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
SMART-CHOICE was a randomized, parallel-group trial comparing P2Y12 antagonist monotherapy with aspirin plus a P2Y12 antagonist after drug-eluting stent implantation in patients with coronary artery disease. The registered primary endpoint was a composite of death, myocardial infarction, or cerebrovascular events at 1 year.
| Feature | SMART-CHOICE |
|---|---|
| Trial identifier | NCT02079194 |
| Condition | Coronary Artery Disease |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 3000 |
| Number of arms | 2 |
| Lead sponsor | Samsung Medical Center |
| Sponsor type | Other |
| Trial phase | NA |
| Status | Unknown |
| Start date | 2014-03-18 |
| Primary completion date | 2018-07-07 |
2. Clinical Question
The central clinical question was whether, after a drug-eluting stent procedure, treatment with a P2Y12 antagonist alone could be compared with dual antiplatelet therapy consisting of aspirin plus a P2Y12 antagonist with respect to a composite of death, myocardial infarction, or cerebrovascular events during the first year.
Population
Patients enrolled in the SMART-CHOICE trial with coronary artery disease undergoing the relevant index procedure involving a drug-eluting stent.
Intervention
P2Y12 antagonist monotherapy.
Comparator
Aspirin plus a P2Y12 antagonist.
Primary question
How do the randomized treatment strategies compare for the registered 1-year composite endpoint of death, myocardial infarction, or cerebrovascular events?
3. Trial Design
The registry describes SMART-CHOICE as a randomized, parallel-group treatment trial with two intervention arms and no masking. The planned enrollment was 3000 participants.
P2Y12 antagonist monotherapy
- P2Y12 antagonist monotherapy
- Drug intervention
- Randomized treatment arm
Aspirin plus P2Y12 antagonist
- Aspirin plus a P2Y12 antagonist
- Drug intervention
- Randomized treatment arm
4. Endpoints
The registry identifies one registered primary endpoint. Its definition and time frame are reproduced below in a compact statistical format.
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Primary endpoint | A composite of death, myocardial infarction, or cerebrovascular events | 1 year |
| Registry description | 12 months after the index procedure | 12 months after the index procedure |
Why the composite endpoint matters
A composite endpoint combines multiple clinically distinct events into a single outcome. In SMART-CHOICE, the components are death, myocardial infarction, and cerebrovascular events. Statistically, the first qualifying component experienced during the specified follow-up would ordinarily determine whether a participant experiences the composite event.
The advantage of a composite is that it can capture more total events than any single component alone, potentially increasing statistical information. The trade-off is interpretability: a difference in the composite does not automatically mean that every component changes to the same degree or has the same clinical importance.
5. Statistical Methodology
The ClinicalTrials.gov record identifies the primary endpoint and its 1-year assessment window but does not post a formal statistical analysis for the endpoint. For a randomized trial with a binary composite outcome assessed at a fixed follow-up time, the analysis would ordinarily compare the proportion of participants experiencing the composite event between randomized treatment groups.
Several related effect measures can be useful for this type of endpoint. A risk ratio compares the event probabilities between groups, while a risk difference describes their absolute difference. An odds ratio compares the corresponding odds. The choice of effect measure should follow the prespecified statistical analysis plan rather than being selected after observing the results.
A negative risk difference would indicate a lower observed event probability in the first group; a positive value would indicate a higher observed event probability. This formula describes the estimand conceptually rather than providing an observed SMART-CHOICE result.
Time-to-event alternative
If the registry's full statistical analysis plan specifies use of time-to-first-event information rather than only the status at 12 months, a time-to-event framework would ordinarily be appropriate. Kaplan-Meier estimation can describe the cumulative event experience over follow-up, while a Cox proportional-hazards model can estimate a hazard ratio.
The distinction is important. A fixed-time analysis asks whether a participant experienced the composite by the specified assessment point. A time-to-event analysis also uses information about when the event occurred and can account for participants who are censored before the end of follow-up.
A hazard ratio is not a risk ratio and should not be interpreted as a percentage of participants who experience or avoid the event.
Intention-to-treat principle
Because SMART-CHOICE is randomized, the principal treatment comparison would ordinarily preserve the randomized assignment of participants. An intention-to-treat analysis compares groups according to the treatment to which participants were randomized, regardless of subsequent treatment changes, discontinuation, or adherence.
This principle protects the benefit of randomization. If participants are selectively removed from their randomized groups after treatment begins, the resulting comparison can become more vulnerable to confounding because post-randomization characteristics may be related to both treatment exposure and outcomes.
6. Statistical Methods Explained
Why use a composite endpoint?
A composite endpoint can increase the number of observed events by combining several clinically relevant outcomes. This can provide greater statistical information than relying on a less frequent individual component. The interpretation, however, remains dependent on understanding which components contribute events and whether the components are clinically coherent.
Why does randomization matter statistically?
Randomization creates the basis for a causal comparison between treatment strategies by making treatment assignment independent of baseline characteristics in expectation. With adequate randomization and appropriate analysis, systematic differences in outcomes between groups can be attributed to the assigned treatment strategy rather than to measured or unmeasured baseline differences, subject to the usual assumptions of randomized-trial inference.
What is the difference between risk ratio and risk difference?
The risk ratio expresses the event probability in one group relative to another. The risk difference expresses the absolute difference in event probability. These answer different questions: a relative effect describes proportional change, whereas an absolute effect describes the number of additional or fewer events per corresponding number of treated participants.
Relative measure
A risk ratio of 0.80 would correspond to an event probability in the treatment group that is 80% of the comparator probability.
Absolute measure
A risk difference of −0.02 would correspond to an absolute reduction of 0.02 in event probability, equivalent to 2 fewer events per 100 participants under the stated comparison.
Why might Kaplan-Meier analysis be useful?
If participants are followed for different lengths of time or some participants do not experience the endpoint before follow-up ends, simply classifying everyone at a single time point can discard information. Kaplan-Meier methods use the observed event times and censoring times to estimate the event-free probability over follow-up.
What does a Cox hazard ratio represent?
A Cox model compares the estimated instantaneous event rates between treatment groups over time. A hazard ratio below 1 indicates a lower modeled instantaneous event rate in the treatment group. It does not directly provide the absolute probability of experiencing the endpoint by 1 year.
Why must the analysis population be specified?
Randomized, treated, and per-protocol populations can answer different questions. The randomized population preserves the treatment assignment created by randomization, whereas an analysis based only on participants who adhered to treatment may describe a more selected subset. The prespecified analysis population is therefore part of the estimand and should not be chosen after seeing the results.
7. Planned Analysis
The registered primary endpoint is a composite of death, myocardial infarction, or cerebrovascular events measured 1 year after the index procedure. For this type of endpoint, a conventional analysis would compare the occurrence of the composite between the two randomized treatment groups using an appropriate effect measure such as a risk ratio, risk difference, or odds ratio. If event timing and censoring are incorporated, Kaplan-Meier and time-to-event methods would ordinarily be considered.
| Statistical question | Typical approach for this endpoint type |
|---|---|
| Did the proportion with the composite differ between randomized groups? | Two-group comparison of binary event risk, with an appropriate effect estimate and confidence interval. |
| How does event experience evolve over follow-up? | Kaplan-Meier estimation of time to first composite event when event timing is available. |
| What is the relative event rate over follow-up? | Cox proportional-hazards modeling may be used when its assumptions and prespecified analysis support it. |
| How should component outcomes be understood? | Death, myocardial infarction, and cerebrovascular events should be examined individually in addition to the composite when component results are available. |
8. Interpreting a Future Composite-Endpoint Result
When a randomized trial reports a result for a composite endpoint, the first question is the effect estimate. The second is its uncertainty. The third is whether the statistical comparison corresponds to the prespecified analysis and population. Only after these are established should the magnitude and clinical meaning of the observed difference be considered.
An effect estimate summarizes the contrast between the two randomized treatment strategies. For a risk ratio, values below 1 indicate a lower event probability in the numerator treatment group relative to the comparator. For a risk difference, negative values indicate lower absolute event probability in the numerator treatment group.
A confidence interval describes statistical uncertainty around the estimated treatment effect under the specified statistical model and sampling framework. A narrow interval indicates greater precision than a wide interval; it does not describe the range of outcomes that individual participants can experience.
A p-value measures how compatible the observed data are with a specified null hypothesis under the statistical model. It does not measure the size of the treatment effect, the clinical importance of the result, or the probability that either treatment is effective.
A statistically significant composite result should not automatically be described as a statistically significant effect on every component. Component-specific event counts and effect estimates are needed to determine how the overall composite result was generated.
9. Analysis of the Composite Endpoint
The SMART-CHOICE primary endpoint combines three event types: death, myocardial infarction, and cerebrovascular events. This structure creates an important hierarchy of interpretation.
| Level | Statistical question | Interpretation |
|---|---|---|
| Composite | Did participants experience any qualifying component? | The primary registered endpoint is defined at this combined level. |
| Components | Which individual event types contributed to the composite? | Component-specific results clarify the clinical composition of the overall effect. |
| Timing | When did the first qualifying event occur? | Time-to-event methods can use event timing when available. |
| Absolute effect | How large was the difference in event probability? | Risk differences communicate absolute treatment effects. |
| Relative effect | How large was the proportional difference? | Risk ratios or hazard ratios communicate relative effects under their respective definitions. |
For clinical interpretation, the distinction between these levels is essential. A composite endpoint is not simply a larger version of a single endpoint; it represents a particular statistical definition in which the first qualifying event determines whether the participant contributes an event to the composite.
10. Multiplicity and Endpoint Interpretation
The registry identifies one registered primary endpoint: a composite of death, myocardial infarction, or cerebrovascular events at 1 year. No additional registered primary endpoints or formal multiplicity strategy is described in the information reported here.
Multiplicity becomes relevant if investigators formally test several endpoints, several treatment contrasts, multiple time points, or multiple subgroup hypotheses. The more hypotheses that are tested under a common familywise error framework, the greater the need for prespecified error control.
Primary endpoint
The registered primary endpoint is one composite outcome assessed at 1 year after the index procedure.
Component outcomes
Death, myocardial infarction, and cerebrovascular events are components of the composite rather than separate registered primary endpoints in the record.
Component-specific analyses can be clinically informative, but their inferential interpretation depends on the statistical analysis plan. A result for an individual component should not automatically be treated as an independently powered confirmatory test simply because it is clinically important.
11. Missing Data, Censoring, and Follow-Up
The registry defines the primary endpoint at 1 year, but the information reported here does not specify a missing-data or imputation strategy. For a fixed 1-year binary analysis, the handling of participants whose 1-year outcome status is unavailable is a key methodological issue because different assumptions about missing outcomes can affect the estimated treatment comparison.
If the analysis instead uses time-to-first-event methods, censoring becomes central. A censored participant contributes information until the censoring time but is not treated as having experienced the event at that time. Standard survival methods rely on assumptions about the relationship between censoring and the underlying event process.
12. Non-Inferiority Considerations
The registered information identifies the comparison between P2Y12 antagonist monotherapy and aspirin plus a P2Y12 antagonist, but it does not report a non-inferiority margin in the information presented here.
If the statistical objective is non-inferiority, the interpretation differs from an ordinary superiority test. Non-inferiority asks whether the treatment effect is not worse than a prespecified clinically acceptable margin. That margin must be specified before the results are examined and must be justified clinically as well as statistically.
The margin, confidence level, effect scale, and direction of the comparison all matter. A non-significant superiority test is not, by itself, proof of non-inferiority.
Because the registry information reported here does not provide a non-inferiority margin or formal analysis specification, the statistical conclusion for such a framework cannot be established from the registry endpoint definition alone.
13. Safety and Adverse Events
The registry information reported here identifies the trial interventions and primary efficacy endpoint but does not provide posted serious adverse-event counts by treatment arm. Consequently, no arm-specific safety comparison can be presented from the registry information described here.
Statistically, safety is usually summarized separately from the primary efficacy endpoint. Serious adverse events, treatment-emergent adverse events, treatment discontinuations, and deaths may have different denominators and different exposure definitions. Combining them into a single safety percentage can therefore obscure clinically important distinctions.
14. Trial Timeline
Trial start
The registered study start date was 2014-03-18.
Randomized trial period
The study was designed as a randomized, parallel-group comparison with 3000 planned participants and two treatment arms.
Primary completion
The registered primary completion date was 2018-07-07.
15. Statistical Interpretation of the Primary Endpoint
The primary endpoint is a composite of death, myocardial infarction, or cerebrovascular events measured 12 months after the index procedure. It therefore captures whether a participant experienced at least one of these specified clinical events during the registered assessment period.
The composite endpoint does not, by itself, distinguish the relative importance or frequency of its individual components. A composite event rate also does not provide a direct measure of treatment adherence, symptom improvement, or other outcomes unless those outcomes are separately defined and analyzed.
For a 1-year binary endpoint, an absolute risk difference is especially useful because it expresses the change in the probability of experiencing the composite. Relative measures can look substantial even when absolute event probabilities are small, while absolute differences depend directly on the underlying event risk.
Any estimated treatment effect should be accompanied by a confidence interval. The interval communicates the precision of the estimate and helps distinguish a relatively precise estimate from one that remains compatible with a broad range of treatment effects.
16. Important Limitations and Interpretation Issues
- No posted statistical analyses: the ClinicalTrials.gov record does not post estimates, confidence intervals, or p-values for the registered primary endpoint in the information reported here.
- Composite endpoint: death, myocardial infarction, and cerebrovascular events may contribute differently to the overall composite, so the composite should not be interpreted as though it were a single homogeneous clinical outcome.
- Open-label design: the registry identifies masking as none. Treatment awareness can influence some aspects of clinical management and outcome ascertainment, depending on the endpoint and assessment process.
- Missing follow-up: the registry information does not specify how missing 1-year endpoint observations were handled.
- Censoring assumptions: if a time-to-event analysis is used, interpretation depends on the validity of the censoring assumptions underlying the survival analysis.
- Non-inferiority framework: the information reported here does not specify a non-inferiority margin, so the margin-based interpretation of the treatment comparison cannot be reconstructed from the endpoint definition alone.
- Analysis population: the registry information reported here does not specify the complete prespecified efficacy and safety analysis populations.
- Statistical model: the registry information reported here does not identify a specific primary statistical model for the composite endpoint.
17. Why This Trial Matters Statistically
SMART-CHOICE is a useful statistical teaching case because it places a common randomized-trial question into a setting where the endpoint is a clinically meaningful composite and the treatment comparison involves two different antiplatelet strategies.
| Concept | How it appears in SMART-CHOICE |
|---|---|
| Randomization | Participants were assigned using a randomized allocation. |
| Parallel-group design | The registry identifies a parallel design with two arms. |
| Fixed-time endpoint | The primary endpoint is assessed at 1 year, described as 12 months after the index procedure. |
| Composite endpoint | Death, myocardial infarction, or cerebrovascular events are combined into one primary endpoint. |
| Absolute effect | A risk difference can communicate the absolute difference in 1-year composite-event probability. |
| Relative effect | A risk ratio can communicate proportional differences in event probability; a hazard ratio can be used when a time-to-event framework is prespecified. |
| Time-to-event analysis | Kaplan-Meier and Cox methods are typical approaches when event timing and censoring are part of the analysis. |
| Intention-to-treat | Randomized treatment assignment provides the natural basis for the primary causal comparison. |
| Open-label design | The registry identifies masking as none. |
| Non-inferiority logic | A non-inferiority conclusion, if that is the prespecified objective, requires a justified margin and appropriate confidence-interval interpretation. |
18. A Practical Statistical Reading of SMART-CHOICE
A statistically rigorous reading of SMART-CHOICE begins with the estimand rather than the p-value. The estimand is the treatment contrast of interest: the difference between the randomized P2Y12 antagonist monotherapy strategy and aspirin plus P2Y12 antagonist strategy for the registered 1-year composite endpoint.
The next step is to establish the analysis scale. If the primary analysis is a binary 1-year comparison, the event probability in each randomized group and an absolute risk difference would provide a direct clinical description. A relative risk measure could supplement that absolute effect. If the analysis uses time-to-event information, the timing of events and censoring become part of the estimand, and a hazard ratio may be appropriate.
The third step is to examine uncertainty. An effect estimate without a confidence interval gives an incomplete picture because the estimate is subject to sampling variability. The interval should be interpreted in relation to the prespecified null hypothesis or non-inferiority margin, depending on the trial's formal statistical objective.
Finally, the composite endpoint should be unpacked. A treatment strategy may affect death, myocardial infarction, and cerebrovascular events differently. The overall composite therefore provides one important summary, but the clinical meaning of the result depends on the contribution and seriousness of its individual components.
19. What Would Constitute a Complete Statistical Report?
A complete results report for this trial's primary endpoint would ordinarily allow readers to identify the event count and event proportion in each randomized treatment group, the effect estimate, its confidence interval, and the corresponding hypothesis test if one was prespecified.
| Reporting element | Why it matters |
|---|---|
| Events in each arm | Shows the observed number of participants contributing to the composite endpoint. |
| Risk in each arm | Provides the absolute event probabilities. |
| Risk difference | Communicates the absolute treatment effect. |
| Relative effect | Communicates the proportional treatment effect. |
| 95% confidence interval | Shows statistical precision around the effect estimate. |
| P-value | Addresses compatibility with the specified null hypothesis under the statistical model. |
| Component outcomes | Shows how death, myocardial infarction, and cerebrovascular events contributed to the composite. |
| Analysis population | Defines which participants contributed to the reported estimate. |
| Missing-data handling | Shows how incomplete 1-year outcome information was treated. |
| Prespecified analysis method | Connects the reported inference to the trial's planned statistical framework. |
These elements are particularly important for a composite endpoint because a single summary number can conceal meaningful differences in event frequency, event timing, or component contribution.
20. Related Statistical Concepts
The statistical framework of SMART-CHOICE connects naturally to broader methods for randomized clinical trials, binary outcomes, composite endpoints, and survival analysis.
21. Sources
- ClinicalTrials.gov: SMART-CHOICE, NCT02079194.
Continue through the Clinical Biostats trial and statistics library
Explore broader statistical methods and clinical-trial analyses that build on the concepts used to interpret randomized treatment comparisons.
22. Record Summary
SMART-CHOICE was a randomized, parallel-group trial enrolling 3000 participants with coronary artery disease and comparing P2Y12 antagonist monotherapy with aspirin plus a P2Y12 antagonist. The registered primary endpoint was a composite of death, myocardial infarction, or cerebrovascular events measured 1 year, defined as 12 months after the index procedure.
Statistically, the trial provides a useful framework for understanding how randomized treatment comparisons are constructed around composite clinical outcomes. The most important interpretive distinctions are between absolute and relative effects, fixed-time and time-to-event analyses, composite and component outcomes, and superiority versus non-inferiority inference.