This page separates reported registry information from statistical interpretation. The ClinicalTrials.gov record includes registered endpoints, design characteristics, posted outcome measures, and serious adverse-event counts, but do not provide formal statistical-analysis results such as effect estimates, confidence intervals, or p-values.
1. Trial at a Glance
HORIZONS-AMI was a randomized phase 3 factorial trial in myocardial infarction with 3602 participants. The design simultaneously evaluated a pharmacology comparison involving bivalirudin and unfractionated heparin and a stent comparison involving paclitaxel-eluting and bare metal stents.
| Feature | HORIZONS-AMI |
|---|---|
| Trial name | HORIZONS-AMI |
| Brief title | Harmonizing Outcomes With Revascularization and Stents in Acute Myocardial Infarction |
| Phase | Phase 3 |
| Condition | Myocardial Infarction |
| Allocation | Randomized |
| Design model | Factorial |
| Masking | Single |
| Primary purpose | Treatment |
| Enrollment | 3602 |
| Primary endpoint count | 3 |
| Primary endpoint type | Binary |
| Results posted | Yes |
| Outcome measures posted | 7 |
| Lead sponsor | Cardiovascular Research Foundation, New York |
| Sponsor type | OTHER |
| Study status | COMPLETED |
2. Clinical Question
The HORIZONS-AMI registry record describes a factorial randomized trial addressing two treatment comparisons within participants with myocardial infarction. The statistical structure is important because the trial does not represent a single simple two-group comparison: it contains a pharmacology comparison and a stent comparison within the same overall trial design.
Population
Participants enrolled in a phase 3 trial for the condition of myocardial infarction.
Pharmacology intervention
Bivalirudin.
Pharmacology comparator
Unfractionated heparin.
Stent comparison
Paclitaxel-eluting stent versus bare metal stent.
The registered primary questions can therefore be separated into two domains: whether the pharmacology comparison differs with respect to major adverse ischemic cardiac events and major bleeding at 30 days, and whether the stent comparison differs with respect to ischemic target lesion revascularization and a composite of death, reinfarction, stroke, or stent thrombosis at 1 year.
3. Trial Design
Bivalirudin
- Drug intervention.
- Serious adverse events: 397/1800.
Unfractionated heparin
- Drug comparator.
- Serious adverse events: 439/1802.
Paclitaxel-eluting stent
- Device intervention.
- Serious adverse events: 473/2257.
Bare metal stent
- Device comparator.
- Serious adverse events: 185/749.
The registry identifies the design model as FACTORIAL. A factorial design allows more than one intervention factor to be studied within the same randomized trial structure. Its statistical interpretation therefore differs from that of a trial with only one treatment factor.
4. What the Factorial Design Means
A factorial design crosses treatment factors so that participants can contribute information to more than one comparison. In HORIZONS-AMI, the registry identifies both a pharmacology factor and a stent factor.
| Factor | Levels identified in the registry | Primary endpoint domain |
|---|---|---|
| Pharmacology | Bivalirudin; Unfractionated heparin | Major adverse ischemic cardiac events and major bleeding at 30 days |
| Stent | Paclitaxel-eluting stent; Bare metal stent | Ischemic target lesion revascularization at 1 year; death, reinfarction, stroke, or stent thrombosis at 1 year |
The key statistical advantage is efficiency: instead of requiring an entirely separate trial for every treatment factor, a factorial structure can investigate multiple factors within one trial population. The corresponding caution is that the analysis must respect the factorial structure and, when relevant, examine whether the effect of one factor depends on the level of the other factor.
5. Primary Endpoints
| Endpoint | Time frame | Type | Registered definition |
|---|---|---|---|
| Pharmacology Arm - Major Adverse Ischemic Cardiac Events and Major Bleeding Events | 30 Days | Binary | Number of participants with major adverse cardiovascular events (death, reinfarction, target-vessel revascularization for ischemia, and stroke) and major bleeding (bleeding adjudicated as not related to coronary artery bypass grafting). |
| Stent Arm - Ischemic Target Lesion Revascularization | 1 year | Binary | Number of Participants With Ischemic Target Lesion Revascularization. |
| Stent Arm - Death, Reinfarction, Stroke, or Stent Thrombosis | 1 year | Binary | Number of Participants With Death, Reinfarction, Stroke, or Stent Thrombosis. |
All three registered primary endpoints are classified as binary. That means the registry defines each endpoint in terms of whether the participant meets the specified event criterion within the stated time frame.
6. Statistical Methodology
The ClinicalTrials.gov record identifies three binary primary endpoints and a factorial randomized design, but they do not supply the formal statistical analyses used for the posted results. Consequently, this page does not assign unreported effect measures, confidence intervals, p-values, or model specifications to the trial.
Binary endpoint analysis
For a binary endpoint, the basic analysis compares the proportion of participants experiencing the event in the relevant randomized groups. Depending on the prespecified statistical analysis plan, appropriate measures can include a risk difference, risk ratio, odds ratio, or another model-based treatment effect.
The registry's primary endpoint type indicates that participants are classified according to whether the specified event occurred within the registered time frame. The ClinicalTrials.gov record does not provide the corresponding event counts for the three primary endpoints.
Factorial analysis
Because the design is identified as factorial, the statistical analysis should distinguish the effects of the pharmacology factor from the effects of the stent factor. A factorial analysis also raises the question of interaction between the factors.
Randomization
Randomization provides the design framework for comparing treatment groups while reducing systematic differences in participant characteristics expected from nonrandom allocation. The ClinicalTrials.gov record identifies the allocation as randomized but does not provide the detailed randomization procedure or stratification variables.
Time-specific endpoint definition
The primary endpoints use different time frames: 30 days for the pharmacology endpoint and 1 year for the two stent-arm endpoints. This distinction matters statistically because an event occurring outside the registered endpoint window does not satisfy that endpoint's stated time definition.
Composite endpoint structure
The pharmacology endpoint combines major adverse cardiovascular events and major bleeding, while one stent endpoint combines death, reinfarction, stroke, or stent thrombosis. Composite endpoints require interpretation of both the overall composite and the components because a difference in the composite can arise from different component patterns.
7. Planned Analysis
Pharmacology Arm - Major Adverse Ischemic Cardiac Events and Major Bleeding Events
The registry specifies a binary endpoint assessed at 30 days. A conventional analysis would compare the proportion of participants meeting the composite event definition between the bivalirudin and unfractionated heparin groups. Depending on the prespecified analysis plan, the treatment effect could be expressed using a risk difference, relative risk, odds ratio, or another appropriate binary-outcome model.
The endpoint itself combines two clinically different domains: major adverse cardiovascular events and major bleeding. A complete interpretation would therefore examine the overall endpoint as well as its clinically relevant components, provided those component results were available.
Stent Arm - Ischemic Target Lesion Revascularization
The registry defines this as the number of participants with ischemic target lesion revascularization at 1 year. A standard analysis would compare the event proportion between participants assigned to paclitaxel-eluting stent and those assigned to bare metal stent.
Stent Arm - Death, Reinfarction, Stroke, or Stent Thrombosis
This is a binary composite endpoint assessed at 1 year. A typical analysis would compare the proportion of participants experiencing at least one of the listed events between the relevant stent groups, followed by examination of the individual components if those data are available.
8. Results Posted on ClinicalTrials.gov
The ClinicalTrials.gov record states that results are posted and that 7 outcome measures are posted. However, the ClinicalTrials.gov record does not contain formal statistical-analysis records for those outcomes.
| Registry information | Value |
|---|---|
| Results posted | Yes |
| Outcome measures posted | 7 |
| Primary endpoints | 3 |
| Formal statistical analyses reported | None |
| Primary endpoint type | Binary |
This distinction is important. "Results posted" indicates that outcome information exists in the registry record, but the ClinicalTrials.gov recordset does not include the corresponding formal comparison statistics.
9. Safety
The ClinicalTrials.gov record provides serious adverse-event counts by arm. These counts are reported as affected participants divided by participants at risk.
| Arm | Serious adverse events | Affected / at risk |
|---|---|---|
| Pharmacology Arm - Bivalirudin | Serious AEs | 397/1800 |
| Pharmacology Arm - Unfractionated heparin | Serious AEs | 439/1802 |
| Stent Arm - Paclitaxel-eluting stent | Serious AEs | 473/2257 |
| Stent Arm - Bare metal stent | Serious AEs | 185/749 |
These figures describe serious adverse events as reported in the ClinicalTrials.gov record. They should not be interpreted as a formal comparison of safety without the corresponding statistical-analysis specification, because the ClinicalTrials.gov record does not provide a confidence interval, p-value, or adjusted treatment effect.
The numerator gives the number of participants affected and the denominator gives the number at risk for the reported arm. A direct comparison of these fractions can describe observed event proportions, but a formal inferential conclusion would require the prespecified comparison method and its uncertainty estimate.
10. Statistical Methods Explained
Why use a factorial design?
A factorial design allows more than one treatment factor to be studied within the same trial. Here, the registry identifies a pharmacology comparison and a stent comparison. This can make the trial statistically efficient because information from the same randomized population can contribute to multiple treatment questions.
What is a main effect in a factorial trial?
A main effect describes the average effect associated with one treatment factor across the levels of another factor, under the assumptions of the factorial analysis. For HORIZONS-AMI, one main-effect question concerns pharmacology and another concerns stent type.
What is an interaction?
An interaction asks whether the effect of one treatment factor changes depending on the level of another factor. For example, a formal interaction analysis could ask whether the pharmacology comparison differs according to stent assignment. The ClinicalTrials.gov record does not report such an analysis.
Why are the primary endpoints called binary?
A binary endpoint classifies each participant into an event or no-event state for the defined assessment window. This differs from a continuous endpoint, which measures a quantity such as a laboratory value, and from a time-to-event endpoint, which uses the timing of an event and censoring information.
Why does the 30-day versus 1-year time frame matter?
The endpoint definition includes its time frame. A 30-day binary endpoint and a 1-year binary endpoint answer different questions even when both are represented as event/no-event outcomes. Their event probabilities and clinical interpretations therefore should not be treated as interchangeable.
Why should composite endpoints be examined component by component?
A composite endpoint groups several events into a single binary outcome. The overall result can be driven disproportionately by one component, while other components may behave differently. For HORIZONS-AMI, the registered composites contain several distinct clinical events, so component-level interpretation would be important if those results were available.
Why is the absence of a reported confidence interval important?
A point estimate alone does not describe its statistical precision. A confidence interval provides information about uncertainty around an estimated treatment effect.
11. Understanding Binary Treatment Effects
For binary endpoints, several effect measures can communicate the treatment difference. Each answers a somewhat different question.
| Measure | Conceptual interpretation | Key caution |
|---|---|---|
| Risk difference | Difference in event probabilities between groups. | Depends directly on the baseline event probability. |
| Risk ratio | Ratio of event probabilities between groups. | Relative effects can appear large even when absolute differences are small. |
| Odds ratio | Ratio of event odds between groups. | It is not generally identical to the risk ratio. |
| Number needed to treat | Derived from an absolute risk difference when appropriate. | Requires a clearly defined outcome window and meaningful absolute-risk estimate. |
The registry classification of the three primary endpoints as binary establishes the general statistical structure, but it does not establish which of these effect measures was prespecified. That information should come from the statistical analysis plan rather than being inferred from the endpoint type.
12. Factorial Design and Interaction
The interaction term represents whether the effect of one factor depends on the level of the other factor. This is a conceptual representation of factorial analysis, not a reported model from the registry-reported HORIZONS-AMI data.
The interaction term is particularly important when the scientific question includes whether two interventions work differently in combination than would be expected from their separate effects. A non-significant interaction and an absence of an interaction analysis are not equivalent: the former is evidence from a statistical test, while the latter simply means that the relevant result has not been reported here.
Main-effect question
What is the treatment effect associated with the pharmacology factor when considered within the factorial trial structure?
Second main-effect question
What is the treatment effect associated with the stent factor when considered within the factorial trial structure?
Interaction question
Does the effect of the pharmacology factor depend on stent assignment, or vice versa?
Registry evidence reported
The design is identified as factorial, but no interaction estimate or formal statistical test is reported in the ClinicalTrials.gov record.
13. Analysis Populations and Denominators
The denominators posted on ClinicalTrials.gov for serious adverse events differ between the pharmacology and stent comparisons. This is expected from the way the registry reports those arm-specific safety figures and should be preserved rather than forcing all arms into a single denominator.
| Comparison domain | Reported denominator structure |
|---|---|
| Pharmacology | Bivalirudin: 1800 at risk; Unfractionated heparin: 1802 at risk. |
| Stent | Paclitaxel-eluting stent: 2257 at risk; Bare metal stent: 749 at risk. |
| Overall enrollment | 3602 participants. |
Differences between overall enrollment and the denominators used for particular outcomes are a reminder that an analysis population is defined by the endpoint and the registry's reporting structure. The ClinicalTrials.gov record does not specify exclusions, missing-data handling, or the exact analysis population used for the primary efficacy endpoints.
14. Missing Data and Censoring
The ClinicalTrials.gov record does not describe a missing-data or imputation strategy. Because the three registered primary endpoints are binary and defined at specific time points, a complete statistical analysis would need to specify how participants without an evaluable endpoint assessment were handled.
No formal imputation method is identified in the ClinicalTrials.gov record. This page therefore does not assume that missing outcomes were imputed, carried forward, treated as events, or excluded from analysis.
15. Interim Analysis and Multiplicity
The ClinicalTrials.gov record does not report an interim-analysis plan, alpha-spending method, multiplicity adjustment, or hierarchical testing procedure. These features are therefore not characterized as part of the HORIZONS-AMI statistical design on this page.
| Design topic | Information reported |
|---|---|
| Interim analysis | Not provided in the ClinicalTrials.gov record. |
| Alpha spending | Not provided in the ClinicalTrials.gov record. |
| Multiplicity adjustment | Not provided in the ClinicalTrials.gov record. |
| Endpoint hierarchy | Three registered primary endpoints are identified, but no formal testing hierarchy is reported. |
| Interaction testing | No interaction estimate or test is reported. |
This is an important distinction between describing a trial's registered endpoints and reconstructing its complete confirmatory statistical strategy. The latter requires the prespecified statistical-analysis documentation.
16. Non-Inferiority, Bayesian Methods, and Other Design Features
The ClinicalTrials.gov record does not identify a non-inferiority margin, a Bayesian analysis, or a crossover design. These topics therefore are not treated as features of the HORIZONS-AMI analysis presented here.
Non-inferiority margin
No non-inferiority margin is provided in the trial data.
Bayesian methods
No Bayesian analysis is identified in the ClinicalTrials.gov record.
Crossover
No crossover design feature is identified in the ClinicalTrials.gov record.
Stratification
No stratification variables are provided in the ClinicalTrials.gov record.
17. Timeline and Trial Status
Trial start
The registry profile lists March 2005 as the study start.
Primary completion
The registry profile lists June 2008 as the primary completion date.
Study status
The registry profile identifies the study status as COMPLETED.
18. Serious Adverse Events: Statistical Perspective
The safety figures can be expressed descriptively as observed proportions without turning them into an inferential comparison.
The visualization intentionally displays the reported counts rather than creating a proportional bar from derived percentages. The ClinicalTrials.gov record specifies affected and at-risk counts exactly, and preserving those values avoids implying an unreported formal safety analysis.
19. What a Complete Primary Analysis Would Report
For an educational reconstruction of the HORIZONS-AMI statistical analysis, each primary endpoint would ideally be accompanied by the number analyzed in each randomized group, the number experiencing the event, an appropriate effect estimate, its confidence interval, the hypothesis test and p-value where applicable, and the prespecified analysis population.
| Primary endpoint | Core quantities a complete analysis would report |
|---|---|
| Major Adverse Ischemic Cardiac Events and Major Bleeding Events at 30 Days | Event counts and denominators; treatment-effect measure; confidence interval; formal hypothesis test; analysis population. |
| Ischemic Target Lesion Revascularization at 1 year | Event counts and denominators; treatment-effect measure; confidence interval; formal hypothesis test; analysis population. |
| Death, Reinfarction, Stroke, or Stent Thrombosis at 1 year | Event counts and denominators; treatment-effect measure; confidence interval; formal hypothesis test; analysis population. |
None of these formal statistical-analysis quantities is in the posted results. Accordingly, they are described here as the structure of a complete analysis rather than presented as HORIZONS-AMI results.
20. Clinical Interpretation vs Statistical Interpretation
What the registry establishes
HORIZONS-AMI was a randomized phase 3 factorial trial in myocardial infarction with 3602 participants, two treatment domains, three registered binary primary endpoints, and 7 posted outcome measures.
What the statistical data establish
The ClinicalTrials.gov recordset identifies the endpoint definitions and design structure but does not supply formal treatment-effect estimates, confidence intervals, or p-values.
What can be interpreted
The factorial structure, binary endpoint framework, time-specific definitions, composite endpoints, and reported serious adverse-event denominators can all be examined statistically.
What should not be inferred
A treatment effect, statistical significance, superiority, non-inferiority, or interaction cannot be assigned from the ClinicalTrials.gov record without the missing formal analysis results.
21. Important Limitations and Interpretation Issues
- No formal statistical analyses reported: no formal statistical analyses were posted even though the ClinicalTrials.gov record states that results are posted.
- No effect estimates: the ClinicalTrials.gov record does not contain risk differences, risk ratios, odds ratios, or other primary endpoint effect measures.
- No confidence intervals: uncertainty around the primary treatment effects cannot be described numerically from the ClinicalTrials.gov record.
- No p-values: statistical significance cannot be inferred from the fact that results are posted.
- Factorial interaction: the registry identifies a factorial design, but no interaction estimate or interaction test is reported.
- Analysis populations: the precise primary efficacy analysis population is not identified in the ClinicalTrials.gov record.
- Missing data: no imputation or missing-outcome strategy is provided.
- Multiplicity: no formal multiplicity adjustment or testing hierarchy is posted on ClinicalTrials.gov for the three primary endpoints.
- Safety interpretation: serious adverse-event counts are available by arm, but no inferential safety comparison is reported.
- Composite endpoints: the clinical meaning of the overall composite depends on the contributions of its individual components, which are not separately quantified in the ClinicalTrials.gov record.
22. Why This Trial Matters Statistically
HORIZONS-AMI is a useful statistical teaching case because the registry identifies a factorial randomized design rather than a simple one-factor comparison. That design creates several layers of analysis: the primary treatment factors, the possibility of interaction, multiple binary endpoints, different assessment windows, and safety outcomes reported across the corresponding treatment domains.
| Concept | How it appears in HORIZONS-AMI |
|---|---|
| Randomization | The allocation is identified as RANDOMIZED. |
| Factorial design | The design model is identified as FACTORIAL. |
| Multiple treatment factors | Pharmacology and stent interventions are both identified. |
| Binary endpoints | All 3 registered primary endpoints are identified as binary. |
| Composite endpoint | The pharmacology endpoint and one stent endpoint combine multiple clinical events. |
| Time-specific analysis | The primary endpoints use 30 days and 1 year as their registered time frames. |
| Interaction | The factorial structure creates an interaction question, although no interaction result is reported. |
| Safety analysis | Serious adverse-event counts are reported for the pharmacology and stent arms. |
| Missing statistical output | Results are posted, but the registry-reported statistical-analysis field contains no formal analyses. |
The most important lesson is methodological: a registry record can identify the design and endpoint architecture without providing enough statistical output to reproduce the confirmatory analysis. A rigorous statistical review should preserve that distinction rather than filling gaps with results from memory or unrelated sources.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Statistical Calculators
25. Sources
- ClinicalTrials.gov: HORIZONS-AMI, NCT00433966.
- Linked publication: PubMed record: PMID 28724652.
- Linked publication: PubMed record: PMID 28249084.
- Linked publication: PubMed record: PMID 27287251.
- Linked publication: PubMed record: PMID 27165710.
- Linked publication: PubMed record: PMID 26458411.
The numerical and design statements on this page are restricted to the ClinicalTrials.gov record. The linked publications are listed because they are identified in the ClinicalTrials.gov record; no additional numerical results from those publications are incorporated into this page.
Continue through the Clinical Biostats statistical library
Use the trial's factorial-design and randomization concepts as a starting point for deeper statistical study and practical analysis tools.
26. Record Summary
HORIZONS-AMI is a completed randomized phase 3 factorial trial in myocardial infarction with 3602 enrolled participants. Its statistical structure contains two treatment domains—bivalirudin versus unfractionated heparin and paclitaxel-eluting versus bare metal stents—and three registered binary primary endpoints assessed at either 30 days or 1 year. The registry also reports 7 posted outcome measures and serious adverse-event counts by arm.
The central statistical teaching point is the factorial structure. The trial permits investigation of multiple treatment factors within one randomized framework, while also creating the possibility of an interaction between those factors. The primary endpoints are binary and include composite clinical outcomes, making endpoint definition and component interpretation important.
The ClinicalTrials.gov record does not contain formal statistical analyses for the posted outcomes. Instead, it explains the statistical framework required to interpret the trial while keeping reported information separate from methodological inference.