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
BLAZE-2 was a randomized, double-blind, phase 3 prevention trial evaluating bamlanivimab, with etesevimab also listed among the trial interventions, in nursing home residents and staff. The primary registered endpoint was the percentage of participants with COVID-19 by Week 8 after randomization.
| Feature | BLAZE-2 |
|---|---|
| Phase | Phase 3 |
| Condition | COVID-19; SARS-CoV2 |
| Primary purpose | Prevention |
| Design | Randomized, double-blind, parallel-group |
| Enrollment | 1180 |
| Arms | 9 |
| Interventions | Bamlanivimab; placebo; etesevimab |
| Primary endpoint | Percentage of Participants With COVID-19 |
| Primary endpoint type | Binary |
| Primary analysis | Logistic regression |
| Effect measure | Odds ratio |
| Hypothesis type | Superiority |
| Status | Completed |
| Start | 2020-08-02 |
| Primary completion | 2021-01-16 |
2. Clinical Question
The central question was whether bamlanivimab, compared with placebo, reduced the odds of developing COVID-19 by Week 8 after randomization in the prevention population.
Population
Participants in the prevention population enrolled in a phase 3 study of nursing home residents and staff addressing SARS-CoV-2 infection and COVID-19.
Intervention
Bamlanivimab was one of the registered interventions. The posted primary analysis compared 4200mg bamlanivimab with placebo.
Comparator
Placebo was the comparator in the posted primary analysis.
Primary question
Does 4200mg bamlanivimab reduce the odds of the registered COVID-19 endpoint compared with placebo by Week 8?
3. Trial Design
4200mg Bamlanivimab
- Analyzed against placebo
- Primary efficacy population: all randomized participants from the prevention population
- Primary endpoint assessed by Week 8
- Logistic regression used for the posted primary analysis
Placebo
- Compared with 4200mg bamlanivimab
- Primary efficacy population: all randomized participants from the prevention population
- Primary endpoint assessed by Week 8
- Odds ratio reported as the treatment effect measure
The registry lists nine study arms and three interventions: bamlanivimab, placebo, and etesevimab. The posted statistical analyses shown here specifically compare placebo with 4200mg bamlanivimab.
4. Endpoints
| Endpoint | Time frame | Registry definition / analysis |
|---|---|---|
| Percentage of Participants With COVID-19 | Week 8 after randomization | The first occurrence of COVID-19, defined as detection of SARS-CoV-2 by RT-PCR and mild or worse disease severity within 21 days of detection, by Day 57. The participant needed to test positive on or prior to Week 8 and develop symptoms on or after the positive test date, but no later than 21 days after the positive swab or Week 8, whichever came first. |
| Percentage of Participants With Moderate or Worse Severity COVID-19 | Week 8 after randomization | Secondary binary endpoint analyzed in all randomized participants from the prevention population. |
| Percentage of Participants With SARS-CoV-2 | Week 4 | Secondary binary endpoint analyzed in all randomized participants from the prevention population. |
The primary endpoint is therefore a binary outcome rather than a time-to-event endpoint. That distinction determines the statistical framework: the primary analysis uses logistic regression and reports an odds ratio rather than a hazard ratio.
5. Statistical Methodology
Primary analysis: logistic regression
The registry specifies logistic regression for the primary endpoint. The response variable was occurrence of a primary endpoint event, while treatment and stratification factors such as facility were included as explanatory variables.
Here, Y=1 represents occurrence of the binary endpoint. The treatment coefficient is translated into an odds ratio by exponentiating the corresponding model coefficient.
Odds ratio as the effect measure
The primary treatment effect was reported as an odds ratio comparing 4200mg bamlanivimab with placebo. An odds ratio below 1 indicates lower estimated odds of the binary outcome in the bamlanivimab group relative to placebo, conditional on the variables included in the regression model.
An odds ratio compares odds, not probabilities directly. When an outcome is uncommon, the odds ratio and risk ratio can be numerically similar, but they are not mathematically identical measures.
Analysis population
The posted primary analysis used all randomized participants from the prevention population. This preserves the treatment assignment generated by randomization for the primary efficacy comparison.
Superiority framework
The registry identifies the hypothesis type as superiority. The statistical question is therefore whether the treatment groups differ in the direction specified by the superiority comparison, rather than whether the treatment is merely no worse than a prespecified non-inferiority margin.
6. Primary Result: COVID-19 by Week 8
The primary endpoint was the percentage of participants with COVID-19 at Week 8 after randomization. The posted analysis compared placebo with 4200mg bamlanivimab using logistic regression.
Odds ratio for COVID-19
95% CI: 0.26–0.63 · P < 0.001
Two-sided 95% confidence interval; superiority hypothesis.
| Primary endpoint | Comparison | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|---|
| Percentage of Participants With COVID-19 | 4200mg Bamlanivimab vs Placebo | Logistic regression | OR 0.40 | 0.26–0.63 | <0.001 |
An odds ratio of 0.40 means that the estimated odds of the primary COVID-19 endpoint were 40% of the corresponding odds in the placebo group in the posted logistic regression analysis. Expressed as a relative odds comparison, this corresponds to approximately 60% lower estimated odds in the 4200mg bamlanivimab group.
The odds ratio does not mean that exactly 40% as many participants developed COVID-19, nor does it mean that the absolute probability of COVID-19 was reduced by 60 percentage points. Those interpretations would require the underlying event probabilities.
The 95% confidence interval of 0.26–0.63 describes uncertainty around the estimated odds ratio under the statistical model and sampling framework. It does not describe the range of individual treatment effects across participants.
The P < 0.001 value addresses evidence against the null hypothesis under the specified statistical framework. It does not measure the size, clinical importance, or precision of the treatment effect. The confidence interval and the effect estimate provide that complementary information.
Because the endpoint is binary and the reported model is logistic regression, proportional-hazards assumptions and censoring considerations that arise in survival analysis are not the central issues for this primary comparison.
7. Secondary Endpoint Results
Moderate or Worse Severity COVID-19 at Week 8
Odds ratio
95% CI: 0.27–0.67 · P < 0.001
4200mg Bamlanivimab vs Placebo; logistic regression.
| Secondary endpoint | Comparison | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Percentage of Participants With Moderate or Worse Severity COVID-19 | 4200mg Bamlanivimab vs Placebo | OR 0.43 | 0.27–0.67 | <0.001 |
| Percentage of Participants With SARS-CoV-2 | 4200mg Bamlanivimab vs Placebo | OR 0.66 | 0.46–0.94 | 0.021 |
For moderate or worse severity COVID-19, the odds ratio of 0.43 indicates lower estimated odds in the 4200mg bamlanivimab group than in the placebo group under the posted logistic regression model. The 95% CI of 0.27–0.67 quantifies uncertainty around that estimate.
The Week 4 SARS-CoV-2 endpoint had an odds ratio of 0.66, with a 95% CI of 0.46–0.94 and a two-sided P-value of 0.021. This is a separate binary endpoint at an earlier time point and should not be interpreted as interchangeable with the Week 8 COVID-19 endpoint.
As with the primary result, these odds ratios should not be converted into absolute risk reductions without the corresponding event probabilities. A regression-adjusted odds ratio is also not necessarily the same quantity as an unadjusted risk ratio.
8. Comparing the Reported Odds Ratios
The three reported odds ratios are all below 1, but they correspond to different endpoints and time frames. The primary Week 8 COVID-19 endpoint is not statistically interchangeable with the Week 4 SARS-CoV-2 endpoint, because COVID-19 required both detection of SARS-CoV-2 and the specified symptom-severity timing criteria.
The confidence intervals also provide important context. The primary estimate of 0.40 has a 95% CI extending from 0.26 to 0.63, whereas the Week 4 SARS-CoV-2 estimate of 0.66 has a 95% CI from 0.46 to 0.94. The width and location of each interval reflect the uncertainty associated with its own endpoint and analysis.
9. Serious Adverse Events
The registry reports serious adverse events by arm for the placebo and 4200mg bamlanivimab groups.
| Arm | Affected / at risk |
|---|---|
| Placebo | 30 / 587 |
| 4200mg Bamlanivimab | 39 / 588 |
These counts describe participants affected by serious adverse events and the corresponding at-risk population reported for each arm. They should be distinguished from the efficacy endpoints, which were analyzed using logistic regression for the posted comparisons.
10. Statistical Methods Explained
Why was logistic regression used?
The primary endpoint is binary: each participant either met the registered COVID-19 endpoint by Week 8 or did not. Logistic regression is designed for this type of outcome because it models the probability of an event while allowing treatment and other explanatory variables, including stratification factors such as facility, to enter the model.
What does an odds ratio of 0.40 mean?
An odds ratio of 0.40 means that the estimated odds of the outcome in the 4200mg bamlanivimab group were 0.40 times the odds in the placebo group under the fitted model. This is equivalent to approximately 60% lower estimated odds, but it is not equivalent to saying that the probability of disease was 60% lower.
Why are odds different from probabilities?
If the probability of an event is p, its odds are p/(1-p). For example, a probability of 0.20 corresponds to odds of 0.20/0.80, or 0.25. Because the transformation is nonlinear, an odds ratio cannot generally be interpreted as a simple percentage-point difference in probability.
What does the confidence interval tell us?
The 95% confidence interval expresses statistical uncertainty around the estimated odds ratio. For the primary endpoint, the interval is 0.26–0.63. A narrower interval would indicate greater precision; a wider interval would indicate less precision. The interval does not describe the distribution of individual treatment effects.
Why does the P-value not measure treatment effect size?
The P-value evaluates the compatibility of the observed data with a specified null hypothesis under the statistical model. It depends on both the observed effect and the amount of information available. A very small P-value therefore does not automatically imply a large effect, while a larger P-value does not establish that the effect is absent.
Why include facility or other explanatory variables?
The registry states that the logistic regression model included treatment and stratification factors such as facility as explanatory variables. Incorporating prespecified design factors can account for variation associated with those factors and align the analysis with the randomized trial structure.
Why does randomization matter for the interpretation?
Randomization creates the basis for comparing treatment groups without relying on observational adjustment alone. When participants are analyzed according to their randomized assignments, differences between groups can be interpreted in the context of the randomized treatment comparison rather than simply as associations between an exposure and an outcome.
11. Confidence Intervals and the Primary Estimate
The entire reported interval lies below 1.00. For an odds ratio comparing treatment with placebo, an interval below 1 is consistent with lower estimated odds in the treatment group under the model.
The confidence interval is especially important because the point estimate alone gives an incomplete picture. An odds ratio of 0.40 could appear highly precise or highly uncertain depending on the interval surrounding it. Here, the reported interval supplies the uncertainty information directly.
Point estimate
0.40 is the single best estimate reported by the posted primary analysis.
Precision
0.26–0.63 gives the reported 95% confidence interval around the estimate.
Statistical evidence
P < 0.001 indicates strong evidence against the specified null hypothesis under the reported analysis.
Clinical scale
The odds ratio is a relative measure. Absolute clinical impact requires the underlying event probabilities and their context.
12. Primary vs Secondary Questions
| Endpoint | Role | Time frame | Effect measure | Estimate | P-value |
|---|---|---|---|---|---|
| Percentage of Participants With COVID-19 | Primary | Week 8 after randomization | Odds ratio | 0.40 (95% CI 0.26–0.63) | <0.001 |
| Percentage of Participants With Moderate or Worse Severity COVID-19 | Secondary | Week 8 after randomization | Odds ratio | 0.43 (95% CI 0.27–0.67) | <0.001 |
| Percentage of Participants With SARS-CoV-2 | Secondary | Week 4 | Odds ratio | 0.66 (95% CI 0.46–0.94) | 0.021 |
The distinction between primary and secondary endpoints is important because statistical interpretation depends on the role an endpoint plays in the prespecified trial design. The primary endpoint is the central confirmatory question, while secondary endpoints provide additional information about the intervention's effects on related outcomes.
13. Multiplicity and Multiple Endpoints
The registry reports one primary endpoint and additional secondary endpoints. When a trial evaluates several outcomes, the interpretation of their P-values depends on the prespecified multiplicity strategy and endpoint hierarchy.
A collection of nominally significant P-values should not automatically be interpreted as a collection of independent confirmatory findings. If multiple hypotheses are tested, the overall probability of false-positive findings can increase unless the design specifies appropriate error control.
Primary endpoint
The Week 8 COVID-19 endpoint was the registered primary efficacy question and had a posted formal analysis.
Secondary endpoints
Moderate or worse severity COVID-19 at Week 8 and SARS-CoV-2 at Week 4 were additional binary outcomes with posted analyses.
The registry information in the ClinicalTrials.gov record does not specify an alpha-spending procedure, formal multiplicity-adjustment strategy, or hierarchical testing sequence. Those design elements therefore should not be inferred from the reported P-values.
14. What the Odds Ratio Does — and Does Not — Mean
The primary odds ratio of 0.40 indicates lower estimated odds of the Week 8 COVID-19 endpoint in the 4200mg bamlanivimab group than in the placebo group under the fitted logistic regression model.
It does not mean that 40% of participants were protected, that 60% of participants avoided COVID-19 because of treatment, or that the absolute probability of COVID-19 was reduced by 60 percentage points.
Odds ratios are relative measures. To understand the absolute effect on disease probability, one would need the corresponding event rates or model-based predicted probabilities for the comparison groups.
The primary endpoint was not simply detection of SARS-CoV-2. It required RT-PCR detection together with mild or worse disease severity and a specified symptom timing relationship, by Day 57. Therefore, the primary OR should be interpreted specifically for the registered COVID-19 definition.
15. Logistic Regression in More Detail
For a binary outcome, logistic regression models the log odds of the event as a function of treatment and other explanatory variables. The treatment coefficient is typically exponentiated to produce an odds ratio.
An odds ratio below 1 corresponds to a negative treatment coefficient when treatment is coded as the exposed or intervention group relative to the comparator.
This modeling framework is useful when the trial wants to compare binary outcomes while incorporating design-related variables. In BLAZE-2, the registry specifically states that treatment and stratification factors such as facility were included as explanatory variables.
One advantage of regression over a simple two-by-two table is the ability to account for specified explanatory variables in the treatment-effect model. However, the resulting odds ratio remains an odds-based measure and should not be relabeled as a risk ratio without an appropriate statistical derivation.
16. Randomization and Blinding
Randomization and double blinding are important components of the design because they address different sources of bias. Randomization determines treatment assignment according to the trial design, while blinding can reduce the influence of knowledge of treatment assignment on participant behavior, outcome assessment, and other trial processes.
| Design feature | Statistical relevance |
|---|---|
| Randomized allocation | Provides the basis for the treatment-group comparison and supports causal interpretation under the trial assumptions. |
| Double masking | Reduces the potential influence of treatment knowledge on trial conduct and assessment. |
| Parallel design | Participants remain in their randomized treatment groups for the parallel comparison rather than serving as their own controls. |
| Prevention purpose | The binary efficacy endpoint evaluates occurrence of COVID-19 or related infection outcomes after randomization. |
17. Limitations and Interpretation Issues
- Binary effect measure: the primary result is an odds ratio, so it should not be interpreted as a risk ratio or absolute risk reduction.
- Endpoint definition: the primary COVID-19 endpoint combines laboratory detection with disease severity and symptom-timing requirements. It is therefore narrower than SARS-CoV-2 detection alone.
- Multiple endpoints: the trial reports primary and secondary outcomes. The registry information shown here does not specify a multiplicity-adjustment strategy.
- Multiple arms: the registry lists nine arms, while the posted analyses summarized here focus specifically on placebo versus 4200mg bamlanivimab. Results from that comparison should not automatically be generalized to every study arm.
- Regression adjustment: the reported OR is model-based and depends on the explanatory variables included in the logistic regression.
- Absolute risk: the posted statistical analyses shown here provide odds ratios and confidence intervals rather than corresponding absolute event probabilities.
- Safety comparison: serious adverse-event counts are reported descriptively by arm, without a corresponding formal effect estimate in the posted analysis information.
18. What the Registry Does and Does Not Establish
What is directly reported
The registry reports a randomized, double-blind phase 3 prevention study; a primary binary endpoint; logistic regression; odds ratios; confidence intervals; P-values; and serious-adverse-event counts for specified arms.
What requires caution
The odds ratios do not provide absolute risk differences, and the listed nine-arm design should not be collapsed into a two-arm description of the entire trial. The posted primary analysis is specifically the placebo versus 4200mg bamlanivimab comparison.
The statistical interpretation should therefore remain tied to the endpoint, treatment comparison, analysis population, and regression framework actually reported for each result.
19. Why This Trial Matters Statistically
BLAZE-2 is a useful teaching example because it illustrates how a prevention trial with binary outcomes can be analyzed differently from a conventional time-to-event efficacy trial. The central analysis is logistic regression rather than Kaplan-Meier estimation or a Cox proportional-hazards model.
| Concept | How it appears in BLAZE-2 |
|---|---|
| Randomization | The registry identifies the allocation as randomized. |
| Blinding | The study is identified as double-blind. |
| Binary endpoint | The primary endpoint is the percentage of participants with COVID-19 by Week 8. |
| Logistic regression | Used for the posted primary and secondary binary endpoint analyses. |
| Odds ratio | Used as the reported effect measure. |
| Confidence interval | 95% two-sided confidence intervals accompany the posted odds-ratio estimates. |
| P-value | Reported for each posted statistical analysis. |
| Stratification factors | The primary endpoint model includes treatment and stratification factors such as facility. |
| Superiority | The registry identifies the hypothesis type as superiority. |
| Multiple endpoints | One primary endpoint and additional secondary endpoints were posted. |
20. Related Tutorials
Learn more about the methods used in this trial:
21. Related Calculators
22. Sources
- ClinicalTrials.gov: NCT04497987 — BLAZE-2.
- PubMed: PMID 35713300.
- PubMed: PMID 34473343.
- PubMed: PMID 34374951.
- PubMed: PMID 34081073.
Continue through Clinical Biostats
Connect this trial's statistical concepts to deeper biostatistics tutorials, calculators, and additional clinical-trial analyses.
23. Record Summary
BLAZE-2 illustrates the statistical analysis of a randomized prevention trial with binary outcomes. The primary endpoint was the percentage of participants with COVID-19 by Week 8 after randomization, with COVID-19 defined using both RT-PCR detection of SARS-CoV-2 and the specified disease-severity and symptom-timing criteria. The posted primary analysis used logistic regression and reported an odds ratio of 0.40 with a 95% confidence interval of 0.26–0.63 and P < 0.001 for 4200mg bamlanivimab versus placebo.
The secondary analyses used the same general binary-outcome framework. For moderate or worse severity COVID-19 at Week 8, the reported odds ratio was 0.43 with a 95% confidence interval of 0.27–0.67 and P < 0.001. For SARS-CoV-2 at Week 4, the reported odds ratio was 0.66 with a 95% confidence interval of 0.46–0.94 and P = 0.021.
The central statistical lesson is that these are odds ratios from logistic regression, not hazard ratios from survival analysis. Correct interpretation therefore requires attention to the binary endpoint definition, the treatment comparison, the analysis population, the model covariates, the confidence interval, and the distinction between relative odds and absolute probability.