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
DB2113373 was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating once-daily UMEC/VI 62.5/25 mcg, UMEC 62.5 mcg, VI 25 mcg, and placebo in patients with chronic obstructive pulmonary disease. The primary endpoint was the change from baseline in trough FEV1 on Day 169 (Week 24).
| Feature | DB2113373 |
|---|---|
| Trial name | DB2113373 |
| ClinicalTrials.gov | NCT01313650 |
| Brief title | A 24-week Evaluation of GSK573719/Vilanterol (62.5/25mcg) and Components in COPD |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Respiratory |
| Condition | Chronic obstructive pulmonary disease |
| Enrollment | 1538 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Lead sponsor | GlaxoSmithKline |
2. Clinical Question
The primary statistical question was whether the active treatment groups produced a greater improvement in lung function, measured as change from baseline in trough FEV1 on Day 169 (Week 24), than the relevant comparator group. The registry reports superiority hypotheses for the posted analyses.
Population
Participants enrolled in the phase 3 randomized study of chronic obstructive pulmonary disease.
Combination intervention
UMEC/VI 62.5/25 µg once daily.
Component interventions
UMEC 62.5 µg once daily and VI 25 µg once daily.
Comparator
Placebo.
The posted comparisons allow two different questions to be separated. First, each active regimen can be compared with placebo to quantify the treatment effect relative to no active drug. Second, the combination can be compared directly with each individual component, addressing whether the combination produced additional improvement beyond either component alone.
3. Trial Design
UMEC/VI 62.5/25 µg QD
- Umeclidinium 62.5 µg
- Vilanterol 25 µg
- Once daily
UMEC 62.5 µg QD
- Umeclidinium 62.5 µg
- Once daily
VI 25 µg QD
- Vilanterol 25 µg
- Once daily
Placebo
- Placebo drug
- Once-daily study regimen
4. The Primary Endpoint
The registered primary endpoint was Change From Baseline (BL) in Trough Forced Expiratory Volume in One Second (FEV1) on Day 169 (Week 24), with a time frame of baseline and Day 169.
| Endpoint element | Registry definition |
|---|---|
| Measure | Change from baseline in trough FEV1 on Day 169 (Week 24) |
| Unit | Liters |
| Time frame | Baseline and Day 169 |
| Baseline | Mean of assessments made 30 minutes pre-dose and 5 minutes pre-dose on Treatment Day 1 |
| Trough FEV1 | Mean of FEV1 values obtained at 23 and 24 hours after the previous morning's dosing |
| Day 169 trough | Mean of FEV1 values obtained 23 and 24 hours after the morning dosing on Day 168 |
| Change from baseline | Trough FEV1 at the particular visit minus baseline |
| Repeated assessments | Days 2, 28, 56, 84, 112, 168, and 169 |
FEV1 is a measure of lung function and is defined in the registry as the maximal amount of air that can be forcefully exhaled in one second. The use of trough measurements focuses the endpoint on lung function near the end of the once-daily dosing interval rather than immediately after administration.
5. How the Endpoint Was Measured
The endpoint has an important longitudinal structure. Although the primary reported comparison is the change from baseline at Day 169, trough FEV1 was measured repeatedly throughout the treatment period. That repeated-measures structure is directly relevant to the statistical model.
| Visit | Role in FEV1 assessment |
|---|---|
| Baseline | Mean of the two Treatment Day 1 pre-dose assessments |
| Day 2 | Trough FEV1 assessment |
| Day 28 | Trough FEV1 assessment |
| Day 56 | Trough FEV1 assessment |
| Day 84 | Trough FEV1 assessment |
| Day 112 | Trough FEV1 assessment |
| Day 168 | Trough FEV1 assessment |
| Day 169 | Primary endpoint assessment |
This distinction matters statistically because a repeated-measures model can use the pattern of available longitudinal measurements rather than treating Day 169 as an isolated observation. The registry identifies the posted method as a mixed-model analysis and specifically describes a restricted maximum likelihood (REML)-based repeated measures approach.
6. Statistical Methodology
Mixed-effects repeated-measures model
The primary analyses used a mixed-model repeated-measures approach, abbreviated MMRM in the registry's analysis notes. The model was based on restricted maximum likelihood (REML) and incorporated the longitudinal FEV1 assessments collected during the trial.
For the combination-versus-placebo comparison, the reported least-squares mean difference is the UMEC/VI 62.5/25 µg result minus the placebo result.
Covariate adjustment
The repeated-measures model included treatment, baseline, smoking status, center group, day, and the day-by-baseline and day-by-treatment interactions. These terms allow the analysis to account for important baseline and design variables while modeling how FEV1 changes over the repeated assessment days.
| Model component | Role in the analysis |
|---|---|
| Treatment | Defines the randomized treatment-group comparison. |
| Baseline | Accounts for baseline FEV1 when estimating change over time. |
| Smoking status | Provides covariate adjustment for smoking status. |
| Center group | Accounts for center-group differences. |
| Day | Models the longitudinal assessment time. |
| Day × baseline | Allows the relationship between baseline FEV1 and outcome to vary with day. |
| Day × treatment | Allows treatment differences to vary across assessment days. |
Least-squares mean difference
The effect measure reported by the registry is the least-squares mean difference. This is an adjusted comparison generated from the fitted model rather than a simple arithmetic difference between the raw observed means at Day 169.
Intent-to-treat population
The analysis population was the Intent-to-Treat (ITT) population, defined in the registry as all participants randomized to treatment who received at least one dose of randomized study drug. Participants represented in a particular time-point presentation are those with data available at that time point; however, all participants in the ITT population without missing covariate information and with at least one post-baseline measurement were included in the analysis.
Superiority framework
The posted comparisons were classified as superiority analyses. For these analyses, a positive least-squares mean difference indicates greater adjusted improvement in the first-named active-treatment direction specified in the analysis, and the reported confidence interval and p-value quantify the statistical uncertainty and evidence against the corresponding null hypothesis.
7. Primary Results
The registry contains five posted statistical analyses for the registered primary endpoint. Each analysis uses the mixed-model approach and reports a least-squares mean difference with a two-sided 95% confidence interval.
UMEC 62.5 µg QD vs Placebo
Adjusted difference in change from baseline in trough FEV1
95% CI: 0.076–0.155 · P < 0.001
Least-squares mean difference = UMEC 62.5 µg minus placebo.
| Comparison | Estimate | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|
| UMEC 62.5 µg QD vs Placebo | 0.115 L | 0.076–0.155 | <0.001 | Superiority |
The estimated adjusted difference in change from baseline was 0.115 L in favor of UMEC 62.5 µg QD relative to placebo. In practical terms, the model estimates that the adjusted improvement in trough FEV1 was 0.115 liters greater with UMEC than with placebo at the Day 169 assessment framework.
The 95% confidence interval of 0.076 to 0.155 L describes statistical uncertainty around the estimated mean difference under the fitted model. It does not mean that individual patients' treatment effects must fall within that interval.
The p-value of <0.001 addresses the evidence against the null hypothesis of no difference under the prespecified superiority framework. It does not measure the magnitude or clinical importance of the effect. Effect size is described by the estimate itself, while precision is described by the confidence interval.
Because the estimate comes from a repeated-measures model, it should be interpreted as a model-based adjusted comparison rather than as a simple unadjusted difference between two observed Day 169 means.
VI 25 µg QD vs Placebo
Adjusted difference in change from baseline in trough FEV1
95% CI: 0.032–0.112 · P < 0.001
Least-squares mean difference = VI 25 µg minus placebo.
| Comparison | Estimate | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|
| VI 25 µg QD vs Placebo | 0.072 L | 0.032–0.112 | <0.001 | Superiority |
The estimated adjusted difference was 0.072 L, with VI 25 µg QD producing a greater model-estimated improvement in trough FEV1 than placebo.
The 95% CI of 0.032 to 0.112 L indicates the precision of this estimated mean difference. The interval remains above zero, which is consistent with the reported superiority p-value of <0.001.
The p-value should not be interpreted as saying that there is a probability of less than 0.001 that the null hypothesis is true. It instead quantifies how incompatible the observed statistical result is with the specified null model, under the assumptions of the analysis.
As with the other analyses, the estimate is a least-squares mean difference from a longitudinal mixed model, not a direct statement that every participant experienced a 0.072-liter improvement.
UMEC/VI 62.5/25 µg QD vs Placebo
Adjusted difference in change from baseline in trough FEV1
95% CI: 0.128–0.207 · P < 0.001
Least-squares mean difference = UMEC/VI 62.5/25 µg minus placebo.
| Comparison | Estimate | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|
| UMEC/VI 62.5/25 µg QD vs Placebo | 0.167 L | 0.128–0.207 | <0.001 | Superiority |
The combination produced an estimated adjusted difference of 0.167 L versus placebo in change from baseline in trough FEV1. This is the largest of the three active-versus-placebo estimates reported for the primary endpoint.
The 95% CI of 0.128 to 0.207 L quantifies uncertainty around the model-based estimate. The entire interval is above zero, consistent with the reported P < 0.001 superiority result.
The estimate should not be translated into an individual-level prediction. A least-squares mean difference describes the average adjusted contrast represented by the statistical model. It does not imply that all participants experienced the same change or that the effect is clinically meaningful for every individual.
The repeated-measures structure is particularly relevant here because the trial collected trough FEV1 repeatedly. The combination estimate therefore incorporates the longitudinal modeling framework rather than relying only on an isolated Day 169 comparison.
UMEC 62.5 µg QD vs UMEC/VI 62.5/25 µg QD
Adjusted difference in change from baseline in trough FEV1
95% CI: 0.017–0.087 · P = 0.004
Least-squares mean difference = UMEC/VI 62.5/25 µg minus UMEC 62.5 µg.
| Comparison | Estimate | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|
| UMEC 62.5 µg QD vs UMEC/VI 62.5/25 µg QD | 0.052 L | 0.017–0.087 | 0.004 | Superiority |
The registry defines this contrast as the combination minus UMEC 62.5 µg, yielding an estimated difference of 0.052 L. Thus, the model estimates an additional improvement in trough FEV1 for the combination relative to the UMEC component alone.
The 95% CI of 0.017 to 0.087 L indicates uncertainty around that incremental estimate. The reported P = 0.004 provides evidence against the corresponding null hypothesis under the posted superiority analysis.
This comparison is conceptually different from the combination-versus-placebo analysis. The placebo comparison asks whether the combination differs from no active treatment, whereas the component comparison asks whether adding the second component is associated with additional improvement beyond UMEC alone.
The confidence interval also shows why effect size and statistical significance should be considered separately. The p-value establishes the reported statistical evidence; the 0.052-L estimate and its confidence interval describe the magnitude and precision of the incremental comparison.
VI 25 µg QD vs UMEC/VI 62.5/25 µg QD
Adjusted difference in change from baseline in trough FEV1
95% CI: 0.060–0.130 · P < 0.001
Least-squares mean difference = UMEC/VI 62.5/25 µg minus VI 25 µg.
| Comparison | Estimate | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|
| VI 25 µg QD vs UMEC/VI 62.5/25 µg QD | 0.095 L | 0.060–0.130 | <0.001 | Superiority |
The combination was estimated to improve trough FEV1 by 0.095 L more than VI 25 µg QD under the posted mixed-model analysis.
The 95% CI of 0.060 to 0.130 L gives the model-based precision of that incremental comparison, while the reported P < 0.001 indicates strong statistical evidence against the null hypothesis used for the superiority analysis.
Again, the p-value does not quantify the size of the treatment effect. Nor does the confidence interval represent the range of responses among individual patients. Both are properties of the estimated group-level treatment contrast.
8. Summary of the Posted Primary Analyses
| Comparison | Least-squares mean difference | 95% CI | P-value | Direction of contrast |
|---|---|---|---|---|
| UMEC 62.5 µg QD vs Placebo | 0.115 L | 0.076–0.155 | <0.001 | UMEC − Placebo |
| VI 25 µg QD vs Placebo | 0.072 L | 0.032–0.112 | <0.001 | VI − Placebo |
| UMEC/VI 62.5/25 µg QD vs Placebo | 0.167 L | 0.128–0.207 | <0.001 | UMEC/VI − Placebo |
| UMEC 62.5 µg QD vs UMEC/VI 62.5/25 µg QD | 0.052 L | 0.017–0.087 | 0.004 | UMEC/VI − UMEC |
| VI 25 µg QD vs UMEC/VI 62.5/25 µg QD | 0.095 L | 0.060–0.130 | <0.001 | UMEC/VI − VI |
The five posted analyses form a coherent set of comparisons. The active regimens were each compared with placebo, while the combination was also compared directly with each component. Across these analyses, the estimated least-squares mean differences were positive in the direction specified by the registry, and all five reported confidence intervals excluded zero.
The bar display is scaled to the largest reported estimate and is intended only to visualize relative magnitudes among the five posted estimates. The numerical values and confidence intervals in the table remain the appropriate statistical summaries.
9. Statistical Methods Explained
Why was a mixed-effects model used?
The primary endpoint was measured repeatedly over time: trough FEV1 was assessed on Days 2, 28, 56, 84, 112, 168, and 169. A repeated-measures mixed model is designed for this type of longitudinal structure. Rather than analyzing each visit as an unrelated endpoint, the model represents the trajectory of the outcome over the observed treatment period and estimates treatment differences within that framework.
What is a least-squares mean difference?
A least-squares mean is a model-adjusted mean. It represents the expected mean outcome for a treatment group after accounting for the covariates and model terms included in the analysis. The least-squares mean difference is the contrast between these adjusted means. In this trial, the registry reports the contrast in the change from baseline in trough FEV1.
Why was baseline FEV1 included in the model?
Baseline lung function is directly related to subsequent FEV1 measurements. Including baseline as a covariate can account for baseline differences and improve the precision of the treatment comparison. The model also included a day-by-baseline interaction, allowing the relationship between baseline and the repeated outcome to vary across assessment days.
Why include a day-by-treatment interaction?
The treatment effect does not necessarily have to be constant at every visit. A day-by-treatment interaction allows the estimated treatment difference to vary over the longitudinal assessment period. This is particularly relevant when an outcome is measured repeatedly after treatment initiation.
What does the 0.167 L estimate mean?
For the UMEC/VI 62.5/25 µg versus placebo comparison, the reported least-squares mean difference was 0.167 L. This means that the fitted repeated-measures model estimated a 0.167-liter greater change from baseline in trough FEV1 for the combination relative to placebo, using the model's covariate adjustment and longitudinal structure.
Why does the p-value not measure clinical importance?
A p-value addresses statistical evidence against a null hypothesis. It does not measure how large an effect is, how important that effect is to a patient, or how much an individual patient will improve. Those questions require consideration of the effect estimate, its confidence interval, the clinical context, and patient-level variability.
What does the confidence interval add?
The confidence interval places the estimated treatment difference in a range that reflects statistical uncertainty under the specified model and sampling framework. For example, the 95% CI of 0.128–0.207 L for UMEC/VI versus placebo communicates substantially more information about precision than the p-value <0.001 alone.
10. Intention-to-Treat Analysis
The registry identifies the analysis population as the ITT population. Participants were considered according to their randomized treatment assignment, provided they were randomized and received at least one dose of randomized study drug.
The principal statistical advantage of randomization is preserved when treatment comparisons remain anchored to the randomized groups. The registry's definition additionally requires at least one dose for the ITT population used in these posted analyses.
The registry also states that participants represented at a particular time point are those with data available at that time point, while participants in the ITT population without missing covariate information and with at least one post-baseline measurement were included in the analysis. This distinction separates the analysis population from the number of observations displayed at an individual visit.
11. Repeated Measures and Missing Data
Because FEV1 was collected repeatedly, the statistical analysis could use information from multiple visits rather than relying solely on the Day 169 observation. This is one of the central advantages of a repeated-measures framework in a longitudinal clinical trial.
The registry describes the analysis population and the observations contributing to the analysis, but it does not report a separate missing-data imputation method in the information presented here. The statistical interpretation therefore centers on the posted mixed-model analysis rather than assigning an unreported imputation procedure to the trial.
12. Confidence Intervals and Precision
All five posted primary analyses report two-sided 95% confidence intervals. The width of these intervals provides a useful way to compare precision across the treatment contrasts.
| Comparison | Estimate | 95% CI width | Interpretive point |
|---|---|---|---|
| UMEC vs placebo | 0.115 L | 0.079 L | Confidence interval remains above zero. |
| VI vs placebo | 0.072 L | 0.080 L | Confidence interval remains above zero. |
| UMEC/VI vs placebo | 0.167 L | 0.079 L | Confidence interval remains above zero. |
| UMEC/VI vs UMEC | 0.052 L | 0.070 L | Confidence interval remains above zero. |
| UMEC/VI vs VI | 0.095 L | 0.070 L | Confidence interval remains above zero. |
The similarity of the reported interval widths is consistent with the large randomized trial and common endpoint framework, although interval width alone should not be used to compare clinical importance. Precision and magnitude are different statistical concepts.
13. Comparing the Combination With Its Components
The direct component comparisons are particularly informative statistically because they address a different question from the placebo comparisons.
Combination vs UMEC
The reported adjusted difference was 0.052 L, with a 95% CI of 0.017–0.087 and P = 0.004.
Combination vs VI
The reported adjusted difference was 0.095 L, with a 95% CI of 0.060–0.130 and P < 0.001.
These contrasts should not be interpreted as arithmetic decompositions of the placebo comparison. The statistical model estimates each treatment contrast directly. A difference between two active groups is its own estimand, with its own uncertainty, even when the groups are part of the same randomized trial.
14. Multiplicity and Multiple Comparisons
Five primary-endpoint statistical analyses are posted for the same registered endpoint. The analyses compare placebo with each active regimen and compare the combination with each component.
| Comparison family | Purpose | Posted result |
|---|---|---|
| UMEC vs placebo | Assess active-component superiority over placebo | P < 0.001 |
| VI vs placebo | Assess active-component superiority over placebo | P < 0.001 |
| UMEC/VI vs placebo | Assess combination superiority over placebo | P < 0.001 |
| UMEC/VI vs UMEC | Assess incremental combination effect over UMEC | P = 0.004 |
| UMEC/VI vs VI | Assess incremental combination effect over VI | P < 0.001 |
Multiplicity is important whenever several statistical hypotheses are evaluated. If all hypotheses are treated as one family of confirmatory tests, the probability of at least one false-positive conclusion can exceed the nominal type I error unless an appropriate multiplicity strategy is specified. The registry information presented here identifies the individual posted analyses but does not provide a multiplicity-adjustment procedure for these five comparisons. The individual p-values should therefore be interpreted as the reported results of their respective analyses rather than automatically treating them as a jointly multiplicity-adjusted family.
15. Safety Results
The registry reports serious adverse events by randomized arm. These counts provide an additional descriptive view of the treatment groups, separate from the efficacy analysis.
| Treatment arm | Serious adverse events | At risk | Percentage of arm |
|---|---|---|---|
| Placebo | 9 | 280 | 3.2% |
| UMEC 62.5 µg QD | 27 | 418 | 6.5% |
| VI 25 µg QD | 24 | 421 | 5.7% |
| UMEC/VI 62.5/25 µg QD | 21 | 413 | 5.1% |
The percentages in the final column are simple descriptive proportions calculated from the reported affected and at-risk counts. They should not be confused with a formal comparative safety analysis. The registry information presented here does not provide a confidence interval, adjusted risk ratio, odds ratio, hazard ratio, or hypothesis test for these serious-adverse-event counts.
16. What the Primary Estimate Does — and Does Not — Mean
The combination-versus-placebo estimate of 0.167 L is an adjusted mean difference in change from baseline in trough FEV1. It summarizes the treatment contrast estimated by the specified repeated-measures model.
It does not mean that every participant improved by exactly 0.167 L more than a placebo-treated participant. It is a group-level model-based estimate, not an individual treatment prediction.
The 95% CI of 0.128–0.207 L describes the precision of the estimated treatment contrast under the statistical model and sampling framework. It is not a range containing 95% of individual patient responses.
The reported P < 0.001 indicates strong statistical evidence against the null hypothesis specified for the superiority comparison. It does not quantify effect size, probability of benefit for an individual patient, or clinical importance.
Because the endpoint was analyzed using a repeated-measures mixed model with covariate adjustment, the estimate depends on the model specification, including the treatment, baseline, smoking status, center group, day, and interaction terms.
17. Why This Trial Matters Statistically
DB2113373 is a useful teaching example because the primary endpoint is continuous, repeatedly measured, and analyzed with a longitudinal mixed-effects model rather than a simple two-group comparison at a single time point.
| Statistical concept | How it appears in DB2113373 |
|---|---|
| Randomization | Participants were randomized among four parallel treatment arms. |
| Blinding | The registry describes the trial as quadruple-masked. |
| Continuous endpoint | Change from baseline in trough FEV1 was analyzed in liters. |
| Repeated measures | Trough FEV1 was measured repeatedly from Day 2 through Day 169. |
| Mixed-effects model | The primary analyses used a REML-based repeated-measures model. |
| Covariate adjustment | Treatment, baseline, smoking status, center group, day, and specified interactions were included. |
| Least-squares means | Treatment effects were reported as least-squares mean differences. |
| Confidence intervals | All five posted primary analyses included two-sided 95% confidence intervals. |
| P-values | Each posted superiority analysis included a reported p-value. |
| ITT analysis | The posted analyses used the registry-defined ITT population. |
18. Important Limitations and Interpretation Issues
- Primary endpoint scope: the principal efficacy result represented here is change from baseline in trough FEV1 on Day 169 (Week 24). Other clinical outcomes should not be inferred from this endpoint alone.
- Longitudinal modeling: the estimates depend on the specified mixed-effects repeated-measures model and its covariate structure.
- Missing observations: the registry describes which ITT participants contributed to the analysis but does not provide a separate imputation procedure in the information presented here.
- Multiplicity: five primary-endpoint statistical analyses were posted. The registry information presented here does not identify a multiplicity-adjustment procedure for these comparisons.
- Clinical meaning: statistical evidence that a mean difference is nonzero does not by itself establish how meaningful that difference is to an individual patient.
- Safety interpretation: serious-adverse-event counts are descriptive in the information presented here; formal comparative safety inference would require additional statistical summaries.
- Generalizability: interpretation outside the randomized study population requires consideration of the trial's eligibility criteria and study setting, which are not characterized in detail in the information presented here.
19. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The posted mixed-model analyses estimated positive least-squares mean differences for all five treatment contrasts, with two-sided 95% confidence intervals excluding zero and reported superiority p-values of <0.001 or 0.004.
Clinical interpretation
The primary endpoint measures lung function through trough FEV1. The statistical results describe average adjusted treatment differences in this measure; determining their clinical importance requires context beyond statistical significance alone.
20. Understanding the Five Comparisons
The five posted analyses can be viewed as answering two related but distinct sets of questions.
UMEC, VI, and UMEC/VI were each compared with placebo using the same primary endpoint and mixed-model framework.
The combination was directly compared with UMEC and directly compared with VI.
This structure makes the direct component comparisons statistically important. A combination treatment can have a large difference from placebo, but that alone does not determine whether the combination provides additional improvement beyond an individual component. The direct comparisons address that incremental question.
21. Study Timeline
Study start
The registry lists March 1, 2011 as the study start date.
Repeated trough FEV1 assessments
Trough FEV1 was assessed on Days 2, 28, 56, 84, 112, 168, and 169, with Day 169 serving as the registered primary endpoint time point.
Primary completion
The registry lists April 1, 2012 as the primary completion date.
Posted statistical results
The ClinicalTrials.gov record reports posted results and five statistical analyses for the registered primary endpoint.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: NCT01313650 — DB2113373.
- Linked publication: PubMed record for PMID 23830094.
Continue through Clinical Biostats
Connect this trial's endpoint and statistical methods with deeper biostatistics tutorials and statistical tools.
25. Record Summary
DB2113373 provides a clear example of how a randomized parallel-group trial with a continuous, repeatedly measured endpoint can be analyzed using a mixed-effects repeated-measures model. The primary endpoint was change from baseline in trough FEV1 on Day 169, and the five posted statistical analyses used least-squares mean differences, two-sided 95% confidence intervals, and superiority p-values.
The combination-versus-placebo comparison produced an estimated difference of 0.167 L (95% CI 0.128–0.207; P < 0.001). The component-versus-placebo comparisons were 0.115 L for UMEC and 0.072 L for VI, while the direct combination comparisons were 0.052 L versus UMEC and 0.095 L versus VI. These results illustrate why a complete interpretation should distinguish the estimated effect, its precision, the statistical test, and the question represented by each treatment contrast.