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
ASPEN was a three-arm, randomized, double-blind phase 3 trial in participants with non-cystic fibrosis bronchiectasis. It asked whether brensocatib, given at 10 mg or 25 mg, reduced the rate of pulmonary exacerbations over 52 weeks compared with placebo.
| Feature | ASPEN |
|---|---|
| Phase | Phase 3 |
| Condition | Non-cystic fibrosis bronchiectasis |
| Design | Randomized, parallel-group, double-masked, placebo-controlled |
| Arms | Brensocatib 10 mg; brensocatib 25 mg; placebo |
| Enrollment | 1767 |
| Primary endpoint | Annualized rate of pulmonary exacerbations (PEs), up to Week 52 |
| Primary analysis | Negative binomial regression; rate ratio versus placebo; ITT analysis set |
| Dates | Start 2020-12-01; primary completion 2024-10-28 |
| ClinicalTrials.gov | NCT04594369 |
| Sponsor | Insmed Incorporated (industry) |
2. Clinical Question
Bronchiectasis is characterized by recurrent pulmonary exacerbations, which drive antibiotic use, hospitalization and symptom burden. ASPEN's central question was whether brensocatib could reduce how often these exacerbations occur over a year of treatment, and whether both tested doses did so relative to placebo.
Population
Participants with non-cystic fibrosis bronchiectasis.
Intervention
Brensocatib 10 mg or brensocatib 25 mg (two separate active arms).
Comparator
Placebo, under double masking.
Primary question
Does each brensocatib dose reduce the annualized rate of adjudicated pulmonary exacerbations up to Week 52 compared with placebo?
3. Trial Design
Brensocatib 10 mg
- Active drug, lower dose
- Compared with placebo
Brensocatib 25 mg
- Active drug, higher dose
- Compared with placebo
Placebo
- Masked control
- Shared reference group for both doses
Because the two active doses share a single placebo group, ASPEN produces two correlated comparisons rather than two independent trials. Any chance imbalance in the placebo arm affects both rate ratios in the same direction, which partly explains why the two dose-versus-placebo estimates look so similar.
Randomization and stratification
The statistical models posted to the registry include the randomization stratification factors as fixed effects: geographic region, sputum Pseudomonas aeruginosa status at screening, number of pulmonary exacerbations in the previous 12 months (<3 versus ≥3), and age group. Including stratification factors in the analysis model is recommended practice: it aligns the analysis with the design and usually improves precision, because these factors are strongly prognostic for exacerbation frequency.
| Analysis population | Definition / role |
|---|---|
| ITT analysis set | All participants who were randomized; used for all posted efficacy analyses. |
| Participants with available data | For FEV1 and QOL-B change-from-baseline endpoints, the registry notes that the number analyzed reflects participants with data available. |
| At-risk population for safety | Serious adverse events are reported per arm as affected / at risk. |
4. Endpoints
Primary endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Annualized Rate of Pulmonary Exacerbations (PEs) | A PE was defined as 3 or more of the following symptoms for at least 48 hours resulting in a physician's decision to prescribe antibiotics: increased cough; increased sputum volume or change in consistency; increased sputum purulence; increased breathlessness and/or decreased exercise tolerance; fatigue and/or malaise; hemoptysis. A severe PE required intravenous antibacterial treatment and/or hospitalization. At least 14 days had to separate one exacerbation onset from the next; events within 14 days of a prior exacerbation were not counted as new. An independent adjudication committee of pulmonary physicians assessed reported events against the protocol definition. Analyzed with a negative binomial model. | Up to Week 52 |
Secondary endpoints with posted analyses
| Endpoint | Time frame | Unit | Posted method |
|---|---|---|---|
| Time to First PE | Up to Week 52 | Weeks | Cox proportional hazards; hazard ratio |
| Responder Status for Exacerbation-Free Over the 52-Week Treatment Period | Up to Week 52 | Percentage of participants | Logistic regression with multiple imputation; odds ratio |
| Change From Baseline at Week 52 in Postbronchodilator FEV1 | Baseline, Week 52 | Liter (L) | Linear repeated measures model; difference in LS means |
| Annualized Rate of Severe PEs | Up to Week 52 | Exacerbation per participant-year | Negative binomial regression; rate ratio |
| Change From Baseline at Week 52 in QOL-B Respiratory Symptoms Domain Score in Adult Participants | Baseline, Week 52 | Score on scale | Linear repeated measures model; LS mean difference |
The endpoint set is a useful illustration of how one clinical phenomenon (exacerbations) can be summarized in several statistically distinct ways: as a count per unit time (rate), as a time to first event, and as a binary event-free status. Each uses a different model and answers a slightly different question.
5. Primary Results: Annualized Rate of Pulmonary Exacerbations
Both brensocatib doses were compared with placebo in the ITT analysis set using negative binomial regression, with results expressed as rate ratios (brensocatib divided by placebo) in units of exacerbations per participant-year.
Brensocatib 10 mg vs placebo
Rate ratio for pulmonary exacerbations
95% CI: 0.680–0.916 (two-sided) · P = 0.0019
Negative binomial regression · ITT analysis set · superiority hypothesis
What the estimate means. A rate ratio of 0.789 means that, after adjustment for the stratification factors and accounting for each participant's time at risk, the estimated exacerbation rate in the 10 mg group was 78.9% of the placebo rate, a 21.1% lower estimated rate.
What it does not mean. It does not mean that 21.1% fewer participants had an exacerbation, nor that each participant's own exacerbation frequency fell by 21.1%. The rate ratio compares average event rates across groups; the proportion of participants who remained exacerbation-free is a separate endpoint (see below).
Precision. The 95% confidence interval of 0.680 to 0.916 excludes 1, and its upper limit corresponds to an estimated rate reduction of 8.4%. The data are therefore compatible with effects ranging from modest to fairly substantial, which is worth keeping in mind when judging clinical relevance.
Why the p-value is not the effect size. P = 0.0019 indicates that data this far from a rate ratio of 1 would be unlikely if brensocatib 10 mg had no effect on the exacerbation rate. It says nothing directly about how large the reduction is; a much smaller effect in a larger trial could produce the same p-value.
Cautions. Two dose comparisons were tested against a shared placebo group, and the registry does not describe how the type I error was divided or ordered across them. Both p-values are small, so the conclusion is unlikely to depend on the multiplicity procedure, but formal claims should rest on the prespecified testing strategy. The robust sandwich covariance estimator protects the standard error against some misspecification of the negative binomial variance.
Brensocatib 25 mg vs placebo
Rate ratio for pulmonary exacerbations
95% CI: 0.694–0.936 (two-sided) · P = 0.0046
Negative binomial regression · ITT analysis set · superiority hypothesis
What the estimate means. A rate ratio of 0.806 corresponds to an estimated exacerbation rate in the 25 mg group that was 19.4% lower than in the placebo group, under the same adjusted model and time-at-risk offset.
What it does not mean. The higher dose did not show a larger estimated effect than the lower dose, but that observation alone does not show that the doses are equivalent or that the lower dose is better. The two confidence intervals overlap almost completely, and no direct 25 mg versus 10 mg comparison is posted.
Precision. The 95% CI of 0.694 to 0.936 excludes 1. Its width is very similar to that of the 10 mg comparison, which is expected when arm sizes and follow-up are similar and both comparisons share the placebo group.
Why the p-value is not the effect size. P = 0.0046 is larger than the 10 mg p-value, but that reflects a slightly smaller estimated effect relative to its standard error, not a clinically meaningful difference between doses.
Cautions. The same multiplicity consideration applies. Because the analysis is on the ITT set, participants who discontinued treatment are analyzed in their randomized arm for the time at risk they contributed; how off-treatment follow-up was handled is not described in the registry.
| Comparison | Rate ratio | 95% CI | P-value | Estimated rate reduction |
|---|---|---|---|---|
| Brensocatib 10 mg vs placebo | 0.789 | 0.680–0.916 | 0.0019 | 21.1% |
| Brensocatib 25 mg vs placebo | 0.806 | 0.694–0.936 | 0.0046 | 19.4% |
ClinicalTrials.gov records the between-group rate ratios; group-specific annualized exacerbation rates for each arm are not included in the posted statistical analyses summarized here, so the absolute size of the reduction (exacerbations avoided per participant-year) cannot be read from the rate ratio alone.
6. Secondary Endpoint Results
Time to first pulmonary exacerbation
| Comparison | Hazard ratio | 95% CI | P-value |
|---|---|---|---|
| Brensocatib 10 mg vs placebo | 0.813 | 0.695–0.952 | 0.01 |
| Brensocatib 25 mg vs placebo | 0.825 | 0.703–0.968 | 0.0182 |
The Cox model included treatment, sputum P. aeruginosa status at screening, prior exacerbations (<3 or ≥3 in the last 12 months), region and age group, with a robust sandwich covariance estimator. A hazard ratio of 0.813 means an 18.7% lower estimated instantaneous rate of a first exacerbation with 10 mg; 0.825 corresponds to 17.5% with 25 mg. This endpoint uses only the first event per participant, so it discards information on repeat exacerbations that the rate endpoint retains. The agreement between the two endpoints suggests the rate-ratio result is not driven solely by a few participants with many events.
Exacerbation-free responder status over 52 weeks
| Comparison | Odds ratio | 95% CI | P-value |
|---|---|---|---|
| Brensocatib 10 mg vs placebo | 1.412 | 1.105–1.806 | 0.0059 |
| Brensocatib 25 mg vs placebo | 1.4 | 1.095–1.792 | 0.0074 |
Here an odds ratio above 1 favors brensocatib, because the event being modeled is remaining exacerbation-free. An odds ratio of 1.412 means the estimated odds of being exacerbation-free were 1.412 times higher with 10 mg than with placebo. Odds ratios are not risk ratios: when the outcome is common, the odds ratio overstates the relative change in the proportion of responders. Missing responder status was imputed 100 times, each dataset was analyzed with logistic regression, and results were combined using Rubin's rules.
Annualized rate of severe pulmonary exacerbations
| Comparison | Rate ratio | 95% CI | P-value |
|---|---|---|---|
| Brensocatib 10 mg vs placebo | 0.742 | 0.505–1.089 | 0.1277 |
| Brensocatib 25 mg vs placebo | 0.74 | 0.515–1.062 | 0.1025 |
The point estimates for severe exacerbations (0.742 and 0.74) are numerically lower than those for all exacerbations, but the confidence intervals are much wider and include 1. Severe exacerbations are rarer events, so the count data carry less information and the estimates are less precise. A non-significant result here is not evidence that brensocatib has no effect on severe exacerbations; it means the trial could not estimate that effect precisely enough to exclude no difference.
Postbronchodilator FEV1, change from baseline at Week 52
| Comparison | Difference in LS means (L) | 95% CI | P-value |
|---|---|---|---|
| Brensocatib 10 mg vs placebo | 0.011 | -0.014 to 0.037 | 0.3841 |
| Brensocatib 25 mg vs placebo | 0.038 | 0.011 to 0.065 | 0.0054 |
The repeated-measures model included treatment, visit, the stratification factors as fixed effects and baseline FEV1 as a covariate, with a robust sandwich covariance estimator. The 25 mg comparison shows an estimated difference of 0.038 L with a confidence interval excluding zero; the 10 mg comparison does not. This is the clearest dose-related pattern among the posted results, but differences of a few hundredths of a liter should be interpreted alongside established thresholds for clinical relevance, and the analysis includes only participants with available data.
QOL-B Respiratory Symptoms domain score (adults), change from baseline at Week 52
| Comparison | LS mean difference | 95% CI | P-value |
|---|---|---|---|
| Brensocatib 10 mg vs placebo | 2.031 | -0.081 to 4.143 | 0.0594 |
| Brensocatib 25 mg vs placebo | 3.766 | 1.680 to 5.852 | 0.0004 |
As with FEV1, the 25 mg comparison excludes zero and the 10 mg comparison narrowly does not. A p-value of 0.0594 should not be read as "no effect" and 0.0004 as "large effect"; the two intervals overlap substantially, and the difference between a result just above and just below 0.05 is not itself statistically meaningful.
7. Safety: Serious Adverse Events
| Arm | Participants with serious adverse events (affected / at risk) |
|---|---|
| Brensocatib 10 mg | 101 / 582 |
| Brensocatib 25 mg | 97 / 574 |
| Placebo | 108 / 563 |
Serious adverse event counts were numerically similar across the three arms, with the placebo arm reporting the largest number. These are descriptive counts of participants with at least one serious event; no formal between-arm test is posted. Safety summaries of this kind are not designed to demonstrate equivalence, and rare but important adverse events require event-specific review rather than an aggregate count.
8. Statistical Methodology
Negative binomial regression for exacerbation rates
Exacerbations are counts that accumulate over a follow-up period that differs between participants. A Poisson model would assume the variance equals the mean, but exacerbation counts are typically overdispersed: some participants have many events while most have few. The negative binomial model adds a dispersion parameter that allows for this between-participant heterogeneity.
Yi is the number of exacerbations for participant i, ti is time at risk, entered on the log scale as an offset. The rate ratio is exp(β1).
The offset is what converts a count model into a rate model: a participant followed for half a year contributes proportionally less expected exposure than one followed for the full 52 weeks. This is how the analysis handles unequal follow-up without discarding participants who left early.
Robust (sandwich) covariance estimation
The negative binomial, Cox and FEV1 repeated-measures analyses explicitly used a robust sandwich covariance estimator. This computes standard errors from the observed variability of the data rather than relying entirely on the model's variance assumptions, making confidence intervals and p-values less sensitive to misspecification of the dispersion or correlation structure.
Cox proportional-hazards model
Time to first exacerbation was analyzed with a Cox model adjusted for the stratification factors. The hazard ratio summarizes the relative instantaneous rate of a first event under the assumption that this ratio is roughly constant over the 52 weeks. Participants without an exacerbation are censored at the end of their follow-up.
Logistic regression with multiple imputation
For the exacerbation-free responder endpoint, participants with incomplete follow-up have unknown responder status. Rather than counting them as failures or excluding them, missing status was imputed 100 times, each completed dataset was analyzed by logistic regression, and estimates were combined with Rubin's rules, which add between-imputation variability to within-imputation variability so that the final confidence interval reflects uncertainty from missing data.
Linear repeated measures models
FEV1 and QOL-B were analyzed with repeated-measures models using all post-baseline visits, with treatment-by-visit structure, the stratification factors and baseline value as covariates. The Week 52 treatment difference is estimated as a difference in least-squares (model-adjusted) means. Such models typically use information from participants with partial data under a missing-at-random assumption.
Intention-to-treat principle
All posted efficacy analyses used the ITT analysis set of all randomized participants. This preserves the comparability created by randomization, so the estimated rate ratios reflect the effect of being assigned to brensocatib rather than the effect among those who adhered.
9. Statistical Methods Explained
Why was a negative binomial model used instead of comparing the proportion with an exacerbation?
A proportion throws away information: a participant with one exacerbation and one with five are treated alike. The negative binomial model uses every event and each participant's time at risk, and it accommodates the overdispersion typical of exacerbation data. The proportion-based view was still examined, as the exacerbation-free responder endpoint.
What does a rate ratio of 0.789 mean in practice?
It means the adjusted exacerbation rate with brensocatib 10 mg was estimated at 78.9% of the placebo rate, a 21.1% relative reduction. Translating this into exacerbations avoided per participant-year requires the placebo group's rate; the same relative reduction is worth more in patients who exacerbate frequently.
Why is time at risk included as an offset?
Participants who discontinue early have less opportunity to experience exacerbations. Entering log(time at risk) as an offset models events per unit of exposure, so early dropouts contribute appropriately rather than looking artificially exacerbation-free.
Why does the responder endpoint report an odds ratio above 1 while the other endpoints report ratios below 1?
The direction depends on how the outcome is coded. The rate and hazard ratios model the occurrence of exacerbations, so values below 1 favor brensocatib. The responder endpoint models being exacerbation-free, a favorable outcome, so values above 1 favor brensocatib.
Why were severe exacerbations not statistically significant when all exacerbations were?
Severe events are a subset of all events and occur less often. With fewer events, the estimate is less precise: the severe-PE confidence intervals (for example 0.505 to 1.089) are much wider than the primary ones (0.680 to 0.916). A similar point estimate with a wider interval can cross 1 without implying a smaller true effect.
What are Rubin's rules, and why were 100 imputations used?
Rubin's rules average the estimates from each imputed dataset and combine within- and between-imputation variance into a single standard error. Using many imputations reduces Monte Carlo noise in the combined result, so conclusions do not depend on the random draws used for imputation.
Why does a significant result for 25 mg and a non-significant one for 10 mg on FEV1 not prove a dose effect?
"Significant versus not significant" is not a test of the difference between two effects. Establishing that 25 mg outperforms 10 mg would require a direct comparison of the two doses with its own confidence interval, which is not among the posted analyses.
10. Limitations
- Multiplicity: two doses, a primary endpoint and several secondary endpoints were each tested against placebo. The registry does not describe the hierarchical or alpha-splitting procedure, so secondary p-values should not be treated as independent confirmatory tests.
- Shared control group: both dose comparisons use the same placebo arm, so their results are correlated and not independent replications.
- Relative measures only: the posted analyses report rate, hazard and odds ratios; relative effects alone do not convey the absolute number of exacerbations avoided.
- Missing data assumptions: the repeated-measures and multiple-imputation analyses rely on assumptions (typically missing at random) that cannot be verified from the data; sensitivity analyses are not part of the posted results.
- Event definition: the PE definition requires a physician's decision to prescribe antibiotics, so prescribing practices across regions can influence event counts, although adjudication and masking reduce systematic bias between arms.
- Proportional hazards: the time-to-first-PE hazard ratio is a single summary over 52 weeks; if the treatment effect changed over time, the hazard ratio would average over that change.
- Subgroups and baseline data: ClinicalTrials.gov does not report subgroup analyses or a baseline characteristics comparison in the posted statistical analyses, so consistency across patient types cannot be assessed here.
- Duration: follow-up extends to Week 52; effects beyond one year are not addressed by these results.
11. Why This Trial Matters Statistically
ASPEN is a compact teaching example of how recurrent-event outcomes are analyzed in respiratory trials, and of how a single clinical process can be viewed through several complementary statistical lenses.
| Concept | How it appears in ASPEN |
|---|---|
| Recurrent-event count data | Primary endpoint analyzed as an annualized rate with negative binomial regression |
| Offset for exposure time | Log time at risk converts counts into rates and handles unequal follow-up |
| Rate ratio | 0.789 and 0.806 for the two doses versus placebo |
| Hazard ratio | Time to first PE via Cox proportional-hazards model |
| Odds ratio | Exacerbation-free responder status via logistic regression |
| Multiple imputation | 100 imputations combined with Rubin's rules |
| Repeated measures | FEV1 and QOL-B change from baseline at Week 52 |
| Stratified randomization | Region, P. aeruginosa status, prior PEs and age group in every model |
| Robust variance | Sandwich covariance estimator in the rate, Cox and FEV1 models |
| Multi-arm design | Two active doses against a shared placebo, raising multiplicity and dose-response questions |
| Precision vs significance | Severe PE rate ratios with similar point estimates but wide intervals crossing 1 |
12. Related Tutorials
Learn more about the methods used in this trial:
13. Related Calculators
14. Sources
- ClinicalTrials.gov: NCT04594369: A Study to Assess the Efficacy, Safety, and Tolerability of Brensocatib in Participants With Non-Cystic Fibrosis Bronchiectasis.
- PubMed: PMID 42461571
- PubMed: PMID 41519018
- PubMed: PMID 40267423
- PubMed: PMID 39040578
- PubMed: PMID 35976570
Continue exploring clinical trial statistics
Connect each endpoint and method in this trial to a deeper statistical tutorial, then to an appropriate calculator.
15. Record Summary
ASPEN compared two doses of brensocatib with placebo in 1767 participants with non-cystic fibrosis bronchiectasis. Both doses produced primary rate ratios below 1 with confidence intervals excluding 1 (0.789 and 0.806). Time to first exacerbation and exacerbation-free status pointed in the same direction, while severe exacerbations were estimated imprecisely. For FEV1 and the QOL-B respiratory symptoms score, only the 25 mg comparison had an interval excluding zero. Reading the trial well means combining relative effects, interval precision, the choice of count, time-to-event and binary models, and the multiplicity structure of a two-dose design.