This page separates reported trial results from statistical interpretation. Numerical results are restricted to the information reported in the ClinicalTrials.gov record. The registry provides the official trial record.
1. Trial at a Glance
VENTURE was a randomized, parallel-group, triple-masked phase 3 trial evaluating dupilumab in patients with severe steroid-dependent asthma. The principal statistical question was whether dupilumab increased the reduction in oral corticosteroid (OCS) dose that could be achieved at Week 24 while maintaining asthma control.
| Feature | VENTURE |
|---|---|
| Trial name | VENTURE |
| ClinicalTrials.gov identifier | NCT02528214 |
| Brief title | Evaluation of Dupilumab in Patients With Severe Steroid Dependent Asthma |
| Therapeutic area | Pulmonology |
| Condition | Asthma |
| Phase | Phase 3 |
| Status | Completed |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 210 |
| Lead sponsor | Sanofi |
| Start | October 15, 2015 |
| Primary completion | September 20, 2017 |
2. Clinical Question
The primary clinical question was whether treatment with dupilumab 300 mg every 2 weeks, compared with placebo every 2 weeks, produced a greater reduction from baseline in OCS dose at Week 24 while maintaining asthma control.
Population
Patients with severe steroid-dependent asthma enrolled in the phase 3 VENTURE trial.
Intervention
Dupilumab 300 mg every 2 weeks, alongside the background therapies specified in the registry.
Comparator
Placebo every 2 weeks, alongside the background therapies specified in the registry.
Primary question
Does dupilumab produce a greater percentage reduction in OCS dose at Week 24 while maintaining asthma control?
3. Trial Design
Dupilumab
- Dupilumab 300 mg every 2 weeks
- Background oral corticosteroid therapy
- Inhaled corticosteroid therapy
- Albuterol/Salbutamol or Levalbuterol/Levosalbutamol as registered interventions
Placebo
- Placebo every 2 weeks
- Background oral corticosteroid therapy
- Inhaled corticosteroid therapy
- Albuterol/Salbutamol or Levalbuterol/Levosalbutamol as registered interventions
4. Endpoints
| Endpoint | Time frame | Registry definition / analysis |
|---|---|---|
| Percentage Reduction From Baseline in Oral Corticosteroids (OCS) Dose at Week 24 While Maintaining Asthma Control | Baseline, Week 24 | Percentage reduction calculated as (optimized OCS dose at baseline − final OCS dose at Week 24) / optimized OCS dose at baseline × 100. Result presented as LS Mean (SE) percentage reduction from baseline derived from an ANCOVA model with missing data multiply imputed. |
| Supplementary Presentation of Primary Outcome Measure Data: Median Percentage Reduction From Baseline in Oral Corticosteroids Dose at Week 24 While Maintaining Asthma Control | Baseline, Week 24 | Supplementary presentation of the primary outcome measure, reported as median (inter-quartile range). |
| Percentage of Participants Achieving ≥50% Reduction in Oral Corticosteroids Dose at Week 24 While Maintaining Asthma Control | Week 24 | Binary outcome analyzed using logistic regression in the ITT population. |
| Percentage of Participants Achieving a Reduction in Oral Corticosteroids Dose to <5 mg/Day at Week 24 While Maintaining Asthma Control | Week 24 | Binary outcome analyzed using logistic regression in the ITT population. |
| Percentage of Participants Achieving Maximum Possible Reduction in Oral Corticosteroids Dose Per Protocol at Week 24 While Maintaining Asthma Control | Week 24 | Binary outcome analyzed using logistic regression in the ITT population; per-protocol reduction criterion is part of the endpoint definition. |
| Percentage of Participants Who No Longer Required Oral Corticosteroids Dose at Week 24 While Maintaining Asthma Control | Week 24 | Binary outcome analyzed using logistic regression in the ITT population with baseline OCS dose less than or equal to 30 mg/day. |
The registry lists 2 primary endpoints, with the second described as a supplementary presentation of the primary outcome measure. The posted formal primary analysis is for the LS-mean percentage reduction endpoint; the supplementary median presentation does not have a separate formal statistical comparison in the ClinicalTrials.gov record.
5. Statistical Methodology
Primary ANCOVA analysis
The primary endpoint was analyzed using analysis of covariance (ANCOVA). The response variable was percentage reduction in OCS dose at Week 24. The model included treatment group, baseline eosinophil level, optimized OCS dose at baseline, and region as covariates.
The treatment coefficient represents the adjusted difference between randomized treatment groups after accounting for the covariates specified in the registry analysis.
Logistic regression for binary secondary endpoints
Four posted secondary analyses evaluated whether participants achieved clinically defined OCS-reduction milestones. Each was modeled using logistic regression, with treatment group and specified baseline covariates included in the model.
For these analyses, the exponentiated treatment coefficient is expressed as an odds ratio comparing dupilumab with placebo.
Intention-to-treat analysis
The primary and posted secondary efficacy analyses were conducted in the intent-to-treat (ITT) population. The registry describes this as the randomized population analyzed according to the treatment group allocated by randomization, regardless of whether treatment was received as planned.
Covariate adjustment
Covariate adjustment is especially relevant here because the primary endpoint is a continuous percentage reduction rather than a simple unadjusted treatment-group difference. The ANCOVA model incorporates baseline OCS dose, baseline eosinophil level, and region in the treatment comparison. The secondary logistic models similarly adjust for prespecified baseline characteristics.
Missing data and multiple imputation
The registry states that missing data were imputed using a pattern mixture model by multiple imputation. This means the analysis did not simply discard participants with missing Week 24 information. Instead, missing outcomes were handled through an explicit imputation framework before the statistical comparison was made.
6. Multiplicity and Hierarchical Testing
The registry analysis notes state that a hierarchical testing procedure was used to control the overall type I error. Testing proceeded sequentially in the order the outcome measures were reported and continued when the previous outcome measure was statistically significant at a two-sided 0.05 level.
Why hierarchy matters
Multiple hypothesis tests can increase the chance of a false-positive conclusion if treated as unrelated tests. A prespecified hierarchy provides an ordered framework for determining which subsequent hypotheses can be formally tested.
What the p-values mean here
The reported p-values should be interpreted in the context of this sequential testing strategy rather than as isolated numbers detached from the endpoint hierarchy.
The registry specifically states that the hierarchical procedure continued for the secondary endpoints only when the previous endpoint was statistically significant. This is a different concept from simply declaring every posted p-value independently significant or nonsignificant.
7. Results
Primary endpoint: percentage reduction in OCS dose
The primary formal analysis compared the placebo q2w group with the dupilumab 300 mg q2w group using ANCOVA in the ITT population. The treatment comparison was expressed as a least-squares mean difference.
Adjusted difference in percentage reduction from baseline
95% CI: 15.81–40.67 · P < 0.0001
Effect measure: Least Square Mean Difference · Two-sided 95% CI · Superiority
| Primary analysis component | Reported result |
|---|---|
| Groups compared | Placebo q2w vs Dupilumab 300 mg q2w |
| Analysis population | Intent-to-treat population |
| Method | ANCOVA |
| Effect measure | Least Square Mean Difference |
| Estimate | 28.24 |
| 95% confidence interval | 15.81–40.67 |
| P-value | < 0.0001 |
| Hypothesis type | Superiority |
| Missing-data approach | Pattern mixture model by multiple imputation |
The estimated LS mean difference of 28.24 indicates that the model-estimated percentage reduction in OCS dose was 28.24 percentage points greater for the dupilumab group than for the placebo group, after adjustment for baseline eosinophil level, optimized OCS dose at baseline, and region.
This is an adjusted mean difference, not an odds ratio, relative risk, or probability that an individual patient will achieve a particular OCS reduction. It also does not mean that every patient experienced a 28.24-percentage-point additional reduction.
The 95% CI of 15.81–40.67 describes uncertainty around the estimated adjusted treatment difference under the specified ANCOVA and missing-data framework. It is not an interval containing 95% of individual patient treatment effects.
The p-value < 0.0001 provides evidence against the null hypothesis used for this superiority comparison under the specified statistical framework. A p-value does not measure the magnitude of the treatment effect and does not indicate the probability that the null hypothesis is true.
Because the endpoint was analyzed with missing data multiply imputed under a pattern mixture model, the estimate also depends on the assumptions used to construct the imputed values. The treatment comparison should therefore be understood as a model-based ITT estimate rather than a simple calculation from complete cases alone.
Supplementary presentation of the primary endpoint
The registry also reports a supplementary presentation of the same primary outcome measure as a median percentage reduction from baseline at Week 24 while maintaining asthma control. The registry identifies this as a supplementary presentation and does not provide a separate formal statistical analysis for the median in the ClinicalTrials.gov record.
8. Secondary Endpoint Results
At least 50% reduction in OCS dose
The first posted secondary analysis evaluated the percentage of participants achieving a ≥50% reduction in OCS dose at Week 24 while maintaining asthma control. The ITT population was analyzed using logistic regression.
Odds ratio for achieving ≥50% OCS reduction
95% CI: 2.06–7.67 · P < 0.0001
Two-sided 95% CI · Superiority
An odds ratio of 3.98 means the estimated odds of achieving the ≥50% OCS-reduction criterion were 3.98 times as high in the dupilumab group as in the placebo group, after adjustment for the covariates specified in the logistic model.
The odds ratio is not a probability ratio. An OR of 3.98 cannot be read as saying that the probability of success was 3.98 times higher without knowing the underlying probabilities.
The 95% CI of 2.06–7.67 quantifies uncertainty around the estimated odds ratio. It does not describe the range of responses experienced by individual participants.
The p-value < 0.0001 addresses evidence against the null comparison under the specified model and testing framework. It does not indicate how large or clinically important the observed effect is.
Reduction of OCS dose to <5 mg/day
A second posted secondary endpoint evaluated the percentage of participants achieving a reduction in OCS dose to <5 mg/day at Week 24 while maintaining asthma control.
Odds ratio for achieving OCS dose <5 mg/day
95% CI: 2.39–8.39 · P < 0.0001
Two-sided 95% CI · Superiority
The estimated odds ratio of 4.48 indicates higher estimated odds of reaching the registered <5 mg/day OCS criterion in the dupilumab group than in the placebo group, based on the adjusted logistic regression model.
The estimate should not be interpreted as a 348% increase in probability. Odds and probabilities are related mathematically, but they are not interchangeable measures.
The 95% CI of 2.39–8.39 indicates statistical uncertainty around the odds-ratio estimate. The interval remains above 1, consistent with a positive treatment-group contrast under the model.
The reported p-value < 0.0001 is evidence against the null hypothesis for this sequentially tested endpoint; it is not itself an effect-size measure.
Maximum possible OCS reduction per protocol
The third posted secondary analysis evaluated the percentage of participants achieving the maximum possible reduction in OCS dose per protocol at Week 24 while maintaining asthma control.
Odds ratio for maximum possible OCS reduction
95% CI: 1.4–4.73 · P = 0.0024
Two-sided 95% CI · Superiority
An odds ratio of 2.57 indicates that the estimated odds of achieving the maximum possible protocol-defined OCS reduction were 2.57 times those in the placebo group under the adjusted logistic regression model.
The confidence interval, 1.4–4.73, is wider than would be implied by a highly precise estimate and therefore matters when interpreting the magnitude of the effect. It describes uncertainty around the model-based odds ratio rather than individual patient outcomes.
The p-value of 0.0024 provides evidence against the null comparison under the stated testing framework. It should not be used as a substitute for the effect estimate or its confidence interval.
The registry notes that this endpoint was part of the hierarchical testing sequence. Consequently, its inferential interpretation depends on the preceding endpoint having met the criterion required for continuation of the hierarchy.
No longer requiring OCS
The fourth posted secondary analysis evaluated the percentage of participants who no longer required an OCS dose at Week 24 while maintaining asthma control. The analysis population was the ITT population with baseline OCS dose less than or equal to 30 mg/day.
Odds ratio for no longer requiring OCS
95% CI: 1.47–5.1 · P = 0.0015
Two-sided 95% CI · Superiority
The odds ratio of 2.74 indicates that the estimated odds of no longer requiring OCS at Week 24 were 2.74 times as high in the dupilumab group as in the placebo group among the specified analysis population.
The restriction to participants with baseline OCS dose less than or equal to 30 mg/day is important. This is not the same analysis population as an unrestricted ITT analysis, so its estimate should not be generalized automatically to every randomized participant.
The 95% CI of 1.47–5.1 expresses uncertainty around the adjusted odds-ratio estimate. The p-value of 0.0015 addresses statistical evidence under the specified logistic model and hierarchical testing framework, not effect magnitude or individual probability of success.
9. Results Summary
| Endpoint | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Primary: percentage reduction from baseline in OCS dose at Week 24 while maintaining asthma control | ANCOVA | LS mean difference 28.24 | 15.81–40.67 | < 0.0001 |
| ≥50% reduction in OCS dose | Logistic regression | OR 3.98 | 2.06–7.67 | < 0.0001 |
| OCS dose <5 mg/day | Logistic regression | OR 4.48 | 2.39–8.39 | < 0.0001 |
| Maximum possible OCS reduction per protocol | Logistic regression | OR 2.57 | 1.4–4.73 | 0.0024 |
| No longer requiring OCS, baseline OCS ≤30 mg/day | Logistic regression | OR 2.74 | 1.47–5.1 | 0.0015 |
The statistical pattern is internally coherent: the primary continuous endpoint and the posted binary secondary endpoints all estimate treatment contrasts favoring the dupilumab group under their respective models. However, the effect measures are not directly interchangeable. The ANCOVA result is an adjusted difference in mean percentage reduction, whereas the secondary results are adjusted odds ratios for achieving specific binary milestones.
10. How to Read the Primary Effect Estimate
The primary estimate of 28.24 is a difference in adjusted mean percentage reduction. If the two treatment groups had identical covariate profiles under the fitted model, the estimated difference in their mean percentage reductions would be 28.24 percentage points.
The 95% CI of 15.81–40.67 provides a range of values compatible with the estimated treatment difference under the model and sampling assumptions. A confidence interval is therefore more informative about precision than the p-value alone.
The p-value < 0.0001 quantifies how incompatible the observed result would be with the null hypothesis under the specified statistical model. It does not tell us the probability that dupilumab works, the probability that the effect is clinically important, or the percentage of patients who benefit.
The estimate is not simply the difference between two raw group means. ANCOVA uses the specified baseline covariates to produce a model-adjusted treatment comparison. This can improve precision and account for prognostic baseline information, provided the model is appropriate.
11. How to Read the Odds Ratios
| Odds ratio | Registered endpoint | Statistical meaning |
|---|---|---|
| 3.98 | ≥50% OCS reduction | Estimated odds of meeting the binary criterion were 3.98 times those in the placebo group. |
| 4.48 | OCS dose <5 mg/day | Estimated odds of meeting the <5 mg/day criterion were 4.48 times those in the placebo group. |
| 2.57 | Maximum possible OCS reduction | Estimated odds of meeting the protocol-defined maximum reduction criterion were 2.57 times those in the placebo group. |
| 2.74 | No longer requiring OCS | Estimated odds of no longer requiring OCS were 2.74 times those in the placebo group within the specified analysis population. |
These four odds ratios describe different binary outcomes. They should not be combined into a single summary statistic or treated as interchangeable measures of treatment effect. Each one answers a separate question about OCS reduction.
12. Statistical Methods Explained
Why was ANCOVA used for the primary endpoint?
The primary endpoint is a continuous percentage reduction from baseline. ANCOVA provides a framework for comparing treatment groups while adjusting for prespecified baseline covariates. Here, the registry specifies baseline eosinophil level, optimized OCS dose at baseline, and region as covariates in addition to treatment group.
What does an LS mean difference of 28.24 mean?
It is the model-adjusted difference between the treatment-group least-squares means for percentage reduction in OCS dose. A value of 28.24 means the estimated adjusted percentage reduction was 28.24 percentage points greater in the dupilumab group than in the placebo group. It is not a percentage probability and not a relative risk.
Why use logistic regression for the secondary endpoints?
Each of these endpoints converts the clinical outcome into a binary status: for example, whether a participant achieved at least a 50% OCS reduction. Logistic regression models the probability of that binary outcome and expresses the treatment contrast as an odds ratio.
What does an odds ratio of 3.98 mean?
An odds ratio of 3.98 means the estimated odds of achieving the specified ≥50% OCS reduction were 3.98 times higher in the dupilumab group than in the placebo group, after the model's covariate adjustment. It does not mean that the probability of success was 398% or that the probability itself was multiplied by 3.98.
Why does missing-data imputation matter?
The primary endpoint requires a Week 24 OCS dose. When that information is missing, the observed dataset alone cannot directly provide the participant's unobserved outcome. The registry specifies a pattern mixture model using multiple imputation, so the reported analysis incorporates modeled values for missing observations. Different assumptions about missing outcomes can produce different estimates.
Why does ITT analysis matter?
The ITT principle keeps participants associated with the randomized treatment group for efficacy analysis. This preserves the treatment comparison created by randomization and avoids redefining treatment groups according to what participants subsequently received.
Why does the hierarchical testing procedure matter?
The registry states that testing proceeded sequentially and continued only when the preceding outcome was statistically significant at a two-sided 0.05 level. This controls the overall type I error across the prespecified sequence more effectively than treating every endpoint as an independent hypothesis test.
13. Secondary Analysis: Per-Protocol Criterion Within an ITT Analysis
One of the secondary endpoints is explicitly defined as achievement of the maximum possible reduction in OCS dose per protocol. Importantly, the registry states that the statistical analysis itself was performed in the ITT population.
This distinction is statistically important. A per-protocol analysis population can differ from an endpoint whose definition references protocol-based treatment rules. Here, the registry's analysis population remains ITT while the endpoint itself incorporates a protocol-defined maximum reduction.
14. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm. These are presented as affected participants over participants at risk.
| Safety measure | Placebo q2w | Dupilumab 300 mg q2w |
|---|---|---|
| Serious adverse events | 6/107 | 9/103 |
The ClinicalTrials.gov record does not provide a formal statistical comparison for serious adverse events. Accordingly, the counts are descriptive rather than a reported inferential test of safety differences.
15. Trial Timeline
Trial start
The VENTURE phase 3 study began.
Randomized treatment period
The study evaluated dupilumab 300 mg q2w versus placebo q2w in the randomized parallel-group design.
Primary completion
The registry lists September 20, 2017 as the primary completion date.
Results posted
The registry is marked completed and contains 14 posted outcome measures and 5 posted statistical analyses.
16. What the Trial's Statistical Evidence Shows
Continuous endpoint
The primary ANCOVA estimated a 28.24 percentage-point adjusted difference in OCS dose reduction, with a 95% CI of 15.81–40.67 and p < 0.0001.
Binary milestones
Four secondary endpoints used logistic regression and reported odds ratios of 3.98, 4.48, 2.57, and 2.74.
Missing data
The primary analysis used a pattern mixture model by multiple imputation rather than a complete-case-only approach.
Multiplicity
The registry describes hierarchical testing, with sequential testing continuing when the preceding outcome was statistically significant at a two-sided 0.05 level.
17. Limitations
- Registry-level reporting: this analysis is restricted to the ClinicalTrials.gov record and does not add numerical results from publications or other sources.
- Limited primary endpoint detail: the registry provides the formal LS mean difference and confidence interval for the primary endpoint but does not provide all underlying group-level descriptive statistics in the ClinicalTrials.gov record.
- Supplementary median presentation: the median percentage reduction is identified as a supplementary presentation, but the ClinicalTrials.gov record does not provide a separate formal comparison for it.
- Model dependence: the primary treatment estimate depends on the specified ANCOVA covariates and the assumptions of that model.
- Missing-data assumptions: multiple imputation with a pattern mixture model requires assumptions about the distribution of unobserved outcomes.
- Odds ratios are not risk ratios: the secondary logistic-regression estimates cannot be interpreted as simple probability ratios without additional information about baseline event probabilities.
- Endpoint-specific populations: the analysis for no longer requiring OCS is restricted to participants with baseline OCS dose less than or equal to 30 mg/day.
- Per-protocol terminology: the maximum-possible-reduction endpoint contains a protocol-based criterion but was analyzed in the ITT population according to the ClinicalTrials.gov record.
- Safety inference: serious adverse-event counts are reported by arm, but no formal statistical comparison is reported in the ClinicalTrials.gov record used here.
- Unreported methods: the ClinicalTrials.gov record does not establish stratification factors, crossover procedures, Bayesian methods, or an interim-analysis strategy, so those features are not inferred here.
18. Why This Trial Matters Statistically
VENTURE is a useful teaching case because it illustrates how a clinical trial can use different statistical models for closely related versions of the same clinical question. The primary endpoint is continuous and therefore uses ANCOVA, while several secondary endpoints convert OCS reduction into clinically defined binary milestones and use logistic regression.
| Statistical concept | How it appears in VENTURE |
|---|---|
| Randomization | Randomized parallel-group phase 3 design with 210 enrolled participants. |
| Blinding | Triple masking was specified. |
| Intention-to-treat | Primary and posted secondary efficacy analyses were conducted in the ITT population. |
| ANCOVA | Used for the primary percentage reduction in OCS dose endpoint. |
| Covariate adjustment | Primary model included baseline eosinophil level, optimized OCS dose at baseline, and region. |
| Logistic regression | Used for binary OCS-reduction milestones. |
| Odds ratio | Reported treatment effect for the binary secondary endpoints. |
| Multiple imputation | Missing data handled using a pattern mixture model by multiple imputation. |
| Multiplicity | Hierarchical testing used for sequential endpoint testing. |
| Per-protocol concept | One secondary endpoint used a protocol-defined maximum possible OCS reduction while retaining an ITT analysis population. |
| Confidence intervals | 95% two-sided intervals reported for the primary LS mean difference and secondary odds ratios. |
| Safety analysis | Serious adverse events reported descriptively by randomized arm. |
19. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary ANCOVA estimated a positive adjusted difference in percentage reduction in OCS dose, with a 95% CI of 15.81–40.67 and p < 0.0001. The posted binary secondary analyses also produced odds ratios above 1 with their reported two-sided confidence intervals and p-values.
Clinical interpretation
The endpoint framework focuses on reducing exposure to oral corticosteroids while maintaining asthma control. The trial therefore evaluates both the magnitude of continuous OCS reduction and the proportion of participants reaching specific OCS-reduction milestones.
The statistical evidence should be read at the level of the endpoint being analyzed. A 28.24 percentage-point adjusted difference, an odds ratio of 3.98, and an odds ratio of 4.48 are not different ways of expressing the same numerical quantity. They represent different estimands arising from different outcome definitions and models.
20. A Deeper Look at Missing Data
The primary endpoint requires a final OCS dose at Week 24. Missing Week 24 information can therefore affect the calculation of percentage reduction. The registry explicitly states that missing data were handled using a pattern mixture model with multiple imputation.
The statistical analysis then modeled this percentage reduction using ANCOVA after the specified missing-data procedure.
This is important because the final analysis is not merely a calculation performed independently for every participant with complete data. The analysis combines the endpoint definition, the covariate-adjusted ANCOVA model, and the missing-data assumptions into a single inferential framework.
For educational purposes, this also illustrates why two analysts can use the same randomized trial and obtain different results if they make materially different assumptions about missing outcomes. The appropriate comparison is therefore not only the numerical estimate but also the estimand, analysis population, model, and missing-data strategy.
21. Confidence Intervals Across the Reported Analyses
| Effect | Estimate | 95% CI | What the interval describes |
|---|---|---|---|
| Primary LS mean difference | 28.24 | 15.81–40.67 | Uncertainty around the adjusted difference in percentage reduction. |
| ≥50% OCS reduction OR | 3.98 | 2.06–7.67 | Uncertainty around the adjusted odds ratio. |
| OCS dose <5 mg/day OR | 4.48 | 2.39–8.39 | Uncertainty around the adjusted odds ratio. |
| Maximum possible OCS reduction OR | 2.57 | 1.4–4.73 | Uncertainty around the adjusted odds ratio. |
| No longer requiring OCS OR | 2.74 | 1.47–5.1 | Uncertainty around the adjusted odds ratio in the specified analysis population. |
Confidence intervals are especially useful for comparing the precision of estimates. The width of an interval reflects statistical uncertainty, whereas the location of the interval describes the range of effect estimates compatible with the analysis framework.
22. The Role of the Analysis Population
The registry's repeated use of the ITT population is a major component of the inferential structure. Participants remain classified according to their randomized assignment for efficacy analyses. This maintains the comparison generated by randomization.
The OCS-withdrawal endpoint illustrates why population definitions must always be read alongside effect estimates. Although the analysis remains ITT, it is explicitly restricted to participants whose baseline OCS dose was less than or equal to 30 mg/day. Its odds ratio therefore answers a narrower question than the primary endpoint.
23. Why the Endpoint Hierarchy Matters
The registry describes a sequence in which testing continued only when the previous endpoint was statistically significant at a two-sided 0.05 level. This creates a chain of inferential claims rather than four unrelated secondary tests.
The hierarchy does not make the endpoints equivalent. Rather, it determines how evidence can propagate through the prespecified testing sequence. This is a central distinction between multiplicity control and simply counting how many p-values are below a threshold.
24. Important Statistical Takeaways
Estimate first
The treatment effect should be understood from its effect measure and magnitude before looking at the p-value.
Precision matters
The confidence interval communicates how precisely the trial estimated the treatment contrast.
Model matters
ANCOVA and logistic regression answer different statistical formulations of the treatment question.
Missingness matters
The primary estimate incorporates a pattern mixture multiple-imputation strategy for missing data.
25. Related Tutorials
Learn more about the methods used in this trial:
26. Related Calculators
27. Sources
- ClinicalTrials.gov: VENTURE, NCT02528214.
- PubMed: PMID 29782224.
- PubMed: PMID 36509407.
- PubMed: PMID 36356712.
- PubMed: PMID 35809937.
- PubMed: PMID 35636689.
Continue through the Clinical Biostats statistical library
Explore the statistical methods behind randomized clinical trials, from covariate-adjusted models and logistic regression to confidence intervals and missing-data methods.
28. Record Summary
VENTURE provides a compact example of how randomized clinical-trial evidence can be analyzed at several levels. The primary endpoint uses ANCOVA to estimate an adjusted difference in percentage reduction in OCS dose, while binary OCS-reduction milestones use logistic regression and odds ratios. The ClinicalTrials.gov record also explicitly document ITT analysis, covariate adjustment, pattern mixture multiple imputation, and a hierarchical testing procedure.
The primary formal estimate was an LS mean difference of 28.24, with a two-sided 95% CI of 15.81–40.67 and p < 0.0001. The four posted secondary binary analyses reported odds ratios of 3.98, 4.48, 2.57, and 2.74, respectively, with their corresponding confidence intervals and p-values. Serious adverse events were reported descriptively as 6/107 for placebo q2w and 9/103 for dupilumab 300 mg q2w.