Dose Finding Study of Nemiralisib (GSK2269557) in Subjects With an Acute Moderate or Severe Exacerbation of Chronic Obstructive Pulmonary Disease (COPD): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Dose Finding Study of Nemiralisib (GSK2269557) in Subjects With an Acute Moderate or Severe Exacerbation of Chronic Obstructive Pulmonary Disease (COPD) (NCT03345407), a randomized phase 2 study. The primary outcome is Change From Baseline in Clinic Visit Trough Forced Expiratory Volume in One Second (FEV1) at Day 84 Measured Post Bronchodilator.

2. Clinical Question

The trial addressed dose response, efficacy, safety, and pharmacokinetics in people being treated for an acute COPD exacerbation. Its primary outcome focuses on lung function at the end of the stated treatment period. The registry record also lists outcomes relating to subsequent exacerbations, recovery, lung function, and COPD-related symptoms and resource use. The posted primary results concern the dose-specific adjusted changes and adjusted comparisons with placebo; they do not by themselves answer every question about the course of an exacerbation or later health outcomes.

3. Trial Design

This was a Phase IIb randomized, parallel-group, placebo-controlled dose-finding study sponsored by GlaxoSmithKline. The registry describes double masking with the sponsor open. Nemiralisib was given by inhalation once daily, in addition to standard-of-care therapy, and compared with placebo inhalation plus standard care. Albuterol was available to all participants as rescue medication. The brief summary describes a Screening Period, a 12-Week Treatment Period, and a 12-Week Post-Treatment Follow-Up Period.

The registry record lists enrollment of 943. Its brief summary says approximately 1,250 participants would be randomized, so the record contains both a planned figure and a separate enrollment figure. The record does not explain the difference. The status is terminated. The listed start date is 2017-11-28, and the primary completion date is 2019-01-10; the reason for termination is not reported in the registry record.

The posted primary outcome results include placebo and six nemiralisib dose groups, including a 12.5 mcg group. The intervention list and brief summary do not list that dose, while the posted primary results do. The registry record does not explain this difference in what is listed. The primary endpoint was assessed at Day 84, and the primary analysis description specifies a Bayesian repeated-measures model with adjustment for baseline and prespecified participant and visit characteristics.

Arm sizes differ across outcomes, and the primary endpoint analysis included only participants with data at the specified time point within the stated MITT population. The registry record does not report the reasons for these differences in analyzed numbers. Randomization and a placebo comparator support a structured comparison, but interpretation still depends on the outcome definitions, analysis population, missing data, and model. Details of the protocol’s sample-size justification and the reasons the study ended are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: A total of 943 participants were randomized, and 938 participants who received at least one dose of study treatment were included in the modified intent to treat (MITT) Population. The study included participants enrolled from 16 countries.

Only the first 6 of 7 groups are shown.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: MITT Population consisted of all randomized participants who received at least 1 dose of study treatment.. Only those participants with data available at the specified data points were analyzed.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Only the first 6 of 8 groups are shown.

6. Primary Endpoint

The registry lists the primary outcome as Change From Baseline in Clinic Visit Trough Forced Expiratory Volume in One Second (FEV1) at Day 84 Measured Post Bronchodilator, measured over Baseline and Day 84.

The primary endpoint is change from baseline in clinic-visit trough FEV1 at Day 84, measured after bronchodilator administration. FEV1 is the maximum amount of air forcefully exhaled from the lungs in one second. The Day 84 measurement was taken before dosing and after bronchodilator use; baseline was the latest post-bronchodilator FEV1 measured before the first study-treatment dose. The change is the Day 84 value minus baseline.

FEV1 is an objective measure of airflow and is used in COPD studies to describe lung function. Measuring trough FEV1 before the day’s dose characterizes lung function at that point in the dosing interval, while the post-bronchodilator measurement follows administration of albuterol. These details define what the endpoint captures; they do not make it a direct measure of symptoms, exacerbation recovery, or all aspects of health.

The registry describes the primary summary as a posterior adjusted median change from baseline, with a 95% highest posterior density credible interval. It also reports adjusted median differences between each nemiralisib dose and placebo. The adjustment accounts for baseline-by-visit and treatment-by-visit interactions, smoking status at baseline, region, severity of the index exacerbation, the number of moderate or severe exacerbations in the previous 12 months, and gender. Because this is an adjusted model-based endpoint, its estimates should not be confused with a simple comparison of unadjusted arm averages.

The endpoint is useful for comparing lung-function measurements across randomized groups at the specified visit. It does not alone establish whether participants felt better, recovered sooner, or experienced fewer later exacerbations. Those questions require their own outcome measures and interpretation.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is reported using a Bayesian repeated-measures model. In general, a repeated-measures model uses observations taken across visits rather than treating each measurement as unrelated. It can represent how outcomes change over time and how those patterns may differ between treatment groups. The registry describes adjustment for baseline-by-visit and treatment-by-visit interactions, smoking status at baseline, region, severity of the index exacerbation, the number of moderate or severe exacerbations in the previous 12 months, and gender. It does not report further model details, such as prior distributions or the approach to model checking.

What does an adjusted median represent?

A median describes the middle of a distribution rather than its arithmetic average. An adjusted median is a model-based estimate that accounts for the listed variables and model structure. It is not necessarily the raw median calculated directly from the observed values in an arm. The reported dose-versus-placebo differences are also posterior adjusted median differences, so they summarize the model’s comparison rather than a simple subtraction of the posted arm summaries.

How should the interval be read?

A credible interval describes a range of parameter values supported by the Bayesian model and its assumptions, given the data. The registry calls the primary intervals 95% highest posterior density credible intervals. In general, a wider interval signals less precision than a narrower interval. When an interval for a treatment contrast includes zero, the data and model do not sharply distinguish the contrast from zero at that interval level. That is not the same as demonstrating no difference, and it does not show that two treatments are equivalent. No primary p-values are reported.

Why use repeated measurements and covariate adjustment?

Repeated measurements can use information about the outcome across visits, while covariate adjustment can account for measured characteristics included in the model. Baseline-by-visit and treatment-by-visit interactions allow the relationship with baseline and the treatment contrast to vary by visit. The listed adjustments identify what the registry says was included; they do not establish that every possible source of variation or bias was addressed. In general, randomization helps make groups comparable, while adjustment can improve the precision or fit of a model. The registry record does not state the rationale for each covariate or provide a full statistical analysis plan.

How do the secondary recalculations differ?

For selected binary secondary outcomes, the posted recalculations use event counts and analyzed group sizes to form unadjusted risk differences and associated intervals and p-values. These are simple comparisons based on the posted summaries, not the sponsor’s adjusted model analyses. A risk difference expresses the difference in observed proportions between groups; it does not account for time to event. In contrast, the registry reports posterior median hazard ratios for selected time-to-event outcomes. A hazard ratio summarizes relative event rates over follow-up under its modeling assumptions, and is not interchangeable with a risk difference at a fixed time.

For a risk difference, an interval spanning zero indicates that the unadjusted estimate is imprecise with respect to the direction of the difference. The posted p-values for these recalculations are unadjusted. They do not account for testing multiple outcomes, doses, or comparisons. No multiplicity adjustment or prespecified hierarchy for interpreting these secondary analyses is reported in the registry record. The recalculations therefore offer descriptive context rather than replacing the sponsor’s reported analysis.

What do the planning illustrations mean?

The sample-size illustrations use the observed effect, or a stated fraction of it, with a two-sided significance level of 0.05, equal allocation, and a normal approximation. They are planning illustrations based on the recalculated comparisons, not the trial’s own design justification and not observed power. In general, planning calculations show how a hypothetical study’s size may depend on the effect it aims to detect and assumptions about the outcome. They do not demonstrate that the observed study had adequate power, and the record does not report the trial’s own sample-size justification.

For the time-to-event planning calculations, the stated output is total events; for the other calculations, it is participants per arm. These quantities answer different planning questions and should not be conflated. The registry record does not report interim analyses, and no such procedure should be inferred from the posted results.

10. Analyses Reported by the Sponsor

The sponsor’s primary analysis is a Bayesian repeated-measures analysis of adjusted median FEV1 change, with posterior adjusted median differences and 95% credible intervals. The recalculated primary contrasts are not a separate independent analysis: the posted primary contrast estimates are the sponsor-reported model results. The registry marks no discrepancy for the primary outcome, so there is no stated conflict between the sponsor’s analysis and the values used in the interpretation.

The recalculated secondary comparisons are different. They use posted event counts and group sizes to produce simple unadjusted risk differences for selected binary outcomes. The sponsor’s registry analyses for selected time-to-event outcomes are posterior median hazard ratios with 95% intervals. These approaches target different quantities and use different modeling assumptions. A fixed-period risk difference cannot be read as a substitute for a time-to-event hazard ratio, and a recalculated unadjusted comparison does not supersede the sponsor’s reported analysis.

For the sponsor-reported time-to-event estimates, the registry gives a hazard ratio of 0.455 with a 95% interval from 0.054 to 1.103 for placebo compared with nemiralisib 12.5 mcg, and 0.991 with a 95% interval from 0.580 to 1.500 for placebo compared with nemiralisib 50 mcg, for time to next moderate or severe exacerbation. For time to recovery, the reported values are 1.053 with an interval from 0.477 to 1.765 and 1.200 with an interval from 0.840 to 1.597, respectively. These intervals are broad and include 1. The registry does not report p-values for these analyses.

11. Multiplicity, Missing Data and Analysis Populations

The registry record lists many outcomes and dose comparisons, but it does not report a multiplicity adjustment, a prespecified hierarchy of hypotheses, or a strategy for controlling the overall chance of false-positive findings. This is important because interpretation becomes more difficult when numerous comparisons are considered. An unadjusted p-value describes an individual comparison under its assumptions; it does not account for the broader set of tests. The secondary recalculation p-values should therefore be treated as exploratory rather than as definitive evidence.

The primary outcome uses a Bayesian repeated-measures model and reports posterior estimates and credible intervals. The registry does not report whether the primary dose comparisons were governed by a formal decision rule, how dose response was assessed, or whether interim analyses occurred. Those details are not reported in the registry record and should not be inferred from the presence of multiple dose groups or the study’s terminated status.

The primary analysis population is described as randomized participants who received at least one dose, with only participants having data available at the specified data points included in the analysis. This is the registry’s stated MITT definition. The available record does not describe the amount or pattern of missing primary endpoint data, reasons for missingness, or sensitivity analyses assessing the effect of missing data. A repeated-measures model can use observed measurements under its assumptions, but the registry record does not provide enough detail to assess those assumptions or the robustness of the results to alternative missing-data approaches.

Group sizes vary across endpoints, which reflects different analyzed populations or available data for the reported outcomes; the record does not explain the reasons for those differences. The record also does not report a sample-size justification for the trial, and the posted planning illustrations are not the trial’s design calculations. Consequently, the precision and confirmatory status of the findings cannot be fully assessed from the registry record alone.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: chroni obstructive pulmonary disease (6.6%), dehydration (4.5%), renal impairment (4.5%), pneumonia (3.3%), angina unstable (2.2%).

The registry reports serious adverse events, other adverse events, and deaths by treatment group. For serious adverse events, the reported counts were 23 of 276 for placebo, 2 of 22 for nemiralisib 12.5 mcg, 9 of 91 for 50 mcg, 13 of 92 for 100 mcg, 16 of 90 for 250 mcg, 6 of 89 for 500 mcg, and 26 of 278 for 750 mcg. The registry also reports other adverse events and deaths by group; those counts should be read with their stated group denominators rather than compared as raw counts alone.

The listed serious-event terms include chronic obstructive pulmonary disease, dehydration, renal impairment, pneumonia, unstable angina, and myocardial ischaemia, with percentages reported in the record. These are descriptions of reported events, not evidence that study treatment caused them. The registry figures do not provide a formal comparative safety analysis, uncertainty intervals for the event counts, or enough detail to judge differences in exposure time. Serious events and deaths are important descriptive safety information, but their interpretation depends on the trial context, event definitions, follow-up, and analysis methods.

Counts across treatment groups should not be treated as directly comparable without considering the differing group sizes and the fact that adverse-event denominators differ from those in the primary endpoint analysis. The reason for termination and its relationship, if any, to safety is not reported in the registry record.

13. What the Result Does, and Does Not, Mean

The primary results describe adjusted changes in post-bronchodilator FEV1 at Day 84 in the analyzed participants. The dose-specific adjusted median differences from placebo are close to zero in the reported model estimates, and every reported 95% credible interval for a dose contrast includes zero. This conveys uncertainty about the size and direction of each contrast. The estimates do not establish that the groups are equivalent, nor do they prove that a treatment effect is absent.

The outcome is one measure of lung function at a defined visit. It does not directly establish whether participants recovered faster, experienced fewer future exacerbations, had fewer symptoms, or used fewer health-care resources. Those outcomes are represented by separate endpoints and require their own analysis. The secondary results also do not remove the uncertainty around the primary endpoint, particularly because some comparisons are unadjusted and the record does not describe multiplicity control.

The trial was terminated, but the registry record does not report the reason. Termination alone does not indicate whether the decision was based on efficacy, safety, recruitment, operational issues, or another cause. Nor does it determine whether the posted results are biased or valid. It is a limitation in the available context that should remain separate from interpretation of the numerical estimates.

In general, a randomized placebo-controlled design supports comparisons between assigned groups, but conclusions remain dependent on adherence, outcome availability, the analysis population, missing-data assumptions, and the statistical model. The registry describes some of these features but not all. The reported results therefore support a careful account of the observed and modeled outcomes, not medical advice or a broad conclusion about every potential benefit or harm.

14. Why This Trial Matters Statistically

This trial is statistically informative because it compares multiple nemiralisib dose groups with placebo using a continuous lung-function endpoint and a Bayesian repeated-measures model. The design makes dose-specific contrasts possible, while the adjusted model incorporates baseline, visit, treatment, and participant characteristics listed in the registry. The range of secondary outcomes also illustrates why endpoint definitions and analysis methods matter: a lung-function change, an event count, and time to an event describe different features of the clinical course.

The posted primary estimates highlight the importance of interpreting effect estimates together with uncertainty. Point estimates alone can look precise, but their intervals show the range of values compatible with the model and data. Here, the dose contrasts have intervals that include zero, and no primary p-values are reported. A neutral-looking estimate is not proof of no effect, while an interval spanning zero is not evidence of equivalence.

The registry also shows why analysis populations and missing data deserve attention. The primary results use participants with data at the specified time points within the stated MITT population, and the recorded group sizes differ by endpoint. Without details about missingness, sensitivity analyses, multiplicity, and the reason for termination, readers cannot fully assess the robustness or confirmatory status of every finding.

Finally, the secondary recalculations and sponsor-reported analyses illustrate that different statistical summaries answer different questions. Unadjusted risk differences are not interchangeable with adjusted posterior estimates or hazard ratios. Keeping the estimand and method explicit prevents overreading exploratory comparisons and makes the limits of the registry record clearer.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The registry record is incomplete for several aspects needed to assess the trial’s design and results. It does not report the reason for termination, the sample-size justification, interim analyses, or a full statistical analysis plan. The brief summary’s planned figure of approximately 1,250 participants differs from the enrollment entry of 943, and the record does not explain that difference. It also does not explain why a 12.5 mcg group appears among the posted primary results while the intervention list and brief summary do not list that dose.

The primary analysis includes participants with data available at the specified time points, but the record does not give the extent or reasons for missing outcome data. It does not describe sensitivity analyses for missingness or provide enough information to assess model assumptions, prior choices, or model diagnostics. The posted results include several dose groups, but the record does not specify a multiplicity strategy or hierarchy for interpreting those comparisons.

Secondary results are not uniformly available across the listed outcomes. Some posted summaries use participant counts, while selected sponsor analyses report hazard ratios; the recalculated binary comparisons are unadjusted and cover only selected arms. These estimates are not direct substitutes for one another. The record does not provide comparative safety uncertainty estimates or explain the relation, if any, between adverse events and termination.

These limitations do not invalidate the reported estimates, but they restrict how comprehensively they can be evaluated. Interpretation should remain tied to the endpoint, analysis population, and method actually described in the registry record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Dose Finding Study of Nemiralisib (GSK2269557) in Subjects With an Acute Moderate or Severe Exacerbation of Chronic Obstructive Pulmonary Disease (COPD) — NCT03345407.
• Registry record last updated 2021-07-14; status last verified 2021-07.
• Results first posted 2020-02-10.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

NCT03345407 is a randomized, Phase IIb, double-blind, parallel-group, placebo-controlled dose-finding study of inhaled nemiralisib added to standard care for acute moderate or severe COPD exacerbation. The registry lists enrollment of 943 and status as terminated. The primary endpoint was change from baseline in clinic-visit trough FEV1 at Day 84, measured post-bronchodilator, analyzed using a Bayesian repeated-measures model in the stated MITT population.

The posted adjusted median differences from placebo vary across doses, and all reported 95% credible intervals for those contrasts include zero. Secondary outcomes and adverse-event counts add descriptive context, but the record does not report the reason for termination, a multiplicity plan, the sample-size justification, or detailed missing-data assessments. These omissions limit how strongly the findings can be interpreted beyond the specific estimates reported.