This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
EASE was a randomized, double-blind, parallel phase 3 trial evaluating fosaprepitant versus aprepitant for prevention of chemotherapy-induced nausea and vomiting. The primary efficacy question was whether fosaprepitant was at least as effective as aprepitant for complete response during the overall 120-hour period following initiation of cisplatin chemotherapy.
| Feature | EASE |
|---|---|
| Phase | Phase 3 |
| Condition | Chemotherapy-Induced Nausea and Vomiting (CINV) |
| Primary purpose | Prevention |
| Design | Randomized, double-blind, parallel |
| Enrollment | 2322 |
| Arms | 2 |
| Primary endpoint | Complete response, defined as no vomiting and no use of rescue therapy, overall in the 120 hours following initiation of cisplatin chemotherapy |
| Primary effect measure | Risk difference |
| Hypothesis | Non-inferiority or equivalence |
| Study period | Start: 2008-02 · Primary completion: 2009-06 |
| Lead sponsor | Merck Sharp & Dohme LLC |
| ClinicalTrials.gov | NCT00619359 |
2. Clinical Question
The central question was whether a single-dose fosaprepitant regimen provided a level of complete response that was not clinically meaningfully worse than the aprepitant regimen during the overall 120 hours following initiation of cisplatin chemotherapy.
Population
Patients enrolled in the phase 3 prevention trial for chemotherapy-induced nausea and vomiting who received study therapy, cisplatin chemotherapy, and at least one post-treatment efficacy assessment for the full analysis set.
Intervention
Fosaprepitant dimeglumine, with dexamethasone and ondansetron represented among the registered interventions.
Comparator
Aprepitant, with dexamethasone and ondansetron represented among the registered interventions.
Primary question
Is fosaprepitant at least as effective as aprepitant for complete response during the overall 120 hours following initiation of cisplatin chemotherapy?
3. Trial Design
The trial enrolled 2322 participants and used two randomized, double-blind parallel treatment groups. The statistical structure is fundamentally different from a superiority trial: the prespecified question was framed around non-inferiority, so the interpretation depends on whether the confidence interval for the treatment difference is compatible with a clinically acceptable loss of efficacy.
4. Treatment Groups and Analysis Population
Fosaprepitant regimen
- Fosaprepitant dimeglumine
- Dexamethasone
- Ondansetron
- Evaluated against aprepitant for the registered efficacy outcomes
Aprepitant regimen
- Aprepitant
- Dexamethasone
- Ondansetron
- Comparator regimen for the registered efficacy outcomes
Full Analysis Set
The ClinicalTrials.gov record states that the FAS patient population was used for all efficacy evaluations. Patients were included if they received at least one dose of study therapy, received cisplatin chemotherapy, and had at least one post-treatment efficacy assessment.
5. Endpoints
| Endpoint | Registry definition | Time frame | Role |
|---|---|---|---|
| Complete Response | A Complete Response (no Vomiting and no Use of Rescue Therapy) Overall. The number of patients who reported No Vomiting and No Use of Rescue Therapy. | Overall (in the 120 hours following initiation of cisplatin chemotherapy) | Primary |
| Complete Response in the Delayed Phase | A Complete Response (no Vomiting and no Use of Rescue Therapy) in the Delayed Phase. | 25 to 120 hours following initiation of cisplatin | Secondary |
| No Vomiting Overall | No Vomiting Overall | 120 hours following initiation of cisplatin chemotherapy | Secondary |
All three posted statistical analyses use a binary endpoint and report the treatment effect as a risk difference comparing fosaprepitant with aprepitant. The primary endpoint combines two clinically distinct components—absence of vomiting and absence of rescue therapy—so a patient must satisfy both conditions to be counted as a complete responder.
6. Statistical Methodology
Risk difference as the effect measure
The principal effect measure was the risk difference (RD), defined here as the response probability in the fosaprepitant group minus the response probability in the aprepitant group.
An RD of 0 represents identical response probabilities. A negative RD means the observed response probability was lower in the fosaprepitant group; a positive RD means it was higher.
Non-inferiority framework
The primary analysis was explicitly framed as non-inferiority or equivalence. The registry states that if the confidence interval for the difference in response rates, defined as fosaprepitant minus aprepitant, had a lower limit at least as high as 7 percentage points, fosaprepitant was considered at least as effective as aprepitant for complete response in the overall phase.
The same registry analysis states that the confidence interval was calculated using the methodology of Miettinen and Nurminen and that the study had 90% power to detect non-inferiority for the primary outcome.
Binary outcome analysis
Complete response and no vomiting are binary outcomes: each participant is classified according to whether the specified event occurred. For this type of endpoint, a treatment comparison can be expressed as a difference in proportions, with a confidence interval quantifying uncertainty around that difference.
The registry does not identify a separate normalized hypothesis-test method beyond the reported Miettinen-Nurminen confidence-interval methodology. The most important statistical information that is explicitly reported is therefore the risk difference, its 95% confidence interval, the non-inferiority criterion, and the analysis population.
Intention-to-treat principles
The statistical analysis record explicitly identifies intention-to-treat analysis as an analysis concept, while defining the efficacy population operationally as the FAS. In randomized trials, the intention-to-treat principle generally preserves the treatment assignment created by randomization. Here, the registry's FAS definition additionally requires exposure to study therapy, receipt of cisplatin, and at least one post-treatment efficacy assessment.
7. Primary Result: Overall Complete Response
The primary endpoint was complete response—no vomiting and no use of rescue therapy—during the overall 120 hours following initiation of cisplatin chemotherapy.
Risk difference: Fosaprepitant − Aprepitant
95% CI: −4.1 to 3.3 percentage points
Non-inferiority criterion: lower confidence-limit requirement ≥7 percentage points
| Primary endpoint | Fosaprepitant vs Aprepitant | Effect measure | 95% CI |
|---|---|---|---|
| Complete response overall | −0.4 | Risk difference | −4.1 to 3.3 |
The estimated risk difference of −0.4 percentage points means that the observed complete-response proportion in the fosaprepitant group was 0.4 percentage points lower than in the aprepitant group in the FAS analysis.
The estimate does not mean that fosaprepitant caused a 0.4-percentage-point reduction for every patient, nor does it describe an individual patient's probability of vomiting. It is a group-level difference in the probability of meeting the complete-response definition.
The 95% confidence interval of −4.1 to 3.3 percentage points describes the statistical uncertainty around the estimated difference. It spans both modestly lower and modestly higher response probabilities for fosaprepitant.
The registry's non-inferiority rule is unusually important here. The prespecified requirement was that the lower confidence limit be at least 7 percentage points. Because the reported interval extends from −4.1 to 3.3, its lower limit is below that stated threshold. Therefore, the reported interval itself does not satisfy the registry's stated non-inferiority criterion.
No p-value is reported in the posted primary statistical analysis. In any event, a p-value would not measure the size or clinical importance of the treatment difference; the risk difference and its confidence interval are the quantities that directly describe the comparative response probabilities.
8. Secondary Results
Delayed-Phase Complete Response
The first secondary endpoint evaluated complete response during the delayed phase, defined as 25 to 120 hours following initiation of cisplatin.
Risk difference: Fosaprepitant − Aprepitant
95% CI: −3.5 to 3.7 percentage points
Non-inferiority criterion: lower confidence-limit requirement ≥7.3 percentage points
The estimated risk difference of 0.1 percentage points indicates an observed complete-response probability 0.1 percentage points higher with fosaprepitant than with aprepitant in the analyzed delayed-phase population.
The 95% CI extends from −3.5 to 3.7 percentage points, so the data are compatible with a modestly lower or modestly higher response probability for fosaprepitant. The interval is centered very close to zero, but its lower limit is below the registry's stated non-inferiority requirement of 7.3 percentage points.
The analysis excluded one patient in the aprepitant group who had no delayed-phase data. This is relevant because the analysis population for this endpoint is therefore slightly different from the corresponding overall efficacy population.
No Vomiting Overall
The second secondary endpoint evaluated whether participants experienced no vomiting during the 120 hours following initiation of cisplatin chemotherapy.
Risk difference: Fosaprepitant − Aprepitant
95% CI: −5.3 to 2.0 percentage points
Non-inferiority criterion: lower confidence-limit requirement ≥8.2 percentage points
The estimated risk difference of −1.7 percentage points means that the observed probability of no vomiting was 1.7 percentage points lower in the fosaprepitant group than in the aprepitant group.
The 95% CI of −5.3 to 2.0 percentage points indicates that the uncertainty interval includes both a modestly lower and a modestly higher no-vomiting probability for fosaprepitant.
The endpoint is narrower than complete response because complete response requires both no vomiting and no rescue therapy, whereas this endpoint concerns vomiting alone. Consequently, these two binary outcomes should not be treated as interchangeable measures of efficacy.
The registry states that two patients in the aprepitant group had no vomiting data and were excluded from this analysis. That missingness distinction matters when comparing the denominator structure across the posted analyses.
| Endpoint | Role | Risk difference | 95% CI | Non-inferiority lower-limit criterion |
|---|---|---|---|---|
| Complete response overall | Primary | −0.4 | −4.1 to 3.3 | ≥7 percentage points |
| Complete response, delayed phase | Secondary | 0.1 | −3.5 to 3.7 | ≥7.3 percentage points |
| No vomiting overall | Secondary | −1.7 | −5.3 to 2.0 | ≥8.2 percentage points |
9. Understanding the Non-Inferiority Logic
The most important statistical feature of EASE is that the treatment comparison is not interpreted simply by asking whether the risk difference is close to zero. The registry specifies a directional confidence-interval rule.
The registry states that meeting this criterion would support the conclusion that fosaprepitant was at least as effective as aprepitant for overall complete response.
Because the effect measure is defined as fosaprepitant minus aprepitant, negative values favor the comparator numerically for the endpoint, while positive values favor fosaprepitant. The non-inferiority boundary therefore establishes how much lower the fosaprepitant response could plausibly be before the trial would no longer support non-inferiority.
Why zero is not the boundary
A superiority analysis commonly asks whether a confidence interval excludes zero. A non-inferiority analysis instead asks whether the confidence interval stays above a prespecified negative-performance boundary.
Why the margin matters
The non-inferiority margin defines the largest clinically acceptable loss of efficacy that can still be compatible with the trial's non-inferiority objective.
Why the CI is central
The entire confidence interval is relevant. Looking only at the point estimate can obscure the uncertainty that determines whether a non-inferiority criterion is met.
Why p-values are secondary here
A conventional p-value for equality is not the decision rule specified in the posted analysis. The confidence interval is directly linked to the non-inferiority margin.
10. Statistical Methods Explained
What is a risk difference?
A risk difference compares two event probabilities by subtraction. If 70% of one group and 68% of another experience the endpoint, the risk difference is 2 percentage points. Unlike a relative risk or odds ratio, the RD is expressed directly on the probability scale.
Why is risk difference appropriate for complete response?
Complete response is a binary endpoint: each participant either meets the definition or does not. A risk difference therefore provides a direct description of the absolute separation in response probabilities between the two treatment groups.
Why was a non-inferiority margin used?
The purpose of the comparison was not simply to establish that the two treatments had exactly identical response probabilities. The registry specifies a clinically acceptable boundary and asks whether the uncertainty around the treatment difference is sufficiently favorable to rule out an unacceptable loss of efficacy.
What does the Miettinen-Nurminen method contribute?
The registry identifies the Miettinen and Nurminen methodology for calculating the confidence interval for the difference in response rates. This is a method designed for confidence intervals around differences in binomial proportions and is particularly relevant when the analysis is built around a risk-difference estimand.
Why does the FAS definition matter?
The FAS was used for all efficacy evaluations and required treatment exposure, receipt of cisplatin, and at least one post-treatment efficacy assessment. The analysis therefore does not simply compare all 2322 enrolled participants regardless of whether they contributed efficacy data.
Why should delayed-phase and overall complete response be separated?
The overall endpoint covers the 120 hours following initiation of cisplatin, whereas the delayed-phase endpoint specifically covers hours 25 through 120. They overlap in time but answer different endpoint questions, so their risk differences and confidence intervals should be interpreted separately.
11. Analysis Population and Missing Data
The registry provides specific information about missing efficacy data for the two secondary analyses.
| Endpoint | Analysis population detail |
|---|---|
| Primary complete response overall | FAS: at least one dose of study therapy, cisplatin chemotherapy, and at least one post-treatment efficacy assessment |
| Delayed-phase complete response | FAS; 1 patient in the aprepitant group had no delayed-phase data and was not included |
| No vomiting overall | FAS; 2 patients in the aprepitant group had no vomiting data and were excluded |
These exclusions illustrate an important feature of binary clinical-trial endpoints: the denominator used for a particular analysis can differ when outcome information is unavailable. Even when the number of excluded participants is small, the analysis definition should remain explicit because the estimand depends on which participants contribute observed outcome data.
12. Safety
The registry reports serious adverse events by treatment group using affected participants over participants at risk.
| Safety measure | Fosaprepitant | Aprepitant |
|---|---|---|
| Serious adverse events | 148/1143 | 157/1169 |
Fosaprepitant
148 participants with serious adverse events among 1143 participants at risk.
Aprepitant
157 participants with serious adverse events among 1169 participants at risk.
Serious adverse events are a safety endpoint and should be interpreted independently from the efficacy endpoint of complete response. The reported affected/at-risk counts provide the basic event-frequency information, but the registry data do not provide a posted risk-difference confidence interval or hypothesis test for serious adverse events.
13. Why This Trial Matters Statistically
EASE is a useful teaching case because it illustrates a form of randomized-trial inference in which the statistical question is defined by a non-inferiority margin rather than by a conventional superiority comparison.
| Concept | How it appears in EASE |
|---|---|
| Randomization | Randomized allocation to two treatment groups |
| Blinding | Double-blind design |
| Parallel design | Two parallel treatment groups |
| Binary endpoint | Complete response and no vomiting |
| Risk difference | Primary effect measure for the posted efficacy analyses |
| Confidence interval | 95% intervals quantify uncertainty around each risk difference |
| Non-inferiority | Interpretation uses prespecified lower confidence-limit criteria |
| FAS / ITT concept | FAS used for all efficacy evaluations; intention-to-treat identified as an analysis concept |
| Missing outcome data | Specific exclusions were reported for delayed-phase and no-vomiting analyses |
| Miettinen-Nurminen method | Used for the reported confidence intervals for response-rate differences |
The trial therefore demonstrates why the same numerical estimate can have different implications depending on the hypothesis being tested. A risk difference near zero might look reassuring in an informal comparison, but non-inferiority requires the confidence interval to be evaluated against the prespecified margin.
14. Interpreting the Primary Confidence Interval
The primary risk difference was −0.4 percentage points. This is close to zero on the absolute probability scale, indicating little observed separation between the treatment groups for the primary complete-response endpoint.
The 95% confidence interval ranged from −4.1 to 3.3 percentage points. The interval therefore covers a range that includes modestly lower response with fosaprepitant and modestly higher response with fosaprepitant.
The registry's stated criterion required the lower limit to be at least 7 percentage points. The lower end of the reported interval is below that boundary, so the confidence interval does not satisfy the stated criterion.
The risk difference does not quantify an individual patient's risk of vomiting, does not represent a relative risk reduction, and does not mean that every patient experiences the same treatment difference. It is a population-level comparison of binary response probabilities.
15. Comparing the Three Posted Efficacy Analyses
The three posted analyses tell a consistent descriptive story in the sense that all three point estimates are close to zero: −0.4 for overall complete response, 0.1 for delayed-phase complete response, and −1.7 for no vomiting overall. Their confidence intervals, however, extend across zero and must be interpreted in relation to their respective non-inferiority criteria.
| Endpoint | Estimate | 95% CI | What the interval contains |
|---|---|---|---|
| Complete response overall | −0.4 | −4.1 to 3.3 | Lower and higher response probabilities relative to aprepitant |
| Complete response, delayed phase | 0.1 | −3.5 to 3.7 | Lower and higher response probabilities relative to aprepitant |
| No vomiting overall | −1.7 | −5.3 to 2.0 | Lower and higher no-vomiting probabilities relative to aprepitant |
The three endpoints are related but not identical. Complete response incorporates both vomiting and rescue therapy, while no vomiting isolates the vomiting component. The delayed-phase complete-response endpoint focuses specifically on hours 25 through 120.
16. Limitations and Interpretation Issues
- Non-inferiority interpretation: the central conclusion depends on the prespecified confidence-interval criterion and its corresponding margin rather than on a conventional superiority p-value.
- Analysis population: efficacy analyses used the FAS, which required treatment exposure, cisplatin treatment, and at least one post-treatment efficacy assessment.
- Missing outcome information: one aprepitant participant was excluded from the delayed-phase analysis because of missing delayed-phase data, and two aprepitant participants were excluded from the no-vomiting analysis because vomiting data were unavailable.
- Method reporting: the registry identifies the Miettinen-Nurminen methodology for the confidence intervals but does not report a separate normalized statistical-test method for the posted analyses.
- Endpoint construction: complete response is a composite binary outcome requiring both no vomiting and no rescue therapy; it should not be interpreted as equivalent to the no-vomiting endpoint.
- Safety information: serious adverse events are reported as affected participants over participants at risk, but the posted information does not provide corresponding comparative confidence intervals or hypothesis tests.
- Non-inferiority assumptions: interpretation of a non-inferiority trial depends not only on the numerical margin but also on the credibility of the randomized comparison, endpoint measurement, adherence, and other design features. The registry information available here does not provide enough detail to evaluate every such assumption.
17. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
18. Related Statistical Calculators
19. Sources
- ClinicalTrials.gov: NCT00619359, the official registry record for EASE.
- PubMed: PubMed record for the EASE publication.
Continue through the Clinical Biostats knowledge graph
Connect the trial's design and statistical concepts to deeper tutorials and practical statistical tools.
20. Record Summary
EASE provides a clear example of how a randomized phase 3 trial can be structured around a non-inferiority question. The primary endpoint was a binary complete-response measure over the 120 hours following initiation of cisplatin chemotherapy, and the treatment effect was expressed as the risk difference between fosaprepitant and aprepitant. The reported primary estimate was −0.4, with a 95% confidence interval of −4.1 to 3.3. The registry specified a lower confidence-limit criterion of 7 percentage points for the primary non-inferiority assessment.
The secondary analyses used the same general risk-difference framework for delayed-phase complete response and no vomiting overall. Their estimates were 0.1 and −1.7, respectively, with confidence intervals of −3.5 to 3.7 and −5.3 to 2.0. The registry also reports serious adverse events of 148/1143 with fosaprepitant and 157/1169 with aprepitant.