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 was a randomized, double-blind, parallel-group phase 3 trial evaluating the efficacy and safety of the beta-3 agonist mirabegron (YM178) in patients with symptoms of overactive bladder. The registry reports 2,149 enrolled participants and three study arms: placebo, mirabegron 50 mg, and mirabegron 100 mg.
| Feature | ARIES |
|---|---|
| Phase | Phase 3 |
| Status | Completed |
| Condition | Urinary Bladder, Overactive |
| Design | Randomized, double-blind, parallel |
| Primary purpose | Treatment |
| Enrollment | 2,149 |
| Study period | March 28, 2008 to April 22, 2009 |
| Lead sponsor | Astellas Pharma Inc |
| Primary endpoints | 2 |
| Posted outcome measures | 31 |
| Posted statistical analyses | 10 |
2. Clinical Question
The central question was whether mirabegron improves measures of urinary frequency and incontinence over the treatment period compared with placebo in patients with symptoms of overactive bladder.
Population
Patients with symptoms of overactive bladder, represented in the registry under the condition "Urinary Bladder, Overactive."
Intervention
Mirabegron, evaluated at 50 mg and 100 mg in the posted analyses.
Comparator
Placebo.
Primary question
Does mirabegron produce greater improvement from baseline in mean daily incontinence episodes and micturitions than placebo?
3. Trial Design
Placebo
The placebo arm served as the reference group for the posted efficacy comparisons.
Mirabegron 50 mg
The 50 mg dose was compared with placebo for both coprimary endpoints and the posted secondary endpoints.
Mirabegron 100 mg
The 100 mg dose was compared with placebo for both coprimary endpoints and the posted secondary endpoints.
The design is particularly well suited to an ANCOVA-based comparison because the principal efficacy outcomes are continuous changes from baseline. Rather than simply comparing raw Week 12 means, the registered analyses used baseline as a covariate and incorporated treatment group, gender, and geographic region in the ANCOVA specification for the incontinence endpoint.
4. Endpoints
Primary Endpoint 1: Incontinence Episodes
| Element | Registry definition |
|---|---|
| Endpoint | Change From Baseline to End of Treatment (Final Visit) in Mean Number of Incontinence Episodes Per 24 Hours |
| Time frame | Baseline and Week 12 (Final Visit) |
| Measurement | Average number of incontinence episodes per day, derived from patient-recorded episodes in a micturition diary for 3-days before the Baseline and Week 12 clinic visits |
| Analysis | ANCOVA; stratified rank ANCOVA for the posted hypothesis-testing analysis |
Primary Endpoint 2: Micturitions
| Element | Registry definition |
|---|---|
| Endpoint | Change From Baseline to End of Treatment (Final Visit) in Mean Number of Micturitions Per 24 Hours |
| Time frame | Baseline and Week 12 |
| Measurement | Average number of micturitions per 24 hours, excluding incontinence-only episodes, derived from the 3-day micturition diary before the Baseline and Week 12 clinic visits |
| Analysis | ANCOVA |
Both endpoints are continuous change-from-baseline outcomes. That matters statistically because the effect measure is a least-squares mean difference rather than a hazard ratio, odds ratio, or risk ratio. Negative estimates for the incontinence and micturition endpoints indicate a lower adjusted mean change in the mirabegron group relative to placebo.
5. Statistical Methodology
ANCOVA as the principal analysis
The registry reports ANCOVA as the statistical method across the posted analyses. For the incontinence endpoint, the analysis used a stratified rank ANCOVA for hypothesis testing, with the LS mean difference derived from an ANCOVA model containing treatment group, gender, and geographic region as fixed factors and baseline as a covariate. For micturitions, the registry reports ANCOVA for hypothesis testing.
The purpose of baseline adjustment is to compare treatment groups after accounting statistically for the baseline value of the outcome. The resulting least-squares means represent model-adjusted means rather than simple arithmetic means.
Least-squares mean difference
The primary effect measure was the LS Mean Difference. This is the modeled difference between treatment groups after the covariate and fixed-factor adjustment specified in the ANCOVA model.
For incontinence episodes and micturitions, a negative value indicates fewer episodes or micturitions in the mirabegron group relative to placebo after model adjustment. For volume voided per micturition, a positive value indicates a larger volume in the mirabegron group.
Stratified rank ANCOVA
The registry specifically reports a stratified rank ANCOVA for the incontinence endpoint. A rank-based approach can reduce reliance on the assumptions associated with directly modeling the original continuous measurements when the outcome distribution is not well behaved for a conventional parametric analysis. The registry nevertheless reports the treatment effect itself as an LS mean difference derived from an ANCOVA model.
Analysis populations and LOCF
The full analysis set for the micturition endpoints included all randomized patients who took at least 1 dose of double-blind study drug and had both a baseline and at least 1 post-baseline micturition measurement in the visit diary. For the incontinence endpoint, the full analysis set-incontinence additionally required at least 1 incontinence episode at baseline.
Last observation carried forward (LOCF) was used in the primary Week 12 analyses. The Week 4 secondary analyses explicitly state that LOCF was not used.
Multiplicity control
The registry states that there were 2 coprimary efficacy endpoints and 3 key secondary efficacy endpoints. Multiplicity among these endpoints was controlled at a type I error rate of alpha = 0.05 using a stepwise parallel gatekeeping procedure with 5 stages. In addition, because two mirabegron treatment groups were compared with placebo per endpoint, the Hochberg procedure was used to adjust for multiplicity. An ANCOVA model was used for hypothesis testing.
6. Primary Results: Incontinence Episodes
The first coprimary endpoint was the change from baseline to the end of treatment at the Week 12 final visit in mean number of incontinence episodes per 24 hours.
Mirabegron 50 mg vs Placebo
LS mean difference in change from baseline
95% CI: −0.66 to −0.03 · P = 0.026
Stratified rank ANCOVA; two-sided 95% confidence interval; superiority hypothesis.
The estimated adjusted difference was −0.34 incontinence episodes per 24 hours, meaning the modeled change from baseline was lower with mirabegron 50 mg than with placebo by 0.34 episodes per day. Because the endpoint is a change score, the estimate represents a between-group difference in improvement rather than the number of episodes experienced by an individual patient.
The 95% confidence interval extends from −0.66 to −0.03. Its width reflects uncertainty around the estimated treatment difference, and the fact that the interval remains below zero is consistent with the reported superiority test. The interval does not describe the range of treatment effects that individual patients would experience.
The p-value of 0.026 addresses the strength of evidence against the relevant null hypothesis within the prespecified statistical framework. It does not measure the magnitude or clinical importance of the effect. Effect size and precision are better understood from the LS mean difference and its confidence interval.
Because the analysis used a stratified rank ANCOVA and LOCF, interpretation should remain tied to that analysis population and missing-data approach rather than being generalized to an alternative analysis model.
Mirabegron 100 mg vs Placebo
LS mean difference in change from baseline
95% CI: −0.82 to −0.18 · P < 0.001
Stratified rank ANCOVA; two-sided 95% confidence interval; superiority hypothesis.
The estimated adjusted difference was −0.50 incontinence episodes per 24 hours for mirabegron 100 mg relative to placebo. The negative direction indicates a greater reduction in the modeled mean number of daily incontinence episodes with the 100 mg dose.
The 95% confidence interval, −0.82 to −0.18, lies entirely below zero. This indicates that the estimated treatment difference is separated from zero under the model and analysis framework used for the registry result. It does not imply that every patient would experience a reduction of exactly 0.50 episodes.
The reported P < 0.001 is evidence against the null hypothesis in the prespecified testing framework. It should not be interpreted as a probability that the treatment effect is real, nor as a measure of the size of the treatment effect.
As with the 50 mg comparison, the estimate depends on the specified analysis population, baseline adjustment, fixed factors, stratified rank ANCOVA methodology, and LOCF handling.
| Comparison | LS mean difference | 95% CI | P-value |
|---|---|---|---|
| Mirabegron 50 mg vs placebo | −0.34 | −0.66 to −0.03 | 0.026 |
| Mirabegron 100 mg vs placebo | −0.50 | −0.82 to −0.18 | <0.001 |
7. Primary Results: Micturitions
The second coprimary endpoint was the change from baseline to Week 12 in the mean number of micturitions per 24 hours.
Mirabegron 50 mg vs Placebo
LS mean difference in change from baseline
95% CI: −0.98 to −0.24 · P = 0.001
ANCOVA; two-sided 95% confidence interval; superiority hypothesis.
The adjusted treatment difference was −0.61 micturitions per 24 hours, indicating a lower modeled change in daily micturition frequency with mirabegron 50 mg than with placebo.
The 95% CI of −0.98 to −0.24 quantifies uncertainty around the estimated between-group difference. Because the entire interval is below zero, the estimate is directionally consistent with fewer daily micturitions under mirabegron 50 mg in this analysis.
The p-value of 0.001 indicates strong evidence against the corresponding null hypothesis within the trial's multiplicity-controlled testing framework. It does not quantify how clinically meaningful a reduction of 0.61 micturitions is for an individual patient.
Mirabegron 100 mg vs Placebo
LS mean difference in change from baseline
95% CI: −1.07 to −0.33 · P < 0.001
ANCOVA; two-sided 95% confidence interval; superiority hypothesis.
The estimated adjusted difference was −0.70 micturitions per 24 hours for mirabegron 100 mg relative to placebo. The negative estimate indicates a greater reduction in the modeled daily micturition count.
The 95% CI of −1.07 to −0.33 gives the uncertainty range around the estimated treatment difference under the specified ANCOVA model. It is not a prediction interval for individual patients.
The reported P < 0.001 supports a statistically detectable treatment difference under the prespecified testing framework. A small p-value should not be confused with a large treatment effect; the LS mean difference and its confidence interval provide the information about magnitude and precision.
The analysis used the full analysis set and LOCF, so its interpretation depends on those analysis choices and should not automatically be transferred to a different missing-data strategy.
| Comparison | LS mean difference | 95% CI | P-value |
|---|---|---|---|
| Mirabegron 50 mg vs placebo | −0.61 | −0.98 to −0.24 | 0.001 |
| Mirabegron 100 mg vs placebo | −0.70 | −1.07 to −0.33 | <0.001 |
8. Secondary Endpoint Results
Volume Voided Per Micturition
The registry reports change from baseline to the final visit in mean volume voided per micturition as a secondary endpoint, measured in milliliters.
| Comparison | LS mean difference | 95% CI | P-value |
|---|---|---|---|
| Mirabegron 50 mg vs placebo | 11.1 mL | 4.4 to 17.9 | 0.001 |
| Mirabegron 100 mg vs placebo | 11.0 mL | 4.2 to 17.7 | 0.002 |
50 mg
The LS mean difference was 11.1 mL, with a 95% CI of 4.4 to 17.9 mL and P = 0.001.
100 mg
The LS mean difference was 11.0 mL, with a 95% CI of 4.2 to 17.7 mL and P = 0.002.
Unlike the incontinence and micturition endpoints, the volume endpoint has a positive effect estimate. Here, a positive LS mean difference indicates a greater adjusted increase in mean volume voided per micturition relative to placebo. The confidence intervals for both dose comparisons remain above zero.
Week 4 Incontinence Episodes
| Comparison | LS mean difference | 95% CI | P-value |
|---|---|---|---|
| Mirabegron 50 mg vs placebo | −0.48 | −0.80 to −0.15 | 0.003 |
| Mirabegron 100 mg vs placebo | −0.46 | −0.79 to −0.13 | <0.001 |
These Week 4 analyses used the full analysis set-incontinence and a stratified rank ANCOVA. Unlike the primary Week 12 incontinence analysis, the registry states that LOCF was not used for these Week 4 analyses.
Week 4 Micturitions
| Comparison | LS mean difference | 95% CI | P-value |
|---|---|---|---|
| Mirabegron 50 mg vs placebo | −0.42 | −0.77 to −0.06 | 0.022 |
| Mirabegron 100 mg vs placebo | −0.60 | −0.96 to −0.24 | 0.001 |
The Week 4 micturition analyses used ANCOVA and did not use LOCF. Both estimates were negative, indicating lower modeled micturition counts with mirabegron than placebo at this assessment point.
9. Statistical Results in One View
| Endpoint | Comparison | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Incontinence episodes, Week 12 | 50 mg vs placebo | −0.34 | −0.66 to −0.03 | 0.026 |
| Incontinence episodes, Week 12 | 100 mg vs placebo | −0.50 | −0.82 to −0.18 | <0.001 |
| Micturitions, Week 12 | 50 mg vs placebo | −0.61 | −0.98 to −0.24 | 0.001 |
| Micturitions, Week 12 | 100 mg vs placebo | −0.70 | −1.07 to −0.33 | <0.001 |
| Volume voided, Week 12 | 50 mg vs placebo | 11.1 mL | 4.4 to 17.9 | 0.001 |
| Volume voided, Week 12 | 100 mg vs placebo | 11.0 mL | 4.2 to 17.7 | 0.002 |
| Incontinence episodes, Week 4 | 50 mg vs placebo | −0.48 | −0.80 to −0.15 | 0.003 |
| Incontinence episodes, Week 4 | 100 mg vs placebo | −0.46 | −0.79 to −0.13 | <0.001 |
| Micturitions, Week 4 | 50 mg vs placebo | −0.42 | −0.77 to −0.06 | 0.022 |
| Micturitions, Week 4 | 100 mg vs placebo | −0.60 | −0.96 to −0.24 | 0.001 |
Across the posted analyses, the direction of the estimates is consistent with fewer daily incontinence episodes and micturitions with mirabegron, while the volume voided per micturition estimates are positive. The numerical magnitudes differ by endpoint and dose, so the estimates should not be treated as interchangeable measures of one underlying effect.
10. Statistical Methods Explained
Why was ANCOVA used?
ANCOVA is well suited to a continuous outcome measured at baseline and again after treatment. By including baseline as a covariate, the analysis compares treatment groups after accounting for baseline outcome level. This can improve precision compared with a simple comparison that ignores baseline measurements. In ARIES, the registry also specified treatment group, gender, and geographic region as fixed factors for the ANCOVA used to derive the incontinence LS mean difference.
What does an LS mean difference of −0.50 mean?
For the Week 12 incontinence endpoint, an LS mean difference of −0.50 for mirabegron 100 mg versus placebo means that the model-estimated change from baseline was 0.50 incontinence episodes per 24 hours lower in the mirabegron group than in the placebo group. It is an adjusted group-level estimate, not a statement that each patient had exactly 0.50 fewer episodes.
Why are some estimates negative and the volume estimate positive?
The sign reflects the direction of the outcome. Fewer incontinence episodes and fewer micturitions correspond to negative treatment-minus-placebo differences. For volume voided per micturition, an increase corresponds to a positive difference. Therefore, the sign cannot be interpreted without knowing which endpoint is being measured.
Why does the confidence interval matter?
The confidence interval communicates precision around the estimated treatment difference. For example, the Week 12 100 mg incontinence estimate was −0.50 with a 95% CI of −0.82 to −0.18. The interval indicates that the estimate is uncertain rather than exact. It does not describe the range of responses among individual patients.
Why was multiplicity adjustment needed?
ARIES had 2 coprimary efficacy endpoints, 3 key secondary efficacy endpoints, and two mirabegron-versus-placebo comparisons per endpoint. If every hypothesis were tested independently at 0.05, the chance of at least one false-positive result across the family would be higher than 0.05. The registry therefore reports a stepwise parallel gatekeeping procedure and Hochberg adjustment to control multiplicity. An ANCOVA model was used for hypothesis testing.
What does LOCF change about interpretation?
LOCF replaces a missing later measurement with the patient's most recent available observation. The primary Week 12 analyses used LOCF, whereas the Week 4 secondary analyses did not. Because the handling of missing observations can affect estimates, the reported results are specific to the analysis strategy used in the registry.
11. Missing Data and Analysis Populations
The registry's analysis populations are narrower than a pure all-randomized intention-to-treat definition. For the full analysis set, patients had to be randomized, receive at least 1 dose of double-blind study drug, and have baseline and post-baseline micturition measurements. The incontinence-specific population also required at least 1 incontinence episode at baseline.
| Endpoint | Analysis population | Missing-data approach |
|---|---|---|
| Week 12 incontinence | Full analysis set-incontinence; randomized, at least 1 dose, baseline and at least 1 post-baseline diary measurement, at least 1 baseline incontinence episode | LOCF used |
| Week 12 micturitions | Full analysis set; randomized, at least 1 dose, baseline and at least 1 post-baseline diary measurement | LOCF used |
| Week 4 incontinence | Full analysis set-incontinence | LOCF not used |
| Week 4 micturitions | Full analysis set | LOCF not used |
| Week 12 volume voided | Full analysis set | LOCF used |
The difference between the Week 12 and Week 4 missing-data strategies is important. A reader should not assume that all ARIES analyses used the same imputation convention. The primary analyses used LOCF, while the posted Week 4 analyses did not. Consequently, comparisons across time points should be understood as comparisons of analyses with different handling of missing observations.
12. Multiplicity: Gatekeeping and Hochberg Adjustment
The registry provides unusually useful detail about multiplicity. There were two coprimary efficacy endpoints and three key secondary efficacy endpoints. The familywise type I error rate was controlled at the alpha = 0.05 level through a stepwise parallel gatekeeping procedure with five stages.
Within each endpoint, two active-dose comparisons were made against placebo. The registry states that the Hochberg procedure was used to adjust for this multiplicity. An ANCOVA model was used for hypothesis testing.
Gatekeeping
A gatekeeping procedure creates a prespecified sequence for moving statistical testing across families of endpoints while maintaining the desired overall type I error rate.
Hochberg
The Hochberg procedure provides multiplicity adjustment for the two mirabegron dose comparisons with placebo within an endpoint family.
Alpha = 0.05
The registry states that the overall type I error was controlled at the 0.05 level for the endpoint family.
Why this changes interpretation
The p-values should be read as components of the prespecified multiple-testing strategy rather than as ten unrelated hypothesis tests.
13. Safety
The registry reports serious adverse events by treatment arm. The available figures are affected patients divided by patients at risk within each arm.
| Arm | Serious adverse events | Affected / at risk |
|---|---|---|
| Placebo | Serious adverse events | 9 / 453 |
| Mirabegron 50 mg | Serious adverse events | 11 / 442 |
| Mirabegron 100 mg | Serious adverse events | 14 / 433 |
These counts are descriptive safety results. They should not be interpreted as a formal hypothesis test or converted into a comparative safety conclusion without the corresponding prespecified safety analysis, exposure definitions, follow-up, and statistical framework.
The denominators differ across the three reported groups: 453 for placebo, 442 for mirabegron 50 mg, and 433 for mirabegron 100 mg. A simple comparison of the numerators alone would therefore be misleading. Safety interpretation should use the affected and at-risk counts together and should account for the fact that safety outcomes are generally analyzed differently from efficacy endpoints.
14. Dose Comparisons and Statistical Interpretation
The posted results allow the two mirabegron doses to be compared descriptively through their separate placebo-controlled estimates. They do not, by themselves, constitute a direct statistical test of 50 mg versus 100 mg.
| Endpoint | 50 mg estimate | 100 mg estimate |
|---|---|---|
| Week 12 incontinence episodes | −0.34 | −0.50 |
| Week 12 micturitions | −0.61 | −0.70 |
| Week 12 volume voided | 11.1 mL | 11.0 mL |
| Week 4 incontinence episodes | −0.48 | −0.46 |
| Week 4 micturitions | −0.42 | −0.60 |
For example, the Week 12 incontinence estimate is numerically more negative for 100 mg than for 50 mg, whereas the Week 12 volume-voided estimates are very similar. But the separate confidence intervals and p-values are not a substitute for a formal direct dose-comparison test. A descriptive difference between two treatment estimates should therefore not be labeled a statistically significant dose effect unless a corresponding comparison has been performed.
15. Why the Diary Structure Matters Statistically
Both coprimary endpoints were derived from patient-recorded micturition diaries covering 3 days before the relevant clinic visit. The endpoint is therefore not a single observation taken at the clinic; it is an average constructed from repeated daily observations within a short diary period.
Incontinence
The endpoint counts involuntary leakage episodes recorded by the patient and expresses them as an average number per day.
Micturition
The endpoint counts urinations while excluding episodes classified as incontinence-only episodes.
Baseline
The baseline value was incorporated into the ANCOVA as a covariate rather than ignored in the primary analysis.
Week 12
The primary efficacy assessment used the diary information recorded before the Week 12 final visit.
This construction also explains why the effect estimates should be described as differences in mean daily counts. They are not hazard ratios, event probabilities, or percentages of patients achieving a response.
16. What the P-Values Do — and Do Not — Mean
A p-value quantifies the compatibility of the observed data with a specified null hypothesis under the statistical model and testing procedure. In ARIES, the registry reports p-values after a multiplicity-control strategy that accounts for the endpoint family and the two active-dose comparisons.
A p-value is not a measure of effect size. For example, P < 0.001 for the Week 12 micturition comparison at 100 mg does not mean that the treatment effect is "larger" than an effect associated with P = 0.026. Effect magnitude is described by the LS mean difference, while precision is described by the confidence interval.
17. What the Confidence Intervals Do — and Do Not — Mean
The primary results provide two-sided 95% confidence intervals for each LS mean difference. These intervals are useful because they present both direction and precision rather than reducing the evidence to a binary p-value threshold.
| Result | Estimate | 95% CI | Interpretive direction |
|---|---|---|---|
| Incontinence, 50 mg | −0.34 | −0.66 to −0.03 | Lower daily incontinence count |
| Incontinence, 100 mg | −0.50 | −0.82 to −0.18 | Lower daily incontinence count |
| Micturitions, 50 mg | −0.61 | −0.98 to −0.24 | Lower daily micturition count |
| Micturitions, 100 mg | −0.70 | −1.07 to −0.33 | Lower daily micturition count |
None of these confidence intervals should be read as a range containing 95% of individual patient responses. They quantify uncertainty around the estimated group-level treatment difference under the corresponding statistical framework.
18. Stratification and Covariate Adjustment
The posted incontinence analysis is described as a stratified rank ANCOVA. The registry states that the LS mean difference was derived from an ANCOVA model with treatment group, gender, and geographic region as fixed factors and baseline as a covariate.
These terms separate the treatment comparison from variation associated with baseline measurement and the specified fixed factors. The model therefore estimates a treatment contrast after adjustment rather than simply subtracting two unadjusted arithmetic means.
It is important not to confuse this analysis adjustment with a claim that gender or geographic region were randomization strata. The registry information supports their inclusion as fixed factors in the ANCOVA model; it does not establish that they were used as randomization strata.
19. Secondary Endpoints and Endpoint Hierarchy
The registry identifies two primary endpoints and reports additional secondary analyses. The multiplicity strategy is especially relevant because secondary endpoints were incorporated into the five-stage gatekeeping procedure.
| Endpoint family | Examples posted | Statistical approach |
|---|---|---|
| Primary | Incontinence episodes; micturitions | ANCOVA / stratified rank ANCOVA |
| Secondary | Volume voided per micturition | ANCOVA |
| Secondary | Week 4 incontinence episodes | Stratified rank ANCOVA |
| Secondary | Week 4 micturitions | ANCOVA |
The posted results provide a coherent statistical framework across the efficacy outcomes: continuous diary-derived endpoints, placebo-controlled dose comparisons, baseline-adjusted LS mean differences, and explicit multiplicity control.
20. Important Limitations and Interpretation Issues
- Analysis population: the full analysis sets required treatment exposure and post-baseline measurements, so they are not identical to a strict all-randomized intention-to-treat population.
- LOCF: the primary Week 12 analyses used last observation carried forward, an approach that can affect results when observations are missing.
- Different Week 4 handling: the posted Week 4 analyses explicitly state that LOCF was not used, so the Week 4 and Week 12 analyses do not use identical missing-data conventions.
- Patient-reported diary data: the primary outcomes were derived from patient-recorded 3-day micturition diaries, so measurement depends on diary completion and recording.
- Multiple endpoints: the trial evaluated multiple efficacy endpoints and two active doses, making the reported multiplicity-control strategy important for interpretation.
- Direct dose comparison: separate comparisons against placebo do not establish that 50 mg and 100 mg differ significantly from one another.
- Clinical magnitude: statistical evidence that an adjusted difference is nonzero does not by itself determine whether the magnitude is clinically meaningful to an individual patient.
- Safety denominators: the reported serious-adverse-event counts use different at-risk denominators across arms and should not be interpreted from numerators alone.
21. Why This Trial Matters Statistically
ARIES is a useful teaching example because it demonstrates how a clinical trial with relatively straightforward continuous endpoints can require several layers of statistical planning. The central analysis is not simply a comparison of two means. It combines randomization, double blinding, baseline adjustment, fixed-factor adjustment, rank-based hypothesis testing for selected endpoints, multiple treatment comparisons, endpoint-level multiplicity control, and explicit missing-data conventions.
| Concept | How it appears in ARIES |
|---|---|
| Randomization | Randomized parallel-group phase 3 design |
| Blinding | Double-blind design |
| ANCOVA | Primary normalized analysis method for continuous efficacy outcomes |
| Covariate adjustment | Baseline incorporated as a covariate; gender and geographic region included as fixed factors in the incontinence ANCOVA |
| LS mean difference | Primary effect measure for placebo-controlled comparisons |
| Stratified rank ANCOVA | Used for hypothesis testing for the posted incontinence analyses |
| Multiplicity | Five-stage stepwise parallel gatekeeping plus Hochberg adjustment |
| Missing data | LOCF used in the primary Week 12 analyses; not used in posted Week 4 analyses |
| Confidence intervals | Two-sided 95% intervals reported for the LS mean differences |
| P-values | Reported within the prespecified superiority framework |
22. A Practical Reading of the ARIES Results
The most useful way to read the primary results is to start with the endpoint definition, then the direction of the effect, then its precision, and only afterward the p-value.
Step 1 · Identify the endpoint
Ask whether the outcome is incontinence episodes, micturitions, or volume voided. The clinical meaning of a positive or negative estimate depends on this definition.
Step 2 · Read the estimate
The LS mean difference describes the adjusted between-group difference in change from baseline.
Step 3 · Read the CI
The 95% CI communicates uncertainty and precision around the estimated difference.
Step 4 · Read the p-value
The p-value provides evidence against the null hypothesis within the trial's multiplicity-controlled testing framework; it is not the effect size.
For example, the Week 12 micturition result for 100 mg can be read as an LS mean difference of −0.70, with a 95% CI of −1.07 to −0.33 and P < 0.001. That sequence communicates the estimated magnitude, the uncertainty around it, and the strength of evidence against the null hypothesis without reducing the result to the p-value alone.
23. Statistical Methods Explained in Context
Why use a baseline-adjusted model instead of comparing Week 12 means?
The registered ANCOVA explicitly incorporates baseline. This is useful because patients can begin the trial with different levels of the outcome, and baseline adjustment accounts for that starting value when estimating the treatment difference. The target remains the between-group difference in outcome change, but the estimate is obtained through a prespecified model.
Why does the analysis use LS means?
Once covariates and fixed factors are included, a simple arithmetic group mean no longer fully represents the model comparison. LS means are model-based adjusted means evaluated under the specified ANCOVA structure. Their difference provides the treatment contrast reported by the registry.
Why is the incontinence analysis described as rank ANCOVA?
The registry specifically identifies a stratified rank ANCOVA for hypothesis testing of the incontinence endpoint. Rank-based methods work with the ordering of observations and can provide a more robust testing framework when distributional assumptions for a conventional analysis are a concern. The reported treatment effect is nevertheless presented as an LS mean difference from an ANCOVA model.
Why does multiplicity matter even when each p-value is below 0.05?
Multiple hypothesis tests create a greater opportunity for at least one false-positive result. ARIES addressed this through a stepwise parallel gatekeeping procedure for the endpoint hierarchy and Hochberg adjustment for the two active-dose comparisons per endpoint. The individual p-values therefore need to be understood as part of a larger testing strategy.
Why should Week 4 and Week 12 results not be treated as identical analyses?
The endpoint definitions are related, but the registry specifies different missing-data handling. LOCF was used for the primary Week 12 analyses and was not used for the posted Week 4 analyses. That difference can affect estimates and their interpretation.
Why can't two placebo comparisons establish a significant dose difference?
Suppose one dose has an estimate of −0.34 and another has an estimate of −0.50. The difference between those estimates is descriptive unless a formal 50 mg-versus-100 mg hypothesis test is conducted. Comparing whether each dose differs from placebo answers a different question from comparing the two doses directly.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Calculators
26. Sources
- ClinicalTrials.gov: ARIES, NCT00662909.
- Linked publication: PubMed record for PMID 24246044: PMID 24246044.
Continue through Clinical Biostats
Connect this trial's statistical methods with deeper tutorials and practical statistical tools.
27. Record Summary
ARIES provides a clear example of how continuous clinical-trial endpoints can be analyzed within a randomized, double-blind phase 3 framework. The two primary outcomes were derived from patient micturition diaries and analyzed using baseline-adjusted ANCOVA methods, with stratified rank ANCOVA used for the posted incontinence hypothesis tests. The principal treatment effect was expressed as an LS mean difference, with two-sided 95% confidence intervals and p-values reported within a prespecified multiplicity-control strategy.
The primary Week 12 estimates were negative for both incontinence episodes and micturitions, indicating lower modeled changes with mirabegron than placebo, while the volume-voided endpoint showed positive LS mean differences. The Week 4 analyses showed the same general directional pattern for incontinence and micturitions, although the registry used different missing-data handling at that time point. The statistical interpretation therefore depends not only on the numerical estimates but also on the endpoint definition, analysis population, covariate adjustment, multiplicity procedure, and missing-data strategy.