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
SPS3 was a phase 3 randomized, double-blind, parallel-group treatment trial focused on the secondary prevention of small subcortical strokes. The registry identifies cerebrovascular accident and hypertension as the trial conditions and records an enrollment of 3020 participants.
| Feature | SPS3 |
|---|---|
| Trial acronym | SPS3 |
| Full title | Secondary Prevention of Small Subcortical Strokes Trial |
| Phase | Phase 3 |
| Status | Completed |
| Study start | 2003-02 |
| Primary completion | 2012-04 |
| Lead sponsor | University of British Columbia |
| Sponsor type | OTHER |
| Conditions | Cerebrovascular Accident; Hypertension |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 3020 |
| Number of arms | 2 |
2. Clinical Question
The registry identifies the study as the Secondary Prevention of Small Subcortical Strokes Trial, with cerebrovascular accident and hypertension listed as conditions. The primary statistical question is therefore framed around the randomized comparison of the trial interventions for the prevention of subsequent clinically defined stroke outcomes during follow-up.
Population
The registry identifies the study population through the trial title and lists cerebrovascular accident and hypertension as the conditions under study.
Interventions
The registry lists aspirin, clopidogrel, Target of Blood Pressure, and placebo among the trial interventions.
Comparator structure
The registry records a randomized two-arm design and lists placebo among the interventions. The listed intervention field does not provide a complete arm-by-arm description of how each component was combined.
Primary question
Does the randomized treatment strategy affect the occurrence of clinically defined ischemic stroke or hemorrhagic stroke during the registered follow-up period?
3. Trial Design
Intervention structure recorded by the registry
Drug interventions
- Aspirin
- Clopidogrel
Other interventions
- Target of Blood Pressure
- Placebo
The randomized parallel structure is statistically important because the treatment comparison is anchored to the randomized assignment. Double masking is also relevant to the design because knowledge of treatment assignment can influence behavior, assessment, follow-up, or other aspects of trial conduct.
4. Endpoints
The registry lists two primary endpoints. Both are stroke outcomes, but they represent different clinical event types and use distinct definitions.
| Primary endpoint | Registry definition | Time frame |
|---|---|---|
| Clinically defined ischemic stroke | Evidence of clinically defined ischemic stroke (focal neurological deficits persisting for more than 24 hours) confirmed by non-investigational CT or MRI | Mean follow up of 4 years |
| Hemorrhagic stroke | Evidence of hemorrhagic stroke; a neurologic deficit associated with intraparenchymal or subarachnoid space lesion on CT/MRI or cerebral hemorrhage demonstrated by surgery or autopsy. | within mean follow-up of 4 years |
Why the endpoint definitions matter
The ischemic-stroke endpoint requires both a clinical criterion and confirmation by non-investigational CT or MRI. The clinical component is explicitly based on focal neurological deficits persisting for more than 24 hours.
The hemorrhagic-stroke endpoint is defined differently. It includes a neurologic deficit associated with an intraparenchymal or subarachnoid-space lesion on CT/MRI, or cerebral hemorrhage demonstrated by surgery or autopsy.
These definitions illustrate a central principle of clinical-trial statistics: the statistical analysis begins with an operational definition of an event. A treatment comparison is only as interpretable as the outcome definition, ascertainment process, and follow-up framework used to classify participants as having or not having that event.
5. Statistical Methodology
The registry specifies the outcomes and their follow-up timeframes but does not report posted statistical analyses for these endpoints. For a stroke endpoint observed during follow-up, the natural statistical framework is a time-to-event analysis, because participants can experience an event at different times and some participants may complete follow-up without experiencing the event.
Time-to-event analysis
A time-to-event analysis uses both whether an event occurred and when it occurred. This is different from simply calculating the proportion of participants who experienced stroke, because a participant followed for a shorter period contributes less observation time than a participant followed for the full study period.
Here, S(t) represents the probability that the event time T exceeds time t. In a clinical trial, the estimated survival function can describe the probability of remaining free of the specified event through follow-up.
Kaplan-Meier estimation
Kaplan-Meier estimation is commonly used to summarize time-to-event outcomes. It accommodates right censoring, allowing participants who have not experienced the endpoint by their last observed follow-up to contribute information up to that point.
For SPS3, this framework is relevant because the registered primary endpoints are defined as events occurring during follow-up rather than as measurements taken at one fixed visit.
Comparing randomized groups
A randomized two-arm time-to-event comparison is commonly evaluated with a log-rank test and summarized using a hazard ratio from a Cox proportional-hazards model. These methods use the ordering and timing of observed events rather than treating every participant as if they had identical follow-up.
A hazard ratio compares the estimated instantaneous event rates between two groups within a time-to-event model. An HR below 1 indicates a lower estimated instantaneous event rate in the numerator group, while an HR above 1 indicates a higher estimated instantaneous event rate.
Separate analysis of the two primary endpoints
The two registered primary endpoints represent different types of stroke. They should therefore be treated as distinct statistical outcomes unless a prespecified composite or hierarchical strategy is documented. The registry record identifies both endpoints but does not report a multiplicity strategy or endpoint hierarchy.
Event confirmation and statistical validity
Endpoint confirmation is part of the statistical design, not merely a clinical detail. For ischemic stroke, confirmation by non-investigational CT or MRI is part of the registered definition. For hemorrhagic stroke, imaging, surgery, or autopsy can provide the evidence specified by the definition.
6. Statistical Methods Explained
Why is a time-to-event method appropriate for stroke?
The registered endpoints are events occurring during follow-up. Participants may experience an event at different times, while others may remain event-free through the end of observation. Time-to-event methods preserve information about when events occur and accommodate censoring.
What does Kaplan-Meier estimation add?
Kaplan-Meier estimation provides an estimate of the probability of remaining free of an event over time. It is particularly useful when follow-up duration varies or when not every participant experiences the event during observation.
What does a hazard ratio describe?
A hazard ratio is a relative measure comparing instantaneous event rates between randomized groups within a specified time-to-event model. It is not itself an absolute risk difference, a probability of stroke, or a statement that every participant experiences the same relative change.
Why does the timing of stroke matter statistically?
Two groups could have the same eventual proportion of participants experiencing an event but differ substantially in when those events occur. A time-to-event analysis can distinguish those patterns because event timing contributes directly to the analysis.
Why does censoring matter?
A participant who remains free of stroke until the last observed follow-up does not necessarily have an event-free outcome for all future time. Censoring records the period during which that participant was observed without the event. Valid time-to-event analysis depends on appropriate handling and interpretation of censoring.
Why are the two stroke endpoints analytically distinct?
Ischemic stroke and hemorrhagic stroke have different registered clinical definitions. Treating them as separate outcomes preserves the distinction specified by the registry. Combining them into a single endpoint would require an explicit composite-endpoint rationale and prespecified analysis plan.
7. Planned Analysis
The registry identifies two primary outcomes to be measured during approximately four years of mean follow-up: clinically defined ischemic stroke and hemorrhagic stroke. No formal statistical analyses are posted to ClinicalTrials.gov for these endpoints.
| Primary endpoint | What is measured | Typical statistical approach |
|---|---|---|
| Clinically defined ischemic stroke | Focal neurological deficits persisting for more than 24 hours, confirmed by non-investigational CT or MRI | Time-to-event methods such as Kaplan-Meier estimation, a log-rank comparison, and a Cox model are typical approaches for randomized follow-up data |
| Hemorrhagic stroke | Neurologic deficit associated with intraparenchymal or subarachnoid space lesion on CT/MRI, or cerebral hemorrhage demonstrated by surgery or autopsy | Time-to-event methods are typically appropriate when the occurrence and timing of hemorrhagic stroke are observed during follow-up |
The statistical estimand for a conventional time-to-event comparison would describe the difference in event experience between the randomized groups over follow-up. Kaplan-Meier curves can show the event-free probability over time, while a hazard ratio can summarize the relative event rate under a Cox model.
The key statistical feature of SPS3 is that the primary outcomes are clinical events occurring over time, rather than continuous measurements taken at a single time point. This makes survival-analysis concepts such as censoring, risk sets, event timing, and hazard rates central to understanding an appropriate analysis.
What a complete analysis would need to establish
- The number and timing of ischemic stroke events in each randomized group.
- The number and timing of hemorrhagic stroke events in each randomized group.
- The censoring and follow-up experience of participants who did not experience the relevant endpoint.
- The prespecified statistical model and testing framework used for the two primary endpoints.
- Any multiplicity adjustment or endpoint hierarchy governing simultaneous inference for the primary outcomes.
The registry record identifies the outcomes and their definitions but does not report these additional statistical-analysis details.
8. Interpreting the Registered Follow-Up
The ischemic-stroke endpoint has a registered time frame of mean follow up of 4 years. The hemorrhagic-stroke endpoint is registered as occurring within mean follow-up of 4 years.
Mean follow-up is not a median
A mean follow-up duration summarizes follow-up across participants. It does not describe the midpoint of the follow-up distribution.
Follow-up is not an event rate
A four-year mean follow-up describes observation time. It does not indicate how many participants experienced stroke.
Event timing remains important
Participants can experience stroke at different times during follow-up, which is why time-to-event methods can be informative.
Endpoint definition controls classification
The registered clinical and imaging criteria determine which neurological events qualify as the primary endpoints.
9. Randomization and Causal Interpretation
Randomization is one of the central design features of SPS3. With randomized allocation, treatment groups are created by an assignment mechanism rather than by allowing participants or investigators to choose their treatment.
The value of randomization is primarily causal: differences in outcomes between randomized groups can be interpreted as treatment-group differences under the trial's design assumptions. This is distinct from an observational comparison, where treatment selection may be associated with prognostic characteristics.
For time-to-event outcomes, the corresponding treatment contrast may be expressed using survival probabilities, cumulative event probabilities, event-time summaries, or model-based measures such as a hazard ratio.
Randomization does not eliminate every source of uncertainty. Statistical inference remains necessary because the observed trial participants constitute one realization of the randomized experiment. The confidence interval, when reported for an effect estimate, quantifies uncertainty under the relevant statistical model and sampling framework.
10. Double Masking and Outcome Assessment
The registry classifies SPS3 as double masked. Masking is a design feature intended to reduce the influence of treatment knowledge on aspects of trial conduct and assessment.
For an outcome such as ischemic or hemorrhagic stroke, objective diagnostic evidence is particularly important. The registered ischemic-stroke definition explicitly requires confirmation by non-investigational CT or MRI, while the hemorrhagic-stroke definition specifies CT/MRI, surgery, or autopsy as possible sources of confirmation.
11. Two Primary Endpoints and Multiplicity
SPS3 has two registered primary endpoints: clinically defined ischemic stroke and hemorrhagic stroke. Whenever a trial has multiple primary endpoints, statistical interpretation depends not only on the individual endpoint analyses but also on how the collection of primary hypotheses is handled.
| Issue | Statistical implication |
|---|---|
| Multiple primary endpoints | Two distinct primary hypotheses are registered, so the overall confirmatory testing framework matters. |
| Endpoint definitions | Each endpoint has its own clinical and diagnostic definition and should be analysed according to that definition. |
| Multiplicity control | The registry record does not report a multiplicity adjustment or testing hierarchy for the two endpoints. |
| Interpretation | A formal claim about the overall primary objective requires knowing how simultaneous inference was specified. |
Multiplicity is not a reason to discard an individual endpoint result. Rather, it determines how confidently multiple statistical tests can be interpreted together. If two primary hypotheses are tested independently at the same nominal significance level, the probability of at least one false-positive conclusion can differ from the nominal level for a single test.
12. Statistical Interpretation of Stroke-Free Survival
For a time-to-event endpoint, an event-free survival curve can be interpreted as the estimated probability of remaining free from the specified stroke through a given time.
If an estimated curve for ischemic stroke is higher in one randomized group, that group has a higher estimated probability of remaining free from the registered ischemic-stroke endpoint at the corresponding times, subject to the assumptions of the analysis.
An event-free survival estimate is not the probability that a particular individual will definitely remain stroke-free in the future. It is a population-level estimate based on the observed trial follow-up and statistical model.
Participants who are censored contribute information through their last observed time. They are not automatically treated as having experienced the endpoint, nor does censoring establish that they will remain event-free indefinitely.
13. Hazard Ratios: A Statistical Framework for Time-to-Event Outcomes
If a Cox proportional-hazards model is used for the registered stroke endpoints, the hazard ratio would compare the modeled instantaneous event rates between randomized groups.
The hazard ratio is a relative measure. It does not directly report the absolute probability of stroke, the number needed to treat, or the percentage of participants who avoid stroke.
Why the proportional-hazards assumption matters
A conventional Cox hazard ratio has a particularly straightforward interpretation when the relative hazards are reasonably stable over time. If hazards change substantially relative to one another, a single hazard ratio can become less descriptive of the complete event-time pattern.
This is why a time-to-event analysis is often best understood through several complementary quantities: the event-free survival curve, clinically meaningful time-specific event probabilities, numbers of events, and a model-based relative effect estimate.
14. What the Registry Does and Does Not Establish Statistically
| Feature | Registered information |
|---|---|
| Randomized allocation | Yes |
| Parallel design | Yes |
| Double masking | Yes |
| Phase | Phase 3 |
| Primary purpose | Treatment |
| Enrollment | 3020 |
| Primary endpoints | Clinically defined ischemic stroke and hemorrhagic stroke |
| Endpoint follow-up | Mean follow-up of 4 years / within mean follow-up of 4 years |
| Posted statistical analyses | None |
| Reported effect estimates | None in the posted statistical analyses |
| Reported confidence intervals | None in the posted statistical analyses |
| Reported p-values | None in the posted statistical analyses |
| Reported subgroup results | None in the posted statistical analyses |
This distinction is important for statistical reading. A registry can establish the design, intervention categories, and endpoint definitions without providing the numerical outcome analysis required to quantify a treatment effect.
15. Limitations and Interpretation Issues
- No posted statistical analyses: the ClinicalTrials.gov record contains the registered primary endpoints but no posted statistical analyses with effect estimates, confidence intervals, or p-values.
- Multiplicity information: two primary endpoints are registered, but the record does not specify the multiplicity-control strategy or hierarchy used to interpret them jointly.
- Arm-level treatment detail: the intervention list identifies aspirin, clopidogrel, Target of Blood Pressure, and placebo, but the listed record does not provide a complete arm-by-arm mapping of these interventions.
- Follow-up summary: the endpoint time frames specify mean follow-up of 4 years, but a mean follow-up duration does not provide the complete distribution of participant observation times.
- Event-time information: the registry record does not provide the individual event timing or censoring information needed to reconstruct time-to-event curves.
- Endpoint-specific interpretation: ischemic and hemorrhagic stroke are separately defined outcomes and should not be silently combined into an unregistered composite.
- Model assumptions: if a Cox proportional-hazards model is used, the proportional-hazards assumption and other model diagnostics are relevant to interpretation.
- Generalizability: the registry information identifies the conditions and trial design but does not by itself provide the detailed participant characteristics needed to determine how broadly the findings apply to other populations.
16. Why This Trial Matters Statistically
SPS3 provides a useful statistical teaching case because it combines randomized treatment allocation with clinically meaningful stroke endpoints that occur over follow-up. The design naturally leads to questions about event definitions, censoring, time-to-event analysis, and simultaneous interpretation of multiple primary outcomes.
| Concept | How it appears in SPS3 |
|---|---|
| Randomization | The registry identifies randomized allocation. |
| Parallel design | The trial is registered with a parallel design model and two arms. |
| Double masking | The registry identifies the study as double masked. |
| Time-to-event outcomes | Both primary endpoints are stroke events occurring during follow-up. |
| Event definition | Each primary endpoint has a specific clinical and diagnostic definition. |
| Kaplan-Meier methods | A standard framework for estimating event-free survival over follow-up. |
| Cox regression | A standard model for estimating relative event rates over time. |
| Censoring | Relevant when participants complete observation without experiencing the endpoint or have incomplete event follow-up. |
| Multiple primary endpoints | Both ischemic stroke and hemorrhagic stroke are registered as primary endpoints. |
| Multiplicity | The joint interpretation of two primary hypotheses requires an explicit error-control strategy. |
17. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial is structured to compare randomized treatment groups on two defined stroke outcomes observed during follow-up. Time-to-event methods are a natural analytical framework because both endpoints are events occurring over time.
Clinical interpretation
The clinical meaning of the trial depends on whether and how the randomized strategies change the occurrence of the two defined stroke outcomes. That interpretation requires the corresponding numerical outcome analyses.
Keeping these two levels separate is important. Statistical methodology describes how a treatment contrast is estimated and tested; clinical interpretation asks what the resulting evidence means for the population represented by the trial.
18. Understanding the Two Stroke Outcomes
Ischemic stroke
The registered definition requires focal neurological deficits persisting for more than 24 hours and confirmation by non-investigational CT or MRI.
Hemorrhagic stroke
The registered definition includes a neurologic deficit associated with an intraparenchymal or subarachnoid space lesion on CT/MRI or cerebral hemorrhage demonstrated by surgery or autopsy.
The distinction is statistically consequential because changing an endpoint definition changes which observations are classified as events. A treatment comparison is therefore tied to the precise endpoint definition specified before or during the trial's registration and conduct.
19. Trial History
Study begins
The registry records the SPS3 study start in February 2003.
Randomized clinical-trial design
The study is registered as a phase 3 randomized, double-masked, parallel trial with two arms and a treatment purpose.
Enrollment
The registry records enrollment of 3020 participants.
Primary completion
The registry records primary completion in April 2012.
20. Statistical Reading Guide for SPS3
A useful way to read the SPS3 design is to move from the clinical endpoint definition to the statistical estimand and then to the randomized comparison.
This sequence helps prevent a common statistical error: starting with a p-value or hazard ratio before establishing exactly what the event represents and how follow-up was handled. The endpoint definition is the foundation of the analysis.
21. Important Distinctions in Time-to-Event Analysis
| Quantity | What it tells you | What it does not tell you |
|---|---|---|
| Event count | How many participants experienced the specified event | When those events occurred or how long event-free participants were observed |
| Event-free survival probability | Estimated probability of remaining free of the event through a specified time | The future outcome of a particular individual |
| Hazard ratio | Relative modeled instantaneous event rate between groups | An absolute risk difference or individual-level treatment effect |
| Confidence interval | Uncertainty around an estimated effect under the statistical framework | The distribution of treatment effects across individual patients |
| p-value | Evidence against a specified null hypothesis under the test's assumptions | The size or clinical importance of an effect |
These distinctions are particularly important for SPS3 because its primary outcomes are clinical events observed over time. A complete statistical report would ordinarily bring these complementary perspectives together rather than relying on a single summary statistic.
22. Sources
- ClinicalTrials.gov: SPS3 — NCT00059306.
Continue through the Clinical Biostats clinical-trial library
Clinical trial analyses connect study design, endpoint definitions, statistical models, uncertainty, and interpretation into a single statistical framework.
23. Record Summary
SPS3 is a phase 3 randomized, double-masked, parallel clinical trial with 3020 enrolled participants and two randomized arms. The registry identifies cerebrovascular accident and hypertension as conditions and lists aspirin, clopidogrel, Target of Blood Pressure, and placebo among the interventions. Its two primary endpoints are clinically defined ischemic stroke, requiring focal neurological deficits persisting for more than 24 hours and confirmation by non-investigational CT or MRI, and hemorrhagic stroke defined through specified neurologic, imaging, surgical, or autopsy evidence. Both endpoints are evaluated over a mean follow-up of approximately four years.
Statistically, the trial is a useful example of how randomized clinical-trial design intersects with time-to-event analysis. The most important concepts are precise event definition, observation time, censoring, event-free survival, hazard-based comparisons, uncertainty, and the interpretation of multiple primary endpoints. The registry records the design and endpoint framework but has no posted statistical analyses for these outcomes.