This page separates reported trial results from statistical interpretation. Numerical results on this page are restricted to the ClinicalTrials.gov record for NCT00235495. Where the registry does not provide a requested detail, it is not inferred from external publications or other sources.
1. Trial at a Glance
ALIAS was a randomized, parallel, triple-masked phase 3 trial comparing albumin with saline in participants with ischemic stroke. The primary endpoint was a binary favorable outcome at 3 months, defined using the National Institute of Health Stroke Scale (NIHSS) and/or Modified Rankin Scale (mRS).
| Feature | ALIAS |
|---|---|
| Brief title | Albumin in Acute Ischemic Stroke Trial |
| Phase | Phase 3 |
| Condition | Ischemic Stroke |
| Design | Randomized, parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 841 |
| Interventions | Albumin (biological); Saline (drug) |
| Primary endpoint type | Binary |
| Results posted | Yes |
| Outcome measures posted | 22 |
| Statistical analyses posted | 22 |
| Lead sponsor | University of Miami |
| Trial status | Terminated |
2. Clinical Question
The central statistical question was whether participants randomized to albumin had a greater proportion of favorable outcomes at 3 months than participants randomized to saline, where favorable outcome was defined as a National Institute of Health Stroke Scale (NIHSS) score of 0-1 and/or a Modified Rankin Scale (mRS) score of 0-1.
Population
Participants with ischemic stroke enrolled in the phase 3 ALIAS trial.
Intervention
Albumin, identified in the registry as a biological intervention.
Comparator
Saline, identified in the registry as a drug intervention.
Primary question
Does albumin produce a greater proportion of participants with the prespecified favorable neurologic outcome at 3 months?
3. Trial Design
Albumin
- Albumin
- Biological intervention
- Compared with saline under the randomized parallel design
Saline
- Saline
- Drug intervention according to the registry classification
- Comparator for the randomized treatment comparison
The registry identifies the allocation as randomized, the design model as parallel, and masking as triple. Randomization provides the principal basis for comparing outcomes between treatment assignments, while triple masking is intended to reduce the opportunity for knowledge of treatment assignment to influence trial conduct or outcome assessment.
4. Trial Timeline and Registry Status
Trial start
The registry lists June 2006 as the trial start date.
Primary completion
The registry lists February 2013 as the primary completion date.
Terminated
The ClinicalTrials.gov record identifies the study status as terminated.
5. Primary Endpoint
| Endpoint | Definition | Time frame | Type |
|---|---|---|---|
| Favorable outcome | The Number of Participants With Favorable Outcome Defined as National Institute of Health Stroke Scale (NIHSS) Score of 0-1 and/or Modified Rankin Scale (mRS) of 0-1. | at 3 months | Binary |
The registry definition combines two clinically meaningful scales. The NIHSS provides a quantitative assessment of stroke-related neurologic deficit, while the mRS provides a functional disability scale. The binary endpoint therefore converts these underlying assessments into a yes/no favorable-outcome classification.
Once this endpoint is defined, the primary comparison becomes a comparison of proportions: the proportion meeting the favorable-outcome definition in the albumin group versus the proportion in the saline group.
6. Statistical Methodology
Intention-to-treat analysis
The primary analysis used the Intent to Treat population. The registry defines this as all subjects randomized into the study, included and analyzed based on the treatment arm to which they were randomized. This preserves the treatment comparison created by randomization.
Logistic regression
The reported primary method was logistic regression. Logistic regression is appropriate when the outcome is binary because it models the relationship between treatment assignment and the probability of the outcome occurring. In this trial, the registry reports the effect measure as a risk ratio (RR) rather than an odds ratio.
The primary analysis notes that adjustment was made for baseline NIHSS and Thrombolysis stratum. Thus, the reported effect was not simply an unadjusted comparison of two raw proportions; it incorporated prespecified baseline information identified in the registry analysis notes.
An RR of 1 corresponds to equal outcome proportions. An RR below 1 means the observed favorable-outcome risk was lower in the albumin group relative to saline, while an RR above 1 means it was higher.
Wilcoxon rank sum test
Several continuous outcomes were analyzed using the Wilcoxon rank sum test with normal approximation. This is the two-group version commonly associated with the Mann-Whitney framework. Rather than relying on a normal-distribution assumption for the raw outcome values, the method works through the relative ordering, or ranks, of observations.
Different endpoints require different statistical summaries
The ALIAS results illustrate why a clinical trial does not have one universal statistical test. Binary endpoints such as favorable outcome, NIHSS thresholds, mRS thresholds, and adverse events were analyzed with logistic regression and reported using risk ratios. Continuous outcomes such as Trailmaking A and B were analyzed using the Wilcoxon rank sum approach and reported with rank-sum statistics and p-values.
7. Primary Result: Favorable Outcome at 3 Months
The registry reports one formal primary endpoint analysis comparing albumin with saline in the intent-to-treat population.
Risk ratio for favorable outcome
95% CI: 0.84–1.10
Two-sided confidence interval · Logistic regression · Intent-to-treat population
| Primary endpoint | Effect measure | Estimate | 95% CI | Analysis |
|---|---|---|---|---|
| Favorable outcome at 3 months | Risk Ratio | 0.96 | 0.84–1.10 | Logistic regression |
The estimated risk ratio of 0.96 means that the estimated proportion with the favorable binary outcome was 0.96 times the corresponding proportion in the saline group under the reported analysis. Expressed descriptively, this is an estimated relative difference of approximately 4% lower in the albumin group.
The estimate does not mean that albumin reduced every patient's chance of favorable recovery by 4%, nor does it provide an individual-patient probability. It is a group-level relative effect estimate.
The 95% confidence interval of 0.84–1.10 communicates uncertainty around the estimated risk ratio. Because the interval includes 1, the ClinicalTrials.gov record does not establish a statistically distinguishable difference in the favorable-outcome risk at the conventional null value of equal risk.
The ClinicalTrials.gov record does not report a p-value for the primary analysis. A p-value should therefore not be inferred or reconstructed from the confidence interval.
The analysis was adjusted for baseline NIHSS and Thrombolysis stratum according to the registry analysis notes. Interpretation should therefore remain tied to the reported adjusted model rather than treating 0.96 as a simple unadjusted ratio of raw proportions.
8. Secondary Efficacy Results
The registry reports additional binary neurologic, functional, quality-of-life, and cognitive endpoints. Most were analyzed with logistic regression in the intent-to-treat population and reported as risk ratios with two-sided confidence intervals.
| Secondary endpoint | Time frame | Estimate | CI | Method |
|---|---|---|---|---|
| Composite of mRS 0-1 and/or NIHSS 0-1 and/or decrease in NIHSS from baseline by 10 or more points | at 3 months | RR 0.93 | 99% CI 0.80–1.09 | Logistic regression |
| NIHSS 0-1 | at 24 hours | RR 0.99 | 99% CI 0.66–1.49 | Logistic regression |
| NIHSS 0-1 | at 90 days | RR 0.88 | 99% CI 0.71–1.10 | Logistic regression |
| mRS 0-1 | at 90 days | RR 1.03 | 99% CI 0.82–1.28 | Logistic regression |
| mRS 0-2 | at 90 days | RR 0.99 | 99% CI 0.85–1.13 | Logistic regression |
| Favorable outcome per mRS | 90 days | RR 0.98 | 99% CI 0.87–1.11 | Logistic regression |
| Barthel Index 95-100 | at 90 days | RR 0.95 | 99% CI 0.83–1.10 | Logistic regression |
| EuroQol (EQ-5D) favorable score < 0.78 | at 90 days | RR 0.99 | 99% CI 0.84–1.17 | Logistic regression |
| Stroke Specific Quality of Life Scale (SSQOL) score ≥3 | at 90 days | RR 0.93 | 99% CI 0.84–1.02 | Logistic regression |
The repeated use of 99% confidence intervals for these secondary binary endpoints is statistically important. A 99% interval is wider than a 95% interval for the same underlying estimate and therefore reflects a more stringent confidence level. The ClinicalTrials.gov record does not provide a separate multiplicity-adjustment description for these endpoints, so the confidence level should not be treated as proof of a particular multiple-testing procedure.
How to read the secondary risk ratios
The estimates range from 0.88 to 1.03 across the reported neurologic, functional, and quality-of-life binary endpoints. The corresponding confidence intervals vary substantially in width. For example, the 24-hour NIHSS endpoint has a 99% CI of 0.66–1.49, indicating substantially more uncertainty than the 90-day mRS 0-2 endpoint with a 99% CI of 0.85–1.13.
These results should be read endpoint by endpoint. A risk ratio is meaningful only in the context of its exact outcome definition and time frame. The 0.88 estimate for NIHSS 0-1 at 90 days is not interchangeable with the 0.99 estimate for mRS 0-2 at 90 days because the binary outcomes are different.
9. Cognitive Outcomes: Trailmaking A and B
Two continuous secondary endpoints were analyzed using the Wilcoxon rank sum test with normal approximation.
| Endpoint | Time frame | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Trailmaking A | at 90 days | Rank Sum | 73098 | 95% CI reported | 0.913 |
| Trailmaking B | at 90 days | Rank Sum | 44853.5 | 95% CI reported | 0.923 |
The registry identifies Trailmaking A as measured in minutes and states that participants were excluded if the Tail Making Test was not assessed. Trailmaking B was analyzed in the intent-to-treat population according to the ClinicalTrials.gov record.
The reported statistics of 73098 and 44853.5 are rank-sum statistics rather than mean differences or risk ratios. They should therefore not be interpreted as minutes of treatment effect.
The reported p-values are 0.913 for Trailmaking A and 0.923 for Trailmaking B. A p-value describes compatibility of the observed data with the null hypothesis under the specified statistical procedure; it does not measure the magnitude or clinical importance of an effect.
Because the registry supplies rank-sum statistics rather than a direct difference in means or medians, these results do not by themselves quantify how many minutes faster or slower one group performed.
10. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized arm using affected participants over the number at risk. The safety sample is described in the registry analyses as consisting of 830 subjects who received at least 20% of the intended dose of study drug.
| Safety measure | Albumin | Saline |
|---|---|---|
| Serious adverse events | 212/422 | 178/419 |
| Safety analysis population described in registry | 830 subjects who received at least 20% of the intended dose of study drug | |
The serious-adverse-event figures are descriptive affected/at-risk counts reported in the ClinicalTrials.gov record. The ClinicalTrials.gov record does not provide a formal confidence interval or p-value for this summary, so no inferential comparison is added here.
11. Reported Safety and Clinical Event Analyses
The registry also reports binary analyses for neurologic deterioration, neurologic death, recurrent ischemic stroke, atrial fibrillation, pulmonary edema, shortness of breath, intracerebral hemorrhage, and death. These were analyzed with logistic regression and expressed as risk ratios.
| Endpoint | Time frame | RR | 95% CI |
|---|---|---|---|
| Neurological deterioration | within 48 hours | 1.18 | 0.79–1.76 |
| Neurological death | within 7 days | 1.01 | 0.47–2.15 |
| Recurrent ischemic stroke | within 30 days | 1.16 | 0.39–3.41 |
| Atrial fibrillation | within 48 hours | 1.69 | 0.98–2.94 |
| Pulmonary edema | within 48 hours | 10.8 | 4.37–26.72 |
| Shortness of breath | within 48 hours | 2.58 | 1.09–6.12 |
| Symptomatic intracerebral hemorrhage (ICH) | within 24 hours | 2.42 | 1.02–5.78 |
| Asymptomatic ICH | within 24 hours | 1.14 | 0.68–2.01 |
| Death | within 30 days | 1.05 | 0.68–1.61 |
| Death | within 90 days | 1.10 | 0.74–1.63 |
These estimates illustrate the difference between a point estimate and its precision. The pulmonary-edema estimate is 10.8, accompanied by a 95% CI of 4.37–26.72. The point estimate is large, but the interval is also wide. The recurrent-ischemic-stroke estimate of 1.16 has a much wider relative uncertainty, with a 95% CI of 0.39–3.41.
A risk ratio above 1 indicates a higher estimated event risk in the albumin group relative to saline for the particular endpoint. It does not automatically establish causality for an individual adverse event, and the magnitude of the estimate must be considered alongside the confidence interval, endpoint definition, event frequency, and analysis population.
The registry reports a risk ratio of 2.42 for symptomatic ICH within 24 hours with a 95% CI of 1.02–5.78. The interval lies above 1, whereas the interval for asymptomatic ICH, 0.68–2.01, includes 1. These are separate endpoints and should not be combined into one conclusion.
12. Statistical Interpretation of the Primary Risk Ratio
The primary estimate of 0.96 is a relative comparison of the probability of the prespecified favorable outcome between albumin and saline. A value of 1 would represent equal estimated risk; 0.96 is below that reference value.
It does not mean that albumin causes a 4% decrease in every participant's probability of recovery. It also does not mean that exactly 4% fewer participants had a favorable outcome. The absolute difference in proportions is not reported in the ClinicalTrials.gov record and therefore is not reconstructed here.
The 95% CI of 0.84–1.10 shows the statistical uncertainty around the estimated relative effect. The interval contains 1, so the reported estimate is compatible with equal risks as well as with relative effects on either side of the null within that interval.
The trial data provide a primary estimate and confidence interval but no primary p-value. A confidence interval and p-value are related to the same inferential framework, but the absent p-value should not be reverse-engineered from rounded values. The safest interpretation is the one directly supported by the reported estimate and interval.
13. Why Logistic Regression Was Used
The primary endpoint was binary: each randomized participant either met the favorable-outcome definition or did not. Logistic regression is designed for this type of outcome because it models the probability of a binary event while allowing covariate adjustment.
This is a conceptual representation of the adjustment described in the registry analysis notes. The registry reports logistic regression with adjustment for baseline NIHSS and Thrombolysis stratum; it does not provide the complete fitted equation in the ClinicalTrials.gov record.
The important statistical point is that adjustment can improve precision and account for prespecified prognostic variables. It does not turn an observational comparison into a randomized one: the fundamental treatment comparison still comes from the randomized allocation.
14. Why the Wilcoxon / Mann-Whitney Test Was Used
Trailmaking A and B were continuous outcomes and were analyzed using the Wilcoxon rank sum test with normal approximation. This approach compares the ordering of observations between two independent groups rather than requiring the analysis to be based directly on means.
Rank-based comparison
Observations are converted into ranks across the groups, and the treatment comparison is based on those ranks.
Normal approximation
The registry method explicitly identifies a normal approximation for the Wilcoxon rank sum statistic.
Not a mean difference
A rank-sum statistic such as 73098 or 44853.5 is not itself a treatment difference measured in minutes.
Missing assessment
For Trailmaking A, the registry specifically states that participants were excluded if the Tail Making Test was not assessed.
15. Statistical Methods Explained
Why was intention-to-treat analysis used?
In an intention-to-treat analysis, participants remain associated with the treatment group to which they were randomized. The registry definition explicitly includes all randomized subjects and analyzes them according to randomized treatment arm. This protects the interpretability of the randomized comparison and avoids selectively removing participants because of what happened after randomization.
Why is the primary endpoint binary?
The registered endpoint converts neurologic and functional assessments into a favorable-versus-not-favorable outcome using NIHSS 0-1 and/or mRS 0-1. Once defined this way, the statistical question becomes a comparison of two proportions, making categorical-data methods appropriate.
What does a risk ratio of 0.96 mean?
A risk ratio of 0.96 means the estimated probability of the specified favorable outcome in the albumin group was 0.96 times that in the saline group under the reported model. It is a relative measure, not an absolute percentage-point difference.
Why does the confidence interval include 1?
For a risk ratio, 1 is the null value corresponding to equal risks. The primary 95% CI of 0.84–1.10 therefore spans the null. This means the data do not provide a sufficiently precise estimate to exclude equal risk at the confidence level represented by that interval.
Why shouldn't a p-value be inferred?
The registry-reported primary analysis gives an estimate and a 95% confidence interval but does not give a p-value. Although p-values and confidence intervals are mathematically related under particular inferential assumptions, reconstructing a p-value from displayed rounded limits would introduce an unsupported result.
Why use the Wilcoxon test for Trailmaking outcomes?
The registry reports Wilcoxon rank sum testing with normal approximation for Trailmaking A and B. Rank-based methods are useful when comparing two independent groups without making the analysis depend directly on normality of the raw measurements.
Why are the safety and efficacy populations different?
The primary efficacy analysis is explicitly intention-to-treat, while the registry-reported safety analyses describe a safety sample of subjects who received at least 20% of the intended dose. These populations answer different questions: randomized efficacy comparisons preserve treatment assignment, whereas exposure-based safety summaries focus on participants who received study drug to the specified extent.
16. Confidence Intervals and Precision Across Endpoints
The ALIAS registry results provide a useful demonstration that point estimates should never be interpreted without their uncertainty intervals.
| Endpoint | Estimate | Confidence interval | What the interval shows |
|---|---|---|---|
| Primary favorable outcome | RR 0.96 | 95% CI 0.84–1.10 | Moderate uncertainty around a point estimate close to 1 |
| NIHSS 0-1 at 24 hours | RR 0.99 | 99% CI 0.66–1.49 | Wide uncertainty around the relative effect |
| mRS 0-1 at 90 days | RR 1.03 | 99% CI 0.82–1.28 | Estimate close to 1 with uncertainty on both sides |
| Pulmonary edema within 48 hours | RR 10.8 | 95% CI 4.37–26.72 | Large point estimate with a wide interval |
| Death within 90 days | RR 1.10 | 95% CI 0.74–1.63 | Substantial uncertainty around a modest point estimate |
The width of a confidence interval reflects statistical precision, not clinical importance. A narrow interval around an unimportant effect can be highly precise, while a wide interval around a large point estimate may leave substantial uncertainty about the true magnitude.
17. Multiple Endpoints and Statistical Context
The ClinicalTrials.gov record contains a large collection of efficacy, functional, cognitive, and safety outcomes. The primary endpoint is explicitly identified, while the remaining reported analyses are secondary.
Primary endpoint
One binary favorable-outcome endpoint at 3 months, analyzed with logistic regression in the intent-to-treat population.
Secondary endpoints
Additional neurologic, functional, quality-of-life, cognitive, and clinical-event outcomes were analyzed separately.
Different confidence levels
The primary analysis reports a 95% CI, while many secondary binary analyses report 99% CIs.
Interpretation caution
Multiple endpoint estimates should not automatically be treated as independent confirmatory tests. The ClinicalTrials.gov record does not provide a complete multiplicity strategy.
Because the ClinicalTrials.gov record does not specify a multiplicity-control procedure for the full family of secondary endpoints, this page does not assign confirmatory status to individual secondary findings merely because a confidence interval excludes 1.
18. Analysis Populations and Missing Data
The ClinicalTrials.gov record identifies two important analysis-population rules.
| Population / rule | Registry description | Statistical implication |
|---|---|---|
| Intent-to-treat | All subjects randomized; analyzed according to randomized treatment arm | Preserves the randomized treatment comparison for efficacy |
| Trailmaking A | Participants were excluded if the Tail Making Test was not assessed | The analyzed sample can differ from the full randomized population for this endpoint |
| Safety sample | 830 subjects who received at least 20% of the intended dose of study drug | Exposure-based population differs from the ITT efficacy population |
The ClinicalTrials.gov record does not describe a general missing-data imputation method such as multiple imputation, last observation carried forward, or a specific missing-at-random model. Accordingly, no imputation procedure is attributed to the trial here.
19. Stratification and Covariate Adjustment
The primary analysis notes specify adjustment for baseline NIHSS and Thrombolysis stratum. Baseline NIHSS is directly relevant to stroke severity, while stratification variables can represent design information that should be retained in the analysis.
The ClinicalTrials.gov record supports these adjustment variables. It does not provide enough information to reconstruct every coefficient, standard error, or model diagnostic.
Adjustment should not be confused with post-randomization subgrouping. The primary analysis remains an overall treatment comparison in the intent-to-treat population, with the reported baseline and stratification variables incorporated into the model.
20. Other Design Topics: What the Registry Does and Does Not Report
| Design topic | Supported by the ClinicalTrials.gov record? | Interpretation |
|---|---|---|
| Randomization | Yes | Allocation is explicitly randomized. |
| Blinding | Yes | Masking is explicitly triple. |
| Parallel design | Yes | The design model is parallel. |
| Intention-to-treat | Yes | Used for the primary analysis and many secondary analyses. |
| Logistic regression | Yes | Primary and many secondary binary analyses use logistic regression. |
| Wilcoxon / Mann-Whitney | Yes | Used for Trailmaking A and B. |
| Non-inferiority margin | No | The ClinicalTrials.gov record identifies superiority, not a non-inferiority design. |
| Crossover | No | No crossover information is reported. |
| Factorial design | No | The design model is reported as parallel; no factorial structure is reported. |
| Bayesian methods | No | No Bayesian analysis is identified in the registry-reported methods. |
| Interim-analysis method | No | No interim-analysis procedure is reported in the ClinicalTrials.gov record. |
| Multiplicity procedure | Not specified | The data show multiple secondary endpoints but do not provide a complete multiplicity strategy. |
| Imputation method | Not specified | No general missing-data imputation procedure is reported. |
21. Why the Primary Analysis Is a Superiority Analysis
The registry identifies the hypothesis type as superiority. In a superiority framework, the question is whether the treatment groups differ in the specified direction or, depending on the formal alternative, whether the intervention provides evidence of a greater favorable outcome.
This differs fundamentally from a non-inferiority trial. In non-inferiority testing, a prespecified margin defines how much worse the experimental treatment could be while still being considered acceptably similar. No non-inferiority margin is posted on ClinicalTrials.gov for ALIAS, and the primary analysis should therefore not be interpreted through non-inferiority logic.
22. Safety Event Interpretation
The safety endpoints demonstrate another important statistical issue: relative measures can become very large when an event is uncommon, and the resulting confidence interval may still be wide.
Pulmonary edema within 48 hours
95% CI: 4.37–26.72
Logistic regression · Two-sided confidence interval
The reported risk ratio of 10.8 indicates a substantially higher estimated relative risk of the specified pulmonary-edema event in the albumin group compared with saline under the reported analysis.
The estimate does not mean that 10.8 times as many participants necessarily experienced the event, because a risk ratio compares probabilities rather than event counts alone and the underlying denominators are not posted on ClinicalTrials.gov for this endpoint.
The 95% CI of 4.37–26.72 remains above 1, while also showing considerable uncertainty about the exact magnitude of the relative effect.
Shortness of breath within 48 hours
95% CI: 1.09–6.12
Logistic regression · Two-sided confidence interval
The reported risk ratio of 2.58 means the estimated relative risk of shortness of breath within 48 hours was 2.58 times as high in the albumin group as in the saline group under the reported analysis.
The 95% CI of 1.09–6.12 is relatively wide. The point estimate therefore should not be treated as a highly precise estimate of the underlying relative risk.
23. What the Registry Results Tell Us Statistically
Several patterns are visible in the analyses posted on ClinicalTrials.gov without requiring unsupported conclusions about outcomes that are not directly reported.
| Statistical feature | What the ClinicalTrials.gov record shows |
|---|---|
| Primary efficacy estimand | Risk ratio for favorable binary outcome at 3 months |
| Primary point estimate | 0.96 |
| Primary uncertainty | 95% CI 0.84–1.10 |
| Primary model | Logistic regression |
| Primary population | Intent-to-treat |
| Primary adjustment | Baseline NIHSS and Thrombolysis stratum |
| Secondary binary outcomes | Generally analyzed with logistic regression and risk ratios |
| Continuous cognitive outcomes | Wilcoxon rank sum test with normal approximation |
| Reported p-values | 0.913 for Trailmaking A and 0.923 for Trailmaking B |
| Safety sample | 830 participants receiving at least 20% of intended dose |
The most important statistical distinction is between effect magnitude and effect precision. The primary RR of 0.96 is close to the null value of 1, while the confidence interval extends on both sides of 1. For some safety outcomes, the point estimates are considerably farther from 1, but their intervals can still be wide. Both pieces of information are necessary for a balanced statistical reading.
24. Important Limitations and Interpretation Issues
- Registry-level reporting: this analysis is restricted to the numerical and methodological information from the ClinicalTrials.gov record. It does not add unreported baseline characteristics, survival estimates, event counts, or publication-derived analyses.
- Primary p-value: the ClinicalTrials.gov record does not report a p-value for the primary favorable-outcome analysis, so none is inferred.
- Different analysis populations: the primary efficacy analysis uses intention-to-treat, while safety analyses use a dose-exposure-based safety sample.
- Endpoint multiplicity: the registry contains numerous secondary endpoints, but the ClinicalTrials.gov record does not specify a complete multiplicity-control strategy.
- Secondary confidence levels: many secondary binary endpoints use 99% confidence intervals, while the primary endpoint uses a 95% confidence interval.
- Missing data: the registry specifies an exclusion rule for Trailmaking A but does not supply a general imputation strategy.
- Rank-sum interpretation: Trailmaking estimates are rank-sum statistics and should not be interpreted as mean or median treatment differences.
- Safety interpretation: a large relative risk does not by itself provide the absolute event burden; absolute risks or event counts are needed for that interpretation.
- Multiple safety endpoints: the large collection of safety analyses means individual estimates should be interpreted in their endpoint-specific context rather than treated as isolated confirmatory tests.
25. Why This Trial Matters Statistically
ALIAS is a useful teaching case because it combines randomized clinical-trial design with several different statistical estimands and methods. The same randomized comparison generates binary neurologic outcomes, continuous cognitive outcomes, and safety-event outcomes, each requiring a different analytical representation.
| Concept | How it appears in ALIAS |
|---|---|
| Randomization | Randomized allocation in a parallel phase 3 design |
| Blinding | Triple masking |
| Intention-to-treat analysis | Primary favorable-outcome analysis uses all randomized subjects according to randomized arm |
| Binary endpoint | Favorable outcome defined by NIHSS and/or mRS thresholds |
| Logistic regression | Primary and numerous secondary binary analyses |
| Risk ratio | Reported effect measure for the binary comparisons |
| Confidence interval | 95% for the primary endpoint and 99% for many secondary endpoints |
| Covariate adjustment | Primary analysis adjusted for baseline NIHSS and Thrombolysis stratum |
| Wilcoxon / Mann-Whitney | Trailmaking A and B analyses |
| P-values | Reported for the Trailmaking analyses as 0.913 and 0.923 |
| Safety population | 830 subjects receiving at least 20% of the intended dose |
| Endpoint multiplicity | Numerous secondary efficacy and safety outcomes require careful interpretation |
26. Related Tutorials
Learn more about the methods used in this trial:
27. Related Calculators
28. Sources
- ClinicalTrials.gov: ALIAS — NCT00235495. Official registry record and source for the trial data analyzed on this page.
- PubMed: PMID 27462118.
- PubMed: PMID 26325387.
- PubMed: PMID 24076337.
- PubMed: PMID 21546491.
Continue with the underlying statistical methods
Explore the clinical-trial methods represented in ALIAS, from randomization and intention-to-treat analysis to logistic regression, risk ratios, confidence intervals, and nonparametric testing.
29. Record Summary
ALIAS provides a compact example of how statistical analysis changes with the structure of a clinical endpoint. The primary outcome was binary and was analyzed with logistic regression in the intent-to-treat population, with adjustment for baseline NIHSS and Thrombolysis stratum and a reported risk ratio of 0.96 with a 95% CI of 0.84–1.10. Secondary binary outcomes used the same broad modeling framework, while Trailmaking A and B used Wilcoxon rank sum testing with normal approximation. Safety analyses were based on a separate exposure-defined population, and the registry reports serious adverse events of 212/422 for albumin and 178/419 for saline.
The most important educational lesson is that statistical interpretation requires more than reading a point estimate. The endpoint definition, randomized analysis population, model, effect measure, confidence level, missing-data rule, and multiplicity context all affect what a reported number means. In ALIAS, a risk ratio of 0.96, a rank-sum statistic of 73098, and a safety risk ratio of 10.8 are fundamentally different quantities and cannot be interpreted on the same scale.