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
ILA115938 was a dose-finding study. Its purpose was to characterize how lung function responds as the dose of UMEC added to a fixed inhaled corticosteroid dose (FF 100 mcg) increases, using a within-participant crossover design so that each participant served as their own control across three treatment periods.
| Feature | ILA115938 |
|---|---|
| Phase | Phase 2 |
| Condition | Asthma |
| Design | Randomized, double-blind, crossover |
| Treatments | FF 100 mcg; FF 100 mcg + UMEC 15.6, 31.25, 62.5, 125 or 250 mcg; FF 100 mcg + VI 25 mcg |
| Periods | Three treatment periods; trough assessment on Day 15 of each period |
| Enrollment | 421 |
| Primary endpoints | Model-predicted change from baseline in trough FEV1 (dose-response); percentage chance of exceeding target FEV1 responses; mean change from baseline in trough FEV1 on Day 15 |
| Primary analysis model | Mixed model with treatment, period, period baseline FEV1 and mean baseline FEV1 as fixed effects and participant as a random effect |
| Multiplicity control | Overall F-test gate, then step-down closed testing from the highest UMEC dose |
| ClinicalTrials.gov | NCT01573624 |
| Lead sponsor | GlaxoSmithKline (industry) |
2. Clinical Question
The study asked two linked questions. First, does adding UMEC to FF 100 mcg improve trough FEV1 compared with FF 100 mcg alone, and at which doses? Second, what is the shape of the dose-response relationship, so that a dose can be selected for later development? FF 100 mcg plus VI 25 mcg was included as an active benchmark: a combination of an inhaled corticosteroid with a long-acting bronchodilator, against which the size of the UMEC response could be calibrated.
Population
Participants with asthma, studied in three sequential treatment periods.
Intervention
FF 100 mcg combined with UMEC at 15.6, 31.25, 62.5, 125 or 250 mcg.
Comparators
FF 100 mcg alone (reference for the dose-response) and FF 100 mcg + VI 25 mcg (active benchmark).
Primary question
How much does each UMEC dose, added to FF, change trough FEV1 from period baseline relative to FF alone after 14 days of dosing?
3. Trial Design
Each participant received three treatments in sequence, but seven treatments were studied. The design is therefore an incomplete-block crossover: no participant received every treatment, and each pairwise comparison draws on the participants who happened to receive both treatments, together with indirect information linked through the other treatments. The safety counts confirm the structure: the numbers at risk across the seven treatments sum to 1,263, which is three treatment periods for each of the 421 participants.
| Treatment | Role | Treatment periods at risk |
|---|---|---|
| FF 100 mcg | Reference (inhaled corticosteroid alone) | 187 |
| FF 100 mcg + UMEC 15.6 mcg | Dose-ranging | 183 |
| FF 100 mcg + UMEC 31.25 mcg | Dose-ranging | 179 |
| FF 100 mcg + UMEC 62.5 mcg | Dose-ranging | 180 |
| FF 100 mcg + UMEC 125 mcg | Dose-ranging | 176 |
| FF 100 mcg + UMEC 250 mcg | Dose-ranging | 186 |
| FF 100 mcg + VI 25 mcg | Active benchmark | 172 |
4. Analysis Population and Covariates
All posted analyses used the ITT population, with the registry noting that only participants with data available at the indicated time points were analyzed. In a crossover mixed model, a participant who completes only one or two periods still contributes the periods they completed; the model does not require complete data per participant.
| Model term | Type | Purpose |
|---|---|---|
| Treatment | Fixed effect | Estimates the effect of each of the seven treatments |
| Period | Fixed effect | Adjusts for systematic drift across periods 1, 2 and 3 (seasonal change, learning effects, disease course) |
| Period baseline FEV1 | Fixed-effect covariate | Adjusts each period's change for that period's starting FEV1 |
| Mean baseline FEV1 | Fixed-effect covariate | Captures a participant's overall lung-function level across periods |
| Participant | Random effect | Models correlation among repeated measurements on the same participant |
5. Endpoints
All three primary endpoints are built on the same measurement. FEV1 is the maximal amount of air forcefully exhaled in one second; the highest of three acceptable efforts was recorded between 5:00 and 11:00 AM after withholding albuterol (salbutamol) for at least 4 hours. Baseline was the pre-dose FEV1 on Day 1 of each period, and trough was the FEV1 obtained 24 hours after the Day 14 morning dose (that is, on Day 15).
| Endpoint | Registry definition (shortened) | Time frame |
|---|---|---|
| Primary: Model-predicted change from baseline trough FEV1 | Slope-intercept on log dose model used to predict trough FEV1 change from baseline for each FF + UMEC dose, adjusted by FF 100 mcg alone; mean expected response with 95% CI | Baseline (Day 1) and Day 15 of each treatment period |
| Primary: Percentage chance that FF-corrected FEV1 response exceeds a target | Percentage chance that the FF 100 mcg–corrected change from baseline trough FEV1 would exceed 50 mL, 75 mL, 100 mL and 150 mL, by UMEC dose | Baseline (Day 1) and Day 15 of each treatment period |
| Primary: Mean change from baseline in trough FEV1 on Day 15 | Mixed model with treatment, period, period baseline FEV1 and mean baseline FEV1 as fixed effects and participant as random effect; post hoc analysis to confirm nullification of UMEC carryover | Baseline (Day 1) and Day 15 of each treatment period |
| Secondary: Mean change in daily morning PEF | Pre-dose and pre-rescue bronchodilator morning peak expiratory flow | Baseline (Week 0) and last 7 days of each treatment period |
| Secondary: Mean change in daily evening PEF | Pre-dose and pre-rescue bronchodilator evening peak expiratory flow | Baseline (Week 0) and last 7 days of each treatment period |
| Secondary: Mean change in rescue albuterol/salbutamol use | Number of puffs | Baseline (Week 0) and last 7 days of each treatment period |
The first two primary endpoints are model-based summaries of the dose-response curve rather than pairwise comparisons. The registry does not post hypothesis-test analyses for them; the formal treatment comparisons with confidence intervals and p-values are posted for the third primary endpoint, the mixed-model mean change in trough FEV1.
6. Primary Results: Trough FEV1, FF + UMEC versus FF Alone
Each estimate below is the adjusted mean difference in change from baseline trough FEV1, in litres, for the combination minus FF 100 mcg alone. Positive values indicate a larger improvement with the combination.
| Comparison vs FF 100 mcg | Mean difference (L) | 95% CI | P-value | Status in step-down testing |
|---|---|---|---|---|
| FF + UMEC 250 mcg | 0.055 | 0.010 to 0.101 | 0.018 | Tested first; significant |
| FF + UMEC 125 mcg | 0.055 | 0.010 to 0.100 | 0.018 | Tested second; significant |
| FF + UMEC 62.5 mcg | 0.026 | −0.020 to 0.071 | 0.271 | Tested third; not significant, testing stops |
| FF + UMEC 31.25 mcg | 0.033 | −0.013 to 0.079 | 0.165 | Not reached in the hierarchy |
| FF + UMEC 15.6 mcg | 0.026 | −0.019 to 0.071 | 0.264 | Not reached in the hierarchy |
| FF + VI 25 mcg | 0.078 | 0.032 to 0.124 | <0.001 | Benchmark comparison outside the UMEC hierarchy |
Highest UMEC doses versus FF alone
UMEC 250 mcg: 95% CI 0.010–0.101 · P = 0.018
UMEC 125 mcg: 95% CI 0.010–0.100 · P = 0.018
What the estimate means. After adjustment for period and baseline lung function, the change from baseline in trough FEV1 was estimated to be 0.055 L (55 mL) larger with FF + UMEC 125 mcg or 250 mcg than with FF alone. Because the design is within-participant, this is best read as the average difference in response when the same kind of participant receives one treatment rather than the other.
What it does not mean. It does not mean that every participant improved by 55 mL, nor that the 125 and 250 mcg doses are proven equal. Identical point estimates at the two top doses suggest a plateau, but the trial was not designed to test the difference between those doses directly, and none is posted.
Precision. The confidence intervals are wide relative to the estimate: the data are compatible with differences as small as 10 mL or as large as about 100 mL. The lower bound excludes zero, but only just, so the size of the benefit is uncertain even where its direction is supported.
P-values and effect size. Both comparisons have P = 0.018, and the lower three doses have p-values between 0.165 and 0.271. These p-values describe how compatible each estimate is with no difference; they do not rank doses by clinical value. The 31.25 mcg dose has a numerically larger estimate (0.033 L) than the 62.5 mcg dose (0.026 L) and a smaller p-value, which is sampling noise in a flat region of the curve, not evidence that the lower dose is better.
Cautions. Significance at the 125 and 250 mcg doses is confirmed under the step-down closed testing procedure. Because the 62.5 mcg comparison was not significant, the hierarchy stopped there; the 31.25 and 15.6 mcg comparisons are not formally tested, and their p-values are descriptive only. The registry does not report the result of the gating overall F-test for treatment. Estimates also rely on the crossover assumption of negligible carryover, addressed by a post hoc analysis described in the registry.
7. Primary Results: Trough FEV1, FF + UMEC versus FF + VI
The registry also posts comparisons of each UMEC dose with the FF + VI benchmark. Here each estimate is FF + UMEC minus FF + VI, so negative values indicate a smaller trough FEV1 response with the UMEC combination.
| Comparison vs FF + VI 25 mcg | Mean difference (L) | 95% CI | P-value |
|---|---|---|---|
| FF + UMEC 15.6 mcg | −0.052 | −0.097 to −0.007 | 0.023 |
| FF + UMEC 31.25 mcg | −0.045 | −0.091 to 0.000 | 0.051 |
| FF + UMEC 62.5 mcg | −0.053 | −0.098 to −0.007 | 0.023 |
| FF + UMEC 125 mcg | −0.023 | −0.068 to 0.021 | 0.306 |
| FF + UMEC 250 mcg | −0.023 | −0.068 to 0.023 | 0.330 |
What the estimates mean. Every UMEC dose had a numerically smaller trough FEV1 response than FF + VI. For the lower doses the shortfall was about 45 to 53 mL; for 125 and 250 mcg it was 23 mL.
What they do not mean. The 125 and 250 mcg comparisons are not statistically significant, but that does not show that these doses are equivalent to FF + VI. Their intervals (−0.068 to 0.021 and −0.068 to 0.023) include deficits of up to 68 mL. Demonstrating equivalence or non-inferiority would require a prespecified margin, and these comparisons were framed as superiority tests.
Precision. All five intervals are roughly 90 mL wide, consistent with the within-participant precision of the model and the incomplete-block allocation.
P-values and effect size. The 31.25 mcg comparison illustrates a boundary case: P = 0.051 with an upper confidence limit of 0.000. An estimate sitting at the edge of conventional significance is not meaningfully different from one at P = 0.023; the underlying effects (−0.045 versus −0.052 L) are very similar.
Cautions. The registry states that no multiplicity adjustments were made for comparisons between FF + UMEC doses and FF + VI. These five comparisons are therefore nominal and supportive, and should not be interpreted with the same error control as the step-down comparisons with FF alone.
8. Dose-Response and Target-Probability Endpoints
Two of the primary endpoints approach the same data from a modelling perspective rather than through pairwise tests.
Slope-intercept on log dose
A model in which the FF-corrected trough FEV1 response changes linearly with the logarithm of the UMEC dose. The fitted line yields a predicted mean response, with 95% CI, at each dose studied, borrowing strength across all doses rather than estimating each separately.
Probability of exceeding a target
For each UMEC dose, the percentage chance that the FF-corrected change in trough FEV1 exceeds 50, 75, 100 or 150 mL. This reframes the question from "is there an effect?" to "how likely is an effect of a size that matters?"
The doses studied (15.6, 31.25, 62.5, 125, 250 mcg) are each double the previous one, which makes them equally spaced on the log scale and well suited to a log-dose model. The pairwise results above are broadly consistent with a response that rises between the middle and upper doses and then flattens; a log-linear model summarizes that trend with two parameters, though it cannot itself represent a plateau.
Probability-of-exceeding-target summaries are typically derived from the uncertainty distribution of the modelled response, for example a Bayesian posterior or a sampling distribution from the fitted model. The registry description does not state which framework was used. Their value for dose selection is that a dose with a high probability of exceeding a clinically relevant threshold is more informative than a dose that is merely statistically different from zero.
9. Secondary Results: Peak Expiratory Flow
Daily morning and evening PEF were recorded by participants over the last 7 days of each treatment period. Differences are in litres per minute.
Morning PEF
| Comparison | Mean difference (L/min) | 95% CI | P-value |
|---|---|---|---|
| UMEC 15.6 vs FF alone | 15.9 | 9.5 to 22.3 | <0.001 |
| UMEC 31.25 vs FF alone | 17.4 | 10.9 to 23.9 | <0.001 |
| UMEC 62.5 vs FF alone | 18.6 | 12.1 to 25.1 | <0.001 |
| UMEC 125 vs FF alone | 22.9 | 16.5 to 29.4 | <0.001 |
| UMEC 250 vs FF alone | 22 | 15.5 to 28.4 | <0.001 |
| VI 25 vs FF alone | 27 | 20.6 to 33.5 | <0.001 |
| UMEC 15.6 vs FF + VI | −11.2 | −17.6 to −4.7 | <0.001 |
| UMEC 31.25 vs FF + VI | −9.6 | −16.1 to −3.1 | 0.004 |
| UMEC 62.5 vs FF + VI | −8.4 | −14.9 to −1.9 | 0.012 |
| UMEC 125 vs FF + VI | −4.1 | −10.6 to 2.3 | 0.209 |
| UMEC 250 vs FF + VI | −5.1 | −11.6 to 1.4 | 0.124 |
Evening PEF
| Comparison | Mean difference (L/min) | 95% CI | P-value |
|---|---|---|---|
| UMEC 15.6 vs FF alone | 16.2 | 9.5 to 22.9 | <0.001 |
| UMEC 31.25 vs FF alone | 17.2 | 10.5 to 24.0 | <0.001 |
| UMEC 62.5 vs FF alone | 21.2 | 14.4 to 28.0 | <0.001 |
| UMEC 125 vs FF alone | 28.8 | 22.1 to 35.5 | <0.001 |
| UMEC 250 vs FF alone | 22.9 | 16.2 to 29.6 | <0.001 |
| VI 25 vs FF alone | 26.6 | 19.8 to 33.4 | <0.001 |
| UMEC 15.6 vs FF + VI | −10.3 | −17.2 to −3.5 | 0.003 |
| UMEC 31.25 vs FF + VI | −9.3 | −16.2 to −2.5 | 0.007 |
| UMEC 62.5 vs FF + VI | −5.4 | −12.3 to 1.5 | 0.126 |
| UMEC 125 vs FF + VI | 2.2 | −4.6 to 9.1 | 0.523 |
| UMEC 250 vs FF + VI | −3.7 | −10.5 to 3.1 | 0.290 |
PEF shows a clearer dose gradient than trough FEV1. Every UMEC dose improved morning and evening PEF relative to FF alone with P < 0.001, so under the step-down procedure every dose passes, provided the gating F-test was significant. The benchmark comparisons repeat the FEV1 pattern: lower UMEC doses produced smaller PEF gains than FF + VI, while the 125 and 250 mcg doses were not statistically distinguishable from it. The evening PEF estimate for UMEC 125 mcg is the only UMEC-versus-VI point estimate that is positive (2.2 L/min), within an interval of −4.6 to 9.1.
Why would PEF separate doses more sharply than trough FEV1? Daily PEF is averaged over seven days of home measurements, which reduces noise compared with a single clinic spirometry value, and it captures morning and evening bronchodilation over the dosing interval rather than a single trough time point. Greater precision relative to the effect size produces smaller p-values even when the relative pattern across doses is similar.
10. Secondary Results: Rescue Medication Use
| Comparison | Mean difference (puffs) | 95% CI | P-value | Status in step-down testing |
|---|---|---|---|---|
| UMEC 250 vs FF alone | −0.3 | −0.5 to 0.0 | 0.023 | Significant |
| UMEC 125 vs FF alone | −0.3 | −0.5 to −0.1 | 0.012 | Significant |
| UMEC 62.5 vs FF alone | −0.1 | −0.3 to 0.1 | 0.335 | Not significant; testing stops |
| UMEC 31.25 vs FF alone | −0.2 | −0.5 to 0.0 | 0.064 | Not reached |
| UMEC 15.6 vs FF alone | −0.2 | −0.5 to 0.0 | 0.036 | Not reached |
| VI 25 vs FF alone | −0.4 | −0.7 to −0.2 | <0.001 | Benchmark |
| UMEC 15.6 vs FF + VI | 0.2 | 0.0 to 0.4 | 0.104 | Unadjusted |
Rescue-use differences were small, a few tenths of a puff. The rescue-use results show why a closed testing hierarchy matters: the 15.6 mcg comparison has a nominal P = 0.036, below 0.05, yet it is not a confirmed finding because the hierarchy stopped at 62.5 mcg (P = 0.335). Several intervals also show a bound of 0.0 at one decimal place alongside a p-value below 0.05; this reflects rounding of the displayed limits, not a contradiction.
11. Safety: Serious Adverse Events
| Treatment | Participants with SAE / at risk |
|---|---|
| FF 100 mcg | 0 / 187 |
| FF 100 mcg + UMEC 15.6 mcg | 0 / 183 |
| FF 100 mcg + UMEC 31.25 mcg | 1 / 179 |
| FF 100 mcg + UMEC 62.5 mcg | 0 / 180 |
| FF 100 mcg + UMEC 125 mcg | 0 / 176 |
| FF 100 mcg + UMEC 250 mcg | 0 / 186 |
| FF 100 mcg + VI 25 mcg | 1 / 172 |
Two serious adverse events were reported in total, one during FF + UMEC 31.25 mcg and one during FF + VI 25 mcg. With counts this low, no comparison between treatments is informative, and zero events in a group does not establish zero risk. A common approximation, the "rule of three", gives an upper 95% bound of roughly 3/n for the true event rate when no events are seen in n exposures. In a crossover trial, SAEs are also attributed to the treatment period in which they occurred, which assumes that events are not caused by carryover from an earlier period. Each treatment was given for only about two weeks, so these data say little about longer-term safety.
12. Statistical Methodology
Mixed-effects model for a crossover
The primary and secondary continuous endpoints were analyzed with a linear mixed model. Treatment and period were fixed effects, with two baseline covariates, and participant was a random effect.
Yij is the change from baseline for participant i in period j; τt(i,j) is the effect of the treatment received; πj is the period effect; Bij is the period baseline FEV1 and B̄i the participant's mean baseline across periods; ui ~ N(0, σ2u) is the participant random effect; εij ~ N(0, σ2) is within-participant error.
Including both the period baseline and the mean baseline is a deliberate choice. In a crossover, a period baseline mixes two sources of information: how a participant differs from other participants (between-participant) and how that period's starting point differs from the participant's own average (within-participant). Adding the mean baseline separates these, so that the period-baseline coefficient reflects within-participant adjustment only. This prevents between-participant baseline differences from leaking into treatment contrasts, which matters in an incomplete-block design where not every participant receives every treatment.
Gatekeeping and step-down closed testing
For the comparisons of FF + UMEC with FF alone, the registry describes a two-stage procedure:
- An overall F-test for treatment effect acted as a gate. No pairwise comparisons were made unless it was significant at the 5% level.
- Pairwise comparisons then followed a step-down closed testing procedure: UMEC 250 mcg versus FF was tested first, and each lower dose was tested only if the preceding comparison was significant at the 5% level.
Because each test is performed at the full 5% level only when all earlier tests in the sequence have succeeded, the familywise type I error across the dose comparisons is controlled at 5% without splitting alpha.
The ordering from the highest dose downward reflects the prior expectation of a monotone dose-response: if any dose works, the highest one is most likely to. The cost is that once a dose fails, lower doses cannot be declared effective even if their nominal p-values are small.
Comparisons with FF + VI
The registry states that no multiplicity adjustments were made for the comparisons between FF + UMEC doses and FF + VI. These are outside the controlled hierarchy.
Carryover
The registry notes that a post hoc analysis was performed to confirm nullification of the UMEC carryover effect. Carryover, where the effect of a treatment persists into the next period, would bias crossover treatment estimates. A post hoc check is reassuring but carries less weight than a design-based safeguard, and tests for carryover are typically low-powered.
Hypothesis framework
All posted comparisons were framed as superiority tests with two-sided 95% confidence intervals.
13. Statistical Methods Explained
Why is a mixed model used rather than a paired t-test?
A paired t-test compares two treatments in participants who received both. With seven treatments and three periods per participant, most pairs of treatments are shared by only a fraction of participants. A mixed model uses all observations at once, adjusts for period and baseline, handles participants with missing periods, and recovers information about each contrast both within and between participants through the random participant effect.
Why did the UMEC 15.6 mcg dose not count as effective for rescue use when P = 0.036?
In a fixed-sequence procedure, a test is only confirmatory if every earlier test in the sequence succeeded. The sequence ran from 250 mcg downwards, and the 62.5 mcg comparison failed (P = 0.335). The 31.25 and 15.6 mcg comparisons were therefore not formally tested. Their p-values are nominal, and treating them as confirmatory would reintroduce the type I error inflation the procedure was designed to prevent.
What does a difference of 0.055 L mean for trough FEV1?
It means the adjusted mean change from baseline in trough FEV1 was 55 mL greater with the combination than with FF alone. It is an absolute difference in means, not a percentage. Its practical importance depends on the clinical threshold chosen; the study's target-probability endpoint makes this explicit by asking how likely the FF-corrected response is to exceed 50, 75, 100 or 150 mL.
Why include both period baseline and mean baseline as covariates?
They separate within-participant variation from between-participant variation in baseline FEV1. Adjusting for period baseline alone would mix the two and could bias treatment comparisons in an incomplete-block crossover. With both terms, the model adjusts each period for how that day's starting point differs from the participant's usual lung function.
Does a non-significant comparison with FF + VI mean the high UMEC doses are as good as FF + VI?
No. For UMEC 125 and 250 mcg, the intervals include shortfalls of up to 68 mL relative to FF + VI. A superiority test that fails to reject the null cannot establish equivalence; that would require a prespecified margin and a confidence interval lying entirely within it.
Why model the dose-response on the log-dose scale?
Bronchodilator dose-response curves typically rise steeply at low doses and flatten at higher doses. The doses in this trial double at each step, so they are evenly spaced on the log scale, and a straight line in log dose is a simple, stable summary. It pools information across all five doses to predict the response at each, which is typically more efficient than estimating every dose independently.
14. Limitations and Interpretation Issues
- Short treatment periods: each period involved 14 days of dosing, which is appropriate for bronchodilator dose-finding but not for exacerbations, long-term control or long-term safety.
- Carryover: crossover estimates assume that each period's response is not influenced by the previous treatment. The registry reports a post hoc analysis addressing UMEC carryover, but post hoc checks are typically underpowered.
- Incomplete-block allocation: each participant received three of seven treatments, so each pairwise contrast is estimated with partial within-participant information and relies on the model structure.
- Precision: trough FEV1 intervals of roughly 90 mL leave the magnitude of the UMEC effect uncertain even where it is statistically significant.
- Unadjusted benchmark comparisons: FF + UMEC versus FF + VI comparisons carry no multiplicity adjustment and should be read as supportive.
- Gating result not reported: the registry does not report the overall F-test statistic that gated the pairwise comparisons with FF alone.
- Complete-case ITT: only participants with data at the relevant time points were analyzed; the mixed model's validity with missing periods rests on a missing-at-random assumption.
- Sparse safety data: with two SAEs in total, the trial cannot distinguish safety between doses.
15. Why This Trial Matters Statistically
ILA115938 is a compact example of how phase 2 dose-finding combines efficient designs, model-based estimation and structured multiplicity control.
| Concept | How it appears in ILA115938 |
|---|---|
| Crossover design | Three-period, within-participant comparisons across seven treatments |
| Incomplete block | Each participant receives three of seven treatments; 1,263 treatment periods in total |
| Mixed-effects model | Participant random effect with treatment, period and two baseline covariates |
| Baseline adjustment | Separation of period-level and participant-level baseline FEV1 |
| Gatekeeping | Overall F-test required before pairwise comparisons |
| Step-down closed testing | Highest-dose-first sequence; testing stops at 62.5 mcg for trough FEV1 |
| Nominal vs confirmatory p-values | Rescue use at 15.6 mcg: P = 0.036 but outside the confirmed hierarchy |
| Dose-response modelling | Slope-intercept on log-dose model for predicted FEV1 response |
| Probability of target effect | Chance of exceeding 50, 75, 100 and 150 mL responses |
| Active benchmark | FF + VI comparisons, explicitly unadjusted for multiplicity |
| Carryover assessment | Post hoc analysis of UMEC carryover |
16. Trial Timeline
Study start
Enrollment into the three-period crossover began.
Primary completion
Final data collection for the primary endpoints; 421 participants enrolled.
Results posted
Trough FEV1, PEF, rescue-use and safety results are posted on ClinicalTrials.gov.
17. Related Tutorials
Learn more about the methods used in this trial:
18. Related Calculators
19. Sources
- ClinicalTrials.gov: NCT01573624: Evaluate the Safety, Efficacy and Dose Response of GSK573719 in Combination With Fluticasone Furoate in Subjects With Asthma.
- PubMed: PMID 25452139.
Explore more trial statistics
Each trial page connects its endpoints and methods to tutorials and calculators so the statistical reasoning can be followed step by step.
20. Record Summary
ILA115938 enrolled 421 participants with asthma into a randomized, double-blind, three-period crossover that compared five UMEC doses added to FF 100 mcg with FF alone and with FF + VI 25 mcg. For trough FEV1, the 250 and 125 mcg doses each improved change from baseline by 0.055 L relative to FF alone (P = 0.018 for both), and the step-down hierarchy stopped at 62.5 mcg (P = 0.271). FF + VI improved trough FEV1 by 0.078 L relative to FF alone, and no UMEC dose exceeded it. Morning and evening PEF favoured every UMEC dose over FF alone. The most useful reading combines the within-participant mixed-model estimates, the structure of the closed testing hierarchy, the width of the confidence intervals, and the distinction between confirmatory and nominal p-values.