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
TENSION was a randomized, parallel-group clinical trial evaluating thrombectomy with best medical care versus best medical care for stroke with extended lesion and extended time window. The registry lists 253 enrolled participants and a primary endpoint based on the modified Rankin Scale at 90 days.
| Feature | TENSION |
|---|---|
| Trial acronym | TENSION |
| ClinicalTrials.gov identifier | NCT03094715 |
| Official title | Efficacy and Safety of Thrombectomy in Stroke With Extended Lesion and Extended Time Window |
| Condition | Stroke, Acute; Cerebral Stroke; Cerebrovascular Stroke; Apoplexy; Brain |
| Phase | NA |
| Status | TERMINATED |
| Enrollment | 253 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Single |
| Primary purpose | Treatment |
| Lead sponsor | University Hospital Heidelberg |
| Start date | 2018-06-20 |
| Primary completion date | 2023-05-31 |
2. Clinical Question
The clinical question registered for TENSION is whether thrombectomy with best medical care should be compared with best medical care alone for patients represented by the trial's acute-stroke condition and the extended lesion and extended time-window setting specified in the trial title.
Population
Patients enrolled in the TENSION trial for acute stroke, with the trial focused on an extended lesion and extended time window.
Intervention
Thrombectomy, identified in the registry as a device intervention, together with best medical care.
Comparator
Best medical care, identified in the registry as an other intervention.
Primary question
How does the distribution of modified Rankin Scale outcomes at 90 days differ between the randomized treatment groups?
The wording of the primary endpoint is important. TENSION does not register a primary endpoint as a simple binary event such as death versus survival. Instead, it specifies the modified Rankin Scale (mRS) outcome at 90 days, an ordinal functional-outcome measure. The registry further specifies an ordinal mRS shift analysis based on the difference in linear trends between treatment groups.
3. Trial Design
Thrombectomy
- Thrombectomy is identified as the device intervention.
- Best medical care is part of the comparison framework.
Best medical care
- Best medical care is identified as the other intervention.
- No additional arm-level enrollment count is reported here.
The registry's design information establishes randomization, two parallel arms, single masking, and a treatment purpose. It does not, in the ClinicalTrials.gov record, provide arm-specific enrollment counts, a randomization ratio, or additional design features such as a factorial structure, crossover, or interim-analysis framework.
4. Endpoints
| Endpoint | Time frame | Registry definition / analysis |
|---|---|---|
| Clinical outcome-modified Rankin Scale at 90 days | 90 days | The primary endpoint is the modified Rankin Scale (mRS) outcome at 90 days post-stroke. The primary effectiveness endpoint analysis is a chi-square test of the difference in linear trends in ordinal mRS outcomes at 90 days post-procedure between treatment groups ("mRS shift analysis"). |
Why the endpoint is ordinal
The modified Rankin Scale is used here as an ordered functional-outcome measure. The statistical feature that matters most is not merely that the endpoint has several possible categories, but that those categories have a meaningful order. A patient can therefore be represented as occupying a point along an ordered functional-outcome scale rather than simply being classified into a yes/no outcome.
That structure is the reason the registered analysis is described as an mRS shift analysis. The statistical question is about how the distribution of ordinal outcomes shifts between randomized groups, rather than reducing the endpoint immediately to a single binary cutoff.
5. Primary Endpoint: Statistical Structure
The registry specifies a chi-square test of the difference in linear trends in ordinal mRS outcomes at 90 days post-procedure between the treatment groups. This is the key statistical feature of TENSION's primary analysis.
The analysis compares the pattern of ordered mRS outcomes between randomized treatment groups rather than treating the outcome as an unordered collection of independent categories.
What an mRS shift analysis asks
Suppose the possible mRS categories are ordered from better functional status to worse functional status. A shift analysis asks whether the treatment groups have systematically different distributions across that ordered scale. Evidence of a favorable shift would mean that the treatment-group distribution is displaced toward better functional outcomes relative to the comparator distribution.
This is conceptually different from asking only whether the percentage of patients below one selected threshold differs. A dichotomized endpoint discards information from the rest of the ordinal scale. An ordinal analysis attempts to use the ordering of the outcome categories more fully.
What is compared?
The ordered pattern of 90-day mRS outcomes between the randomized treatment groups.
What is tested?
The registry specifies a chi-square test of the difference in linear trends between groups.
Why preserve ordering?
mRS categories have an ordered clinical interpretation, so the analysis can use more information than a single binary cutoff.
Why 90 days?
The registry specifies 90 days as the time frame for the primary clinical-outcome assessment.
6. Planned Analysis
The registry identifies the modified Rankin Scale outcome at 90 days as the primary endpoint and specifies a chi-square test of the difference in linear trends in ordinal mRS outcomes between treatment groups. This is the formal statistical framework recorded for the primary effectiveness endpoint.
Primary analysis framework
90-day functional outcome · chi-square test of difference in linear trends
For an ordinal outcome of this type, the central statistical task is to compare the distribution of outcomes while respecting their ordering. The analysis should therefore retain the categories of the mRS outcome rather than automatically collapsing them into a binary endpoint.
The registry does not report a numerical primary-endpoint estimate, confidence interval, p-value, or arm-specific mRS outcome distribution in the record represented here. No formal statistical analyses were posted to ClinicalTrials.gov.
7. Statistical Methodology
Randomization
Randomization is the foundation of the comparative design. Because participants are randomly allocated to the two parallel treatment groups, the treatment assignment itself is determined independently of the participant's subsequent outcome. In principle, this allows differences in outcomes between groups to be interpreted as treatment-group differences under the assumptions of the randomized design.
Randomization does not guarantee that every baseline characteristic will be numerically identical between groups. Its statistical role is instead to make treatment assignment independent of prognostic characteristics in expectation and thereby reduce systematic allocation bias.
Ordinal outcome analysis
The primary endpoint is not simply a binary event. The mRS is an ordered outcome, and the registry explicitly preserves that structure through an mRS shift analysis. An ordinal comparison can potentially use information across the full outcome distribution rather than selecting one threshold and discarding the remaining categories.
The statistical analysis uses the fact that the categories have an order. The exact category distribution is needed to quantify the observed shift.
Chi-square test of linear trend
The registry specifies a chi-square test of the difference in linear trends. A chi-square statistic measures how strongly the observed pattern differs from what would be expected under the null hypothesis represented by the test.
For an ordered endpoint, a trend-based chi-square approach differs from an ordinary Pearson chi-square test that treats categories as nominal. The trend formulation incorporates the ordering of categories into the comparison.
Why a shift analysis is appropriate to the endpoint
An mRS shift analysis is naturally aligned with the clinical structure of an ordinal functional-outcome endpoint. If the entire treatment-group distribution moves toward better outcomes, that pattern can be reflected without requiring the analyst to decide in advance that only one particular threshold matters.
There is an important statistical distinction, however: an ordinal analysis describes a distributional shift, while a clinical interpretation asks what that shift means for individual patients. Those are related questions, but they are not identical.
Time-fixed assessment
The primary endpoint is assessed at a specified time point: 90 days. This makes the primary analysis different from a time-to-event analysis such as overall survival. There is no hazard ratio, Kaplan-Meier estimator, or Cox proportional-hazards model specified for the primary endpoint in the registry information represented here.
8. Statistical Methods Explained
Why was an mRS shift analysis used?
The registry defines the primary outcome as the modified Rankin Scale at 90 days and explicitly specifies an mRS shift analysis. Because mRS is ordinal, comparing the distribution across its ordered categories can preserve more of the endpoint's structure than reducing the result to a single binary cutoff.
What does "difference in linear trends" mean?
A linear-trend comparison asks whether there is evidence of a systematic difference across ordered outcome categories between the treatment groups. Instead of treating every category as unrelated to every other category, the analysis uses their ordered structure.
Why use a chi-square statistic?
A chi-square statistic provides a way to test whether the observed pattern is sufficiently different from the pattern expected under the null hypothesis. In TENSION, the registry specifically identifies a chi-square test for the difference in linear trends in the ordinal mRS outcomes.
Why not automatically analyze mRS as a continuous variable?
The numerical labels assigned to ordinal categories indicate order, but they do not necessarily imply equal distances between adjacent categories. Treating an ordinal scale as continuous therefore imposes assumptions about the spacing of categories that an ordinal analysis does not require in the same way.
What would a favorable shift mean?
A favorable shift would mean that the treatment-group distribution is displaced toward better positions on the ordered mRS scale relative to the comparator distribution. The concept concerns the distribution of outcomes across categories, not simply the percentage of patients achieving one selected threshold.
What would a p-value mean for this analysis?
If a formal p-value were reported, it would quantify the evidence against the null hypothesis represented by the specified statistical test. It would not measure the size of the treatment effect, the probability that the treatment works, or the probability that the null hypothesis is true.
9. What the Primary Endpoint Can and Cannot Tell Us
The primary endpoint concerns the distribution of functional outcomes on the mRS at 90 days. A statistical difference would indicate that the treatment groups have different patterns of ordinal outcomes at that time point.
The chi-square test provides evidence about a difference under its specified null hypothesis. It does not, by itself, state how large the treatment-group difference is in a clinically interpretable unit.
For an ordinal endpoint, two treatment groups could have the same percentage at one selected category threshold while still having meaningfully different distributions elsewhere on the scale. A shift analysis is designed to examine the ordered outcome pattern rather than only one cutoff.
A statistical difference in an ordinal functional outcome should be interpreted alongside the actual distribution of mRS categories and the clinical context. Statistical evidence and clinical magnitude are complementary rather than interchangeable.
10. Why the Analysis Is Different From a Survival Trial
TENSION's registered primary endpoint is a 90-day functional outcome, not a time-to-event endpoint. This distinction determines the appropriate statistical framework.
| Feature | TENSION primary endpoint | Time-to-event endpoint |
|---|---|---|
| Outcome structure | Ordinal mRS outcome | Event time |
| Primary time frame | 90 days | Variable follow-up until event or censoring |
| Registered analysis | Chi-square test of difference in linear trends | Not specified for TENSION's primary endpoint |
| Typical effect measure | Depends on the ordinal model or summary used | May include a hazard ratio |
| Kaplan-Meier curve | Not the registered primary analysis | Commonly used for time-to-event distributions |
This distinction is more than terminology. Applying a survival-analysis framework to a fixed-time ordinal endpoint would change the estimand and the assumptions being made. The analysis should follow the structure of the endpoint actually defined in the trial.
11. Analysis Population and Randomization
The registry identifies TENSION as randomized with two parallel arms, but the ClinicalTrials.gov record does not specify a separate analysis-population definition such as an intention-to-treat population, per-protocol population, modified intention-to-treat population, or safety population.
What is established
Allocation was randomized, the design was parallel, and the total enrollment was 253.
What is not specified
The ClinicalTrials.gov record does not identify arm-specific enrollment counts or a separate named analysis population.
For a randomized clinical trial, the population used for the primary analysis is important because exclusions after randomization can alter the protection provided by randomization. The exact analysis population should therefore be distinguished from the enrollment population whenever the registry specifies it.
12. Missing Data and Ordinal Outcomes
The primary endpoint is measured at 90 days, so completeness of the 90-day mRS assessment is an important statistical consideration. Participants without a valid primary endpoint assessment create a missing-data problem that can affect the estimated treatment-group distributions.
The ClinicalTrials.gov record does not specify a missing-data or imputation method for the primary endpoint. Several approaches are possible in clinical-trial analysis, but they make different assumptions about why observations are missing. The appropriate method depends on the prespecified statistical analysis plan and the observed pattern of missingness.
13. Multiplicity and Interim Analysis
The ClinicalTrials.gov record identifies one registered primary endpoint: clinical outcome measured by the modified Rankin Scale at 90 days. It specifies the primary statistical test but does not report an interim-analysis framework, alpha-spending procedure, multiplicity strategy, or hierarchy of additional confirmatory endpoints.
| Design topic | What the registry information establishes |
|---|---|
| Primary endpoint | mRS outcome at 90 days |
| Primary statistical test | Chi-square test of the difference in linear trends |
| Multiplicity strategy | Not reported in the ClinicalTrials.gov record |
| Interim analysis | Not reported in the ClinicalTrials.gov record |
| Alpha-spending method | Not reported in the ClinicalTrials.gov record |
These distinctions matter because repeated looks at accumulating data or multiple confirmatory hypotheses can change the interpretation of nominal statistical significance. Without a documented multiplicity or interim framework, those features should not be inferred from the existence of a randomized trial alone.
14. Safety and Adverse Events
The registry information identifies thrombectomy and best medical care as the two interventions but does not provide arm-specific serious adverse-event counts or other numerical safety results in the information represented here.
| Safety item | Registry information represented here |
|---|---|
| Serious adverse events by arm | Not reported |
| Arm-specific adverse-event counts | Not reported |
| Numerical safety estimates | Not reported |
Safety is analytically distinct from the primary functional-outcome endpoint. A treatment comparison can be evaluated across efficacy and safety outcomes separately, with each endpoint requiring its own definition, analysis population, and interpretation.
15. Trial Timeline
Trial start
The registry lists June 20, 2018 as the start date for TENSION.
Randomized parallel-group trial
The registry lists an enrollment of 253 participants across two randomized parallel treatment arms.
Primary completion
The registry lists May 31, 2023 as the primary completion date.
Current registry status
The registry identifies the study status as terminated.
16. Interpreting the 90-Day mRS Endpoint
The modified Rankin Scale is clinically attractive for stroke trials because functional recovery is multidimensional and can be represented along an ordered scale. From a statistical perspective, however, that same feature creates a choice: should the scale be collapsed into a binary endpoint, analyzed as an ordinal endpoint, or modeled in some other way?
TENSION's registry entry answers that question at the level of the primary analysis: it specifies an mRS shift analysis and a chi-square test of the difference in linear trends. That means the registered analysis is explicitly oriented toward the ordered distribution of outcomes.
Information retained
An ordinal analysis can use information across multiple ordered mRS categories instead of retaining only one threshold.
Information lost by dichotomization
Turning an ordinal scale into a binary outcome can make several clinically distinct categories appear identical for the statistical analysis.
Interpretation challenge
A distributional shift must ultimately be translated into a clinical understanding of functional outcomes.
Statistical challenge
The validity of the chosen test depends on the assumptions and implementation specified in the statistical analysis plan.
17. What a Complete Result Would Normally Contain
For a primary ordinal endpoint such as the one registered for TENSION, a statistically informative results presentation would normally show the observed mRS distribution in each randomized treatment group together with the prespecified test result and an appropriate measure of uncertainty or treatment-effect magnitude, where specified by the analysis plan.
| Result component | Statistical purpose |
|---|---|
| mRS distribution by treatment group | Shows where the observed functional outcomes fall across the ordered scale. |
| Primary test statistic | Quantifies evidence against the null hypothesis represented by the registered test. |
| P-value | Quantifies statistical evidence under the specified test and assumptions. |
| Effect estimate | Provides a measure of the magnitude and direction of the treatment-group difference when defined by the analysis. |
| Confidence interval | Describes statistical uncertainty around an estimated treatment effect when an appropriate interval is reported. |
These components answer different questions. The distribution describes the actual pattern of functional outcomes; the test statistic and p-value address statistical evidence; and an effect estimate with its confidence interval addresses magnitude and precision. No single component substitutes for the others.
18. Why This Trial Matters Statistically
TENSION is a useful statistical teaching case because its primary endpoint illustrates a central problem in clinical-trial analysis: the most clinically meaningful outcome is not always naturally binary or continuous. Functional outcomes can be ordinal, and the statistical analysis needs to respect that structure.
| Concept | How it appears in TENSION |
|---|---|
| Randomization | Participants were allocated randomly to treatment groups. |
| Parallel-group design | The trial has two parallel treatment arms. |
| Ordinal endpoint | The primary endpoint is the modified Rankin Scale at 90 days. |
| Shift analysis | The registry specifies an mRS shift analysis. |
| Trend testing | The primary analysis is a chi-square test of the difference in linear trends. |
| Fixed-time assessment | The primary endpoint is evaluated at 90 days rather than as a time-to-event endpoint. |
| Statistical interpretation | The analysis concerns the distribution of functional outcomes, not merely a single event rate. |
19. Limitations
- No numerical primary results in the registry information represented here: the registry specifies the endpoint and primary analysis but does not provide a numerical estimate, confidence interval, or p-value for that analysis.
- No arm-specific enrollment counts: the total enrollment is 253, but the ClinicalTrials.gov record does not provide the number randomized to each arm.
- No observed mRS distribution: the category-level 90-day functional outcomes are not reported in the ClinicalTrials.gov record.
- Missing-data strategy not specified: the registry information represented here does not describe how missing 90-day mRS assessments are handled.
- Multiplicity not specified: the ClinicalTrials.gov record does not identify a multiplicity adjustment or hierarchy beyond the registered primary endpoint.
- Interim monitoring not specified: no interim-analysis or alpha-spending framework is reported in the ClinicalTrials.gov record.
- Safety results not numerically reported: arm-specific serious adverse-event counts and other numerical safety results are not provided here.
- Analysis-population details not specified: the registry information represented here does not name a separate intention-to-treat, per-protocol, or safety analysis population.
These limitations affect what can be concluded statistically from the registry record. They do not change the registered endpoint or its planned analysis; they determine how much of the eventual statistical result can be independently described from the information available in the record.
20. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary analysis is designed to test whether the ordered 90-day mRS outcome distribution differs between the randomized treatment groups using a chi-square test of the difference in linear trends.
Clinical interpretation
The clinical meaning lies in whether the observed distribution of functional outcomes represents a meaningful difference in recovery and disability at 90 days.
The distinction is important. Statistical analysis determines whether an observed pattern is sufficiently inconsistent with a specified null hypothesis. Clinical interpretation asks what that pattern means for patients. A statistically detectable shift does not automatically tell us whether the magnitude is large enough to be clinically important, while a clinically meaningful difference may require a sufficiently precise estimate before its statistical evidence can be evaluated.
21. How to Read a Future mRS Shift Result
First determine which direction on the mRS represents better functional outcome and whether the observed distribution shifts toward that direction in the thrombectomy group.
Next examine the actual distributions and any prespecified effect measure. A p-value alone cannot establish how large the difference is.
Where a confidence interval is reported for an effect estimate, examine its width. A narrow interval indicates greater statistical precision than a wide interval, although precision and clinical importance remain separate concepts.
Finally, consider how the observed distribution translates into functional outcomes at 90 days. Statistical evidence should be interpreted together with the size and shape of the observed shift.
22. Important Statistical Distinctions
Ordinal versus binary analysis
A binary analysis asks whether an outcome falls above or below one threshold. An ordinal analysis asks whether the entire ordered distribution differs. These approaches answer related but different questions.
Statistical significance versus effect size
A test statistic or p-value addresses evidence against a null hypothesis. It does not directly quantify the size of a treatment effect. Effect estimates and confidence intervals are needed for that purpose when the analysis defines an appropriate effect measure.
Association versus randomized treatment comparison
Because TENSION is randomized, the primary comparison is anchored to treatment assignment rather than simply comparing naturally occurring groups. That is a major distinction from an observational comparison.
Fixed-time outcome versus time-to-event outcome
The primary endpoint is evaluated at 90 days. This differs from an endpoint such as time to death, for which censoring and the timing of events become central components of the analysis.
23. Sources
- ClinicalTrials.gov: TENSION — NCT03094715.
Continue through Clinical Biostats
Explore additional clinical-trial analyses, statistical tutorials, and statistical calculators.
24. Record Summary
TENSION is a randomized, parallel-group trial with 253 enrolled participants evaluating thrombectomy with best medical care versus best medical care in acute stroke with an extended lesion and extended time window. Its registered primary endpoint is the modified Rankin Scale outcome at 90 days, and the registry specifies a chi-square test of the difference in linear trends in ordinal mRS outcomes between treatment groups as the primary effectiveness analysis.
The statistical importance of the trial design lies in the alignment between the clinical endpoint and its analysis. The mRS is ordinal, so the registered shift analysis retains the ordering of functional outcomes rather than automatically reducing the endpoint to a binary threshold. Correct interpretation therefore requires attention to the complete outcome distribution, the statistical test, the magnitude of any estimated treatment effect, and the precision of that estimate.
The registry identifies the study as terminated, with a start date of June 20, 2018 and a primary completion date of May 31, 2023. The numerical primary-endpoint results, arm-specific mRS distributions, confidence intervals, p-values, and numerical safety results are not reported in the registry information represented here. The methodological record nevertheless provides a clear example of how randomized treatment comparisons can be structured around an ordinal functional outcome.