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
ARIES-2 was a phase 3, randomized, double-blind, placebo-controlled study designed to assess the safety and efficacy of ambrisentan in subjects with pulmonary arterial hypertension. The registry records enrollment of 186 subjects and identifies change from baseline in six-minute walk distance after 12 weeks of therapy compared with placebo as the primary endpoint.
| Feature | ARIES-2 |
|---|---|
| Study acronym | ARIES-2 |
| Phase | Phase 3 |
| Condition | Pulmonary Arterial Hypertension |
| Design | Randomized, double-blind, placebo-controlled |
| Allocation | Randomized |
| Design model | Single group |
| Primary purpose | Treatment |
| Enrollment | 186 |
| Intervention | Ambrisentan |
| Status | Completed |
| Start date | 2003-12 |
| Lead sponsor | Gilead Sciences |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT00423202 |
2. Clinical Question
The central question was whether treatment with ambrisentan produced a difference in functional exercise capacity, measured as change from baseline in six-minute walk distance after 12 weeks of therapy, compared with placebo in subjects with pulmonary arterial hypertension.
Population
Subjects with pulmonary arterial hypertension.
Intervention
Ambrisentan.
Comparator
Placebo.
Primary question
What is the difference in change from baseline in six-minute walk distance after 12 weeks of therapy compared with placebo?
3. Trial Design
The registry also identifies the design model as single group. That field sits alongside a separate description of the study as randomized and placebo-controlled. The most direct description of the clinical comparison is therefore the registry's title and allocation fields: a randomized, double-blind, placebo-controlled phase 3 study evaluating ambrisentan.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Primary endpoint | Change from baseline in the six-minute walk distance evaluated after 12 weeks of therapy compared to placebo. | After 12 weeks of therapy |
The six-minute walk distance endpoint is a continuous outcome. Unlike a time-to-event endpoint, it is not inherently summarized by a hazard ratio or Kaplan-Meier curve. The statistical comparison concerns the amount of change from baseline and the difference between treatment groups at the prespecified assessment time.
5. Statistical Methodology
Primary endpoint as a continuous outcome
The registered primary endpoint is the change from baseline in six-minute walk distance. For a continuous endpoint of this type, a typical randomized-trial analysis compares the mean change between treatment groups. The resulting treatment contrast is naturally expressed in the same units as the six-minute walk distance.
A positive value would represent a greater mean increase from baseline in the ambrisentan group under this direction of subtraction. The registry does not report an estimated treatment contrast.
Change-from-baseline analysis
Change from baseline is calculated for each subject as the post-baseline six-minute walk distance minus that subject's baseline distance. A group comparison can then be performed on these individual changes.
The primary endpoint defines the change from baseline at the 12-week assessment; the registry does not provide the individual observations or a posted statistical estimate.
ANCOVA as a common alternative
For a continuous post-treatment outcome, an analysis of covariance, or ANCOVA, is often used in randomized trials. Rather than analyzing change scores alone, ANCOVA models the 12-week measurement while adjusting for the baseline measurement. This can improve statistical efficiency when baseline values explain meaningful variation in the outcome.
Here, the treatment coefficient represents an adjusted between-group difference in the 12-week outcome. This is a description of a standard analytical approach for this type of endpoint, not a claim that the registry specifies this model for ARIES-2.
Why baseline adjustment can matter
Suppose two subjects have the same 12-week walking distance but different starting distances. Their clinical trajectories are different even though their final measurements are identical. Incorporating baseline information allows the analysis to distinguish the final outcome from the starting point.
In a randomized trial, baseline adjustment does not replace randomization. Randomization establishes the primary comparison; baseline adjustment can improve precision and account for prognostic variation that is present before treatment.
6. Statistical Methods Explained
Why is the six-minute walk distance a continuous endpoint?
The endpoint is a measured distance rather than a binary response such as success versus failure. That means information is retained across the range of observed distances. A continuous analysis can therefore estimate the average treatment-group difference rather than reducing the measurement to categories.
What does "change from baseline" mean?
It measures how much each subject's six-minute walk distance differs from their own baseline measurement. This focuses the analysis on improvement or decline over the treatment period rather than only on the absolute distance at follow-up.
Why compare change between randomized groups?
The placebo group supplies a contemporaneous estimate of change in the absence of the active intervention. The treatment effect is therefore based on the difference between the two randomized groups rather than on the change observed in the ambrisentan group alone.
Why might ANCOVA be used instead of a simple comparison of change scores?
ANCOVA can use the baseline measurement as a covariate while comparing the 12-week outcome between randomized groups. When baseline values are prognostically informative, this can reduce residual variation and produce a more precise estimate of the treatment difference.
What would a confidence interval mean for this endpoint?
A confidence interval around the treatment-group difference would describe statistical uncertainty surrounding the estimated mean difference. A narrow interval would indicate greater precision, whereas a wide interval would indicate less precision. It would not describe the range of individual patient responses.
Why does a p-value not measure the size of the treatment effect?
A p-value addresses compatibility of the observed data with a specified null hypothesis under the statistical model. It does not tell the reader how large or clinically important the treatment difference is. For a continuous endpoint, the estimated difference and its confidence interval are essential for understanding magnitude and precision.
7. Planned Analysis
The registry identifies one registered primary endpoint: change from baseline in the six-minute walk distance evaluated after 12 weeks of therapy compared to placebo. Results have not been posted on ClinicalTrials.gov.
What would normally be estimated?
For this continuous endpoint, a standard analysis would estimate the between-group difference in mean change from baseline. An adjusted analysis such as ANCOVA could instead estimate the treatment-group difference in the 12-week outcome while accounting for baseline distance.
| Statistical quantity | Typical interpretation |
|---|---|
| Mean change in each group | Average within-subject change from baseline at the 12-week assessment. |
| Between-group mean difference | Difference in average change between ambrisentan and placebo. |
| 95% confidence interval | Uncertainty around the estimated treatment-group difference. |
| P-value | Evidence against a prespecified null hypothesis of no treatment difference, under the chosen statistical model. |
What the analysis would not establish by itself
A statistically significant mean difference would not establish that every subject improved. Likewise, a nonsignificant result would not establish that the treatments are equivalent. The distribution of individual responses, the magnitude of the estimated difference, and the confidence interval would all contribute to interpretation.
8. Interpreting a Continuous Treatment Effect
For a continuous endpoint, the most direct effect measure is an absolute difference between groups. This is different from the hazard ratio used for survival analysis or an odds ratio used for binary outcomes.
If the estimated mean change were expressed as a difference between ambrisentan and placebo, the quantity would retain the original units of the six-minute walk distance. Its sign would indicate the direction of the estimated difference, while its magnitude would describe the average separation between groups.
A confidence interval would indicate the statistical precision of the estimated treatment difference. It should be interpreted as uncertainty around the population-level effect estimate, not as a prediction interval for where an individual subject's response will fall.
The p-value would address the null hypothesis specified for the primary comparison. It should not be used as a substitute for reporting the treatment difference itself, because statistical significance and effect magnitude are different concepts.
9. Analysis Population and Missing Data
The registry record identifies an enrollment of 186 subjects but does not provide an analysis-population definition for the primary endpoint in the information summarized here.
For a 12-week continuous endpoint, missing follow-up measurements can affect the treatment comparison if the subjects with missing assessments differ systematically from those with observed assessments. A prespecified statistical analysis plan would normally define which subjects contribute to the primary analysis and how missing 12-week measurements are handled.
Observed outcomes
Subjects with a recorded 12-week assessment can contribute directly to a comparison of their observed change from baseline.
Missing outcomes
The treatment effect can depend on assumptions about subjects whose 12-week outcome is unavailable.
Common approaches include model-based analyses that use available longitudinal information, prespecified imputation procedures, or sensitivity analyses under alternative missing-data assumptions. The appropriate choice depends on the protocol and statistical analysis plan; the registry information does not specify one.
10. Safety and Secondary Endpoints
The study title states that ARIES-2 was designed to assess safety and efficacy. However, the registered ClinicalTrials.gov record identifies only one primary endpoint and does not provide posted safety results, serious adverse-event counts by treatment group, or secondary endpoint results.
| Evidence category | Registry information available |
|---|---|
| Primary endpoint | Change from baseline in six-minute walk distance after 12 weeks of therapy compared to placebo. |
| Secondary endpoint results | Not reported in the registry information. |
| Serious adverse events by arm | Not reported in the registry information. |
| Posted statistical analyses | None. |
Safety and efficacy are statistically distinct questions. An efficacy analysis asks whether the randomized groups differ on the prespecified clinical outcome. A safety analysis instead describes adverse events and other safety measures, often with event counts and percentages by treatment exposure. Those quantities should not be substituted for one another.
11. Design Features That Affect Statistical Interpretation
Randomization
Randomized allocation is important because treatment assignment is determined independently of subjects' baseline prognostic characteristics, apart from chance imbalance. This supports a causal comparison of the randomized groups when the trial is analyzed according to assignment and the protocol is appropriately followed.
Double-blinding
Blinding can reduce differential expectations and behavior between treatment groups. For a functional outcome such as six-minute walk distance, this is particularly relevant because performance-based measurements can potentially be influenced by factors other than the underlying physiological state.
Placebo control
A placebo-controlled design provides a contemporaneous reference for the change observed during the same study period. The treatment comparison is therefore not simply "change versus no change"; it is the difference between randomized treatment groups.
Multiplicity
The registry information identifies one registered primary endpoint. It does not report a multiplicity strategy, multiple-primary-endpoint hierarchy, or additional confirmatory testing framework. If additional endpoints or repeated hypothesis tests were prespecified elsewhere, their interpretation would depend on the corresponding statistical analysis plan.
Interim analysis
The registry information does not report an interim efficacy analysis or an alpha-spending procedure. Without such documentation, no interim-monitoring method should be attributed to the trial.
Bayesian methods
The registry information does not report Bayesian statistical methods. A conventional analysis of the registered continuous endpoint would ordinarily be framed through an estimated treatment difference, confidence interval, and hypothesis test rather than a posterior probability, but the specific analysis plan is not reported here.
12. Why Baseline Adjustment Is a Useful Statistical Teaching Point
ARIES-2 provides a useful example of why baseline measurement matters when the primary endpoint is a change in a continuous functional measure.
| Approach | What it compares | Statistical interpretation |
|---|---|---|
| Change-score analysis | Mean individual change between randomized groups | Directly targets the registered change-from-baseline endpoint. |
| ANCOVA | 12-week outcome between groups, adjusted for baseline | Targets the between-group difference at follow-up after accounting for baseline variation. |
| Unadjusted post-treatment comparison | 12-week measurements alone | Can ignore prognostic information contained in baseline measurements. |
These methods answer closely related questions but are not numerically identical in every dataset. The choice should therefore be prespecified rather than selected after looking at the results.
13. Practical Interpretation of the Six-Minute Walk Endpoint
The six-minute walk distance is measured on a continuous scale, so the statistical analysis preserves more information than a binary responder definition would. A subject who changes by a small amount and a subject who changes by a large amount remain distinguishable in the analysis.
The principal estimand is a population-level comparison of change. That distinction matters because an average treatment difference does not imply that every subject experiences the same response. Individual trajectories can vary substantially around the group mean.
Group-level question
Did the randomized groups differ, on average, in change from baseline after 12 weeks?
Individual-level question
How much did a particular subject improve or worsen?
Precision question
How precisely can the population-level treatment difference be estimated?
Clinical question
Is the magnitude of the observed difference meaningful in the clinical context?
These questions should remain conceptually separate. Statistical evidence addresses uncertainty about the treatment comparison, while clinical interpretation requires consideration of the magnitude and relevance of the measured change.
14. Limitations
- No posted efficacy analysis: the ClinicalTrials.gov record does not contain a formal statistical analysis for the registered primary endpoint.
- No treatment estimate: an estimated difference in six-minute walk distance, confidence interval, and p-value are not reported in the registry information.
- Analysis population not specified: the registry reports enrollment of 186 but does not provide the primary analysis population definition in the available record information.
- Missing-data strategy not specified: the registry information does not identify an imputation or model-based missing-data procedure.
- Multiplicity strategy not specified: the registry information does not describe alpha allocation or adjustment for multiple hypotheses.
- Interim monitoring not specified: no interim analysis or alpha-spending approach is reported.
- Secondary results unavailable: the registry information does not provide secondary endpoint estimates or safety event counts by treatment group.
- Design-model field: the registry identifies the design model as single group while separately describing the study as randomized and placebo-controlled; this distinction should be retained rather than silently replacing one registry field with another.
15. Why This Trial Matters Statistically
ARIES-2 is a useful teaching case because its registered primary endpoint illustrates a common clinical-trial problem: how to analyze a continuous outcome measured both before and after treatment.
| Concept | How it appears in ARIES-2 |
|---|---|
| Randomization | The registry identifies randomized allocation. |
| Double-blinding | The registry identifies a double-blind study. |
| Placebo control | The study title identifies a placebo-controlled design. |
| Continuous endpoint | Six-minute walk distance is measured on a continuous scale. |
| Baseline adjustment | The primary endpoint is defined as change from baseline. |
| Between-group comparison | The endpoint is evaluated compared with placebo. |
| Effect estimation | A natural effect measure is a difference in mean change. |
| Confidence intervals | Would quantify uncertainty around a continuous treatment difference. |
| Missing data | Loss of the 12-week measurement can affect a change-from-baseline analysis. |
| Clinical interpretation | The magnitude of a continuous treatment difference matters in addition to statistical significance. |
The statistical lesson is straightforward but important: a trial's endpoint determines the natural scale of its primary treatment effect. For ARIES-2, the registered endpoint points toward a continuous-data framework rather than a survival-analysis framework.
16. Trial Timeline
Study start
The registry lists the study start date as 2003-12.
Randomized controlled study
The study is described as a phase 3, randomized, double-blind, placebo-controlled study of ambrisentan in subjects with pulmonary arterial hypertension.
Primary assessment
The registered primary endpoint evaluates change from baseline in six-minute walk distance after 12 weeks of therapy compared with placebo.
Study status
The registry lists the study status as completed.
17. What a Complete Primary Analysis Would Report
For a continuous primary endpoint such as change from baseline in six-minute walk distance, a statistically informative results section would normally place the estimated treatment effect at the center of the presentation.
| Result component | Purpose |
|---|---|
| Sample size analyzed | Defines the population contributing to the primary comparison. |
| Mean change by treatment group | Describes the observed direction and magnitude of within-group change. |
| Between-group difference | Provides the principal estimate of treatment effect. |
| 95% confidence interval | Shows statistical precision around the estimated difference. |
| P-value | Addresses the prespecified hypothesis test. |
| Missing-data handling | Shows how unavailable 12-week measurements were incorporated into the analysis. |
Presenting all of these elements together is preferable to reporting a p-value alone. The treatment difference answers "how much?", the confidence interval addresses "how precisely?", and the p-value addresses the specified hypothesis test.
18. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registered primary endpoint is a continuous change-from-baseline outcome. Its natural treatment effect is a between-group difference in mean change, potentially estimated with baseline adjustment.
Clinical interpretation
The clinical meaning of any observed difference would depend on its magnitude, precision, the context of pulmonary arterial hypertension, and the relationship between six-minute walk performance and patient-relevant benefit.
These interpretations should not be collapsed into a single numerical conclusion. Statistical significance does not establish clinical importance, while clinical importance cannot be determined from a p-value alone.
19. Important Questions for Reading the Primary Result
Was the analysis based on randomized assignment?
For a randomized comparison, analyzing subjects according to their assigned treatment generally preserves the principal advantage of randomization. The registry information does not specify the analysis population used for ARIES-2.
Was baseline distance included in the model?
Baseline adjustment can improve precision for a continuous endpoint. Whether ANCOVA or another baseline-adjusted model was prespecified is not reported in the registry information.
How were missing 12-week measurements handled?
This is important because the endpoint requires both baseline and post-treatment information. Different assumptions about missing outcomes can produce different estimates. The registry information does not specify a missing-data method.
Was the analysis clinically meaningful as well as statistically significant?
The answer would require examining the estimated magnitude of the difference and its uncertainty, then considering its clinical context. A statistically detectable difference is not automatically a clinically important difference.
20. Record Summary
ARIES-2 was a completed phase 3 study of ambrisentan in subjects with pulmonary arterial hypertension, with randomized allocation, double-blinding, and placebo control identified in the registry. Enrollment was 186, and the registered primary endpoint was change from baseline in the six-minute walk distance evaluated after 12 weeks of therapy compared to placebo.
Statistically, this endpoint belongs to the continuous-outcome family. The central treatment contrast would ordinarily be expressed as a difference in mean change, with a confidence interval describing precision and a hypothesis test addressing the prespecified null hypothesis. Baseline-adjusted methods such as ANCOVA are common for this class of endpoint, although the registry information does not specify the analytical model used for ARIES-2.
The ClinicalTrials.gov record does not post a formal statistical analysis for the primary endpoint, and it does not provide treatment estimates, confidence intervals, p-values, secondary endpoint results, or serious adverse-event counts by treatment group in the information available here. The resulting statistical lesson is therefore methodological: understanding how a randomized trial should analyze a continuous change-from-baseline endpoint, while keeping the distinction between a planned analytical framework and an observed treatment result clear.