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
The COVID-19 trial registered as NCT04322682 was a randomized, parallel-group, triple-masked phase 3 treatment trial comparing colchicine with placebo in participants with Corona Virus Infection. The registry reports 4506 enrolled participants, one binary primary endpoint, and five posted statistical analyses.
| Feature | COVID-19 / COLCORONA |
|---|---|
| ClinicalTrials.gov identifier | NCT04322682 |
| Brief title | Colchicine Coronavirus SARS-CoV2 Trial (COLCORONA) |
| Phase | Phase 3 |
| Status | Terminated |
| Condition | Corona Virus Infection |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Treatment |
| Lead sponsor | Montreal Heart Institute |
| Enrollment | 4506 |
| Primary endpoint type | Binary |
| Results posted | Yes |
| Statistical analyses posted | 5 |
2. Clinical Question
The central question was whether colchicine changed the probability of the prespecified 30-day composite of death or hospitalization due to COVID-19 infection compared with placebo following randomization.
Population
Participants enrolled in a phase 3 treatment trial for Corona Virus Infection.
Intervention
Colchicine, administered as the study drug.
Comparator
Placebo oral tablet.
Primary question
Does colchicine reduce the odds of death or hospitalization due to COVID-19 infection within 30 days following randomization?
3. Trial Design
Colchicine
- Study intervention: colchicine
- Drug intervention
- Compared with placebo oral tablet
Placebo
- Placebo oral tablet
- Drug intervention category in the registry
- Reference group for the reported odds ratios
4. Endpoints
The registry identifies one primary endpoint and reports additional secondary and prespecified analyses. All of the reported outcomes below are binary participant-level outcomes assessed over the 30 days following randomization.
| Endpoint | Registry definition | Time frame | Type |
|---|---|---|---|
| Primary | Number of Participants Who Died or Were Hospitalized Due to COVID-19 Infection in the 30 Days Following Randomization. The primary endpoint will be the composite of death or hospitalization due to COVID-19 infection in the 30 days following randomization. | 30 days post randomization | Binary |
| Secondary | Number of Deaths in the 30 Days Following Randomization. | 30 days post randomization | Binary |
| Secondary | Number of Participants Who Were Hospitalized Due to COVID-19 Infection in the 30 Days Following Randomization. | 30 days post randomization | Binary |
| Secondary | Number of Participants Who Required Mechanical Ventilation in the 30 Days Following Randomization. | 30 days post randomization | Binary |
| Other pre-specified | Number of Participants Who Died or Were Hospitalized Due to COVID-19 Infection in the 30 Days Following Randomization in the Subgroup of Patients With PCR-confirmed COVID-19. | 30 Days post randomization | Binary |
5. Statistical Methodology
Chi-squared testing for binary outcomes
The primary and three secondary analyses reported on the registry used the chi-squared test. This is a natural approach when the outcome is binary and the analysis compares the frequency of an event between two randomized treatment groups.
The chi-squared framework tests whether the observed distribution of the binary outcome differs between the treatment groups under the null hypothesis of no association between treatment assignment and the outcome.
Odds ratio as the effect measure
The registry reports odds ratios for the primary endpoint and each of the three secondary binary outcomes analyzed with the chi-squared method. An odds ratio below 1 indicates lower estimated odds of the event in the colchicine group relative to placebo.
An OR of 1 corresponds to equal odds. An OR below 1 indicates lower estimated odds with colchicine, whereas an OR above 1 indicates higher estimated odds.
Logistic regression for the PCR-confirmed subgroup
The prespecified PCR-confirmed COVID-19 subgroup analysis used logistic regression rather than the chi-squared method. The registry states that the model included the treatment group, the PCR-confirmed COVID-19 subgroup factor (yes/no), and their interaction.
This is statistically important because the interaction term allows the analysis to address whether the treatment effect differs according to PCR-confirmed subgroup status. The reported odds ratio of 0.75 is therefore associated with a model that explicitly included treatment, subgroup status, and treatment-by-subgroup interaction.
Confidence intervals
The primary analysis reports a 95.1% two-sided confidence interval, while the other posted analyses report 95% two-sided confidence intervals. A confidence interval quantifies statistical uncertainty around an estimated odds ratio; it is not a range containing the individual treatment effects experienced by patients.
Superiority hypothesis
The registry classifies the reported hypotheses as superiority. This differs from a non-inferiority framework: the goal is to test whether the treatment groups differ in the specified direction rather than to establish that one treatment is no worse than another within a predefined margin.
6. Primary Endpoint Result
The primary endpoint was the composite of death or hospitalization due to COVID-19 infection in the 30 days following randomization. The registry reports a chi-squared analysis comparing colchicine with placebo.
Death or hospitalization due to COVID-19
95.1% CI: 0.61–1.03 · P = 0.081
Two-sided chi-squared test; superiority hypothesis.
| Primary endpoint | Colchicine vs placebo |
|---|---|
| Outcome | Death or hospitalization due to COVID-19 infection |
| Time frame | 30 days post randomization |
| Method | Chi-squared test |
| Effect measure | Odds ratio |
| Estimate | 0.79 |
| Confidence interval | 95.1% two-sided CI: 0.61–1.03 |
| P-value | 0.081 |
| Hypothesis | Superiority |
The estimated odds ratio of 0.79 means that the estimated odds of the composite endpoint were about 21% lower with colchicine than with placebo in this randomized comparison. This is an odds comparison, not a 21% reduction in absolute probability and not a statement that 21% of patients benefited.
The 95.1% confidence interval of 0.61–1.03 indicates uncertainty around the estimated odds ratio. Because the interval includes 1, the registry result does not exclude equal odds between the groups at that confidence level.
The P-value of 0.081 describes the statistical evidence against the null hypothesis under the specified testing framework; it does not measure the size or clinical importance of the estimated effect. A P-value should therefore be interpreted together with the odds ratio and confidence interval rather than as a standalone measure of treatment effect.
The analysis is a binary endpoint comparison, so concerns specific to proportional hazards do not apply to this primary odds-ratio analysis. The composite endpoint also combines two clinically distinct outcomes—death and hospitalization—so its interpretation is necessarily an interpretation of the combined event definition rather than either component alone.
7. Secondary Endpoint Results
Death within 30 days
Number of deaths
95% CI: 0.19–1.67 · P = 0.291
Two-sided chi-squared test; superiority hypothesis.
The estimated odds ratio of 0.56 corresponds to lower estimated odds of death in the colchicine group relative to placebo. However, the confidence interval is wide, extending from 0.19 to 1.67, which reflects substantial statistical uncertainty around this estimate.
The P-value of 0.291 does not provide evidence against the null hypothesis at conventional significance thresholds. Importantly, the P-value does not establish that the treatment groups are equivalent; the wide confidence interval shows that materially different odds ratios remain compatible with the estimate.
Hospitalization due to COVID-19
Number hospitalized
95% CI: 0.60–1.03 · P = 0.077
Two-sided chi-squared test; superiority hypothesis.
The estimated odds ratio of 0.79 indicates lower estimated odds of hospitalization in the colchicine group relative to placebo. The 95% confidence interval of 0.60–1.03 includes 1, so the reported result is compatible with both a reduction and little or no difference in odds.
The P-value of 0.077 summarizes the evidence against the null hypothesis under the stated chi-squared framework. It should not be interpreted as a measure of the magnitude of the treatment effect.
Mechanical ventilation within 30 days
Number requiring mechanical ventilation
95% CI: 0.25–1.09 · P = 0.080
Two-sided chi-squared test; superiority hypothesis.
The estimated odds ratio of 0.53 corresponds to approximately 47% lower estimated odds of mechanical ventilation with colchicine relative to placebo. The confidence interval, however, ranges from 0.25 to 1.09 and therefore includes 1.
The P-value of 0.080 indicates that the reported result did not meet a conventional 0.05 threshold. More importantly for interpretation, the confidence interval demonstrates that the estimate remains imprecise enough that the true treatment effect could differ materially from the point estimate.
8. Summary of Posted Statistical Analyses
| Endpoint | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Death or hospitalization due to COVID-19 | Chi-squared | OR 0.79 | 0.61–1.03 (95.1%) | 0.081 |
| Death | Chi-squared | OR 0.56 | 0.19–1.67 | 0.291 |
| Hospitalization due to COVID-19 | Chi-squared | OR 0.79 | 0.60–1.03 | 0.077 |
| Mechanical ventilation | Chi-squared | OR 0.53 | 0.25–1.09 | 0.080 |
| PCR-confirmed COVID-19 composite | Logistic regression | OR 0.75 | 0.57–0.99 | 0.042 |
The five posted analyses therefore use two related approaches for binary outcomes: chi-squared tests for the primary and three secondary analyses, and logistic regression for the prespecified PCR-confirmed subgroup analysis.
9. PCR-Confirmed COVID-19 Subgroup Analysis
The registry reports a separate prespecified analysis of the primary composite endpoint among patients with PCR-confirmed COVID-19. Unlike the other posted analyses, this analysis used logistic regression.
PCR-confirmed COVID-19 subgroup
95% CI: 0.57–0.99 · P = 0.042
Two-sided logistic regression; superiority hypothesis.
| Model component | Registry description |
|---|---|
| Outcome | Death or hospitalization due to COVID-19 infection in the 30 days following randomization |
| Population | Subgroup of patients with PCR-confirmed COVID-19 |
| Method | Logistic regression |
| Treatment factor | Treatment group |
| Subgroup factor | PCR-confirmed COVID-19 subgroup factor, yes/no |
| Interaction | Treatment group × PCR-confirmed COVID-19 subgroup factor |
| Effect measure | Odds ratio |
| Estimate | 0.75 |
| 95% CI | 0.57–0.99 |
| P-value | 0.042 |
An odds ratio of 0.75 corresponds to approximately 25% lower estimated odds of the composite endpoint with colchicine relative to placebo within the PCR-confirmed subgroup. The 95% confidence interval of 0.57–0.99 lies below 1, although its upper bound is close to 1.
The P-value of 0.042 is a test result from the specified logistic-regression analysis. It does not quantify the magnitude of benefit, and it should not be interpreted as the probability that colchicine works or the probability that the observed result occurred by chance.
The registry explicitly includes an interaction between treatment and PCR-confirmed subgroup status. That feature matters because an interaction is the formal modeling mechanism for assessing whether treatment effects differ according to subgroup status. The subgroup odds ratio itself should not be treated as proof of effect modification without considering the interaction result and the full model.
10. Statistical Methods Explained
Why was a chi-squared test used?
The primary endpoint and three secondary endpoints are binary outcomes: each participant either experienced the specified event during the 30-day period or did not. A chi-squared test provides a direct comparison of the distribution of events between the two randomized groups.
What does an odds ratio of 0.79 mean?
An odds ratio of 0.79 means that the estimated odds of the event in the colchicine group were 0.79 times the estimated odds in the placebo group. Equivalently, the estimated odds were approximately 21% lower. This does not mean that the event probability was 21 percentage points lower, nor does it necessarily correspond to a 21% reduction in probability.
Why is an odds ratio different from a risk ratio?
A risk ratio compares probabilities directly, whereas an odds ratio compares odds. When events are uncommon, the two measures can be numerically similar, but they are not interchangeable. The registry reports odds ratios for the posted analyses, so the results should be interpreted on the odds scale.
Why use logistic regression for the PCR-confirmed analysis?
Logistic regression models a binary outcome while allowing multiple explanatory variables to be included simultaneously. Here, the registry specifies treatment group, PCR-confirmed subgroup status, and their interaction. That structure allows the model to represent both the treatment comparison and whether the treatment effect varies according to PCR-confirmed status.
What does the confidence interval tell us?
The confidence interval describes the uncertainty around the estimated odds ratio under the statistical model and sampling framework. For the primary endpoint, the 95.1% interval is 0.61–1.03. For the PCR-confirmed subgroup analysis, the 95% interval is 0.57–0.99. A confidence interval crossing 1 means that equal odds remain compatible with the interval; it does not prove that the treatment groups are equivalent.
Why should the P-value not be treated as an effect size?
A P-value depends on the estimated effect, its uncertainty, and the amount of information contributing to the analysis. It does not directly describe the magnitude of the treatment effect. The odds ratio and confidence interval are therefore essential for understanding both direction and precision.
11. Safety Results
The registry reports serious adverse events by randomized treatment arm. These figures provide the number affected and the number at risk for each group.
| Treatment arm | Serious adverse events | Affected / at risk |
|---|---|---|
| Colchicine | Serious adverse events | 108 / 2195 |
| Placebo | Serious adverse events | 139 / 2217 |
The affected and at-risk counts are presented as reported by the registry. The registry does not post a formal comparative statistical analysis for serious adverse events in the information summarized here, so these figures should be read descriptively rather than as a separate hypothesis test.
12. Understanding the Primary Composite Endpoint
The primary endpoint combines death and hospitalization due to COVID-19 infection into one binary outcome. A participant meeting either component contributes an event to the composite.
Why composites are used
A composite endpoint can capture more events than a rare individual component, increasing the amount of statistical information available for a comparison.
What the composite does not show
The composite odds ratio does not tell us whether any observed difference is driven primarily by death, hospitalization, or both components.
Component analyses
The registry separately reports analyses for death and hospitalization, allowing the components to be examined individually.
Interpretation
The primary endpoint should therefore be interpreted as the combined probability of meeting either prespecified event during the 30-day period.
13. Odds Ratios and Absolute Risk
The posted results are expressed as odds ratios rather than absolute risk differences. This distinction is important when communicating clinical-trial results.
An odds ratio describes a relative comparison on the odds scale. It cannot, by itself, be translated into an absolute percentage-point difference without the underlying event probabilities.
For example, the primary OR of 0.79 indicates a relative comparison of odds, but the clinical meaning of that ratio depends on the baseline event probability in the placebo group. Two studies can have the same odds ratio while producing different absolute risk differences if their baseline risks differ.
14. Confidence Intervals and Statistical Precision
| Analysis | Estimate | Confidence interval | What the interval communicates |
|---|---|---|---|
| Primary composite | OR 0.79 | 0.61–1.03 | Uncertainty includes equal odds at the 1.0 reference value. |
| Death | OR 0.56 | 0.19–1.67 | Wide interval; substantial uncertainty around the point estimate. |
| Hospitalization | OR 0.79 | 0.60–1.03 | Uncertainty includes equal odds. |
| Mechanical ventilation | OR 0.53 | 0.25–1.09 | Wide enough to include equal odds. |
| PCR-confirmed composite | OR 0.75 | 0.57–0.99 | Interval lies below 1, with the upper bound close to the null value. |
The point estimate is only one part of an analysis. The width of the confidence interval provides information about precision. The death analysis, for example, has an interval extending from 0.19 to 1.67, illustrating considerably more uncertainty around the point estimate than would be conveyed by the OR of 0.56 alone.
Confidence intervals should not be interpreted as the range of plausible outcomes for an individual participant. They describe uncertainty in the estimated population-level treatment effect under the statistical framework used for the analysis.
15. P-values and Hypothesis Testing
All five posted analyses are classified in the registry as superiority hypotheses. The primary endpoint has a two-sided P-value of 0.081. The secondary P-values are 0.291 for death, 0.077 for hospitalization, and 0.080 for mechanical ventilation. The PCR-confirmed subgroup logistic-regression analysis reports 0.042.
| Endpoint | P-value | Interpretive point |
|---|---|---|
| Death or hospitalization | 0.081 | Primary superiority test; P-value above 0.05. |
| Death | 0.291 | Secondary binary endpoint; substantial uncertainty in the reported OR. |
| Hospitalization | 0.077 | Secondary binary endpoint; confidence interval includes 1. |
| Mechanical ventilation | 0.080 | Secondary binary endpoint; confidence interval includes 1. |
| PCR-confirmed composite | 0.042 | Prespecified subgroup logistic-regression analysis. |
16. Logistic Regression and the Interaction Term
The PCR-confirmed analysis is the only posted analysis that explicitly uses logistic regression. The model included treatment group, PCR-confirmed COVID-19 subgroup status, and their interaction.
The interaction term represents whether the treatment association with the binary outcome changes according to PCR-confirmed subgroup status.
This is different from simply performing two unrelated analyses in two subgroups. A formal interaction term provides a direct statistical framework for evaluating effect modification. The presence of an interaction term in the model therefore makes the analysis structurally suited to asking whether treatment effects vary by PCR-confirmed status.
17. Trial Timeline
Trial start
The registry lists March 23, 2020 as the study start date.
Randomized phase 3 study
The trial evaluated colchicine versus placebo in a randomized, parallel-group, triple-masked treatment design.
Primary completion
The registry lists January 21, 2021 as the primary completion date.
Early stopping
The registry states that the study was stopped early for logistical, human, and budgetary considerations and to allow communication of results after approximately 4500 participants had completed 30-day follow-up.
18. Important Limitations and Interpretation Issues
- Early stopping: the registry states that the study was stopped early for logistical, human, and budgetary considerations. Early stopping can affect the amount of information accumulated and therefore the precision of estimated treatment effects.
- Composite endpoint: the primary endpoint combines death and hospitalization, which are clinically distinct outcomes.
- Odds-ratio interpretation: odds ratios are not risk ratios and should not be described as absolute percentage changes in event probability.
- Confidence-interval width: several secondary analyses have relatively wide confidence intervals, particularly the death endpoint.
- Multiple analyses: the registry reports one primary endpoint and several additional analyses. Their statistical roles differ and should not be treated as interchangeable.
- Subgroup analysis: the PCR-confirmed analysis is a subgroup analysis and includes an interaction term. Subgroup estimates generally require more cautious interpretation than the principal randomized comparison.
- Safety interpretation: serious adverse-event counts are reported by arm, but the registry information summarized here does not provide a corresponding formal comparative test.
- Time frame: the reported efficacy endpoints concern the 30 days following randomization. They should not be interpreted as measures of longer-term outcomes beyond that registered time frame.
19. Why This Trial Matters Statistically
This trial is a useful teaching example because it demonstrates how a relatively simple binary clinical endpoint can require several layers of statistical interpretation.
| Concept | How it appears in this trial |
|---|---|
| Randomization | Participants were assigned in a randomized parallel-group design. |
| Blinding | The registry specifies triple masking. |
| Binary endpoint | The primary outcome records whether death or hospitalization occurred within 30 days. |
| Chi-squared test | Used for the primary and three secondary binary endpoint analyses. |
| Odds ratio | Used as the reported effect measure for all five posted analyses. |
| Confidence interval | Provides uncertainty around each estimated odds ratio. |
| P-value | Quantifies evidence against the null hypothesis under each specified testing framework. |
| Logistic regression | Used for the PCR-confirmed COVID-19 subgroup analysis. |
| Interaction | The PCR-confirmed analysis included a treatment-by-subgroup interaction. |
| Composite endpoint | Death and hospitalization were combined into the primary endpoint. |
| Early stopping | The registry describes early study termination for logistical, human, and budgetary considerations. |
20. Overall Statistical Interpretation
The primary analysis estimated an odds ratio of 0.79 for death or hospitalization due to COVID-19 within 30 days, with a 95.1% confidence interval of 0.61–1.03 and a P-value of 0.081. The point estimate is below 1, while the confidence interval includes 1.
The reported odds ratios were 0.56 for death, 0.79 for hospitalization, and 0.53 for mechanical ventilation. Their confidence intervals were 0.19–1.67, 0.60–1.03, and 0.25–1.09, respectively. Each interval includes 1, and the corresponding P-values were 0.291, 0.077, and 0.080.
The prespecified PCR-confirmed COVID-19 subgroup analysis used logistic regression with treatment, subgroup status, and their interaction. It reported an odds ratio of 0.75, with a 95% confidence interval of 0.57–0.99 and a P-value of 0.042.
These estimates do not by themselves describe absolute risk reduction, the number needed to treat, individual patient benefit, or outcomes beyond the registered 30-day assessment period. Those quantities require information beyond the reported odds ratios and P-values.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: NCT04322682.
- Publication: PubMed record 39240567.
- Publication: PubMed record 34658014.
- Publication: PubMed record 34051877.
Continue through the Clinical Biostats statistical library
Connect the endpoints and methods in this trial to deeper statistical tutorials and analysis tools.
24. Record Summary
The NCT04322682 trial provides a focused example of statistical analysis for randomized binary clinical outcomes. Its primary endpoint was a 30-day composite of death or hospitalization due to COVID-19 infection, analyzed with a chi-squared test and reported as an odds ratio of 0.79 with a 95.1% confidence interval of 0.61–1.03 and P = 0.081. Secondary analyses addressed death, hospitalization, and mechanical ventilation, while a prespecified PCR-confirmed COVID-19 subgroup was analyzed using logistic regression with treatment, subgroup status, and their interaction.
The most useful statistical reading combines the direction and magnitude of the odds ratios, the precision of their confidence intervals, the role of each P-value, and the design features that determine how each analysis should be interpreted. In particular, the composite endpoint, multiple analyses, subgroup interaction, and early study termination are important parts of the statistical context.