This page separates reported registry results from statistical interpretation. Trial-specific numerical information is restricted to the ClinicalTrials.gov data posted on ClinicalTrials.gov for NCT01205776. The registry provides two posted statistical analyses for secondary time-to-event analyses of outcomes that overlap with the registered primary endpoint measure.
1. Trial at a Glance
EXCEL was a completed randomized, parallel-design clinical trial comparing percutaneous coronary intervention (PCI) with coronary artery bypass grafting (CABG). The trial enrolled 1905 participants and registered a binary primary endpoint concerning all-cause death, protocol defined myocardial infarction, or protocol defined stroke at 5 years. The registry also posts two statistical analyses of related time-to-event outcomes at 30 days and from 0 to 3 years.
| Feature | EXCEL |
|---|---|
| Trial name | EXCEL |
| Brief title | EXCEL Clinical Trial |
| Therapeutic area | Cardiology |
| Status | COMPLETED |
| Phase | NA |
| Design | Randomized, parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 1905 |
| Interventions | Percutaneous Coronary Intervention (device); CABG (procedure) |
| Lead sponsor | Abbott Medical Devices |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT01205776 |
2. Clinical Question
The registry describes a randomized comparison between percutaneous coronary intervention and coronary artery bypass grafting in a population associated with unprotected left main coronary artery disease and related coronary disease conditions. The registered primary endpoint asks whether participants experience all-cause death, protocol defined myocardial infarction, or protocol defined stroke during a 5-year time frame.
Population
The registered conditions include unprotected left main coronary artery disease, chronic coronary occlusion, coronary artery disease, coronary artery stenosis, coronary disease, myocardial ischemia, vascular disease, coronary restenosis, and stent thrombosis.
Intervention
Percutaneous Coronary Intervention (PCI), identified in the registry as a device intervention.
Comparator
Coronary Artery Bypass Graft (CABG), identified in the registry as a procedure.
Primary question
For the registered 5-year endpoint, how do the randomized PCI and CABG groups compare with respect to all-cause death, protocol defined MI, or protocol defined stroke?
3. Trial Design
Percutaneous Coronary Intervention
- Intervention type: device
- Compared directly with CABG
- Serious adverse events: 535 affected / 948 at risk
Coronary Artery Bypass Graft
- Intervention type: procedure
- Compared directly with PCI
- Serious adverse events: 550 affected / 957 at risk
4. Trial Timeline
Trial start
The registry lists September 29, 2010 as the study start date.
Registered primary endpoint time frame
The primary endpoint is defined over 5 years and counts all-cause death, protocol defined MI, or protocol defined stroke.
Primary completion
The registry lists May 3, 2019 as the primary completion date.
5. Primary Endpoint
| Endpoint | Registered definition | Time frame | Type |
|---|---|---|---|
| Number of Participants With All-cause Death, Protocol Defined MI or Protocol Defined Stroke | All deaths includes Cardiac death, Vascular death and Non-cardiovascular death. MI includes Q wave MI, defined by development of a new pathological Q wave on the ECG, and non-Q wave MI, defined by elevation of CK levels to ≥ two times the upper limit of normal (ULN) with elevated CK-MB in the absence of new pathological Q waves. Stroke is defined as the rapid onset of a new persistent neurologic deficit. | 5 years | Binary |
The primary endpoint is registered as binary, but the statistical analyses posted for the same outcome measure are expressed as time-to-event analyses. That distinction matters. A binary endpoint asks whether the event occurred within a specified period; a time-to-event analysis additionally uses information about when the event occurred and can accommodate right-censored observations.
The registry therefore combines three clinically distinct event types into one endpoint. For a participant who experiences more than one component, interpretation of the composite depends on the prespecified event definitions and timing rules; the ClinicalTrials.gov record does not provide component-specific numerical results.
6. Statistical Methodology
Com-Nougue approach
The registry's statistical-analysis field reports the Com-Nougue approach. In the ClinicalTrials.gov record, this is described as a non-inferiority approach for two Kaplan-Meier failure rates. The analysis text refers to a p-value and a 95% confidence interval of the difference.
This is important because the reported framework is not simply a conventional superiority comparison. In a non-inferiority analysis, the question is whether the evidence is compatible with the new intervention being no worse than the comparator by more than a prespecified clinically acceptable margin.
Kaplan-Meier failure rates
The posted analysis describes comparison of two Kaplan-Meier failure rates. Kaplan-Meier methods estimate the event-free or survival experience over time while retaining information from participants who are censored before the end of follow-up.
Here di represents events at an event time and ni represents participants at risk immediately before that time. The registry's Com-Nougue analysis uses the resulting Kaplan-Meier failure-rate information in its non-inferiority comparison.
Hazard ratio reporting
The registry normalizes the posted effect measure to hazard ratio and identifies the effect measure in both statistical analyses as “Hazard Ratio (HR).” However, the numerical estimate posted on ClinicalTrials.gov for the 30-day analysis is -3.1, and the numerical estimate posted on ClinicalTrials.gov for the 0-to-3-year analysis is 4. Because the ClinicalTrials.gov record does not provide enough information to reconcile these estimates with the HR label, this page reports the registry values exactly as reported in the registry rather than transforming or reinterpreting them.
Intention-to-treat analysis
The second posted statistical analysis explicitly identifies the ITT population. Under the intention-to-treat principle, participants remain associated with the randomized treatment group for the primary treatment comparison, regardless of subsequent treatment changes. This preserves the principal protection provided by randomization.
Non-inferiority hypothesis
Both posted statistical analyses are labeled as non-inferiority analyses. The registry provides a non-inferiority margin of 2% for the 30-day analysis and 8.4% for the 0-to-3-year analysis, each described as being evaluated at the 0.05 significance level.
7. Results
ClinicalTrials.gov posts two statistical analyses. Both concern secondary analyses of time-to-event outcomes, rather than a statistical-analysis record explicitly labeled as the 5-year primary endpoint. The distinction is important: the registered primary endpoint has a 5-year time frame, while the posted analyses reported here use 30 days and 0 to 3 years.
30-Day Analysis
All-cause death, protocol defined MI or protocol defined stroke
Effect measure as posted: Hazard Ratio (HR) · two-sided · P < 0.0001
Non-inferiority analysis using the Com-Nougue approach; margin: 2% at the 0.05 level of significance.
| Analysis feature | 30-day result |
|---|---|
| Outcome | Number of Participants With All-cause Death, Protocol Defined MI or Protocol Defined Stroke |
| Time frame | 30 days |
| Population | Not specified in the ClinicalTrials.gov record |
| Groups compared | PCI vs CABG |
| Method | Com-Nougue |
| Effect measure as reported | Hazard Ratio (HR) |
| Estimate | -3.1 |
| Confidence interval | The registry analysis comment references a 95% CI of the Difference, but numerical bounds are not provided in the ClinicalTrials.gov record. |
| P-value | <0.0001 |
| Hypothesis | Non-inferiority |
| Non-inferiority margin | 2% |
The registry reports an estimate of -3.1 for this analysis and labels the effect measure as a hazard ratio. Because a conventional hazard ratio is ordinarily a positive ratio, the registry-reported value and the effect-measure label should not be silently reconciled. The appropriate educational reading is therefore that -3.1 is the value posted in the registry analysis record, not that it should be converted into a different effect measure.
The P < 0.0001 value indicates strong statistical evidence under the specified non-inferiority testing framework. A p-value does not itself measure the size of an effect, the clinical importance of an effect, or the probability that one treatment is beneficial.
The registry states that the non-inferiority calculation uses a 2% margin and a 95% confidence interval of the difference. However, the numerical CI bounds are not contained in the ClinicalTrials.gov record, so its precision cannot be quantified numerically here.
Because this is a non-inferiority analysis, interpretation depends on the prespecified margin and the direction of the comparison. It is not equivalent to demonstrating superiority. The time-to-event framework also depends on the handling of censoring and the assumptions underlying the analysis.
0-to-3-Year Analysis
Death, protocol defined MI, protocol defined stroke or unplanned revascularization for ischemia
Effect measure as posted: Hazard Ratio (HR) · two-sided · P = 0.011
ITT analysis using the Com-Nougue approach; non-inferiority margin: 8.4% at the 0.05 level of significance.
| Analysis feature | 0-to-3-year result |
|---|---|
| Outcome | Number of Participants With Death, Protocol Defined MI, Protocol Defined Stroke or Unplanned Revascularization for Ischemia |
| Time frame | 0 to 3 years |
| Analysis population | ITT population |
| Groups compared | PCI vs CABG |
| Method | Com-Nougue Approach |
| Effect measure as reported | Hazard Ratio (HR) |
| Estimate | 4 |
| Confidence interval | The registry analysis comment references a 95% CI of the Difference, but numerical bounds are not provided in the ClinicalTrials.gov record. |
| P-value | 0.011 |
| Hypothesis | Non-inferiority |
| Non-inferiority margin | 8.4% |
The registry reports an estimate of 4 for the 0-to-3-year analysis and labels the effect measure as a hazard ratio. The ClinicalTrials.gov record does not provide enough information to determine whether this numerical estimate is intended to be interpreted as a conventional hazard ratio, a difference expressed elsewhere in the Com-Nougue calculation, or another registry representation. It should therefore be reported exactly as posted rather than converted.
The P = 0.011 value is evidence against the relevant null hypothesis under the posted statistical framework, but it does not quantify the magnitude of the treatment effect. In particular, it should not be read as an “11% effect” or as a probability that the treatment works.
The registry specifies an 8.4% non-inferiority margin. Non-inferiority is assessed relative to this prespecified margin, using the confidence-interval framework described in the registry, rather than by looking at the p-value alone.
The analysis is explicitly conducted in the ITT population. That preserves randomization as the basis of the comparison, but the ClinicalTrials.gov record does not provide details about censoring rules, missing data, or the handling of treatment changes.
8. Secondary Endpoint Results in Context
The ClinicalTrials.gov record is both designated Secondary. The first evaluates the registered composite outcome at 30 days. The second expands the event definition to include unplanned revascularization for ischemia and evaluates it from 0 to 3 years.
| Feature | 30 days | 0 to 3 years |
|---|---|---|
| Endpoint role | Secondary | Secondary |
| Composite outcome | Death, protocol defined MI or protocol defined stroke | Death, protocol defined MI, protocol defined stroke or unplanned revascularization for ischemia |
| Analysis population | Not specified in the ClinicalTrials.gov record | ITT population |
| Method | Com-Nougue | Com-Nougue Approach |
| Effect measure as reported | Hazard Ratio (HR) | Hazard Ratio (HR) |
| Estimate | -3.1 | 4 |
| P-value | <0.0001 | 0.011 |
| Non-inferiority margin | 2% | 8.4% |
These are not interchangeable estimands. The addition of unplanned revascularization for ischemia changes the composite endpoint itself, while the follow-up window also changes from 30 days to 0 to 3 years. Consequently, a statistical comparison across these two records would require a clearly defined estimand and cannot be inferred from the posted summary values alone.
9. Safety Results
The ClinicalTrials.gov record provides serious adverse-event counts by randomized intervention group. These are presented as affected participants over participants at risk.
| Safety measure | PCI | CABG |
|---|---|---|
| Serious adverse events | 535/948 | 550/957 |
| Numerator | 535 affected | 550 affected |
| Denominator | 948 at risk | 957 at risk |
The appropriate statistical interpretation of a safety count is different from interpretation of a time-to-event efficacy estimate. The ClinicalTrials.gov record identifies the number affected and number at risk, but does not provide a formal between-group statistical analysis, confidence interval, p-value, severity breakdown, or specific serious-adverse-event categories.
The values 535/948 and 550/957 are counts of participants with serious adverse events divided by participants at risk. They should not be confused with hazard ratios or with the estimates from the non-inferiority analyses.
A comparison of the two fractions could be performed using an appropriate binary-outcome method if the analysis objective were a simple risk comparison. A time-to-event safety analysis would require additional timing information that is not contained in the ClinicalTrials.gov record.
10. Non-Inferiority Trial Design
Non-inferiority is central to the statistical profile of EXCEL. The registry labels both posted analyses as non-inferiority analyses and specifies different margins for the two analysis periods.
30-day margin
The Com-Nougue analysis uses a non-inferiority margin of 2% at the 0.05 level of significance.
0-to-3-year margin
The corresponding analysis uses a non-inferiority margin of 8.4% at the 0.05 level of significance.
Why the margin matters
In a non-inferiority trial, the margin defines how much worse the experimental intervention could be while still satisfying the trial's prespecified non-inferiority criterion. The statistical question is therefore not simply whether the observed p-value crosses 0.05.
The confidence interval is central because it describes the range of treatment-effect values compatible with the data under the specified statistical framework. A numerical decision boundary cannot be reconstructed here beyond the margins explicitly reported by the registry.
The use of different margins for different analysis periods is especially important. The 2% and 8.4% margins should not be treated as interchangeable thresholds. They belong to separate posted analyses with different time frames and endpoint structures.
11. Statistical Methods Explained
Why use a time-to-event analysis for a composite endpoint?
The registered primary endpoint is defined as a participant experiencing death, protocol defined MI, or protocol defined stroke over a specified period. A time-to-event approach can use both the occurrence and timing of the first relevant event while retaining information from participants who have not experienced an event by the end of their available follow-up.
What does the Com-Nougue approach contribute?
The registry identifies the Com-Nougue approach as the statistical method for both posted analyses. Its accompanying text describes a non-inferiority comparison of two Kaplan-Meier failure rates. Thus, the method connects the time-to-event estimates with a prespecified non-inferiority framework rather than treating the endpoint as an ordinary unadjusted binary proportion.
Why is non-inferiority judged against the margin?
A non-inferiority claim is meaningful only relative to a clinically prespecified amount of acceptable difference. In EXCEL's registry-reported registry analyses, those margins are 2% at 30 days and 8.4% from 0 to 3 years. A p-value by itself does not contain that clinical tolerance threshold.
Why does the ITT population matter?
The 0-to-3-year analysis explicitly uses the ITT population. An ITT analysis retains participants according to randomized assignment and therefore preserves the principal comparison created by randomization. In non-inferiority trials, analysis-population choices are particularly important because departures from assigned treatment can affect the ability to distinguish true treatment differences from treatment switching or other deviations.
What does a hazard ratio mean?
A conventional hazard ratio is a relative comparison of instantaneous event rates under a time-to-event model. A value below 1 would conventionally correspond to a lower modeled hazard in the numerator group, while a value above 1 would conventionally correspond to a higher modeled hazard. However, the registry-reported EXCEL registry records create a specific reporting issue: the effect measure is labeled “Hazard Ratio (HR),” but one estimate is -3.1. That value cannot be interpreted as an ordinary positive hazard ratio without additional source information, so the registry value should be preserved rather than mathematically rewritten.
Why is the confidence interval important?
The confidence interval describes uncertainty around an estimated treatment effect under the specified analysis framework. For both posted analyses, the registry comment refers to a 95% CI of the Difference, but the numerical confidence-interval limits are not contained in the ClinicalTrials.gov record.
Why should the two posted analyses not be pooled?
The analyses use different time frames, different composite definitions, and different non-inferiority margins. The first is a 30-day analysis of death, protocol defined MI, or protocol defined stroke. The second covers 0 to 3 years and adds unplanned revascularization for ischemia. These differences define different estimands.
12. Intention-to-Treat and the Meaning of Randomization
Randomization is the structural foundation of the treatment comparison. The registry identifies the trial as randomized, and the 0-to-3-year statistical analysis explicitly uses the ITT population.
Randomization does not make every observed outcome identical between groups, nor does it eliminate uncertainty. Its principal statistical value is that, in expectation, treatment assignment is not determined by the prognostic factors that would otherwise confound the comparison.
13. Time-to-Event Endpoints and Censoring
The posted statistical analyses are identified as time-to-event endpoints and use a Kaplan-Meier-based Com-Nougue approach. This makes the distinction between event status and follow-up time important.
Event
A participant experiences one of the events specified by the relevant composite endpoint.
Censoring
A participant can contribute information up to the end of observed follow-up without necessarily having experienced the event.
Kaplan-Meier
The method estimates the failure or survival experience over time using the observed event and risk-set structure.
Non-inferiority
The estimated difference is interpreted against a prespecified margin rather than solely against zero.
The ClinicalTrials.gov record does not provide detailed censoring rules or missing-data procedures. Therefore, those elements should not be inferred from the mere presence of a Kaplan-Meier analysis.
14. Multiplicity, Interim Analysis, and Other Design Topics
The ClinicalTrials.gov record identifies a non-inferiority hypothesis, two posted statistical analyses, an ITT population for the 0-to-3-year analysis, and two distinct non-inferiority margins. They do not provide information supporting a description of interim efficacy analyses, alpha spending, a factorial design, Bayesian methods, or a specific multiplicity-adjustment strategy.
| Design topic | Supported by the ClinicalTrials.gov record? | What can be stated |
|---|---|---|
| Randomization | Yes | Allocation is randomized. |
| Parallel design | Yes | Design model is parallel. |
| Non-inferiority | Yes | Both posted statistical analyses use a non-inferiority hypothesis. |
| ITT | Yes | The 0-to-3-year analysis uses the ITT population. |
| Interim analysis | Not reported | No interim-analysis method is provided in the ClinicalTrials.gov record. |
| Multiplicity adjustment | Not reported | No multiplicity procedure is provided in the ClinicalTrials.gov record. |
| Factorial design | Not supported | The trial is identified as parallel rather than factorial. |
| Bayesian methods | Not reported | No Bayesian method is identified. |
| Missing-data/imputation method | Not reported | No imputation procedure is identified. |
| Stratification | Not reported | No stratification factors are identified in the ClinicalTrials.gov record. |
15. Interpreting the P-values
The two posted analyses have p-values of <0.0001 and 0.011. These numbers should be understood in the context of their respective non-inferiority hypotheses and margins.
A p-value measures the compatibility of the observed data with a specified null hypothesis under the statistical model. It does not tell the reader how large the treatment difference is, how clinically important the difference is, or the probability that the null hypothesis is true.
A value such as P = 0.011 should not be translated into a 98.9% probability that PCI or CABG is superior, inferior, or non-inferior. Such a conversion is not what the p-value represents.
The meaning of the p-value depends on the prespecified non-inferiority hypothesis and its margin. In the registry-reported EXCEL analyses, those margins are 2% and 8.4% for the respective analysis periods.
16. Understanding the Composite Endpoints
Composite endpoints can increase the number of observed events by combining several clinically defined outcomes. But the components may differ in clinical meaning and frequency.
| Analysis | Composite components |
|---|---|
| Registered primary endpoint, 5 years | All-cause death; protocol defined MI; protocol defined stroke |
| Posted secondary analysis, 30 days | All-cause death; protocol defined MI; protocol defined stroke |
| Posted secondary analysis, 0 to 3 years | Death; protocol defined MI; protocol defined stroke; unplanned revascularization for ischemia |
The addition of unplanned revascularization for ischemia in the 0-to-3-year endpoint is statistically consequential because it changes the event definition. It means that the second composite is not simply a longer follow-up version of the first analysis.
The registry's definition of all deaths includes cardiac death, vascular death, and non-cardiovascular death. Its protocol-defined MI definition distinguishes Q wave and non-Q wave MI using the ECG and laboratory criteria reported in the ClinicalTrials.gov record. Stroke is defined as the rapid onset of a new persistent neurologic deficit.
17. Safety Versus Efficacy
The EXCEL registry data illustrate why safety and efficacy should be treated as distinct statistical questions.
Efficacy
The posted analyses compare time-to-event composite outcomes between PCI and CABG under a non-inferiority framework.
Safety
The ClinicalTrials.gov record reports serious adverse-event counts by arm: 535/948 for PCI and 550/957 for CABG.
Different estimands
An efficacy hazard-related estimate and a serious-adverse-event count answer different questions and should not be collapsed into a single numerical measure.
Incomplete safety detail
The ClinicalTrials.gov record does not identify specific serious adverse-event categories or provide a formal statistical comparison.
18. Limitations and Interpretation Issues
- Registry-level reporting: this analysis is constrained to the ClinicalTrials.gov record. It does not add numerical results from publications or other external sources.
- Primary versus posted analyses: the registered primary endpoint has a 5-year time frame, while the ClinicalTrials.gov record is secondary analyses at 30 days and 0 to 3 years.
- Effect-measure inconsistency: the registry labels the estimates as hazard ratios, but the registry-reported numerical estimate of -3.1 cannot be interpreted as a conventional positive hazard ratio without additional information.
- Missing numerical confidence intervals: both Com-Nougue analysis comments refer to a 95% CI of the Difference, but the ClinicalTrials.gov record does not provide the CI bounds.
- Different composite definitions: the 0-to-3-year analysis adds unplanned revascularization for ischemia to the composite outcome.
- Different non-inferiority margins: the analyses posted on ClinicalTrials.gov use 2% at 30 days and 8.4% from 0 to 3 years. These are distinct analysis specifications.
- No detailed censoring information: the registry identifies time-to-event and Kaplan-Meier methodology through the Com-Nougue description, but the ClinicalTrials.gov record does not specify detailed censoring rules.
- No detailed missing-data procedure: the ClinicalTrials.gov record does not identify an imputation method or other missing-data strategy.
- No multiplicity procedure reported: the data do not identify an alpha-adjustment or multiplicity-control method for the posted analyses.
- Safety detail: serious adverse-event counts are available, but the ClinicalTrials.gov record does not include event-specific categories or a formal statistical comparison.
- Composite endpoints: combining several event types can make the overall endpoint more statistically efficient while potentially obscuring which component drives the result.
19. Why This Trial Matters Statistically
EXCEL provides a useful teaching example because the ClinicalTrials.gov record brings together randomized allocation, parallel-group comparison, composite time-to-event endpoints, Kaplan-Meier failure rates, an ITT analysis, hazard-ratio reporting, and non-inferiority margins.
| Concept | How it appears in EXCEL |
|---|---|
| Randomization | The trial allocation is randomized. |
| Parallel design | The design model is parallel. |
| Time-to-event endpoint | The two posted statistical analyses are identified as time-to-event analyses. |
| Kaplan-Meier estimation | The Com-Nougue description explicitly concerns two Kaplan-Meier failure rates. |
| Hazard ratio | The effect measure is reported in the registry as Hazard Ratio (HR). |
| Non-inferiority | Both posted analyses use a non-inferiority hypothesis. |
| Non-inferiority margin | Margins of 2% and 8.4% are specified for the two analyses. |
| ITT analysis | The 0-to-3-year analysis explicitly uses the ITT population. |
| Composite endpoint | Death, MI, and stroke form the registered primary composite; revascularization is added in the 0-to-3-year secondary composite. |
| P-values | The posted analyses report P < 0.0001 and P = 0.011. |
| Safety analysis | Serious adverse-event counts are reported separately for PCI and CABG. |
The particularly useful statistical lesson is that the estimand comes before the number. An estimate cannot be interpreted correctly without knowing the endpoint definition, time frame, analysis population, effect-measure definition, and hypothesis being tested. EXCEL's the ClinicalTrials.gov record demonstrate this directly because the two posted analyses differ in all of these dimensions.
20. Reading the Registry Results Correctly
Start with the endpoint
Determine exactly which event components are being counted and whether the endpoint is the registered 5-year primary measure or one of the posted secondary analyses.
Then identify the time frame
The analyses posted on ClinicalTrials.gov cover 30 days and 0 to 3 years. They should not be treated as equivalent follow-up periods.
Then identify the population
The 0-to-3-year analysis explicitly uses ITT. The registry-reported 30-day analysis does not specify an analysis population.
Then identify the hypothesis
Both analyses are labeled non-inferiority, with margins of 2% and 8.4%, respectively.
Only after those elements are established should the estimate and p-value be interpreted. This ordering prevents a common statistical error: treating a numerical estimate as self-explanatory when its meaning depends on the underlying estimand.
21. A Note on the Reported Effect Estimates
The two posted records deserve explicit documentation because the registry's effect-measure field and numerical estimate are not fully self-consistent in the ClinicalTrials.gov record.
| Analysis | Registry effect-measure label | Registry estimate | Registry analysis comment |
|---|---|---|---|
| 30 days | Hazard Ratio (HR) | -3.1 | p-value and 95% CI of the Difference derived from the Com-Nougue approach of non-inferiority for two Kaplan-Meier failure rates, with a 2% margin. |
| 0 to 3 years | Hazard Ratio (HR) | 4 | p-value and 95% CI of the Difference derived from the Com-Nougue approach of non-inferiority for two Kaplan-Meier failure rates, with an 8.4% margin. |
For an educational statistical record, the safest approach is to preserve both fields exactly and explain the discrepancy. A negative number cannot ordinarily be a conventional hazard ratio, while the registry's accompanying description refers to a difference. Without the underlying registry raw analysis record or statistical report, assigning a different meaning would go beyond the ClinicalTrials.gov record.
22. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial was randomized and parallel. The posted secondary analyses use a Com-Nougue non-inferiority framework for Kaplan-Meier failure rates, with p-values of <0.0001 and 0.011 and margins of 2% and 8.4%.
Clinical interpretation
The registry compares two different coronary revascularization interventions in a population with the listed coronary disease conditions. The ClinicalTrials.gov record alone do not establish a treatment recommendation.
Keeping these interpretations separate is important. Statistical non-inferiority is a formal property of an analysis relative to a prespecified margin; it is not itself a statement that the two procedures are identical, interchangeable in every respect, or preferable for every patient.
23. What the ClinicalTrials.gov record Do Not Establish
- The ClinicalTrials.gov record does not provide a numerical 5-year estimate for the registered primary endpoint.
- The ClinicalTrials.gov record does not provide numerical confidence-interval bounds for either posted statistical analysis.
- The ClinicalTrials.gov record does not provide median time-to-event values.
- The ClinicalTrials.gov record does not provide subgroup analyses or interaction tests.
- The ClinicalTrials.gov record does not provide a Kaplan-Meier curve or event-by-time dataset.
- The ClinicalTrials.gov record does not provide an interim-analysis procedure.
- The ClinicalTrials.gov record does not provide a multiplicity-adjustment procedure.
- The ClinicalTrials.gov record does not provide a missing-data or imputation method.
- The ClinicalTrials.gov record does not provide a stratification scheme.
- The ClinicalTrials.gov record does not provide Bayesian analyses.
These omissions are important because they define the boundary between what can be documented from the registry data and what would require additional source material. A statistically rigorous trial summary should preserve that boundary rather than filling it with plausible but unverified details.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Statistical Calculators
26. Sources
- ClinicalTrials.gov: NCT01205776 — EXCEL Clinical Trial.
- PubMed: PubMed record — PMID 41664927.
- PubMed: PubMed record — PMID 39632005.
- PubMed: PubMed record — PMID 38502720.
- PubMed: PubMed record — PMID 38465592.
- PubMed: PubMed record — PMID 37256598.
Continue through the Clinical Biostats statistical pathway
Explore tutorials and calculators for survival analysis, non-inferiority design, randomized trials, and related clinical-trial methods.
27. Record Summary
EXCEL is a randomized, parallel clinical trial comparing PCI with CABG, with 1905 participants and a registered 5-year primary composite endpoint of all-cause death, protocol defined MI, or protocol defined stroke. The ClinicalTrials.gov record additionally contain two secondary time-to-event analyses using a Com-Nougue approach for Kaplan-Meier failure rates under non-inferiority hypotheses. The 30-day analysis reports an estimate of -3.1 and P < 0.0001 with a 2% margin; the 0-to-3-year analysis, conducted in the ITT population, reports an estimate of 4 and P = 0.011 with an 8.4% margin. Serious adverse events are reported as 535/948 for PCI and 550/957 for CABG.