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
EVADE was a randomized, quadruple-masked, parallel-group phase 2/3 trial evaluating ADG20 versus placebo for the prevention of COVID-19. The registry reports 2,582 enrolled participants and two intervention arms.
| Feature | EVADE |
|---|---|
| Brief title | Evaluation of ADG20 for the Prevention of COVID-19 |
| Phase | Phase 2/3 |
| Condition | COVID-19 |
| Therapeutic area | Infectious Disease; Respiratory |
| Primary purpose | Prevention |
| Design | Randomized, parallel-group |
| Masking | Quadruple |
| Enrollment | 2,582 |
| Interventions | ADG20 and placebo |
| Primary endpoint type | Binary |
| Primary endpoints registered | 4 |
| Outcome measures posted | 26 |
| Statistical analyses posted | 6 |
| Lead sponsor | Invivyd, Inc. |
| ClinicalTrials.gov | NCT04859517 |
2. Clinical Question
The central question was whether ADG20 could reduce the occurrence of RT-PCR-confirmed symptomatic COVID-19 compared with placebo during prespecified periods of follow-up.
Population
Participants enrolled in a randomized phase 2/3 prevention trial for COVID-19. The primary efficacy analyses used prespecified modified Full Analysis Sets focused on participants without specified evidence of SARS-CoV-2 infection at baseline.
Intervention
ADG20 administered intramuscularly, with the registry identifying ADG20 as a drug intervention.
Comparator
Placebo administered intramuscularly.
Primary question
Does ADG20 reduce the percentage of participants developing RT-PCR-confirmed symptomatic COVID-19 during the prespecified PEP and PrEP assessment periods?
3. Trial Design
ADG20 IM
- ADG20
- Intramuscular administration
- Evaluated in Cohort A and Cohort B analyses
Placebo IM
- Placebo
- Intramuscular administration
- Evaluated in Cohort A and Cohort B analyses
The registry identifies two principal efficacy settings. PEP analyses used a population without current SARS-CoV-2 infection at baseline, whereas PrEP analyses used a more restrictive population without prior or current SARS-CoV-2 infection at baseline. The pre-Omicron analysis populations were restricted to participants randomized on or prior to 30-Nov-2021 and were designed around follow-up through the emergence of Omicron on 15-Dec-2021.
4. Analysis Populations
| Population | Registry definition and role |
|---|---|
| Pre-Omicron mFAS-1 | Participants without current SARS-CoV-2 infection at baseline based on central tests, defined as baseline RT-PCR-negative regardless of serostatus. Participants were randomized on or prior to 30-Nov-2021, allowing a minimum of 2 weeks of follow-up through 15-Dec-2021. |
| Pre-Omicron mFAS | Participants without prior or current SARS-CoV-2 infection at baseline based on central tests, defined as baseline RT-PCR-negative and seronegative. Participants were randomized on or prior to 30-Nov-2021, allowing a minimum of 2 weeks of follow-up through 15-Dec-2021. |
The distinction between these populations is statistically important. The mFAS-1 definition excludes current infection at baseline while allowing baseline seropositivity, whereas the mFAS definition excludes both prior and current infection using the specified RT-PCR and serology criteria. Consequently, estimates from the two analyses address somewhat different target populations.
5. Primary Endpoints
| Endpoint | Time frame | Definition | Type |
|---|---|---|---|
| Percentage of Participants With RT-PCR Confirmed Symptomatic COVID-19 (PEP) | Day 1 through Day 28 or emergence of Omicron (15-Dec-2021), whichever is earlier | Positive central or local RT-PCR test from a nasopharyngeal or saliva sample accompanied by protocol-defined COVID-19 symptoms occurring within 14 days from sample collection. COVID-19-related hospitalization with a positive local SARS-CoV-2 test within 14 days or all-cause death is counted toward the endpoint. | Binary |
| Percentage of Participants With RT-PCR Confirmed Symptomatic COVID-19 (PrEP) | Day 1 through 3 Months or emergence of Omicron (15-Dec-2021), whichever is earlier | Positive central or local RT-PCR test from a nasopharyngeal or saliva sample accompanied by protocol-defined COVID-19 symptoms occurring within 14 days from sample collection. COVID-19-related hospitalization with a positive local SARS-CoV-2 test within 14 days or all-cause death is counted toward the endpoint. | Binary |
| Solicited Injection Site Reactions (PEP, PrEP) | Day 1 through Day 4 | Percentage of participants with at least one solicited injection site reaction. | Binary |
| Participants With at Least 1 Treatment Emergent Adverse Events (PEP, PrEP) | Day 1 through 14 Months or 25-Jul-2022, whichever is earlier | Any untoward medical occurrence in a clinical investigation participant administered a pharmaceutical product, which does not necessarily have a causal relationship with treatment; includes solicited injection site reactions. | Binary |
The four registered primary endpoints therefore include two efficacy endpoints and two safety/tolerability endpoints. The two efficacy endpoints differ principally in the baseline infection criteria and follow-up duration represented by the PEP and PrEP populations.
6. Statistical Methodology
Logistic regression for binary outcomes
The registry reports logistic regression for the two primary symptomatic COVID-19 analyses and for two secondary binary endpoints. For the primary PEP analysis, the treatment comparison was expressed as a risk difference rather than as a logistic regression odds ratio.
A negative risk difference indicates a lower estimated event proportion in the ADG20 group than in the placebo group. The magnitude is expressed in percentage points when the underlying proportions are expressed as percentages.
Covariate adjustment
For the primary PEP analysis, the registry states that the logistic-regression approach used predefined prognostic factors to obtain a population-level estimate of the treatment difference. The analysis notes that a statistical method described by Ge et al. (2011) was used and that the variance of the estimated treatment difference or relative risk reduction was estimated using the delta method.
This matters because the reported risk difference is an adjusted population-level treatment comparison rather than simply the subtraction of two unadjusted percentages. Covariate adjustment can improve precision when prognostic factors explain variation in the outcome, while the interpretation remains tied to the prespecified analysis population and model.
Log-rank testing for time-to-event outcomes
The registry reports log-rank tests for the two secondary time-to-event analyses evaluating time from randomization to first RT-PCR-confirmed symptomatic COVID-19. The associated effect measure was the hazard ratio.
A hazard ratio below 1 indicates a lower estimated instantaneous rate of first symptomatic COVID-19 in the ADG20 group during the analyzed follow-up.
Intention-to-treat analysis
The registry identifies intention-to-treat analysis as an analysis concept for the reported primary and secondary efficacy comparisons. The principle preserves the randomized treatment assignment as the basis for comparing groups. In this trial, however, the named efficacy populations are modified Full Analysis Sets with specific baseline infection criteria, so "intention-to-treat" should not be interpreted as meaning that every one of the 2,582 enrolled participants necessarily entered every efficacy analysis.
7. Primary Results: PEP Symptomatic COVID-19
The first primary efficacy analysis evaluated RT-PCR-confirmed symptomatic COVID-19 from Day 1 through Day 28 or the emergence of Omicron on 15-Dec-2021, whichever was earlier.
Risk difference: ADG20 vs placebo
95% CI: -8.87 to -1.08 · P = 0.0123
Two-sided 95% confidence interval; superiority hypothesis.
| Feature | Primary PEP analysis |
|---|---|
| Population | Pre-Omicron mFAS-1 |
| Comparison | ADG20 IM - Cohort A vs Placebo IM - Cohort A |
| Endpoint | RT-PCR-confirmed symptomatic COVID-19 |
| Time frame | Day 1 through Day 28 or emergence of Omicron (15-Dec-2021), whichever is earlier |
| Method | Logistic regression |
| Effect measure | Risk difference |
| Estimate | -5.0 |
| 95% CI | -8.87 to -1.08 |
| P-value | 0.0123 |
| Hypothesis | Superiority |
The estimated risk difference of -5.0 means that the estimated proportion experiencing the endpoint was 5.0 percentage points lower in the ADG20 group than in the placebo group, on the scale used for the reported analysis.
The negative direction is consistent with fewer symptomatic COVID-19 events under ADG20 in this analysis population. It does not mean that every participant experienced a 5.0-percentage-point reduction in individual risk, nor does it establish that ADG20 prevented COVID-19 in every person treated.
The two-sided 95% confidence interval from -8.87 to -1.08 describes uncertainty around the estimated treatment difference under the specified analysis framework. Because the entire interval is below zero, the interval is consistent with a lower event proportion for ADG20 relative to placebo.
The P = 0.0123 value addresses compatibility with the null hypothesis under the specified statistical test; it does not measure the size or clinical importance of the treatment effect. The effect size is conveyed by the risk difference and its confidence interval.
The estimate is specific to the pre-Omicron mFAS-1 population and the Day 1 through Day 28-or-Omicron time frame. It should not automatically be generalized to later variants or to populations with different baseline infection status.
Why the risk difference is useful here
A risk difference is an absolute measure. Unlike a relative effect measure, it directly describes the difference in event probability between treatment groups. That makes it particularly useful for prevention endpoints, where the practical meaning of the comparison depends not only on the relative treatment effect but also on the underlying event frequency.
8. Primary Results: PrEP Symptomatic COVID-19
The second primary efficacy analysis evaluated RT-PCR-confirmed symptomatic COVID-19 from Day 1 through 3 Months or the emergence of Omicron on 15-Dec-2021, whichever was earlier.
Risk difference: ADG20 vs placebo
95% CI: -5.75 to -2.01 · P < 0.0001
Two-sided 95% confidence interval; superiority hypothesis.
| Feature | Primary PrEP analysis |
|---|---|
| Population | Pre-Omicron mFAS |
| Comparison | ADG20 IM - Cohort B vs Placebo IM - Cohort B |
| Endpoint | RT-PCR-confirmed symptomatic COVID-19 |
| Time frame | Day 1 through 3 Months or emergence of Omicron (15-Dec-2021), whichever is earlier |
| Method | Logistic regression |
| Effect measure | Risk difference |
| Estimate | -3.9 |
| 95% CI | -5.75 to -2.01 |
| P-value | <0.0001 |
| Hypothesis | Superiority |
The estimated risk difference of -3.9 indicates a 3.9-percentage-point lower estimated event proportion in the ADG20 group than in the placebo group for the analyzed PrEP population and time frame.
The 95% confidence interval, -5.75 to -2.01, remains below zero. It therefore describes an estimated treatment difference consistently in the direction of fewer symptomatic COVID-19 events with ADG20 within the statistical uncertainty represented by this interval.
The reported P < 0.0001 indicates strong statistical evidence against the null hypothesis under the reported analysis. It does not mean that the probability that ADG20 works is less than 0.0001, and it does not quantify the magnitude of benefit. The risk difference and confidence interval provide that information.
As with the PEP result, the estimate is tied to a specific modified Full Analysis Set and a specific pre-Omicron observation period. The more restrictive baseline infection criteria in the mFAS also distinguish this result from the mFAS-1 PEP analysis.
9. Secondary Endpoint Results
Secondary PEP Symptomatic COVID-19 Analysis
The registry also reports a secondary analysis of the PEP symptomatic COVID-19 endpoint using the pre-Omicron mFAS rather than the mFAS-1 used for the corresponding primary analysis.
Standardized risk difference
95% CI: -13.74 to -1.46 · P = 0.0153
| Feature | Secondary PEP analysis |
|---|---|
| Population | Pre-Omicron mFAS |
| Comparison | ADG20 IM - Cohort A vs Placebo IM - Cohort A |
| Method | Logistic regression |
| Effect measure | Standardized risk difference |
| Estimate | -7.6 |
| 95% CI | -13.74 to -1.46 |
| P-value | 0.0153 |
The registry describes this as a population-level treatment difference in proportion adjusted for predefined prognostic factors. The confidence interval remains below zero, while the estimate differs from the primary PEP result because the analysis population and standardized analysis differ.
Secondary SARS-CoV-2 Infection Endpoint
A separate secondary binary endpoint evaluated SARS-CoV-2 infection, asymptomatic or symptomatic, as determined by positive RT-PCR or serology.
Risk difference
95% CI: -9.03 to 0.21 · P = 0.0612
| Feature | Secondary infection analysis |
|---|---|
| Population | Pre-Omicron mFAS-1 |
| Comparison | ADG20 IM - Cohort A vs Placebo IM - Cohort A |
| Endpoint | SARS-CoV-2 infection, asymptomatic or symptomatic, determined by positive RT-PCR or serology |
| Time frame | Day 1 through Day 28 or emergence of Omicron (15-Dec-2021), whichever is earlier |
| Method | Logistic regression |
| Effect measure | Risk difference |
| Estimate | -4.4 |
| 95% CI | -9.03 to 0.21 |
| P-value | 0.0612 |
The point estimate is negative, but the 95% confidence interval extends from a negative value to 0.21. Thus, under the reported analysis, the interval includes no difference. The P-value of 0.0612 is also above conventional 0.05 thresholds, although the meaning of a P-value should not be reduced to a binary significant/non-significant label.
10. Secondary Time-to-Event Results
PEP: Time to First Symptomatic COVID-19
The registry reports a secondary time-to-event analysis of time from randomization to first RT-PCR-confirmed symptomatic COVID-19. Follow-up was an average of 2 months up to 7 months through a cutoff date prior to emergence of Omicron on 15-Dec-2021.
Hazard ratio
95% CI: 0.06–0.76 · P = 0.0077
A hazard ratio of 0.22 means that the estimated instantaneous rate of first RT-PCR-confirmed symptomatic COVID-19 was approximately 22% of the corresponding rate in the placebo group during the analyzed follow-up, under the time-to-event framework.
Equivalently, 0.22 corresponds to an estimated 78% lower hazard in the ADG20 group relative to placebo. This is a relative instantaneous-rate interpretation, not a statement that 78% of participants were protected or that individual risk was reduced by exactly 78%.
The 95% CI of 0.06 to 0.76 reflects substantial uncertainty around the hazard-ratio estimate. It remains below 1, supporting a treatment effect in the direction of fewer events under the reported model.
The P-value of 0.0077 addresses evidence against the null hazard comparison and does not measure the magnitude of the treatment effect. The hazard ratio and confidence interval provide the effect estimate and its precision.
Because the analysis is based on a time-to-event model, interpretation also depends on the way censoring and event times are handled and on the appropriateness of summarizing the treatment comparison with a hazard ratio over follow-up.
PrEP: Time to First Symptomatic COVID-19
The corresponding PrEP time-to-event endpoint evaluated time from randomization to first RT-PCR-confirmed symptomatic COVID-19 over an average of 2.5 months up to 7.5 months through a cutoff date prior to emergence of Omicron on 15-Dec-2021.
Hazard ratio
95% CI: 0.14–0.49 · P < 0.0001
A hazard ratio of 0.27 corresponds to an estimated instantaneous event rate approximately 27% of that under placebo. On a relative hazard scale, that is approximately a 73% lower estimated hazard for first symptomatic COVID-19 in the ADG20 group.
The estimate does not mean that 73% of participants avoided COVID-19, nor does it mean that each participant experienced a 73% reduction in personal probability of infection. Hazard ratios summarize relative event rates over time under a time-to-event analysis.
The 95% CI of 0.14 to 0.49 remains below 1 and provides a range of values compatible with the estimated relative treatment effect under the analysis framework.
The reported P < 0.0001 indicates strong statistical evidence against the null comparison, but it is not a measure of effect size. The effect magnitude is described by the hazard ratio and its confidence interval.
11. Primary and Secondary Results at a Glance
| Endpoint | Role | Effect | 95% CI | P-value |
|---|---|---|---|---|
| RT-PCR-confirmed symptomatic COVID-19, PEP | Primary | RD -5.0 | -8.87 to -1.08 | 0.0123 |
| RT-PCR-confirmed symptomatic COVID-19, PrEP | Primary | RD -3.9 | -5.75 to -2.01 | <0.0001 |
| RT-PCR-confirmed symptomatic COVID-19, PEP | Secondary | Standardized RD -7.6 | -13.74 to -1.46 | 0.0153 |
| SARS-CoV-2 infection, asymptomatic or symptomatic | Secondary | RD -4.4 | -9.03 to 0.21 | 0.0612 |
| Time to first symptomatic COVID-19, PEP | Secondary | HR 0.22 | 0.06 to 0.76 | 0.0077 |
| Time to first symptomatic COVID-19, PrEP | Secondary | HR 0.27 | 0.14 to 0.49 | <0.0001 |
The table illustrates why the choice of effect measure matters. The two primary efficacy analyses use absolute risk differences, whereas the secondary time-to-event analyses use hazard ratios. These estimates should not be compared numerically as if they were interchangeable measures.
12. Statistical Methods Explained
Why was logistic regression used?
The primary symptomatic COVID-19 endpoints are binary: each participant either meets the endpoint definition during the specified period or does not. Logistic regression is therefore a natural modeling framework for binary outcomes. In EVADE, the registry identifies logistic regression and, for the PEP analysis, describes adjustment for predefined prognostic factors.
What does a risk difference of -5.0 mean?
A risk difference of -5.0 means the estimated event proportion in the ADG20 group was 5.0 percentage points lower than the corresponding estimated proportion in the placebo group for the analyzed population and period. It is an absolute effect measure, not a relative risk or hazard ratio.
Why does the confidence interval matter?
The point estimate is only one estimate of the treatment difference. The 95% confidence interval describes statistical uncertainty around that estimate under the specified model and sampling framework. For the primary PEP analysis, the interval from -8.87 to -1.08 remains below zero; for the secondary infection analysis, the interval from -9.03 to 0.21 includes zero.
What does a hazard ratio of 0.22 mean?
A hazard ratio of 0.22 indicates that the estimated instantaneous event rate under ADG20 was approximately 22% of the rate under placebo during the analyzed follow-up. It can also be expressed as approximately a 78% lower estimated hazard. It does not mean that 78% of participants were protected.
Why use a log-rank test for time-to-event endpoints?
The log-rank test compares the timing of events between treatment groups while accounting for censoring. That makes it suitable for endpoints such as time from randomization to first symptomatic COVID-19, where not every participant necessarily experiences the event during observation.
Why are the PEP and PrEP analyses different?
The registry defines different analysis populations and follow-up windows. The PEP primary analysis uses the pre-Omicron mFAS-1, which requires baseline RT-PCR negativity regardless of serostatus. The PrEP primary analysis uses the pre-Omicron mFAS, which requires both baseline RT-PCR negativity and seronegativity. Their time frames also differ: Day 28 versus 3 months, respectively, with both constrained by emergence of Omicron on 15-Dec-2021.
13. Confidence Intervals, P-values, and Effect Size
The EVADE results demonstrate three related but distinct statistical ideas: effect size, precision, and evidence against a null hypothesis.
Effect size
The risk difference or hazard ratio describes the estimated magnitude and direction of the treatment comparison.
Precision
The confidence interval shows how uncertain the estimated treatment effect is under the statistical framework used for the analysis.
P-value
The P-value evaluates compatibility of the observed data with a specified null hypothesis. It is not a probability that the null hypothesis is true.
Clinical meaning
Statistical evidence does not by itself determine the practical importance of an effect. Absolute event differences, endpoint definitions, follow-up, and population all matter.
For example, the primary PEP risk difference is -5.0 with a 95% CI of -8.87 to -1.08. The estimate communicates the size of the observed treatment difference, while the interval communicates its uncertainty. The P-value of 0.0123 provides a different piece of information: evidence against the corresponding null hypothesis under the specified analysis.
The same logic applies to the PrEP analysis, where the risk difference is -3.9 with a 95% CI of -5.75 to -2.01 and P < 0.0001.
14. Time-to-Event Interpretation
The two secondary time-to-event analyses complement the binary endpoint analyses by incorporating the timing of symptomatic COVID-19 events rather than reducing the observation period to a simple event/no-event classification.
For EVADE, the reported hazard ratios were 0.22 for the PEP analysis and 0.27 for the PrEP analysis. These are relative time-to-event measures and should not be interpreted as absolute risk reductions.
Time-to-event methods also permit participants who do not experience symptomatic COVID-19 during their available observation period to contribute information until censoring. This is one reason a survival-analysis framework can be informative in prevention trials with differing follow-up times.
The hazard ratio should nevertheless be interpreted as a model-based summary. A single hazard ratio is most straightforward when the relative hazards are reasonably represented by a stable proportional relationship over time. The registry does not provide a separate assessment of the proportional-hazards assumption in the information reported here.
15. Safety Results
The registry reports serious adverse events by arm for four cohort-specific groups.
| Group | Serious adverse events affected | At risk |
|---|---|---|
| ADG20 IM - Cohort A | 4 | 240 |
| Placebo IM - Cohort A | 8 | 241 |
| ADG20 IM - Cohort B | 46 | 1,001 |
| Placebo IM - Cohort B | 46 | 1,001 |
These counts describe serious adverse events by arm and cohort. They should not be conflated with the separate registered primary endpoint of participants with at least one treatment-emergent adverse event, whose registry definition includes solicited injection site reactions.
Cohort A
The registry reports 4 serious adverse-event-affected participants among 240 in the ADG20 group and 8 among 241 in the placebo group.
Cohort B
The registry reports 46 serious adverse-event-affected participants among 1,001 in each of the ADG20 and placebo groups.
Serious adverse-event counts are descriptive safety information here. A complete comparative safety analysis would require the relevant analysis population, event definitions, exposure information, and the prespecified statistical framework for safety comparisons.
16. The Role of Baseline Infection Status
Baseline infection status is central to the interpretation of EVADE's efficacy analyses because the endpoint concerns prevention of symptomatic COVID-19 after randomization.
| Analysis | Baseline criterion | Why it matters |
|---|---|---|
| mFAS-1 | Baseline RT-PCR-negative regardless of serostatus | Excludes current infection while allowing participants who may have evidence of prior exposure. |
| mFAS | Baseline RT-PCR-negative and seronegative | Excludes both current and prior infection according to the specified central testing criteria. |
These definitions create different estimands. The mFAS-1 analysis asks about symptomatic COVID-19 among participants without current infection, while the mFAS analysis narrows the population further by excluding baseline seropositivity. That distinction is important when comparing the resulting risk differences or hazard ratios.
17. The Omicron Cutoff and Interpretation of Follow-Up
Several efficacy endpoints use the emergence of Omicron on 15-Dec-2021 as an upper boundary for the prespecified pre-Omicron analysis period.
Trial start
The registry lists 27-Apr-2021 as the trial start date.
Randomization cutoff for pre-Omicron efficacy populations
The reported mFAS and mFAS-1 analyses include participants randomized on or prior to 30-Nov-2021.
Emergence of Omicron
The registry uses this date as the cutoff for several pre-Omicron efficacy endpoint definitions.
Safety endpoint cutoff
The treatment-emergent adverse-event endpoint extends through 14 months or 25-Jul-2022, whichever is earlier.
Primary completion
The registry lists 04-Nov-2022 as the primary completion date.
The cutoff is statistically consequential because treatment effects are conditional on the population, calendar period, circulating virus context, and follow-up represented in the analysis. A pre-Omicron estimate should therefore not be interpreted as an unrestricted estimate across every subsequent phase of the pandemic.
18. Multiplicity and Multiple Endpoints
The registry lists four primary endpoints: two symptomatic COVID-19 efficacy endpoints and two safety/tolerability endpoints. The registry also reports six posted statistical analyses, of which two are primary-endpoint analyses.
| Endpoint group | Role | Statistical type |
|---|---|---|
| PEP symptomatic COVID-19 | Primary | Binary; logistic regression; risk difference |
| PrEP symptomatic COVID-19 | Primary | Binary; logistic regression; risk difference |
| Solicited injection site reactions | Primary | Binary |
| Treatment-emergent adverse events | Primary | Binary |
Multiple endpoints create a broader statistical context than any single P-value. Whether and how type I error is controlled across multiple primary endpoints depends on the prespecified testing strategy. The ClinicalTrials.gov record identifies the endpoints and reported analyses but does not provide an alpha-allocation or hierarchical testing scheme.
Accordingly, individual P-values should be read as the results of their specified analyses rather than automatically treating every endpoint as an independent confirmatory hypothesis test with an identical error-control interpretation.
19. Why the Primary Endpoint Is Binary
The primary efficacy endpoint asks whether participants develop RT-PCR-confirmed symptomatic COVID-19 during a specified interval. This naturally produces a binary outcome: endpoint met or endpoint not met during the defined period.
Binary analysis
Logistic regression compares the probability of the binary endpoint between randomized treatment groups while permitting covariate adjustment.
Risk difference
The reported effect measure translates the model into an absolute difference in event probability between ADG20 and placebo.
Time-to-event analysis
The secondary analysis retains information about when the first symptomatic COVID-19 event occurred and uses a hazard ratio to compare event rates over time.
Complementary evidence
Binary and time-to-event analyses answer related but distinct questions and should be interpreted on their own scales.
20. What the Risk Difference Does — and Does Not — Mean
A risk difference of -3.9 means that the estimated proportion experiencing the specified endpoint was 3.9 percentage points lower in the ADG20 group than in the placebo group for that analysis.
It does not mean that every individual participant's probability was reduced by exactly 3.9 percentage points, nor does it imply that all participants receive the same absolute benefit.
A 95% confidence interval describes uncertainty around the estimated risk difference. The interval from -5.75 to -2.01 indicates a narrower range of compatible treatment differences than the interval from -13.74 to -1.46 reported for the secondary standardized risk difference.
A P-value such as <0.0001 is evidence against a specified null hypothesis under the reported statistical model. It is not the probability that the treatment is effective, and it is not a measure of how large the effect is.
21. What the Hazard Ratio Does — and Does Not — Mean
The hazard ratio of 0.22 indicates an estimated instantaneous event rate approximately 22% of that in the placebo group. This corresponds to approximately a 78% lower estimated hazard under the model.
The hazard ratio of 0.27 indicates an estimated instantaneous event rate approximately 27% of that in the placebo group. This corresponds to approximately a 73% lower estimated hazard under the model.
Neither hazard ratio is an absolute risk reduction, and neither states the proportion of participants who avoided COVID-19. Absolute event probabilities and risk differences answer a different question.
22. Important Limitations and Interpretation Issues
- Pre-Omicron analysis: several important efficacy analyses are explicitly bounded by emergence of Omicron on 15-Dec-2021. Their results therefore describe the specified pre-Omicron analysis period.
- Modified analysis populations: the PEP and PrEP primary analyses use different baseline infection criteria, so their estimates apply to different analysis populations.
- Different time frames: the PEP primary endpoint runs through Day 28 or emergence of Omicron, while the PrEP primary endpoint runs through 3 months or emergence of Omicron. The estimates should not be compared as if they represented identical follow-up.
- Binary versus time-to-event measures: risk differences and hazard ratios are different effect measures and cannot be interpreted as interchangeable numerical quantities.
- Covariate adjustment: the primary PEP analysis uses predefined prognostic factors, so its adjusted risk difference should not be assumed to equal a simple unadjusted difference between raw event proportions.
- Hazard-ratio assumptions: the registry reports hazard ratios from log-rank analyses but does not provide a separate assessment of proportional hazards in the reported information.
- Multiple endpoints: four primary endpoints are registered. The statistical interpretation of individual P-values depends on the prespecified multiplicity strategy, and the registry information here does not provide an alpha-allocation scheme.
- Safety population context: serious adverse-event counts are reported by cohort and arm and should not be substituted for the separate treatment-emergent adverse-event primary endpoint.
- Generalizability: the efficacy estimates apply specifically to the registry-defined analysis populations and follow-up periods.
23. Why This Trial Matters Statistically
EVADE is a useful statistical teaching case because it combines randomized prevention-trial design with both absolute and time-to-event efficacy measures. The primary efficacy analyses use adjusted risk differences from logistic regression, while secondary analyses use log-rank tests and hazard ratios to incorporate event timing.
| Concept | How it appears in EVADE |
|---|---|
| Randomization | Participants were randomized to ADG20 or placebo in a parallel-group design. |
| Blinding | The registry identifies quadruple masking. |
| Binary endpoints | The primary efficacy endpoints classify whether symptomatic COVID-19 occurred during defined periods. |
| Logistic regression | Used for the reported primary efficacy analyses and selected secondary binary outcomes. |
| Risk difference | The two primary efficacy analyses use risk difference as the effect measure. |
| Covariate adjustment | The primary PEP analysis adjusted for predefined prognostic factors. |
| Intention-to-treat principle | Identified as an analysis concept for the reported efficacy analyses. |
| Log-rank test | Used for secondary time-to-event analyses of time to symptomatic COVID-19. |
| Hazard ratio | Used to quantify the relative treatment effect in the time-to-event analyses. |
| Confidence intervals | Reported as two-sided 95% intervals for the posted efficacy analyses. |
| Multiple endpoints | Four primary endpoints are registered. |
| Variant-era cutoff | Several efficacy analyses use 15-Dec-2021, the emergence of Omicron, as a prespecified boundary. |
24. Clinical Biostats Interpretation
The most important statistical feature of EVADE is the convergence of several analysis perspectives without treating them as interchangeable. The primary PEP and PrEP analyses report absolute risk differences, while the secondary time-to-event analyses report hazard ratios. Together, these measures describe treatment differences from different statistical angles.
The primary PEP risk difference of -5.0 has a two-sided 95% confidence interval of -8.87 to -1.08. The primary PrEP risk difference of -3.9 has a corresponding interval of -5.75 to -2.01. Both intervals are below zero, but the estimates arise from different analysis populations and time frames.
The secondary time-to-event analyses report hazard ratios of 0.22 and 0.27. These estimates indicate lower estimated instantaneous rates of first symptomatic COVID-19 under ADG20 in the respective analyses. Their interpretation is conditional on the time-to-event framework rather than representing absolute prevention probabilities.
The secondary SARS-CoV-2 infection analysis provides an important counterpoint: its risk difference is -4.4, but the 95% confidence interval extends from -9.03 to 0.21, and the P-value is 0.0612. This demonstrates why the point estimate, confidence interval, and P-value should be considered together rather than relying on the direction of the point estimate alone.
Finally, the baseline definitions and calendar cutoff matter. The registry-defined efficacy analyses focus on participants meeting specific infection-status criteria and on a pre-Omicron period. Statistical interpretation therefore depends not only on the estimated treatment effect but also on precisely who was analyzed and when the outcome was observed.
25. Related Tutorials
Learn more about the methods used in this trial:
26. Related Calculators
27. Sources
- ClinicalTrials.gov: EVADE, NCT04859517.
- PubMed: PMID 37496612.
- PubMed: PMID 35713300.
- PubMed: PMID 34473343.
Continue with Clinical Biostats
Connect the statistical methods used in EVADE to deeper tutorials, statistical calculators, and additional clinical trial analyses.
28. Record Summary
EVADE provides a useful example of statistical analysis for a randomized prevention trial in which both binary and time-to-event endpoints were used. The primary efficacy analyses used logistic regression with risk difference as the effect measure, while secondary time-to-event analyses used log-rank testing with hazard ratios. The interpretation depends critically on the prespecified modified Full Analysis Sets, the distinction between PEP and PrEP populations, and the pre-Omicron cutoff used for several efficacy analyses.
The primary PEP analysis reported a risk difference of -5.0 (95% CI -8.87 to -1.08; P = 0.0123), while the primary PrEP analysis reported a risk difference of -3.9 (95% CI -5.75 to -2.01; P < 0.0001). Secondary time-to-event analyses reported hazard ratios of 0.22 (95% CI 0.06–0.76; P = 0.0077) for PEP and 0.27 (95% CI 0.14–0.49; P < 0.0001) for PrEP.
The secondary infection endpoint illustrates the importance of uncertainty: its risk difference was -4.4, but its 95% confidence interval ranged from -9.03 to 0.21. This is a useful reminder that treatment-effect interpretation should incorporate the full confidence interval and the analysis population rather than relying solely on the direction of an estimate.