This page separates reported trial results from statistical interpretation. Numerical trial results are restricted to the ClinicalTrials.gov record. The registry provides one formal statistical analysis for the primary binary endpoint.
1. Trial at a Glance
ROSE was a randomized, parallel-group, phase 3 trial evaluating early neuromuscular blockade versus no routine early neuromuscular blockade in acute respiratory distress syndrome. The registered primary endpoint was hospital mortality to day 90, defined as the percentage of subjects alive at study day 90.
| Feature | ROSE |
|---|---|
| Trial name | Reevaluation Of Systemic Early Neuromuscular Blockade |
| Phase | Phase 3 |
| Condition | Acute Respiratory Distress Syndrome |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 1008 |
| Arms | 2 |
| Intervention | Cisatracurium Besylate |
| Primary endpoint | Hospital Mortality to Day 90 |
| Primary endpoint type | Binary |
| Hypothesis type | Superiority |
| Statistical method reported | Wald test for the difference of two proportions |
| Effect measure | Risk Difference (RD) |
| ClinicalTrials.gov | NCT02509078 |
| Lead sponsor | Massachusetts General Hospital |
2. Clinical Question
The primary statistical question was whether early neuromuscular blockade differed from no routine early neuromuscular blockade with respect to hospital mortality to day 90 in patients with acute respiratory distress syndrome.
Population
Participants enrolled in the ROSE phase 3 trial with the condition acute respiratory distress syndrome.
Intervention
Early Neuromuscular Blockade (NMB), with cisatracurium besylate listed as the intervention.
Comparator
Control: No Routine Early NMB.
Primary question
Does early neuromuscular blockade produce a difference in hospital mortality to day 90 compared with no routine early neuromuscular blockade?
3. Trial Design
Early Neuromuscular Blockade (NMB)
- Early neuromuscular blockade
- Cisatracurium besylate listed as the intervention
- Serious adverse events reported for this arm: 28/501 affected/at risk
Control: No Routine Early NMB
- No routine early neuromuscular blockade
- Serious adverse events reported for this arm: 21/505 affected/at risk
These design features matter because the statistical comparison is anchored to randomized treatment assignment, while the absence of masking means that participants and study personnel were not blinded to assignment. The ClinicalTrials.gov record does not report a stratification scheme, crossover procedure, interim analysis, missing-data method, imputation strategy, Bayesian method, or multiplicity adjustment.
4. Endpoints
| Endpoint | Registry definition | Time frame | Type |
|---|---|---|---|
| Hospital Mortality to Day 90 | The percentage of subjects alive at study day 90. Those subjects discharged home prior to day 90 were counted as alive at day 90. | 90 days after randomization | Binary |
The registry identifies one primary endpoint and describes it as binary. This is important statistically: the primary comparison is therefore based on proportions of participants meeting the endpoint definition rather than on a continuous measurement or a time-to-event hazard ratio.
5. Statistical Methodology
Primary endpoint: difference between two proportions
The registry reports a Wald test for the difference of two proportions, a normal-approximation (z) test. The effect measure is the risk difference (RD).
For a binary outcome, the risk difference compares the observed or estimated event proportions directly on an absolute probability scale. Here the registry reports the effect as a percentage-point difference for hospital mortality to day 90.
The key distinction is between an absolute difference and a relative measure. A risk difference of −0.3 means that the estimated mortality proportion in the early NMB group was 0.3 percentage points lower than the corresponding proportion in the control group, based on the direction of the reported estimate. It does not mean that mortality was reduced by 0.3 relative units, nor does it describe an individual patient's probability of benefit.
Wald testing
The Wald framework standardizes an estimated difference by its estimated standard error. Conceptually, the resulting statistic has the form:
For a superiority comparison of two proportions, the null hypothesis is that the treatment groups have no difference in the target proportion. The reported P-value then quantifies the compatibility of the observed result with that null model under the specified test.
Confidence interval
The registry reports a two-sided 95% confidence interval for the risk difference. The interval ranges from −6.4 to 5.9. This interval describes statistical uncertainty around the estimated difference; it is not a range containing 95% of individual patient outcomes.
Superiority hypothesis
The registry classifies the hypothesis type as superiority. In a superiority framework, the question is whether the randomized groups differ rather than whether one treatment stays within a prespecified non-inferiority margin. The ClinicalTrials.gov record does not report a non-inferiority margin because one was not part of the stated statistical analysis profile.
Why randomization matters
Randomization provides the foundation for interpreting a difference between treatment groups as a randomized treatment comparison rather than simply an association between treatment received and outcome. The statistical test does not itself create the benefits of randomization; it quantifies uncertainty around the observed group difference under the statistical model.
6. Primary Result: Hospital Mortality to Day 90
The only formal statistical analysis posted in the ClinicalTrials.gov record is for the primary endpoint, hospital mortality to day 90. The comparison was Early Neuromuscular Blockade (NMB) versus Control: No Routine Early NMB.
Risk difference
95% CI: −6.4 to 5.9 · P = 0.93
Two-sided Wald test; estimate reported as a percentage.
| Primary endpoint | Early NMB vs Control | Statistical method | 95% CI | P-value |
|---|---|---|---|---|
| Hospital Mortality to Day 90 | Risk Difference = −0.3 | Wald / z-test | −6.4 to 5.9 | 0.93 |
The reported risk difference of −0.3 indicates an estimated absolute difference of 0.3 percentage points in the direction of lower hospital mortality for the Early NMB group relative to the control group. Because the registry states that the estimate is a percentage, the estimate is interpreted on the percentage-point scale.
The estimate does not mean that early neuromuscular blockade reduced an individual's probability of death by exactly 0.3%, and it does not establish that every patient experienced a treatment effect of that size. A group-level risk difference summarizes the contrast between the two randomized groups.
The 95% confidence interval from −6.4 to 5.9 is substantially wider than the point estimate and crosses zero. In statistical terms, the interval includes both negative and positive differences, so the data are compatible with a range of possible underlying group differences under the confidence-interval framework.
The P-value of 0.93 does not measure the size of the treatment effect. A P-value addresses evidence against the null hypothesis under the specified statistical model; it should not be read as a probability that the treatment works, a probability that the null hypothesis is true, or a measure of clinical importance.
The registry identifies this as a superiority analysis. Because the confidence interval includes zero and the reported P-value is 0.93, the formal statistical result does not provide evidence of a difference under this reported superiority test. That statement concerns the statistical comparison; it is not a claim that the two interventions are mathematically identical or that all possible clinically relevant effects have been excluded.
Reading the result on an absolute scale
7. What the Risk Difference Does — and Does Not — Mean
What it measures
The risk difference measures the absolute difference between the outcome proportions in the two groups. Its scale is percentage points when the underlying proportions are expressed as percentages.
What −0.3 means
The reported estimate places the Early NMB group 0.3 percentage points below the control group for the primary outcome, based on the direction specified by the reported comparison.
What the CI adds
The 95% CI of −6.4 to 5.9 shows that the point estimate is uncertain and that the interval includes zero as well as both negative and positive differences.
What the P-value adds
The P-value of 0.93 summarizes the compatibility of the observed test statistic with the null hypothesis under the reported Wald framework. It is not an effect-size measure.
This distinction between effect estimate, confidence interval, and P-value is central to reading randomized clinical trials. The effect estimate answers "how large is the observed difference?" The confidence interval addresses "how precisely is that difference estimated?" The P-value addresses a different question: "how much evidence against the null hypothesis is reported by this test statistic?"
8. Serious Adverse Events
The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected participants over participants at risk. These figures should be kept separate from the primary mortality analysis because they describe a safety outcome rather than the registered primary efficacy endpoint.
| Safety measure | Affected / at risk |
|---|---|
| Early Neuromuscular Blockade (NMB) | 28/501 |
| Control: No Routine Early NMB | 21/505 |
The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or P-value for these serious adverse-event counts. Accordingly, the figures are presented descriptively rather than as evidence of a statistically tested between-group difference.
The denominators matter. "28/501" and "21/505" identify the number affected and the number at risk for the reported safety measure. Without a formal analysis reported in the ClinicalTrials.gov record, the appropriate interpretation is descriptive rather than inferential.
Safety outcomes also require careful attention to the definition of an event, exposure time, severity, attribution, and competing events. The ClinicalTrials.gov record does not provide those additional analytical details, so they should not be inferred.
9. Statistical Methods Explained
Why was a Wald / z-test used?
The registry identifies the primary endpoint as binary and reports a Wald test for the difference of two proportions. A Wald framework provides a standardized test of an estimated group difference using its estimated standard error. Here, the parameter of interest is the risk difference rather than a ratio.
What does a risk difference of −0.3 mean?
Because the registry states that the estimate is a percentage, −0.3 represents a 0.3-percentage-point difference in the direction of lower mortality for Early NMB relative to the control group. It is an absolute measure, not a relative percentage reduction.
Why is zero important for the confidence interval?
For a risk difference, zero represents no difference between the two group proportions. The reported 95% CI, −6.4 to 5.9, crosses zero. That means the interval includes values corresponding to a lower outcome proportion in Early NMB, no difference, and a higher outcome proportion in Early NMB.
Why doesn't the P-value measure effect size?
The P-value depends on both the estimated difference and its statistical uncertainty. A very small P-value can occur for a modest effect when the estimate is precise, while a clinically noticeable estimate can have a larger P-value when uncertainty is substantial. Therefore the risk difference and confidence interval should be examined directly rather than using the P-value as a proxy for magnitude.
Why does randomization matter to the interpretation?
Randomization establishes the intended comparison between the intervention and control groups before outcomes are observed. It helps balance measured and unmeasured prognostic factors in expectation. The Wald test then addresses statistical uncertainty around the resulting group comparison; it does not substitute for randomization.
Why is this not a hazard-ratio analysis?
The registered primary endpoint is binary at study day 90, and the posted analysis uses a difference of two proportions. A hazard ratio would answer a different question about relative instantaneous event rates over time and would require time-to-event information. The ClinicalTrials.gov record does not identify a hazard-ratio analysis for the primary endpoint.
10. Analysis Population and Censoring Considerations
The ClinicalTrials.gov record identifies the primary endpoint as a binary measurement at study day 90. They do not provide a separate named analysis population definition, nor do they report a formal censoring rule for the primary statistical analysis beyond the endpoint definition itself.
The definition contains an important outcome-handling rule: subjects discharged home prior to day 90 were counted as alive at day 90. This makes the endpoint a status assessment at a fixed time point rather than a conventional time-to-death analysis in which discharge might simply be another follow-up state.
11. Multiplicity, Interim Analysis, and Other Design Issues
The registry profile posted on ClinicalTrials.gov for ROSE contains one registered primary endpoint and one posted statistical analysis. It does not report a multiplicity adjustment, an interim analysis, an alpha-spending strategy, or a separate endpoint hierarchy.
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Primary endpoint count | 1 |
| Primary endpoint | Hospital Mortality to Day 90 |
| Formal statistical analyses posted | 1 |
| Multiplicity adjustment | Not reported in the ClinicalTrials.gov record |
| Interim analysis | Not reported in the ClinicalTrials.gov record |
| Non-inferiority margin | Not applicable to the reported superiority hypothesis |
| Bayesian methods | Not reported in the ClinicalTrials.gov record |
| Crossover | Not reported in the ClinicalTrials.gov record |
| Missing-data / imputation method | Not reported in the ClinicalTrials.gov record |
| Stratification factors | Not reported in the ClinicalTrials.gov record |
This absence of information should not be interpreted as proof that a feature was absent from the full protocol or statistical analysis plan. It means only that the ClinicalTrials.gov record does not report the feature. For this page, that distinction is important because adding an unreported method would turn an educational reconstruction into an unsupported assertion about the trial.
12. Why the Confidence Interval Is Central
The primary estimate is −0.3, but the confidence interval extends from −6.4 to 5.9. The interval therefore provides considerably more information than the point estimate alone.
The point estimate is close to zero, while the confidence interval spans both sides of zero. The interval represents uncertainty about the estimated population difference under the statistical framework; it does not state that every value in the interval is equally likely.
A useful statistical reading therefore starts with the effect measure and its scale, then examines the confidence interval, and only then considers the P-value. The P-value of 0.93 is consistent with the confidence interval crossing the null value, but it should not replace the interval as the primary description of precision.
13. Superiority Testing in This Trial
The registry identifies the hypothesis as superiority. That designation determines the type of inferential question being asked. The analysis is designed to test for a difference rather than to establish that the intervention is no worse than control within a prespecified margin.
Superiority question
Is there evidence of a difference between Early NMB and Control for hospital mortality to day 90?
Null value for RD
For a risk difference, the null value is zero: no difference between the two group proportions.
Reported result
The estimate is −0.3, with a two-sided 95% CI of −6.4 to 5.9 and P = 0.93.
Not a non-inferiority analysis
No non-inferiority margin is reported because the registry classifies the hypothesis as superiority.
This distinction prevents a common interpretive error. A result that does not demonstrate superiority should not automatically be described as proof of equivalence or non-inferiority. Those are different statistical claims requiring different hypotheses and, for non-inferiority, a prespecified margin.
14. Interpreting Statistical and Clinical Evidence Separately
Statistical interpretation
The posted primary analysis estimates a risk difference of −0.3, with a two-sided 95% confidence interval from −6.4 to 5.9 and a P-value of 0.93 from the reported Wald test.
Clinical interpretation
The ClinicalTrials.gov record does not provide a separate clinical-effect threshold or additional outcome results from which to determine the practical importance of particular values within the confidence interval.
Keeping these statements separate is important. Statistical significance is a property of an inference under a specified model and testing framework. Clinical importance depends on the magnitude of an effect, the outcome, the clinical context, patient priorities, treatment burden, and other evidence. The ClinicalTrials.gov record supports the statistical comparison but do not justify adding an independent clinical threshold.
15. Limitations
- Limited posted statistical detail: the ClinicalTrials.gov record contains one formal statistical analysis for the primary endpoint. They do not provide a full statistical analysis plan.
- Underlying arm-specific mortality values are not reported: the registry analysis gives the risk difference, confidence interval, and P-value, but the ClinicalTrials.gov record does not include the two arm-specific primary mortality proportions.
- No additional efficacy estimates: the ClinicalTrials.gov record does not report median survival, hazard ratios, subgroup estimates, longitudinal survival curves, or other efficacy estimates for this page.
- No reported stratification factors: the ClinicalTrials.gov record does not identify stratification variables, so no stratified analysis is described.
- No reported missing-data strategy: the ClinicalTrials.gov record does not specify imputation or sensitivity analyses for missing primary-endpoint information.
- No reported interim-analysis framework: the registry-reported statistical profile does not identify an interim look, alpha spending, or group-sequential boundary.
- No reported crossover: crossover is not identified in the ClinicalTrials.gov record and is therefore not incorporated into interpretation.
- No masking: the registry describes the study as having no masking. This is a design characteristic that differs from a blinded randomized comparison and should be considered when interpreting outcomes that may be susceptible to knowledge of assignment.
- Binary endpoint: hospital mortality to day 90 reduces the primary outcome to a fixed-time status measure. It does not describe the exact timing of deaths before day 90.
- Safety analysis is descriptive here: serious adverse-event counts are provided by arm, but the ClinicalTrials.gov record does not contain a formal statistical comparison for that safety endpoint.
16. Why This Trial Matters Statistically
ROSE is a useful teaching case because its primary statistical analysis is comparatively direct: a randomized parallel-group phase 3 comparison, a binary endpoint at a fixed follow-up time, a risk-difference effect measure, and a Wald test. That simplicity makes the trial particularly useful for understanding how an absolute treatment effect, its confidence interval, and its P-value answer different questions.
| Concept | How it appears in ROSE |
|---|---|
| Randomization | The trial is registered as randomized. |
| Parallel design | The trial has 2 parallel arms. |
| Binary endpoint | Hospital Mortality to Day 90 is registered as binary. |
| Fixed-time outcome | The endpoint is assessed 90 days after randomization. |
| Risk difference | The treatment effect is reported as Risk Difference (RD). |
| Wald / z-test | The registry reports a Wald test for the difference of two proportions. |
| Confidence interval | The primary RD has a two-sided 95% CI of −6.4 to 5.9. |
| P-value | The reported two-sided P-value is 0.93. |
| Superiority testing | The registry identifies the hypothesis type as superiority. |
| Safety comparison | Serious adverse events are reported descriptively by arm. |
17. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
18. Related Statistical Calculators
19. Primary Sources
- ClinicalTrials.gov: ROSE — NCT02509078.
- Linked PubMed record: PMID 31112383.
- Linked PubMed record: PMID 27788018.
- Linked PubMed record: PMID 27779896.
Continue through the Clinical Biostats statistical pathway
Use the trial's primary analysis to explore confidence intervals, P-values, randomization, risk differences, and Wald / z-tests in greater depth.
20. Record Summary
ROSE provides a clear example of a randomized clinical-trial analysis built around a binary primary endpoint. The registered primary outcome was hospital mortality to day 90, defined as the percentage of subjects alive at study day 90, with subjects discharged home before day 90 counted as alive. The posted analysis used a Wald test for the difference of two proportions and reported a risk difference of −0.3, a two-sided 95% confidence interval of −6.4 to 5.9, and a P-value of 0.93.
The most important statistical lesson is that these three quantities should be interpreted together. The risk difference describes the estimated absolute contrast, the confidence interval describes uncertainty around that estimate, and the P-value addresses the evidence against the null hypothesis under the reported Wald testing framework. The interval crossing zero and the P-value of 0.93 indicate that the posted superiority analysis did not demonstrate a statistically detectable difference in the primary endpoint under that test.