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
ISCHEMIA-EXTEND is an extended follow-up study associated with a large clinical trial population in cardiovascular disease. The registry identifies two arms and an enrollment of 5391 participants, with mortality evaluated through Year 10 over a range of 6 to 13 years.
| Feature | ISCHEMIA-EXTEND |
|---|---|
| Registry title | ISCHEMIA-EXTEND (Extended Follow-up) |
| ClinicalTrials.gov identifier | NCT04894877 |
| Status | Active, not recruiting |
| Start date | 2012-07 |
| Primary completion date | 2026-03-31 |
| Enrollment | 5391 |
| Number of arms | 2 |
| Lead sponsor | NYU Langone Health |
| Sponsor type | OTHER |
| Conditions | Cardiovascular Diseases; Coronary Disease; Coronary Artery Disease; Heart Diseases; Myocardial Ischemia |
2. Clinical Question
The registry's primary statistical question is centered on long-term mortality. ISCHEMIA-EXTEND measures cumulative event rates for death from all causes, cardiovascular causes, and non-cardiovascular causes at Year 10, with the registry specifying a time-frame range of 6 to 13 years.
Population
The registry identifies the study conditions as cardiovascular diseases, coronary disease, coronary artery disease, heart diseases, and myocardial ischemia.
Interventions
The registry lists cardiac catheterization, coronary artery bypass graft surgery, percutaneous coronary intervention, lifestyle, and medication.
Study structure
The registry identifies two arms and an enrollment of 5391 participants.
Primary question
What are the cumulative event rates for all-cause death, cardiovascular death, and non-cardiovascular death at Year 10?
3. Trial Design
ISCHEMIA-EXTEND is registered as an extended follow-up study. The registry records a study start date of July 2012, an enrollment of 5391 participants, two arms, and a primary completion date of March 31, 2026. The current registry status is active, not recruiting.
The registry lists cardiac catheterization, coronary artery bypass graft surgery, percutaneous coronary intervention, lifestyle, and medication among the interventions. The registry information available here does not specify the allocation of those interventions to the two arms, so the statistical interpretation should not assign them to particular treatment groups.
4. Endpoints
| Primary endpoint | Registry definition | Time frame |
|---|---|---|
| Cumulative Event Rate of Death: All-Cause | Cumulative event rate of death from all causes. | Year 10 (Range: 6-13 years) |
| Cumulative Event Rate of Cardiovascular (CV) Death | Cumulative event rate of death from cardiovascular causes. | Year 10 (Range 6-13 years) |
| Cumulative Event Rate of Non-CV Death | Cumulative event rate of death from non-cardiovascular causes. | Year 10 (Range 6-13 years) |
What these endpoints measure
The three endpoints separate mortality into clinically meaningful categories. All-cause mortality counts deaths regardless of cause. Cardiovascular mortality restricts the event definition to cardiovascular causes, while non-cardiovascular mortality captures deaths attributed to causes outside the cardiovascular category.
This distinction is statistically important. All-cause death is a single event type in which competing causes do not need to be separated. Cause-specific mortality, by contrast, requires attention to the fact that death from one cause prevents a participant from subsequently experiencing death from another cause.
5. Statistical Methodology
No formal statistical analyses are posted in the registry record. The endpoint definitions nevertheless identify a clear time-to-event framework: participants are followed over an extended period and the cumulative occurrence of specified mortality events is evaluated at Year 10.
All-cause mortality
For all-cause mortality, a standard analysis would treat death from any cause as the event of interest. Kaplan-Meier estimation is commonly used to describe the probability of remaining alive over time, from which the estimated cumulative event probability can be expressed as one minus the survival probability.
Here, S(t) represents the estimated probability of surviving beyond time t. The corresponding quantity, F(t), represents the estimated cumulative probability of death by that time.
Cause-specific mortality
Cardiovascular and non-cardiovascular death introduce competing-risk considerations. Once a participant dies from one cause, that participant can no longer experience death from another cause. A cause-specific analysis therefore needs to distinguish the event of interest from competing causes of death.
Unlike an all-cause mortality analysis, a cause-specific cumulative incidence calculation must account for the fact that competing causes can occur first and permanently remove a participant from risk of the event being studied.
Time-to-event censoring
Extended follow-up commonly includes participants whose outcome status is known only through a particular observation time. In a time-to-event analysis, such participants can contribute information up to the point of censoring. Correct handling of censoring is central to estimating long-term mortality rates without treating incomplete follow-up as if it were an observed survival outcome.
Comparing two arms
Because the registry identifies two arms, an eventual comparative analysis could describe cumulative mortality separately by arm and compare the resulting time-to-event distributions. Depending on the prespecified statistical analysis plan, methods could include Kaplan-Meier curves, log-rank testing, Cox proportional-hazards models, or competing-risk methods for cause-specific endpoints.
The choice among these approaches matters. A hazard ratio describes a relative instantaneous event rate under a specified model, whereas a cumulative event rate at Year 10 answers an absolute time-point question. For cause-specific mortality, cumulative incidence methods can be particularly important because competing causes of death directly affect the probability of observing the event of interest.
6. Planned Analysis
The registry identifies three primary endpoints to be evaluated at Year 10: cumulative event rate of all-cause death, cumulative event rate of cardiovascular death, and cumulative event rate of non-cardiovascular death. The specified time frame for each endpoint is Year 10, with a range of 6 to 13 years.
| Endpoint | Typical statistical approach | Primary interpretation |
|---|---|---|
| All-cause death | Kaplan-Meier estimation or another prespecified survival method | Cumulative probability of death from any cause by the specified follow-up time |
| Cardiovascular death | Competing-risk cumulative incidence or a prespecified cause-specific survival approach | Cumulative probability of cardiovascular death while accounting for other causes of death |
| Non-cardiovascular death | Competing-risk cumulative incidence or a prespecified cause-specific survival approach | Cumulative probability of non-cardiovascular death while accounting for cardiovascular death as a competing event |
The absence of posted numerical analyses is particularly relevant for a long-term extension study. A Year 10 cumulative event rate is not simply a proportion obtained by dividing the number of deaths by enrollment. The timing of events, the timing of censoring, and, for cause-specific mortality, competing causes of death all affect the appropriate statistical estimate.
7. Statistical Methods Explained
What is a cumulative event rate?
A cumulative event rate describes how much of an event has accumulated by a specified point in follow-up. For mortality, the quantity can be interpreted as the estimated probability that a participant has experienced the specified death event by a particular time.
This is different from an incidence rate, which describes the number of events relative to accumulated person-time. The registry's wording explicitly identifies a cumulative event rate, so the natural interpretation is a probability-based time-to-event quantity rather than a simple events-per-person-year rate.
Why is Year 10 important?
The primary time frame is Year 10, with a range of 6 to 13 years. Long-term follow-up changes the statistical problem compared with a short trial endpoint. More participants may experience the event, but others may have different lengths of observable follow-up. The analysis therefore needs to preserve the timing of events and censoring rather than treating every participant as having identical follow-up.
Why does the 6-to-13-year range matter?
A range of 6 to 13 years indicates that the registry's Year 10 endpoint is embedded within a broader range of observed follow-up durations. Participants do not necessarily contribute the same amount of follow-up time. Survival-analysis methods are designed to use the information available from each participant while respecting when observation begins and ends.
Why are cardiovascular and non-cardiovascular death different from all-cause death?
All-cause death has one event definition: death from any cause. Cause-specific endpoints divide mortality into categories. A participant who dies from cardiovascular disease can no longer subsequently die from a non-cardiovascular cause within the study's follow-up. That makes the two cause-specific outcomes competing events from a statistical perspective.
What is a competing risk?
A competing risk is an event that prevents the event of interest from occurring. For cardiovascular death, non-cardiovascular death is a competing event. For non-cardiovascular death, cardiovascular death is a competing event. Ignoring this relationship can produce an inappropriate interpretation of the cumulative probability of a specific cause of death.
Why is a hazard ratio different from a cumulative event rate?
A hazard ratio is a relative comparison of event rates over time under a statistical model. A cumulative event rate is an absolute probability accumulated to a specified time. If both were reported, they would answer complementary questions: the hazard ratio would describe the relative treatment or arm difference in instantaneous event rates, while the cumulative event rate would describe the absolute frequency of events by Year 10.
8. All-Cause, Cardiovascular, and Non-Cardiovascular Mortality
The three primary endpoints form a useful statistical hierarchy because they progressively narrow the definition of death. All-cause mortality is the broadest endpoint. Cardiovascular and non-cardiovascular mortality partition deaths into cause-specific categories.
All-cause death
Counts death regardless of cause. Because every death qualifies, there is no competing cause-of-death category within the endpoint itself.
Cardiovascular death
Counts death attributed to cardiovascular causes. Non-cardiovascular death prevents a participant from later experiencing cardiovascular death.
Non-cardiovascular death
Counts death attributed to non-cardiovascular causes. Cardiovascular death is a competing event for this endpoint.
Why the distinction matters
Cause-specific mortality cannot be interpreted correctly by simply treating all other causes as ordinary censored observations without considering the competing-risk structure.
9. Interpreting Long-Term Follow-up
Extended follow-up changes the information available for mortality analysis. The registry specifies a Year 10 endpoint with a follow-up range of 6 to 13 years. That means the statistical analysis needs to account for unequal observation periods and the possibility that participants enter the endpoint analysis with different amounts of available follow-up.
A long-term cumulative event estimate is therefore fundamentally a time-to-event quantity. It should not be interpreted as though every participant were observed continuously through the same calendar date. Participants who remain alive and under observation contribute survival information, while participants whose follow-up ends before an event contribute information through their last known observation under the assumptions of the chosen censoring method.
The same principle applies to cause-specific mortality, but with an additional layer: a participant who experiences a competing cause of death is no longer available to experience the primary cause of interest.
10. Why the Endpoint Definitions Matter
| Statistical feature | Why it matters |
|---|---|
| Year 10 | Defines the principal time point at which cumulative mortality is assessed. |
| 6-13 year range | Signals extended and potentially unequal follow-up duration. |
| All-cause mortality | Provides a broad mortality endpoint unaffected by cause classification. |
| Cardiovascular mortality | Requires attention to non-cardiovascular death as a competing event. |
| Non-cardiovascular mortality | Requires attention to cardiovascular death as a competing event. |
| Cumulative event rate | Expresses event occurrence over time rather than simply counting events. |
The endpoint wording also helps distinguish the scientific question from the statistical model. The registry defines what is to be measured; the statistical method determines how that measurement is estimated from longitudinal observations.
11. Statistical Interpretation of Cumulative Event Rates
A cumulative event rate at Year 10 is an absolute measure. If an estimated cumulative event rate were available, it would describe the estimated probability of experiencing the specified endpoint by that time under the analysis framework.
A cumulative event rate is not a hazard ratio and does not describe an instantaneous relative rate. It also does not imply that every participant has the same underlying risk or that the event probability is constant over time.
With follow-up extending across 6 to 13 years, the amount of information available at any particular time can differ among participants. Time-to-event methods use the observed timing of events and censoring rather than requiring identical follow-up for everyone.
For cardiovascular and non-cardiovascular mortality, the estimated probability of one cause is affected by the occurrence of the other cause. A cause-specific cumulative probability therefore cannot be interpreted in isolation from the competing-risk structure.
12. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary endpoints are long-term cumulative mortality measures evaluated at Year 10. Their appropriate analysis requires methods that preserve the time dimension of follow-up and, for cause-specific mortality, account for competing causes of death.
Clinical interpretation
The endpoints address whether long-term mortality differs between the study's two arms and whether mortality is attributable primarily to cardiovascular or non-cardiovascular causes. Numerical treatment effects are not reported in the registry record.
The distinction is important because a statistical endpoint is not merely a label attached to a number. Its definition determines which observations count as events, which observations remain at risk, and how competing outcomes influence the final estimate.
13. Limitations
- No posted statistical analyses: the registry does not report numerical estimates, confidence intervals, or p-values for the three primary endpoints.
- Two-arm structure without arm-specific intervention assignments: the registry identifies two arms but the available record does not specify which listed interventions correspond to each arm.
- Extended follow-up: the endpoint time frame spans a range of 6 to 13 years, making unequal observation time an important consideration.
- Competing risks: cardiovascular and non-cardiovascular deaths prevent one another from subsequently occurring and therefore require an appropriate competing-risk framework.
- Cause-of-death classification: cause-specific mortality depends on how deaths are classified into cardiovascular and non-cardiovascular categories.
- Model dependence: different statistical approaches answer somewhat different questions. A cause-specific hazard model, a subdistribution approach, and a cumulative incidence estimate should not be treated as interchangeable quantities.
- Long-term censoring: participants with incomplete observation contribute information only through the time at which their follow-up ends under the assumptions of the analysis.
- Registry status: the study is listed as active, not recruiting, with a primary completion date of March 31, 2026.
14. Why This Trial Matters Statistically
ISCHEMIA-EXTEND provides a useful teaching example of how clinical-trial statistics change when the primary question moves from short-term treatment effects to long-term mortality. The key issue is not simply how many deaths occurred, but how those deaths accumulated over time and how the cause of death changes the statistical structure.
| Concept | How it appears in ISCHEMIA-EXTEND |
|---|---|
| Extended follow-up | Primary mortality endpoints are evaluated at Year 10 with a 6-13 year time-frame range. |
| Time-to-event analysis | Mortality is accumulated over longitudinal follow-up rather than assessed as a simple cross-sectional proportion. |
| Cumulative incidence | The registry defines the primary outcomes as cumulative event rates. |
| All-cause mortality | Death from all causes is evaluated as a primary endpoint. |
| Competing risks | Cardiovascular and non-cardiovascular deaths form competing event categories. |
| Absolute effects | Cumulative event rates describe absolute event probability by a specified time. |
| Arm comparison | The study has two registered arms, allowing the long-term mortality endpoints to be framed as between-arm outcomes. |
| Follow-up heterogeneity | The registry's 6-13 year range emphasizes that participants may contribute different amounts of observable follow-up. |
15. Sponsor and Study Context
The lead sponsor listed in the registry is NYU Langone Health, with sponsor type recorded as OTHER. The study concerns cardiovascular diseases, coronary disease, coronary artery disease, heart diseases, and myocardial ischemia.
The intervention list includes cardiac catheterization, coronary artery bypass graft surgery, percutaneous coronary intervention, lifestyle, and medication. These descriptions establish the intervention vocabulary in the registry, but the available record does not provide enough arm-specific detail to construct an intervention-versus-control table without adding information beyond the registry record.
16. Timeline
Study start
The registry records July 2012 as the study start date.
Long-term mortality assessment
The three primary endpoints are defined at Year 10, with a specified time-frame range of 6 to 13 years.
Primary completion
The registry lists March 31, 2026 as the primary completion date.
Active, not recruiting
The study is currently listed as active, not recruiting.
17. A Statistical Framework for the Three Primary Endpoints
CV death → cause-specific mortality + non-CV competing event
Non-CV death → cause-specific mortality + CV competing event
The three endpoints should be analyzed according to their distinct event structures rather than assuming that one statistical method automatically answers all three questions.
For all-cause mortality, the event definition collapses all causes of death into a single endpoint. A standard survival curve can therefore represent the probability of remaining alive, and one minus that survival probability gives the cumulative probability of death.
For cardiovascular mortality, non-cardiovascular death is not ordinary loss to follow-up. It is an event that changes the future probability of cardiovascular death to zero for that participant. The same relationship operates in the opposite direction for non-cardiovascular mortality.
This is why long-term cause-specific mortality is a particularly useful setting for understanding competing-risk analysis. The statistical question is not merely whether a participant eventually died, but which event occurred first and how the occurrence of competing events changes the population's remaining risk.
18. What Would a Complete Statistical Report Contain?
A complete numerical analysis of these endpoints would normally distinguish the cumulative event estimate at the prespecified time point from measures of relative treatment effect and from measures of uncertainty.
| Component | Statistical purpose |
|---|---|
| Cumulative event estimate | Quantifies the absolute probability of the specified event by the target time. |
| Confidence interval | Describes statistical uncertainty around the estimated cumulative event probability. |
| Between-arm comparison | Quantifies whether and how the mortality experience differs between the two study arms. |
| Cause-specific analysis | Separates cardiovascular from non-cardiovascular mortality while respecting competing events. |
| Time-to-event curve | Shows how mortality accumulates over follow-up rather than reducing the analysis to one time point. |
| Sensitivity analysis | Can examine whether conclusions depend materially on modeling assumptions, censoring, or cause-of-death handling. |
These components answer different questions. A confidence interval is not a measure of clinical importance, and a p-value, if reported, would not quantify the magnitude of an effect. The cumulative event estimate itself is the direct measure of absolute event probability at the specified time.
19. Statistical Concepts in This Trial
The most important statistical concepts illustrated by ISCHEMIA-EXTEND are long-term time-to-event analysis, cumulative incidence, competing risks, censoring, and interpretation of absolute mortality probabilities.
Time-to-event data
Participants can contribute information for different lengths of time, so event timing and censoring are integral to the analysis.
Cumulative incidence
The primary endpoints are explicitly defined as cumulative event rates at Year 10.
Competing risks
Cardiovascular and non-cardiovascular death are mutually competing event types.
Absolute risk
A cumulative event rate describes the probability of experiencing the specified endpoint by a defined time.
20. Record Summary
ISCHEMIA-EXTEND is registered as an extended follow-up study with 5391 enrolled participants and two arms. The primary endpoints are cumulative event rates of all-cause death, cardiovascular death, and non-cardiovascular death at Year 10, with each endpoint assigned a time-frame range of 6 to 13 years.
Statistically, the most important feature of the endpoint structure is the distinction between all-cause mortality and cause-specific mortality. All-cause death can be analyzed as a single time-to-event outcome, whereas cardiovascular and non-cardiovascular death require explicit consideration of competing events. The long follow-up range also makes censoring and unequal observation time central to the interpretation.
No formal statistical analyses are posted to the ClinicalTrials.gov record for these endpoints. Consequently, the registry establishes the questions to be measured but does not provide numerical mortality estimates, confidence intervals, or p-values for interpreting the two-arm comparison.
21. Sources
- ClinicalTrials.gov: NCT04894877 — ISCHEMIA-EXTEND.
- PubMed: PubMed record.
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical trial statistical analyses and the methods used to interpret time-to-event, efficacy, safety, and long-term follow-up data.