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
MANDALA was a randomized, quadruple-masked, phase 3, parallel-group treatment trial in adults and children 4 years of age or older with asthma. The study compared two doses of budesonide/albuterol sulfate metered-dose inhaler with albuterol sulfate metered-dose inhaler, with severe asthma exacerbation as the single registered primary endpoint.
| Feature | MANDALA |
|---|---|
| Phase | Phase 3 |
| Condition | Asthma |
| Population | Adults and children 4 years of age or older with asthma |
| Design | Randomized, parallel-group, quadruple-masked |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Primary endpoint | Number of Participants With a Severe Asthma Exacerbation Event |
| Primary endpoint type | Time-to-event |
| Enrollment | 3,132 |
| Trial status | Completed |
| Start | December 13, 2018 |
| Primary completion | August 19, 2021 |
2. Clinical Question
The central statistical question was whether either dose of budesonide/albuterol sulfate metered-dose inhaler reduced the hazard of a first severe asthma exacerbation compared with albuterol sulfate metered-dose inhaler during the randomized treatment period.
Population
Adults and children 4 years of age or older with asthma enrolled in a phase 3 randomized treatment trial.
Interventions
Budesonide/albuterol sulfate metered-dose inhaler 160/180 μg and 80/180 μg.
Comparator
Albuterol sulfate metered-dose inhaler 180 μg.
Primary question
Does either budesonide/albuterol dose reduce time to first severe asthma exacerbation relative to albuterol?
3. Trial Design
Budesonide/albuterol combination
- Budesonide/albuterol sulfate metered-dose inhaler
- Dose: 160/180 μg
- Primary comparison restricted to participants ≥12 years of age because children <12 years were not randomized to this dose
Budesonide/albuterol combination
- Budesonide/albuterol sulfate metered-dose inhaler
- Dose: 80/180 μg
- Compared with the full albuterol comparator analysis population
Albuterol comparator
- Albuterol sulfate metered-dose inhaler
- Dose: 180 μg
- Comparator for both primary treatment comparisons
4. Endpoint Definition and Follow-up
The registered primary endpoint was the Number of Participants With a Severe Asthma Exacerbation Event. The time frame was from randomization up to discontinuation of randomized treatment or a change in maintenance therapy. The mean and median reporting period for all participants was 44 and 48 weeks, respectively.
| Endpoint component | Registry definition |
|---|---|
| Primary event | Time to first severe asthma exacerbation. |
| Time origin | Randomization. |
| End of observation | Discontinuation of randomized treatment or a change in maintenance therapy. |
| Severe exacerbation | An exacerbation resulting in at least one specified systemic-corticosteroid, emergency-room/urgent-care, or inpatient-hospitalization criterion. |
| Reporting period | Mean 44 weeks; median 48 weeks. |
What counted as a severe exacerbation?
The registry defines a severe exacerbation as an asthma exacerbation resulting in at least one of the following: a temporary bolus or burst of systemic corticosteroids for at least 3 consecutive days to treat worsening asthma symptoms; an emergency-room or urgent-care visit of less than 24 hours for evaluation and treatment of asthma requiring systemic corticosteroids; or an inpatient hospitalization involving admission to an inpatient facility and/or at least 24 hours in a healthcare facility because of asthma.
This definition matters statistically because the primary endpoint is not simply a symptom score or any worsening of asthma. It is a clinically defined event that occurs at a particular time, making time-to-event methodology appropriate.
5. Statistical Analysis Populations
The posted primary analyses used a Full Analysis Set. The registry defines the analysis population as participants who were randomized, took at least one puff of randomized treatment, and had at least one efficacy assessment, excluding participants confirmed as duplicates.
| Population feature | Registry description |
|---|---|
| Randomization | Participants had to be randomized. |
| Treatment exposure | Participants had to take at least one puff of randomized treatment. |
| Efficacy assessment | Participants had to have at least one efficacy assessment. |
| Duplicate records | Participants confirmed as duplicates were excluded; the registry identifies n = 4. |
| Age restriction for 160/180 μg comparison | Because children <12 years of age were not randomized to BDA MDI 160/180 μg, the primary comparison used participants ≥12 years of age within the AS MDI 180 μg arm. |
This population definition is important when interpreting the hazard ratios. The reported 160/180 μg comparison is not a comparison across every enrolled participant: it is restricted to the age range in which both treatment groups were randomized.
6. Primary Endpoint Results
The registry contains formal statistical analyses for both primary comparisons. Both used Cox proportional-hazards models with treatment group, age group, region, and number of severe exacerbations in the last 12 months before randomization as factors.
BDA MDI 160/180 μg vs AS MDI 180 μg
Hazard ratio for first severe asthma exacerbation
95% CI: 0.611–0.879 · P < 0.001
Two-sided superiority analysis.
The estimated hazard ratio of 0.733 means that, under the fitted Cox model, the estimated instantaneous hazard of a first severe asthma exacerbation was approximately 26.7% lower with BDA MDI 160/180 μg than with AS MDI 180 μg. This follows directly from 1 − 0.733 = 0.267.
The hazard ratio does not mean that 26.7% of participants avoided an exacerbation, nor does it mean that every participant experienced a 26.7% reduction in individual risk. It is a relative time-to-event measure based on the model.
The 95% confidence interval of 0.611–0.879 describes statistical uncertainty around the estimated hazard ratio. Because the entire interval is below 1, the estimated treatment effect remains below the null value throughout the reported interval.
The P < 0.001 value addresses compatibility with the null hypothesis under the specified statistical model and testing framework. It does not measure the magnitude or clinical importance of the treatment effect. Interpretation also depends on censoring and on the proportional-hazards model being an appropriate description of the treatment comparison over follow-up.
BDA MDI 80/180 μg vs AS MDI 180 μg
Hazard ratio for first severe asthma exacerbation
95% CI: 0.702–0.992 · P = 0.041
Two-sided superiority analysis.
The estimated hazard ratio of 0.835 corresponds to an approximately 16.5% lower estimated instantaneous hazard of a first severe asthma exacerbation for BDA MDI 80/180 μg relative to AS MDI 180 μg, under the fitted Cox model.
Again, this is not a statement that 16.5% of patients were protected from an exacerbation or that each individual patient had exactly that reduction in risk. The hazard ratio summarizes a relative comparison of event hazards over the observed follow-up.
The 95% confidence interval of 0.702–0.992 is relatively close to the null value of 1 at its upper boundary. It therefore indicates more statistical uncertainty around this estimate than would an interval concentrated much farther below 1.
The reported P = 0.041 is evidence against the specified null hypothesis under the reported two-sided model. It should not be interpreted as a measure of effect size, and its interpretation should be considered together with the confidence interval and the prespecified sequential testing strategy used for secondary endpoints.
| Primary comparison | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| BDA MDI 160/180 μg vs AS MDI 180 μg | Hazard ratio | 0.733 | 0.611–0.879 | <0.001 |
| BDA MDI 80/180 μg vs AS MDI 180 μg | Hazard ratio | 0.835 | 0.702–0.992 | 0.041 |
7. Statistical Model for the Primary Endpoint
The registry reports a Cox regression model for both primary comparisons. The model incorporated treatment group, age group, region, and the number of severe exacerbations in the last 12 months before randomization as factors.
The exponentiated treatment coefficient provides the estimated hazard ratio comparing the treatment groups after accounting for the listed factors.
Covariate adjustment is useful here because age group, region, and prior severe-exacerbation history may be associated with the timing of future exacerbations. Including these factors in the model adjusts the estimated treatment comparison for their prespecified contribution to the fitted hazard.
Why the endpoint is naturally analyzed as time-to-event
Simply counting whether each participant ever experienced a severe exacerbation would discard information about when the first event occurred and about participants who remained event-free through their observation period. A time-to-event analysis uses both types of information while allowing participants to have different amounts of follow-up.
The registry specifies that follow-up ended at discontinuation of randomized treatment or a change in maintenance therapy. This creates a defined observation window for the primary analysis and makes censoring an important part of the statistical interpretation.
8. Secondary Endpoint Results: Annualized Severe Exacerbation Rate
The annualized severe exacerbation rate was analyzed using negative binomial regression. The model included treatment, age group, region, and the number of severe exacerbations in the last 12 months before randomization as categorical covariates. The logarithm of time at risk was included as an offset variable.
BDA MDI 160/180 μg vs AS MDI 180 μg
Rate ratio for annualized severe exacerbation rate
95% CI: 0.62–0.93 · P = 0.008
A rate ratio of 0.76 indicates that the estimated annualized severe exacerbation rate was 24% lower in the BDA MDI 160/180 μg group relative to the AS MDI 180 μg group under the negative binomial model.
This is a rate comparison rather than a time-to-first-event comparison. It considers the number of severe exacerbations in relation to time at risk, whereas the primary Cox analysis focuses on the time until the first severe exacerbation.
The 95% CI of 0.62–0.93 indicates uncertainty around the estimated rate ratio. The interval remains below 1, while still allowing a range of plausible relative rate differences.
The P-value of 0.008 describes evidence against the null hypothesis under the specified model. It does not quantify the size or clinical importance of the rate difference.
BDA MDI 80/180 μg vs AS MDI 180 μg
Rate ratio for annualized severe exacerbation rate
95% CI: 0.66–0.98 · P = 0.028
The rate ratio of 0.8 corresponds to an estimated 20% lower annualized severe exacerbation rate with BDA MDI 80/180 μg compared with AS MDI 180 μg under the fitted negative binomial model.
The confidence interval, 0.66–0.98, remains below 1 but comes relatively close to the null boundary. The estimate should therefore be read together with its uncertainty rather than as an exact 20% reduction.
The P-value of 0.028 is evidence against the specified null hypothesis under the reported analysis. It is not an effect-size measure and does not indicate the probability that the treatment is effective.
| Secondary endpoint | Comparison | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|---|
| Annualized severe exacerbation rate | BDA 160/180 vs AS 180 μg | Negative binomial regression | Rate ratio 0.76 | 0.62–0.93 | 0.008 |
| Annualized severe exacerbation rate | BDA 80/180 vs AS 180 μg | Negative binomial regression | Rate ratio 0.8 | 0.66–0.98 | 0.028 |
Why the rate analysis complements the primary endpoint
Two participants can have the same first-exacerbation time but very different subsequent event burdens. The annualized rate endpoint addresses the broader frequency of severe exacerbations during time at risk. The Cox endpoint and the negative binomial endpoint therefore answer related but distinct questions.
9. Secondary Endpoint Results: ACQ-5 Responders at Week 24
The registry also reports the number of participants who were responders on the Asthma Control Questionnaire-5 at Week 24. Logistic regression was used, with baseline ACQ-5 as a continuous covariate and age group, region, and number of severe exacerbations in the 12 months before randomization as categorical covariates.
| Comparison | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| BDA MDI 160/180 μg vs AS MDI 180 μg | Odds ratio | 1.221 | 1.016–1.467 | 0.033 |
| BDA MDI 80/180 μg vs AS MDI 180 μg | Odds ratio | 1.132 | 0.946–1.353 | 0.175 |
An odds ratio above 1 indicates higher estimated odds of being a responder in the BDA group relative to the AS group. The interpretation is conditional on the logistic regression model and its covariate adjustment.
For the 160/180 μg comparison, the estimated odds ratio was 1.221, with a 95% CI of 1.016–1.467 and P = 0.033. For the 80/180 μg comparison, the estimated odds ratio was 1.132, with a 95% CI of 0.946–1.353 and P = 0.175.
10. Secondary Endpoint Results: AQLQ+12 Responders at Week 24
The Asthma Quality of Life Questionnaire for Participants Aged 12 Years and Older (AQLQ+12) responder endpoint was assessed from baseline to 24 weeks. Logistic regression used baseline AQLQ-12 as a continuous covariate together with age group, region, and number of severe exacerbations in the 12 months before randomization as categorical covariates.
| Comparison | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| BDA MDI 160/180 μg vs AS MDI 180 μg | Odds ratio | 1.228 | 1.022–1.475 | 0.028 |
| BDA MDI 80/180 μg vs AS MDI 180 μg | Odds ratio | 1.111 | 0.925–1.335 | 0.26 |
For the 160/180 μg comparison, the estimated odds ratio was 1.228. The 95% CI of 1.022–1.475 lies above 1, while the P-value was 0.028. For the 80/180 μg comparison, the estimated odds ratio was 1.111, with a 95% CI of 0.925–1.335 and P = 0.26.
Because these are responder endpoints, the regression models compare the odds of belonging to the responder category rather than comparing a continuous questionnaire score directly. Baseline adjustment is intended to account for the baseline ACQ-5 or AQLQ-12 measurement while estimating the treatment association at Week 24.
11. Secondary Endpoint Results: Annualized Systemic Corticosteroid Dose
Total annualized dose of systemic corticosteroid was analyzed with a Wilcoxon rank sum test. The registry reports P-values but does not provide an effect estimate or confidence interval for this endpoint.
| Comparison | Method | P-value |
|---|---|---|
| BDA MDI 160/180 μg vs AS MDI 180 μg | Wilcoxon rank sum | 0.002 |
| BDA MDI 80/180 μg vs AS MDI 180 μg | Wilcoxon rank sum | 0.06 |
The Wilcoxon rank sum test compares the distributions of the outcome between two groups without requiring the same distributional assumptions as a standard two-sample t-test. It is especially useful when an outcome is skewed or contains values for which a rank-based comparison is more appropriate.
The reported P-value of 0.002 for the 160/180 μg comparison provides evidence against the specified null hypothesis under the rank-based test. The P-value of 0.06 for the 80/180 μg comparison is closer to the conventional null boundary. Neither P-value by itself describes the magnitude of the difference in systemic corticosteroid dose.
12. Sequential Testing and Multiplicity
The registry states that a sequential testing strategy was used for secondary endpoints. Under that strategy, a null hypothesis for a later secondary endpoint can only be rejected if all preceding null hypotheses are also rejected.
The registry specifies that a null hypothesis can only be rejected when all preceding null hypotheses in the sequence have also been rejected.
This is an important distinction from reading a table of P-values as a collection of independent hypothesis tests. A P-value of 0.028, for example, is not by itself sufficient to establish a formal sequentially significant result if the endpoint occupies a later position in the prespecified testing sequence and an earlier hypothesis was not rejected.
Multiplicity is therefore part of the interpretation of the secondary endpoints. The reported effect estimates and confidence intervals remain useful descriptions of the observed statistical comparisons, but formal hypothesis-testing conclusions depend on the sequence specified by the analysis.
13. Safety Results
The registry reports serious adverse events by randomized treatment arm. The reported counts are presented as affected participants over participants at risk.
| Treatment arm | Serious adverse events | Affected / at risk |
|---|---|---|
| BDA MDI 160/180 μg | Serious adverse events | 53 / 1,015 |
| BDA MDI 80/180 μg | Serious adverse events | 40 / 1,055 |
| AS MDI 180 μg | Serious adverse events | 48 / 1,057 |
These counts describe serious adverse events by randomized arm. They should be kept conceptually separate from the efficacy analyses: an efficacy hazard ratio addresses time to severe asthma exacerbation, whereas a serious-adverse-event count summarizes a safety outcome.
14. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
The primary endpoint records the time from randomization until the first severe asthma exacerbation. A Cox model is designed for this type of time-to-event outcome because it can use information from participants who experience an event and from participants who remain event-free until the end of their observation period or another censoring point. The model expresses the treatment comparison as a hazard ratio.
What does a hazard ratio of 0.733 mean?
A hazard ratio of 0.733 indicates that the fitted model estimates a lower instantaneous hazard of the first severe asthma exacerbation in the BDA MDI 160/180 μg group than in the AS MDI 180 μg group. The relative difference is approximately 26.7% when expressed as 1 − HR. It does not describe a 26.7% reduction for every individual participant, nor does it directly provide an absolute risk difference.
Why was negative binomial regression used for exacerbation rates?
The annualized severe exacerbation endpoint is a count of events observed over time at risk. Negative binomial regression is a count-data model that can accommodate greater variability in event counts than a simple Poisson model. Here, the registry specifies that the logarithm of time at risk was included as an offset, allowing the model to estimate a rate comparison while accounting for differences in observation time.
What does a rate ratio of 0.76 mean?
A rate ratio of 0.76 means that the fitted model estimates the annualized severe exacerbation rate to be 76% of the comparator rate, corresponding to an estimated 24% lower rate. This is different from saying that 24% fewer participants experienced at least one exacerbation. The primary Cox analysis and the rate-ratio analysis summarize different aspects of exacerbation burden.
Why was logistic regression used for responder endpoints?
The ACQ-5 and AQLQ+12 outcomes were defined as the number of participants who were responders at Week 24. A responder endpoint is binary: each participant is classified as a responder or not a responder. Logistic regression is designed for this structure and produces an odds ratio. The models also adjusted for the relevant baseline questionnaire score and prespecified categorical factors.
Why does an odds ratio of 1.221 not mean a 22.1% increase in response probability?
An odds ratio compares odds, not probabilities. Odds are defined as probability divided by one minus probability. Consequently, the same odds ratio can correspond to different absolute changes in probability depending on the baseline response probability. The odds ratio of 1.221 therefore should not be translated directly into a 22.1-percentage-point or 22.1-percent increase in response probability.
Why was a Wilcoxon rank sum test used for systemic corticosteroid dose?
The registry reports a Wilcoxon rank sum analysis for total annualized systemic corticosteroid dose. A rank-based test compares the relative ordering of observations between groups and does not require the same parametric assumptions as a conventional two-sample mean comparison. Its P-value tests the specified distributional comparison; it does not provide an effect-size estimate unless an additional measure is calculated.
15. Confidence Intervals and P-values
The MANDALA registry analyses illustrate why effect estimates, confidence intervals, and P-values should be read together.
| Measure | What it tells the reader | What it does not tell the reader |
|---|---|---|
| Hazard ratio | Relative comparison of event hazards under a Cox model. | The absolute probability of an event or the probability that an individual patient benefits. |
| Rate ratio | Relative comparison of modeled event rates. | The absolute difference in the number of events per participant. |
| Odds ratio | Relative comparison of odds for a binary outcome. | A direct percentage change in response probability. |
| 95% confidence interval | Statistical uncertainty around an estimated parameter under the specified model. | The range of individual patient effects or the probability that the true value lies inside the interval. |
| P-value | Compatibility of the observed result with a specified null hypothesis under the statistical model. | Effect size, clinical importance, or probability that the null hypothesis is true. |
For the two primary Cox analyses, the confidence intervals were 0.611–0.879 and 0.702–0.992. Their positions relative to 1 provide an immediate visual indication of the uncertainty surrounding the corresponding hazard ratios.
16. Covariate Adjustment
Covariate adjustment appears repeatedly in the MANDALA analyses. The primary Cox model included treatment group, age group, region, and the number of severe exacerbations in the last 12 months before randomization. The negative binomial models used the same treatment, age, region, and prior-exacerbation factors. The logistic models additionally incorporated the relevant baseline questionnaire score.
Primary survival analysis
Treatment group + age group + region + prior severe exacerbations.
Exacerbation-rate analysis
Treatment group + age group + region + prior severe exacerbations, with log time at risk as an offset.
ACQ-5 responder analysis
Treatment group + baseline ACQ-5 + age group + region + prior severe exacerbations.
AQLQ+12 responder analysis
Treatment group + baseline AQLQ-12 + age group + region + prior severe exacerbations.
Adjustment does not turn a randomized comparison into an observational study. Instead, the prespecified factors are incorporated into the statistical model used to estimate the treatment contrast. The interpretation remains conditional on the validity and specification of the model.
17. Time-to-Event Interpretation
The primary endpoint and its analysis make censoring central to the statistical story. A participant who does not experience a severe exacerbation before the end of the observation period does not simply become a missing observation. That participant contributes information about remaining free of the event through the time at which observation ends.
The analysis compares the event-time distributions using a Cox proportional-hazards model rather than simply comparing the proportion of participants who ever experienced an event.
The proportional-hazards interpretation deserves particular care. A Cox hazard ratio is most straightforward when the relative hazard between treatment groups can reasonably be represented as approximately constant over time. The registry reports the Cox model but does not provide a separate assessment of the proportional-hazards assumption in the information summarized here.
18. Primary and Secondary Statistical Story
| Endpoint | Comparison | Method | Effect | P-value |
|---|---|---|---|---|
| Severe asthma exacerbation | BDA 160/180 vs AS 180 μg | Cox proportional-hazards model | HR 0.733 (95% CI 0.611–0.879) | <0.001 |
| Severe asthma exacerbation | BDA 80/180 vs AS 180 μg | Cox proportional-hazards model | HR 0.835 (95% CI 0.702–0.992) | 0.041 |
| Annualized severe exacerbation rate | BDA 160/180 vs AS 180 μg | Negative binomial regression | RR 0.76 (95% CI 0.62–0.93) | 0.008 |
| Annualized severe exacerbation rate | BDA 80/180 vs AS 180 μg | Negative binomial regression | RR 0.8 (95% CI 0.66–0.98) | 0.028 |
| ACQ-5 responder, Week 24 | BDA 160/180 vs AS 180 μg | Logistic regression | OR 1.221 (95% CI 1.016–1.467) | 0.033 |
| ACQ-5 responder, Week 24 | BDA 80/180 vs AS 180 μg | Logistic regression | OR 1.132 (95% CI 0.946–1.353) | 0.175 |
| AQLQ+12 responder, Week 24 | BDA 160/180 vs AS 180 μg | Logistic regression | OR 1.228 (95% CI 1.022–1.475) | 0.028 |
| AQLQ+12 responder, Week 24 | BDA 80/180 vs AS 180 μg | Logistic regression | OR 1.111 (95% CI 0.925–1.335) | 0.26 |
| Total annualized SCS dose | BDA 160/180 vs AS 180 μg | Wilcoxon rank sum | Effect estimate not reported | 0.002 |
| Total annualized SCS dose | BDA 80/180 vs AS 180 μg | Wilcoxon rank sum | Effect estimate not reported | 0.06 |
This table demonstrates why a clinical-trial statistical analysis should not collapse every endpoint into a single P-value. MANDALA used several endpoint structures and therefore several corresponding effect measures. The hazard ratio describes time to first exacerbation, the rate ratio describes exacerbation frequency over time, the odds ratio describes responder odds, and the Wilcoxon analysis provides a rank-based hypothesis test for systemic corticosteroid dose.
19. Limitations
- Different estimands: the primary time-to-first-event endpoint and annualized exacerbation rate measure related but distinct aspects of severe asthma exacerbation burden.
- Model dependence: Cox, negative binomial, and logistic regression estimates depend on the assumptions and specification of their respective models.
- Proportional hazards: interpretation of the Cox hazard ratio is most direct when the relative hazards can reasonably be summarized by a proportional-hazards model.
- Censoring: the primary endpoint is observed only through discontinuation of randomized treatment or a change in maintenance therapy, making the handling and interpretation of censored observations important.
- Sequential testing: secondary endpoint P-values cannot all be interpreted as independent confirmatory tests because the registry specifies a sequential testing strategy.
- Analysis population: the primary analysis population required randomized treatment exposure and at least one efficacy assessment, with duplicate records excluded.
- Age-specific comparison: the 160/180 μg primary comparison was restricted to participants ≥12 years of age because children <12 years were not randomized to that dose.
- Effect measures are not interchangeable: hazard ratios, rate ratios, and odds ratios answer different statistical questions and should not be interpreted as equivalent forms of relative risk.
- Safety versus efficacy: serious adverse-event counts describe a different outcome domain from severe asthma exacerbation efficacy analyses and should not be combined into a single summary measure.
20. Why This Trial Matters Statistically
MANDALA is a useful teaching example because its registry results show how a single randomized trial can require several distinct statistical frameworks. The primary endpoint is time-to-event, while secondary outcomes include recurrent-event rates, binary responder endpoints, and a rank-based comparison of systemic corticosteroid dose.
| Concept | How it appears in MANDALA |
|---|---|
| Randomization | Randomized allocation across 3 parallel treatment arms. |
| Blinding | Quadruple masking. |
| Time-to-event endpoint | Time to first severe asthma exacerbation. |
| Cox proportional-hazards model | Primary hazard-ratio analysis. |
| Hazard ratio | Relative comparison of time to first severe exacerbation. |
| Negative binomial regression | Annualized severe exacerbation rate. |
| Rate ratio | Relative comparison of annualized exacerbation rates. |
| Logistic regression | ACQ-5 and AQLQ+12 responder analyses. |
| Odds ratio | Effect measure for binary responder endpoints. |
| Covariate adjustment | Age group, region, prior severe exacerbations, and baseline questionnaire scores where applicable. |
| Wilcoxon / Mann-Whitney test | Analysis of total annualized systemic corticosteroid dose. |
| Sequential testing | Prespecified approach to formal secondary-endpoint hypothesis testing. |
21. Interpreting the Two Primary Comparisons Together
The two primary analyses provide separate treatment comparisons against the same albuterol comparator. The higher budesonide dose comparison produced an estimated hazard ratio of 0.733, while the lower dose comparison produced an estimated hazard ratio of 0.835.
These estimates should be interpreted as separate model-based treatment contrasts rather than as a dose-response curve. The registry reports two randomized comparisons, but the information summarized here does not provide a formal statistical test comparing the two budesonide/albuterol doses directly with each other.
The confidence intervals also provide useful context. The 160/180 μg estimate has a 95% CI of 0.611–0.879, while the 80/180 μg estimate has a 95% CI of 0.702–0.992. Confidence intervals that overlap or differ in width should not be used alone to claim that treatment effects differ between groups; a formal interaction or direct between-dose comparison would be required for that question.
The primary results are best understood as two prespecified superiority comparisons, each producing a hazard ratio with a two-sided 95% confidence interval and P-value. The numerical difference between 0.733 and 0.835 is descriptive; it does not, by itself, establish that the two BDA doses have different effects.
22. What the Registry Results Do and Do Not Establish
What the primary analyses establish statistically
They provide model-based estimates of the relative hazard of a first severe asthma exacerbation for each BDA dose versus AS, with corresponding confidence intervals and P-values.
What the secondary analyses add
They describe exacerbation rates, responder odds, and systemic corticosteroid dose using methods matched to those endpoint structures.
What a hazard ratio cannot provide alone
It does not give an absolute event probability, number needed to treat, or individual-level treatment effect.
What a P-value cannot provide
It does not quantify effect size, clinical importance, or the probability that a treatment hypothesis is true.
23. Trial Timeline
Randomized phase 3 study begins
The MANDALA trial began on December 13, 2018, evaluating budesonide/albuterol metered-dose inhalers and albuterol in asthma.
Primary study period completed
The registry lists August 19, 2021 as the primary completion date.
Statistical analyses posted
The registry reports 5 outcome measures and 10 statistical analyses, including 2 formal primary-endpoint analyses.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Calculators
26. Sources
- ClinicalTrials.gov: NCT03769090 — MANDALA.
- PubMed: PMID 35569035.
- PubMed: PMID 40664332.
- PubMed: PMID 36161875.
Continue through the Clinical Biostats statistical library
Explore the underlying statistical methods through tutorials, calculators, and additional clinical-trial analyses.
27. Record Summary
MANDALA provides a compact example of how endpoint structure determines statistical methodology. The primary outcome was a time-to-first-event endpoint analyzed with Cox proportional hazards models and reported as hazard ratios. Secondary analyses extended the statistical picture using negative binomial regression for annualized severe exacerbation rates, logistic regression for ACQ-5 and AQLQ+12 responder outcomes, and the Wilcoxon rank sum test for total annualized systemic corticosteroid dose.
The most informative interpretation therefore combines the effect estimate, its confidence interval, the corresponding P-value, the analysis population, the covariate structure, and the statistical assumptions associated with each method. The registry's sequential testing strategy is also important when interpreting formal claims about secondary endpoints.