This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
SOCRATES was a randomized, double-blind, parallel-group phase 3 trial evaluating ticagrelor versus acetylsalicylic acid (ASA) for prevention after acute ischaemic stroke or transient ischaemic attack. The primary endpoint was a composite of stroke, myocardial infarction, or death from randomization through day 97.
| Feature | SOCRATES |
|---|---|
| Phase | Phase 3 |
| Therapeutic area | Neurology |
| Population | Participants with acute ischaemic stroke or transient ischaemic attack |
| Design | Randomized, double-blind, parallel-group |
| Primary purpose | Prevention |
| Interventions | Ticagrelor (drug) and acetylsalicylic acid (ASA) (drug) |
| Primary endpoint | Number of participants with composite of stroke/MI/death |
| Primary endpoint time frame | From randomization up to 97 days |
| Primary endpoint type | Binary in the registered endpoint field; analyzed as a time-to-event endpoint |
| Hypothesis type | Superiority |
| Results posted | Yes |
| ClinicalTrials.gov | NCT01994720 |
2. Clinical Question
The statistical question was whether ticagrelor 90 mg differed from ASA 100 mg with respect to the time to the first occurrence of the registered composite of stroke, myocardial infarction, or death during the period from randomization through day 97.
Population
Participants with acute ischaemic stroke or transient ischaemic attack enrolled in a phase 3 prevention trial.
Intervention
Ticagrelor 90 mg.
Comparator
Acetylsalicylic acid 100 mg.
Primary question
Does ticagrelor produce a different time-to-event profile for the composite of stroke, MI, or death compared with ASA through 97 days?
3. Trial Design
Ticagrelor 90 mg
- Randomized intervention: ticagrelor 90 mg.
- Efficacy analyses compared this group with ASA 100 mg.
- Primary efficacy analysis used the full analysis set, including all randomized patients.
- Safety analyses used the safety analysis set.
ASA 100 mg
- Randomized comparator: acetylsalicylic acid 100 mg.
- Efficacy analyses compared this group with ticagrelor 90 mg.
- Primary efficacy analysis used the full analysis set, including all randomized patients.
- Safety analyses used the safety analysis set.
4. Trial Timing and Status
Trial start
The registry lists 2014-01-07 as the study start date.
Primary completion
The registry lists 2016-03-02 as the primary completion date.
Registry status
The trial is listed as COMPLETED.
5. Analysis Populations
The registry explicitly distinguishes the population used for efficacy analyses from the population used for safety analyses. That distinction matters because the estimand represented by a randomized efficacy analysis is different from the question addressed by an exposure-based safety analysis.
| Population | Definition in the registry | Use on this page |
|---|---|---|
| Full analysis set | Included all randomized patients. | Primary and reported efficacy analyses. |
| Safety analysis set | Included all patients who received at least 1 dose of randomized ticagrelor or ASA and for whom post-dose data are available. | Safety analyses and serious adverse-event summaries. |
The full analysis set is closely aligned with the intention-to-treat principle because treatment assignment remains the organizing basis for the efficacy comparison. In contrast, the safety set is defined by treatment exposure, which is appropriate for describing adverse events occurring after patients actually receive study treatment.
6. Primary Endpoint
| Endpoint | Registered definition | Time frame | Analysis |
|---|---|---|---|
| Number of Participants With Composite of Stroke/MI/Death | Participants with stroke, MI or death. If no event, censoring occurs at the minimum of last date of event assessment, end of treatment date, or day 97. | From randomization up to 97 days | Cox proportional-hazards model |
The registry describes the endpoint in the primary endpoint field as binary, but the posted statistical analysis treats it as a time-to-event endpoint. That distinction is important: the analysis does not merely ask whether a participant ever experienced the composite during the study period. The Cox model uses the timing of the event and accommodates censoring.
The analysis population was the full analysis set, which included all randomized patients. The comparison was ticagrelor 90 mg versus ASA 100 mg.
7. Primary Result
Composite of Stroke, MI, or Death
Hazard ratio
95% CI: 0.78–1.01 · P = 0.0670
Comparison: Ticagrelor 90 mg vs ASA 100 mg
Time frame: From randomization up to 97 days
The primary analysis used a Cox proportional-hazards model in the full analysis set. The estimated hazard ratio of 0.89 indicates that the estimated instantaneous rate of experiencing the composite endpoint was approximately 11% lower in the ticagrelor group than in the ASA group under the fitted model.
What the estimate means: An HR of 0.89 is a relative time-to-event estimate. It compares the modeled instantaneous event rate between the randomized groups over the analyzed period. Expressed descriptively, 0.89 corresponds to an estimated 11% lower hazard for ticagrelor relative to ASA.
What it does not mean: It does not mean that 11% fewer participants necessarily experienced the endpoint, nor does it mean that every participant had an 11% lower probability of stroke, MI, or death. A hazard ratio is not an absolute risk difference or a risk ratio.
What the confidence interval says: The 95% CI of 0.78–1.01 quantifies uncertainty around the estimated hazard ratio. It includes 1.00, so the data are compatible with a range of relative effects that includes no hazard difference under the model.
Why the p-value is different from the effect size: The P-value of 0.0670 describes the statistical evidence against the relevant null hypothesis under the specified analysis. It does not measure how large or clinically important the treatment effect is. Effect magnitude and statistical evidence are separate quantities.
Time-to-event caution: The interpretation relies on the Cox proportional-hazards framework. A single hazard ratio is most straightforward when the proportional-hazards assumption is reasonable over the analyzed period. The ClinicalTrials.gov record does not provide a formal proportional-hazards diagnostic.
8. Secondary Efficacy Results
The registry contains several secondary time-to-event analyses using the same broad comparison of ticagrelor 90 mg versus ASA 100 mg. Each efficacy analysis used the full analysis set, meaning all randomized patients were included in the stated analysis population.
| Secondary endpoint | Analysis | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Ischaemic stroke | Cox proportional-hazards model | HR 0.87 | 0.76–1.00 | 0.0462 |
| Net Clinical Outcome | Cox proportional-hazards model | HR 0.90 | 0.79–1.02 | 0.0928 |
| Composite of Ischaemic Stroke, MI and CV Death | Cox proportional-hazards model | HR 0.89 | 0.78–1.01 | 0.0771 |
| All-Cause Death | Cox proportional-hazards model | HR 1.18 | 0.83–1.67 | 0.3641 |
| CV Death | Cox proportional-hazards model | HR 1.18 | 0.75–1.85 | 0.4828 |
| MI | Cox proportional-hazards model | HR 1.20 | 0.67–2.14 | 0.5457 |
| Stroke | Cox proportional-hazards model | HR 0.86 | 0.75–0.99 | 0.0342 |
| Fatal Stroke | Cox proportional-hazards model | HR 1.06 | 0.55–2.06 | 0.8557 |
| Disabling Stroke | Cox proportional-hazards model | HR 0.90 | 0.77–1.06 | 0.2126 |
Reading the secondary hazard ratios
The secondary estimates illustrate why the hazard ratio should always be read together with its confidence interval. For example, the estimated HR for ischaemic stroke was 0.87 with a 95% CI of 0.76–1.00, whereas the estimate for MI was 1.20 with a 95% CI of 0.67–2.14. The latter interval is substantially wider relative to its estimate, indicating greater statistical uncertainty about the treatment comparison for that endpoint.
The direction of an estimate also should not be confused with certainty. An HR below 1 is an estimate favoring the ticagrelor group in the modeled comparison, while an HR above 1 is an estimate favoring the ASA group in the opposite direction. Neither direction alone establishes that a true treatment difference exists.
9. Secondary Functional Outcome Analysis
Severity of Stroke and Overall Disability
Odds ratio
95% CI: 0.85–1.02 · P = 0.1393
Analysis: Logistic regression
The registry analyzed the number of participants by severity of stroke and overall disability using logistic regression. The analysis population was the full analysis set, and the reported effect measure was an odds ratio comparing ticagrelor 90 mg with ASA 100 mg.
What the OR means: An odds ratio of 0.93 indicates that the modeled odds associated with the analyzed binary outcome were estimated to be 7% lower with ticagrelor than with ASA, under the logistic regression model.
Odds are not probabilities: An odds ratio of 0.93 should not be read as a 7% absolute reduction in the probability of the outcome. Odds are defined as probability divided by one minus probability, and the odds ratio compares those odds.
Precision: The 95% CI of 0.85–1.02 includes 1.00. This means the interval includes the possibility of no odds difference under the specified model.
P-value: The P-value of 0.1393 quantifies statistical evidence under the model and null hypothesis; it is not a measure of the magnitude or clinical importance of the odds ratio.
10. Safety Results
The registry reports serious adverse events by randomized arm and also provides formal time-to-event analyses for PLATO Major Bleeding Event and premature discontinuation due to any bleeding adverse event.
| Safety measure | ASA 100 mg | Ticagrelor 90 mg |
|---|---|---|
| Serious adverse events | 533 / 6581 affected / at risk | 532 / 6549 affected / at risk |
These counts are descriptive safety results. They should not be converted into an unadjusted treatment comparison by dividing the two affected counts or by treating the displayed fractions as if they were a formal effect estimate. The registry separately reports model-based safety analyses for selected endpoints.
PLATO Major Bleeding Event
Hazard ratio
95% CI: 0.52–1.34 · P = 0.4511
Time frame: From randomization up to 97 days
This analysis used a Cox proportional-hazards model in the safety analysis set. The safety set included patients who received at least 1 dose of randomized ticagrelor or ASA and for whom post-dose data were available.
An HR of 0.83 is an estimated relative hazard measure for PLATO Major Bleeding Event, not a statement that the absolute probability of major bleeding was 17% lower. The 95% CI of 0.52–1.34 is wide and includes 1.00, indicating substantial uncertainty around the estimated hazard ratio. The P-value of 0.4511 is not a measure of effect size.
Premature Discontinuation Due to Any Bleeding Adverse Event
Hazard ratio
95% CI: 1.53–3.34 · P < 0.0001
Time frame: From first dose and up to and including 7 days following the date of last dose of the study
The registry reports a Cox proportional-hazards analysis for premature discontinuation of study drug due to any bleeding adverse event. The estimated hazard ratio of 2.26 means that the modeled instantaneous rate of this discontinuation event was estimated to be more than twice as high in the ticagrelor group as in the ASA group.
The confidence interval of 1.53–3.34 lies above 1.00, so the estimated effect is directionally separated from the null value under the reported model. The P-value of <0.0001 indicates strong statistical evidence against the null hypothesis for this analysis. It does not, however, tell us how many participants discontinued, nor does it quantify the absolute excess risk.
The time frame also differs from the primary efficacy endpoint: this safety endpoint is measured from first dose through 7 days after the last dose. Comparisons across endpoints therefore require attention to the endpoint definition and observation window rather than treating all hazard ratios as interchangeable.
11. Statistical Methodology
Cox Proportional-Hazards Model
The principal statistical method used for the primary endpoint and most reported secondary endpoints was the Cox proportional-hazards model. This is appropriate for endpoints defined by time until an event, particularly when some participants do not experience the event during the observation period and are therefore censored.
The hazard ratio associated with a treatment indicator is represented by exp(β). The model compares instantaneous event rates while allowing the baseline hazard h0(t) to remain unspecified.
For SOCRATES, the reported analysis compares ticagrelor 90 mg with ASA 100 mg. The resulting HR is therefore a relative measure of the event hazard under the Cox model, rather than an absolute measure of treatment benefit.
Censoring
The primary endpoint definition explicitly states that participants without an event are censored at the minimum of the last date of event assessment, the end of treatment date, or day 97. Censoring allows participants who have not experienced the endpoint to contribute information for the portion of follow-up during which their event status is known.
Why censoring matters
A participant who remains event-free until censoring contributes information up to that time but does not provide an observed event after the censoring point.
Why follow-up matters
The primary endpoint is explicitly limited to the period from randomization through 97 days. The HR therefore describes the randomized comparison over that defined analysis window.
Logistic Regression
One reported secondary analysis used logistic regression for the number of participants by severity of stroke and overall disability. Logistic regression models the relationship between covariates and the log odds of a binary outcome and naturally produces an odds ratio for a treatment comparison.
For a treatment indicator, exp(β) is the modeled odds ratio comparing the treatment group with the comparator group.
Intention-to-Treat Analysis
The registry states that the efficacy population was the full analysis set, which included all randomized patients. This corresponds to the central intention-to-treat idea: randomized participants remain part of the efficacy comparison according to their assigned group.
The importance of this approach is causal rather than merely administrative. Randomization establishes the treatment groups before subsequent events occur. Preserving randomized assignment helps maintain the comparability created by randomization, although post-randomization events can still complicate interpretation.
Superiority Testing
The registry identifies the hypothesis type as superiority. In a superiority framework, the null hypothesis generally corresponds to no treatment difference for the specified effect measure, while the alternative permits a difference in either direction for a two-sided test.
The primary analysis reports a two-sided 95% confidence interval and a P-value of 0.0670. The confidence interval and P-value should be interpreted together with the prespecified analysis framework rather than using the P-value as a stand-alone measure of clinical importance.
12. Hazard Ratios: How to Read the SOCRATES Results
| HR | Descriptive interpretation | What it does not establish |
|---|---|---|
| Below 1 | Lower estimated instantaneous event rate in the ticagrelor group under the fitted model. | It does not establish an absolute risk reduction or that every patient benefits. |
| Equal to 1 | No estimated hazard difference between groups. | It does not prove the treatments are clinically identical. |
| Above 1 | Higher estimated instantaneous event rate in the ticagrelor group under the fitted model. | It does not by itself establish a clinically important harm. |
The SOCRATES results include examples on both sides of 1.00. The primary composite has HR 0.89, all-cause death has HR 1.18, and MI has HR 1.20. These values are estimates, and the uncertainty around each estimate is reflected by its confidence interval.
Do not rank endpoints by their hazard ratio alone. An HR of 0.86 with a 95% CI of 0.75–0.99 and an HR of 0.87 with a 95% CI of 0.76–1.00 are close in estimated magnitude, while their precision and statistical evidence also depend on the corresponding confidence intervals and P-values. Conversely, a larger-looking point estimate can be much less informative when its confidence interval is wide.
13. Statistical Methods Explained
Why was a Cox model used for the primary endpoint?
The primary endpoint is defined by the occurrence of stroke, MI, or death and specifies censoring for participants without an event. That makes the timing of the event relevant. A Cox proportional-hazards model uses both event timing and censoring information rather than reducing every participant to a simple event/no-event indicator.
What does an HR of 0.89 mean?
Under the fitted Cox model, the estimated instantaneous rate of the primary composite was 0.89 times the rate in the ASA group for ticagrelor, corresponding descriptively to an approximately 11% lower estimated hazard. It is not an 11-percentage-point reduction in risk and does not imply that 11% of participants benefited.
Why does the confidence interval matter?
The point estimate is only one possible summary of the treatment effect. The 95% CI of 0.78–1.01 shows the range of values compatible with the uncertainty of the estimate under the model and sampling framework. Because the interval reaches across 1.00, the data do not exclude a null hazard ratio at that confidence level.
Why is a P-value not an effect size?
The P-value is influenced by the observed data and the statistical model, and it quantifies evidence against the null hypothesis. It does not tell us whether an effect is large, small, clinically important, or practically meaningful. Those questions require the effect estimate, its confidence interval, and clinical context.
How is an odds ratio different from a hazard ratio?
A hazard ratio compares instantaneous event rates over time, whereas an odds ratio compares odds for a binary outcome. The SOCRATES registry uses HRs for time-to-event endpoints and an OR for the logistic regression analysis of stroke severity and overall disability. The two measures should not be interpreted as interchangeable.
Why does the analysis population matter?
The efficacy analyses used the full analysis set containing all randomized patients, while the safety analyses used a treatment-exposure-based safety set. The former preserves the randomized comparison for efficacy; the latter focuses on patients who actually received study treatment and had post-dose data available.
Why should the 97-day time frame be kept visible?
A time-to-event estimate is tied to the period over which events and censoring are observed. The registered primary endpoint is explicitly measured from randomization up to 97 days. Extending the interpretation beyond that period would require evidence from an appropriately defined longer-term analysis.
14. Primary and Secondary Results in Context
| Endpoint | Measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Primary composite: stroke / MI / death | HR | 0.89 | 0.78–1.01 | 0.0670 |
| Ischaemic stroke | HR | 0.87 | 0.76–1.00 | 0.0462 |
| Net Clinical Outcome | HR | 0.90 | 0.79–1.02 | 0.0928 |
| Ischaemic stroke / MI / CV death | HR | 0.89 | 0.78–1.01 | 0.0771 |
| All-cause death | HR | 1.18 | 0.83–1.67 | 0.3641 |
| CV death | HR | 1.18 | 0.75–1.85 | 0.4828 |
| MI | HR | 1.20 | 0.67–2.14 | 0.5457 |
| Stroke | HR | 0.86 | 0.75–0.99 | 0.0342 |
| Fatal stroke | HR | 1.06 | 0.55–2.06 | 0.8557 |
| Disabling stroke | HR | 0.90 | 0.77–1.06 | 0.2126 |
| Stroke severity / overall disability | OR | 0.93 | 0.85–1.02 | 0.1393 |
This table should be read horizontally, not as a ranking of outcomes. Each endpoint has its own definition, event process, uncertainty, and statistical question. The primary endpoint remains the prespecified composite of stroke, MI, and death, while the other analyses provide additional information about individual or related outcomes.
15. Understanding the Primary Composite
The primary endpoint combines three clinically distinct event types: stroke, MI, and death. Composite endpoints can increase the number of observed events and therefore potentially improve statistical information, but they also require careful interpretation because a treatment effect on the composite does not necessarily imply the same effect on every component.
Composite endpoint
The primary outcome counts the first occurrence of stroke, MI, or death during the defined observation period.
Component endpoints
The registry separately reports analyses for stroke, MI, all-cause death, CV death, fatal stroke, and disabling stroke.
Statistical consequence
A composite HR summarizes the time to the first qualifying component and should not be interpreted as an average effect across the three components.
Interpretive consequence
Examining component results helps explain what contributes to the composite estimate, while recognizing that secondary analyses have their own uncertainty.
16. Safety Analysis Population and Exposure
The safety analysis set consisted of all patients who received at least 1 dose of randomized ticagrelor or ASA and for whom post-dose data were available. This definition differs from the full analysis set used for efficacy.
| Analysis feature | Efficacy | Safety |
|---|---|---|
| Population basis | Randomization | Treatment exposure |
| Registry population | Full analysis set | Safety analysis set |
| Definition | All randomized patients | Patients receiving at least 1 dose with post-dose data available |
| Primary method represented in registry-reported analyses | Cox proportional-hazards model | Cox proportional-hazards model for selected safety endpoints |
This distinction is important when comparing efficacy and safety evidence. A randomized efficacy analysis asks about outcomes according to assigned treatment, while an exposure-based safety analysis asks about outcomes among participants who actually received study treatment.
17. What the Confidence Intervals Tell Us
The confidence intervals reported for SOCRATES vary substantially in width. That variation is informative because it reflects the precision of the corresponding estimates.
| Endpoint | Estimate | 95% CI | Precision observation |
|---|---|---|---|
| Primary composite | HR 0.89 | 0.78–1.01 | Relatively concentrated around the point estimate; interval includes 1.00. |
| Ischaemic stroke | HR 0.87 | 0.76–1.00 | Interval reaches the null value. |
| MI | HR 1.20 | 0.67–2.14 | Wide interval; considerable uncertainty around the point estimate. |
| All-cause death | HR 1.18 | 0.83–1.67 | Interval spans values below and above 1.00. |
| Bleeding-related discontinuation | HR 2.26 | 1.53–3.34 | Interval remains above 1.00. |
A confidence interval should not be interpreted as a range containing a fixed percentage of individual patient effects. Instead, it describes uncertainty around the estimated population-level treatment effect under the statistical model and sampling framework.
18. Limitations
- Composite endpoint: the primary endpoint combines stroke, MI, and death. The composite HR does not establish that the treatment has the same relative effect on each component.
- Multiple secondary analyses: the ClinicalTrials.gov record contains numerous secondary analyses but do not specify a multiplicity-adjustment procedure for the posted results. Secondary P-values therefore require caution.
- Proportional-hazards assumption: Cox interpretation depends on the model's assumptions. The ClinicalTrials.gov record does not provide a formal diagnostic of proportional hazards.
- Time horizon: the primary endpoint is defined from randomization up to 97 days. Longer-term conclusions are outside the registry-reported primary endpoint definition.
- Confidence intervals: several secondary endpoints have wide intervals, particularly MI, fatal stroke, and CV death, indicating substantial uncertainty around their point estimates.
- Analysis-population distinction: efficacy and safety use different analysis populations. Mixing them would change the meaning of the reported estimates.
- Secondary endpoint interpretation: secondary results are useful for understanding the overall evidence pattern but should not automatically be treated as independently confirmatory findings.
- Registry-level information: the ClinicalTrials.gov record does not provide a complete statistical analysis plan, baseline characteristic table, subgroup estimates, Kaplan-Meier coordinates, or detailed missing-data strategy. Those elements are therefore not reconstructed here.
19. Why This Trial Matters Statistically
SOCRATES is a useful teaching example because its registry record brings together randomized trial design, double blinding, time-to-event endpoints, censoring, Cox regression, logistic regression, hazard ratios, odds ratios, intention-to-treat principles, safety analysis populations, and the interpretation of multiple secondary outcomes.
| Concept | How it appears in SOCRATES |
|---|---|
| Randomization | The trial is randomized and uses a parallel design with 13307 participants. |
| Blinding | The trial is double-blind. |
| Intention-to-treat principle | The efficacy population is the full analysis set containing all randomized patients. |
| Time-to-event endpoint | The primary composite is analyzed from randomization through day 97 with censoring specified for participants without an event. |
| Cox model | Used for the primary endpoint and multiple secondary endpoints. |
| Hazard ratio | Used to quantify relative time-to-event effects. |
| Logistic regression | Used for the stroke-severity and overall-disability analysis. |
| Odds ratio | Reported for the logistic regression analysis. |
| Safety population | Based on treatment exposure and post-dose data availability. |
| Composite endpoint | Primary endpoint combines stroke, MI, and death. |
| Multiple endpoints | One primary endpoint and multiple secondary analyses are reported. |
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary Cox analysis estimated an HR of 0.89 with a 95% CI of 0.78–1.01 and P = 0.0670. The interval quantifies uncertainty around the estimate and includes 1.00.
Endpoint interpretation
The primary endpoint is a composite time-to-first-event measure. Its estimate should be understood as a comparison of the modeled event hazard, not as an absolute percentage of patients benefiting.
Secondary evidence
Several secondary endpoints have HR estimates below 1, while others have estimates above 1. Their confidence intervals vary substantially in width and should be considered alongside the point estimates.
Safety interpretation
Serious adverse events were reported descriptively by arm, while selected bleeding endpoints were analyzed using Cox regression in the safety analysis set.
21. A Practical Reading of the Primary Result
A statistically disciplined reading of the primary result starts with the endpoint definition rather than the P-value. The outcome was time from randomization to the first occurrence of stroke, MI, or death, with specified censoring for participants without an event. The analysis therefore uses a survival-analysis framework.
The estimated HR of 0.89 points below 1.00, indicating a lower estimated instantaneous event rate for ticagrelor under the Cox model. The 95% CI of 0.78–1.01 shows that the estimate is uncertain enough that the null value remains within the interval. The P-value of 0.0670 provides the corresponding measure of statistical evidence under the reported two-sided analysis.
These three quantities answer different questions:
| Quantity | Question it addresses | SOCRATES primary result |
|---|---|---|
| Hazard ratio | What is the estimated relative event hazard? | 0.89 |
| 95% confidence interval | How uncertain is the estimated relative effect? | 0.78–1.01 |
| P-value | How much statistical evidence is present against the null under the specified test? | 0.0670 |
Keeping these concepts separate prevents a common statistical error: treating a P-value as though it were a measure of treatment benefit.
22. Endpoint-Specific Statistical Lessons
The primary outcome demonstrates why survival methods are useful when the timing of an event and censoring are important. A simple binary event rate would discard some of that temporal information.
The separate stroke, MI, and death analyses show why a composite endpoint should not be interpreted as though every component has the same treatment effect. Component estimates have their own confidence intervals and P-values.
The registry uses hazard ratios for time-to-event endpoints and an odds ratio for a logistic regression analysis. These measures arise from different statistical models and answer different questions.
The full analysis set and safety analysis set serve different purposes. The distinction should be preserved whenever results are summarized or compared.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: SOCRATES, NCT01994720.
- Linked publication: PubMed record: PMID 35538232.
- Linked publication: PubMed record: PMID 32666831.
- Linked publication: PubMed record: PMID 31296654.
- Linked publication: PubMed record: PMID 31092152.
- Linked publication: PubMed record: PMID 30776996.
Continue through the Clinical Biostats statistical pathway
Explore the underlying survival-analysis, regression, confidence-interval, and clinical-trial concepts connected to this analysis.
26. Record Summary
SOCRATES provides a useful example of randomized clinical-trial analysis in which the primary endpoint is a censored time-to-event composite and the principal treatment effect is expressed as a hazard ratio. The registry reports an HR of 0.89 for the composite of stroke, MI, or death, with a 95% CI of 0.78–1.01 and P = 0.0670. Secondary analyses use Cox regression for several individual and composite outcomes and logistic regression for stroke severity and overall disability.
The statistical story is broader than the primary P-value. The endpoint definition determines the survival-analysis framework; censoring determines how incomplete event histories contribute information; the analysis population determines the treatment comparison; confidence intervals describe uncertainty; and different effect measures answer different statistical questions. Safety results require a separate exposure-based analysis population, and multiple secondary analyses require appropriate caution when interpreting individual P-values.