This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial facts on this page are limited to the ClinicalTrials.gov record data. The ClinicalTrials.gov record contains no formal statistical-analyses results for the primary endpoint.
1. Trial at a Glance
STOP COVID 2 was a randomized, parallel, triple-masked phase 3 treatment trial evaluating fluvoxamine versus placebo in participants with Covid-19 and Coronavirus as the registered conditions. The study enrolled 670 participants and registered one binary primary endpoint measured over approximately 15 days.
| Feature | STOP COVID 2 |
|---|---|
| Trial name | STOP COVID 2 |
| Brief title | Fluvoxamine for Early Treatment of Covid-19 (Stop Covid 2) |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Infectious Disease |
| Conditions | Covid19; Coronavirus |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 670.0 |
| Interventions | Fluvoxamine; Placebo |
| Lead sponsor | Washington University School of Medicine |
| Results posted | Yes |
| ClinicalTrials.gov | NCT04668950 |
2. Clinical Question
The registered trial question can be expressed statistically as whether randomized assignment to fluvoxamine, compared with placebo, changes the occurrence of the prespecified clinical deterioration endpoint during approximately 15 days of follow-up.
Population
Participants enrolled in the STOP COVID 2 phase 3 trial for the registered conditions Covid19 and Coronavirus.
Intervention
Fluvoxamine, registered as a drug intervention.
Comparator
Placebo, registered as a drug intervention.
Primary question
Does randomized assignment to fluvoxamine versus placebo affect the number of participants experiencing clinical deterioration during approximately 15 days?
3. Trial Design
Fluvoxamine
- Fluvoxamine registered as a drug intervention.
- Participants were assigned within the randomized parallel design.
- The primary endpoint was assessed over approximately 15 days.
Placebo
- Placebo registered as a drug intervention.
- Participants were assigned within the randomized parallel design.
- The primary endpoint was assessed over approximately 15 days.
4. Trial Timeline
Study start
The registered study start date was December 22, 2020.
Primary completion
The registered primary completion date was September 28, 2021.
Registry status
The ClinicalTrials.gov record classifies STOP COVID 2 as completed.
5. Primary Endpoint
| Endpoint | Registered definition | Time frame | Type |
|---|---|---|---|
| Number of Participants With Clinical Deterioration | Defined as the number of participants who experienced both of the following: 1) presence of dyspnea and/or hospitalization for shortness of breath or pneumonia, and 2) decrease in O2 saturation (<92% on room air) and/or supplemental oxygen requirement to keep O2 saturation ≥92%. | RCT-approximately 15 days | Binary |
The endpoint is explicitly registered as binary. Conceptually, each participant contributes one of two endpoint states during the specified assessment window: clinical deterioration meeting the registry definition, or no clinical deterioration meeting that definition within the relevant follow-up framework.
The registry defines the endpoint through a combination of respiratory symptoms or hospitalization and oxygenation-related criteria. The statistical analysis therefore concerns the probability or odds of experiencing the prespecified event, rather than a continuous measurement such as oxygen saturation itself.
6. Results Reporting Status
This distinction is important. A posted outcome measure does not necessarily mean that the ClinicalTrials.gov record contains the complete inferential analysis required to quantify the difference between randomized groups.
7. Planned Analysis
Because the primary endpoint is binary and is assessed over a defined follow-up period, the natural primary comparison would be a comparison of the proportion of participants experiencing clinical deterioration in the fluvoxamine and placebo groups.
Risk difference
The absolute difference in event proportions directly describes how much the observed probability of clinical deterioration differs between randomized groups.
Risk ratio
The ratio of event probabilities describes the relative frequency of clinical deterioration in the fluvoxamine group compared with the placebo group.
Odds ratio
An odds ratio compares the odds of clinical deterioration between randomized groups. It should not be described as though it were a risk ratio when the event is not rare.
Confidence interval
An interval around the chosen effect estimate would quantify statistical uncertainty and would be essential for interpreting the precision of the treatment comparison.
What a formal analysis would normally report
For a randomized trial with a binary primary endpoint, a complete statistical presentation would ordinarily identify the number and proportion of participants with the event in each randomized arm, specify the treatment-effect measure, provide a confidence interval, and report the prespecified hypothesis test or corresponding p-value when applicable.
The ClinicalTrials.gov record does not identify which particular inferential method was prespecified for STOP COVID 2. Therefore, this page describes the statistical framework appropriate to a binary randomized endpoint rather than attributing an unreported method to the trial.
8. Statistical Interpretation of a Binary Clinical Endpoint
A negative risk difference would indicate a lower observed event probability in the fluvoxamine group; a positive value would indicate a higher observed event probability. The direction alone does not establish statistical significance.
A risk ratio of 1 would indicate equal event probabilities. A value below 1 would indicate a lower event probability in the fluvoxamine group, while a value above 1 would indicate a higher event probability.
The odds ratio is based on odds rather than probabilities. When events are not rare, the numerical value of an odds ratio can differ materially from the corresponding risk ratio.
For this endpoint, the most transparent clinical presentation would generally place the event count and event proportion beside an absolute effect measure, while also reporting a relative measure when prespecified or otherwise appropriate. The ClinicalTrials.gov record does not contain the event counts for clinical deterioration in either randomized arm, so those quantities cannot be reconstructed here.
9. Safety Results
The ClinicalTrials.gov record reports serious adverse events by arm. These are presented as affected participants divided by the number at risk in the corresponding arm.
| Safety measure | Fluvoxamine | Placebo |
|---|---|---|
| Serious adverse events, affected / at risk | 13 / 272 | 12 / 275 |
The serious-adverse-event denominators reported by the registry are 272 and 275, respectively. They are not the same as the overall enrollment of 670, so the safety figures should be reported exactly with their registry-reported denominators rather than being presented as though every enrolled participant contributed to these particular safety denominators.
10. Statistical Methods Explained
Why is the primary endpoint binary?
The registered endpoint is based on whether a participant experiences a prespecified clinical deterioration event during approximately 15 days. That naturally produces a binary outcome: the participant either meets the endpoint definition or does not. Binary endpoints are analyzed differently from continuous outcomes such as a laboratory measurement or change from baseline.
Why does randomization matter for a binary endpoint?
Randomization creates the treatment groups before the outcome is observed. In a well-conducted randomized comparison, the treatment assignment mechanism provides the foundation for attributing differences in the endpoint distribution to treatment assignment rather than to systematic differences created by the assignment process itself.
What does triple masking contribute?
Triple masking is a registered design feature intended to reduce opportunities for knowledge of treatment assignment to influence aspects of trial conduct or outcome assessment. The ClinicalTrials.gov record identifies the masking level but does not specify which three groups or roles were masked.
Why report a confidence interval for a binary treatment effect?
A point estimate alone gives only one numerical summary of the observed treatment difference. A confidence interval communicates how precisely the effect has been estimated under the specified statistical framework. A narrow interval indicates greater statistical precision than a wide interval, although precision and clinical importance are different concepts.
Why is a p-value not a measure of effect size?
A p-value addresses compatibility of the observed data with a specified null hypothesis under the chosen statistical model. It does not state how large the treatment difference is. Effect size should therefore be communicated through quantities such as a risk difference, risk ratio, or odds ratio, accompanied by uncertainty measures.
Would logistic regression necessarily be the analysis?
Not necessarily. Logistic regression is one standard method for binary outcomes, but the appropriate primary analysis depends on the prespecified statistical analysis plan. A trial may instead use a direct comparison of proportions, a stratified method, a model incorporating prespecified covariates, or another approach. The ClinicalTrials.gov record does not specify the formal statistical analysis.
Why should the approximately 15-day time frame be retained?
The time frame defines the observation window for the registered endpoint. Changing it would change the question being answered. A binary event rate at one follow-up window is not interchangeable with a cumulative event rate measured over a different period.
11. Randomization and Masking as Sources of Statistical Protection
STOP COVID 2 was registered as randomized, with a parallel design and triple masking. These are not simply descriptive labels; they are design mechanisms that affect the credibility of the treatment comparison.
Randomization and masking address different sources of bias. Randomization concerns how treatment assignment is generated; masking concerns knowledge of that assignment during the conduct of the study. Neither feature eliminates every possible source of uncertainty, and neither substitutes for prespecified outcome definitions and appropriate statistical analysis.
12. What Is and Is Not Reported in the Supplied Record
| Item | Supplied registry information |
|---|---|
| Primary endpoint | Yes — one binary endpoint is registered. |
| Primary endpoint definition | Yes — clinical deterioration is explicitly defined. |
| Primary endpoint time frame | Yes — RCT-approximately 15 days. |
| Results posted | Yes. |
| Outcome measures posted | 1. |
| Formal statistical analyses | None in the posted results. |
| Primary treatment-effect estimate | Not reported. |
| Primary confidence interval | Not reported. |
| Primary p-value | Not reported. |
| Serious adverse events by arm | Yes — 13/272 for fluvoxamine and 12/275 for placebo. |
This distinction prevents a common statistical reporting error: treating the existence of a posted outcome measure as equivalent to having a complete inferential analysis. The available data support description of the endpoint and design, but they do not support fabrication of a treatment-effect estimate that is absent from the registry-reported statistical-analyses field.
13. Limitations
- No formal primary analysis reported: no formal statistical analyses were posted, so the primary treatment-effect estimate, confidence interval, and p-value cannot be reported from the ClinicalTrials.gov record.
- Results-posted status is not sufficient to reconstruct the analysis: the record states that results are posted and one outcome measure is posted, but the ClinicalTrials.gov record does not include the numerical primary endpoint results.
- No treatment-effect reconstruction: without clinical-deterioration event counts by randomized arm, absolute and relative treatment effects cannot be calculated without introducing unsupported numbers.
- Safety denominator distinction: serious adverse event data use denominators of 272 and 275, which differ from the overall enrollment of 670.
- Method not attributed beyond the registry: although several statistical methods could reasonably analyze a binary endpoint, the ClinicalTrials.gov record does not identify which formal inferential method was prespecified.
- Masking details are limited: the record identifies triple masking but does not specify the three masked roles in the ClinicalTrials.gov record.
- No subgroup results: the ClinicalTrials.gov record contains no subgroup estimates, confidence intervals, or interaction analyses.
- No longitudinal efficacy results: the ClinicalTrials.gov record does not contain additional follow-up estimates, survival analyses, or other time-to-event results.
14. Why This Trial Matters Statistically
STOP COVID 2 provides a useful teaching example of a randomized clinical trial built around a clinically defined binary endpoint. Its statistical structure is different from trials whose primary outcomes are continuous measurements or time-to-event endpoints.
| Concept | How it appears in STOP COVID 2 |
|---|---|
| Randomization | The allocation is registered as randomized. |
| Parallel design | The design model is registered as parallel. |
| Masking | The study is registered as triple masked. |
| Binary endpoint | The primary endpoint type is registered as binary. |
| Clinical endpoint definition | Clinical deterioration requires both respiratory/hospitalization and oxygenation-related criteria. |
| Fixed assessment window | The primary endpoint uses an RCT-approximately 15-day time frame. |
| Absolute treatment effect | A risk difference would directly quantify the difference in event probability. |
| Relative treatment effect | A risk ratio or odds ratio could describe the relative comparison if specified by the analysis plan. |
| Confidence interval | Would quantify uncertainty around the chosen treatment-effect estimate. |
| Safety analysis | Serious adverse events are reported by arm with affected/at-risk counts. |
The educational value is particularly strong because the absence of a registry-reported formal statistical analysis makes clear why statistical interpretation must be grounded in actual reported estimates. A reader should never substitute a remembered result, independently recomputed value, or assumed statistical model for the analysis actually reported by the trial record.
15. A Framework for Reading the Primary Endpoint
The event is clinical deterioration, not merely an isolated oxygen saturation measurement or a hospitalization by itself. The registry requires both components of its definition to be present.
The registered time frame is RCT-approximately 15 days. The event proportion therefore needs to be interpreted within that specified trial window.
The relevant comparison is between participants assigned to fluvoxamine and participants assigned to placebo. The randomized comparison should remain the organizing principle of the efficacy analysis.
A complete analysis would report event counts and proportions in each arm, followed by an absolute and/or relative treatment-effect estimate with a confidence interval.
A numerical treatment effect should be reported only when supported by the trial record. In this the ClinicalTrials.gov recordset, the formal statistical-analysis results are absent, so the correct interpretation is limited to the registered endpoint, design, and reported safety counts.
16. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial uses a randomized, parallel, triple-masked phase 3 design and a binary primary endpoint measured over approximately 15 days. The registry-reported statistical-analyses field does not contain a formal effect estimate, confidence interval, or p-value.
Clinical interpretation
The registered clinical question concerns whether participants experience the defined form of clinical deterioration during the specified trial window. The available data do not provide the numerical efficacy result needed to quantify the difference between the two randomized groups.
17. Serious Adverse Events: How to Read the Reported Counts
The reported serious adverse-event figures are 13/272 for fluvoxamine and 12/275 for placebo. The numerator represents affected participants and the denominator represents participants at risk according to the registry field.
These figures are descriptive safety data. They do not by themselves establish whether the difference between groups is statistically significant or clinically meaningful.
A formal safety comparison would require a prespecified statistical framework and consideration of the number and type of safety endpoints being examined. Multiple safety events can also be analyzed at different levels, including participants with at least one event, individual event types, event severity, or event timing. None of those additional analyses is reported here.
18. Multiplicity, Interim Analysis, and Other Design Topics
The ClinicalTrials.gov record does not report a non-inferiority margin, crossover design, factorial design, multiplicity procedure, interim-analysis method, missing-data or imputation method, stratification factors, or Bayesian analysis.
| Design topic | Supported by the ClinicalTrials.gov record? | Interpretation |
|---|---|---|
| Non-inferiority margin | No | No non-inferiority margin is reported. |
| Crossover | No | No crossover feature is reported. |
| Factorial design | No | The registry identifies a parallel design with two arms. |
| Multiplicity adjustment | No | No multiplicity procedure is reported. |
| Interim analysis | No | No interim-analysis method is reported. |
| Missing-data / imputation method | No | No imputation method is reported. |
| Stratification | No | No stratification factors are reported. |
| Bayesian methods | No | No Bayesian method is reported. |
These omissions are important because statistical methods should be attributed to the trial only when they are documented. A generic method that is common in clinical-trial practice should not be presented as though it were necessarily the method used in STOP COVID 2.
19. What a Complete Binary-Endpoint Result Would Look Like
For educational purposes, a complete report of the primary endpoint would normally allow the reader to move from raw event counts to an interpretable treatment effect.
The registry-reported STOP COVID 2 data stop before the numerical efficacy result in this sequence.
20. Primary Endpoint: What the Definition Means
The clinical deterioration endpoint is intentionally composite within its definition. The first component concerns dyspnea and/or hospitalization for shortness of breath or pneumonia. The second concerns oxygen saturation below 92% on room air and/or supplemental oxygen requirement to keep oxygen saturation at or above 92%.
| Component | Registered criterion |
|---|---|
| Clinical/respiratory component | Presence of dyspnea and/or hospitalization for shortness of breath or pneumonia |
| Oxygenation component | Decrease in O2 saturation (<92% on room air) and/or supplemental oxygen requirement to keep O2 saturation ≥92% |
| Overall endpoint rule | Both components must be experienced |
| Endpoint type | Binary |
| Time frame | RCT-approximately 15 days |
Because both components are required, the endpoint should not be casually described as "hospitalization," "dyspnea," or "low oxygen saturation." Those are components of the registered definition rather than interchangeable names for the primary endpoint.
21. Why This Trial Is a Useful Binary-Endpoint Example
Many clinical-trial readers are accustomed to survival endpoints such as overall survival or progression-free survival. STOP COVID 2 illustrates a different statistical pattern: a prespecified clinical event assessed over a relatively short, defined window.
Unlike a time-to-event endpoint
The registry-reported primary endpoint is registered as binary. The record does not provide a time-to-event analysis, hazard ratio, or survival curve for this endpoint.
Unlike a continuous endpoint
The primary outcome is not a numerical scale analyzed through a mean difference or similar continuous-outcome statistic.
Defined clinical event
The endpoint combines respiratory/hospitalization criteria with oxygenation criteria, making the exact definition central to interpretation.
Randomized comparison
The intervention and placebo groups are compared within a randomized phase 3 parallel design.
22. Sources
- ClinicalTrials.gov: STOP COVID 2 (NCT04668950). Official registry record for the trial design, endpoints, status, enrollment, interventions, and posted outcome information.
- Linked publication identified in the registry data: PubMed PMID 37622035. The ClinicalTrials.gov record identifies this publication as linked to STOP COVID 2; no additional numerical trial facts from the publication are incorporated into this page.
Continue through the Clinical Biostats statistical pathway
Use the trial design as a starting point for learning how randomization and blinding affect the interpretation of clinical-trial evidence, then connect those concepts to statistical tools.
23. Record Summary
STOP COVID 2 is a completed randomized, parallel, triple-masked phase 3 trial of fluvoxamine versus placebo for the registered conditions Covid19 and Coronavirus. The study enrolled 670 participants and registered one binary primary endpoint: the number of participants with clinical deterioration during an RCT-defined period of approximately 15 days. The endpoint requires both a clinical/respiratory component and an oxygenation component.
The ClinicalTrials.gov record states that results are posted, but they contain no entries in the formal the posted statistical analyses. Consequently, the available evidence supports a detailed discussion of the endpoint, design, appropriate binary-outcome methods, and reported serious adverse-event counts, but it does not support reporting an unverified efficacy estimate, confidence interval, or p-value.