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
COVID-19 PEP was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating hydroxychloroquine versus placebo in the setting of post-exposure prophylaxis / preemptive therapy for SARS-CoV-2. The registry reports an enrollment of 1312 participants and two primary endpoints assessed over 14 days.
| Feature | COVID-19 PEP |
|---|---|
| Trial name | COVID-19 PEP |
| Brief title | Post-exposure Prophylaxis / Preemptive Therapy for SARS-Coronavirus-2 |
| Phase | Phase 3 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 1312 |
| Interventions | Hydroxychloroquine; placebo |
| Lead sponsor | University of Minnesota |
| Status | Completed |
| Study start | March 17, 2020 |
| Primary completion | May 20, 2020 |
| ClinicalTrials.gov | NCT04308668 |
2. Clinical Question
The central question was whether hydroxychloroquine, compared with placebo, affected the development of active COVID-19 disease among participants who were asymptomatic at baseline and affected the change in disease severity over 14 days among participants who were symptomatic at baseline.
Population
Participants enrolled in the phase 3 COVID-19 PEP trial for post-exposure prophylaxis / preemptive therapy for SARS-CoV-2.
Intervention
Hydroxychloroquine.
Comparator
Placebo.
Primary questions
Did the treatment groups differ in active COVID-19 disease at Day 14 among participants asymptomatic at baseline, and did they differ in change in disease severity over 14 days among participants symptomatic at baseline?
3. Trial Design
Hydroxychloroquine
- Hydroxychloroquine
- Randomized treatment assignment
- Quadruple-masked trial design
Placebo
- Placebo
- Randomized comparator assignment
- Quadruple-masked trial design
The design is a conventional two-arm randomized parallel comparison, with masking intended to reduce the influence of treatment knowledge on participant behavior, clinical assessment, and other aspects of trial conduct. The registry identifies the study as quadruple-masked but does not provide, in the information summarized here, a breakdown of which four participant or study roles were masked.
4. Trial Population and Analysis Structure
The two primary endpoints address different baseline clinical states. This distinction is statistically important because the first endpoint is restricted to participants who were asymptomatic at baseline, whereas the second concerns participants who were symptomatic at baseline.
| Primary endpoint population | Outcome | Time frame | Endpoint type |
|---|---|---|---|
| Participants asymptomatic at baseline | Number of participants with active COVID-19 disease at Day 14 | 14 days | Binary |
| Participants symptomatic at baseline | Change in disease severity over 14 days | Baseline and 14 days | Continuous / longitudinal |
This separation means the two primary analyses are not simply two measurements of the same outcome. The first asks about a disease-status event at a defined time point. The second evaluates change in a 0–10 symptom-severity scale over time. The registry therefore reports different statistical methods for the two endpoints.
5. Primary Endpoints
| Endpoint | Registry definition | Time frame | Statistical method | Effect measure |
|---|---|---|---|---|
| Number of Participants With Active COVID-19 Disease at Day 14 Among Those Who Were Asymptomatic at Baseline | Number of participants at 14 days post enrollment with active COVID19 disease among those who were asymptomatic at baseline. | 14 days | Fisher exact test | Mean Difference (Net); values represent percentages |
| Change in Disease Severity Over 14 Days Among Those Who Are Symptomatic at Baseline | Visual Analog Scale 0-10 score of rating overall symptom severity (0 = no symptoms; 10 = most severe) | Baseline and 14 days | Mixed Models Analysis | Mean Difference (Net) |
The two endpoints illustrate why the choice of statistical method should follow the structure of the outcome. A binary disease-status endpoint is naturally handled as a categorical comparison, while a symptom score measured across baseline and follow-up has a longitudinal structure that can be handled with a mixed-effects model.
6. Statistical Methodology
Fisher exact test for the binary primary endpoint
The registry identifies Fisher exact testing as the method used for the Day 14 active-COVID-19 endpoint among participants who were asymptomatic at baseline. Fisher's exact test evaluates the association between treatment assignment and a binary outcome in a contingency table and is particularly useful when exact inference is desired rather than relying on a large-sample approximation.
The statistical comparison asks whether the observed allocation of participants with and without active COVID-19 disease differs between the hydroxychloroquine and placebo groups.
The posted effect measure is a mean difference (net), with the analysis notes specifying that the values represent percentages. Thus, the reported estimate of -2.4 can be read as a treatment-minus-placebo difference of 2.4 percentage points for the endpoint as reported by the registry.
Mixed-effects model for disease-severity change
The second primary endpoint uses a mixed models analysis. The outcome is a Visual Analog Scale from 0 to 10, where 0 represents no symptoms and 10 represents the most severe symptoms. Measurements are specified at baseline and 14 days, creating a repeated-measure structure within participants.
A mixed-effects framework allows observations from the same participant to be modeled as correlated rather than treating repeated measurements as independent. The registry does not provide the detailed covariance structure or fixed-effect specification in the summarized record.
For this endpoint, the posted effect measure is a mean difference. The estimate of -0.27 represents the reported net difference in the modeled change in the disease-severity score between treatment groups.
Confidence intervals
Both posted primary analyses include two-sided 95% confidence intervals. A confidence interval describes the statistical uncertainty surrounding the estimated treatment difference under the corresponding analysis framework. It is not a statement that there is a 95% probability that the true effect lies inside the particular interval after the data have been observed.
P-values
The registry reports a P-value for each primary analysis. A P-value quantifies how incompatible the observed result, or a result more extreme, would be with the null hypothesis under the specified statistical test. It does not measure the size or clinical importance of the treatment effect.
Superiority framework
The registry identifies both primary analyses as superiority hypotheses. The statistical question is therefore whether the treatment groups differ, rather than whether hydroxychloroquine satisfies a prespecified non-inferiority margin.
7. Results: Active COVID-19 Disease at Day 14
The first primary endpoint was the number of participants with active COVID-19 disease at 14 days post enrollment among those who were asymptomatic at baseline. The registry reports a Fisher exact analysis comparing hydroxychloroquine with placebo.
Mean difference in percentage points
95% CI: -7.0 to 2.2 · P = 0.35
Values represent percentages; experimental treatment was compared with placebo.
| Primary endpoint | Hydroxychloroquine vs placebo | 95% CI | P-value | Method |
|---|---|---|---|---|
| Active COVID-19 disease at Day 14 among participants asymptomatic at baseline | -2.4 percentage points | -7.0 to 2.2 | 0.35 | Fisher exact test |
The reported estimate of -2.4 means that the percentage of participants with active COVID-19 disease at Day 14 was estimated to be 2.4 percentage points lower in the hydroxychloroquine group than in the placebo group, using the registry's reported effect measure.
The estimate is surrounded by a two-sided 95% confidence interval from -7.0 to 2.2. Because the interval crosses 0, the data are compatible with both a lower percentage and a higher percentage in the hydroxychloroquine group under the statistical framework used.
The P = 0.35 value is evidence about the statistical compatibility of the observed comparison with the null hypothesis; it is not a measure of whether the 2.4-percentage-point estimate is clinically large or small. The confidence interval is more informative about the range of treatment differences supported by the observed data.
This endpoint is binary, so Fisher's exact test evaluates the treatment-group comparison of disease status. The interpretation should remain tied to the defined population—participants who were asymptomatic at baseline—and the specified Day 14 assessment rather than being generalized to the entire enrolled population.
8. Results: Change in Disease Severity Over 14 Days
The second primary endpoint evaluated change in disease severity among participants who were symptomatic at baseline. Severity was recorded using a Visual Analog Scale from 0 to 10, with 0 corresponding to no symptoms and 10 corresponding to the most severe symptoms.
Net mean difference in change
95% CI: -0.61 to 0.27 · P = 0.117
Mixed models analysis; baseline and 14-day assessments.
| Primary endpoint | Estimate | 95% CI | P-value | Method |
|---|---|---|---|---|
| Change in disease severity over 14 days among participants symptomatic at baseline | -0.27 | -0.61 to 0.27 | 0.117 | Mixed-effects model |
The estimated treatment difference in change in symptom severity was -0.27. On the registry's 0–10 severity scale, a negative value indicates a lower modeled change in the hydroxychloroquine group relative to placebo under the reported effect-measure convention.
The two-sided 95% confidence interval extends from -0.61 to 0.27. Because the interval crosses 0, the observed data are compatible with either direction of the treatment difference within the uncertainty represented by this interval.
The reported P = 0.117 does not describe the magnitude of the treatment effect. It addresses the statistical evidence against the null hypothesis in the specified mixed-model analysis. The confidence interval provides the more direct description of precision around the estimated mean difference.
The longitudinal nature of the endpoint is important. The mixed-effects framework is designed to account for repeated measurements within participants, but the registry summary does not provide the full model specification, such as the covariance structure. The result therefore should be interpreted as the registry's reported mixed-model comparison rather than as a reconstruction of a more detailed model.
9. Putting the Two Primary Results Together
The two primary endpoints address complementary aspects of the trial question. The first concerns whether participants who were asymptomatic at baseline had active COVID-19 disease at Day 14. The second concerns change in symptom severity among participants who were symptomatic at baseline.
| Primary endpoint | Estimate | 95% CI | P-value | Statistical structure |
|---|---|---|---|---|
| Active COVID-19 disease at Day 14 among those asymptomatic at baseline | -2.4 | -7.0 to 2.2 | 0.35 | Binary; Fisher exact test |
| Change in disease severity over 14 days among those symptomatic at baseline | -0.27 | -0.61 to 0.27 | 0.117 | Continuous longitudinal outcome; mixed-effects model |
Both confidence intervals include zero, and both reported P-values are above 0.05. Descriptively, the point estimates are negative, but the uncertainty intervals encompass zero and, for the first endpoint, also encompass positive values. These results should therefore be read as estimates with uncertainty rather than as proof of a particular direction or magnitude of treatment effect.
10. Serious Adverse Events
The registry reports serious adverse events by randomized treatment arm as affected participants divided by participants at risk.
| Safety measure | Hydroxychloroquine | Placebo |
|---|---|---|
| Serious adverse events | 6 / 658 | 11 / 654 |
The reported serious-adverse-event counts are descriptive safety information. They should not be interpreted as an efficacy endpoint or combined with the primary efficacy results into a single overall treatment effect.
The denominators also matter. The registry reports 6 affected participants among 658 at risk in the hydroxychloroquine group and 11 among 654 at risk in the placebo group. The denominators differ from the overall enrollment of 1312, illustrating why a safety analysis population or risk set should be identified separately from the headline randomized enrollment when interpreting adverse-event summaries.
11. Statistical Methods Explained
Why was Fisher's exact test used for the first primary endpoint?
The first endpoint is binary: participants either had active COVID-19 disease at Day 14 or they did not, within the baseline-asymptomatic population. Fisher's exact test provides an exact categorical-data comparison of treatment groups without relying on the same large-sample approximation used by some other contingency-table procedures.
What does a mean difference of -2.4 mean for the first endpoint?
The registry states that the values represent percentages. Therefore, the reported -2.4 can be read as a net treatment-minus-placebo difference of 2.4 percentage points in the reported percentage of participants with active COVID-19 disease at Day 14. It is not a hazard ratio, odds ratio, or relative risk.
Why was a mixed-effects model used for symptom severity?
The symptom-severity endpoint is measured at baseline and again at 14 days. Measurements from the same participant are related, so treating them as independent observations would not appropriately represent the repeated-measure structure. A mixed-effects model provides a framework for modeling the longitudinal observations while accounting for participant-level variation.
What does a mean difference of -0.27 mean?
The reported estimate is the net mean difference in change in the 0–10 disease-severity score between hydroxychloroquine and placebo. A negative value indicates a lower modeled change in the hydroxychloroquine group according to the reported treatment comparison. The confidence interval, from -0.61 to 0.27, shows that the estimate is not precise enough to exclude zero under the stated analysis.
Why does the P-value not measure treatment effect size?
A P-value is determined by both the observed effect and the amount of statistical information available. A small P-value does not necessarily imply a large or clinically important effect, and a larger P-value does not establish that the treatment groups are identical. Effect estimates and confidence intervals are needed to describe magnitude and precision.
Why are the baseline populations important?
The two primary endpoints are explicitly defined in different baseline populations. The Day 14 disease-status endpoint applies to participants who were asymptomatic at baseline, whereas the disease-severity endpoint applies to participants who were symptomatic at baseline. Mixing those populations would change the question being analyzed.
12. Confidence Intervals and the Meaning of Uncertainty
Both primary analyses report two-sided 95% confidence intervals. These intervals are central to interpretation because they show how much uncertainty surrounds the point estimates.
First endpoint
The estimate is -2.4, with a 95% CI from -7.0 to 2.2. The interval spans zero and includes both negative and positive treatment differences.
Second endpoint
The estimate is -0.27, with a 95% CI from -0.61 to 0.27. The interval also spans zero and includes both directions of the treatment difference.
The first confidence interval covers a wider range of percentage-point differences than the point estimate alone communicates. The second covers a range of changes in the symptom-severity score. Neither interval should be interpreted as a range containing individual patient responses; both describe uncertainty around a population-level treatment comparison.
The fact that both intervals include zero is consistent with the corresponding P-values of 0.35 and 0.117. The confidence intervals add information that the P-values alone do not provide: they show the direction and scale of effects compatible with the observed data under the specified analyses.
13. Why Randomization Matters
Randomization is a foundational feature of the trial design. Participants were assigned to hydroxychloroquine or placebo through a randomized allocation, creating the basis for a treatment-group comparison that is less dependent on systematic differences in measured and unmeasured baseline characteristics than a nonrandomized comparison would be.
Randomization does not guarantee identical groups in any particular sample. Its principal value is that, in expectation, treatment assignment is independent of prognostic factors before treatment is given.
The quadruple-masked design adds another layer of protection against bias arising from knowledge of treatment assignment. The registry identifies the masking level but does not specify the four masked roles in the summarized information.
14. Why the Trial Uses Two Different Statistical Methods
A useful statistical lesson from COVID-19 PEP is that a single trial can require fundamentally different analyses for different primary outcomes.
| Feature | Day 14 active disease | Change in disease severity |
|---|---|---|
| Baseline population | Asymptomatic | Symptomatic |
| Outcome structure | Binary | Continuous / longitudinal |
| Time structure | Day 14 assessment | Baseline and 14 days |
| Registry method | Fisher exact test | Mixed models analysis |
| Effect measure | Mean difference (net), values represent percentages | Mean difference (net) |
| Estimate | -2.4 | -0.27 |
| 95% CI | -7.0 to 2.2 | -0.61 to 0.27 |
| P-value | 0.35 | 0.117 |
The statistical method should be driven by the measurement scale and design of the endpoint. Applying the same test to both outcomes would discard important information about the way the data were collected.
15. Interpretation of the First Primary Endpoint
The reported mean difference of -2.4 indicates a 2.4-percentage-point lower estimated percentage in the hydroxychloroquine group relative to placebo for active COVID-19 disease at Day 14 among participants who were asymptomatic at baseline.
The 95% confidence interval of -7.0 to 2.2 is substantially wider than the point estimate and crosses zero. Thus, the observed estimate does not by itself establish a treatment-group difference in either direction.
The reported P = 0.35 indicates that the Fisher exact analysis did not provide strong statistical evidence against the null hypothesis at conventional significance thresholds. It should not be interpreted as the probability that the treatment has no effect.
The result applies to the registry-defined endpoint population—participants who were asymptomatic at baseline. It should not be relabeled as an analysis of all 1312 enrolled participants.
16. Interpretation of the Second Primary Endpoint
The reported mean difference in change was -0.27 on the disease-severity scale. The negative direction represents a lower modeled change in the hydroxychloroquine group under the registry's reported comparison.
The 95% confidence interval extends from -0.61 to 0.27. The interval crosses zero, so the observed estimate is compatible with a range of treatment differences in either direction under the model.
The reported P = 0.117 is a measure of statistical evidence under the mixed-model analysis, not a measure of effect magnitude. The estimated difference should be considered together with its confidence interval and the 0–10 scale used for the endpoint.
Because the endpoint is measured at baseline and 14 days, the model accounts for the repeated nature of the observations. The registry summary does not provide enough detail to independently reconstruct every component of the mixed-effects model.
17. Design Features Relevant to Statistical Interpretation
The registry information summarized here does not identify a non-inferiority margin, crossover procedure, factorial structure, Bayesian analysis, interim-analysis plan, or missing-data/imputation strategy. Those topics therefore are not part of the statistical interpretation of the posted primary analyses on this page.
18. Results Posted on ClinicalTrials.gov
The registry reports that results were posted for the study. It lists nine posted outcome measures and two posted statistical analyses, corresponding to the two primary endpoints described above.
| Registry result feature | Reported value |
|---|---|
| Results posted | Yes |
| Outcome measures posted | 9 |
| Statistical analyses posted | 2 |
| Primary-endpoint analyses | 2 |
| Primary analyses with estimate + CI | 2 |
| Normalized methods | Fisher exact test; mixed-effects model |
| Effect measures | Mean difference |
| Hypothesis type | Superiority |
The presence of a formal analysis for each primary endpoint is important because the page can distinguish between the registry's descriptive outcome reporting and the posted inferential comparisons. For both primary endpoints, the registry provides an estimate, a two-sided 95% confidence interval, and a P-value.
19. Limitations
- Different endpoint populations: the two primary endpoints apply to different baseline clinical groups. Results should not be generalized across those populations without qualification.
- Limited model detail: the registry identifies a mixed models analysis for symptom severity but does not provide the complete model specification in the summarized record.
- Effect-measure terminology: the registry labels both analyses with a mean-difference effect measure, while the first endpoint is binary and its analysis notes clarify that the values represent percentages. Interpretation should therefore retain the registry's percentage-point context.
- Confidence-interval width: both primary confidence intervals cross zero, indicating uncertainty around the direction of the estimated treatment difference.
- Safety interpretation: the serious-adverse-event counts are descriptive and should not be interpreted as a formal comparative safety hypothesis test without additional statistical information.
- Population restriction: the first endpoint is restricted to participants asymptomatic at baseline, while the second is restricted to those symptomatic at baseline.
- Outcome measurement: the second primary endpoint uses a 0–10 Visual Analog Scale, so the numerical interpretation of a mean difference depends on the scale and its clinical meaning.
- Registry-level reporting: the ClinicalTrials.gov record provides the posted estimates and methods, but the summarized information does not contain every detail of the underlying statistical analysis plan.
20. Why This Trial Matters Statistically
COVID-19 PEP is a useful teaching case because its two primary endpoints require two different statistical frameworks within the same randomized trial. The first endpoint is categorical and assessed at a defined time point, while the second is a repeated continuous measure evaluated over time.
| Concept | How it appears in COVID-19 PEP |
|---|---|
| Randomization | Participants were randomly allocated to hydroxychloroquine or placebo. |
| Blinding | The registry identifies the trial as quadruple-masked. |
| Binary outcome analysis | Day 14 active COVID-19 disease among participants asymptomatic at baseline was analyzed with Fisher exact testing. |
| Continuous outcome | Disease severity was recorded on a 0–10 Visual Analog Scale. |
| Repeated measurements | Severity was evaluated at baseline and 14 days. |
| Mixed-effects modeling | The longitudinal disease-severity endpoint was analyzed with a mixed models analysis. |
| Confidence intervals | Both primary estimates include two-sided 95% confidence intervals. |
| P-values | Both primary analyses report P-values for the corresponding statistical tests. |
| Superiority testing | The registry classifies both primary hypotheses as superiority. |
| Endpoint-specific populations | The two primary endpoints apply to different baseline symptom-status groups. |
The broader lesson is that statistical analysis begins with the clinical question and measurement structure. A binary outcome, a repeated symptom score, and a safety event count are different statistical objects even when they arise in the same randomized trial.
21. A Closer Look at Fisher's Exact Test
For the first primary endpoint, the relevant information can conceptually be arranged in a 2 × 2 table: treatment group by active COVID-19 disease status at Day 14.
| Active COVID-19 disease | No active COVID-19 disease | |
|---|---|---|
| Hydroxychloroquine | Endpoint event | No endpoint event |
| Placebo | Endpoint event | No endpoint event |
Fisher's exact test evaluates the probability of the observed allocation, and tables at least as extreme according to the specified test, under the null hypothesis of no association between treatment group and outcome. Its exact nature makes it useful for categorical comparisons where the distribution of counts may not be well represented by an asymptotic approximation.
The posted estimate is more clinically interpretable than the P-value alone because it gives a direction and magnitude: -2.4 percentage points. The corresponding 95% confidence interval, -7.0 to 2.2, communicates the uncertainty around that estimate.
22. A Closer Look at the Mixed-Effects Model
The second primary endpoint differs fundamentally because the same participant contributes a severity measurement at baseline and another at 14 days. Those observations are expected to be correlated because they come from the same individual.
A longitudinal model recognizes that repeated observations from one participant are not equivalent to observations from two unrelated participants.
A mixed-effects model can represent both population-level treatment effects and participant-level variation. This allows the analysis to use the longitudinal structure rather than discarding the relationship between baseline and follow-up measurements.
The registry's reported result is a mean difference of -0.27, with a two-sided 95% confidence interval of -0.61 to 0.27 and P = 0.117. These quantities should be interpreted together rather than using the P-value as a standalone summary.
23. Mean Difference Versus Relative Effect Measures
The trial's posted statistical analyses use mean differences rather than hazard ratios or odds ratios. That distinction matters because the interpretation is tied to the scale of the endpoint.
Percentage-point difference
For the binary endpoint, the registry specifies that the values represent percentages. The -2.4 estimate therefore describes a difference in percentage points.
Score difference
For the disease-severity endpoint, -0.27 represents a mean difference in change on the 0–10 symptom-severity scale.
Neither result is a relative risk, odds ratio, hazard ratio, or percentage reduction. Keeping the effect measure tied to the outcome scale prevents a common statistical error: translating an absolute difference into a relative effect without the information required to justify that conversion.
24. What the Confidence Intervals Do Not Say
The first interval, -7.0 to 2.2, is not the range of individual participants' disease outcomes. The second interval, -0.61 to 0.27, is not the range of individual changes in symptom severity.
Instead, the intervals quantify uncertainty around the respective population-level treatment estimates. Individual patients can experience outcomes far outside the interval around a group-level mean difference.
25. Statistical Interpretation Versus Clinical Interpretation
Statistical interpretation
The two posted primary analyses produced estimates of -2.4 and -0.27, respectively. Both two-sided 95% confidence intervals cross zero, with P-values of 0.35 and 0.117.
Clinical interpretation
The clinical meaning of a percentage-point difference in Day 14 disease status and a 0–10 symptom-severity difference requires consideration of the endpoint definitions, populations, uncertainty, and the clinical context of the scale.
A statistical result does not automatically determine clinical importance. Conversely, a clinically meaningful difference should be evaluated with its statistical uncertainty. The most informative reading of the trial therefore keeps the effect estimate, confidence interval, P-value, endpoint definition, and analysis population together.
26. Primary Results Summary
Primary endpoint 1
Active COVID-19 disease at Day 14 among participants asymptomatic at baseline
95% CI: -7.0 to 2.2 · P = 0.35 · Fisher exact test
Primary endpoint 2
Change in disease severity over 14 days among participants symptomatic at baseline
95% CI: -0.61 to 0.27 · P = 0.117 · Mixed-effects model
The two primary estimates point in a negative direction, but both confidence intervals include zero. The first analysis concerns a percentage-based binary endpoint, while the second concerns change in a 0–10 symptom-severity score. The appropriate interpretation is therefore endpoint-specific rather than a single combined treatment conclusion.
27. Related Tutorials
Learn more about the methods used in this trial:
28. Related Calculators
29. Sources
- ClinicalTrials.gov: NCT04308668.
- PubMed: PMID 32492293.
- PubMed: PMID 32673060.
- PubMed: PMID 33204764.
- PubMed: PMID 33721539.
- PubMed: PMID 33274360.
Continue through the Clinical Biostats knowledge graph
Connect the endpoints and statistical methods in this trial to deeper biostatistics tutorials and practical statistical tools.
30. Record Summary
COVID-19 PEP provides a compact example of how a randomized clinical trial can use different statistical methods for different primary endpoints. The first primary endpoint is a binary Day 14 disease-status outcome among participants who were asymptomatic at baseline and was analyzed with Fisher's exact test. The second is a longitudinal disease-severity outcome among participants symptomatic at baseline and was analyzed using a mixed models approach.
The reported treatment differences were -2.4 percentage points for active COVID-19 disease at Day 14 and -0.27 for change in disease severity. Their two-sided 95% confidence intervals were -7.0 to 2.2 and -0.61 to 0.27, respectively, with P-values of 0.35 and 0.117. Both intervals include zero, emphasizing the importance of reporting uncertainty alongside point estimates.
The trial is also a useful illustration of why endpoint definitions, baseline populations, measurement scales, and statistical models must be interpreted together. A Fisher exact test and a mixed-effects model answer different statistical questions, and the effect measures should remain tied to the scales on which the outcomes were defined.