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
CAIRO was a randomized, parallel-group, triple-masked phase 4 treatment trial enrolling 188 participants with ischemic cerebrovascular accident. The registry identifies clopidogrel and aspirin as the interventions and lists two primary neurologic endpoints: change in NIH stroke scale score from baseline through 1 week and neurologic outcome assessed with the Modified Ranking Scale at 3 months.
| Feature | CAIRO |
|---|---|
| Acronym | CAIRO |
| ClinicalTrials.gov identifier | NCT02776540 |
| Official title | Clopidogrel for Acute Ischaemia of Recent Onset |
| Condition | Ischemic Cerebrovascular Accident |
| Phase | Phase 4 |
| Status | Completed |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 188 |
| Number of arms | 3 |
| Lead sponsor | Ain Shams University |
2. Clinical Question
The registry frames CAIRO as a treatment study in ischemic cerebrovascular accident. Its two primary questions concern neurologic change during the first week after baseline and neurologic outcome at 3 months.
Population
Patients with ischemic cerebrovascular accident.
Interventions
Clopidogrel and aspirin are identified as the trial interventions.
Design
Randomized, parallel-group, triple-masked treatment trial with 3 arms.
Primary questions
How does neurologic status change from baseline through 1 week, and what is the neurologic outcome at 3 months?
3. Trial Design
4. Trial Timeline
Trial start
The registry lists 2016-06-01 as the study start date.
Primary completion
The registry lists 2019-02-01 as the primary completion date.
Current registry status
The study is listed as completed, with enrollment of 188 participants.
5. Primary Endpoints
The registry specifies two primary endpoints. They measure neurologic status on different time scales, so they should be interpreted as related but distinct outcomes rather than as interchangeable measures.
| Endpoint | Time frame | Registry description |
|---|---|---|
| The change of NIH stroke scale score | Baseline and up to 1 week | Patients will be assessed for early neurologic improvement or deterioration using the change of NIH stroke scale score. Early neurological outcome |
| Neurologic outcome | 3 months | Patients will be assessed for neurologic outcome using the Modified Ranking Scale at 3 months after onset. |
Two endpoints, two statistical questions
The first endpoint is a change-from-baseline measure. Its natural statistical focus is the difference in neurologic-score change between randomized treatment groups. The second endpoint is a 3-month neurologic outcome measured using the Modified Ranking Scale, which is an ordinal clinical outcome. Its statistical analysis should therefore respect the ordered nature of the scale rather than automatically treating it as a continuous measurement.
Early neurologic outcome
The NIH stroke scale endpoint focuses on change from baseline through 1 week, allowing improvement or deterioration to be represented through a change score.
Three-month outcome
The Modified Ranking Scale endpoint assesses neurologic outcome at 3 months after onset and has an ordered categorical structure.
6. Planned Analysis
The registry contains the endpoint definitions and time frames but does not post formal statistical analyses or numerical outcome results. Consequently, the appropriate statistical framework can be described, but no treatment effect estimate can be reported from the registry.
Primary endpoint 1: change in NIH stroke scale score
The endpoint is defined as the change of NIH stroke scale score from baseline and up to 1 week. For a randomized comparison, an analysis would typically compare the distribution of change scores across the treatment groups while accounting for the baseline measurement when appropriate.
The sign and clinical interpretation of the change depend on the direction in which the NIH stroke scale is defined. The registry identifies the endpoint as a change score intended to assess early neurologic improvement or deterioration.
An ANCOVA-style analysis could compare follow-up neurologic scores between randomized groups while adjusting for baseline score, provided that the statistical assumptions and prespecified analysis plan support that approach. Alternatively, the change score itself could be analysed directly. The choice should be driven by the prespecified statistical analysis plan and the distributional properties of the observed outcome.
Primary endpoint 2: neurologic outcome at 3 months
The registry defines neurologic outcome using the Modified Ranking Scale at 3 months after onset. Because this is an ordered outcome, an ordinal regression model such as a proportional-odds model is a natural candidate when its assumptions are reasonable.
Here, Y represents an ordered neurologic outcome, k represents an outcome threshold, and X represents treatment assignment. The model evaluates treatment across the ordered outcome structure rather than collapsing the scale into a binary endpoint without justification.
A proportional-odds analysis assumes that the treatment effect is sufficiently consistent across the cumulative thresholds of the ordinal outcome. If that assumption does not hold, alternative ordinal models or prespecified dichotomization strategies may be considered.
7. Statistical Methodology
Randomization
Randomization creates treatment groups whose differences in outcomes can be interpreted within the framework of a randomized experiment. The principal statistical comparison should preserve the randomized treatment assignment rather than selecting patients according to their observed post-randomization outcomes.
Change-from-baseline analysis
The NIH stroke scale endpoint is explicitly defined as a change from baseline through 1 week. Change scores quantify how much an individual's measured neurologic status differs from baseline. A treatment comparison can then focus on the difference between groups in these changes.
The quantity δ represents a treatment-group contrast in average change when a mean-based analysis is appropriate. The registry does not report a value for this contrast.
ANCOVA as a potential framework
For a continuous outcome measured after treatment, ANCOVA can compare groups at follow-up while adjusting for the corresponding baseline measurement. This can improve statistical efficiency and directly account for baseline differences in the measured score.
For this trial, the key distinction is between analysing change and analysing the follow-up score adjusted for baseline. These are related approaches but are not mathematically identical in every setting, particularly when baseline and follow-up variability or model assumptions differ.
Ordinal regression for the Modified Ranking Scale
The Modified Ranking Scale is an ordered outcome. Ordinal regression makes use of the ordering without assuming that the distance between adjacent categories is numerically equal. This is important because an ordinal scale communicates ranking but does not automatically imply equal spacing between categories.
Confidence intervals
A confidence interval quantifies uncertainty around an estimated treatment contrast under the specified statistical model and sampling framework. It is more informative than a point estimate alone because it shows the range of values compatible with the analysis at the stated confidence level.
P-values
A p-value addresses the compatibility of the observed data with a specified null hypothesis under the statistical model. It is not a measure of the magnitude of a treatment effect and does not indicate the probability that the null hypothesis is true.
8. Statistical Methods Explained
Why analyse change in the NIH stroke scale?
The first primary endpoint is explicitly defined as a change from baseline. Analysing change focuses attention on the amount of neurologic improvement or deterioration occurring during the specified period rather than relying only on the absolute score at follow-up.
Why might baseline adjustment be useful?
Patients can enter a randomized trial with different baseline neurologic scores even when randomization is properly performed. Adjusting for the baseline score can account for prognostic information already available before treatment and can make the treatment comparison more statistically efficient when the model is appropriate.
Why is the Modified Ranking Scale different from a continuous outcome?
The Modified Ranking Scale is an ordered clinical outcome. Its categories have a meaningful order, but the numerical labels should not automatically be interpreted as measurements with equal distances between categories. Ordinal methods preserve the ordering without requiring that equal-spacing assumption.
What does a proportional-odds model assume?
An ordinal proportional-odds model assumes that the treatment effect is approximately consistent across the cumulative thresholds of the ordered outcome. If that assumption is violated, the model's single treatment-effect parameter may not adequately describe the data.
Why are the 1-week and 3-month endpoints not interchangeable?
The NIH stroke scale endpoint evaluates change from baseline through 1 week, whereas the Modified Ranking Scale endpoint evaluates neurologic outcome at 3 months after onset. They therefore describe neurologic status at different times and using different measurement structures.
Why does randomization matter for interpretation?
Randomization establishes the treatment comparison before the outcomes are observed. This reduces systematic differences in treatment assignment that could otherwise confound an observational comparison. The statistical analysis then estimates the contrast between randomized groups within that experimental framework.
9. Analysis of the Early Neurologic Endpoint
The first primary endpoint is the change of NIH stroke scale score from baseline and up to 1 week. The registry describes this endpoint as an assessment of early neurological improvement or deterioration.
Outcome structure
Repeated neurologic assessment over the interval from baseline through 1 week is summarized as a change in NIH stroke scale score.
Primary comparison
The central comparison would be the difference in neurologic change between the randomized treatment groups.
Baseline role
The baseline score is part of the endpoint definition and is also an important candidate covariate for adjusted analysis.
Clinical interpretation
The endpoint is designed to distinguish early neurologic improvement or deterioration rather than long-term neurologic status.
A useful analysis would report an effect estimate with a confidence interval rather than relying only on a hypothesis-test result. If the score distribution is substantially non-normal or strongly skewed, distribution-free or ordinal approaches may be more appropriate than a simple mean comparison.
10. Analysis of the Three-Month Neurologic Endpoint
The second primary endpoint is neurologic outcome at 3 months, measured using the Modified Ranking Scale. This endpoint extends the assessment beyond the first week and uses an ordered outcome scale.
| Feature | Statistical implication |
|---|---|
| Time point | 3 months after onset |
| Outcome | Neurologic outcome |
| Measurement | Modified Ranking Scale |
| Scale structure | Ordered categorical outcome |
| Natural modelling framework | Ordinal regression when model assumptions are appropriate |
For an ordinal analysis, the treatment effect can be expressed as an odds ratio comparing the cumulative probability of being at or below successive outcome thresholds. The direction of the odds ratio must be stated together with the coding convention because the numerical interpretation depends on which outcome direction is modelled as favorable.
11. Missing Data and Follow-Up
The registry identifies the assessment time frames for the two primary endpoints but does not provide a posted missing-data or imputation strategy. For a longitudinal neurologic endpoint, missing follow-up measurements can affect both the estimated treatment contrast and its uncertainty.
A prespecified analysis plan would ordinarily define how participants with incomplete 1-week assessments and incomplete 3-month assessments are handled. Potential approaches depend on the endpoint, the reason for missingness, the amount of missing information, and the assumptions considered scientifically credible.
Why missingness matters
If follow-up measurements are unavailable for some participants, the analysed population may differ from the randomized population in ways that affect inference.
Imputation is not automatic
An imputation method should reflect the outcome structure and a defensible assumption about the mechanism generating missing observations.
For the Modified Ranking Scale, imputation should also respect its ordinal structure. For the NIH stroke scale change endpoint, any imputation strategy should account for the baseline and follow-up measurement structure.
12. Multiplicity and Two Primary Endpoints
CAIRO has two registered primary endpoints. When a trial has more than one primary endpoint, the interpretation of statistical significance depends on how the endpoints were incorporated into the prespecified hypothesis-testing strategy.
| Primary endpoint | Time frame | Statistical issue |
|---|---|---|
| Change of NIH stroke scale score | Baseline and up to 1 week | Continuous or score-based change analysis, with baseline considerations |
| Neurologic outcome using Modified Ranking Scale | 3 months | Ordinal outcome analysis and appropriate multiplicity control |
If both endpoints are intended to support a confirmatory conclusion, the statistical plan should specify whether success requires both endpoints, either endpoint, or a multiplicity-adjusted combination of hypotheses. Without the prespecified testing strategy, a numerical p-value for one endpoint cannot by itself establish how the complete primary-endpoint family should be interpreted.
13. Randomization and Three-Arm Interpretation
The trial has 3 arms and randomized allocation. A three-arm design creates several possible treatment contrasts, but the scientifically meaningful comparison must be defined by the prespecified hypotheses rather than selected after looking at observed outcomes.
With 3 arms, a complete statistical analysis should distinguish the overall comparison among groups from any specific pairwise comparison. If several pairwise tests are performed, the familywise interpretation depends on the prespecified multiplicity strategy.
14. Masking and Bias Control
The registry describes CAIRO as triple-masked. Masking is particularly relevant when outcomes involve clinical assessment, because knowledge of treatment assignment can influence how symptoms or neurologic status are evaluated.
Measurement protection
Masking can reduce the potential for treatment knowledge to influence outcome assessment.
Interpretation
Triple masking is a design characteristic; it does not itself quantify the magnitude of any treatment effect.
The presence of masking should therefore be considered part of the trial's bias-control framework rather than treated as a statistical result. The randomized comparison still depends on appropriate analysis of the prespecified outcomes.
15. Statistical Interpretation of the Planned Endpoints
The first endpoint asks whether neurologic status changes differently between randomized groups during the period from baseline through 1 week. An effect estimate should describe the between-group difference in change, accompanied by a confidence interval that quantifies statistical uncertainty.
The second endpoint asks whether the distribution of neurologic outcomes at 3 months differs between randomized groups. Because the Modified Ranking Scale is ordered, an ordinal treatment-effect measure can retain more information than an arbitrary binary split.
Even if a formal analysis produced a statistically significant p-value, that would not quantify the size or clinical importance of the treatment effect. The magnitude, confidence interval, outcome scale, and absolute distribution of outcomes would all remain important for interpretation.
A confidence interval would express uncertainty around the estimated treatment contrast under the chosen model. It would not describe the range of outcomes expected for individual patients.
16. Why Change Scores and Follow-Up Scores Are Different
The first endpoint explicitly uses a change from baseline. This makes the baseline measurement part of the outcome calculation. A follow-up-only analysis asks a different question unless it adjusts for the baseline measurement.
| Approach | Question addressed |
|---|---|
| Change from baseline | How much did neurologic status change during the specified period? |
| Follow-up adjusted for baseline | How do treatment groups differ at follow-up after accounting for baseline status? |
| Unadjusted follow-up comparison | How do treatment groups differ in observed follow-up scores without baseline adjustment? |
These approaches can produce similar conclusions in appropriate settings, but they rely on different statistical formulations. A rigorous analysis should use the method specified before outcome data are examined.
17. Interpreting an Ordinal Treatment Effect
Suppose an ordinal model were used for the 3-month Modified Ranking Scale. Its treatment effect would generally summarize the relative odds of being at or below a specified outcome threshold across the ordered scale, depending on the direction and coding of the model.
The exact clinical direction of an odds ratio depends on how the Modified Ranking Scale categories are coded and which cumulative event is modelled.
This illustrates why the numerical effect estimate cannot be interpreted independently of the outcome coding. The same underlying clinical comparison can yield reciprocal odds ratios under opposite coding conventions.
18. What the Registry Reports — and What It Does Not
| Element | Registry status |
|---|---|
| Trial title | Reported |
| Enrollment | 188 |
| Randomized allocation | Reported |
| Parallel design | Reported |
| Triple masking | Reported |
| Primary endpoints | Reported |
| Endpoint time frames | Reported |
| Formal statistical analyses | Not posted |
| Numerical primary endpoint results | Not posted |
This distinction is important because trial registration information and trial results serve different purposes. The registry establishes the study's design and planned outcome framework, while a statistical results report would provide estimates, uncertainty, and formal hypothesis tests.
19. Important Limitations and Interpretation Issues
- No posted statistical results: the registry does not provide numerical treatment-effect estimates, confidence intervals, or p-values for the primary endpoints.
- Three-arm design: the presence of 3 arms means that the prespecified treatment contrasts and multiplicity strategy are important to interpreting any eventual statistical tests.
- Two primary endpoints: the NIH stroke scale change endpoint and the 3-month Modified Ranking Scale endpoint address different aspects and time frames of neurologic outcome.
- Ordinal outcome: the Modified Ranking Scale should be analysed with methods appropriate to ordered categorical data unless a prespecified alternative is justified.
- Baseline dependence: the NIH stroke scale endpoint is defined through change from baseline, making the baseline measurement central to its interpretation and analysis.
- Missing follow-up: incomplete measurements can affect estimates and uncertainty, particularly for outcomes assessed at defined follow-up times.
- Model assumptions: ordinal regression, ANCOVA, and other model-based analyses require assumptions that should be assessed against the observed data.
- Time-specific interpretation: an early neurologic endpoint through 1 week should not be treated as equivalent to neurologic outcome at 3 months.
20. Why This Trial Matters Statistically
CAIRO is a useful teaching case because its registered design connects several fundamental clinical-trial concepts: randomized treatment allocation, a three-arm parallel structure, triple masking, change-from-baseline analysis, an ordinal neurologic endpoint, multiple primary endpoints, and distinct early and later outcome assessments.
| Concept | How it appears in CAIRO |
|---|---|
| Randomization | Randomized allocation of 188 participants |
| Parallel design | 3-arm parallel-group trial |
| Masking | Triple-masked design |
| Change from baseline | Change of NIH stroke scale score from baseline and up to 1 week |
| Ordinal analysis | Modified Ranking Scale at 3 months |
| Multiple primary endpoints | Early neurologic change and 3-month neurologic outcome |
| Multiplicity | Multiple arms and primary endpoints require prespecified hypothesis-testing strategy |
| Missing data | Follow-up assessments create potential missing-outcome considerations |
| Clinical timing | Early assessment through 1 week versus outcome at 3 months |
The statistical lesson is that endpoint definition comes before statistical testing. The NIH stroke scale endpoint has a change-score structure, while the Modified Ranking Scale endpoint has an ordinal structure. Treating both as generic numerical outcomes would discard important information about how the outcomes were defined.
21. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registry establishes a randomized, triple-masked, three-arm design with two primary neurologic endpoints but does not post formal statistical results. The appropriate analysis depends on the outcome structure and prespecified hypothesis-testing framework.
Clinical interpretation
The two endpoints were designed to examine neurologic status at different stages: early neurologic change through 1 week and neurologic outcome at 3 months. They therefore provide distinct clinical perspectives on the treatment comparison.
22. A Practical Statistical Analysis Framework
If a complete patient-level dataset and prespecified statistical analysis plan were available, a rigorous analysis could proceed in a sequence such as the following:
The first step would confirm that the analysis population and treatment coding preserve the randomized design. The second would characterize the baseline score relevant to the change endpoint. The third would estimate the treatment contrast for the NIH stroke scale change. The fourth would analyse the Modified Ranking Scale using an ordinal framework when appropriate. The final step would interpret both primary endpoints under the prespecified multiplicity strategy.
23. Educational Takeaways
Endpoint structure drives analysis
A change score and an ordinal outcome should not automatically be analysed using the same statistical method.
Randomization defines the comparison
The principal treatment comparison should follow randomized assignment rather than post-randomization outcome status.
Time matters
An endpoint measured through 1 week answers a different question from an endpoint measured at 3 months.
Multiplicity matters
Three treatment arms and two primary endpoints make the prespecified testing strategy important for formal inference.
24. Sources
- ClinicalTrials.gov: NCT02776540 — CAIRO: Clopidogrel for Acute Ischaemia of Recent Onset.
Continue through Clinical Biostats
Explore statistical tutorials, calculators, and additional clinical trial analyses covering study design, endpoint selection, regression, and clinical-trial inference.
25. Record Summary
CAIRO is a completed phase 4 randomized parallel-group trial of 188 participants with ischemic cerebrovascular accident. The registry identifies 3 arms, triple masking, and two primary neurologic endpoints: change of NIH stroke scale score from baseline and up to 1 week, and neurologic outcome using the Modified Ranking Scale at 3 months after onset.
From a statistical perspective, the trial illustrates why the measurement scale and timing of an endpoint should determine the analysis. The early endpoint is naturally framed around change from baseline, whereas the 3-month endpoint is an ordered neurologic outcome for which ordinal modelling can be appropriate. The three-arm structure and two primary endpoints also make the prespecified hypothesis-testing strategy central to formal inference.