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
ANGEL-ASPECT is a completed randomized, parallel-group clinical trial comparing endovascular therapy with best medical management in patients with acute ischemic stroke. The registry identifies neurofunctional deficit, defined using the modified Rankin Scale (mRS), as the primary endpoint at 90±7 days after randomization.
| Feature | ANGEL-ASPECT |
|---|---|
| Trial acronym | ANGEL-ASPECT |
| ClinicalTrials.gov identifier | NCT04551664 |
| Condition | Acute Ischemic Stroke |
| Study title | Study of Endovascular Therapy in Acute Anterior Circulation Large Vessel Occlusive Patients With a Large Infarct Core |
| Status | Completed |
| Enrollment | 456 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Single |
| Number of arms | 2 |
| Primary purpose | Health Services Research |
| Lead sponsor | Beijing Tiantan Hospital |
2. Clinical Question
The clinical question is whether endovascular therapy, compared with best medical management, affects neurofunctional deficit in patients represented by the trial population described in the study title: acute anterior circulation large vessel occlusive patients with a large infarct core.
Population
Patients with acute ischemic stroke described by the registry study title as having acute anterior circulation large vessel occlusion with a large infarct core.
Intervention
Endovascular therapy, classified in the registry as a procedure.
Comparator
Best medical management, classified in the registry as a drug intervention.
Primary question
What is the neurofunctional outcome, as defined by the modified Rankin Scale, at 90±7 days after randomization?
3. Trial Design
Endovascular therapy
- Intervention type: procedure
- Randomized as one of the two parallel treatment arms
Best medical management
- Intervention type: drug
- Randomized as the comparator arm
Because the registry information summarized here does not provide arm-specific enrollment counts, the two arms should not be assigned numerical sample sizes beyond the overall enrollment of 456.
4. Endpoints
Primary endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Neurofunctional deficit defined as modified Rankin Scale (mRS) | 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. | 90±7 days after randomization |
Understanding the mRS scale
The mRS is an ordered outcome. Its values have a natural progression from 0 through 6, but the distance between adjacent scores should not automatically be treated as numerically equal. That distinction is central to statistical analysis of stroke disability outcomes.
| mRS score | Registry description | Statistical meaning |
|---|---|---|
| 0 | No symptoms | Lowest disability category |
| 1 | No significant disability; able to carry out all usual activities, despite some symptoms | Ordered category |
| 2 | Slight disability; able to look after own affairs without assistance, but unable to carry out all previous activities | Ordered category |
| 3 | Moderate disability; requires some help, but able to walk unassisted | Ordered category |
| 4 | Moderately severe disability; unable to attend to own bodily needs without assistance, and unable to walk unassisted | Ordered category |
| 5 | Severe disability; requires constant nursing care and attention, bedridden, incontinent | Ordered category |
| 6 | Dead | Highest disability category |
5. Statistical Methodology
The primary endpoint is an mRS outcome measured at a prespecified follow-up window rather than a time-to-event endpoint. That distinction determines the statistical framework that would ordinarily be considered for the primary comparison.
Ordinal outcome analysis
Because mRS takes ordered values from 0 through 6, an analysis that preserves the ordering can be statistically attractive. A common approach for an ordinal endpoint is an ordinal logistic regression model, often called a proportional-odds model when the proportional-odds assumption is imposed.
Here, Y represents the ordered mRS outcome, k indexes the cumulative cut points, and X represents treatment assignment. The treatment coefficient describes a common shift across the cumulative outcome thresholds when the proportional-odds assumption holds.
The registry record summarized here does not post a formal statistical analysis for the primary endpoint. It therefore does not establish which specific model, effect measure, covariate structure, or hypothesis-testing framework was used for the primary mRS comparison.
Alternative binary analysis
Stroke trials sometimes define a clinically relevant mRS threshold and compare the proportion of participants at or below that threshold. Such an analysis changes the statistical question: instead of using the full ordered outcome, it asks whether the probability of falling into a specified favorable category differs between randomized groups.
The distinction matters because dichotomizing an ordinal endpoint discards information. Two patients whose mRS values lie on opposite sides of a chosen threshold are treated as different, while patients with different mRS values on the same side of the threshold are treated as equivalent.
Why the choice of analysis matters
An ordinal analysis and a binary analysis do not estimate exactly the same quantity. An ordinal model uses the complete ordering of the mRS categories, whereas a binary model focuses on one particular cutoff. Consequently, the treatment effect and its interpretation depend on the prespecified statistical estimand.
Preserve ordering
An ordinal method can use information from all seven mRS categories rather than reducing the endpoint to two groups.
Clinical threshold
A binary approach can provide an easily communicated probability comparison when a specific favorable-outcome threshold is prespecified.
6. Statistical Methods Explained
Why is mRS considered an ordinal endpoint?
The mRS categories have a meaningful order from 0 through 6: higher scores represent greater disability, with 6 representing death. However, the scale does not imply that moving from 1 to 2 represents exactly the same quantitative change as moving from 4 to 5. This makes ordinary continuous-outcome methods inappropriate unless their assumptions are specifically justified.
Why might an ordinal logistic model be useful?
An ordinal logistic model can use the complete ordering of the mRS categories. Rather than selecting one cutoff, it considers cumulative probabilities across the ordered scale. This can make the analysis more statistically efficient when the model assumptions are reasonable.
What is the proportional-odds assumption?
The proportional-odds assumption states, in simplified terms, that the treatment effect is sufficiently consistent across the cumulative splits of the ordinal outcome. For example, the treatment effect distinguishing mRS ≤0 from mRS >0 is constrained to relate to the treatment effect distinguishing mRS ≤1 from mRS >1 through the same treatment coefficient. If that assumption is inappropriate, the resulting common odds ratio may not adequately describe the data.
Why could a binary mRS analysis be used?
A binary analysis can answer a focused clinical question such as whether participants achieve a prespecified level of functional outcome. Its interpretation is straightforward, but it does not use the full ordering of all mRS categories.
Why does randomization matter statistically?
Randomization establishes the treatment assignment mechanism before outcome information is observed. In a well-conducted randomized comparison, baseline differences arise through random variation rather than systematic treatment-selection decisions. This supports causal interpretation of an appropriately specified between-arm comparison.
Why is the 90±7-day window important?
The endpoint is explicitly tied to 90±7 days after randomization. Defining the assessment window in advance helps establish when the primary outcome is measured and reduces ambiguity about which observations belong to the primary endpoint analysis.
Why should mRS not automatically be analysed as a continuous variable?
Assigning numerical values of 0 through 6 does not establish equal spacing between categories. Treating the scale as continuous therefore imposes assumptions that are not implied merely by the registry definition. An analysis should respect the ordinal nature of the outcome unless a different treatment is justified.
7. Planned Analysis
The registry identifies a single primary endpoint: neurofunctional deficit defined as modified Rankin Scale (mRS), assessed at 90±7 days after randomization. The endpoint is a seven-category ordered functional outcome ranging from 0, no symptoms, through 6, death.
| Statistical question | Appropriate analytical framework |
|---|---|
| Does treatment shift the distribution of mRS toward lower disability? | An ordinal regression framework can directly model the ordered outcome. |
| Does treatment increase the probability of a favorable mRS category? | A binary regression framework can be used if a specific mRS threshold is prespecified. |
| Are the complete mRS categories relevant? | An ordinal analysis retains more information than dichotomization. |
| Is the endpoint measured over time? | The primary registry endpoint is assessed at a fixed follow-up window, 90±7 days, rather than being defined as time to an event. |
What a complete primary analysis would ordinarily report
A complete statistical report for an ordinal mRS endpoint would normally identify the analysis population, treatment contrast, estimand, statistical model, effect measure, confidence interval, hypothesis test, and treatment of missing or unavailable 90±7-day assessments. If an ordinal proportional-odds model were used, the proportional-odds assumption and the direction of the odds ratio would also require clear interpretation.
For a binary mRS analysis, the report would additionally need to state the exact cutoff defining a favorable outcome. Without that prespecified threshold, the binary estimand is not uniquely defined.
8. Interpreting Effect Measures for an mRS Endpoint
Unlike a hazard ratio, an ordinal treatment effect does not inherently describe a rate of an event occurring over time. The interpretation depends on the model used for the ordered functional outcome.
If a proportional-odds model produces an odds ratio greater than 1 for the treatment group under a coding scheme in which lower mRS categories are favorable, that estimate would indicate greater odds of being in a lower-disability category across the cumulative mRS comparisons. The direction must always be stated with the model's coding convention.
If a binary favorable-outcome analysis is used, the risk difference compares the probability of meeting the prespecified favorable mRS criterion between randomized groups. It directly describes an absolute probability difference, unlike an odds ratio.
A confidence interval describes statistical uncertainty around the estimated treatment effect under the specified model and sampling framework. It does not describe the range of individual patient outcomes or guarantee that the true effect lies within the interval.
A p-value addresses the compatibility of the observed data with a specified null hypothesis under the statistical model. It does not measure the magnitude of the treatment effect, the clinical importance of an effect, or the probability that a treatment works.
9. Analysis Population and Randomization
The registry identifies the study as randomized with an enrollment of 456 and two arms. Randomization is the central design feature for making a treatment comparison because treatment assignment is determined before the primary outcome is observed.
| Design element | Statistical significance |
|---|---|
| Randomized allocation | Supports comparison of outcomes according to assigned treatment rather than treatment choice. |
| Two treatment arms | Creates a direct comparison between endovascular therapy and best medical management. |
| Parallel design | Participants remain in their randomized treatment groups rather than sequentially receiving both randomized interventions. |
| 456 enrolled | Defines the overall registered study size; the registry summary does not provide arm-specific enrollment counts. |
For a primary efficacy analysis, the intention-to-treat principle would ordinarily preserve randomized treatment assignment. The ClinicalTrials.gov record summarized here does not post a formal analysis population definition for the primary endpoint, so the exact population used for an eventual analysis cannot be established from this record alone.
10. Missing Data and Follow-Up
The primary endpoint has a defined assessment window of 90±7 days after randomization. That creates a natural point at which outcome data must be available for the primary analysis.
However, the registry information summarized here does not specify a missing-data or imputation method. Several statistically different strategies could be considered in a clinical trial with an ordinal functional endpoint, including complete-case analysis, prespecified imputation approaches, or sensitivity analyses under alternative missing-data assumptions.
Why missing mRS matters
If participants with unavailable primary outcomes differ systematically from those with observed outcomes, a complete-case analysis can introduce bias.
Why sensitivity analysis matters
Different assumptions about missing outcomes can produce different treatment-effect estimates. A robust statistical report therefore makes those assumptions explicit.
The absence of a posted missing-data method in the registry means that the primary endpoint's missing-data handling is not specified in the record summarized here.
11. Masking and Potential Bias
The registry classifies ANGEL-ASPECT as single-masked. Masking can reduce the influence of knowledge of treatment assignment on outcome assessment or other trial procedures, depending on which party is masked.
For an outcome such as mRS, assessment procedures are particularly important because the endpoint involves functional status and disability. A statistically appropriate model cannot by itself remove bias introduced by systematic differences in outcome assessment.
Randomization addresses treatment-selection bias, while masking and standardized outcome assessment can address other pathways through which knowledge of treatment assignment might influence measured outcomes.
The registry summary does not identify which participants, investigators, assessors, or other trial personnel were masked. That detail should therefore not be inferred from the classification alone.
12. Why the mRS Endpoint Is Statistically Interesting
The modified Rankin Scale is a useful teaching example because it combines several statistical features in one endpoint: ordered categories, a clinically interpretable direction, a fixed follow-up window, and a possible choice between preserving the ordinal distribution and collapsing it into a binary outcome.
| Feature | Statistical implication |
|---|---|
| Seven ordered categories | The outcome is ordinal rather than inherently continuous. |
| 0 to 6 scale | Higher scores correspond to greater disability, with death at 6. |
| Fixed follow-up | The primary endpoint is evaluated at 90±7 days after randomization. |
| Functional outcome | Outcome assessment can be sensitive to how disability is measured and classified. |
| Potential dichotomization | A threshold can simplify interpretation but discards information from the remaining categories. |
| Ordinal regression | Can preserve the ordering of the complete mRS scale when its assumptions are appropriate. |
13. Limitations
- No posted statistical analyses: the ClinicalTrials.gov record summarized here does not provide a primary treatment-effect estimate, confidence interval, p-value, or named statistical model.
- Limited endpoint detail: the registry identifies mRS and the 90±7-day time frame but does not specify a primary estimand or model in the information summarized here.
- Unknown arm sizes: overall enrollment is 456, but the record does not provide arm-specific enrollment counts in the information available here.
- Missing-data method not specified: the summarized registry information does not identify an imputation strategy or sensitivity-analysis framework for unavailable primary endpoint assessments.
- Masking details: the study is classified as single-masked, but the summarized record does not identify the masked party.
- Ordinal-model assumptions: if a proportional-odds model were used, its assumptions would need to be evaluated rather than taken for granted.
- Binary threshold dependence: a dichotomized mRS analysis depends on the selected cutoff, so the threshold is part of the estimand and should be prespecified.
- Incomplete statistical specification: the registry record alone does not establish whether covariate adjustment, stratification, interaction terms, or other modeling features were used.
14. Why This Trial Matters Statistically
ANGEL-ASPECT illustrates an important distinction in clinical-trial statistics: the clinical endpoint and the statistical estimand are not the same thing. The mRS provides the clinical measurement scale. The statistical method determines how the randomized treatment comparison is expressed and what quantity is being estimated.
| Concept | How it appears in ANGEL-ASPECT |
|---|---|
| Randomization | The trial uses randomized allocation between two interventions. |
| Parallel-group design | The registry identifies a parallel design rather than a crossover design. |
| Ordinal outcome | The primary endpoint is the modified Rankin Scale from 0 through 6. |
| Fixed follow-up | The primary endpoint is assessed 90±7 days after randomization. |
| Functional outcome | The endpoint measures disability or dependence in daily activities after stroke. |
| Ordinal regression | A potential framework for preserving the ordered structure of mRS. |
| Dichotomization | A potential alternative that focuses on a prespecified favorable-outcome threshold. |
| Confidence intervals | Needed to quantify uncertainty around a treatment-effect estimate. |
| Missing data | Important because the endpoint requires an assessment within a defined follow-up window. |
15. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary endpoint is an ordered functional outcome. The appropriate analysis should define a treatment estimand and use a model that respects the structure of the mRS scale or a prespecified clinically meaningful dichotomization.
Clinical interpretation
The mRS is intended to describe the degree of disability or dependence following stroke. A treatment comparison should therefore be understood in terms of functional outcome rather than as an abstract statistical score alone.
A statistically significant result, if reported, would address evidence against a specified null hypothesis. It would not by itself establish that every patient benefits, that the effect is clinically important for every individual, or that the treatment produces the same shift throughout the mRS distribution.
16. Statistical Questions a Complete Analysis Should Answer
| Question | Why it matters |
|---|---|
| What is the treatment estimand? | It defines exactly which treatment effect is being estimated and under what outcome and treatment conditions. |
| Is mRS analysed ordinally or dichotomized? | The choice determines which information from the seven-category endpoint is used. |
| If ordinal, is proportional odds appropriate? | The validity of a proportional-odds estimate depends on the model assumption. |
| If binary, what mRS threshold is used? | Different thresholds answer different clinical questions. |
| What analysis population is used? | Excluding or including participants in different ways can change the estimand and potential bias. |
| How are missing 90±7-day assessments handled? | Missing outcome assumptions can materially affect treatment-effect estimation. |
| What is the confidence interval? | It describes the statistical precision of the treatment-effect estimate. |
| What hypothesis is tested? | The p-value is meaningful only relative to a specified null hypothesis and analysis framework. |
17. Trial Timeline
Study start
The registry lists October 2, 2020 as the trial start date.
Randomized parallel-group evaluation
The trial enrolled 456 participants and compared endovascular therapy with best medical management in two parallel arms.
Primary endpoint assessment
Neurofunctional deficit was defined using the modified Rankin Scale at this prespecified time frame.
Primary completion
The registry lists August 15, 2022 as the primary completion date.
18. What the Modified Rankin Scale Does — and Does Not — Mean
The mRS records functional status on an ordered scale from 0 to 6. It is not itself a measure of treatment effect. The treatment effect comes from comparing the distribution or probability of mRS outcomes between randomized groups using a defined statistical estimand.
The difference between adjacent scores should not automatically be interpreted as a fixed numerical quantity. A change from 5 to 4 and a change from 2 to 1 both represent one category of movement on the scale, but the clinical meaning of those movements is not necessarily equivalent.
The registry defines mRS 6 as death. Consequently, an analysis of the full mRS distribution incorporates the most severe possible outcome directly into the ordered endpoint.
An estimated treatment effect summarizes the comparison between randomized groups. It does not imply that every participant experiences the same change in functional status.
19. Limitations of Reading Only a Registry Record
A registry record is useful for identifying the registered clinical question, design, endpoint definitions, and study status. It is not necessarily a complete statistical analysis plan or results report.
For ANGEL-ASPECT, the record identifies the primary endpoint and its timing but does not post formal statistical analyses in the information summarized here. As a result, several important statistical details remain unspecified, including the exact estimand, primary model, treatment-effect measure, confidence interval, p-value, missing-data method, and analysis population.
This distinction is important for statistical interpretation. Knowing that an endpoint is measured with the mRS does not by itself determine whether an ordinal regression, binary regression, or another method was prespecified. Those choices define different statistical questions.
20. Sources
- ClinicalTrials.gov: NCT04551664 — ANGEL-ASPECT.
Continue through the Clinical Biostats clinical-trial library
Clinical trial analysis connects study design, endpoint definition, estimands, statistical models, uncertainty, and interpretation into one coherent statistical story.
21. Record Summary
ANGEL-ASPECT is a completed randomized, parallel-group trial enrolling 456 participants with acute ischemic stroke. The registry describes two interventions: endovascular therapy and best medical management. Its primary endpoint is neurofunctional deficit defined as the modified Rankin Scale, assessed at 90±7 days after randomization.
Statistically, the central feature of the trial is the ordinal nature of the mRS endpoint. A complete analysis must distinguish the clinical scale from the treatment estimand used to compare randomized groups. An ordinal model can preserve information across the full mRS distribution, while a binary analysis can focus on a prespecified favorable-outcome threshold. Each approach answers a different question and requires its own assumptions and interpretation.