This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
AIR-BX1 was a randomized, double-masked, parallel phase 3 trial evaluating the safety and effectiveness of AZLI, an inhaled antibiotic, compared with placebo in adults with non-cystic fibrosis bronchiectasis.
| Feature | AIR-BX1 |
|---|---|
| Trial name | AIR-BX1 |
| NCT ID | NCT01313624 |
| Therapeutic area | Pulmonology |
| Condition | Bronchiectasis |
| Phase | Phase 3 |
| Design | Randomized, double-masked, parallel |
| Enrollment | 266 |
| Interventions | AZLI; Placebo |
| Primary purpose | Treatment |
| Lead sponsor | Gilead Sciences |
| Sponsor type | Industry |
| Trial status | Completed |
| Start | 2011-04 |
| Primary completion | 2013-03 |
2. Clinical Question
The central question was whether AZLI produced a difference in the registered QOL-B Respiratory Symptoms Score compared with placebo in adults with non-cystic fibrosis bronchiectasis.
Population
Adults with non-cystic fibrosis bronchiectasis.
Intervention
AZLI, an inhaled antibiotic.
Comparator
Placebo.
Primary question
What was the difference between AZLI and placebo in change in the QOL-B Respiratory Symptoms Score from baseline to Day 28?
3. Trial Design
The ClinicalTrials.gov record identifies AIR-BX1 as a randomized, double-masked, parallel phase 3 trial with 266 enrolled participants and two interventions: AZLI and placebo. The ClinicalTrials.gov record does not provide arm-specific randomized enrollment counts, dosing details, treatment duration, stratification factors, or an allocation ratio.
4. Trial Arms
AZLI
- AZLI was the intervention.
- It is described in the trial title as an inhaled antibiotic.
- The primary comparison was AZLI versus placebo.
Placebo
- Placebo was the comparator intervention.
- The primary statistical analysis compared AZLI-AZLI with Placebo-AZLI.
- The ClinicalTrials.gov record does not provide the randomized arm sizes for this comparison.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Change in QOL-B Respiratory Symptoms Score at Day 28 | The mean (SD) change in the Respiratory Symptoms score on the Quality of Life Questionnaire-Bronchiectasis (QOL-B) was measured from baseline to the end of Course 1 (Day 28). The QOL-B respiratory symptoms score was transformed onto a scale of 0-100, with higher scores representing a better quality of life. | Baseline to Day 28 |
This is a continuous patient-reported outcome. The statistical target is therefore a difference in mean change between randomized groups rather than a hazard ratio, odds ratio, or event-time comparison.
6. Statistical Methodology
Intention-to-treat analysis set
The posted primary analysis used participants in the ITT Analysis Set with scores at both baseline and Day 28. The registry therefore identifies both the randomized-treatment framework and the availability requirement used for this posted analysis.
Mixed model for repeated measures
The registry reports MMRM, or mixed model for repeated measures, as the statistical method. MMRM is designed for longitudinal continuous outcomes in which measurements are obtained repeatedly over time. Rather than reducing a longitudinal dataset to a single crude comparison, the model can represent the pattern of measurements across scheduled assessment times.
The posted effect measure was a difference in least squares mean, normalized here as a mean difference. A positive estimate indicates a higher modeled mean change in the AZLI group than in the placebo group.
Why least squares means are useful
Least squares means are model-based adjusted means. They are especially useful in regression and longitudinal models because the comparison is derived from the fitted statistical model rather than simply taking the arithmetic difference between two unadjusted sample means.
Confidence interval
The primary analysis reports a two-sided 95% confidence interval. For a mean difference, the interval describes statistical uncertainty around the estimated between-group difference under the fitted model and its assumptions.
Hypothesis type
The registry identifies the primary analysis as a superiority hypothesis. Thus, the statistical question was whether the AZLI group differed from placebo in the specified direction of a superiority framework, rather than whether AZLI was merely within a prespecified non-inferiority margin.
7. Primary Result: QOL-B Respiratory Symptoms Score at Day 28
The primary endpoint was analyzed from baseline to Day 28 using an MMRM in participants in the ITT Analysis Set with scores at both baseline and Day 28.
Difference in least squares mean change
95% CI: -3.1 to 4.7 · P = 0.68
AZLI-AZLI vs Placebo-AZLI · Baseline to Day 28
| Endpoint | AZLI-AZLI vs Placebo-AZLI | 95% CI | P-value | Method |
|---|---|---|---|---|
| Change in QOL-B Respiratory Symptoms Score at Day 28 | Difference in least squares mean: 0.8 | -3.1 to 4.7 | 0.68 | MMRM |
The estimated difference in least squares mean change was 0.8 QOL-B units, with AZLI-AZLI compared with Placebo-AZLI. Because the QOL-B respiratory symptoms score is scaled from 0-100 and higher scores represent better quality of life, the positive estimate corresponds to a modeled higher change in the AZLI group.
The estimate does not mean that AZLI improved every patient's respiratory symptoms by 0.8 units. It is a group-level model-based comparison, not an individual treatment effect.
The 95% confidence interval extends from -3.1 to 4.7. This interval includes zero, meaning that the registry's reported uncertainty is compatible with a small difference favoring either group as well as with a larger positive difference within the interval. The width of the interval is also important: the estimate of 0.8 should not be interpreted without considering the uncertainty surrounding it.
The P = 0.68 is a measure used for the statistical hypothesis test; it is not a measure of the size or clinical importance of the estimated difference. A p-value does not tell us that the treatment effect is "68%" or that there is a 68% probability that the null hypothesis is true.
Because the reported analysis is an MMRM, interpretation also depends on the model specification and assumptions. The ClinicalTrials.gov record identifies the method but do not provide the complete model specification, covariance structure, or all estimation details. The result should therefore be interpreted as the posted registry analysis rather than as a reconstruction of the underlying statistical model.
8. Secondary Endpoint Result: QOL-B Respiratory Symptoms Score at Day 84
The registry also posted a secondary analysis of the change in QOL-B Respiratory Symptoms Score from baseline to Day 84. Like the primary endpoint, this was analyzed with an MMRM in the ITT Analysis Set among participants with scores at both baseline and Day 84.
Difference in least squares mean change
95% CI: -3.0 to 5.6 · P = 0.56
AZLI-AZLI vs Placebo-AZLI · Baseline to Day 84
| Endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Change in QOL-B Respiratory Symptoms Score at Day 84 | Difference in least squares mean | 1.3 | -3.0 to 5.6 | 0.56 |
The Day 84 estimate of 1.3 represents a modeled difference in mean change between AZLI-AZLI and Placebo-AZLI. As with the Day 28 result, it is not an individual-level estimate and should not be interpreted as the amount by which every participant's score changed.
The 95% confidence interval of -3.0 to 5.6 includes zero. Thus, the reported estimate has substantial uncertainty relative to its magnitude, and the interval encompasses both a small negative difference and a positive difference.
The reported P = 0.56 describes the statistical evidence under the specified superiority hypothesis test. It does not quantify treatment benefit, probability of benefit, or clinical importance.
The Day 84 analysis is a secondary endpoint and should also be distinguished from the primary Day 28 analysis. The ClinicalTrials.gov record does not provide a multiplicity-adjustment procedure for the secondary endpoint, so no stronger confirmatory interpretation should be assigned to its p-value than the registry data support.
9. Comparing the Day 28 and Day 84 Estimates
| Time frame | Estimate | 95% CI | P-value | Endpoint role |
|---|---|---|---|---|
| Baseline to Day 28 | 0.8 | -3.1 to 4.7 | 0.68 | Primary |
| Baseline to Day 84 | 1.3 | -3.0 to 5.6 | 0.56 | Secondary |
The point estimates are both positive, with 0.8 at Day 28 and 1.3 at Day 84. However, the corresponding confidence intervals both include zero. The ClinicalTrials.gov record therefore support describing the numerical estimates and their uncertainty, rather than treating the difference between the two point estimates as evidence of a time trend.
It is also important not to compare the two p-values as though the smaller value represented a stronger treatment effect. A p-value depends on the estimate, its uncertainty, sample information, and the statistical testing framework. Effect size and statistical evidence are different quantities.
10. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment sequence and study phase. These figures should be kept separate from the randomized efficacy comparison because the safety categories include both double-blind and open-label groups.
| Group | Serious adverse events affected / at risk |
|---|---|
| AZLI-AZLI (Double-Blind) | 28/134 |
| Placebo-AZLI (Double-Blind) | 13/132 |
| AZLI-AZLI (Open-Label) | 9/96 |
| Placebo-AZLI (Open-Label) | 18/122 |
The registry data therefore identify serious adverse event counts and denominators for four groups. The double-blind groups are particularly important for understanding safety during the randomized blinded comparison, while the open-label groups represent a different study context.
11. Statistical Methods Explained
Why was an MMRM used?
The primary outcome is a continuous QOL-B score measured over time, making a longitudinal modeling framework appropriate. MMRM is designed for repeated continuous measurements and estimates treatment-group differences using a fitted model rather than relying only on a simple unadjusted comparison at one time point.
What does a difference in least squares mean of 0.8 mean?
It means the fitted model estimated the AZLI-AZLI group's mean change to be 0.8 QOL-B units higher than the Placebo-AZLI group's modeled mean change for the Day 28 analysis. It is not a statement that every patient experienced a 0.8-unit improvement.
Why is the confidence interval important?
The 95% confidence interval of -3.1 to 4.7 places the point estimate in the context of uncertainty. Because zero lies inside the interval, the ClinicalTrials.gov record does not exclude no between-group difference under the corresponding statistical framework. The interval also communicates more information about plausible effect size than the p-value alone.
What does P = 0.68 tell us?
The p-value measures the degree of statistical evidence against the null hypothesis under the specified testing framework. It does not measure how large the treatment effect is, whether the effect is clinically important, or the probability that the null hypothesis is true.
Why does intention-to-treat analysis matter?
Randomized clinical trials generally use the ITT principle to preserve the comparison created by randomization. In this posted analysis, the registry specifically states that participants in the ITT Analysis Set with scores at both the relevant baseline and follow-up time were analyzed. The availability requirement should therefore be distinguished from an unrestricted analysis of every enrolled participant.
Why does double masking matter statistically?
Double masking is a design feature intended to reduce the opportunity for knowledge of treatment assignment to influence participant behavior, outcome reporting, or other aspects of trial conduct. It is especially relevant for a patient-reported quality-of-life endpoint because expectations about treatment can potentially affect reported symptoms.
Why is randomization important?
Randomization establishes the treatment assignment mechanism before outcomes are observed. In a properly conducted randomized comparison, it provides the principal design basis for attributing differences between groups to treatment rather than to systematic pre-treatment assignment differences. The registry identifies AIR-BX1 as randomized but does not provide the detailed randomization procedure in the ClinicalTrials.gov record.
12. Understanding the QOL-B Endpoint
The registered primary endpoint is a change score on the Quality of Life Questionnaire-Bronchiectasis (QOL-B) Respiratory Symptoms domain. The registry states that the score was transformed onto a 0-100 scale, with higher scores representing a better quality of life.
The reported treatment effect is therefore interpreted according to the direction of this scale. A positive between-group change difference favors the AZLI group numerically, while a negative difference favors the placebo group numerically.
This directionality is important because statistical signs are not intrinsically "good" or "bad." Their meaning depends on how the endpoint is coded. The registry explicitly supplies the coding direction for this endpoint.
13. Confidence Intervals vs P-values
Estimate
The estimate describes the modeled magnitude and direction of the between-group difference: 0.8 at Day 28 and 1.3 at Day 84.
Confidence interval
The 95% intervals describe uncertainty around those estimates: -3.1 to 4.7 at Day 28 and -3.0 to 5.6 at Day 84.
P-value
The p-values, 0.68 and 0.56, quantify statistical evidence under the respective hypothesis tests rather than effect magnitude.
Clinical meaning
Statistical estimates should be interpreted alongside the endpoint's 0-100 scale and its direction, not by p-value alone.
A useful statistical reading therefore proceeds in sequence: identify the endpoint, identify the treatment contrast, inspect the point estimate, examine the confidence interval, and then consider the p-value in the context of the prespecified hypothesis. This prevents a common mistake in which the p-value becomes the entire interpretation of the result.
14. Analysis Population and Missing Data Considerations
The registry specifies that the Day 28 primary analysis included participants in the ITT Analysis Set with scores at both baseline and Day 28. The Day 84 secondary analysis similarly included participants in the ITT Analysis Set with scores at both baseline and Day 84.
| Endpoint | Analysis population stated in registry |
|---|---|
| Day 28 primary | Participants in the ITT Analysis Set with scores at both baseline and Day 28. |
| Day 84 secondary | Participants in the ITT Analysis Set with scores at both baseline and Day 84. |
This distinction matters because an ITT analysis set and an analysis restricted to participants with observed endpoint measurements are not conceptually identical. The ClinicalTrials.gov record does not specify the missing-data mechanism, imputation procedure, covariance structure, sensitivity analyses, or other detailed MMRM assumptions. Those aspects therefore should not be inferred from the method name alone.
15. Multiplicity and Hypothesis Testing
The ClinicalTrials.gov record identifies one registered primary endpoint and one posted secondary endpoint. Both have formal statistical analyses, and both use superiority hypotheses.
| Endpoint | Role | Hypothesis type | Reported P-value |
|---|---|---|---|
| Change in QOL-B Respiratory Symptoms Score at Day 28 | Primary | Superiority | 0.68 |
| Change in QOL-B Respiratory Symptoms Score at Day 84 | Secondary | Superiority | 0.56 |
The ClinicalTrials.gov record does not report a multiplicity-adjustment strategy. Accordingly, the Day 84 p-value should be described as the posted secondary analysis result rather than as an independently confirmed hypothesis with an assumed multiplicity procedure.
16. What the Mean Difference Does — and Does Not — Mean
The Day 28 estimate of 0.8 is a model-based difference in least squares mean change between AZLI-AZLI and Placebo-AZLI. Because higher QOL-B scores represent better quality of life, the positive sign corresponds numerically to greater modeled improvement in the AZLI group.
It does not mean that 0.8% of patients benefited, that patients experienced an 0.8% improvement, or that the treatment caused exactly 0.8 units of change in each individual.
The 95% confidence interval of -3.1 to 4.7 describes uncertainty around the estimated difference. It includes zero and therefore encompasses both a small negative difference and a positive difference under the model's uncertainty framework.
The reported P = 0.68 is evidence from a statistical test, not a numerical measure of treatment benefit. The magnitude of the observed estimate and the confidence interval are needed to understand the estimated effect itself.
17. Primary vs Secondary Evidence
| Dimension | Day 28 | Day 84 |
|---|---|---|
| Endpoint role | Primary | Secondary |
| Time frame | Baseline to Day 28 | Baseline to Day 84 |
| Effect measure | Difference in least squares mean | Difference in least squares mean |
| Estimate | 0.8 | 1.3 |
| 95% CI | -3.1 to 4.7 | -3.0 to 5.6 |
| P-value | 0.68 | 0.56 |
| Method | MMRM | MMRM |
The distinction between primary and secondary endpoints is fundamental. The primary endpoint is the principal registered efficacy question, while the Day 84 result provides an additional longitudinal assessment. A secondary result should not be silently substituted for the primary endpoint simply because it occurs later in time.
18. Blinding, Randomization, and Causal Interpretation
Randomization
The registry identifies the allocation as randomized. Randomization is the design feature that supports a causal comparison between AZLI and placebo.
Double masking
The trial is identified as double-masked, which is particularly relevant for a patient-reported quality-of-life endpoint.
Parallel design
The registry identifies a parallel design, meaning the principal comparison is between groups rather than a crossover comparison within the same participants.
Analysis population
The posted efficacy analyses use the ITT Analysis Set subject to the registry's requirement for observed baseline and follow-up scores.
These design features operate together. Randomization establishes the treatment comparison, masking can reduce expectation-related influences, and the prespecified longitudinal analysis provides a statistical framework for comparing the continuous endpoint over time.
19. What the Registry Data Do Not Establish
The registry-reported AIR-BX1 data are sufficient to describe the principal design and posted statistical analyses, but they do not provide every component of a full statistical analysis plan. In particular, the ClinicalTrials.gov record does not report:
- Randomized sample size by efficacy arm.
- Allocation ratio.
- Detailed randomization or stratification procedures.
- Detailed MMRM covariance structure or full model specification.
- Specific missing-data or imputation assumptions.
- A multiplicity-adjustment procedure.
- A Bayesian analysis.
- An interim analysis procedure.
- A non-inferiority margin.
- A crossover design.
- Additional formal statistical analyses beyond the two registry-reported analyses.
These omissions are not evidence that such procedures were absent from the complete protocol or statistical analysis plan. They simply are not supported by the ClinicalTrials.gov record and therefore are not treated as established features of the trial on this page.
20. Limitations
- Registry-level detail: the ClinicalTrials.gov record identifies MMRM but do not provide the complete model specification, covariance structure, or all estimation details.
- Observed-score requirement: the posted efficacy analyses were restricted to participants in the ITT Analysis Set with scores at both baseline and the relevant follow-up time.
- Missing-data assumptions: the ClinicalTrials.gov record does not state an imputation method or sensitivity-analysis framework.
- Multiplicity: the ClinicalTrials.gov record does not specify how the primary and secondary endpoint tests were incorporated into an overall type I error strategy.
- Safety interpretation: serious adverse event data are reported for double-blind and open-label groups, so they should not be collapsed into a single randomized comparison.
- Arm sizes: the ClinicalTrials.gov record reports total enrollment of 266 but does not give randomized arm sizes for the efficacy analysis.
- Generalizability: the ClinicalTrials.gov record describes adults with non-cystic fibrosis bronchiectasis, but it does not provide the detailed eligibility criteria needed to characterize the full study population.
- Endpoint scope: the statistical ClinicalTrials.gov record cover the posted primary Day 28 and secondary Day 84 QOL-B respiratory symptoms analyses; no additional outcomes are inferred.
21. Why This Trial Matters Statistically
AIR-BX1 is a useful teaching example because it illustrates a different statistical pattern from trials dominated by survival endpoints. The primary outcome is a continuous quality-of-life measure, and the reported analysis uses an MMRM with a difference in least squares means.
| Concept | How it appears in AIR-BX1 |
|---|---|
| Randomization | The trial is identified as randomized. |
| Double masking | The trial is identified as double-masked. |
| Parallel design | The registry identifies a parallel study model. |
| Intention-to-treat analysis | The efficacy analyses use the ITT Analysis Set subject to the stated observed-score requirement. |
| MMRM | The primary and secondary posted analyses use mixed model for repeated measures. |
| Least squares mean | The effect measure is the difference in least squares mean. |
| Confidence interval | Both reported estimates include two-sided 95% confidence intervals. |
| P-value | The primary and secondary analyses report p-values of 0.68 and 0.56. |
| Continuous endpoint | The QOL-B Respiratory Symptoms Score is measured on a 0-100 scale. |
| Longitudinal assessment | The endpoint is evaluated at Day 28 and Day 84 from baseline. |
| Safety populations | Serious adverse events are reported for double-blind and open-label groups. |
The trial is particularly instructive for understanding why the statistical method should follow the structure of the endpoint. A continuous repeated quality-of-life measure calls for different analytical thinking from a binary response endpoint or a time-to-event endpoint. Here, the key statistical quantities are the modeled mean difference, its confidence interval, and the corresponding hypothesis test.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: AIR-BX1, NCT01313624.
- PubMed: PubMed record, PMID 25154045.
Continue through the Clinical Biostats statistical pathway
Use the related tutorials and calculators to explore randomization, masking, confidence intervals, p-values, intention-to-treat analysis, and mixed models for repeated measures.
25. Record Summary
AIR-BX1 was a randomized, double-masked, parallel phase 3 trial enrolling 266 adults with non-cystic fibrosis bronchiectasis. Its registered primary endpoint was change in the QOL-B Respiratory Symptoms Score from baseline to Day 28, measured on a 0-100 scale in which higher scores represent better quality of life. The posted primary analysis used an MMRM in the ITT Analysis Set among participants with scores at both baseline and Day 28, reporting a difference in least squares mean of 0.8, with a two-sided 95% CI of -3.1 to 4.7 and P = 0.68.
A secondary Day 84 analysis used the same general statistical framework and reported a difference in least squares mean of 1.3, with a two-sided 95% CI of -3.0 to 5.6 and P = 0.56. The principal statistical lesson is that these results should be read through the complete set of quantities: the direction and magnitude of the estimated mean difference, the confidence interval, the hypothesis-test result, the analysis population, and the assumptions underlying the longitudinal model.