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
TRAIL1 was a randomized, parallel-group, triple-masked phase 2 treatment study comparing pirfenidone with placebo in patients with rheumatoid arthritis interstitial lung disease. The registry reports an enrollment of 123 participants and one registered primary endpoint: the number of participants who developed any element of a composite endpoint at 52 weeks.
| Feature | TRAIL1 |
|---|---|
| Trial acronym | TRAIL1 |
| ClinicalTrials.gov identifier | NCT02808871 |
| Phase | Phase 2 |
| Status | Completed |
| Condition | Rheumatoid Arthritis Interstitial Lung Disease |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 123 |
| Number of arms | 2 |
| Interventions | Pirfenidone; Placebo |
| Lead sponsor | Brigham and Women's Hospital |
| Sponsor type | Other |
2. Clinical Question
The central statistical question in TRAIL1 was whether participants assigned to pirfenidone differed from those assigned to placebo with respect to development of either component of the registered composite endpoint by 52 weeks.
Population
Patients with rheumatoid arthritis interstitial lung disease.
Intervention
Pirfenidone.
Comparator
Placebo.
Primary question
How many participants developed any element of the composite endpoint by 52 weeks in the randomized comparison?
3. Trial Design
Pirfenidone
- Study intervention: pirfenidone
- Randomized treatment arm
- Safety information reported for 62 participants at risk for serious adverse events
Placebo
- Study comparator: placebo
- Randomized control arm
- Safety information reported for 60 participants at risk for serious adverse events
The study was randomized, parallel-group, and triple-masked. These design features are important statistically because randomization creates the basis for comparing outcome distributions between treatment assignments, while masking can reduce the influence of treatment knowledge on trial conduct and assessment.
4. Trial Timing and Status
Study start
The registry lists 2017-04-07 as the study start date.
Primary completion
The registry lists 2021-04-07 as the primary completion date.
Registry status
The ClinicalTrials.gov record lists the study status as completed.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Number of Participants Who Developed Any Element of the Composite Endpoint | Number of participants who developed any element of the composite endpoint of decline in percent predicted FVC of 10% or greater or death. | 52 weeks |
This endpoint combines two clinically distinct events into a single binary outcome: a decline in percent predicted FVC of 10% or greater, or death. A participant needs to experience only one component to meet the composite endpoint.
The logical OR is important. The endpoint counts participants who experience either component; a participant does not need to experience both components.
6. Understanding the Composite Endpoint
Composite endpoints are often used when investigators want one outcome to capture more than one clinically relevant event. In TRAIL1, the two components reflect different aspects of disease progression: a decline in pulmonary function measured through percent predicted FVC and death.
FVC component
The first component is a decline in percent predicted FVC of 10% or greater. The threshold is part of the registered endpoint definition.
Death component
Death is the second component. It is automatically an event for the composite endpoint because the endpoint uses an OR rule.
One event is sufficient
A participant meeting either component qualifies as having developed an element of the composite endpoint.
Components are not identical
FVC decline and death differ in clinical meaning and frequency. That distinction matters when interpreting a combined endpoint.
7. Planned Analysis
The ClinicalTrials.gov record reports that results are posted for the primary endpoint, but no formal statistical analysis is posted in the ClinicalTrials.gov record. The registry reports the endpoint as a count of participants who developed any element of the composite endpoint at 52 weeks.
For an endpoint defined this way, the primary descriptive analysis would normally begin with the number and proportion of participants meeting the composite endpoint in each randomized treatment group. A treatment comparison could then be expressed using an absolute risk difference, a risk ratio, an odds ratio, or another prespecified binary-outcome measure, with an appropriate confidence interval.
The treatment comparison should be based on the prespecified analysis population and denominator. The statistical meaning of an effect estimate depends on exactly which participants are included in each arm.
Because the registered endpoint is defined at 52 weeks, a simple comparison of 52-week event proportions is conceptually different from a time-to-event analysis. If the statistical analysis instead used time to first occurrence of either component, censoring and event timing would become part of the analysis and methods such as Kaplan-Meier estimation and Cox regression could be appropriate. The endpoint definition alone does not establish which inferential method was used.
8. Statistical Methodology
Binary endpoint analysis
The registered primary endpoint is framed as the number of participants who developed any element of the composite endpoint by 52 weeks. For a fixed follow-up assessment, this naturally produces a binary outcome for each participant: composite endpoint occurred or did not occur.
The basic event proportion for an arm can be written as:
where E is the number of participants meeting the composite endpoint and N is the number of participants included in the relevant analysis denominator.
Absolute risk difference
An absolute risk difference compares the observed event proportions directly:
The absolute difference is expressed in probability units. A negative value would indicate fewer composite events in the pirfenidone group; a positive value would indicate more.
Risk ratio
A risk ratio compares the probability of the composite endpoint between randomized groups:
An RR of 1 represents equal event probabilities. Values below 1 indicate a lower observed event probability in the pirfenidone group, while values above 1 indicate a higher observed event probability.
Confidence intervals
A confidence interval accompanies an effect estimate to describe its statistical precision. For a risk difference, the interval is expressed in absolute probability units; for a risk ratio or odds ratio, it is expressed on a relative scale.
Time-to-event alternative
If the analysis uses the actual time until the first component of the composite endpoint occurs rather than only whether it occurred by 52 weeks, survival-analysis methods become relevant. Kaplan-Meier estimation can account for right censoring, while a Cox proportional-hazards model can estimate a relative hazard under its modeling assumptions.
9. Statistical Methods Explained
Why use a composite endpoint?
A composite endpoint can capture more than one clinically important outcome in a single analysis. In TRAIL1, the endpoint combines a defined decline in percent predicted FVC with death, so a participant can meet the endpoint through either pathway.
What does the 10% FVC threshold do statistically?
The threshold converts a continuous measure of lung function change into a predefined event criterion. Instead of analyzing every numerical change in FVC as a continuous outcome, the registered endpoint classifies participants according to whether the decline reaches 10% or greater.
Why is the 52-week time frame important?
The time frame establishes when the registered endpoint is assessed. A participant's classification at 52 weeks is therefore different from an analysis that uses the complete timing of events throughout follow-up.
Why can a composite endpoint be difficult to interpret?
The two components may differ in clinical importance, frequency, and susceptibility to treatment. A treatment effect on the composite can therefore be driven primarily by one component. Component-specific results are valuable for understanding what produced the overall finding.
What is the difference between risk ratio and risk difference?
The risk ratio expresses the treatment-group probability relative to the comparator probability. The risk difference expresses the absolute separation between the two probabilities. Relative measures can look substantial even when absolute differences are small, so both scales can be informative.
When would Kaplan-Meier analysis be appropriate?
Kaplan-Meier estimation is appropriate when the analysis uses the time until the composite event and some participants are censored before experiencing the event. It is not required simply because an endpoint contains the word "death"; the actual analysis definition determines whether time-to-event methods are appropriate.
What does a Cox hazard ratio mean?
If a Cox model were used for time to the composite endpoint, its hazard ratio would describe a relative difference in the modeled instantaneous event rate. It would not directly equal a difference in 52-week event proportions and would rely on the assumptions of the fitted model.
10. Analysis of the Posted Safety Information
The registry reports serious adverse events by randomized arm using the number affected and the number at risk. This provides an arm-specific safety count and denominator.
| Safety measure | Pirfenidone | Placebo |
|---|---|---|
| Serious adverse events | 10 / 62 | 11 / 60 |
The two reported figures are counts over their respective reported risk sets. They should not be treated as an efficacy result or combined with the primary endpoint into a single benefit-risk statistic.
Pirfenidone
10 participants were affected among 62 at risk for the reported serious-adverse-event measure.
Placebo
11 participants were affected among 60 at risk for the reported serious-adverse-event measure.
The safety denominators reported by the registry are distinct from the overall enrollment of 123. That distinction matters because an adverse-event rate must use the denominator specified for the safety analysis rather than automatically using the randomized enrollment.
11. Randomization and Masking: Statistical Importance
Randomization is central to the causal interpretation of a treatment comparison. By assigning participants to pirfenidone or placebo through a randomized process, the study creates treatment groups whose differences in outcomes can be interpreted in relation to treatment assignment, subject to the usual assumptions and limitations of randomized trials.
Triple masking adds another layer of protection against differential expectations or assessment. The registry identifies the study as triple-masked but does not specify in the ClinicalTrials.gov record which three groups of individuals were masked. The masking designation therefore should not be expanded beyond the registry's wording.
The statistical comparison should preserve the randomized treatment assignment when estimating the primary efficacy effect unless the prespecified analysis explicitly defines another population.
12. Interpreting the Primary Endpoint Correctly
The endpoint measures whether a participant developed any element of the composite outcome by 52 weeks: a decline in percent predicted FVC of 10% or greater or death.
It does not, by definition, measure the magnitude of an FVC decline as a continuous outcome, the time to death alone, or the time to FVC decline alone. Those would be distinct statistical outcomes.
An event count without its analysis denominator cannot establish an event rate. A valid between-arm comparison requires knowing which participants were included in each analysis population.
A composite result is most informative when the contribution of its components can also be understood. A treatment difference dominated by FVC decline would have a different statistical story from one dominated by death.
13. What a Formal Effect Estimate Would Tell Us
If the primary analysis reports an absolute risk difference, that estimate answers the question of how far apart the two 52-week composite-event probabilities are. If it reports a risk ratio, the estimate answers a relative question. If it reports an odds ratio, the comparison is based on odds rather than probabilities.
| Measure | Question answered | Key interpretation |
|---|---|---|
| Risk difference | How many more or fewer events occur per unit of population probability? | Absolute separation between groups. |
| Risk ratio | How does event probability in pirfenidone compare with placebo? | Relative event probability. |
| Odds ratio | How do event odds compare between groups? | Relative odds, which differs from relative risk when events are not rare. |
| Hazard ratio | How does the modeled instantaneous event rate compare over time? | Time-to-event relative measure requiring a time-to-event analysis. |
These measures should not be treated as interchangeable. In particular, a hazard ratio is not simply another name for a risk ratio, and neither one directly describes the absolute number of participants experiencing the endpoint.
14. Confidence Intervals and Statistical Precision
For a formal treatment-effect estimate, the confidence interval is an essential part of the result. It communicates how precisely the analysis estimates the underlying contrast under the chosen statistical model and sampling framework.
A narrow confidence interval indicates greater statistical precision than a wide interval, all else being equal. Precision depends on factors such as the number of participants, number of events, variability, and the analysis method.
15. P-Values and Effect Size
If a formal hypothesis test is reported for the primary endpoint, its p-value addresses compatibility of the observed data with the null hypothesis under the specified test. It does not measure the size of the treatment effect.
For example, a small p-value can accompany a modest effect when the analysis is highly precise, while a clinically meaningful-looking estimate can have substantial statistical uncertainty when the event count is limited. Effect estimates and confidence intervals therefore provide information that a p-value alone cannot provide.
These are related but different statistical questions.
16. Serious Adverse Events and Comparative Interpretation
The reported serious-adverse-event information is 10/62 for pirfenidone and 11/60 for placebo. These figures show the affected count and the corresponding reported risk set for each arm.
Safety interpretation should preserve the distinction between descriptive event rates and formal comparative inference. A formal comparison would require an explicitly defined safety estimand, denominator, follow-up framework, and statistical method.
It is also important not to infer that a difference in serious adverse events establishes either overall safety superiority or equivalence. A single safety category captures only one dimension of tolerability, and the registry information reported here does not provide a complete safety profile.
17. Statistical Issues Specific to the Composite Endpoint
| Issue | Why it matters |
|---|---|
| Different components | FVC decline and death have different clinical meanings. |
| Component frequency | The more frequent component can contribute more events to the composite. |
| Threshold definition | The FVC component depends on the prespecified 10% or greater decline threshold. |
| Fixed time frame | The registry specifies assessment at 52 weeks. |
| Analysis denominator | Event counts must be interpreted relative to the participants included in the analysis. |
| Time-to-event distinction | A 52-week binary endpoint differs from an analysis using the timing of events and censoring. |
The composite endpoint therefore has two layers of statistical interpretation: first, whether the combined event differs between randomized groups; second, what the component-level pattern says about the meaning of that difference.
18. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary endpoint is a prespecified composite binary outcome assessed at 52 weeks. Its comparison requires event counts, appropriate denominators, an effect measure, and uncertainty assessment.
Clinical interpretation
The endpoint is intended to capture either a meaningful decline in pulmonary function or death, but the clinical meaning of the overall result depends on which component contributes the observed events.
Keeping these two levels separate prevents a common statistical error: treating a numerical difference between groups as self-explanatory without examining what was measured, how it was analyzed, and which component generated the outcome.
19. Important Limitations and Interpretation Issues
- Composite endpoint: the endpoint combines FVC decline and death, which are clinically different events.
- Component contribution: interpretation of a composite treatment effect depends on the relative contribution of its individual components.
- Fixed-time versus time-to-event analysis: a 52-week endpoint classification is not equivalent to an analysis of the exact time until an event.
- Analysis denominator: the relevant denominator must come from the prespecified efficacy analysis population rather than being assumed from overall enrollment.
- Safety denominator: the reported serious-adverse-event risk sets are 62 and 60, respectively, and therefore differ from the total enrollment of 123.
- Formal inferential detail: the registry record reports the primary endpoint and posted results status but does not provide a posted formal statistical analysis in the record represented here.
- Masking detail: the registry identifies the study as triple-masked but does not specify in the ClinicalTrials.gov record which three groups were masked.
- Generalizability: the trial population is defined by rheumatoid arthritis interstitial lung disease and the study's eligibility criteria; results should therefore be interpreted within that study population.
20. Why This Trial Matters Statistically
TRAIL1 is a useful teaching case because it illustrates how a randomized clinical trial can reduce a complex clinical question to a prespecified composite outcome and how that outcome must then be translated into an appropriate statistical comparison.
| Concept | How it appears in TRAIL1 |
|---|---|
| Randomization | Participants were assigned in a randomized parallel-group design. |
| Control group | Placebo provides the comparator for pirfenidone. |
| Masking | The study is identified as triple-masked. |
| Composite endpoint | FVC decline of 10% or greater or death. |
| Fixed assessment time | The primary endpoint has a 52-week time frame. |
| Binary outcome analysis | The registered endpoint can be represented as whether a participant developed any component. |
| Absolute effect | A risk difference can describe the absolute separation between treatment groups. |
| Relative effect | A risk ratio or odds ratio can describe a relative comparison of fixed-time event probabilities. |
| Time-to-event distinction | Kaplan-Meier and Cox methods would apply if the analysis uses event timing and censoring rather than only the 52-week classification. |
| Safety analysis | Serious adverse events are reported separately by arm with their own denominators. |
21. A Practical Statistical Reading of TRAIL1
A rigorous reading of the primary endpoint should proceed in a fixed sequence. First identify the population and randomized treatment assignment. Next identify exactly what qualifies as an event. Then determine the analysis time frame and denominator. Only after those steps should the treatment effect be summarized numerically.
Step 1 · Define the event
Either a decline in percent predicted FVC of 10% or greater or death.
Step 2 · Define the clock
The registered primary endpoint is assessed over a 52-week time frame.
Step 3 · Define the denominator
Use the participants included in the prespecified efficacy analysis rather than assuming the enrollment number is automatically the denominator.
Step 4 · Choose the estimand
Use a fixed-time risk measure for a fixed-time binary analysis or a time-to-event measure if event timing and censoring are incorporated.
This sequence is broadly useful in clinical-trial statistics. It prevents the common mistake of selecting an effect measure before establishing precisely what constitutes an event and when that event is being evaluated.
22. What the Registry Does and Does Not Establish
The ClinicalTrials.gov record establishes that TRAIL1 was a completed, randomized, parallel, triple-masked phase 2 study with 123 enrolled participants, two interventions, and a registered primary endpoint evaluated at 52 weeks.
The record represented here also identifies the primary endpoint as having posted results, while no formal statistical analysis is posted in the registry record represented here. Consequently, the methodological discussion above distinguishes the endpoint itself from statistical procedures that are commonly appropriate for analyzing that type of endpoint.
23. Sources
- ClinicalTrials.gov: NCT02808871 — TRAIL1.
- PubMed: PubMed record 36075242.
- PubMed: PubMed record 31515704.
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical-trial analyses and statistical resources for understanding study design, endpoints, effect measures, and interpretation.
24. Record Summary
TRAIL1 was a randomized, parallel, triple-masked phase 2 study of pirfenidone versus placebo in rheumatoid arthritis interstitial lung disease. The registry reports 123 enrolled participants and a primary endpoint defined as the number of participants who developed any element of a composite outcome consisting of a decline in percent predicted FVC of 10% or greater or death, assessed over 52 weeks.
Statistically, the central issue is the distinction between the clinical definition of the composite endpoint and the method used to compare it. A fixed-time analysis can summarize the event probability in each randomized group and express the contrast through an absolute or relative effect measure. A time-to-event analysis would answer a different question by incorporating the timing of events and censoring. The distinction is fundamental when interpreting clinical-trial results.
The reported serious-adverse-event information is also presented separately, with 10 of 62 participants affected in the pirfenidone group and 11 of 60 in the placebo group. These safety data use their own reported risk sets and should remain separate from the efficacy estimand.