This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial facts on this page are restricted to the ClinicalTrials.gov record. The registry lists posted outcome measures, but it does not provide formal statistical analyses for the two registered primary endpoints in the ClinicalTrials.gov record.
1. Trial at a Glance
CHEST-2 was a phase 3, single-group, unmasked long-term extension study of riociguat in patients with pulmonary hypertension. The registry reports enrollment of 237 participants and two registered primary binary endpoints: treatment-emergent adverse events and death.
| Feature | CHEST-2 |
|---|---|
| Trial name | CHEST-2 |
| Brief title | BAY63-2521 - Long-term Extension Study in Patients With Chronic Thromboembolic Pulmonary Hypertension |
| Phase | Phase 3 |
| Therapeutic area | Pulmonology |
| Condition | Pulmonary Hypertension |
| Intervention | Riociguat (Adempas, BAY63-2521) |
| Allocation | NA |
| Design model | SINGLE_GROUP |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Enrollment | 237 |
| Arms | 1 |
| Status | COMPLETED |
| Start | 2009-07-01 |
| Primary completion | 2019-08-19 |
| Lead sponsor | Bayer |
| Sponsor type | INDUSTRY |
2. Clinical Question
The registry describes CHEST-2 as a long-term extension study of riociguat in patients with pulmonary hypertension. Statistically, the central questions represented by the registered primary endpoints are whether participants experience treatment-emergent adverse events during the specified treatment period and how many participants die during the specified safety follow-up period.
Population
Patients enrolled in the long-term extension study with pulmonary hypertension. The registry title identifies the study as a long-term extension in patients with chronic thromboembolic pulmonary hypertension.
Intervention
Riociguat (Adempas, BAY63-2521).
Study structure
A phase 3, single-group, unmasked extension study with one study arm and enrollment of 237 participants.
Primary questions
How many participants experience treatment-emergent adverse events, and how many participants die during the registered endpoint time frames?
3. Trial Design
Study structure and statistical consequence
A single-group extension study has a different inferential target from a parallel randomized controlled trial. There is no contemporaneous randomized comparator represented in the registry-reported CHEST-2 design data. Consequently, the most direct statistical quantities are counts, proportions, and—in a fuller dataset—time-to-event summaries or other descriptive measures.
This distinction is particularly important for safety. A high or low observed adverse-event frequency does not, by itself, establish that riociguat caused the event at a particular rate relative to an untreated or alternative-treatment population. Background disease, prior therapies, patient selection, duration of observation, and other factors can all affect the observed frequency.
4. Intervention
Riociguat
- Riociguat
- Adempas
- BAY63-2521
- Drug intervention
No randomized comparator
- Single-group design
- Allocation recorded as NA
- Masking recorded as none
- No separate randomized control arm is specified in the ClinicalTrials.gov record
The registry's serious-adverse-event summary is nevertheless divided into two historical-treatment categories: Former Riociguat 1.0-2.5 mg and Former Placebo. These categories should not be relabeled as current randomized arms of CHEST-2. The registry-reported design explicitly identifies CHEST-2 as a single-group study.
5. Primary Endpoints
| Endpoint | Registry definition / time frame | Endpoint type | Formal analysis posted |
|---|---|---|---|
| Number of Participants With Treatment-emergent Adverse Events (TEAE) | From administration of first dose of study medication up to 2 days after end of treatment with study medication, up to 10 years | Binary | No statistical analysis reported |
| Number of Participants With Death | From baseline to end of safety follow-up visit, up to 10 years (1 month more than End of study visit) | Binary | No statistical analysis reported |
TEAE endpoint
The first primary endpoint is the number of participants with treatment-emergent adverse events. The registered observation window begins with administration of the first dose of study medication and extends through 2 days after the end of treatment, with the registry giving the overall time frame as up to 1.
The registry definition further states that analyses of drug-related TEAEs were based on the assessment of causal relationship to study medication. This is a distinction between whether an adverse event occurred and whether investigators assessed a causal relationship to the study medication.
Death endpoint
The second primary endpoint is the number of participants with death, measured from baseline to the end of the safety follow-up visit, with a registered time frame of up to 10 years. The registry states that the safety follow-up visit was to be performed 30 days after the last dose of riociguat.
The registry also states that analyses of deaths were based on assessment of causal relationship to study medication. Again, this does not turn the binary death endpoint into a causal treatment-effect estimate; it describes how relationship to study medication was assessed.
6. Statistical Analysis Status
This is an important distinction between posted outcome data and a posted formal statistical comparison. The registry can report how many participants experienced an event without supplying a model-based estimate, hypothesis test, or confidence interval.
For a single-group binary endpoint, the natural descriptive quantity is the observed proportion:
where x is the number of participants experiencing the event and n is the number of participants in the relevant analysis population.
A confidence interval for a single proportion can then quantify statistical uncertainty around the observed proportion. The choice between exact binomial and large-sample methods should be prespecified or justified based on the sample size and event frequency. Neither such calculation nor a corresponding p-value is reported in the ClinicalTrials.gov record, so none is presented here as a CHEST-2 result.
7. Planned Analysis: Treatment-Emergent Adverse Events
The registry reports the TEAE endpoint as a binary outcome: each participant can be classified according to whether a treatment-emergent adverse event occurred within the registered observation window.
| Statistical component | Appropriate interpretation for this endpoint |
|---|---|
| Outcome | Participant-level binary indicator: TEAE occurred or did not occur. |
| Primary descriptive measure | Number and proportion of participants with a TEAE. |
| Uncertainty | A confidence interval for the proportion would quantify sampling uncertainty if calculated. |
| Causal assessment | Drug-related TEAE analyses were based on assessment of causal relationship to study medication. |
| Comparator | No randomized comparator is represented by the registry-reported CHEST-2 single-group design. |
For a single-group safety extension, a descriptive proportion is usually more directly aligned with the design than a between-arm hypothesis test. If a historical benchmark were introduced, that would create a different inferential question and would require careful justification of the benchmark population and its comparability.
The TEAE endpoint answers a straightforward descriptive question: how many participants experienced at least one treatment-emergent adverse event during the defined observation period? It does not, by itself, measure the severity of every event, the duration of each event, or the probability that riociguat caused an event.
The registry's separate assessment of drug-related TEAEs is therefore important. An event can be treatment-emergent without necessarily being judged related to study medication. Conversely, causal attribution is an assessment rather than a randomized estimate of attributable risk.
A confidence interval around a proportion would describe uncertainty in the estimated proportion; it would not mean that individual patients have probabilities confined to that interval. Likewise, a p-value—if one were specified against an external benchmark—would test the corresponding null hypothesis rather than quantify the magnitude or clinical importance of the adverse-event rate.
8. Planned Analysis: Death
The registered death endpoint is also classified as binary. The registry defines the endpoint from baseline through the end of the safety follow-up visit, up to 10 years, and states that the safety follow-up visit was to occur 30 days after the last dose of riociguat.
| Statistical component | Appropriate interpretation for this endpoint |
|---|---|
| Outcome | Participant-level indicator of death during the registered follow-up period. |
| Primary descriptive measure | Number and proportion of participants with death. |
| Follow-up | From baseline to end of safety follow-up, up to 10 years. |
| Causal assessment | The registry states that analyses of deaths were based on assessment of causal relationship to study medication. |
| Randomized comparison | Not represented in the registry-reported single-group CHEST-2 design. |
Death is inherently a time-dependent event, so a richer analysis could preserve information about when deaths occurred rather than reducing follow-up to a binary indicator. For example, a time-to-event analysis would normally use each participant's event or censoring time and could estimate a survival function. However, the registered primary endpoint reported here is explicitly classified as binary, and no formal time-to-event statistical analysis is in the posted results.
A binary death endpoint tells us whether a participant died during the specified follow-up period. It does not distinguish between deaths occurring early and late within that period. That distinction matters when follow-up varies across participants or when the timing of events is scientifically important.
The registered time frame of up to 10 years gives the endpoint a long observation horizon, but a binary summary alone would discard the exact timing information. A Kaplan-Meier analysis or another time-to-event method would normally preserve that information if event dates and censoring information were available.
Because CHEST-2 is a single-group study in the ClinicalTrials.gov record, an observed mortality proportion should not be interpreted as a randomized estimate of excess or reduced mortality caused by riociguat. A comparison with another population would require assumptions about comparability that are not established by the ClinicalTrials.gov record.
9. Reported Safety Results
The ClinicalTrials.gov record provides a serious-adverse-event summary by historical treatment category. These figures are the available arm-level safety counts in the ClinicalTrials.gov record.
| Historical treatment category | Serious adverse events: affected / at risk |
|---|---|
| Former Riociguat 1.0-2.5 mg | 96 / 155 |
| Former Placebo | 56 / 82 |
The two denominators are important. The registry reports 96/155 for former riociguat and 56/82 for former placebo. These categories describe prior treatment history rather than two randomized CHEST-2 arms. Therefore, the figures should be presented as reported safety categories rather than as a randomized comparison within CHEST-2.
10. Statistical Methods Explained
Why are the primary endpoints called binary?
A binary endpoint classifies each participant into one of two states for the defined analysis window—for example, whether a treatment-emergent adverse event occurred or did not occur. The resulting analysis naturally starts with a count and proportion rather than a mean or a continuous treatment difference.
Why is a single-group study different from a randomized trial?
Randomization creates comparability between treatment groups in expectation and provides the foundation for a causal treatment comparison. In a single-group extension, there is no contemporaneous randomized comparator. Consequently, an observed event frequency primarily describes the participants who entered and remained under observation rather than directly estimating a randomized treatment effect.
What would a confidence interval for a binary endpoint tell us?
A confidence interval for a single proportion would quantify statistical uncertainty around the observed proportion under the chosen sampling model. It would not describe the range of outcomes that individual patients will experience, nor would it establish causality.
Why does the death endpoint raise a time-to-event issue?
The registered death endpoint is binary, but death occurs at a particular time. Converting the entire follow-up period into a yes/no outcome can discard timing information. If exact event and censoring times are available, Kaplan-Meier estimation or another survival-analysis method can use that information more efficiently.
Why shouldn't the serious-adverse-event categories be treated as treatment groups?
The registry design identifies CHEST-2 as SINGLE_GROUP with one arm. The reported categories—Former Riociguat 1.0-2.5 mg and Former Placebo—therefore represent historical treatment categories in the reported safety summary. They are not evidence that CHEST-2 randomized participants between those categories.
What does a drug-related adverse-event assessment mean?
The registry states that analyses of drug-related TEAEs were based on assessment of causal relationship to study medication. This is an attribution assessment. It should not be confused with a randomized estimate of attributable risk, because CHEST-2's registry-reported design does not contain a randomized comparator.
11. Descriptive Proportions Versus Comparative Effects
One of the most important statistical distinctions in CHEST-2 is between describing an event rate and estimating a treatment effect.
Descriptive question
What proportion of enrolled participants experienced a specified event during the registered observation period?
Comparative question
How does the event probability under one treatment compare with the event probability under another treatment?
CHEST-2 structure
The registry-reported design supports the descriptive question directly because the study is single-group.
What is missing for causal comparison
A contemporaneous randomized comparator or a separately justified comparative framework would be needed for a direct treatment-effect interpretation.
This distinction is not merely semantic. A proportion such as x/n answers what happened among the observed study population. A risk difference, risk ratio, odds ratio, or hazard ratio answers a comparative question and therefore requires two populations or another explicit reference framework.
These measures require a defined comparison between two populations. They should not be inferred simply because a safety table contains two historical treatment categories.
12. Interpreting the Serious-Adverse-Event Counts
The reported serious-adverse-event data contain both numerators and denominators, which is statistically useful because the raw counts alone would not account for the different numbers at risk in the two historical categories.
| Reported category | Affected | At risk | What the data establish |
|---|---|---|---|
| Former Riociguat 1.0-2.5 mg | 96 | 155 | 96 participants were reported as affected among 155 at risk. |
| Former Placebo | 56 | 82 | 56 participants were reported as affected among 82 at risk. |
The ClinicalTrials.gov record does not provide a formal statistical analysis comparing these historical categories. In particular, no confidence interval, p-value, risk ratio, risk difference, odds ratio, or adjusted model is included in the statistical-analyses field.
The denominators allow the reader to understand the scale of the reported serious-adverse-event experience, but the categories must remain tied to their registry labels. Because they are historical treatment categories within a single-group extension study, a difference between the two observed proportions would not automatically represent a causal treatment difference.
A formal comparative analysis would require assumptions about how participants entered the extension study, what treatment they previously received, their exposure duration, and other differences between the historical groups. Those assumptions are not reported in the ClinicalTrials.gov record and therefore should not be silently imposed.
13. Follow-Up and Endpoint Windows
Study start
The registry lists 2009-07-01 as the study start date.
First dose through 2 days after treatment
The TEAE endpoint is measured from administration of the first dose of study medication through 2 days after the end of treatment, with the registered overall time frame given as up to 1.
Baseline through safety follow-up
The death endpoint is measured from baseline to the end of the safety follow-up visit, up to 10 years.
Primary completion
The registry lists 2019-08-19 as the primary completion date.
The different endpoint windows illustrate why endpoint definitions matter. A TEAE analysis is anchored to treatment exposure and extends through a specified short post-treatment period, whereas the death endpoint extends from baseline through safety follow-up. The two endpoints therefore do not have interchangeable denominators or observation windows.
14. Analysis Population and Denominator Discipline
For binary outcomes, the denominator is as important as the numerator. The interpretation of a count such as the number of participants with an adverse event depends on exactly who was considered at risk and which observation period was used.
TEAE denominator
The relevant population must correspond to participants observed from first dose through the registered TEAE window.
Death denominator
The relevant population must correspond to participants included in the baseline-to-safety-follow-up definition.
Safety-category denominator
The registry-reported serious-adverse-event categories explicitly report their own at-risk denominators: 155 and 82.
Why denominators can differ
Different endpoint definitions and analysis populations can produce different denominators. Combining them without checking the endpoint definition can create misleading rates.
The ClinicalTrials.gov record does not specify a separate formal analysis population definition beyond the registered endpoint descriptions and reported safety denominators.
15. Missing Data and Censoring Considerations
Binary endpoints and time-to-event endpoints handle incomplete follow-up differently. For a simple binary analysis, participants need a sufficiently defined observation window to be classified correctly. Participants without complete ascertainment can create ambiguity if the endpoint definition does not specify how incomplete observations are handled.
The registry-reported CHEST-2 data do not report a statistical imputation strategy for missing binary outcomes. No formal missing-data method is included in the registry-reported statistical-analyses field.
The death endpoint is also conceptually compatible with survival analysis because participants can have different follow-up durations. A survival analysis can treat participants who remain alive at their last known follow-up as right-censored rather than forcing them into an artificial binary classification at a common time point. However, the registry definition and endpoint type are binary, and the page does not substitute an unreported survival model for the registered analysis.
16. What the Registry Data Do and Do Not Establish
| Question | Supported by the ClinicalTrials.gov record? | Interpretation |
|---|---|---|
| Was CHEST-2 a phase 3 study? | Yes | the ClinicalTrials.gov record identifies the phase as 3. |
| Was it single-group? | Yes | The design model is SINGLE_GROUP. |
| Were there 237 enrolled participants? | Yes | Enrollment is 237. |
| Were two primary endpoints registered? | Yes | TEAE and death. |
| Were the primary endpoints binary? | Yes | Both are classified as binary. |
| Were results posted? | Yes | the ClinicalTrials.gov record says Results Posted: Yes. |
| Are formal statistical analyses reported? | No | No formal statistical analyses were posted. |
| Is there a randomized treatment comparison? | No evidence in registry-reported design | The registry-reported design is single-group with one arm. |
| Are serious adverse events reported by historical treatment category? | Yes | 96/155 and 56/82 are reported. |
| Can those historical categories be treated as randomized CHEST-2 arms? | No | That would contradict the registry-reported single-group design. |
17. Statistical Methods Explained in Context
Why not report a hazard ratio for CHEST-2?
A hazard ratio requires a defined time-to-event comparison, generally between two groups or between covariate-defined hazards. The registry-reported CHEST-2 statistical analyses contain no hazard-ratio estimate. Although the death endpoint has a long follow-up window and could conceptually be analyzed with survival methods, reporting an HR here would create a result that is not present in the ClinicalTrials.gov record.
Why not report a p-value for the TEAE endpoint?
A p-value requires a specified null hypothesis and statistical test. For a single-group proportion, the null might involve comparison with a prespecified benchmark, but no such benchmark or formal test is reported here.
Could the 96/155 and 56/82 values be compared directly?
They could be described as two observed historical-category proportions, but the registry-reported study design does not establish that they are randomized treatment groups. A formal comparison would require a statistical model and assumptions appropriate to the historical treatment structure. Those are not reported.
Why does the endpoint definition matter for adverse events?
The TEAE endpoint is explicitly tied to the period from first dose through 2 days after the end of treatment. That definition determines which events belong in the endpoint. A different safety window could produce a different event count and therefore a different estimated proportion.
Why is causal attribution separate from event occurrence?
The registry says that drug-related TEAE and death analyses were based on assessment of causal relationship to study medication. The occurrence variable and the causality assessment answer different questions. Statistical summaries should preserve that distinction rather than treating all treatment-emergent events as necessarily caused by riociguat.
Why can a long follow-up period make binary death summaries less informative?
Two studies could have the same proportion of participants who die by the end of follow-up while having very different timing of those deaths. Time-to-event methods preserve that timing information and can also account for censoring. The registered CHEST-2 endpoint, however, is classified as binary in the ClinicalTrials.gov record.
18. Limitations
- No randomized comparator: the registry-reported design identifies CHEST-2 as a single-group study with one arm, limiting direct causal treatment comparisons.
- No formal statistical analyses reported: the registry data identify posted outcomes but provide no entries in the posted results.
- Binary endpoint classification: the primary death endpoint does not, by itself, preserve the exact timing of deaths within the follow-up period.
- Historical treatment categories: the reported serious-adverse-event categories should not be interpreted as randomized CHEST-2 treatment arms.
- Causal attribution: investigator assessment of drug-relatedness is not equivalent to a randomized estimate of attributable risk.
- Missing-data methodology: the ClinicalTrials.gov record does not identify an imputation strategy for missing binary outcomes.
- Unreported model assumptions: because no formal statistical analyses are reported, assumptions for comparative models, survival models, or multiplicity procedures cannot be attributed to the trial.
- Long-term extension context: participants in an extension study can represent a selected population relative to the broader population initially considered for treatment, so extension-study safety experience should be interpreted in that context.
19. Why This Trial Matters Statistically
CHEST-2 is useful as a statistical teaching case because it demonstrates a problem that is easy to overlook: a trial can have posted outcome data without having a readily available formal comparative analysis.
| Concept | How it appears in CHEST-2 |
|---|---|
| Single-group design | The registry identifies the design model as SINGLE_GROUP. |
| Binary endpoints | Both registered primary endpoints are classified as binary. |
| Safety analysis | TEAE is one of the two registered primary endpoints. |
| Long-term mortality | Death is registered from baseline through safety follow-up, up to 10 years. |
| Denominator discipline | Serious adverse events are reported with explicit affected/at-risk counts. |
| Causal attribution | Drug-related TEAEs and deaths were assessed for causal relationship to study medication. |
| Historical treatment groups | Safety results distinguish former riociguat and former placebo categories without changing the single-group design. |
| Missing formal inference | No formal statistical analyses were posted. |
| Endpoint definition | Different endpoints use different observation windows and therefore require different denominator discipline. |
The broader statistical lesson is that the sophistication of an analysis must match the information actually available.
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The ClinicalTrials.gov record describes a phase 3, single-group extension study with two binary primary endpoints and posted outcomes, but no formal statistical analyses are posted on ClinicalTrials.gov for those endpoints.
Clinical interpretation
The study provides long-term safety and mortality outcome information for participants receiving riociguat in the extension setting. The ClinicalTrials.gov record does not establish a randomized comparative treatment effect.
Safety interpretation
The reported serious-adverse-event counts are 96/155 for Former Riociguat 1.0-2.5 mg and 56/82 for Former Placebo.
Inference boundary
The historical treatment categories should not be converted into randomized arms or used to construct an unreported causal comparison.
21. Sources
- ClinicalTrials.gov: NCT00910429 — CHEST-2.
- PubMed: PMID 34353714.
- PubMed: PMID 27162632.
- PubMed: PMID 27067478.
- PubMed: PMID 23883377.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Record Summary
CHEST-2 is a phase 3, single-group, unmasked long-term extension study of riociguat in pulmonary hypertension with enrollment of 237 participants. Its two registered primary endpoints are binary measures of treatment-emergent adverse events and death. The TEAE endpoint is defined from first dose through 2 days after the end of treatment, while the death endpoint extends from baseline through the safety follow-up visit, up to 10 years.
The ClinicalTrials.gov record indicates that results have been posted, but they do not supply formal statistical analyses for the primary endpoints. The available serious-adverse-event summary reports 96/155 affected participants for Former Riociguat 1.0-2.5 mg and 56/82 for Former Placebo. Those categories are historical treatment classifications and should not be interpreted as randomized CHEST-2 arms because the study is recorded as single-group.
Statistically, the most important lesson is therefore one of inferential discipline: The appropriate interpretation depends on the endpoint definition, observation window, denominator, single-group design, and the distinction between event occurrence and investigator-assessed causal relationship to study medication.
Continue through the Clinical Biostats statistical library
Explore the statistical methods behind binary endpoints, safety analysis, single-arm studies, and time-to-event data.
25. Methodological Takeaway
CHEST-2 demonstrates why clinical-trial statistical interpretation begins with the design rather than with the numbers. A phase 3 designation does not by itself imply randomization, and a posted safety table does not by itself imply a formal treatment comparison. Here, the registry design is single-group, the two primary endpoints are binary, and no formal statistical analyses are reported.
For the TEAE and death endpoints, the appropriate starting point is therefore transparent reporting of event counts and clearly defined denominators, followed by an explicit description of uncertainty if the underlying analysis data permit it. More complex methods such as risk ratios, hazard ratios, or regression models should only be introduced when the design and available data support the corresponding inferential question.
That separation between what was observed, what was formally analyzed, and what can reasonably be inferred is central to responsible clinical biostatistics.