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
BATURA was a randomized, double-blind, parallel-group phase 3 trial asking whether an as-needed BDA metered-dose inhaler (PT027) reduces severe asthma exacerbations compared with an as-needed AS metered-dose inhaler (PT007). Its primary endpoint was the time to the first severe asthma exacerbation, analyzed with a Cox regression model at a prespecified interim analysis.
| Feature | BATURA |
|---|---|
| Registry title | A Comparison of PT027 vs PT007 Used as Needed in Participants With Asthma |
| Phase | Phase 3 |
| Condition | Asthma |
| Design | Randomized, parallel-group, double-masked, treatment purpose |
| Arms | 2: BDA MDI (PT027) 160/180 μg vs AS MDI (PT007) 180 µg, both used as needed |
| Enrollment | 2,516 |
| Primary endpoint | Time to first severe asthma exacerbation, While on Treatment strategy, interim analysis data cut-off (participants aged ≥12 years) |
| Primary analysis | Cox regression; hazard ratio with two-sided 95% CI |
| Status | Completed; results posted |
| Key dates | Start 2022-09-02; interim analysis data cut-off 22 April 2024; primary completion 2024-08-22 |
| Sponsor | Bond Avillion 2 Development LP (industry) |
| ClinicalTrials.gov | NCT05505734 |
2. Clinical Question
The trial compared two as-needed (reliever) inhaler strategies. The central question was whether using BDA MDI as needed lengthens the time until a participant experiences a severe asthma exacerbation, compared with using AS MDI as needed. Because both arms received an active reliever, this is a head-to-head comparison of two reliever strategies rather than a comparison against placebo.
Population
Participants with asthma. The primary analysis used the full analysis set of participants aged ≥12 years; supporting analyses were repeated in participants aged ≥18 years.
Intervention
BDA MDI (PT027) 160/180 μg, a metered-dose inhaler used as needed.
Comparator
AS MDI (PT007) 180 µg, a metered-dose inhaler used as needed.
Primary question
Does as-needed BDA MDI reduce the hazard of a first severe asthma exacerbation relative to as-needed AS MDI while participants remain on their assigned treatment?
3. Trial Design
BDA MDI (PT027) 160/180 μg
- Metered-dose inhaler
- Used as needed
- Experimental arm in all posted comparisons
AS MDI (PT007) 180 µg
- Metered-dose inhaler
- Used as needed
- Active comparator (reference group for hazard and rate ratios)
The registry reports the number of participants at risk for adverse events in each arm but does not report the number randomized per arm, the allocation ratio or the number analyzed in each efficacy population. The near-identical at-risk counts are consistent with approximately equal-sized arms.
4. Analysis Populations and Estimands
All posted efficacy analyses used the full analysis set (FAS), split into two age-defined populations: participants aged ≥12 years and participants aged ≥18 years. The ≥18-year population is a subset of the ≥12-year population, so the two sets of results are not independent replications; they largely share the same participants and events.
BATURA is also a clear example of the estimand framework in practice. The same endpoint was analyzed under two strategies for handling intercurrent events such as treatment discontinuation or a step-up in maintenance therapy:
| Strategy | How follow-up ends | Question answered |
|---|---|---|
| While on Treatment | Censored at treatment discontinuation, a step-up in maintenance therapy, study completion or (for interim analyses) the data cut-off | What is the effect while participants remain on their assigned reliever and their maintenance therapy is unchanged? |
| Treatment Policy | Followed until study completion or withdrawal only, regardless of treatment discontinuation or step-up in maintenance therapy | What is the effect of being assigned to the reliever strategy, whatever happens afterwards? |
The While on Treatment strategy was used for the primary endpoint. The Treatment Policy strategy is closer in spirit to an intention-to-treat analysis, because it keeps counting events after participants deviate from their assigned treatment. Agreement between the two strategies is informative: it suggests that the estimated effect is not driven by how intercurrent events were handled.
5. Endpoints
Primary endpoint
| Endpoint | Definition | Time frame |
|---|---|---|
| Time to First Severe Asthma Exacerbation, While on Treatment Strategy, Interim Analysis Data Cut-off (Participants Aged ≥12 Years) | Length in days from treatment initiation until the date of the first severe asthma exacerbation, in the FAS. Participants were censored at treatment discontinuation, a step-up in maintenance therapy, or interim analysis DCO. An exacerbation was severe if it resulted in at least one of: use of systemic steroids for at least 3 days; inpatient hospitalization due to asthma or an emergency room visit that required systemic steroids; or death due to asthma. | From treatment initiation until the earliest of study completion, treatment discontinuation, step-up in maintenance therapy, or interim analysis DCO, up to one year following treatment initiation. |
The registry notes that the primary outcome was analyzed at the interim analysis (22 April 2024), and that because the primary objective was met at the interim analysis, it was not re-tested at final database lock. It also notes that fewer than 50% of participants had an exacerbation event, so median time to first exacerbation could not be calculated; the number of participants with an event is reported instead.
Secondary endpoints with posted analyses
| Endpoint | Population | Time frame (shortened) | Method |
|---|---|---|---|
| Time to first severe exacerbation, Treatment Policy strategy, interim analysis DCO | FAS, ≥12 y | Until study completion or withdrawal only, up to one year | Cox regression |
| Time to first severe exacerbation, While on Treatment strategy | FAS, ≥18 y | Until study completion, treatment discontinuation or step-up, up to one year | Cox regression |
| Time to first severe exacerbation, Treatment Policy strategy | FAS, ≥18 y | Until study completion or withdrawal only, up to one year | Cox regression |
| Annualized severe asthma exacerbation rate, While on Treatment strategy | FAS, ≥12 y | Until study completion, treatment discontinuation or step-up, up to one year | Negative binomial |
| Annualized severe asthma exacerbation rate, While on Treatment strategy | FAS, ≥18 y | Until study completion, treatment discontinuation or step-up, up to one year | Negative binomial |
| Total systemic glucocorticoid exposure (mg/year), While on Treatment strategy | FAS, ≥12 y | Until study completion, treatment discontinuation or step-up, up to one year | Wilcoxon rank sum |
| Total systemic glucocorticoid exposure (mg/year), While on Treatment strategy | FAS, ≥18 y | Until study completion, treatment discontinuation or step-up, up to one year | Wilcoxon rank sum |
Note that the two ≥18-year time-to-event analyses do not carry the "interim analysis data cut-off" label, which indicates that they use the full follow-up rather than the interim data cut.
6. Primary Result: Time to First Severe Asthma Exacerbation
The primary analysis compared BDA MDI with AS MDI in the FAS aged ≥12 years under the While on Treatment strategy, at the interim analysis data cut-off.
Hazard ratio for first severe asthma exacerbation
Two-sided 95% CI: 0.392–0.730 · P < 0.001
Cox regression with treatment group, pre-study asthma therapy and number of severe exacerbations in the 12 months before randomization as factors. HR < 1 favors BDA MDI.
What the estimate means. A hazard ratio of 0.535 means that, under the fitted Cox model and over the observed follow-up, the estimated instantaneous rate of a first severe exacerbation among participants using BDA MDI was about 53.5% of that among participants using AS MDI, corresponding to a 46.5% lower estimated hazard.
What it does not mean. It does not mean that 46.5% of exacerbations were prevented, that 46.5% of participants benefited, or that each individual's risk fell by that amount. It is a relative measure; it says nothing by itself about the absolute probability of an exacerbation within a year. Because fewer than half of participants had an event, a median time to exacerbation could not be estimated, so the absolute picture has to come from event counts or time-point probabilities rather than medians.
Precision. The 95% confidence interval of 0.392 to 0.730 lies entirely below 1. Values as favorable as a 60.8% lower hazard and as modest as a 27.0% lower hazard are both compatible with the data under the model. The interval excludes no effect, but it is fairly wide, which is typical of a relatively uncommon event analyzed at an interim cut.
Why the p-value is not the effect size. P < 0.001 says the data would be very unlikely if the true hazard ratio were 1. It does not say how large the benefit is; that information is in the hazard ratio and its interval. A much smaller effect in a larger trial could produce the same p-value.
Cautions. (1) A single hazard ratio assumes approximately proportional hazards over the one-year window; the registry does not report a test of this assumption. (2) The result comes from an interim analysis; its confirmatory status depends on the prespecified interim boundary, which the registry does not describe. Estimates from trials stopped or declared successful at an interim look can overstate the true effect somewhat. (3) Under the While on Treatment strategy, participants were censored at treatment discontinuation or a step-up in maintenance therapy. If those events were related to worsening asthma, censoring is not independent of the outcome, which is why the Treatment Policy analysis is an important check.
7. Secondary Results
Time-to-event analyses under alternative estimands and populations
| Analysis | Population | HR (95% CI) | P-value |
|---|---|---|---|
| While on Treatment, interim DCO (primary) | FAS ≥12 y | 0.535 (0.392–0.730) | <0.001 |
| Treatment Policy, interim DCO | FAS ≥12 y | 0.539 (0.397–0.733) | <0.001 |
| While on Treatment | FAS ≥18 y | 0.541 (0.405–0.724) | <0.001 |
| Treatment Policy | FAS ≥18 y | 0.541 (0.406–0.720) | <0.001 |
The four hazard ratios are almost identical (0.535 to 0.541). The Treatment Policy estimate of 0.539 at the interim cut shows that keeping participants in the analysis after they stopped treatment or stepped up maintenance therapy did not dilute the estimated effect. The ≥18-year analyses, which use longer follow-up, have slightly narrower intervals, consistent with more accumulated events.
Annualized severe exacerbation rate
| Population | Rate ratio (95% CI) | P-value | Reading |
|---|---|---|---|
| FAS ≥12 y, While on Treatment | 0.47 (0.34–0.64) | <0.001 | 53% lower estimated annualized rate |
| FAS ≥18 y, While on Treatment | 0.46 (0.33–0.63) | <0.001 | 54% lower estimated annualized rate |
The rate analysis counts all severe exacerbations, not just the first. A rate ratio of 0.47 means the model-estimated number of severe exacerbations per year of exposure with BDA MDI was 47% of that with AS MDI. The similarity between the rate ratio and the hazard ratio suggests that the benefit extends beyond delaying the first event to reducing the overall burden of exacerbations.
Total systemic glucocorticoid exposure
Total systemic glucocorticoid exposure (mg/year) was compared between arms with a Wilcoxon rank sum test in both the ≥12-year and ≥18-year populations, each with P < 0.001. The posted analyses give only the p-value for this endpoint, with no accompanying effect estimate or confidence interval. The result therefore indicates that the distributions of exposure differed between arms, but not by how much.
A rank-based test is a natural choice here. Systemic glucocorticoid exposure is heavily skewed: many participants have no exacerbation and therefore zero or near-zero exposure, while a few have repeated courses. A t-test on means would be dominated by those few participants; the Wilcoxon test compares the whole distribution through ranks and is robust to that skew. The registry does not state the direction of the difference alongside the p-value, although the fewer severe exacerbations with BDA MDI, which by definition involve systemic steroid use, make lower exposure in that arm the expected pattern.
8. Safety: Serious Adverse Events
| Arm | Participants with serious adverse events | Participants at risk |
|---|---|---|
| BDA MDI (PT027) 160/180 μg | 37 | 1209 |
| AS MDI (PT007) 180 µg | 37 | 1212 |
The same number of participants (37) experienced at least one serious adverse event in each arm, out of almost identical at-risk populations. The registry does not post a formal statistical comparison of serious adverse events. With counts this small and equal, any confidence interval for a between-arm difference would be centered near zero and would still be wide enough to leave uncertainty about rare harms; safety comparisons in efficacy trials are generally descriptive rather than hypothesis tests.
9. Statistical Methodology
Cox proportional-hazards regression
All four time-to-first-exacerbation analyses used a Cox regression model with treatment group, pre-study asthma therapy and number of severe exacerbations in the 12 months before randomization as factors. The Cox model estimates the ratio of hazards between arms without assuming any particular shape for the baseline hazard over time.
The reported hazard ratio is exp(β1): the ratio of hazards for BDA MDI versus AS MDI, holding the two baseline factors fixed. Adjusting for strong prognostic factors such as prior exacerbation history typically increases precision and gives a conditional estimate for participants with the same baseline profile.
Censoring
Participants without an exacerbation were censored at the end of their observed follow-up. Under the While on Treatment strategy, censoring also occurred at treatment discontinuation and at a step-up in maintenance therapy. The Cox model and Kaplan-Meier-type estimation assume that censoring is non-informative, that is, unrelated to the risk of the event after conditioning on the model factors. Because fewer than 50% of participants had an event, the survival curve never fell below 0.5, which is why a median time to first exacerbation is not estimable.
Negative binomial regression for exacerbation rates
Annualized exacerbation rates and rate ratios were estimated from a negative binomial model with the same factors: treatment, pre-study asthma therapy and prior-year exacerbation count. Exacerbation counts are overdispersed: most participants have zero events while some have several. A Poisson model would assume the variance equals the mean and would understate uncertainty; the negative binomial adds a dispersion parameter to account for between-participant heterogeneity.
The log of each participant's follow-up time enters as an offset, so the model estimates rates per unit time and handles unequal follow-up. The rate ratio is exp(β1).
Wilcoxon rank sum test
Systemic glucocorticoid exposure was compared with a Wilcoxon rank sum (Mann-Whitney) test, a nonparametric comparison of two distributions based on ranks. It makes no normality assumption and is not dominated by extreme values, but it yields a p-value rather than a directly interpretable difference in mg/year.
Interim analysis
The primary endpoint was analyzed at an interim analysis with a data cut-off of 22 April 2024. Because the primary objective was met at that interim analysis, it was not re-tested at final database lock. In group-sequential designs, interim efficacy analyses are typically judged against a prespecified boundary stricter than the final-analysis threshold so that the overall type I error is preserved; the registry does not state the boundary or the alpha allocated to the interim look.
Superiority framework and two-sided inference
Every posted analysis is labelled as a superiority comparison, with two-sided 95% confidence intervals. The aim was to show that BDA MDI is better than AS MDI, not merely no worse, so no non-inferiority margin is involved.
10. Statistical Methods Explained
Why was a Cox model used for the primary endpoint rather than simply comparing the proportion of participants with an exacerbation?
Participants had different lengths of follow-up: some were censored at the interim cut-off soon after starting, others at treatment discontinuation or step-up. Comparing raw proportions would treat a participant followed for a few weeks the same as one followed for a year. The Cox model uses each participant's observed time at risk and handles censoring correctly, and it also allows adjustment for prior therapy and prior exacerbation history.
What does a hazard ratio of 0.535 mean in practice?
At any point during follow-up, among participants who had not yet had a severe exacerbation, the model estimates that those using BDA MDI experienced a first severe exacerbation at roughly 53.5% of the rate seen with AS MDI, a 46.5% lower estimated hazard. It is a relative rate of first events, not a percentage of participants who were protected.
Why report both a hazard ratio and a rate ratio?
They answer different questions. The hazard ratio (0.535) concerns the first exacerbation only. The rate ratio (0.47) concerns all severe exacerbations per year of exposure, including second and later events. A treatment could delay a first event without reducing recurrences, or vice versa; in BATURA the two measures point in the same direction and are of similar size.
Why a negative binomial model instead of Poisson regression for exacerbation counts?
Exacerbations cluster in a minority of participants, so the variance of the counts exceeds the mean. Poisson regression ignores this and produces confidence intervals that are too narrow. The negative binomial model adds a dispersion parameter, giving more honest uncertainty for the rate ratio of 0.47 (0.34–0.64).
What is the difference between the While on Treatment and Treatment Policy results?
The While on Treatment strategy stops counting at treatment discontinuation or a step-up in maintenance therapy; the Treatment Policy strategy continues counting regardless. The first estimates the effect under continued use of the assigned reliever, the second the effect of being assigned to it. The hazard ratios, 0.535 and 0.539 at the interim cut, are almost identical, so the conclusion does not depend on the estimand choice.
Why does the glucocorticoid comparison have a p-value but no effect estimate?
The Wilcoxon rank sum test compares the ranks of values between arms. It produces a test statistic and a p-value but not a mean difference. An effect size, such as a Hodges-Lehmann shift estimate with its confidence interval, can accompany a rank test, but none is posted for this endpoint in the registry, so the size of the difference in mg/year cannot be read from the posted analysis.
Why could median time to first exacerbation not be calculated?
A median event time is the point at which the estimated probability of remaining event-free falls to 50%. Because fewer than half of participants had a severe exacerbation during follow-up, that point was never reached in either arm. This is common for relatively infrequent events and is not a sign of a problem with the analysis; it simply means the hazard ratio and event counts are more informative summaries than medians.
11. Statistical vs Clinical Interpretation
Statistical interpretation
The primary Cox analysis gave HR 0.535 (95% CI 0.392–0.730, P < 0.001), with consistent estimates across two estimands, two age populations and a separate rate analysis. The evidence against no difference is strong and the direction is consistent across all posted efficacy analyses.
Clinical interpretation
The size of the benefit for a given patient depends on their baseline risk of severe exacerbation, which the relative measures alone do not convey. Serious adverse events were equal in number between arms. Weighing benefit and harm requires absolute event rates and fuller safety detail than the relative estimates provide.
12. Limitations
- Interim-analysis estimate: the primary result comes from an interim data cut. Effects estimated at a successful interim look can be somewhat larger than the true effect, and the interim boundary is not described in the registry.
- Relative measures only: the posted analyses report hazard and rate ratios. Without absolute exacerbation rates per arm, the number of participants who would need to use BDA MDI to avoid one severe exacerbation cannot be judged from these estimates alone.
- Proportional hazards: a single hazard ratio summarizes the effect over up to one year; the registry does not report a check of the proportional-hazards assumption.
- Informative censoring: under the While on Treatment strategy, censoring at discontinuation or step-up in maintenance therapy may be related to asthma control. The similar Treatment Policy estimate is reassuring but does not remove this concern entirely.
- Overlapping populations: the ≥18-year analyses use a subset of the ≥12-year participants, so their agreement is expected and should not be counted as independent confirmation.
- Multiplicity: the registry does not describe the testing hierarchy across secondary endpoints and populations, so secondary p-values should not be treated as separately confirmatory.
- Glucocorticoid endpoint: only a p-value is posted, so the magnitude of the reduction in systemic steroid exposure cannot be assessed from the posted analysis.
- Missing detail: the registry does not report baseline characteristics, subgroup analyses or the numbers randomized and analyzed per arm alongside the effect estimates, limiting assessment of generalizability and effect heterogeneity.
13. Why This Trial Matters Statistically
BATURA is a compact teaching case for how a single clinical question can be examined through several complementary statistical lenses, and for how estimand choices are made explicit in a modern trial.
| Concept | How it appears in BATURA |
|---|---|
| Time-to-event endpoint | Time to first severe asthma exacerbation, measured in days from treatment initiation |
| Censoring | At interim cut-off, treatment discontinuation, step-up in maintenance therapy or study completion |
| Cox model with covariates | HR adjusted for pre-study asthma therapy and prior-year exacerbation count |
| Non-estimable median | Fewer than 50% had an event, so event counts replace median times |
| Estimands | While on Treatment versus Treatment Policy strategies give HRs of 0.535 and 0.539 |
| Count data and overdispersion | Negative binomial rate ratios of 0.47 and 0.46 |
| Nonparametric testing | Wilcoxon rank sum test for skewed glucocorticoid exposure |
| Interim analysis | Primary objective met at the 22 April 2024 interim cut and not re-tested at final lock |
| Confidence intervals | All posted two-sided 95% intervals exclude 1 |
| Safety reporting | Serious adverse events: 37/1209 vs 37/1212, summarized descriptively |
14. Related Tutorials
Learn more about the methods used in this trial:
15. Related Calculators
16. Sources
- ClinicalTrials.gov: NCT05505734 — A Comparison of PT027 vs PT007 Used as Needed in Participants With Asthma (BATURA).
- PubMed: PMID 40388330.
Continue learning with Clinical Biostats
Explore the statistical methods behind this trial in more depth, or apply them with our calculators.
17. Record Summary
BATURA compared two as-needed reliever inhalers in 2,516 participants with asthma. The primary Cox analysis of time to first severe exacerbation gave a hazard ratio of 0.535 (95% CI 0.392–0.730, P < 0.001) at the interim analysis, and every posted secondary analysis pointed the same way: hazard ratios of 0.539 to 0.541 across estimands and age populations, exacerbation rate ratios of 0.47 and 0.46, and Wilcoxon tests of systemic glucocorticoid exposure with P < 0.001. Serious adverse events occurred in 37 participants in each arm. The most useful reading combines the relative effect estimates, their confidence intervals, the estimand behind each analysis and the interim-analysis context, while recognizing that absolute benefit depends on each patient's baseline exacerbation risk.