This page separates reported trial information from statistical interpretation. The registry contains 21 posted outcome measures, but the ClinicalTrials.gov record contains no formal statistical-analysis results.
1. Trial at a Glance
SELECT2 was a completed randomized, parallel-group clinical trial in acute ischemic stroke. The trial enrolled 352 participants and compared endovascular thrombectomy plus medical management with medical management alone. The registered primary endpoint was degree of disability/dependence measured by the modified Rankin Scale at 90 days.
| Feature | SELECT2 |
|---|---|
| Trial name | SELECT2: A Randomized Controlled Trial to Optimize Patient's Selection for Endovascular Treatment in Acute Ischemic Stroke |
| Condition | Acute Ischemic Stroke |
| Phase | NA |
| Status | COMPLETED |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | SINGLE |
| Primary purpose | TREATMENT |
| Enrollment | 352 |
| Interventions | Endovascular Thrombectomy (device); Medical Management (other) |
| Lead sponsor | University Hospitals Cleveland Medical Center |
| Sponsor type | OTHER |
| ClinicalTrials.gov | NCT03876457 |
2. Clinical Question
The statistical question is whether outcomes measured after randomization differ between patients assigned to endovascular thrombectomy plus medical management and patients assigned to medical management. The registered primary outcome focuses specifically on disability or dependence at 90 days using the modified Rankin Scale.
Population
Participants enrolled in SELECT2 with acute ischemic stroke.
Intervention
Endovascular Thrombectomy (device) plus the trial's medical-management component.
Comparator
Medical Management (other).
Primary question
Does randomized assignment to endovascular thrombectomy plus medical management produce a different degree of disability/dependence at 90 days than medical management?
3. Trial Design
Endovascular thrombectomy plus medical management
- Endovascular Thrombectomy (device)
- Medical-management component
Medical management
- Medical Management (other)
4. Trial Timeline
Trial start
The registered start date for SELECT2 was October 11, 2019.
Primary completion
The registered primary completion date was November 20, 2022.
Completed
The ClinicalTrials.gov record identifies SELECT2 as completed and reports results as posted.
5. Primary Endpoint
| Endpoint | Time frame | Registry definition in the ClinicalTrials.gov record |
|---|---|---|
| Degree of Disability/Dependence as Measured by the Modified Rankin Scale (mRS) Score | 90 days | The modified Rankin Scale (mRS) is a commonly used scale for measuring the degree of disability or dependence in the daily activities of people who have suffered a stroke or other causes of neurological disability. The scale runs from 0-6 with "0" being perfect health without symptoms to "6" being death. Score 0: No symptoms Score 1: No significant disability. Able to carry out all usual activities, despite some symptoms. Score 2: Slight disability. Able to look after own affairs without assistance, but unable to carry out all previous activities. Score 3: Moderate disability. Requires some help, but able to walk unassisted. Score 4: Moderately severe disability. Unable to attend to own bodily needs without assistance, and unable to walk unassisted. Score 5: Severe disability. Requires constant nursing care and attention, bedridden, incontinent. Score 6: Dead |
6. Endpoint Measurement: Why the Modified Rankin Scale Is Statistically Interesting
The modified Rankin Scale is especially important statistically because it is an ordinal outcome. The categories have a natural order from less disability to greater disability, but the numerical distance between adjacent categories should not automatically be treated as equal.
That distinction affects the choice of statistical model. Treating an mRS score as though it were an ordinary continuous measurement can impose assumptions about distances between categories that are not inherent in the scale. Conversely, converting the scale immediately into a simple binary endpoint discards some of the ordering information.
Ordinal information
An mRS score retains an ordering of disability levels. A statistical method can potentially use that ordering without assuming that the difference between every pair of adjacent scores is identical.
Clinical meaning
The endpoint is intended to describe disability or dependence in daily activities at a prespecified time point, rather than measuring survival time or a laboratory value.
Why the time frame matters
The registered assessment is at 90 days. A comparison at a different follow-up time would answer a different statistical question.
Why the full distribution matters
For an ordinal endpoint, the distribution across all score categories can contain information that would be lost if the outcome were reduced to only one threshold.
7. Planned Analysis
The ClinicalTrials.gov record reports that results are posted and that 21 outcome measures are available, but the statistical-analysis field contains no formal statistical analyses. Therefore, no treatment-effect estimate, confidence interval, or p-value is reported on this page.
For the registered primary endpoint, the registry measures the modified Rankin Scale score at 90 days. An endpoint of this type would ordinarily be analyzed using a method appropriate for an ordinal outcome, with the exact model determined by the prespecified statistical analysis plan and the assumptions judged appropriate for the observed data.
In a cumulative-logit framework, Y represents the ordered outcome, k represents an mRS threshold, and X represents the treatment indicator. The proportional-odds assumption would need to be evaluated or otherwise justified before interpreting a single common treatment coefficient across thresholds.
This equation is presented as an educational example of a commonly used ordinal-analysis framework, not as a claim that SELECT2 used this particular model. The ClinicalTrials.gov record does not identify the formal statistical model, covariate adjustment, treatment-effect parameter, hypothesis-testing procedure, missing-data strategy, or multiplicity procedure.
8. How an Ordinal Analysis Could Be Interpreted
If a cumulative-odds model were prespecified and its assumptions were satisfied, a treatment-effect estimate would summarize the relative odds of being at or below successive outcome thresholds. The interpretation would depend on exactly how the categories and treatment indicator were coded.
An ordinal treatment effect is different from a mean difference in mRS. It uses the ordered structure of the scale rather than assuming that moving from one score to the next represents the same quantitative change at every point on the scale.
An odds ratio from an ordinal model would not automatically mean that every individual patient has the same probability of improvement. It would summarize a population-level treatment contrast under the specified model.
For a proportional-odds model, the treatment effect is assumed to be consistent across the cumulative thresholds represented by the model. If that assumption is inappropriate, another ordinal-analysis strategy may be preferable.
Even when a single summary measure is used for the formal analysis, displaying the observed distribution of mRS categories can make the clinical meaning of the treatment comparison easier to understand.
9. Statistical Methodology
Randomization
SELECT2 used randomized allocation. Randomization is the foundation of the causal comparison because treatment assignment is determined by the trial's allocation process rather than by a clinician simply selecting treatment for each participant.
In conceptual terms, randomization creates a framework in which differences in outcomes between randomized groups can be attributed to treatment assignment, subject to the usual requirements of valid trial conduct, follow-up, and analysis.
Parallel-group comparison
The registry identifies the design model as parallel. Each participant is assigned within the randomized trial to one of the two intervention strategies, and outcomes are subsequently compared between the groups.
Ordinal outcome analysis
The primary endpoint is an mRS score ranging from 0 to 6 in the registry definition. Because this is ordered categorical information, an ordinal regression framework is a natural class of methods to consider. Other valid approaches can also be used depending on the prespecified analysis plan and the scientific question.
Confidence intervals
A confidence interval would quantify statistical uncertainty around a treatment-effect estimate. The width of the interval would reflect information in the sample and the variability of the estimated effect. A confidence interval is not a prediction interval for individual patients.
Hypothesis testing
If the primary analysis specified a null hypothesis and a corresponding test statistic, the resulting p-value would describe how compatible the observed data are with that null model under the specified assumptions. A p-value would not measure the magnitude or clinical importance of an effect.
Intention-to-treat principle
For a randomized treatment comparison, an intention-to-treat analysis generally preserves the original randomized assignment when analyzing outcomes. The ClinicalTrials.gov record does not specify the formal analysis population used for SELECT2, so this page does not attribute a particular ITT implementation to the trial.
10. Statistical Methods Explained
Why is randomization important?
Randomization provides the basis for a fair comparison between treatment strategies. Before outcomes are observed, assignment is determined by the randomized design rather than by the investigator's preference for a particular treatment. This reduces systematic differences in treatment assignment and supports causal interpretation.
Why is the modified Rankin Scale an ordinal endpoint?
The mRS categories have a meaningful order from less disability to greater disability, but the numerical labels should not automatically be interpreted as equal-interval measurements. That makes ordinal methods particularly relevant when preserving the full scale.
Why not simply compare the average mRS score?
A mean treats the difference between adjacent numerical scores as quantitatively comparable. That assumption is not automatically justified for an ordinal clinical scale. An ordinal model can instead use the ordering of the categories directly.
What does a cumulative-odds model do?
It considers cumulative probabilities across thresholds of the ordered outcome. For example, the model can conceptually distinguish patients at or below one disability threshold from those above it, while simultaneously using information from the other thresholds.
What is the proportional-odds assumption?
In a proportional-odds model, the treatment coefficient is constrained to have the same interpretation across the cumulative thresholds. If the treatment effect varies materially between thresholds, a different model or an alternative summary of the ordinal distribution may be more appropriate.
Why does the 90-day time frame matter?
The registered endpoint is assessed at 90 days. Statistical inference is therefore tied to disability or dependence at that prespecified time point. Moving the assessment to another time would change the estimand and potentially the clinical interpretation.
Why should an empty statistical-analysis field be taken seriously?
Because an observed outcome and a formal treatment comparison are not the same thing.
11. Safety Results
The ClinicalTrials.gov record reports serious adverse events by arm. These are presented as affected participants divided by the number at risk in each arm.
| Safety measure | Endovascular thrombectomy plus medical management | Medical management |
|---|---|---|
| Serious adverse events | 129 / 178 | 123 / 174 |
These figures describe the number affected and the corresponding number at risk as reported in the ClinicalTrials.gov record. They should not be transformed here into a new comparative risk estimate because the task requires trial numbers to be reported exactly as reported in the registry and the registry's formal statistical-analysis field is empty.
12. Results Availability and Interpretation
Results are posted, and the registry reports 21 outcome measures. However, no statistical analyses are in the posted results.
| Evidence item | What the ClinicalTrials.gov record establishes |
|---|---|
| Trial status | COMPLETED |
| Results posted | Yes |
| Outcome measures posted | 21 |
| Registered primary endpoints | 1 |
| Statistical analyses reported | None |
| Primary endpoint | Degree of Disability/Dependence as Measured by the Modified Rankin Scale (mRS) Score |
| Primary endpoint time frame | 90 days |
This distinction is important for reproducible statistical communication. A registry can contain outcome observations without exposing every component needed to reconstruct the formal inferential analysis. A responsible analysis page should preserve that distinction rather than filling the gap with a guessed model or reconstructed p-value.
13. Limitations
- No formal statistical analyses reported: the trial data contain no statistical-analysis entries, so treatment-effect estimates, confidence intervals, and p-values cannot be reported from the ClinicalTrials.gov record.
- Analysis model is unspecified: the ClinicalTrials.gov record identifies the primary endpoint but does not specify the formal statistical model used for it.
- Analysis population is unspecified: the ClinicalTrials.gov record does not identify whether the formal primary analysis used an intention-to-treat, modified intention-to-treat, per-protocol, or another analysis population.
- Missing-data strategy is unspecified: the ClinicalTrials.gov record does not identify how missing 90-day mRS assessments were handled.
- Multiplicity is unspecified: the record does not provide a formal hierarchy or type-I-error allocation across the 21 posted outcome measures.
- Stratification is unspecified: no randomization strata or stratified analysis factors are provided in the ClinicalTrials.gov record.
- Interim analysis is unspecified: the ClinicalTrials.gov record does not report an interim-analysis schedule or alpha-spending procedure.
- Bayesian methods are unspecified: no Bayesian model or Bayesian decision rule is identified in the ClinicalTrials.gov record.
- Serious-adverse-event comparison: the registry supplies affected/at-risk counts, but no formal comparative safety analysis is reported in the statistical-analysis field.
14. Why This Trial Matters Statistically
SELECT2 is a useful statistical teaching case because it combines a randomized parallel-group design with an ordered neurological outcome. The central methodological challenge is not merely calculating a treatment difference; it is choosing an analysis that respects what the modified Rankin Scale actually measures.
| Concept | How it appears in SELECT2 |
|---|---|
| Randomization | The registry identifies allocation as RANDOMIZED. |
| Parallel-group design | The design model is PARALLEL. |
| Single masking | The registry identifies the trial as SINGLE masked. |
| Ordinal outcome | The primary endpoint uses the modified Rankin Scale, with scores from 0-6 in the registry-reported definition. |
| Fixed assessment time | The primary endpoint is assessed at 90 days. |
| Treatment comparison | The two interventions are endovascular thrombectomy plus medical management and medical management. |
| Safety analysis | Serious adverse events are reported by randomized arm as affected/at-risk counts. |
| Registry transparency | Results are posted, but the registry-reported statistical-analysis field contains no formal analyses. |
15. What a Complete Statistical Report Would Normally Need
For a reproducible analysis of the primary mRS endpoint, a complete statistical report would ideally identify the treatment-effect measure, the analysis population, the statistical model, model assumptions, handling of missing assessments, confidence level, hypothesis-testing procedure, and any multiplicity adjustment.
Effect estimate
The report should identify exactly what parameter summarizes the randomized treatment contrast, such as an ordinal odds ratio or another prespecified estimand.
Uncertainty
A confidence interval would show the precision of the estimated treatment effect rather than merely reporting whether a hypothesis test crossed a threshold.
Model assumptions
If an ordinal proportional-odds model were used, the proportional-odds assumption would be an important part of interpretation.
Missing data
Loss of a 90-day mRS assessment can affect an ordinal endpoint directly, making the prespecified missing-data strategy important to reproducibility.
16. Interpreting Safety Counts
The serious-adverse-event data are reported as affected participants divided by participants at risk: 129/178 for endovascular thrombectomy plus medical management and 123/174 for medical management.
Those denominators matter. A safety count without its population at risk can be difficult to interpret because the apparent frequency depends on the number of participants exposed or otherwise eligible for the safety assessment.
For a randomized safety comparison, investigators may additionally report risk differences, risk ratios, incidence rates, or time-to-event safety analyses, depending on the endpoint definition and follow-up structure. None of those comparative measures is reported in the trial's statistical-analysis field here.
17. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial is randomized and parallel-group, and its primary endpoint is an ordered 90-day mRS outcome. The ClinicalTrials.gov record does not report a formal inferential analysis for that endpoint.
Clinical interpretation
The primary clinical outcome concerns disability or dependence at 90 days after acute ischemic stroke. That outcome is distinct from safety outcomes such as serious adverse events.
18. A Practical Analysis Workflow for SELECT2
A statistician approaching the primary endpoint would first confirm the randomized analysis population and the exact mRS coding. The next step would be to identify the prespecified estimand and statistical model, followed by assessment of model assumptions and missing-data handling.
19. Related Tutorials
Learn more about the methods and statistical concepts relevant to this trial:
20. Related Statistical Calculators
21. Sources
- ClinicalTrials.gov: SELECT2 (NCT03876457), the official registry record.
- Linked publication: PubMed record — PMID 41906875.
- Linked publication: PubMed record — PMID 40570276.
- Linked publication: PubMed record — PMID 40151939.
- Linked publication: PubMed record — PMID 39869840.
- Linked publication: PubMed record — PMID 39744837.
Continue through the Clinical Biostats statistical pathway
Explore tutorials on randomized trials, ordinal outcomes, regression, confidence intervals, and related clinical-trial methods.
22. Record Summary
SELECT2 is a randomized, parallel-group trial in acute ischemic stroke with 352 enrolled participants and two intervention arms: endovascular thrombectomy plus medical management and medical management. Its registered primary endpoint is degree of disability/dependence measured by the modified Rankin Scale at 90 days.
The principal statistical teaching point is the distinction between randomized treatment comparison and appropriate analysis of an ordinal outcome. The mRS contains ordered categories, so a formal analysis should respect that structure rather than automatically treating the score as a continuous variable. At the same time, the ClinicalTrials.gov record does not identify the formal statistical model, analysis population, missing-data procedure, multiplicity strategy, or treatment-effect estimate.
The registry reports results as posted and 21 outcome measures, but its registry-reported statistical-analysis array is empty. The ClinicalTrials.gov record shows serious adverse events as 129/178 in the endovascular thrombectomy plus medical-management arm and 123/174 in the medical-management arm; no formal comparative safety analysis is reported.