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
MELODY was a completed, randomized, quadruple-masked, parallel-group phase 3 prevention trial evaluating MEDI8897 versus placebo in healthy late preterm and term infants. The primary efficacy endpoint was medically attended respiratory syncytial virus lower respiratory tract infection through 150 days post dose in the primary cohort.
| Feature | MELODY |
|---|---|
| Trial | MELODY |
| ClinicalTrials.gov identifier | NCT03979313 |
| Phase | Phase 3 |
| Condition | Respiratory Syncytial Virus Infections |
| Primary purpose | Prevention |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Enrollment | 3012 |
| Interventions | MEDI8897 and placebo |
| Primary endpoint type | Binary |
| Primary endpoint analyses posted | 1 |
| Statistical analyses posted | 4 |
| Primary statistical method | Poisson regression with robust variance |
| Primary effect measure | Relative Risk Reduction (RRR) |
| Hypothesis type | Superiority |
| Lead sponsor | AstraZeneca |
2. Clinical Question
The central clinical question was whether MEDI8897 reduced the occurrence of medically attended respiratory syncytial virus lower respiratory tract infection through 150 days post dose compared with placebo in healthy late preterm and term infants.
Population
Healthy late preterm and term infants included in the MELODY prevention trial for respiratory syncytial virus infection.
Intervention
MEDI8897, evaluated for prevention of medically attended RSV lower respiratory tract infection.
Comparator
Placebo.
Primary question
Does MEDI8897 reduce the risk of medically attended RSV lower respiratory tract infection through 150 days post dose?
3. Trial Design
MEDI8897
- MEDI8897 was evaluated for prevention of medically attended RSV lower respiratory tract infection.
- Primary efficacy assessment occurred through 150 days post dose.
- Primary and additional analyses used the ITT population.
Placebo
- Placebo served as the randomized comparator.
- The primary efficacy comparison was MEDI8897 versus placebo.
- The primary hypothesis was superiority.
4. Endpoints
| Endpoint role | Registered endpoint | Time frame | Population |
|---|---|---|---|
| Primary | Number of Participants With MA RSV LRTI Through 150 Days Post Dose (Primary Cohort) | Through 150 Days Post Dose | Primary Cohort; N=1490 Participants |
| Secondary | Number of Participants With MA RSV LRTI With Hospitalisation Through 150 Days Post Dose (Primary Cohort) | Through 150 Days Post Dose | ITT population / Primary Cohort |
| Other pre-specified | Number of Participants With MA RSV LRTI Through 150 Days Post Dose (All Subjects) | Through 150 Days Post Dose | ITT population / All Subjects |
| Other pre-specified | Number of Participants With MA RSV LRTI With Hospitalisation Through 150 Days Post Dose (All Subjects) | Through 150 Days Post Dose | ITT population / All Subjects |
The primary endpoint is binary because each participant either experienced the specified medically attended RSV lower respiratory tract infection outcome within the registered follow-up window or did not. The registry reports four statistical analyses covering the primary endpoint, a secondary hospitalization endpoint, and corresponding analyses in all subjects.
5. Analysis Populations and Statistical Structure
The posted analyses identify the ITT population as the analysis population, with the primary analysis restricted to the primary cohort. The primary cohort contained 1490 participants for the registered primary endpoint.
| Analysis | Population | Role | Method |
|---|---|---|---|
| Primary MA RSV LRTI | ITT / Primary Cohort | Primary | Poisson regression with robust variance |
| MA RSV LRTI with hospitalization | ITT / Primary Cohort | Secondary | Poisson regression with robust variance |
| MA RSV LRTI, all subjects | ITT / All Subjects | Other pre-specified | Poisson regression with robust variance |
| MA RSV LRTI with hospitalization, all subjects | ITT / All Subjects | Other pre-specified | Poisson regression with robust variance |
The registry analysis text indicates that the Poisson regression included robust variance, incorporated age at randomization as a covariate, and obtained the reported confidence interval and p-value after multiple imputation. Stratification factors were included through the analysis specification.
6. Statistical Methodology
Poisson regression for a binary clinical outcome
The posted analyses used a Poisson model with robust variance to estimate the treatment comparison for the binary participant-level endpoints. Although Poisson regression is often introduced for counts or rates, a modified Poisson approach with robust variance can be used for binary outcomes when the objective is direct estimation of a relative risk.
For a binary outcome, the fitted model can be paired with robust variance estimation to obtain inference for a relative risk. The treatment coefficient represents the modeled relative difference between treatment groups on the risk scale.
Relative risk reduction
The registry reports the treatment effect as Relative Risk Reduction (RRR). Conceptually, relative risk reduction expresses the proportional reduction in risk associated with treatment relative to the comparator:
The underlying relative risk compares the probability of the outcome in the MEDI8897 group with the probability in the placebo group. RRR then expresses the corresponding relative reduction.
Covariate adjustment
The posted analysis notes that age at randomization was included as a covariate. Covariate adjustment can improve precision and account for prespecified design variables while retaining the randomized treatment comparison.
Multiple imputation
The registry analysis notes that the confidence intervals and p-values were obtained after multiple imputation. Multiple imputation is a framework for handling missing outcome or covariate information by creating multiple plausible completed datasets, analyzing each dataset, and combining the resulting estimates so that uncertainty associated with missing information is reflected in the inference.
Robust variance
The use of robust variance is important because the ordinary Poisson variance assumptions are not necessarily appropriate when a Poisson model is used for binary outcomes. Robust variance provides a variance estimate appropriate to the treatment-effect inference even though the outcome is not a conventional Poisson count.
Superiority hypothesis
The registered hypothesis type was superiority. The statistical question was therefore whether the observed risk of medically attended RSV lower respiratory tract infection was lower with MEDI8897 than with placebo, rather than whether MEDI8897 was merely no worse than placebo within a prespecified non-inferiority margin.
7. Primary Result
Medically Attended RSV LRTI Through 150 Days Post Dose
The primary endpoint was the number of participants with medically attended RSV lower respiratory tract infection through 150 days post dose in the primary cohort. The posted analysis used the ITT population and primary cohort, with a Poisson model and robust variance.
Relative risk reduction
95% CI: 49.63–87.12 · P < 0.0001
Poisson regression with robust variance, including age at randomization as a covariate, after multiple imputation.
| Primary endpoint | MEDI8897 vs placebo | 95% CI | P-value |
|---|---|---|---|
| MA RSV LRTI through 150 days post dose, primary cohort | RRR 74.53% | 49.63–87.12 | <0.0001 |
The reported 74.53% relative risk reduction means that the modeled relative risk of the primary outcome was substantially lower in the MEDI8897 group than in the placebo group over the specified 150-day period. This is a relative measure: it does not state the absolute probability of infection in either treatment group, the absolute risk difference, or the number of infections prevented per fixed number of treated infants.
The 95% confidence interval of 49.63–87.12 describes uncertainty around the estimated relative risk reduction under the specified statistical model and analysis framework. The interval remains on the reduction side of the effect scale, which is consistent with the reported superiority analysis.
The P < 0.0001 result addresses the statistical evidence against the null hypothesis under the prespecified analysis. It does not measure the magnitude of the treatment effect. Effect magnitude is described by the RRR estimate, while the confidence interval describes its precision.
The interpretation also depends on the analysis population, the covariate adjustment, the multiple-imputation procedure, and the robust variance calculation. Because the endpoint is assessed over a fixed 150-day window rather than as a time-to-event outcome, a hazard-ratio interpretation is not appropriate for this result.
8. Secondary Endpoint Result
Medically Attended RSV LRTI With Hospitalisation Through 150 Days Post Dose
The registered secondary endpoint evaluated the number of participants with medically attended RSV lower respiratory tract infection with hospitalization through 150 days post dose in the primary cohort.
Relative risk reduction
95% CI: -8.57–86.80 · P = 0.0708
Poisson regression with robust variance, including age at randomization as a covariate, after multiple imputation.
| Secondary endpoint | MEDI8897 vs placebo | 95% CI | P-value |
|---|---|---|---|
| MA RSV LRTI with hospitalization through 150 days post dose, primary cohort | RRR 62.15% | -8.57–86.80 | 0.0708 |
The point estimate of 62.15% relative risk reduction indicates a lower estimated risk in the MEDI8897 group than in the placebo group for this hospitalization endpoint. However, the confidence interval extends from -8.57 to 86.80. Because the interval includes zero relative risk reduction, the estimate is compatible with a range extending from a modest increase in risk through a substantial reduction in risk.
The P = 0.0708 value is above the conventional 0.05 threshold, but the p-value should not be interpreted as the probability that the treatment works or does not work. It is a measure of compatibility between the observed result and the null hypothesis under the specified model and testing framework.
The wide confidence interval also shows that this endpoint was estimated less precisely than the primary endpoint. That difference in precision is important: the point estimate alone can make an effect look definitive when the interval demonstrates substantially greater uncertainty.
Because the analysis was specified as a superiority test, the relevant inferential question is whether the data provide sufficient evidence for superiority under the registered statistical framework. The reported estimate should therefore be considered together with its confidence interval and p-value rather than interpreted from the point estimate alone.
9. Additional Pre-Specified Results in All Subjects
The registry also reports two pre-specified analyses in the all-subjects ITT population. These analyses extend the same outcome concepts beyond the primary cohort.
| Endpoint | Population | Effect | 95% CI | P-value |
|---|---|---|---|---|
| MA RSV LRTI through 150 days post dose | ITT / All Subjects | RRR 76.36% | 62.27–85.18 | <0.0001 |
| MA RSV LRTI with hospitalization through 150 days post dose | ITT / All Subjects | RRR 76.84% | 49.36–89.41 | 0.0002 |
All-subjects medically attended RSV LRTI
Relative risk reduction
95% CI: 62.27–85.18 · P < 0.0001
The all-subjects analysis estimates a 76.36% relative risk reduction for medically attended RSV lower respiratory tract infection through 150 days post dose. Its 95% confidence interval, 62.27–85.18, remains above zero relative risk reduction, and the reported p-value is <0.0001.
This is a broader analysis population than the primary cohort analysis. The similarity in direction and the relatively narrow interval provide an additional description of the treatment comparison, but this analysis should retain its registered role as an other pre-specified analysis rather than being relabeled as the primary endpoint.
All-subjects medically attended RSV LRTI with hospitalization
Relative risk reduction
95% CI: 49.36–89.41 · P = 0.0002
The all-subjects hospitalization analysis reports a 76.84% relative risk reduction, with a 95% confidence interval of 49.36–89.41 and a p-value of 0.0002. The interval remains above zero relative risk reduction, indicating that the uncertainty interval is concentrated on the side of reduced risk.
The confidence interval is wider than the corresponding all-subjects LRTI endpoint, illustrating that even with the broader population, hospitalization-related outcomes can carry greater statistical uncertainty. The registry does not provide arm-level event counts in the posted statistical analysis information, so the relative effect cannot be translated into an absolute hospitalization risk reduction from these values alone.
10. Comparing the Four Posted Analyses
The four estimates are all reported on the same relative risk reduction scale, which makes their numerical magnitudes easy to compare. The more important statistical distinction is their role in the trial: one is the registered primary endpoint, one is a secondary endpoint, and two are other pre-specified analyses in all subjects.
The primary cohort hospitalization analysis is the clearest example of why point estimates should not be interpreted independently of their uncertainty. Its point estimate is 62.15%, but the 95% confidence interval extends below zero. By contrast, the other three posted confidence intervals remain above zero.
| Analysis | Role | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Primary cohort, MA RSV LRTI | Primary | RRR 74.53% | 49.63–87.12 | <0.0001 |
| Primary cohort, hospitalized MA RSV LRTI | Secondary | RRR 62.15% | -8.57–86.80 | 0.0708 |
| All subjects, MA RSV LRTI | Other pre-specified | RRR 76.36% | 62.27–85.18 | <0.0001 |
| All subjects, hospitalized MA RSV LRTI | Other pre-specified | RRR 76.84% | 49.36–89.41 | 0.0002 |
11. Intention-to-Treat Analysis
The statistical analyses identify the ITT population as the analysis population. In a randomized trial, the ITT principle preserves the treatment assignment established at randomization rather than redefining treatment groups according to subsequent treatment exposure or adherence.
Why ITT matters
Randomization creates comparable groups in expectation. An ITT comparison maintains that randomized contrast when analyzing outcomes.
What ITT does not guarantee
ITT does not eliminate missing data, nonadherence, protocol deviations, or uncertainty about outcomes. Those issues still require appropriate statistical handling.
For MELODY, the posted analysis explicitly connects the ITT framework with multiple imputation and covariate-adjusted Poisson regression. This combination separates the roles of randomization, missing-data handling, and effect estimation rather than treating them as interchangeable statistical concepts.
12. Covariate Adjustment and Stratification
The registry analysis notes that age at randomization was included as a covariate. It also identifies stratification factors in the analysis text. Covariate adjustment and stratification address related but distinct statistical goals: covariates enter the outcome model, while stratification can preserve aspects of the trial's randomized design in the analysis.
Adjustment does not replace randomization. Instead, it can account for prespecified variables and potentially improve the precision of the estimated treatment effect.
Importantly, a covariate-adjusted relative risk should still be interpreted as a model-based treatment comparison. The adjustment means that the reported estimate is conditional on the statistical model rather than being a simple unadjusted ratio of observed event proportions.
13. Multiple Imputation and Missing Data
The registry specifies that the confidence intervals and p-values for the posted analyses were obtained after multiple imputation. This is an important methodological detail because inference can depend on how incomplete observations are handled.
Why multiple imputation is used
Rather than replacing every missing value with a single deterministic value, multiple imputation creates multiple plausible versions of the incomplete dataset. Each completed dataset is analyzed, and the results are combined. The resulting inference reflects both within-dataset statistical variation and uncertainty associated with the missing information.
Strength
Multiple imputation can preserve uncertainty associated with missing information rather than pretending that an imputed value is known exactly.
Key assumption
Validity depends on the plausibility of the assumptions and variables used in the imputation model. The registry identifies multiple imputation but does not provide the full imputation specification in the statistical analysis information shown here.
The presence of multiple imputation also means that the reported confidence interval and p-value should be viewed as products of the full specified analysis procedure rather than as calculations based solely on the observed participant counts.
14. Statistical Methods Explained
Why was Poisson regression used for a binary endpoint?
The endpoint counts whether a participant experienced a defined clinical event within the 150-day window, so the outcome itself is binary. A modified Poisson approach with robust variance is commonly used in this setting to obtain a directly interpretable relative risk rather than an odds ratio. The robust variance addresses the mismatch between the Poisson variance assumption and the binary outcome structure.
What does a relative risk reduction of 74.53% mean?
It describes the proportional reduction in the modeled risk of the outcome associated with MEDI8897 relative to placebo. It is not an absolute risk reduction. For example, it cannot tell us how many fewer infections occurred per 100 or 1,000 infants without the underlying absolute event risks.
Why is the confidence interval important?
The point estimate is only one possible estimate under repeated sampling. The 95% confidence interval describes the uncertainty around the estimated treatment effect under the model and analysis framework. The primary interval of 49.63–87.12 is substantially narrower than the secondary hospitalization interval of -8.57–86.80, indicating greater precision for the primary endpoint estimate.
Why does the p-value not measure treatment-effect size?
A p-value quantifies statistical evidence against a null hypothesis under the specified model. It does not tell us whether the treatment effect is large, clinically important, or precisely estimated. Those questions require the effect estimate, confidence interval, and clinical context.
What does the hospitalization result of RRR 62.15% mean?
The point estimate indicates a lower modeled risk of hospitalized medically attended RSV lower respiratory tract infection with MEDI8897. However, the 95% confidence interval ranges from -8.57 to 86.80, so the data represented by that interval include both a modest increase and a substantial reduction in relative risk. The estimate therefore carries more uncertainty than the primary endpoint estimate.
Why use the ITT population?
Analyzing participants according to randomized assignment preserves the principal treatment comparison created by randomization. It prevents post-randomization treatment behavior from redefining the groups used for the primary efficacy question.
How does multiple imputation affect interpretation?
Multiple imputation means the reported inferential results incorporate uncertainty associated with missing information under the specified imputation framework. The resulting confidence interval and p-value therefore depend not only on observed outcomes but also on the statistical assumptions used to address missing data.
15. Confidence Intervals Across Endpoints
| Endpoint | Point estimate | 95% confidence interval | Interpretive feature |
|---|---|---|---|
| Primary cohort, MA RSV LRTI | RRR 74.53% | 49.63–87.12 | Interval entirely above zero RRR |
| Primary cohort, hospitalized MA RSV LRTI | RRR 62.15% | -8.57–86.80 | Interval includes zero RRR |
| All subjects, MA RSV LRTI | RRR 76.36% | 62.27–85.18 | Interval entirely above zero RRR |
| All subjects, hospitalized MA RSV LRTI | RRR 76.84% | 49.36–89.41 | Interval entirely above zero RRR |
One useful way to read this table is to distinguish magnitude from precision. The primary endpoint has a large point estimate and a confidence interval that remains above zero. The primary-cohort hospitalization analysis also has a large point estimate, but its confidence interval is much wider and crosses zero. The all-subjects analyses have intervals that remain above zero relative risk reduction.
For an RRR measure, zero represents no relative risk reduction. When a two-sided 95% confidence interval extends below zero, the interval includes values corresponding to no reduction and to an increase in risk as well as values corresponding to reductions. This is why the secondary hospitalization point estimate should not be interpreted in isolation.
16. Safety Results
The registry reports serious adverse events by randomized arm. The affected and at-risk counts are:
| Arm | Participants with serious adverse events | Participants at risk |
|---|---|---|
| Placebo | 83 | 997 |
| MEDI8897 | 149 | 1997 |
These are arm-level serious adverse-event counts as reported by the registry. The unequal arm sizes mean that the raw affected counts should not be compared without considering the corresponding numbers at risk.
Placebo
83 participants with serious adverse events among 997 participants at risk.
MEDI8897
149 participants with serious adverse events among 1997 participants at risk.
17. Primary Endpoint Versus Hospitalization Endpoint
The MELODY statistical record illustrates an important clinical-trial distinction: an overall disease-related endpoint and a more severe manifestation of that endpoint can produce different levels of statistical precision.
| Feature | Primary LRTI endpoint | Hospitalization endpoint |
|---|---|---|
| Clinical event | Medically attended RSV LRTI | Medically attended RSV LRTI with hospitalization |
| Time frame | Through 150 days post dose | Through 150 days post dose |
| Primary cohort RRR | 74.53% | 62.15% |
| Primary cohort 95% CI | 49.63–87.12 | -8.57–86.80 |
| Primary cohort P-value | <0.0001 | 0.0708 |
| Statistical role | Primary | Secondary |
The distinction is not simply that one endpoint is "positive" and the other is "negative." The statistical evidence differs in magnitude and precision. The hospitalization endpoint has a large point estimate but substantially wider uncertainty, demonstrating why a single p-value or point estimate is insufficient for understanding an endpoint.
18. Trial Timeline
Trial start
The MELODY study began on July 23, 2019.
Primary completion
The registry reports March 11, 2021 as the primary completion date.
Trial status
The ClinicalTrials.gov record identifies the study as completed and reports results for seven outcome measures, including four posted statistical analyses.
19. What the Primary RRR Does — and Does Not — Mean
The primary 74.53% RRR is a relative measure comparing the modeled risk of medically attended RSV lower respiratory tract infection between MEDI8897 and placebo. It indicates the relative magnitude of the treatment-associated difference on the risk scale.
It does not mean that 74.53% of infants were protected, that 74.53% of infections were prevented in absolute terms, or that every individual infant had exactly the same reduction in risk.
The 95% CI of 49.63–87.12 gives a range describing statistical uncertainty around the estimated relative risk reduction under the specified analysis. It does not describe the range of effects experienced by individual infants.
The P < 0.0001 result provides evidence against the null hypothesis under the registered superiority analysis. It is not a probability that the null hypothesis is true, and it does not measure clinical importance.
Relative risk reduction and absolute risk reduction answer different questions. Without the underlying arm-specific event probabilities, the absolute number of events prevented cannot be determined from the posted RRR alone.
20. Why the Analysis Uses Relative Risk Rather Than a Hazard Ratio
The registered endpoint is assessed as the number of participants experiencing a binary outcome through a fixed 150-day period. The posted analysis therefore uses Poisson regression and reports relative risk reduction.
| Effect measure | Typical interpretation | MELODY primary endpoint |
|---|---|---|
| Risk ratio | Ratio of cumulative probabilities of an event over a defined period | Underlying effect measure for the posted RRR |
| Relative risk reduction | Proportional reduction derived from the relative risk | Reported as 74.53% |
| Hazard ratio | Relative instantaneous event rate over time in a time-to-event model | Not the posted primary effect measure |
This distinction matters because a hazard ratio and a relative risk are not interchangeable. The hazard ratio is a time-to-event measure, whereas the MELODY primary endpoint is defined as whether the participant experienced the specified medically attended infection through a fixed follow-up period.
21. Statistical Interpretation of the Secondary Hospitalization Result
The secondary hospitalization endpoint provides a particularly useful example of why statistical interpretation should combine the estimate, confidence interval, and p-value.
RRR 62.15% · 95% CI -8.57–86.80 · P = 0.0708
First, the effect estimate: 62.15% is a substantial point estimate of relative risk reduction.
Second, the confidence interval: -8.57 to 86.80 is wide and crosses zero. That means the statistical uncertainty includes no relative reduction as well as substantial relative reductions.
Third, the p-value: 0.0708 does not cross the conventional 0.05 threshold. More importantly, it should be read alongside the wide confidence interval rather than treated as a standalone measure of the treatment effect.
This example illustrates why a large estimated effect can coexist with relatively limited statistical precision. The width of the confidence interval communicates information that the point estimate alone cannot.
22. Important Limitations and Interpretation Issues
- Absolute event risks: the posted statistical analyses report relative risk reduction, confidence intervals, and p-values, but the available analysis information does not provide arm-level event counts for the efficacy endpoints. Absolute risk reduction therefore cannot be determined from these results alone.
- Hospitalization precision: the primary-cohort hospitalization analysis has a wide confidence interval extending from -8.57 to 86.80, demonstrating substantial uncertainty around its point estimate.
- Binary endpoint: the primary outcome is assessed over a fixed 150-day period. A relative risk should not be interpreted as a hazard ratio or as an instantaneous event-rate comparison.
- Model dependence: the reported estimates and inference depend on the Poisson regression model, robust variance, age-at-randomization covariate adjustment, stratification factors, and multiple-imputation procedure.
- Analysis populations: the primary endpoint is defined for the primary cohort, while additional analyses use all subjects. Results from these populations should not be silently treated as identical estimands.
- Multiplicity: the record contains a primary endpoint, a secondary endpoint, and other pre-specified analyses. Their inferential roles differ and should not be collapsed into one undifferentiated set of tests.
- Safety comparison: serious adverse-event counts are available by arm, but a formal comparative safety analysis is not included in the posted statistical analyses summarized here.
- Missing data: multiple imputation was used, so interpretation depends partly on the assumptions underlying the imputation framework.
23. Why This Trial Matters Statistically
MELODY is a useful teaching case because its primary endpoint demonstrates how a randomized prevention trial can use a modified Poisson approach to estimate a directly interpretable relative risk measure, while its secondary hospitalization endpoint shows why point estimates must be interpreted together with their confidence intervals.
| Concept | How it appears in MELODY |
|---|---|
| Randomization | Participants were randomly allocated to MEDI8897 or placebo. |
| Blinding | The study used quadruple masking. |
| ITT analysis | The posted statistical analyses use the ITT population. |
| Binary endpoint | The primary outcome records whether participants experienced MA RSV LRTI through 150 days post dose. |
| Poisson regression | Used with robust variance for the posted analyses. |
| Relative risk reduction | The registered effect measure for the posted analyses. |
| Covariate adjustment | Age at randomization was included as a covariate. |
| Multiple imputation | Used before estimation of confidence intervals and p-values. |
| Stratified analysis | Identified in the posted analysis specification. |
| Confidence intervals | Show the different levels of precision across the four analyses. |
| Secondary endpoint interpretation | The hospitalization analysis illustrates how a large point estimate can have a confidence interval crossing zero. |
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Calculators
26. Sources
- ClinicalTrials.gov: NCT03979313 — MELODY.
- PubMed: PMID 40958705.
- PubMed: PMID 39656748.
- PubMed: PMID 38902186.
- PubMed: PMID 38219024.
- PubMed: PMID 37468530.
Continue through the Clinical Biostats knowledge graph
Connect randomized-trial endpoints to deeper statistical tutorials and practical analysis tools.
27. Record Summary
MELODY provides a clear example of randomized prevention-trial analysis using a binary clinical endpoint and a modified Poisson approach with robust variance. The primary analysis reported a 74.53% relative risk reduction for medically attended RSV lower respiratory tract infection through 150 days post dose in the primary cohort, with a 95% confidence interval of 49.63–87.12 and P < 0.0001.
The secondary hospitalization endpoint illustrates the importance of statistical precision: its point estimate was a 62.15% relative risk reduction, but the 95% confidence interval ranged from -8.57 to 86.80, with P = 0.0708. The two all-subjects analyses reported relative risk reductions of 76.36% and 76.84%, respectively, with confidence intervals remaining above zero relative risk reduction.
The statistical story therefore depends on more than the largest point estimate. Randomization establishes the treatment comparison; ITT analysis preserves that randomized framework; Poisson regression with robust variance provides the relative-risk inference; age at randomization and stratification factors enter the analysis specification; multiple imputation addresses missing information; and confidence intervals distinguish the precision of the different endpoint estimates.