood pressure or a time-to-event endpoint analyzed directly as a survival distribution. The registry labels both primary endpoints as Binary.
A binary endpoint is naturally represented by an indicator variable:
The treatment comparison then concerns the distribution of this indicator across the randomized groups.
Intention-to-treat analysis
The registry explicitly states that the composite primary endpoint uses an Intention To Treat (ITT) analysis of the whole population. Under the ITT principle, randomized participants are analyzed according to the treatment group to which they were assigned rather than being reassigned according to subsequent treatment exposure.
That principle is important because the causal interpretation of a randomized comparison depends on preserving the original allocation. If participants discontinue therapy, deviate from treatment, or otherwise fail to follow the intended treatment pathway, removing them selectively from the primary efficacy analysis can undermine the protection created by randomization.
Event adjudication
Both primary endpoints were adjudicated by an endpoint committee according to the ClinicalTrials.gov record. For the composite endpoint, adjudication determines whether a reported clinical event qualifies as vascular death, MI, or stroke. For major bleeding, adjudication applies the protocol-defined classification system.
Fixed follow-up windows
The first endpoint is assessed from randomization up to 12 months, while the bleeding endpoint is assessed from first dosing up to 12 months. This means the analysis is tied to the registered observation definitions rather than being an unrestricted comparison of events occurring at any time after enrollment.
Choice of effect measure
For a binary endpoint, the effect measure should be selected before looking at the results. A risk difference is often particularly transparent because it expresses the absolute difference in event probability. A risk ratio gives a relative comparison of probabilities, while an odds ratio compares odds. None is universally interchangeable with the others.
Confidence intervals
A confidence interval should accompany the primary effect estimate whenever the formal analysis is reported. The interval describes statistical uncertainty around the estimated treatment effect under the specified analysis framework. It does not describe the range of effects that individual patients will experience.
9. Statistical Methods Explained
Why is the primary endpoint a composite?
A composite endpoint combines multiple clinically defined event types into a single outcome. The statistical advantage is that more participants may experience at least one component of the composite than would experience any single component alone, potentially increasing the amount of information available for the overall comparison. The tradeoff is interpretability: a treatment effect for the composite does not automatically imply the same effect for every component.
Why does the ITT principle matter?
ITT preserves the randomized comparison. If participants were selectively removed after randomization because of treatment discontinuation or other post-randomization events, the resulting groups could differ for reasons introduced after allocation. The registry specifically identifies ITT analysis for the composite primary endpoint, making preservation of randomized assignment part of the registered analysis description.
Why does endpoint adjudication matter?
Clinical events such as MI, stroke, vascular death, and major bleeding require operational definitions. Independent adjudication can standardize the classification process across participants and treatment groups. Statistically, better outcome classification reduces one source of measurement inconsistency, although adjudication cannot eliminate all uncertainty in clinical classification.
What does a risk difference tell us?
If the event probability were 20% in one randomized group and 15% in the other, the risk difference would be 5 percentage points. That is an absolute comparison. It is different from saying that one group has 25% lower relative risk. The actual PLATO event probabilities are not reported in the ClinicalTrials.gov record used for this page, so no numerical treatment effect is inferred here.
What does a p-value tell us?
A p-value is a measure calculated under a specified null hypothesis and statistical model. It addresses how compatible the observed data are with that null model; it does not quantify the magnitude or clinical importance of an effect. A small p-value can accompany a small effect in a large trial, while a clinically important estimate can have substantial statistical uncertainty in a smaller study.
Why should the two primary endpoints be considered separately?
The composite cardiovascular endpoint and major bleeding answer different clinical questions. One concerns vascular death, MI, or stroke; the other concerns protocol-defined major bleeding. Even if both are analyzed as binary outcomes, they should not be collapsed into a single interpretation unless the prespecified trial framework explicitly defines such a combined benefit-risk analysis.
10. Missing Data, Censoring, and Follow-Up
The registry definition for the composite endpoint states that if no event occurs, censoring occurs at the earliest of patient withdrawal consent or the date of scheduled withdrawal from therapy. That statement establishes an explicit rule for participants without an observed qualifying event under the registered endpoint definition.
For a binary endpoint, however, it is important to distinguish an ordinary complete observation from an observation that ends before the planned assessment window. A participant who has no event during the defined period is different from a participant for whom outcome ascertainment becomes unavailable before the relevant period ends.
What the registry specifies
The composite endpoint definition specifies the censoring rule for participants without an event and identifies the ITT population.
What the ClinicalTrials.gov record does not specify
The ClinicalTrials.gov record does not describe a formal missing-data or imputation method for either primary endpoint.
11. Multiplicity and Two Primary Endpoints
PLATO has two registered primary endpoints. That fact matters statistically because the trial is not asking only one primary question. When multiple primary hypotheses are tested confirmatorily, the prespecified testing strategy determines how the overall type I error is controlled.
The ClinicalTrials.gov record does not specify a multiplicity adjustment, hierarchical testing sequence, alpha allocation, or other formal error-control procedure. Accordingly, no particular multiplicity strategy is attributed to PLATO here.
| Issue | What the ClinicalTrials.gov record establishes | What is not established by the ClinicalTrials.gov record |
|---|---|---|
| Number of primary endpoints | 2 | — |
| Endpoint type | Both binary | — |
| Multiplicity strategy | Not reported | Specific alpha adjustment or hierarchy |
| Formal p-values | Not reported in statistical-analysis records | Any numerical p-value |
| Confidence intervals | Not reported in statistical-analysis records | Any numerical confidence interval |
This distinction is especially important when reading a large randomized trial. The existence of two primary endpoints does not itself tell the reader whether both had to meet a particular significance threshold, whether they were tested hierarchically, or whether another prespecified strategy was used. Those details belong to the trial's statistical analysis plan.
12. Blinding and Its Statistical Role
The trial is registered as quadruple masked. Masking is not a substitute for randomization, but it addresses a different source of potential bias: knowledge of treatment assignment.
Randomization
Controls treatment allocation through a randomized assignment mechanism and supports comparability of groups in expectation.
Blinding
Reduces the opportunity for knowledge of treatment assignment to influence trial conduct, outcome assessment, or other study processes.
Adjudication
Provides a structured mechanism for determining whether reported clinical events meet the registered endpoint definitions.
ITT analysis
Preserves the randomized treatment assignment when evaluating the primary efficacy comparison.
These four mechanisms operate at different points in the evidence chain. Randomization addresses allocation, masking addresses information about allocation, adjudication addresses outcome classification, and ITT addresses how the randomized comparison is maintained during analysis.
13. Safety Analysis Framework
The second registered primary endpoint is a safety endpoint measuring participants with any major bleeding event from first dosing up to 12 months. The registry defines major bleeding using a study protocol scale incorporating treatment requirements, number of transfusions, hemoglobin decrease, and other factors.
Because the ClinicalTrials.gov record does not provide event counts or arm-specific estimates, this page does not state a numerical bleeding rate or treatment effect. The appropriate interpretation is therefore methodological rather than numerical.
| Safety-analysis question | PLATO registry information reported |
|---|---|
| Safety endpoint | Participants with any major bleeding event |
| Observation period | First dosing up to 12 months |
| Endpoint classification | Protocol-defined major bleeding |
| Classification factors | Need for treatment, number of transfusions, hemoglobin decrease, and other factors |
| Adjudication | Endpoint committee |
| Formal statistical estimate in the ClinicalTrials.gov record | Not provided |
A safety endpoint should also be interpreted as a separate statistical question from the cardiovascular composite. A treatment comparison can affect efficacy and safety endpoints in different directions, and those endpoints cannot be combined into a single numerical conclusion without a prespecified benefit-risk framework.
14. What Can and Cannot Be Concluded From the registry Results Record
The ClinicalTrials.gov record establishes that PLATO was a completed randomized phase 3 parallel trial with quadruple masking, 18,624 enrolled participants, two drug interventions, and two registered binary primary endpoints. Results were posted, and both primary endpoints have detailed registry definitions and specified time frames.
The registry-reported statistical-analysis array contains no formal treatment-effect estimates, confidence intervals, or p-values. Therefore, this page does not claim a numerical difference between ticagrelor and clopidogrel for either primary endpoint.
A statistically responsible trial-results page should not infer an effect estimate merely because a registry reports that an outcome measure has results. The analysis must be tied to an actual reported statistical comparison or to a clearly identified prespecified analysis method. Here, the ClinicalTrials.gov record supports methodological interpretation but not a numerical efficacy or safety result.
15. Statistical Interpretation of Binary Treatment Effects
Although the ClinicalTrials.gov record does not contain numerical treatment effects, the structure of the endpoints allows the principal statistical quantities to be defined clearly.
Absolute effect
The risk difference compares event probabilities directly:
where pT is the event probability in the ticagrelor group and pC is the event probability in the clopidogrel group.
The sign of the risk difference depends on how the treatment and comparator are ordered. For a harmful event such as major bleeding, a negative risk difference would indicate a lower observed event probability in the treatment group. For an efficacy event such as the composite endpoint, the interpretation depends on whether the event is considered clinically adverse—which it is in this trial.
Relative effect
The risk ratio is:
An RR below 1 would indicate a lower event probability in the ticagrelor group relative to clopidogrel, whereas an RR above 1 would indicate a higher event probability.
Odds-based effect
The odds ratio compares the odds of the event rather than the probabilities themselves. For rare outcomes, odds ratios and risk ratios can be numerically similar, but they are not mathematically identical. The distinction becomes increasingly important as event probabilities become larger.
Confidence intervals and precision
A point estimate alone is incomplete. A confidence interval communicates the statistical precision of the estimated treatment effect. A wide interval indicates greater uncertainty, while a narrower interval indicates greater precision under the specified model and sampling assumptions.
P-values and effect size
A p-value does not measure how large the treatment effect is. A useful results table should therefore present the effect estimate and its confidence interval rather than relying on a significance statement alone.
16. Analysis Populations
The registry explicitly specifies an ITT analysis of the whole population for the composite primary endpoint. This is the key analysis-population statement available in the ClinicalTrials.gov record.
| Population / analysis concept | Evidence in reported PLATO data |
|---|---|
| Intention-to-treat | Explicitly stated for the composite primary endpoint |
| Whole population | Explicitly stated for the composite primary endpoint |
| Per-protocol analysis | Not specified in the ClinicalTrials.gov record |
| As-treated analysis | Not specified in the ClinicalTrials.gov record |
| Safety population definition | Not specified in the ClinicalTrials.gov record |
The absence of a registry-reported per-protocol or as-treated definition does not establish that such analyses were never performed. It means only that they are not part of the ClinicalTrials.gov record. This is an important distinction when separating the registry evidence available here from a broader trial-reporting history.
17. Limitations
- No formal statistical-analysis records reported: the structured statistical-analysis field is empty even though results are posted, so numerical treatment effects cannot be reproduced from the ClinicalTrials.gov record.
- No event counts by arm reported: the available data identify the endpoint definitions but do not provide the number of participants with each primary endpoint in either treatment group.
- No confidence intervals reported: there is no basis in the ClinicalTrials.gov record for reporting a numerical confidence interval.
- No p-values reported: a formal hypothesis-test result cannot be stated without the corresponding reported analysis.
- Multiplicity information is incomplete: the registry data establish two primary endpoints but do not specify an alpha-allocation or multiplicity-control strategy.
- Missing-data methods are incomplete: the composite endpoint includes a censoring rule, but the ClinicalTrials.gov record does not specify an imputation method for missing binary outcomes.
- Masking roles are not specified: the trial is quadruple masked, but the ClinicalTrials.gov record does not identify the four masked groups or roles.
- No subgroup estimates are reported: subgroup-specific treatment effects cannot be evaluated from the ClinicalTrials.gov record.
- No Bayesian analysis is specified: the ClinicalTrials.gov record contains no Bayesian model, prior, posterior estimate, or Bayesian decision criterion.
- No crossover design is specified: the registered design is parallel, and the ClinicalTrials.gov record does not describe crossover between treatments.
- Endpoint timing differs: the composite begins at randomization, whereas major bleeding begins at first dosing; those origins should not be silently treated as identical.
18. Why This Trial Matters Statistically
PLATO is a useful statistical teaching case because the registry description brings together several foundational principles of randomized clinical-trial analysis without requiring a complicated endpoint scale. The primary outcomes are binary, but the design surrounding those outcomes is substantial: randomization, parallel treatment assignment, quadruple masking, ITT analysis, clinical endpoint adjudication, a composite endpoint, a protocol-defined safety endpoint, and two primary questions.
| Concept | How it appears in PLATO |
|---|---|
| Randomization | Participants were randomized between ticagrelor and clopidogrel. |
| Parallel design | The registry identifies the design model as parallel. |
| Blinding | The registry identifies the trial as quadruple masked. |
| Binary endpoints | Both registered primary endpoints are classified as binary. |
| Composite endpoint | Vascular death, MI, and stroke are combined into one primary endpoint. |
| ITT analysis | The composite endpoint specifies ITT analysis of the whole population. |
| Endpoint adjudication | Both primary endpoints were adjudicated by an endpoint committee. |
| Different time origins | The composite begins at randomization; major bleeding begins at first dosing. |
| Multiplicity | Two primary endpoints create a prespecified multiple-testing question. |
| Safety analysis | Major bleeding is a registered primary binary endpoint. |
19. Composite Endpoints: A Closer Statistical Look
The first primary endpoint is a composite of death from vascular causes, MI, and stroke. Composite endpoints can be attractive in large outcome trials because they combine several clinically important events into a single primary analysis.
But the composite should be interpreted as a single endpoint. Suppose a treatment changed one component substantially while having little effect on another. A statistically significant composite result would not establish that every component was affected to the same degree.
Statistical advantage
Combining several event types can increase the number of participants experiencing the composite outcome and therefore increase the information available for the primary comparison.
Interpretive limitation
The composite estimate represents the combined endpoint. It does not automatically describe the treatment effect for each component separately.
Another important issue is whether the components are sufficiently clinically related and whether the event frequencies are balanced enough for the composite to be meaningful. Those questions require component-level results and the prespecified statistical analysis plan, neither of which is contained in the ClinicalTrials.gov record.
20. Why the Major Bleeding Endpoint Is Statistically Different
The major bleeding endpoint is also binary, but its clinical interpretation differs from that of the cardiovascular composite. It begins at first dosing, rather than randomization, and its definition is based on a protocol scale involving clinical consequences and treatment requirements.
That difference illustrates a general principle: two endpoints can share the same statistical data type while having different scientific interpretations. "Binary" describes the mathematical form of the outcome; it does not describe the clinical meaning of the outcome.
Both are binary, but they measure different clinical phenomena and use different registered time origins and definitions.
21. Primary Question vs Statistical Question
The clinical question asks whether outcomes differ between ticagrelor and clopidogrel. The statistical question is more specific: given randomized assignment, how large is the observed difference in the probability of each prespecified endpoint, and how much uncertainty surrounds that estimate?
| Level | Question |
|---|---|
| Clinical | Do the two treatments differ with respect to the registered cardiovascular composite and major bleeding endpoints? |
| Design | Were participants randomized and assigned in a parallel, quadruple-masked trial? |
| Endpoint | Did a participant experience the protocol-defined event during the registered observation period? |
| Estimation | What is the absolute or relative difference in event probability between treatment groups? |
| Uncertainty | How precise is the estimated treatment effect? |
| Inference | What hypothesis-testing framework was prespecified for the two primary endpoints? |
The ClinicalTrials.gov record answer the design and endpoint-definition questions but do not provide the numerical estimation and inference results.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Statistical Calculators
24. Sources
- ClinicalTrials.gov: NCT00391872 — PLATO.
- Linked publication: PubMed PMID 37713684.
- Linked publication: PubMed PMID 36565198.
- Linked publication: PubMed PMID 36302586.
- Linked publication: PubMed PMID 35224730.
- Linked publication: PubMed PMID 34195679.
Continue through Clinical Biostats
Explore clinical-trial methods, biostatistics tutorials, and statistical calculators that build from the design and endpoint concepts illustrated by PLATO.
25. Record Summary
PLATO provides a clear example of how a large randomized phase 3 trial can be understood through its statistical architecture even when the structured registry analysis fields do not contain the numerical treatment comparisons. The ClinicalTrials.gov record establishes a randomized, parallel, quadruple-masked design with 18,624 enrolled participants and two registered binary primary endpoints: a composite of vascular death, MI, and stroke assessed from randomization up to 12 months, and major bleeding assessed from first dosing up to 12 months.
The composite endpoint is specified for an ITT analysis of the whole population, and both endpoints were adjudicated by an endpoint committee. These details are central to interpretation because they define the population, event classification, and observation framework for the primary analyses.
At the same time, the ClinicalTrials.gov record contains no formal effect estimates, confidence intervals, or p-values. A rigorous statistical review should therefore distinguish what the registry establishes from what would require the underlying statistical analysis plan or a formal posted analysis. The appropriate educational lesson is not to fill that gap with reconstructed numbers, but to understand how the binary endpoints should be analyzed and interpreted once the prespecified statistical results are available.