This page separates reported trial results from statistical interpretation. Numerical results are restricted to the ClinicalTrials.gov record for CHEST-1. The registry provides the official trial record.
1. Trial at a Glance
CHEST-1 was a completed, randomized, parallel-group, triple-masked phase 3 treatment trial evaluating riociguat (Adempas, BAY63-2521) versus placebo in pulmonary hypertension. The registered primary endpoint was change in 6-minute walking distance from baseline to week 16.
| Feature | CHEST-1 |
|---|---|
| Trial name | CHEST-1 |
| ClinicalTrials.gov identifier | NCT00855465 |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Pulmonology |
| Condition | Pulmonary Hypertension |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 262 |
| Interventions | Riociguat (Adempas, BAY63-2521) and placebo |
| Primary endpoint | 6 Minutes Walking Distance (6MWD) - Change From Baseline to Week 16 |
| Results posted | Yes |
| Outcome measures posted | 42 |
| Statistical analyses posted | 24 |
| Primary- endpoint analyses posted | 3 |
| Hypothesis type | Superiority |
2. Clinical Question
The central statistical question was whether participants assigned to riociguat had a different change in functional exercise capacity, measured by the 6-minute walking distance, from baseline to week 16 than participants assigned to placebo.
Population
Participants enrolled in the phase 3 CHEST-1 trial for pulmonary hypertension. The ClinicalTrials.gov record identifies the trial population but do not provide a detailed baseline-characteristics table.
Intervention
Riociguat (Adempas, BAY63-2521), with individual dose titration as identified in the statistical analyses.
Comparator
Placebo.
Primary question
Does riociguat produce a different change in 6MWD from baseline to week 16 than placebo under a superiority hypothesis?
3. Trial Design
Riociguat
- Riociguat (Adempas, BAY63-2521)
- Individual dose titration
Placebo
- Placebo
The registry data do not provide randomized arm counts for the full enrolled population. The safety analysis data do provide affected/at-risk counts of 34/173 for riociguat and 14/88 for placebo; these denominators should therefore not be treated as the overall randomized sample sizes without additional evidence.
4. Endpoints
| Endpoint | Registry definition | Time frame | Endpoint type |
|---|---|---|---|
| Primary: 6 Minutes Walking Distance (6MWD) - Change From Baseline to Week 16 | 6-minute walking distance (6MWD) is a measure for the objective evaluation of a participant's functional exercise capacity. | Baseline and week 16 | Continuous |
| Pulmonary Vascular Resistance (PVR) - Change From Baseline to Week 16 | Change in pulmonary vascular resistance. | Baseline and week 16 | Continuous |
| NT-proBNP - Change From Baseline to Week 16 | Change in N-terminal Prohormone of Brain Natriuretic Peptide. | Baseline and week 16 | Continuous |
| Percentage of Participants With Clinical Worsening | Percentage of participants experiencing clinical worsening. | At week 16 | Time-to-event |
| EQ-5D Utility Score - Change From Baseline to Week 16 | Change in EQ-5D utility score. | Baseline and week 16 | Continuous |
| Living With Pulmonary Hypertension (LPH) Questionnaire - Change From Baseline to Week 16 | Change in LPH Questionnaire score. | Baseline and week 16 | Continuous |
| WHO Functional Class - Change From Baseline to Week 16 | Change in WHO Functional Class. | Baseline and week 16 | Binary in the posted statistical analysis |
| Borg CR 10 Scale - Change From Baseline to Week 16 | Change in Borg CR 10 Scale score. | Baseline and week 16 | Continuous |
| Mean Pulmonary Artery Pressure (PAPmean) - Change From Baseline to Week 16 | Change in mean pulmonary artery pressure. | Baseline and week 16 | Continuous |
| Cardiac Index (CI) - Change From Baseline to Week 16 | Change in cardiac index. | Baseline and week 16 | Continuous |
The registry reports 42 outcome measures and 24 statistical analyses. The table above focuses on endpoints for which the ClinicalTrials.gov record contains specific methods or estimates.
5. Statistical Methodology
Primary analysis: ANCOVA
The primary 6MWD comparison used analysis of covariance (ANCOVA). The analysis population was the Intent-to-Treat (ITT) population, defined in the registry analysis record as a randomized participant who was valid for ITT analyses if at least one dose of study medication was administered.
The analysis notes specify that baseline 6MWD was used as a covariate, with treatment group and region as main effects. This is an important distinction from simply comparing the two groups' raw week-16 means: ANCOVA adjusts the final-value comparison for baseline 6MWD and includes the specified treatment and region terms.
The exact fitted coefficients are not reported in the ClinicalTrials.gov record. The model expression illustrates the role of baseline 6MWD, treatment group, and region in the analysis.
Wilcoxon / Mann-Whitney testing
The registry also reports Wilcoxon (Mann-Whitney) analyses for the primary 6MWD endpoint and several secondary endpoints. This is a nonparametric approach that does not rely on the same normal-distribution assumption as a conventional parametric comparison.
For the primary endpoint, the analysis notes describe missing-value handling and state that comparison was done using ANCOVA with baseline 6MWD as a covariate and treatment group and region as main effects. The posted analysis also contains a Wilcoxon result with a P-value of <0.0001.
Log-rank testing
Clinical worsening was treated as a time-to-event endpoint. The registry reports a log-rank analysis, with the test for difference in occurrence of "Any event" stratified by region. The effect measure reported for this analysis was risk difference, based on a Mantel-Haenszel estimate stratified by region.
Normality assessment
Shapiro-Wilk testing is reported for ANCOVA residuals. For the primary endpoint, the posted P-value was 0.0001. The purpose of this analysis is diagnostic: it evaluates the distributional behavior of the residuals rather than directly estimating the treatment effect.
Missing-data procedures
The ClinicalTrials.gov record explicitly describe several endpoint-specific imputation rules. For the primary 6MWD endpoint, participants who withdrew or died before week 16 had missing values imputed with a worst value of 0m in case of death or clinical worsening without a termination visit, and with the last observed value otherwise.
For PVR and NT-proBNP, the registry states that missing week-16 values were imputed using the last available post-baseline observation. For WHO Functional Class, missing values after withdrawal or death were assigned worst values according to the clinical outcome described in the analysis notes. EQ-5D and LPH Questionnaire analyses likewise contain explicit missing-data rules.
6. Analysis Population and Covariate Adjustment
The primary analysis used the Intent-to-Treat (ITT) population. In the registry's definition, a randomized participant was valid for ITT analyses if at least one dose of study medication was administered.
| Element | Role in the registry-reported analysis |
|---|---|
| ITT population | Primary efficacy analysis population. |
| Baseline 6MWD | Covariate in the primary ANCOVA. |
| Treatment group | Main treatment comparison. |
| Region | Main effect in the ANCOVA; also used for stratification in selected analyses. |
| Superiority hypothesis | The reported analyses test for a treatment difference rather than non-inferiority. |
Covariate adjustment can improve statistical efficiency when the covariate is strongly related to the outcome. Here, baseline 6MWD is especially relevant because the endpoint is defined as change from baseline to week 16 and baseline functional capacity can explain part of the variation in final measurements.
7. Primary Result: 6-Minute Walking Distance
The registered primary endpoint was 6 Minutes Walking Distance (6MWD) - Change From Baseline to Week 16, with baseline and week 16 as the specified time frame. The outcome unit was meters.
Adjusted mean difference in final values
95% CI: 24.74 to 66.63 m · P < 0.0001
ANCOVA · ITT population · Superiority hypothesis
| Primary endpoint | Riociguat vs placebo |
|---|---|
| Endpoint | 6MWD - Change From Baseline to Week 16 |
| Outcome unit | Meters |
| Analysis population | Intent to Treat |
| Method | ANCOVA |
| Effect measure | Mean Difference (Final Values) |
| Estimate | 45.69 |
| 95% CI | 24.74 to 66.63 |
| P-value | <0.0001 |
| Hypothesis | Superiority |
The reported mean difference of 45.69 meters means that the ANCOVA-estimated difference in final 6MWD between the riociguat and placebo groups was 45.69 meters, with the comparison adjusted for baseline 6MWD and including treatment group and region as main effects.
It does not mean that every participant increased walking distance by 45.69 meters, nor does it describe the change experienced by any particular individual. It is a group-level model-based comparison.
The 95% confidence interval of 24.74 to 66.63 meters describes statistical uncertainty around the estimated mean difference under the analysis framework. It does not represent the range of individual treatment responses.
The P-value of <0.0001 measures how incompatible the observed data are with the null hypothesis under the specified statistical model. It does not measure the magnitude or clinical importance of the effect. The magnitude is described by the 45.69-meter estimate and its confidence interval.
The interpretation also depends on the specified ITT population, the baseline-covariate adjustment, the handling of missing week-16 observations, and the model assumptions underlying ANCOVA.
Supporting nonparametric analysis
The registry also reports a Wilcoxon (Mann-Whitney) analysis for the same primary endpoint, with a P-value of <0.0001. The registry-reported analysis notes identify the same broad ANCOVA framework for the primary comparison and describe missing-value imputation. This provides a nonparametric analysis record alongside the primary ANCOVA result.
8. Secondary Endpoint Results
Pulmonary Vascular Resistance
Mean difference in PVR at week 16
95% CI: −303.33 to −189.53 dyn*s*cm−5 · P < 0.0001
ANCOVA · ITT population · Superiority hypothesis
The negative mean difference indicates that the model-estimated final PVR was lower in the riociguat group than in the placebo group under the reported comparison. The confidence interval remains below zero, spanning −303.33 to −189.53 dyn*s*cm−5.
The −246.43 estimate is a between-group mean difference, not a percentage reduction and not a statement about each participant's individual change. Its negative direction indicates a lower estimated final PVR for riociguat relative to placebo within the reported analysis.
The 95% CI of −303.33 to −189.53 quantifies uncertainty around the estimated difference. The P-value of <0.0001 addresses evidence against the null hypothesis of no difference; it does not quantify the size of the PVR effect.
NT-proBNP
Mean difference in NT-proBNP at week 16
95% CI: −842.95 to −45.03 pg/mL · P = 0.0293
ANCOVA · ITT population · Superiority hypothesis
The reported mean difference of −443.99 pg/mL indicates a lower model-estimated final NT-proBNP value in the riociguat group relative to placebo under the specified analysis.
The 95% CI, from −842.95 to −45.03 pg/mL, is relatively wide compared with the point estimate. That width is important: the data support a range of plausible between-group differences rather than identifying the true effect as exactly −443.99 pg/mL.
The P-value of 0.0293 is evidence against a null difference under the reported test. It should not be read as a 2.93% probability that the null hypothesis is true, nor as a measure of how large the treatment effect is.
Clinical worsening
Risk difference in clinical worsening
95% CI: −8.72 to 1.99 · P = 0.1724
Log-rank analysis · Region-stratified · Mantel-Haenszel risk difference
The reported risk difference of −3.37 indicates a lower estimated occurrence of the analyzed clinical-worsening event in the riociguat group than in the placebo group under the reported comparison.
The 95% CI ranges from −8.72 to 1.99, crossing zero. Thus, the registry-reported estimate is compatible with both a reduction and a small increase in the event difference under the uncertainty represented by the interval.
The P-value of 0.1724 does not provide evidence of a statistically significant difference under the reported superiority test. Importantly, this does not prove that the two groups are identical; the confidence interval shows the range of differences that remains compatible with the data and analysis.
Because this endpoint was analyzed using a log-rank framework and stratified by region, its interpretation should not be reduced to a simple unadjusted proportion comparison.
EQ-5D Utility Score
Mean difference in EQ-5D Utility Score
95% CI: 0.06 to 0.21 · P = 0.0002
ANCOVA · ITT population · Superiority hypothesis
The reported mean difference of 0.13 represents the between-group difference in final EQ-5D utility score under the ANCOVA analysis. The 95% CI of 0.06 to 0.21 describes uncertainty around that estimated difference.
The P-value of 0.0002 is evidence against a null difference under the reported analysis. It does not by itself establish the practical importance of a 0.13-point difference.
Living With Pulmonary Hypertension Questionnaire
Mean difference in LPH Questionnaire score
95% CI: −10.45 to −1.06 · P = 0.0165
ANCOVA · ITT population · Superiority hypothesis
The estimated difference of −5.76 indicates a lower final LPH Questionnaire score in the riociguat group relative to placebo under the reported ANCOVA. The confidence interval, −10.45 to −1.06, expresses uncertainty around that estimate.
The P-value of 0.0165 addresses the statistical evidence for a difference under the reported test. It is not a measure of effect size and does not tell us how large a difference an individual participant experienced.
Other reported secondary analyses
| Endpoint | Method | Effect / P-value | 95% CI |
|---|---|---|---|
| PVR - Change From Baseline to Week 16 | Wilcoxon / Mann-Whitney | P < 0.0001 | Not reported for this analysis |
| NT-proBNP - Change From Baseline to Week 16 | Wilcoxon / Mann-Whitney | P < 0.0001 | Not reported for this analysis |
| WHO Functional Class - Change From Baseline to Week 16 | Wilcoxon / Mann-Whitney | P = 0.0026 | Not reported for this analysis |
| Borg CR 10 Scale - Change From Baseline to Week 16 | Wilcoxon / Mann-Whitney | P = 0.0035 | Not reported for this analysis |
| EQ-5D Utility Score - Change From Baseline to Week 16 | Wilcoxon / Mann-Whitney | P < 0.0001 | Not reported for this analysis |
| LPH Questionnaire - Change From Baseline to Week 16 | Wilcoxon / Mann-Whitney | P = 0.1220 | Not reported for this analysis |
For these analyses, the registry supplies P-values but not corresponding effect estimates and confidence intervals. The P-values therefore should not be presented as substitutes for quantitative effect sizes.
9. Other Pre-Specified Endpoint Results
Mean Pulmonary Artery Pressure
Mean difference in PAPmean
95% CI: −6.75 to −3.16 mmHg · P < 0.0001
ANCOVA · ITT population · Superiority hypothesis
The estimated mean difference of −4.96 mmHg indicates a lower final mean pulmonary artery pressure in the riociguat group relative to placebo under the reported ANCOVA.
The 95% CI from −6.75 to −3.16 mmHg expresses uncertainty around the estimated difference. The P-value of <0.0001 provides evidence against a null difference under the reported superiority test, but does not measure the magnitude of the pressure difference.
Cardiac Index
Mean difference in cardiac index
95% CI: 0.33 to 0.62 L/min/m2 · P < 0.0001
ANCOVA · ITT population · Superiority hypothesis
The estimated mean difference of 0.47 L/min/m2 indicates a higher final cardiac index in the riociguat group relative to placebo under the reported ANCOVA.
The 95% CI of 0.33 to 0.62 L/min/m2 describes uncertainty around the estimated difference. The P-value of <0.0001 addresses evidence against a null difference and does not represent the probability that the observed effect is clinically important.
| Other pre-specified endpoint | Method | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Mean Pulmonary Artery Pressure (PAPmean) - Change From Baseline to Week 16 | ANCOVA | −4.96 mmHg | −6.75 to −3.16 | <0.0001 |
| Cardiac Index (CI) - Change From Baseline to Week 16 | ANCOVA | 0.47 L/min/m2 | 0.33 to 0.62 | <0.0001 |
Supporting nonparametric analyses were also posted for PAPmean and cardiac index, with P-values of <0.0001 for both. A Shapiro-Wilk analysis of PAPmean ANCOVA residuals produced P = 0.0231; the corresponding cardiac-index Shapiro-Wilk P-value was 0.1160.
10. Statistical Methods Explained
Why was ANCOVA used for the primary 6MWD endpoint?
The endpoint is continuous and measured at baseline and week 16. ANCOVA provides a way to compare final values between treatment groups while accounting for baseline 6MWD. In the registry-reported analysis, baseline 6MWD was explicitly included as a covariate, with treatment group and region as main effects.
This differs from simply calculating two unadjusted mean changes and subtracting them. The ANCOVA estimates the treatment-group difference in final values after accounting for the prespecified baseline covariate.
What does a mean difference of 45.69 meters mean?
It is the estimated difference between the two treatment groups in the analyzed final 6MWD values under the ANCOVA model. A positive value indicates a higher estimated final value in the riociguat group relative to placebo.
It is not an individual-level prediction. Some participants may have improved more, less, or not at all. The group-level estimate summarizes the comparison specified by the statistical model.
Why does the confidence interval matter?
A point estimate alone gives only one value. The 95% confidence interval of 24.74 to 66.63 meters communicates the statistical precision of the primary estimate. A narrower interval would generally provide more precision; a wider interval would indicate greater uncertainty.
The interval should not be interpreted as saying that 95% of individual patients have treatment effects inside that range. It concerns uncertainty in the estimated group-level parameter.
Why was the Wilcoxon / Mann-Whitney test also reported?
The Wilcoxon / Mann-Whitney test provides a nonparametric comparison that does not require the same distributional assumptions as a conventional parametric analysis. CHEST-1 reports this method for several endpoints, including the primary 6MWD endpoint.
Its P-value should not be confused with the ANCOVA effect estimate. The Wilcoxon test result provides evidence for a between-group distributional difference, whereas the ANCOVA result supplies the reported adjusted mean-difference estimate.
Why test the ANCOVA residuals with Shapiro-Wilk?
The Shapiro-Wilk test is a diagnostic for the normality assumption applied to ANCOVA residuals. It does not directly test whether riociguat is better than placebo. A P-value from a residual diagnostic therefore has a different statistical meaning from the primary treatment-effect P-value.
Why does missing-data imputation matter?
The primary endpoint is assessed at week 16, but some participants may not have a week-16 measurement. The registry's analysis did not simply discard every participant with a missing value. For the primary 6MWD endpoint, specific worst-value and last-observation rules were used.
Because different assumptions about missing observations can produce different estimated treatment effects, the imputation strategy is part of the analysis definition rather than a minor technical detail.
What does the clinical-worsening risk difference tell us?
The reported risk difference of −3.37 is an absolute between-group difference in the analyzed occurrence of clinical worsening. Unlike a hazard ratio, a risk difference is expressed on an absolute probability scale. Its confidence interval crosses zero, and its reported P-value is 0.1724.
11. Missing Data and Imputation
CHEST-1 is particularly useful for teaching that "ITT" does not automatically mean "no missing data." The registry analyses explicitly describe how missing observations were handled for several endpoints.
| Endpoint | Reported missing-data approach |
|---|---|
| 6MWD | For participants who withdrew or died before 16 weeks: worst value of 0m for death/clinical worsening without termination visit; last observed value otherwise. |
| PVR | Missing week-16 values imputed using the last available post-baseline observation. |
| NT-proBNP | Missing week-16 values imputed using the last available post-baseline observation. |
| WHO Functional Class | Worst value of IV for clinical worsening without termination visit/measurement and V for death; last observed value otherwise. |
| EQ-5D | Missing baseline values imputed using the last available observation before study treatment; subsequent missing values used a worst value of −0.594 for death or clinical worsening without termination visit and last observation otherwise. |
| LPH Questionnaire | Missing baseline values imputed using the last available observation before study treatment; subsequent missing values used a worst value of 105 for death or clinical worsening without termination visit and last observation otherwise. |
| Borg CR 10 Scale | Missing values after withdrawal or death were assigned a worst value of 10 for death or clinical worsening without termination visit, with the last observed value otherwise. |
12. Stratification and Region
Region appears explicitly in the registry-reported analysis methods. For the primary endpoint, the ANCOVA included region as a main effect. For the clinical-worsening analysis, the log-rank test was stratified by region, and the reported risk difference was based on a Mantel-Haenszel estimate stratified by region.
Stratification recognizes that event rates or outcome distributions can differ across regions. Rather than allowing regional imbalance or heterogeneity to be ignored, the analysis incorporates region into the comparison.
The ClinicalTrials.gov record does not identify the individual regions or provide region-specific treatment estimates. Therefore, no region-specific conclusions can be drawn from this record.
13. Safety Results
The ClinicalTrials.gov record provides serious adverse event counts by treatment arm. They do not provide a complete adverse-event table or rates for all adverse-event categories, so the safety section is limited to the reported serious adverse events.
| Safety measure | Riociguat | Placebo |
|---|---|---|
| Serious adverse events | 34/173 | 14/88 |
The denominators in the safety record are the reported at-risk populations for this safety measure. They differ from the total trial enrollment of 262, so the safety counts should not be presented as though they represented all enrolled participants.
14. What the Primary P-value Does — and Does Not — Mean
The primary ANCOVA P-value was <0.0001. Under the reported superiority analysis, this provides strong statistical evidence against a null hypothesis of no treatment difference in the analyzed primary endpoint.
It does not mean that there is a <0.01% probability that the null hypothesis is true. It also does not say that the treatment has a particular probability of benefiting an individual patient.
The effect size is described by the 45.69-meter mean difference, while the 24.74 to 66.63 meter confidence interval describes uncertainty around that estimate. The P-value and confidence interval answer related but different questions.
The primary estimate is conditional on the specified ANCOVA model, baseline 6MWD adjustment, region effect, ITT population, and missing-data procedures. Statistical significance should therefore be interpreted as a property of the complete analysis framework rather than as a standalone characteristic of the treatment.
15. Understanding the Risk Difference for Clinical Worsening
The clinical-worsening analysis provides a useful contrast with the continuous 6MWD endpoint. Its reported effect measure was a risk difference, rather than a mean difference.
A negative risk difference indicates fewer events in the riociguat group under the specified comparison. The reported estimate was −3.37, with a 95% CI from −8.72 to 1.99.
Because the confidence interval crosses zero and the P-value is 0.1724, the registry-reported analysis does not establish a statistically significant difference in the analyzed clinical-worsening event occurrence under the superiority framework.
This illustrates why it is important not to generalize the primary endpoint result to every other endpoint. Different outcomes can have different estimands, statistical models, levels of precision, and missing-data structures.
16. Multiplicity and Multiple Endpoint Interpretation
The registry contains 42 posted outcome measures and 24 posted statistical analyses, including the primary endpoint and numerous secondary and other pre-specified analyses.
The ClinicalTrials.gov record identifies the hypothesis type as superiority, but they do not provide an explicit multiplicity-adjustment procedure, alpha hierarchy, or endpoint-testing sequence. Accordingly, individual secondary P-values should be described as reported statistical results rather than automatically treated as independently confirmatory findings.
17. Nonparametric and Parametric Analyses Together
CHEST-1 provides a useful example of using different statistical approaches for the same clinical endpoint. The primary 6MWD endpoint has a reported ANCOVA analysis, a Wilcoxon / Mann-Whitney analysis, and a Shapiro-Wilk residual diagnostic.
| Analysis | Purpose | Primary 6MWD result |
|---|---|---|
| ANCOVA | Estimate the adjusted between-group difference in final 6MWD. | Mean difference 45.69; 95% CI 24.74–66.63; P < 0.0001. |
| Wilcoxon / Mann-Whitney | Provide a nonparametric comparison. | P < 0.0001. |
| Shapiro-Wilk | Assess normality of ANCOVA residuals. | P = 0.0001. |
These analyses should not be treated as three independent treatment tests. They have different purposes. ANCOVA provides the reported effect estimate; Wilcoxon provides a nonparametric comparison; and Shapiro-Wilk is a model diagnostic.
18. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary ANCOVA estimated a 45.69-meter between-group difference in final 6MWD, with a 95% CI of 24.74 to 66.63 meters and P < 0.0001 under the reported superiority analysis.
Clinical interpretation
The statistical result concerns a measure of functional exercise capacity. Determining the practical importance of the estimated difference requires clinical context beyond the P-value alone, and the ClinicalTrials.gov record does not provide a clinical threshold for interpretation.
This distinction is important. Statistical evidence answers whether the observed data are compatible with a specified null model. Clinical interpretation asks what the magnitude and direction of the observed difference mean in practice. The two questions overlap but are not interchangeable.
19. Limitations
- Registry-level reporting: this analysis is restricted to the ClinicalTrials.gov record and does not reconstruct information that is absent from the registry record.
- No baseline table reported: the data identify enrollment and treatment arms but do not provide a detailed baseline demographic or clinical-characteristics table.
- No arm-level randomized sample sizes reported: the overall enrollment is 262, while the serious-adverse-event analysis reports safety denominators of 173 and 88. Those safety denominators should not be assumed to equal randomized arm sizes.
- Limited safety detail: the ClinicalTrials.gov record reports serious adverse events by arm but not the full adverse-event profile.
- Multiplicity: the ClinicalTrials.gov record does not identify a formal multiplicity-adjustment strategy for the large collection of outcome measures and analyses.
- Missing-data assumptions: several endpoint analyses use last-observation or worst-value rules. These assumptions can affect the estimated treatment effect.
- Effect estimates are not available for every analysis: several secondary Wilcoxon analyses provide P-values without corresponding confidence intervals or effect estimates.
- Endpoint heterogeneity: continuous, binary, and time-to-event outcomes require different statistical interpretations and should not be combined into a single summary statistic.
- Model diagnostics: Shapiro-Wilk testing evaluates residual normality but does not establish that every assumption underlying ANCOVA is satisfied.
20. Why This Trial Matters Statistically
CHEST-1 is a useful statistical teaching case because the ClinicalTrials.gov record combines randomized treatment comparison, covariate-adjusted continuous-outcome analysis, nonparametric testing, time-to-event analysis, stratification, explicit missing-data rules, and multiple endpoint types.
| Concept | How it appears in CHEST-1 |
|---|---|
| Randomization | Randomized allocation in a two-arm parallel phase 3 trial. |
| Blinding | Triple masking. |
| ITT analysis | Primary efficacy analysis uses the registry-defined ITT population. |
| ANCOVA | Primary 6MWD analysis and several continuous secondary endpoints. |
| Covariate adjustment | Baseline 6MWD included as a covariate for the primary analysis. |
| Stratified analysis | Region included in ANCOVA and used to stratify the clinical-worsening analysis. |
| Wilcoxon / Mann-Whitney | Supporting analyses for the primary and several secondary endpoints. |
| Log-rank test | Clinical-worsening analysis. |
| Risk difference | Effect measure for the clinical-worsening analysis. |
| Missing data | Endpoint-specific worst-value and last-observation rules. |
| Confidence intervals | Reported for the primary ANCOVA and several other mean-difference analyses. |
| P-values | Reported across parametric, nonparametric, time-to-event, and diagnostic analyses. |
21. A Practical Reading of the CHEST-1 Statistical Record
A disciplined reading of the record begins with the endpoint hierarchy. The registered primary endpoint is 6MWD change from baseline to week 16. Its principal reported analysis is ANCOVA, producing a mean difference of 45.69 meters with a 95% confidence interval of 24.74 to 66.63 meters and P < 0.0001.
The next question is how that estimate was produced. The analysis was performed in the ITT population, baseline 6MWD was used as a covariate, and treatment group and region were included as main effects. Missing values were handled using prespecified worst-value or last-observation rules.
Only after understanding that primary analysis should the supporting analyses be considered. The Wilcoxon result provides a nonparametric analysis, while Shapiro-Wilk evaluates the normality of the ANCOVA residuals. These are complementary pieces of the statistical record rather than interchangeable treatment-effect estimates.
Finally, the secondary endpoints should be interpreted individually. PVR, NT-proBNP, EQ-5D, LPH, WHO Functional Class, Borg CR 10, clinical worsening, PAPmean, and cardiac index do not measure the same construct. Their effect measures and methods differ, and the ClinicalTrials.gov record does not specify a multiplicity procedure that would allow every small P-value to be interpreted as independently confirmatory.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: NCT00855465 — CHEST-1.
- PubMed: PMID 23883377.
- PubMed: PMID 23883383.
- PubMed: PMID 27067478.
- PubMed: PMID 27326239.
- PubMed: PMID 27263466.
Continue through the Clinical Biostats statistical library
Explore the statistical methods that appear across randomized clinical trials, from covariate adjustment and nonparametric tests to missing-data analysis and time-to-event methods.
25. Record Summary
CHEST-1 provides a compact but statistically rich example of randomized clinical-trial analysis. The registered primary endpoint was change in 6-minute walking distance from baseline to week 16, and its reported ANCOVA estimated a 45.69-meter mean difference between riociguat and placebo, with a 95% CI of 24.74 to 66.63 meters and P < 0.0001.
The statistical story extends beyond that single number. The analysis used an ITT population, adjusted for baseline 6MWD and region, and specified explicit procedures for missing observations. Supporting Wilcoxon / Mann-Whitney analyses and Shapiro-Wilk residual diagnostics were also reported. Secondary analyses used ANCOVA, nonparametric methods, and a region-stratified log-rank analysis, producing a mixture of mean differences, risk differences, and P-values.
The most informative interpretation therefore combines the effect estimate, confidence interval, P-value, analysis population, covariate structure, missing-data assumptions, and endpoint-specific statistical method. That framework helps distinguish what the registry actually estimates from broader conclusions that cannot be established from a P-value alone.