This page separates information reported in the trial registry from statistical interpretation. The ClinicalTrials.gov record reports the primary endpoint definition and states that results were posted, but it does not provide a formal statistical analysis estimate, confidence interval, or p-value for that endpoint in the ClinicalTrials.gov record.
1. Trial at a Glance
ILLUMINATE was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating QVA149 versus fluticasone/salmeterol in patients with chronic obstructive pulmonary disease. The registry lists 523 participants and one registered primary endpoint: FEV1 AUC 0-12 at week 26.
| Feature | ILLUMINATE |
|---|---|
| Trial name | ILLUMINATE |
| Brief title | QVA149 Versus Fluticasone/Salmeterol in Patients With Chronic Obstructive Pulmonary Disease (COPD) |
| Phase | Phase 3 |
| Condition | Chronic Obstructive Pulmonary Disease (COPD) |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 523 |
| Study status | Completed |
| Start | March 2011 |
| Primary completion | March 2012 |
| Lead sponsor | Novartis Pharmaceuticals |
2. Clinical Question
The central question was whether indacaterol and glycopyrronium (QVA149) should be compared with fluticasone/salmeterol with respect to the registered lung-function endpoint, FEV1 AUC 0-12, measured at week 26.
Population
Patients with chronic obstructive pulmonary disease (COPD) enrolled in the phase 3 ILLUMINATE trial.
Intervention
Indacaterol and glycopyrronium (QVA149), with placebo to fluticasone/salmeterol.
Comparator
Fluticasone/salmeterol, with placebo to indacaterol and glycopyrronium (QVA149).
Primary question
How does the randomized comparison perform on FEV1 AUC 0-12 at week 26?
The ClinicalTrials.gov record does not provide a formal treatment-effect estimate for the primary endpoint. Consequently, the statistical interpretation here focuses on what the endpoint measures, how the stated linear mixed model relates to the endpoint, and what cannot be concluded from the available numerical record.
3. Trial Design
Indacaterol and glycopyrronium
- Indacaterol and glycopyrronium (QVA149)
- Placebo to fluticasone/salmeterol
- Randomized parallel-group assignment
Fluticasone/salmeterol
- Fluticasone/salmeterol
- Placebo to indacaterol and glycopyrronium (QVA149)
- Randomized parallel-group assignment
4. Randomization and Masking
Randomization is important because the comparison is intended to represent the effect of assignment to QVA149 versus assignment to fluticasone/salmeterol rather than a comparison of patients selected into the two groups by investigators or participants.
The trial was quadruple-masked. The ClinicalTrials.gov record does not identify the individual roles included in the masking designation, so this page does not assign specific masking responsibilities beyond the registry's stated classification.
Randomization primarily protects against systematic confounding at treatment assignment. Masking addresses a different problem: it can reduce the potential for knowledge of treatment assignment to affect behavior, assessment, reporting, or other aspects of trial conduct. These are complementary design protections rather than interchangeable concepts.
5. Primary Endpoint
| Endpoint | Time frame | Registry definition | Registered analysis information |
|---|---|---|---|
| Forced Expiratory Volume in 1 Second Area Under the Curve (FEV1 AUC) 0-12 | Week 26 | Forced Expiratory Volume in one second (FEV1) was calculated as the volume of air forcibly exhaled in one second as measured by a spirometer. FEV1 was normalized by 12 hours (divided by time). This outcome measures absolute values at week 26. | Results are obtained from linear mixed model. |
What FEV1 AUC 0-12 represents
FEV1 is a pulmonary function measure based on the volume of air forcibly exhaled in one second. The registered endpoint extends this concept over a specified interval by using the area under the FEV1 curve from 0 to 12 hours and normalizing it by 12 hours.
This distinction matters statistically. A single FEV1 measurement captures lung function at one point in time. An area-under-the-curve endpoint summarizes measurements across a time interval. Normalizing the area by 12 hours puts the result on an average-FEV1 scale over that interval rather than leaving the quantity expressed as an unnormalized area.
The registry explicitly states that FEV1 AUC 0-12 is normalized by 12 hours. The formula above is a conceptual representation of that definition, not a reconstruction of individual patient calculations.
6. Results
The ClinicalTrials.gov record states that results were posted and that 11 outcome measures were posted. However, the ClinicalTrials.gov record contains no formal statistical analyses for the registered primary endpoint.
| Result component | What the ClinicalTrials.gov record reports |
|---|---|
| Results posted | Yes |
| Outcome measures posted | 11 |
| Formal statistical analyses reported | None |
| Primary endpoint | FEV1 AUC 0-12 at week 26 |
| Registered analysis description | Results are obtained from linear mixed model |
| Treatment-effect estimate reported | Not reported in the provided statistical-analysis data |
| Confidence interval reported | Not reported in the provided statistical-analysis data |
| P-value reported | Not reported in the provided statistical-analysis data |
The absence of a registry-reported effect estimate is not evidence that the treatment comparison had no effect. It means only that the numerical statistical comparison needed to quantify that effect is not present in the ClinicalTrials.gov record.
For this endpoint, the clinically and statistically informative quantities would ordinarily include the estimated between-group difference in FEV1 AUC 0-12, its uncertainty interval, and an appropriate hypothesis-test result if a formal confirmatory comparison was specified.
A p-value, if available, would address evidence against a prespecified null hypothesis under the specified analysis. It would not itself measure the magnitude or clinical importance of the treatment effect. Similarly, a confidence interval would describe statistical uncertainty around the estimated difference; it would not describe the range of responses expected for every individual patient.
7. Planned Analysis
Because the ClinicalTrials.gov record contains no formal comparison for the primary endpoint, the appropriate educational approach is to describe the analysis indicated by the registry rather than infer an outcome.
Endpoint structure
FEV1 AUC 0-12 is a continuous lung-function measure derived over a 12-hour interval and evaluated at week 26.
Registered model
The registry states that results are obtained from a linear mixed model.
Typical purpose of the model
A linear mixed model can account for repeated or correlated measurements and estimate treatment effects while representing patient-level variability through random effects.
What is not specified here
The ClinicalTrials.gov record does not identify the fixed-effect structure, covariance structure, baseline adjustment, visit effects, interaction terms, or missing-data assumptions.
A typical longitudinal linear mixed-model analysis would express the observed FEV1 measurements as a function of treatment and relevant time-related terms while allowing observations from the same participant to be correlated. The exact model, however, must come from the trial's prespecified statistical analysis plan or reported analysis rather than being inferred from the phrase "linear mixed model."
Here, Yij represents a measurement for participant i at observation j; Xβ represents fixed effects; bi represents participant-specific random effects; and ε represents residual variation. This is a general educational representation, not the reconstructed ILLUMINATE model.
8. Statistical Methodology
Linear mixed models
A linear mixed model is useful when an outcome is continuous and observations may be correlated because multiple measurements are obtained from the same participant. Ordinary linear regression assumes independent observations; repeated measurements from one patient generally do not satisfy that assumption.
Mixed models address this by combining fixed effects, which describe population-level relationships such as treatment or time, with random effects, which can represent participant-specific departures from the population pattern.
Why the model fits the endpoint structure
The registered FEV1 AUC 0-12 endpoint summarizes lung-function measurements over a defined interval and is evaluated at week 26. The registry specifically states that results are obtained from a linear mixed model. That makes the model part of the documented statistical methodology even though the ClinicalTrials.gov record does not provide its full specification.
The important statistical distinction is between what is documented and what is merely typical. It is documented that a linear mixed model was used. It is not documented in the ClinicalTrials.gov record which covariance structure, fixed effects, random effects, baseline terms, or missing-data strategy were selected.
Continuous treatment-effect estimation
For a continuous endpoint such as FEV1 AUC 0-12, the treatment comparison is naturally expressed on the scale of the endpoint. A treatment effect could therefore be represented by a difference in adjusted mean FEV1 AUC 0-12 between randomized groups, depending on the prespecified model.
A positive or negative value would have to be interpreted according to the exact direction and scale of the registered endpoint. The ClinicalTrials.gov record does not provide the observed value of this contrast.
Randomization as the basis for comparison
Randomization establishes the primary design basis for attributing between-group differences to treatment assignment, subject to the usual assumptions of a properly conducted randomized trial. Statistical modeling can improve precision or account for repeated measurements, but it does not replace the causal role of randomization.
Masking and measurement
FEV1 is measured by spirometer, and the trial is classified as quadruple-masked. Objective measurement does not eliminate all sources of variability, but masking can reduce the possibility that knowledge of assignment affects trial procedures or assessment.
9. Statistical Methods Explained
Why was a linear mixed model used?
The registry explicitly identifies a linear mixed model as the method used to obtain results for FEV1 AUC 0-12. Mixed models are commonly used when continuous outcomes arise from repeated measurements because they can represent correlation among measurements from the same participant. The ClinicalTrials.gov record does not provide the exact model specification, so the rationale here is methodological rather than a reconstruction of the investigators' full statistical analysis plan.
What does AUC 0-12 mean?
AUC stands for area under the curve. Here, the curve is the measured FEV1 over the 0-to-12-hour interval. The registry says the AUC is normalized by 12 hours, meaning the area is divided by the duration of the interval. Conceptually, this makes the result an average FEV1 over the specified period rather than an unnormalized area quantity.
Why use an area-under-the-curve endpoint instead of one FEV1 measurement?
A single measurement can be sensitive to the exact time at which it is taken. An area-under-the-curve summary incorporates information across the specified interval. Statistically, this can provide a measure of overall lung-function performance during that interval rather than focusing on one observation.
What would a treatment difference mean?
If the formal analysis reported an estimated difference in FEV1 AUC 0-12, that difference would describe the estimated separation between randomized treatment groups on the endpoint's measurement scale. It would not automatically translate into a proportional change in symptoms, exacerbations, hospitalization, or other clinical outcomes unless those relationships had been separately established.
What would a confidence interval add?
A confidence interval would quantify statistical uncertainty around the estimated treatment contrast under the specified model and sampling framework. A narrow interval would indicate greater statistical precision than a wide interval, all else equal. The interval would not describe the range of individual patient responses.
Why is a p-value not enough?
A p-value addresses compatibility of the observed data with a null hypothesis under the specified statistical model. It does not communicate the magnitude of the treatment effect by itself. A very small p-value can accompany a small effect in a sufficiently large sample, while an important-looking estimate can remain uncertain when the data are limited. Effect estimates and uncertainty intervals therefore provide information that a p-value alone cannot supply.
Why can the exact mixed-model specification matter?
Different choices about fixed effects, covariance structure, baseline adjustment, time terms, and handling of incomplete measurements can affect both estimation and precision. Because the ClinicalTrials.gov record does not identify those choices, a complete reproduction of the ILLUMINATE primary analysis cannot be made from the registry summary alone.
10. Safety Results
The ClinicalTrials.gov record provides serious adverse-event counts by treatment arm. These are reported as affected participants divided by participants at risk for the safety measure reported in the ClinicalTrials.gov record.
| Safety measure | QVA149 | Fluticasone/Salmeterol |
|---|---|---|
| Serious adverse events | 13 / 258 | 14 / 264 |
The denominators for the registry-reported serious-adverse-event measure are 258 and 264, respectively, and therefore should not be silently replaced with the overall enrollment of 523. The ClinicalTrials.gov record does not provide a confidence interval, hypothesis test, exposure-adjusted incidence rate, or formal comparative safety analysis for these serious adverse events.
The raw counts show the number of affected participants and the corresponding at-risk denominators reported in the ClinicalTrials.gov record. They do not, by themselves, establish that the treatments have identical or different safety profiles.
A formal comparison would depend on the prespecified safety analysis, the event definitions, exposure time, treatment-emergent rules, and potentially recurrent-event considerations. None of those additional analytical details are reported here.
11. Understanding the Primary Endpoint Statistically
Measurement scale
FEV1 AUC 0-12 is a continuous outcome derived from spirometric FEV1 measurements over a 12-hour interval.
Time frame
The registered primary endpoint is assessed at week 26.
Model
The registry states that the outcome results are obtained from a linear mixed model.
Interpretation
The relevant treatment comparison would be a model-based contrast on the FEV1 AUC 0-12 scale, with uncertainty quantified by an appropriate confidence interval.
The statistical meaning of an FEV1 AUC endpoint is different from the meaning of a time-to-event endpoint. There is no hazard ratio to interpret here, no censoring framework inherent in the endpoint definition reported, and no reason to substitute a survival-analysis framework simply because the study is a randomized clinical trial.
Likewise, the endpoint should not be transformed into a binary "responder" measure unless the trial prespecified such a definition. The ClinicalTrials.gov record defines the primary endpoint as an absolute continuous FEV1 AUC 0-12 measurement.
12. Analysis Population and Missing Data
The ClinicalTrials.gov record identifies the enrollment, treatment arms, primary endpoint, and registered linear mixed-model approach, but they do not specify the formal analysis population used for the primary endpoint or the missing-data strategy.
| Methodological topic | Information supported by the ClinicalTrials.gov record |
|---|---|
| Randomization | Yes — allocation is randomized. |
| Analysis population | Not specified in the ClinicalTrials.gov record. |
| Missing-data method | Not specified in the ClinicalTrials.gov record. |
| Imputation method | Not specified in the ClinicalTrials.gov record. |
| Covariance structure | Not specified in the ClinicalTrials.gov record. |
| Baseline adjustment | Not specified in the ClinicalTrials.gov record. |
| Formal multiplicity procedure | Not specified in the ClinicalTrials.gov record. |
| Interim analysis | Not specified in the ClinicalTrials.gov record. |
| Bayesian methods | Not specified in the ClinicalTrials.gov record. |
| Stratification factors | Not specified in the ClinicalTrials.gov record. |
This distinction is important for reproducibility. A statistical analysis page should not fill missing protocol information with assumptions merely because a particular approach is common. For ILLUMINATE, the ClinicalTrials.gov record supports the conclusion that a linear mixed model was used for the registered primary endpoint, but it does not support reconstruction of all modeling decisions.
13. Multiplicity and Confirmatory Inference
The ClinicalTrials.gov record identifies one registered primary endpoint and 11 posted outcome measures. They do not provide a multiplicity-control strategy for the collection of outcomes.
Having one registered primary endpoint simplifies one aspect of the statistical hierarchy: the principal endpoint is clearly identified. It does not, however, automatically establish how secondary outcomes were tested or whether their analyses were confirmatory, supportive, or exploratory.
A formal p-value for the primary endpoint would need to be interpreted in the context of the prespecified testing framework. Because the ClinicalTrials.gov record contains no multiplicity procedure and no formal statistical-analysis result, this page does not assign confirmatory significance to any unreported comparison.
14. Crossover, Factorial, and Bayesian Design Considerations
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Crossover | Not specified; the design model is parallel. |
| Factorial design | Not specified; the registry identifies a parallel two-arm design. |
| Non-inferiority margin | Not specified. |
| Interim analysis | Not specified. |
| Bayesian analysis | Not specified. |
These omissions are not methodological defects in the trial record; they are boundaries on what can responsibly be inferred from the ClinicalTrials.gov recordset. In particular, a parallel randomized trial should not be described as a crossover or factorial trial simply because those designs are common in other therapeutic studies.
15. Limitations
- No formal primary-effect estimate in the ClinicalTrials.gov record: although results are reported as posted, the registry-reported statistical-analysis section contains no numerical treatment comparison, confidence interval, or p-value for FEV1 AUC 0-12.
- Incomplete model specification: the registry states that a linear mixed model was used but does not, in the ClinicalTrials.gov record, specify its fixed effects, random effects, covariance structure, or other modeling details.
- Unspecified missing-data strategy: the ClinicalTrials.gov record does not identify how incomplete FEV1 measurements were handled.
- Unspecified analysis population: the ClinicalTrials.gov record does not identify whether the primary analysis used all randomized participants, a treated population, or another prespecified population.
- Limited safety detail: serious adverse-event counts are reported, but no formal comparative safety analysis, confidence interval, p-value, or exposure-adjusted rate is provided.
- Outcome-level interpretation: FEV1 AUC 0-12 is a lung-function endpoint. Its statistical result should not automatically be interpreted as an effect on every clinical outcome in COPD.
- Registry-level evidence: the ClinicalTrials.gov record is sufficient to describe the trial design and registered endpoint, but not to reconstruct the complete statistical analysis plan.
16. Why This Trial Matters Statistically
ILLUMINATE is a useful teaching case because its primary endpoint illustrates a different branch of clinical-trial statistics from the survival-analysis examples commonly encountered in oncology. Instead of asking when an event occurs, the primary analysis concerns a continuous physiological measurement summarized over a defined interval.
| Concept | How it appears in ILLUMINATE |
|---|---|
| Randomization | Participants were randomized to two parallel treatment groups. |
| Masking | The trial was quadruple-masked. |
| Continuous endpoint | FEV1 AUC 0-12 is a continuous lung-function measure. |
| Area under the curve | FEV1 is summarized over the 0-to-12-hour interval and normalized by 12 hours. |
| Longitudinal modeling | The registry states that results were obtained from a linear mixed model. |
| Between-group contrast | The statistical question concerns the difference between randomized treatment groups on the endpoint scale. |
| Uncertainty | A formal confidence interval would quantify uncertainty around the treatment estimate, but none is reported here. |
| Missing-data considerations | The ClinicalTrials.gov record does not specify the missing-data or imputation strategy. |
| Safety analysis | Serious adverse events are reported as affected participants over at-risk participants by arm. |
| Statistical reproducibility | The registry's identification of a linear mixed model does not by itself provide enough information to recreate the full model. |
The most important statistical lesson is that the analysis method must match the endpoint structure. A continuous repeated-measures lung-function endpoint calls for a different framework from a binary response endpoint or a time-to-event endpoint. Understanding that distinction is central to interpreting clinical-trial evidence correctly.
17. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial was randomized and quadruple-masked, and the registered primary endpoint was FEV1 AUC 0-12 at week 26. The registry states that a linear mixed model was used, but the ClinicalTrials.gov record does not provide the numerical treatment contrast or its uncertainty.
Clinical interpretation
FEV1 AUC 0-12 is intended to characterize lung function over a defined 12-hour period rather than a single instantaneous measurement. The ClinicalTrials.gov record does not support a numerical conclusion about the magnitude of the treatment difference.
This separation is important. A statistically complete interpretation requires an effect estimate, uncertainty, and an understanding of the prespecified analysis. A clinically complete interpretation additionally requires consideration of what the measured endpoint means for patients. Neither layer should be filled in by inference when the underlying numerical result is absent.
18. How to Read the Missing Primary Result
The primary endpoint is explicitly defined as FEV1 AUC 0-12 at week 26, and the registry states that the results are obtained from a linear mixed model.
The provided statistical-analysis data do not include the estimated treatment difference, confidence interval, standard error, p-value, or complete fitted-model specification.
It would be inappropriate to infer the direction, magnitude, statistical significance, or clinical importance of the QVA149 treatment effect solely from the fact that the trial posted results or from the fact that a linear mixed model was used.
This is an important principle for clinical-trial interpretation: methodological information is not a substitute for the result itself. Knowing that an appropriate model was used tells us how the investigators analyzed the endpoint; it does not tell us what estimate that model produced.
19. Trial Timeline
Trial start
The ILLUMINATE phase 3 trial began in March 2011 according to the ClinicalTrials.gov record.
Primary completion
The registry lists March 2012 as the primary completion date.
Study status
the ClinicalTrials.gov record identifies the study status as completed.
20. Serious Adverse Events: Statistical Perspective
The serious-adverse-event data illustrate why denominators and analysis populations matter. The registry-reported counts are 13/258 for QVA149 and 14/264 for fluticasone/salmeterol.
| Arm | Affected | At risk | What the ClinicalTrials.gov record permits |
|---|---|---|---|
| QVA149 | 13 | 258 | Description of the reported affected/at-risk count |
| Fluticasone/Salmeterol | 14 | 264 | Description of the reported affected/at-risk count |
Even when two raw counts appear numerically similar, formal statistical interpretation requires more information. For adverse events, investigators may need to consider follow-up duration, treatment exposure, event definitions, whether multiple events per patient are possible, and the prespecified safety population. The ClinicalTrials.gov record does not provide those details, so the appropriate interpretation remains descriptive.
21. What a Complete Statistical Result Would Contain
For educational purposes, a fully reported primary analysis of a continuous endpoint such as FEV1 AUC 0-12 would ordinarily allow a reader to identify at least four elements:
- The treatment-group estimates on the endpoint scale.
- The treatment contrast, such as a model-based difference between groups.
- An uncertainty measure, typically a confidence interval or standard error.
- The inferential result, such as a p-value or other prespecified decision criterion.
The registry-reported ILLUMINATE data provide the endpoint definition and the model class, but not these numerical inferential components. This page therefore treats the primary endpoint as a planned-analysis framework rather than a reconstructed efficacy result.
A complete report would pair this contrast with an appropriate uncertainty interval and inferential result. No numerical value is reported in the ClinicalTrials.gov record provided for ILLUMINATE.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: ILLUMINATE, NCT01315249.
- Linked publication: PubMed record for PMID 24321804.
Continue through the Clinical Biostats statistical library
Explore clinical-trial methods, statistical tutorials, and calculators that build from the concepts used in randomized clinical research.
25. Record Summary
ILLUMINATE provides a useful example of a randomized phase 3 clinical trial in which the primary endpoint is a continuous physiological measure rather than a time-to-event outcome. The trial randomized 523 participants in a parallel, quadruple-masked design to compare indacaterol and glycopyrronium (QVA149) with fluticasone/salmeterol in patients with COPD. Its registered primary endpoint was FEV1 AUC 0-12 at week 26, with the registry stating that results were obtained from a linear mixed model.
The ClinicalTrials.gov record does not include the numerical primary treatment comparison, confidence interval, or p-value. That limitation is central to the interpretation of this record. The appropriate statistical conclusion is therefore about the design and analysis framework, not an inferred efficacy result. The record supports discussion of randomization, quadruple masking, continuous endpoint analysis, area-under-the-curve construction, and linear mixed modeling, while leaving the exact treatment effect and its uncertainty unreported in the ClinicalTrials.gov record.