This page separates registry-reported trial information from statistical interpretation. The ClinicalTrials.gov record reports results for its registered outcomes, but the ClinicalTrials.gov record contains no formal statistical-analysis records with effect estimates, confidence intervals, or p-values.
1. Trial at a Glance
ISCHEMIA was a completed, randomized, parallel-group clinical trial with 5179 participants and two study arms. The registered interventions included cardiac catheterization, coronary artery bypass graft surgery, percutaneous coronary intervention, lifestyle intervention, and medication. The principal registered outcomes include a composite cardiovascular endpoint and several five-year component and cumulative-event measures.
| Feature | ISCHEMIA |
|---|---|
| Trial name | International Study of Comparative Health Effectiveness With Medical and Invasive Approaches (ISCHEMIA) |
| NCT identifier | NCT01471522 |
| Therapeutic area | Cardiology |
| Status | COMPLETED |
| Phase | NA |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Enrollment | 5179 |
| Lead sponsor | NYU Langone Health |
| Primary endpoints registered | 12 |
| Outcome measures posted | 12 |
| Results posted | Yes |
2. Clinical Question
The registry describes ISCHEMIA as a randomized treatment trial involving cardiovascular and coronary disease, coronary artery disease, heart diseases, and myocardial ischemia. The comparison is between two study strategies identified in the safety data as the Invasive Strategy (INV) and the Conservative Strategy.
Population
Participants enrolled in ISCHEMIA with the registered conditions of cardiovascular diseases, coronary disease, coronary artery disease, heart diseases, and myocardial ischemia.
Interventions
Registered interventions included cardiac catheterization, coronary artery bypass graft surgery, percutaneous coronary intervention, lifestyle intervention, and medication.
Comparator
The registry identifies an Invasive Strategy and a Conservative Strategy as the two study groups.
Primary question
How do the randomized strategies compare with respect to the registered composite cardiovascular outcome and its five-year component and cumulative-event measures?
3. Trial Design
Invasive Strategy (INV)
- Serious adverse events: 318 affected participants
- Participants at risk: 2588
- Registered interventions include cardiac catheterization, coronary artery bypass graft surgery, and percutaneous coronary intervention.
Conservative Strategy
- Serious adverse events: 352 affected participants
- Participants at risk: 2591
- Registered interventions include lifestyle and medication.
The ClinicalTrials.gov record identifies ISCHEMIA as randomized, parallel, and unmasked. They do not provide a treatment allocation ratio, detailed randomization procedure, stratification variables, crossover rules, interim-analysis plan, non-inferiority margin, missing-data strategy, or Bayesian analysis specification. Those design features therefore are not assumed here.
4. Endpoints
The registry contains 12 registered primary endpoints. The endpoint definitions and time frames below preserve the registry wording in the ClinicalTrials.gov record.
| # | Registered primary endpoint | Time frame | Type |
|---|---|---|---|
| 1 | Primary Composite Outcome: Death From Cardiovascular Causes, Myocardial Infarction, or Hospitalization for Unstable Angina, Heart Failure, or Resuscitated Cardiac Arrest | 3.2 year follow-up (median) | Other / unclear |
| 2 | Cumulative Event Rate of Primary Composite Outcome (Death From Cardiovascular Causes, Myocardial Infarction, or Hospitalization for Unstable Angina, Heart Failure, or Resuscitated Cardiac Arrest) | 5 years | Binary |
| 3 | Number of Participants That Experienced Death From Cardiovascular Causes or Myocardial Infarction | 5 years | Binary |
| 4 | Cumulative Event Rate of Death From Cardiovascular Causes or Myocardial Infarction | 5 years | Binary |
| 5 | Number of Participants That Experienced Death From Any Cause | 5 years | Binary |
| 6 | Cumulative Event Rate of Death From Any Cause | 5 years | Other / unclear |
| 7 | Number of Participants That Experienced Myocardial Infarction | 5 years | Binary |
| 8 | Cumulative Event Rate of Myocardial Infarction | 5 years | Other / unclear |
| 9 | Estimated Difference in Cumulative Event Rate (%) of Primary Composite Outcome: Invasive Minus Conservative | 5 years | Binary |
| 10 | Estimated Difference in Cumulative Event Rate of Death From Cardiovascular Causes: Invasive Minus Conservative or Myocardial Infarction Between Invasive and Conservative Strategies | 5 years | Other / unclear |
| 11 | Estimated Difference in Cumulative Event Rate of Death From Any Cause: Invasive Minus Conservative | 5 years | Other / unclear |
| 12 | Estimated Difference in Cumulative Event Rate of Myocardial Infarction: Invasive Minus Conservative | 5 years | Other / unclear |
5. Results Status and Planned Analysis
ClinicalTrials.gov reports results for all 12 registered primary endpoints, but no formal statistical analyses were posted. Consequently, this page does not report treatment-effect estimates, confidence intervals, or p-values. The appropriate statistical interpretation is therefore to distinguish what the registry says was measured from what a formal comparative analysis would ordinarily estimate.
Primary composite outcome
The primary composite outcome is defined as death from cardiovascular causes, myocardial infarction, or hospitalization for unstable angina, heart failure, or resuscitated cardiac arrest, with a 3.2 year follow-up (median) time frame.
For a composite outcome observed over follow-up, a typical analysis would account for the time until the first qualifying component occurs and for censoring of participants who have not experienced an event by the end of observation. The ClinicalTrials.gov record does not state which formal comparative model was used for this endpoint.
Five-year cumulative event rate
The registry specifically describes the five-year cumulative event rate as an estimated cumulative probability within each treatment group. A formal comparison would ordinarily present the two group-specific cumulative probabilities together with an appropriate measure of between-group difference and its uncertainty.
A cumulative event rate answers an absolute probability question: what proportion of participants are estimated to have experienced the endpoint by a specified time? It is different from a hazard ratio and does not describe an instantaneous event rate.
The registry's definition also makes clear why this is not necessarily equivalent to simply dividing the number of observed events by the number enrolled. A time-to-event cumulative estimate can account for different follow-up times and censoring. Without the numerical registry result, however, no numerical event probability or treatment difference can be stated here.
Estimated difference: Invasive Minus Conservative
Several registered five-year endpoints explicitly define the contrast as Invasive Minus Conservative. This direction matters: a negative difference would indicate a lower estimated cumulative event rate in the invasive strategy, whereas a positive difference would indicate a higher estimated cumulative event rate. The ClinicalTrials.gov record does not provide the numerical differences.
An absolute difference is generally easier to translate into an event-probability comparison than a relative hazard measure. For example, if the invasive-minus-conservative difference were negative, the estimated five-year cumulative event probability would be lower in the invasive group. The magnitude, confidence interval, and statistical significance would still be required before drawing a formal inference about the comparison.
6. Statistical Methodology
Randomization
Randomization is the central design feature posted on ClinicalTrials.gov for ISCHEMIA. With 5179 enrolled participants allocated across two parallel study arms, randomization provides the basis for comparing outcomes between strategies while reducing systematic differences in treatment assignment.
Randomization does not guarantee identical groups on every characteristic. Its statistical role is to make treatment assignment independent of measured and unmeasured baseline factors in expectation, so that systematic outcome differences can be attributed to treatment assignment under the assumptions of the trial design.
Time-to-event analysis
Several ISCHEMIA endpoints are naturally expressed as events occurring during follow-up. The primary composite has a median follow-up time frame of 3.2 years, while several registered outcomes are defined at five years.
For a time-to-event endpoint, the Kaplan-Meier estimator is a standard way to estimate the probability of remaining event-free over time. It handles right-censored observations by using each participant's available follow-up information up to the time of censoring.
Here, di is the number of events at event time ti and ni is the number of participants at risk immediately before that time.
Cumulative event probability
The registry's definition of the five-year cumulative event rate is closely related to the Kaplan-Meier framework. Conceptually, the cumulative event probability by time t is one minus the probability of remaining free of the event through time t:
This relationship explains why a cumulative event rate is a time-specific absolute measure rather than a hazard ratio. It expresses the estimated probability of having experienced the endpoint by a specified time.
Between-group comparison
For randomized time-to-event data, common comparative approaches include a log-rank test and a Cox proportional-hazards model. A log-rank test evaluates evidence of a difference between survival curves, while a Cox model can summarize the relative event rate through a hazard ratio.
The ClinicalTrials.gov record does not identify a particular log-rank test, Cox model, hazard-ratio specification, covariate adjustment, stratification scheme, or proportional-hazards assessment for ISCHEMIA. Those methods therefore should be treated here as general statistical approaches rather than documented ISCHEMIA analysis specifications.
Binary endpoint analysis
The registry classifies several five-year measures as binary endpoints, including the number of participants experiencing death from cardiovascular causes or myocardial infarction, death from any cause, and myocardial infarction. A simple binary comparison could use an absolute risk difference, risk ratio, odds ratio, or a suitable regression model. For a five-year endpoint in which follow-up and censoring matter, however, a time-to-event approach may contain information that a simple binary analysis at five years discards.
7. Safety Results
The ClinicalTrials.gov record provides serious adverse events by randomized strategy. These are the only numerical safety results available in the ClinicalTrials.gov record.
| Safety measure | Invasive Strategy (INV) | Conservative Strategy |
|---|---|---|
| Participants affected by serious adverse events | 318 | 352 |
| Participants at risk | 2588 | 2591 |
The serious-adverse-event counts should be interpreted with their corresponding numbers at risk. The ClinicalTrials.gov record does not provide a formal statistical comparison, follow-up time for the safety measure, event rates, confidence intervals, or p-values. It is therefore not appropriate to convert the counts into an inferential treatment comparison here.
The registry reports 318 affected participants among 2588 at risk in the invasive strategy and 352 among 2591 at risk in the conservative strategy. These are descriptive arm-level counts. They do not, by themselves, establish that one strategy has a higher or lower serious-adverse-event risk because the ClinicalTrials.gov record does not specify the formal safety analysis, timing of events, censoring rules, or statistical uncertainty.
8. Statistical Methods Explained
Why does randomization matter statistically?
Randomization determines the treatment assignment mechanism before outcomes are observed. In a sufficiently large randomized trial, this tends to balance prognostic factors between groups in expectation. The resulting comparison is therefore fundamentally different from an observational comparison in which treatment choice may be related to prognosis.
Why is a cumulative event rate different from a raw event percentage?
A cumulative event rate is an estimated probability of experiencing the endpoint by a specified time. When follow-up is incomplete or varies between participants, a time-to-event estimator can use information up to censoring rather than treating every participant as if five years of complete observation were available.
What does "Invasive Minus Conservative" mean?
The registry explicitly defines several five-year contrasts in this direction. Mathematically, the comparison is the estimated cumulative event rate in the invasive group minus the estimated cumulative event rate in the conservative group. The sign therefore carries information about direction: negative values favor a lower event probability in the invasive group for that endpoint, while positive values indicate a higher estimated event probability in the invasive group.
Why would a hazard ratio be useful for these endpoints?
A hazard ratio summarizes a relative difference in the instantaneous event rate under a proportional-hazards model. It can use the timing of events rather than reducing every participant to a yes/no status at a single time point. But a hazard ratio is not an absolute risk difference, and its interpretation depends on the model and its assumptions.
Why should a five-year estimate be shown alongside an effect measure?
A time-specific cumulative event estimate answers an absolute question, while a relative effect measure describes the contrast between groups. Presenting both can make the statistical result more interpretable. A relative measure alone does not tell the reader the underlying absolute probability of an event at a clinically meaningful time point.
Why are confidence intervals important?
A confidence interval quantifies statistical uncertainty around an estimated treatment effect under the specified statistical framework. A narrow interval generally indicates greater precision than a wide interval. It does not describe the range of effects that individual participants experience, nor does it substitute for understanding the clinical magnitude of the estimated difference.
Why can the registry report results without providing a formal statistical-analysis record?
A clinical-trial registry can contain posted outcome-measure data without the ClinicalTrials.gov recordset containing a separate formal statistical-analysis record. In this ISCHEMIA dataset, the results-posted field is "Yes" and 12 outcome measures are posted, while no formal statistical analyses were posted. That combination means the existence of posted outcomes should not be treated as evidence for a particular statistical model, p-value, or confidence interval.
9. What the Registered Endpoint Measures Do — and Do Not — Mean
Composite endpoint
The primary composite combines several clinically distinct events: cardiovascular death, myocardial infarction, and hospitalization for unstable angina, heart failure, or resuscitated cardiac arrest. A composite therefore summarizes first occurrence of any qualifying component rather than measuring one outcome alone.
Component endpoints
The registry separately includes death from cardiovascular causes or myocardial infarction, death from any cause, and myocardial infarction. These measures allow the broader composite to be considered alongside selected components.
Cumulative event rate
The five-year cumulative event rate is an estimated probability of experiencing the endpoint within five years. It is an absolute measure and is not itself a hazard ratio.
Estimated difference
The registry defines several differences as invasive minus conservative. The numerical estimate is not reported in the ClinicalTrials.gov record, so its magnitude and uncertainty cannot be reported.
When multiple events are combined into one endpoint, the overall composite result can be driven by components that occur more frequently than others. Therefore, a composite result should be interpreted together with its prespecified components whenever those component results are available. ISCHEMIA's registry structure explicitly includes several such component measures, but the ClinicalTrials.gov record does not provide their numerical results.
10. Design Features That Are and Are Not Documented Here
| Design topic | What the ClinicalTrials.gov record establishes | Statistical implication |
|---|---|---|
| Randomization | Allocation is RANDOMIZED | Supports a randomized between-strategy comparison. |
| Parallel design | Design model is PARALLEL | Participants are assigned to one of two parallel study strategies. |
| Masking | Masking is NONE | The registry describes the trial as unmasked. |
| Non-inferiority margin | Not reported | No non-inferiority interpretation should be imposed on the endpoints. |
| Crossover | Not reported | No crossover adjustment or dilution analysis should be inferred. |
| Factorial design | Not reported; design is PARALLEL | No factorial interaction structure should be assumed. |
| Interim analysis | Not reported | No alpha-spending or sequential-monitoring claim should be made. |
| Missing data / imputation | Not reported | No imputation method should be attributed to the trial. |
| Stratification | Not reported | No stratified model should be attributed to the trial without supporting documentation. |
| Bayesian methods | Not reported | No Bayesian prior, posterior estimate, or Bayesian decision rule should be inferred. |
This distinction is important in statistical reporting. A familiar clinical-trial method may be appropriate for a particular endpoint, but appropriateness is not evidence that the investigators actually used that method. The ClinicalTrials.gov record is the evidentiary boundary for this page.
11. Limitations
- No formal statistical-analysis records reported: the statistical-analysis field is empty even though results are posted.
- No numerical outcome results reported: the data identify the registered endpoints and their time frames but do not provide numerical cumulative event rates, event counts for the five-year efficacy endpoints, estimated differences, confidence intervals, or p-values.
- No model specification: the ClinicalTrials.gov record does not identify a Cox model, log-rank test, logistic model, competing-risk model, or other formal inferential procedure used for the posted outcomes.
- No randomization ratio: the trial has two arms, but the ClinicalTrials.gov record does not state the allocation ratio.
- No stratification information: the ClinicalTrials.gov record does not identify stratification factors or whether stratification was incorporated into the primary analysis.
- No crossover information: the ClinicalTrials.gov record does not specify whether participants crossed between strategies or how such events were handled analytically.
- No multiplicity plan: 12 registered primary endpoints are listed, but the ClinicalTrials.gov record does not describe an alpha allocation or multiplicity-adjustment strategy.
- Composite endpoint interpretation: the primary composite contains several different clinical events, and the relative contribution of each component cannot be assessed from the registry-reported numerical data.
- Safety interpretation: serious-adverse-event counts are available by arm, but formal event rates and inferential comparisons are not reported.
12. Why This Trial Matters Statistically
ISCHEMIA provides a useful teaching example because the registry combines randomized treatment allocation with a broad set of time-based cardiovascular endpoints. The structure illustrates how a clinical question can be translated into a primary composite outcome, component outcomes, cumulative probabilities, and explicitly defined between-group differences.
| Concept | How it appears in ISCHEMIA |
|---|---|
| Randomization | Allocation is RANDOMIZED across two study arms. |
| Parallel-group design | The registered design model is PARALLEL. |
| Composite endpoint | The primary composite combines cardiovascular death, myocardial infarction, and specified hospitalizations or resuscitated cardiac arrest. |
| Time-to-event follow-up | The primary composite uses a 3.2 year follow-up (median) time frame. |
| Fixed-time cumulative event rate | Several registered endpoints estimate cumulative event probabilities at 5 years. |
| Absolute treatment difference | Several five-year measures explicitly define the contrast as Invasive Minus Conservative. |
| Component outcomes | Death from cardiovascular causes or myocardial infarction, death from any cause, and myocardial infarction are separately registered. |
| Safety analysis | Serious adverse events are reported descriptively by strategy. |
| Statistical reporting discipline |
A high-quality trial-results page should not turn an endpoint definition into an observed result. The registry tells us that ISCHEMIA posted results for its registered outcomes; the ClinicalTrials.gov record does not tell us the numerical values of those outcomes. Maintaining that distinction prevents a statistical explanation from accidentally becoming a fabricated result.
For readers learning clinical-trial statistics, this is itself an important lesson: an endpoint definition, an observed estimate, and a formal inferential test are three different pieces of information. They should not be silently substituted for one another.
13. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registry describes a randomized, parallel comparison with a 3.2-year median follow-up for the primary composite and multiple five-year cumulative-event and component endpoints. Formal treatment-effect estimates are not contained in the ClinicalTrials.gov record.
Clinical interpretation
The registered clinical question concerns outcomes spanning cardiovascular death, myocardial infarction, hospitalization for unstable angina or heart failure, resuscitated cardiac arrest, and other five-year component measures. The ClinicalTrials.gov record does not support a numerical conclusion about which strategy produced fewer events.
This separation is especially important when reading registry results. A trial can be statistically sophisticated without every detail appearing in the registry's structured statistical-analysis field. Conversely, a page should not infer a treatment effect simply because the trial has a familiar design or because results are marked as posted.
14. Longitudinal Trial History
ISCHEMIA begins
The ClinicalTrials.gov record gives a start date of 2012-07.
Primary completion date
The registry lists 2019-06-30 as the primary completion date.
Results posted
The trial is listed as COMPLETED, with results posted and 12 outcome measures posted.
15. Understanding the Five-Year Endpoint Framework
Four groups of registered endpoints deserve to be distinguished. The first is the primary composite outcome itself. The second consists of five-year event counts for selected components. The third consists of five-year cumulative event rates. The fourth consists of explicit estimated differences between the invasive and conservative strategies.
| Endpoint family | What it asks | Statistical perspective |
|---|---|---|
| Primary composite | Did a participant experience any qualifying component during follow-up? | Natural time-to-event framework for a composite first event. |
| Five-year event counts | How many participants experienced a specified event by five years? | Descriptive binary summary unless a formal comparative model is specified. |
| Five-year cumulative event rates | What is the estimated probability of experiencing the endpoint by five years? | Time-specific absolute risk measure related to survival analysis. |
| Estimated differences | How does the five-year cumulative event rate differ between strategies? | Absolute treatment contrast; uncertainty is needed for inference. |
An absolute difference directly describes how far apart two event probabilities are at a specified time. This can be more immediately interpretable than a relative measure, but it is also time-specific. A five-year difference is not automatically the same as a difference at another follow-up time.
16. A Note on Hazard Ratios and Proportional Hazards
Hazard ratios are common in randomized time-to-event trials, but the registry-reported ISCHEMIA data do not report a hazard ratio or state that a Cox proportional-hazards model was used. It is therefore useful to understand the method without attributing it to this trial.
For a binary treatment indicator, exp(β) is the model-based hazard ratio. Under the proportional-hazards assumption, that ratio is interpreted as constant over time on the hazard scale.
A hazard ratio below 1 would conventionally indicate a lower modeled instantaneous event rate in the treatment group, while a hazard ratio above 1 would indicate a higher modeled instantaneous event rate. But a hazard ratio is not a five-year cumulative risk difference, and it does not say that the same proportional reduction applies to every individual's risk.
17. Serious Adverse Events: How to Read the Reported Counts
The ClinicalTrials.gov record reports 318 affected participants among 2588 at risk in the Invasive Strategy and 352 affected participants among 2591 at risk in the Conservative Strategy.
| Group | Affected | At risk | What is directly supported |
|---|---|---|---|
| Invasive Strategy (INV) | 318 | 2588 | Descriptive count of affected participants and corresponding risk set. |
| Conservative Strategy | 352 | 2591 | Descriptive count of affected participants and corresponding risk set. |
The ClinicalTrials.gov record does not specify whether these serious adverse events were analyzed as time-to-first event, whether participants could contribute multiple events, the safety follow-up window, or the formal statistical method. Those omissions prevent a more detailed inferential safety analysis without importing outside information.
Counts alone do not provide a complete comparison because the number of participants at risk differs between groups. Even when denominators are available, a rigorous comparison may also need the timing of events and censoring. Therefore, the safest interpretation of the ClinicalTrials.gov record is descriptive rather than inferential.
18. Sources
- ClinicalTrials.gov: NCT01471522 — ISCHEMIA.
- Linked publication: PubMed PMID 42641674.
- Linked publication: PubMed PMID 42554061.
- Linked publication: PubMed PMID 42384892.
- Linked publication: PubMed PMID 42303606.
- Linked publication: PubMed PMID 42132177.
Continue through the Clinical Biostats statistical pathway
Explore the statistical concepts behind randomized comparisons, time-to-event endpoints, cumulative event rates, and treatment-effect estimation.
19. Related Tutorials
Learn more about the methods used to understand this trial:
20. Related Calculators
21. Record Summary
ISCHEMIA is a randomized, parallel-group trial with 5179 participants and two study strategies. Its registry record contains 12 registered primary endpoints, including a composite outcome assessed at a 3.2 year median follow-up and multiple five-year measures covering cumulative event rates, component events, and estimated differences defined as invasive minus conservative. The ClinicalTrials.gov record also report serious adverse events by strategy.
The statistical lesson is as important as the clinical endpoint structure. The registry reports that results are posted, but the registry-reported statistical-analysis field contains no formal analyses and no numerical efficacy estimates. A rigorous interpretation therefore stops where the evidence stops: