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
NextCOVE was a randomized, parallel, quadruple-masked phase 3 prevention trial comparing mRNA-1283 formulations with mRNA-1273 formulations in participants aged 12 years and older. The registry reports 13,553 participants enrolled and five posted primary statistical analyses, including antibody-response comparisons and a time-to-event analysis of CDC-defined COVID-19.
| Feature | NextCOVE |
|---|---|
| Trial name | NextCOVE |
| Phase | Phase 3 |
| Condition | COVID-19 |
| Population | Participants ≥12 years of age |
| Design | Randomized, parallel, quadruple-masked |
| Primary purpose | Prevention |
| Enrollment | 13,553.0 |
| Primary endpoints registered | 11 |
| Outcome measures posted | 18 |
| Statistical analyses posted | 5 |
| ClinicalTrials.gov | NCT05815498 |
2. Clinical Question
The central question was whether mRNA-1283 vaccine formulations could provide immune responses and protection against COVID-19 that were non-inferior to the corresponding mRNA-1273 formulations, while characterizing solicited and unsolicited adverse reactions and longer safety outcomes.
Population
Participants aged 12 years and older enrolled in a phase 3 COVID-19 prevention study.
Intervention
mRNA-1283.222 in Part 1 and mRNA-1283.815 in Part 3.
Comparator
mRNA-1273.222 in Part 1 and mRNA-1273.815 in Part 3.
Primary question
Are the immune responses and vaccine efficacy associated with mRNA-1283 non-inferior to those associated with mRNA-1273?
3. Trial Design
mRNA-1283.222 vs mRNA-1273.222
- mRNA-1283.222 was compared with mRNA-1273.222.
- Primary immune-response assessments were performed at Day 29.
- CDC-defined COVID-19 was assessed from 14 days after injection up to Day 365.
- Safety endpoints included solicited reactions through Day 7, unsolicited AEs through Day 28, and selected events from Day 1 up to Day 365.
mRNA-1283.815 vs mRNA-1273.815
- mRNA-1283.815 was compared with mRNA-1273.815.
- Solicited local and systemic adverse reactions were assessed through Day 7.
- Unsolicited adverse events were assessed through Day 28.
- Selected serious and other safety outcomes were assessed from Day 1 up to Day 181.
The trial is a parallel-group design rather than a crossover design. The quadruple-masked structure is intended to reduce the influence of treatment knowledge on participant reporting, clinical assessment, and other study processes. Randomization provides the basis for comparing outcomes between vaccine formulations, while the non-inferiority framework determines whether the new formulation is sufficiently close to the comparator on prespecified measures.
4. Endpoints
The registry lists 11 primary endpoints. They cover immunogenicity, COVID-19 efficacy, and safety across Parts 1 and 3.
| Primary endpoint | Time frame | Definition / assessment |
|---|---|---|
| Part 1: GM of Omicron BA.4/5 | Day 29 | Antibody values below LLOQ were replaced by 0.5*LLOQ; values above ULOQ were replaced by ULOQ when actual values were unavailable. LLOQ was 103 AU/mL and ULOQ was 28571 AU/mL. |
| Part 1: SRR Omicron BA.4/5 | Day 29 | Seroresponse was based on the prespecified fold-rise rules using the pre-booster baseline. LLOQ was 103 AU/mL and ULOQ was 28571 AU/mL. |
| Part 1: GM of ancestral SARS-CoV-2 D614G | Day 29 | Antibody values below LLOQ were replaced by 0.5*LLOQ; values above ULOQ were replaced by ULOQ when actual values were unavailable. LLOQ was 10 AU/mL and ULOQ was 111433 AU/mL. |
| Part 1: SRR of ancestral SARS-CoV-2 D641G | Day 29 | Seroresponse used the prespecified fold-rise rules from the pre-booster baseline. LLOQ was 10 AU/mL and ULOQ was 111433 AU/mL. |
| Part 1: First event of CDC-defined COVID-19 | From 14 days after injection up to Day 365 | At least 1 CDC-listed symptom and a positive RT-PCR test on a respiratory sample. |
| Part 1: Solicited local and systemic ARs | Up to Day 7 | Daily electronic-diary assessment of prespecified local and systemic reactions. |
| Part 3: Solicited local and systemic ARs | Up to Day 7 | Daily electronic-diary assessment of prespecified local and systemic reactions. |
| Part 1: Unsolicited AEs | Up to Day 28 | Any AE reported by the participant that was not specified as a solicited AR in the protocol. |
| Part 3: Unsolicited AEs | Up to Day 28 | Any AE reported by the participant that was not specified as a solicited AR in the protocol. |
| Part 1: Any SAEs, MAAEs, AEs leading to withdrawal, and AESIs | Day 1 up to Day 365 | Composite safety assessment including serious, medically attended, withdrawal-related, and adverse events of special interest. |
| Part 3: Any SAEs, MAAEs, AEs leading to withdrawal, and AESIs | Day 1 up to Day 181 | Composite safety assessment including serious, medically attended, withdrawal-related, and adverse events of special interest. |
How the antibody endpoints were constructed
The geometric-mean endpoints are continuous antibody measurements. Because antibody concentrations are typically right-skewed, geometric means provide a multiplicative summary that is naturally interpreted through ratios. The registry also specifies handling of measurements outside the assay's quantification range, which is important because the treatment comparison depends on how those observations enter the analysis dataset.
The seroresponse endpoints are binary classifications derived from the change from baseline. The definition is deliberately conditional on the baseline antibody level: a participant starting below the LLOQ requires a rise to at least 4*LLOQ, whereas participants with measurable baseline antibody require prespecified fold increases.
How the COVID-19 endpoint differs
The COVID-19 endpoint is a time-to-first-event outcome. Participants contribute follow-up beginning 14 days after injection and are evaluated through Day 365. The event definition combines a clinical component with virologic confirmation, requiring at least 1 CDC-listed symptom and a positive RT-PCR test on a respiratory sample.
5. Statistical Methodology
Non-inferiority framework
All five posted primary statistical analyses are identified in the registry as non-inferiority analyses. This changes the interpretation of the confidence interval. The relevant question is not simply whether an estimated effect differs from zero or one, but whether the entire estimate is sufficiently favorable relative to a prespecified margin.
| Endpoint | Effect measure | Estimate | 95% CI | Non-inferiority margin |
|---|---|---|---|---|
| Omicron BA.4/5 GM | GMR | 1.335 | 1.194–1.492 | 0.667 (1/1.5) |
| Omicron BA.4/5 SRR | SRR difference | 14.4 | 9.3–19.4 | 10% |
| Ancestral D614G GM | GMR | 1.240 | 1.128–1.362 | 0.667 (1/1.5) |
| Ancestral D641G SRR | SRR difference | 10.7 | 6.0–15.4 | 10% |
| CDC-defined COVID-19 | rVE | 9.31 | −6.58–22.83 | 10% |
Geometric mean ratios
The registry reports a ratio of geometric means for the two continuous antibody endpoints. A GMR of 1.335 means that the geometric mean antibody level for mRNA-1283.222 was estimated to be 1.335 times that for mRNA-1273.222 at Day 29. The corresponding 95% confidence interval was 1.194 to 1.492.
For a non-inferiority analysis, the direction and margin matter. Here the prespecified lower non-inferiority margin was 0.667. A confidence interval entirely above that margin supports the non-inferiority criterion for the ratio.
Seroresponse-rate differences
The SRR analyses use an absolute difference between treatment groups. The reported estimate of 14.4 for Omicron BA.4/5 is expressed in percentage-point units according to the registry's effect-measure field. The 95% confidence interval is 9.3 to 19.4, and the non-inferiority margin is 10%.
The ancestral endpoint has the same structure: the reported SRR difference is 10.7, with a 95% confidence interval of 6.0 to 15.4 and a non-inferiority margin of 10%.
Cox proportional-hazards model
The registry reports a Cox proportional-hazards model for the first event of CDC-defined COVID-19. This is appropriate for an endpoint where both the occurrence and timing of the first event are relevant and where participants may have different lengths of event-free follow-up.
The registry defines relative vaccine efficacy as 1 minus the hazard ratio for mRNA-1283.222 versus mRNA-1273.222. The posted estimate was 9.31.
The analysis population was the Per-Protocol Set for Efficacy (PPSE), consisting of randomized participants who received the planned dose of study drug and had no major protocol deviations that impacted vaccine efficacy data. The antibody analyses used the PPIS, which required the planned dose, baseline and Day 29 neutralizing-antibody data, and no major protocol deviations affecting key or critical data. 'Overall number of participants analyzed' = participants evaluable for this outcome measure.
What the registry does not specify for the antibody analyses
The statistical method field for the four antibody primary analyses is listed as not reported. The registry does, however, identify the effect measures as GMRs or SRR differences and provides the non-inferiority margins and confidence intervals. For a continuous antibody endpoint, a ratio of geometric means is naturally associated with analysis on the logarithmic scale followed by back-transformation, but the specific model or calculation should not be attributed to the registry beyond what it reports.
6. Results: Antibody Response
Omicron BA.4/5 Geometric Mean
Geometric mean ratio at Day 29
95% CI: 1.194–1.492 · Non-inferiority margin: 0.667 (1/1.5)
mRNA-1283.222 versus mRNA-1273.222
The GMR of 1.335 means that the estimated geometric mean Omicron BA.4/5 antibody level in the mRNA-1283.222 group was 1.335 times the corresponding geometric mean in the mRNA-1273.222 group at Day 29.
The 95% confidence interval, 1.194–1.492, quantifies statistical precision around the estimated ratio. It does not describe the range of individual antibody responses. It is entirely above the non-inferiority margin of 0.667, which is the relevant comparison for the registry's non-inferiority framework.
There is no p-value reported for this posted analysis. In a non-inferiority setting, the margin and confidence interval provide the direct framework for determining whether the prespecified criterion is met; a p-value would not itself describe the magnitude of the antibody difference.
Omicron BA.4/5 Seroresponse Rate
Seroresponse-rate difference at Day 29
95% CI: 9.3–19.4 · Non-inferiority margin: 10%
mRNA-1283.222 minus mRNA-1273.222
The reported SRR difference of 14.4 represents the difference in seroresponse rates between the two groups under the registry's prespecified seroresponse definition. The confidence interval extends from 9.3 to 19.4.
The non-inferiority margin is 10%. Because the lower confidence-limit value is 9.3, the interval itself crosses the margin. That is the key statistical feature to recognize when interpreting this particular posted estimate rather than focusing only on the point estimate of 14.4.
The confidence interval describes uncertainty around the between-group difference; it does not describe variability among individual participants. No p-value is posted for this analysis, so the non-inferiority interpretation should remain anchored to the prespecified margin and confidence interval.
Ancestral SARS-CoV-2 D614G Geometric Mean
Geometric mean ratio at Day 29
95% CI: 1.128–1.362 · Non-inferiority margin: 0.667 (1/1.5)
mRNA-1283.222 versus mRNA-1273.222
The estimated GMR of 1.240 indicates that the geometric mean ancestral SARS-CoV-2 D614G antibody level was estimated at 1.240 times that of the comparator group at Day 29.
The 95% confidence interval of 1.128–1.362 is relatively narrow compared with the magnitude of the point estimate and lies above the non-inferiority margin of 0.667. The interval represents uncertainty in the estimated group ratio, not the distribution of antibody concentrations in individual participants.
No p-value is reported. For this non-inferiority analysis, the relationship of the confidence interval to the prespecified margin is more directly informative than a generic significance test.
Ancestral SARS-CoV-2 D641G Seroresponse Rate
Seroresponse-rate difference at Day 29
95% CI: 6.0–15.4 · Non-inferiority margin: 10%
mRNA-1283.222 minus mRNA-1273.222
The estimated SRR difference is 10.7, meaning the observed seroresponse-rate difference favored the mRNA-1283.222 group by the reported estimate. The 95% confidence interval ranges from 6.0 to 15.4.
The lower confidence limit of 6.0 is below the 10% non-inferiority margin. Thus, the point estimate alone should not be used to characterize the non-inferiority criterion; the full confidence interval is essential to that assessment.
Again, no p-value is posted. A p-value would address a null hypothesis chosen for a specific test, whereas the non-inferiority question is defined around the prespecified clinical margin.
7. Results: First CDC-Defined COVID-19 Event
Relative vaccine efficacy
99.4% CI: −6.58–22.83 · P = 0.0005
mRNA-1283.222 versus mRNA-1273.222
| Endpoint | Analysis population | Method | Effect | Confidence interval | P-value |
|---|---|---|---|---|---|
| First event of CDC-defined COVID-19 from 14 days after injection up to Day 365 | PPSE | Cox proportional-hazards model | rVE 9.31 | 99.4% CI −6.58–22.83 | = 0.0005 |
The registry defines rVE as 1 minus the hazard ratio. An rVE estimate of 9.31 therefore corresponds to a hazard ratio of approximately 0.9069 under that definition, meaning the fitted model estimates a lower instantaneous rate of the first CDC-defined COVID-19 event in the mRNA-1283.222 group relative to mRNA-1273.222.
The 99.4% confidence interval is −6.58 to 22.83. The interval is wide enough to include both a negative relative vaccine-efficacy value and values above the point estimate, illustrating substantial uncertainty around the estimated treatment difference.
The p-value of 0.0005 should not be interpreted as the size of the vaccine effect. A p-value measures evidence against a specified null hypothesis under the statistical model; it does not tell the reader how clinically large the treatment difference is. For this non-inferiority analysis, the prespecified 10% rVE margin and the confidence interval are central to interpretation.
Because the registry identifies a Cox proportional-hazards model, interpretation also depends on the usual proportional-hazards framework. The hazard ratio is a relative time-to-event measure, not a direct statement about the proportion of participants who experienced COVID-19.
8. Safety Results
The registry reports serious adverse events by arm using affected participants divided by participants at risk. These figures are reported separately for Parts 1 and 3.
| Part | Intervention | Serious adverse events | Interpretation of denominator |
|---|---|---|---|
| Part 1 | mRNA-1283.222 | 203/5706 | Affected / at risk |
| Part 1 | mRNA-1273.222 | 197/5710 | Affected / at risk |
| Part 3 | mRNA-1283.815 | 15/1052 | Affected / at risk |
| Part 3 | mRNA-1273.815 | 20/1039 | Affected / at risk |
The safety endpoints were broader than serious adverse events alone. Solicited local and systemic reactions were recorded daily through Day 7 using electronic diaries. Unsolicited adverse events were assessed through Day 28. Part 1 also included serious adverse events, medically attended adverse events, adverse events leading to withdrawal, and adverse events of special interest from Day 1 up to Day 365; the corresponding Part 3 endpoint extended from Day 1 up to Day 181.
Solicited local reactions
Injection-site pain, erythema, swelling or induration, and ipsilateral axillary swelling or tenderness were recorded during the 7-day follow-up period.
Solicited systemic reactions
Fever, headache, fatigue, myalgia, arthralgia, nausea/vomiting, and chills were recorded using electronic diaries.
Unsolicited adverse events
An unsolicited AE was an AE reported by a participant that was not specified as a solicited adverse reaction in the protocol.
Serious adverse events
SAEs included events resulting in death, life-threatening events, hospitalization or prolonged hospitalization, persistent or significant incapacity, congenital anomaly or birth defect, or medically important events.
Not all solicited adverse reactions were considered adverse events. The investigator reviewed whether a solicited reaction should also be recorded as an AE. This distinction matters statistically because a diary-based solicited-reaction endpoint and a broader AE endpoint have different ascertainment mechanisms and clinical meanings.
9. Statistical Methods Explained
Why was a non-inferiority design used?
A non-inferiority trial asks whether a new intervention is not unacceptably worse than an established comparator according to a prespecified margin. In NextCOVE, the registry reports non-inferiority hypotheses for the posted primary analyses. The margin is therefore part of the endpoint definition rather than an after-the-fact interpretation of the observed estimate.
What does a GMR of 1.335 mean?
A GMR of 1.335 means the geometric mean antibody level in the mRNA-1283.222 group was estimated to be 1.335 times that in the mRNA-1273.222 group. Because the endpoint is a ratio, 1 represents equality between groups. The non-inferiority margin of 0.667 represents the lower boundary against which the confidence interval is evaluated.
Why are geometric means used for antibody concentrations?
Antibody measurements can span a wide multiplicative range. Geometric means summarize such data on a multiplicative rather than additive scale, making ratios between treatment groups natural effect measures. The registry's use of GMR therefore provides a way to compare relative antibody levels without treating the raw concentration scale as approximately symmetric.
Why is the confidence interval more important than the point estimate in non-inferiority?
A point estimate can be comfortably above a non-inferiority margin while the confidence interval still extends beyond that margin. The SRR analyses illustrate this principle: the point estimates were 14.4 and 10.7, but their lower confidence limits were 9.3 and 6.0, respectively. Non-inferiority conclusions therefore depend on the prespecified margin and the uncertainty around the estimate.
Why was a Cox model used for COVID-19?
The CDC-defined COVID-19 endpoint records the first event over a defined follow-up period. A Cox proportional-hazards model uses both whether an event occurred and when it occurred, while allowing participants who remain event-free through their available follow-up to contribute censored information. This is different from simply comparing the proportion of participants with COVID-19 at one fixed time point.
What does rVE of 9.31 mean?
The registry defines rVE as 1 minus the hazard ratio for mRNA-1283.222 versus mRNA-1273.222. The reported rVE of 9.31 therefore represents the estimated relative reduction in the instantaneous hazard of a first CDC-defined COVID-19 event under that definition. It does not mean that exactly 9.31% of participants were protected, nor does it represent an absolute risk reduction.
10. Analysis Populations and Why They Matter
| Population | Registry definition | Primary role |
|---|---|---|
| PPIS | Randomized participants who received the planned dose, had baseline and Day 29 nAb data, and had no major protocol deviations impacting key or critical data. 'Overall number of participants analyzed' = participants evaluable for this outcome measure. | Antibody primary endpoints |
| PPSE | Randomized participants who received the planned dose of study drug and had no major protocol deviations impacting vaccine efficacy data. | First CDC-defined COVID-19 event |
The distinction between these populations is statistically important. Randomization establishes comparability at the start of the trial, but a per-protocol analysis applies additional eligibility conditions after randomization. Such restrictions can improve alignment with the biological question of vaccine performance under protocol adherence, while also creating the possibility that the analyzed population differs from the original randomized population.
For the antibody endpoints, evaluability requires baseline and Day 29 neutralizing-antibody measurements. For the efficacy endpoint, the PPSE excludes participants with major protocol deviations affecting vaccine-efficacy data. These definitions should therefore be considered when interpreting the reported estimates and their generalizability.
11. Understanding the Non-inferiority Margins
| Endpoint family | Effect scale | Margin | What the margin means |
|---|---|---|---|
| Omicron BA.4/5 GM | GMR | 0.667 (1/1.5) | The lower acceptable boundary for the ratio in the registry's non-inferiority framework. |
| Ancestral D614G GM | GMR | 0.667 (1/1.5) | The lower acceptable boundary for the ratio in the registry's non-inferiority framework. |
| Omicron BA.4/5 SRR | SRR difference | 10% | The prespecified margin on the reported SRR-difference scale. |
| Ancestral D641G SRR | SRR difference | 10% | The prespecified margin on the reported SRR-difference scale. |
| CDC-defined COVID-19 | rVE | 10% | The prespecified margin on the relative-vaccine-efficacy scale. |
These margins are not interchangeable. A GMR is a ratio, an SRR difference is an absolute difference in response rates, and rVE is defined from a hazard ratio. The numerical value of a margin only has meaning on its corresponding effect scale.
12. Why the P-value and Effect Estimate Answer Different Questions
The posted COVID-19 efficacy analysis contains both an effect estimate and a p-value: rVE 9.31, with a 99.4% confidence interval of −6.58 to 22.83 and P = 0.0005. These quantities should not be treated as interchangeable.
Effect estimate
The rVE estimate describes the magnitude and direction of the relative vaccine-efficacy effect on the registry's defined scale.
Confidence interval
The 99.4% interval describes uncertainty around the estimate and shows the range of effects compatible with the statistical analysis at that confidence level.
P-value
The p-value quantifies evidence against a specified null hypothesis under the statistical model; it is not a measure of effect size.
Non-inferiority margin
The margin defines how much loss of efficacy would be considered unacceptable for the non-inferiority question.
This distinction is particularly important when an analysis uses a nonstandard confidence level such as 99.4%. The confidence interval should be interpreted at the stated confidence level rather than silently converted to a different level.
13. Antibody Measurement and Quantification Limits
The registry explicitly defines how antibody values outside the assay's quantification range were handled.
| Antibody endpoint | LLOQ | ULOQ | Out-of-range handling |
|---|---|---|---|
| Omicron BA.4/5 | 103 AU/mL | 28571 AU/mL | Below LLOQ: 0.5*LLOQ; above ULOQ: ULOQ if actual value unavailable. |
| Ancestral SARS-CoV-2 D614G | 10 AU/mL | 111433 AU/mL | Below LLOQ: 0.5*LLOQ; above ULOQ: ULOQ if actual value unavailable. |
This preprocessing step matters because geometric means are sensitive to the values included in the analysis. Replacing measurements below the quantification limit with a specified fraction of the LLOQ avoids treating those observations as missing, while replacing unquantifiable high values with the ULOQ prevents an unavailable actual measurement from entering the calculation as though its precise concentration were known.
14. Safety Endpoint Structure
The safety strategy uses several complementary observation windows rather than a single adverse-event endpoint. The shortest window focuses on solicited reactions after vaccination, while the Day 28 window captures unsolicited adverse events. Longer windows capture serious, medically attended, withdrawal-related, and special-interest events.
| Safety assessment | Part 1 | Part 3 |
|---|---|---|
| Solicited local and systemic ARs | Up to Day 7 | Up to Day 7 |
| Unsolicited AEs | Up to Day 28 | Up to Day 28 |
| SAEs / MAAEs / withdrawal / AESIs | Day 1 up to Day 365 | Day 1 up to Day 181 |
These endpoints have different ascertainment periods and definitions. A solicited reaction is actively elicited through an electronic diary, whereas an unsolicited AE is participant-reported without being one of the protocol-specified solicited reactions. Consequently, comparing their frequencies as though they were the same statistical outcome would obscure important differences in measurement.
15. Limitations
- Different analysis populations: the antibody analyses use PPIS and the COVID-19 efficacy analysis uses PPSE. These populations are not equivalent to an all-randomized intention-to-treat analysis.
- Per-protocol selection: requiring protocol adherence and evaluable antibody measurements can affect which randomized participants contribute to a particular endpoint.
- Confidence-interval interpretation: several non-inferiority intervals extend across the corresponding prespecified margin, so the point estimates alone do not capture the full statistical conclusion.
- Antibody-model specification: the registry does not report the specific statistical method used for the four posted antibody analyses, although it reports the effect measures, estimates, confidence intervals, and non-inferiority margins.
- Time-to-event assumptions: the COVID-19 analysis uses a Cox proportional-hazards model, whose hazard-ratio interpretation depends on the proportional-hazards framework.
- Endpoint multiplicity: the registry lists 11 registered primary endpoints and five posted primary statistical analyses. The available information does not specify a complete multiplicity-adjustment strategy across all registered endpoints.
- Safety comparisons: serious-adverse-event counts are reported for affected participants and participants at risk, while the broader safety endpoint definitions use different observation windows and ascertainment methods.
- Part-specific populations: Part 1 and Part 3 involve different vaccine formulations and different safety follow-up windows, so their results should not be treated as though they arose from a single homogeneous comparison.
16. Why This Trial Matters Statistically
NextCOVE is a useful teaching example because it combines two distinct statistical questions within a single randomized prevention program: whether antibody responses are sufficiently preserved relative to a comparator and whether the vaccine formulation changes the time to a clinically defined COVID-19 event.
| Statistical concept | How it appears in NextCOVE |
|---|---|
| Randomization | The study is randomized and parallel-group. |
| Blinding | The registry describes the study as quadruple-masked. |
| Non-inferiority | All five posted primary statistical analyses use a non-inferiority hypothesis. |
| Geometric mean ratio | Used for the two continuous antibody endpoints. |
| Seroresponse-rate difference | Used for the two binary antibody-response endpoints. |
| Cox proportional hazards | Used for the first CDC-defined COVID-19 event. |
| Per-protocol analysis | PPIS and PPSE populations are used for primary analyses. |
| Confidence intervals | 95% intervals are posted for the antibody analyses; a 99.4% interval is posted for rVE. |
| Time-to-event analysis | COVID-19 efficacy is defined as the first event over a specified follow-up period. |
| Safety ascertainment | Solicited, unsolicited, serious, medically attended, withdrawal-related, and special-interest events use distinct windows. |
The trial therefore demonstrates why a clinical-trial statistical analysis cannot be reduced to a single p-value. The estimand changes with the endpoint: a ratio of geometric means answers an immunogenicity question, an SRR difference answers a binary response question, and a Cox-model-derived rVE addresses the relative rate of a first clinical event over time.
It also illustrates an important principle in non-inferiority trials: the clinical margin is part of the inferential framework. An estimate that looks favorable in isolation may have a confidence interval that remains compatible with an effect beyond the prespecified margin.
17. Overall Statistical Interpretation
The posted geometric-mean analyses report GMR estimates of 1.335 for Omicron BA.4/5 and 1.240 for ancestral SARS-CoV-2 D614G, with confidence intervals of 1.194–1.492 and 1.128–1.362, respectively. Both intervals lie above the corresponding GMR non-inferiority margin of 0.667.
The posted SRR differences are 14.4 for Omicron BA.4/5 and 10.7 for ancestral SARS-CoV-2 D641G. Their confidence intervals are 9.3–19.4 and 6.0–15.4, respectively, compared with a 10% non-inferiority margin. The lower confidence limits therefore provide important context that is not visible from the point estimates alone.
The Cox-model analysis reports rVE of 9.31 with a 99.4% confidence interval of −6.58–22.83 and P = 0.0005. The registry defines rVE as 1 minus the hazard ratio. Because the confidence interval spans negative and positive rVE values, the estimate should be interpreted together with the prespecified 10% non-inferiority margin and the per-protocol analysis population rather than by the p-value alone.
Serious adverse events were reported as 203/5706 for mRNA-1283.222 versus 197/5710 for mRNA-1273.222 in Part 1, and 15/1052 versus 20/1039 for mRNA-1283.815 versus mRNA-1273.815 in Part 3. These serious-event summaries complement, rather than replace, the separate solicited and unsolicited safety endpoints.
18. Related Tutorials
Learn more about the methods used in this trial:
19. Related Calculators
20. Sources
- ClinicalTrials.gov: NextCOVE, NCT05815498.
- Linked publication: PubMed record for PMID 40639387.
Continue through Clinical Biostats
Explore statistical tutorials and calculators covering the methods used to design, analyze, and interpret randomized clinical trials.