This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. View the AIR-BX2 ClinicalTrials.gov record.
1. Trial at a Glance
AIR-BX2 was a randomized, double-blind, parallel phase 3 trial in adults with non-cystic fibrosis bronchiectasis. The registry reports an enrollment of 274 participants and two study arms: AZLI and placebo. The posted primary efficacy analysis compared change in the QOL-B Respiratory Symptoms Score from baseline to Day 28 using a mixed model for repeated measures.
| Feature | AIR-BX2 |
|---|---|
| Trial name | AIR-BX2 |
| NCT identifier | NCT01314716 |
| Phase | Phase 3 |
| Condition | Bronchiectasis |
| Population | Adults with non-cystic fibrosis bronchiectasis |
| Design | Randomized, double-blind, parallel |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 274 |
| Interventions | AZLI (drug); Placebo (drug) |
| Lead sponsor | Gilead Sciences |
| Sponsor type | Industry |
| Trial status | Completed |
| Start | 2011-04 |
| Primary completion | 2013-04 |
2. Clinical Question
The central statistical question was whether participants randomized to AZLI had a different change in the QOL-B Respiratory Symptoms Score from baseline to Day 28 compared with participants randomized to placebo.
Population
Adults with non-cystic fibrosis bronchiectasis.
Intervention
AZLI, an inhaled antibiotic.
Comparator
Placebo.
Primary question
Does AZLI produce a greater improvement in the QOL-B Respiratory Symptoms Score from baseline to Day 28 than placebo?
The registered endpoint is a patient-reported quality-of-life measure rather than a mortality or time-to-event endpoint. That distinction determines the appropriate statistical framework: the primary analysis is based on change in a continuous score, not on a hazard ratio or survival probability.
3. Trial Design
AZLI
- AZLI (drug)
- Compared with placebo in the randomized, double-blind parallel design
Placebo
- Placebo (drug)
- Served as the randomized comparator
4. Endpoints
| Endpoint | Time frame | Registry definition | Analysis |
|---|---|---|---|
| Change in QOL-B Respiratory Symptoms Score at Day 28 | Baseline to 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. | MMRM; difference in least squares mean |
| Change in QOL-B Respiratory Symptoms Score at Day 84 | Baseline to Day 84 | Change in the QOL-B Respiratory Symptoms Score from baseline to Day 84. | MMRM; difference in least squares mean |
The registered primary endpoint is the Day 28 change in the QOL-B Respiratory Symptoms Score. The Day 84 analysis is reported as a secondary endpoint in the posted statistical analyses.
The registry states that the respiratory symptoms score is transformed onto a 0-100 scale, with higher scores representing a better quality of life. Consequently, a positive between-group difference in change is directionally favorable to AZLI in the reported AZLI-AZLI versus Placebo-AZLI comparison.
5. Statistical Methodology
Mixed model for repeated measures
The posted statistical analyses used a mixed model for repeated measures (MMRM). MMRM is designed for longitudinal outcomes in which each participant can contribute repeated measurements over time. Rather than analyzing every time point as an entirely separate observation, the model treats the repeated measurements as part of a longitudinal structure.
For AIR-BX2, the reported endpoint is a continuous QOL-B score. The statistical result is expressed as a difference in least squares means between AZLI-AZLI and Placebo-AZLI. In normalized terms, this is a mean difference.
The reported estimate is the modeled difference between the two randomized groups for the Day 28 change in the QOL-B Respiratory Symptoms Score. Because higher QOL-B scores represent better quality of life, the positive estimate favors the AZLI-AZLI group on this endpoint.
Intention-to-treat analysis set
The primary statistical analysis was conducted among participants in the ITT Analysis Set with scores at both baseline and Day 28. The secondary Day 84 analysis similarly used participants in the ITT Analysis Set with scores at both baseline and Day 84.
This wording is important. The trial's analysis population is described as an ITT Analysis Set, but the posted analysis also specifies availability of both the baseline and relevant follow-up score. Therefore, the reported analysis is not equivalent to simply taking every randomized participant regardless of whether a corresponding outcome measurement was available.
Least squares means
A least squares mean is a model-based adjusted mean. In an MMRM, it represents the estimated group mean under the fitted longitudinal model rather than necessarily being identical to the raw arithmetic average of observed changes.
Why this matters
The reported 4.6 is a model-based treatment contrast, not a statement that every participant improved by exactly 4.6 QOL-B units.
What the CI adds
The 95% confidence interval of 1.1 to 8.2 describes uncertainty around the estimated group difference.
6. Primary Result: Day 28 QOL-B Respiratory Symptoms Score
The registry reports a formal statistical analysis for the primary endpoint, comparing AZLI-AZLI with Placebo-AZLI from baseline to Day 28.
Difference in least squares mean
95% CI: 1.1–8.2 · P = 0.011
Two-sided confidence interval; superiority hypothesis; MMRM analysis.
| Primary endpoint | AZLI-AZLI vs Placebo-AZLI |
|---|---|
| Outcome | Change in QOL-B Respiratory Symptoms Score at Day 28 |
| Time frame | Baseline to Day 28 |
| Endpoint type | Continuous |
| Analysis population | Participants in the ITT Analysis Set with scores at both baseline and Day 28 |
| Method | MMRM |
| Effect measure | Difference in least squares mean |
| Estimate | 4.6 |
| 95% CI | 1.1–8.2 |
| P-value | 0.011 |
| Hypothesis | Superiority |
The estimated difference of 4.6 QOL-B units means that the modeled change in the respiratory symptoms score was 4.6 units higher for AZLI-AZLI than for Placebo-AZLI at Day 28. Because higher QOL-B scores represent better quality of life, the direction of this estimate is favorable to the AZLI-AZLI group.
This does not mean that every participant experienced a 4.6-unit improvement, nor does it mean that AZLI improved every aspect of quality of life by 4.6 units. It is a between-group model-based estimate for the registered respiratory symptoms endpoint.
The 95% confidence interval extends from 1.1 to 8.2. This interval describes statistical uncertainty around the estimated group difference; it is not a range containing 95% of individual patient responses.
The P-value of 0.011 quantifies the evidence against the relevant null hypothesis under the specified statistical model. It does not measure the size, clinical importance, or probability of the treatment effect. The magnitude of the estimate and its confidence interval must be considered separately.
The analysis is based on participants meeting the stated analysis-population requirement of having both baseline and Day 28 scores. That feature matters when considering missing outcome measurements. The ClinicalTrials.gov record does not specify the detailed covariance structure, fixed-effect terms, missing-data assumptions, or imputation procedure used by the MMRM, so those aspects should not be inferred.
7. Secondary Result: Day 84 QOL-B Respiratory Symptoms Score
The registry also reports a formal MMRM analysis of the change in QOL-B Respiratory Symptoms Score from baseline to Day 84.
Difference in least squares mean
95% CI: -2.7–5.0 · P = 0.56
Two-sided confidence interval; superiority hypothesis; MMRM analysis.
| Secondary endpoint | AZLI-AZLI vs Placebo-AZLI |
|---|---|
| Outcome | Change in QOL-B Respiratory Symptoms Score at Day 84 |
| Time frame | Baseline to Day 84 |
| Endpoint type | Continuous |
| Analysis population | Participants in the ITT Analysis Set with scores at both baseline and Day 84 |
| Method | MMRM |
| Effect measure | Difference in least squares mean |
| Estimate | 1.1 |
| 95% CI | -2.7–5.0 |
| P-value | 0.56 |
| Hypothesis | Superiority |
The estimated Day 84 difference was 1.1 QOL-B units in favor of AZLI-AZLI in the reported comparison. This is substantially closer to zero than the Day 28 estimate of 4.6.
The 95% confidence interval ranges from -2.7 to 5.0, so the interval includes zero. The data therefore do not provide a statistically clear separation between the randomized groups for this particular Day 84 comparison under the reported model.
The P-value of 0.56 should not be interpreted as the probability that the treatment has no effect. Nor does it demonstrate equivalence between the groups. It is a measure of statistical evidence under the specified null hypothesis and model.
The Day 84 result also illustrates why a single p-value should not be used as a substitute for the estimated effect and its confidence interval. The estimate, interval, endpoint definition, and analysis population together provide the more informative statistical description.
8. Day 28 and Day 84 Results Side by Side
| Feature | Day 28 primary | Day 84 secondary |
|---|---|---|
| Time frame | Baseline to Day 28 | Baseline to Day 84 |
| Analysis | MMRM | MMRM |
| Effect measure | Difference in least squares mean | Difference in least squares mean |
| Estimate | 4.6 | 1.1 |
| 95% CI | 1.1–8.2 | -2.7–5.0 |
| P-value | 0.011 | 0.56 |
| Hypothesis type | Superiority | Superiority |
The two analyses should not be collapsed into a single conclusion. They answer related but distinct questions at different time frames. The Day 28 estimate is positive and its 95% confidence interval does not include zero, whereas the Day 84 estimate is smaller and its confidence interval includes zero.
9. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment sequence and study period. These are presented exactly as provided rather than converted into rates or recomputed percentages.
| Study group | Serious adverse events | Affected / at risk |
|---|---|---|
| AZLI-AZLI (Double-Blind) | Serious adverse events | 18/135 |
| Placebo-AZLI (Double-Blind) | Serious adverse events | 16/137 |
| AZLI-AZLI (Open-Label) | Serious adverse events | 8/112 |
| Placebo-AZLI (Open-Label) | Serious adverse events | 9/110 |
These data contain four group-period descriptions rather than a simple two-arm randomized safety table. The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or p-value for these serious adverse-event counts.
10. Statistical Methods Explained
Why was an MMRM used?
The outcome is a continuous QOL-B respiratory symptoms score measured over time. An MMRM is well suited to repeated longitudinal measurements because it models the outcome across visits while accounting for the fact that observations from the same participant are related. It can therefore provide model-based comparisons at specific time points without treating repeated observations as independent.
What does a least squares mean represent?
A least squares mean is an estimated mean generated by the statistical model. The reported treatment effect is the difference between the modeled least squares means for the two groups. It should not be confused with a simple difference between two raw arithmetic averages unless the model happens to produce that same result.
What does the 4.6 estimate mean?
The estimate of 4.6 is the modeled difference in change in the QOL-B Respiratory Symptoms Score between AZLI-AZLI and Placebo-AZLI at Day 28. Because the score is coded so that higher values represent better quality of life, a positive value points toward a greater improvement in the AZLI-AZLI group for this endpoint.
Why does the confidence interval matter?
The 95% confidence interval of 1.1 to 8.2 shows the statistical uncertainty around the Day 28 estimate. It provides substantially more information than the point estimate alone. Importantly, it describes uncertainty in the estimated group difference, not variability among individual patients and not the probability that a particular patient's response falls within that interval.
Why does the p-value not measure effect size?
The Day 28 p-value of 0.011 provides evidence against the relevant null hypothesis under the specified model, but it does not say whether an effect is large or small. A p-value can be influenced by sample size and variability. Effect size is conveyed by the estimated 4.6-unit difference, while the confidence interval communicates its precision.
What does the Day 84 result tell us?
The Day 84 estimate of 1.1 has a 95% confidence interval of -2.7 to 5.0 and a p-value of 0.56. The interval includes zero, so the analysis does not provide statistically clear evidence of a between-group difference at that time point under the reported MMRM analysis. This should not be converted into a claim that the two treatments are equivalent.
11. Intention-to-Treat Analysis and Outcome Availability
The posted analyses explicitly refer to the ITT Analysis Set. This is important because randomization establishes the original treatment comparison. However, the registry's analysis-population wording adds a second requirement: participants needed scores at both baseline and the relevant follow-up visit.
Randomization principle
Randomization assigns treatment independently of subsequent outcomes and is the foundation of the treatment comparison.
Outcome requirement
The reported analyses included participants with both the baseline and relevant follow-up score.
Why missingness matters
When repeated outcomes are unavailable, the assumptions and model used to handle incomplete longitudinal data become important.
What is not reported
The trial data do not specify the detailed missing-data assumptions or imputation procedure, so no such procedure is attributed to AIR-BX2 here.
This distinction is a useful statistical lesson. "ITT analysis" does not automatically mean that every randomized participant appears as an observed outcome value at every time point. Longitudinal models are often designed specifically to use the available repeated measurements under stated assumptions about the missing-data process.
12. Confidence Intervals and the Primary Finding
The primary Day 28 estimate is 4.6, with a two-sided 95% confidence interval from 1.1 to 8.2.
The entire reported interval is above zero. For this continuous outcome, zero represents no estimated difference between the two groups on the modeled change score. The interval therefore provides statistical evidence of a positive between-group difference at Day 28 under the reported analysis.
The interval should not be treated as a prediction interval for future patients. It also does not state that there is a 95% probability that the true treatment effect lies between 1.1 and 8.2. Rather, it is a confidence interval generated under the statistical framework used for estimation.
At Day 84, the corresponding interval is -2.7 to 5.0. Its inclusion of zero provides a useful contrast with the primary Day 28 result and illustrates why confidence intervals should be examined alongside p-values.
13. P-values and Superiority Testing
Both posted statistical analyses are identified as superiority hypotheses and use two-sided confidence intervals. The Day 28 p-value is 0.011; the Day 84 p-value is 0.56.
| Analysis | Estimate | 95% CI | P-value | Statistical reading |
|---|---|---|---|---|
| Day 28 primary | 4.6 | 1.1–8.2 | 0.011 | Evidence of a positive between-group difference under the reported superiority analysis |
| Day 84 secondary | 1.1 | -2.7–5.0 | 0.56 | No statistically clear between-group difference under the reported analysis |
The phrase "no statistically clear difference" is intentionally different from "no difference." A nonsignificant p-value does not establish that the true effect is exactly zero. The confidence interval is the appropriate place to examine which effect sizes remain compatible with the statistical uncertainty represented by the analysis.
14. Multiplicity, Interim Analysis, and Other Design Features
The ClinicalTrials.gov record identifies the Day 28 QOL-B Respiratory Symptoms Score as the registered primary endpoint and provide one additional secondary statistical analysis at Day 84. However, the ClinicalTrials.gov record does not specify a multiplicity-adjustment procedure, alpha-spending strategy, interim analysis, stopping boundary, or formal endpoint hierarchy beyond the designation of the Day 28 endpoint as primary.
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Multiplicity | No multiplicity-adjustment procedure is specified. |
| Interim analysis | No interim analysis procedure is specified. |
| Alpha spending | No alpha-spending method is specified. |
| Bayesian methods | No Bayesian method is specified. |
| Stratification | No stratification factors are specified. |
| Non-inferiority margin | Not applicable to the reported superiority hypothesis; no non-inferiority margin is provided. |
| Crossover | No crossover procedure is specified in the ClinicalTrials.gov record. |
| Factorial design | The trial is described as parallel rather than factorial. |
| Missing-data/imputation details | No detailed imputation procedure or missing-data assumption is specified. |
These omissions are not evidence that a procedure was absent from the underlying protocol or statistical analysis plan. They mean only that the ClinicalTrials.gov record does not provide the information needed to describe such procedures accurately.
15. What the Mean Difference Does — and Does Not — Mean
The primary estimate of 4.6 is a difference in least squares means for change in the QOL-B Respiratory Symptoms Score between AZLI-AZLI and Placebo-AZLI at Day 28.
It does not mean that every participant experienced a 4.6-unit improvement. It also does not imply that the treatment effect is identical at all time points or across all domains of quality of life.
The 95% CI of 1.1–8.2 describes uncertainty around the estimated treatment contrast. Because the interval is entirely above zero, the reported Day 28 analysis supports a positive between-group difference under the specified model.
The p-value of 0.011 addresses statistical evidence against the relevant null hypothesis. It is not an effect-size measure and should not be used to decide whether a 4.6-unit difference is clinically important without additional clinical context.
16. Day 28 Versus Day 84: A Longitudinal Perspective
AIR-BX2 provides a compact example of why clinical-trial results should be interpreted across the registered time frames rather than through a single isolated number.
The visual comparison is based only on the two reported point estimates and is intended as an educational display, not as a reconstruction of the underlying repeated-measures model. The Day 28 estimate is 4.6, whereas the Day 84 estimate is 1.1.
17. Important Limitations and Interpretation Issues
- Outcome-population restriction: the posted analyses specify participants in the ITT Analysis Set with scores at both baseline and the relevant follow-up time point.
- Missing-data details: the ClinicalTrials.gov record does not state the detailed missing-data assumptions or imputation method underlying the MMRM.
- Model specification: the registry identifies MMRM but does not provide the full fixed-effect, covariance, or estimation specification in the ClinicalTrials.gov record.
- Clinical importance: statistical evidence for a positive difference does not by itself establish the clinical importance of a 4.6-unit QOL-B difference.
- Time-frame interpretation: Day 28 and Day 84 are distinct endpoint analyses. Their estimates should not be combined into a single effect.
- Multiplicity: the ClinicalTrials.gov record does not specify whether or how multiplicity was controlled across the primary and secondary analyses.
- Safety interpretation: serious adverse-event data are reported in four study-period/group categories, without a registry-reported formal comparative analysis.
- Generalizability: the ClinicalTrials.gov record establishes the condition and adult population but do not provide a detailed baseline-characteristics table for evaluating representativeness.
- Unreported design details: randomization ratio, stratification factors, detailed treatment schedule, interim-analysis plan, and other protocol features are not specified in the ClinicalTrials.gov record.
18. Why This Trial Matters Statistically
AIR-BX2 is a useful teaching case because it demonstrates how a randomized clinical trial can use a continuous patient-reported outcome rather than a time-to-event endpoint. Its posted analyses also illustrate the role of longitudinal modeling, least squares means, confidence intervals, p-values, and analysis-population definitions.
| Concept | How it appears in AIR-BX2 |
|---|---|
| Randomization | Participants were randomized to AZLI or placebo. |
| Blinding | The trial is described as double-blind. |
| Parallel design | The registry identifies a parallel design model. |
| Continuous endpoint | The primary QOL-B Respiratory Symptoms Score is analyzed as a continuous outcome. |
| Change from baseline | The primary endpoint measures change from baseline to Day 28. |
| MMRM | The posted analyses use a mixed model for repeated measures. |
| Least squares mean | The treatment contrast is reported as a difference in least squares means. |
| Confidence interval | The primary estimate has a two-sided 95% CI of 1.1–8.2. |
| P-value | The primary analysis reports P = 0.011. |
| ITT analysis set | The posted analysis population is based on the ITT Analysis Set with relevant scores available. |
| Repeated time frame | A secondary analysis evaluates the same QOL-B respiratory symptoms construct at Day 84. |
| Safety analysis | Serious adverse events are reported by study group and period. |
19. Statistical Concepts in This Trial
The AIR-BX2 analysis connects directly to several core biostatistical concepts. Learn more about the methods used in this trial:
20. Related Statistical Calculators
21. Sources
- ClinicalTrials.gov: AIR-BX2 — NCT01314716. Trial registration and posted outcome/statistical-analysis data.
- PubMed: PubMed record for PMID 25154045.
Continue through the Clinical Biostats statistical pathway
Use the trial's methods as a starting point for deeper study of randomized trials, repeated-measures models, confidence intervals, and statistical inference.
22. Record Summary
AIR-BX2 is a phase 3 randomized, double-blind, parallel trial in adults with non-cystic fibrosis bronchiectasis, with 274 participants enrolled and AZLI compared with placebo. Its registered primary endpoint was change in the QOL-B Respiratory Symptoms Score from baseline to Day 28. The posted primary analysis used an MMRM in participants in the ITT Analysis Set with scores at both baseline and Day 28, reporting a difference in least squares mean of 4.6, with a two-sided 95% confidence interval of 1.1–8.2 and P = 0.011.
The secondary Day 84 analysis used the same general MMRM framework and reported a difference in least squares mean of 1.1, with a 95% confidence interval of -2.7–5.0 and P = 0.56. The contrast between these estimates illustrates why statistical interpretation should consider the magnitude of the estimated effect, its uncertainty, the endpoint time frame, and the analysis population rather than relying on p-values alone.
The trial is particularly useful statistically because it demonstrates the analysis of a continuous longitudinal patient-reported outcome in a randomized trial. It also highlights an important practical distinction: an analysis described as ITT may still specify the availability of baseline and follow-up measurements for the particular model-based endpoint analysis.