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
STOIC was a randomized, open-label phase 2 trial evaluating budesonide dry powder inhaler in coronavirus infection. The registry describes a parallel two-arm treatment study with 146 participants enrolled and a primary endpoint focused on emergency department attendance or hospitalisation related to COVID-19 during the first 28 days.
| Feature | STOIC |
|---|---|
| Trial acronym | STOIC |
| Full title | STerOids in COVID-19 Study |
| Phase | Phase 2 |
| Status | Terminated |
| Condition | Coronavirus Infection |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 146 |
| Number of arms | 2 |
| Intervention | Budesonide dry powder inhaler |
| Lead sponsor | University of Oxford |
| Start date | 2020-07-16 |
| Primary completion date | 2021-01-12 |
| ClinicalTrials.gov | NCT04416399 |
2. Clinical Question
The central question is whether treatment with a budesonide dry powder inhaler affects the occurrence of emergency department attendance or hospitalisation related to COVID-19 during Day 1 to day 28 in people with coronavirus infection.
Population
Participants enrolled in a phase 2 clinical trial for coronavirus infection.
Intervention
Budesonide dry powder inhaler.
Comparator
The study used two randomized arms, but the registry fields summarized here identify only the budesonide dry powder inhaler intervention and do not specify the other arm.
Primary question
Does the randomized intervention affect emergency department attendance or hospitalisation related to COVID-19 from Day 1 to day 28?
3. Trial Design
4. Trial Timeline
Study start
The registered study start date was July 16, 2020.
Primary completion
The registered primary completion date was January 12, 2021.
Terminated
The ClinicalTrials.gov record identifies the study status as terminated.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Emergency department attendance of hospitalisation related to COVID-19 | Evaluate the effect of intervention on emergency department attendance or hospitalisation related to COVID-19 | Day 1 to day 28 |
The primary endpoint is a clinically consequential event-based outcome rather than a continuous measurement. Its statistical interpretation therefore depends on whether each participant experiences the specified emergency department attendance or hospitalisation outcome during the defined Day 1 to day 28 observation window.
This binary representation is useful for understanding the basic statistical structure of the endpoint. The registry wording itself defines the clinical event and its observation period.
6. Statistical Methodology
No formal statistical analyses were posted to ClinicalTrials.gov for this record. The registry therefore establishes the primary endpoint and its time frame, but it does not provide numerical treatment-effect estimates, confidence intervals, p-values, or a posted statistical analysis of that endpoint.
Primary analysis of a binary event endpoint
For an endpoint defined by whether an event occurs during a fixed follow-up window, a natural primary comparison is the proportion of participants experiencing the event in each randomized arm. The fundamental quantities are the event risk in each arm and a between-arm measure such as a risk difference or risk ratio.
The risk difference compares the two proportions directly, while the risk ratio expresses the intervention-group risk relative to the comparator-group risk.
Risk difference
The risk difference is calculated by subtracting the comparator event probability from the intervention event probability. It is particularly useful when the clinical question concerns the absolute change in the probability of an emergency department attendance or hospitalisation during the specified period.
A negative value would correspond to fewer observed events in the intervention arm; a positive value would correspond to more. The magnitude is expressed in absolute probability units.
Risk ratio
The risk ratio provides a relative comparison between the two randomized groups. A value of 1 corresponds to equal event probabilities. Values below 1 indicate a lower observed event probability in the intervention group relative to the comparator, while values above 1 indicate a higher probability.
Confidence intervals
A confidence interval around a risk difference or risk ratio quantifies sampling uncertainty around the estimated treatment effect. It should be presented alongside the point estimate because the same estimated effect can have very different levels of precision depending on the number of observed events and participants.
Hypothesis testing
A formal comparison could use a two-sample test for proportions, with an exact method considered when event counts are sparse. A corresponding confidence interval for the chosen effect measure provides information about precision that a p-value alone does not provide.
7. Statistical Methods Explained
Why is the primary endpoint naturally analysed as a binary outcome?
The endpoint asks whether emergency department attendance or hospitalisation related to COVID-19 occurs during a specified period. Each participant can therefore be classified according to whether the endpoint occurred during Day 1 to day 28. This creates a binary outcome suitable for comparing event proportions between randomized groups.
Why report a risk difference?
A risk difference gives an absolute interpretation. For example, if two groups had event probabilities that differed by a particular number of percentage points, that difference would describe the change in the probability of experiencing the endpoint during the defined observation period. It is therefore directly connected to the clinical event being measured.
Why also consider a risk ratio?
Relative effects can answer a different question from absolute effects. A risk ratio describes how the probability of the endpoint in one group compares proportionally with the probability in the other group. Reporting both absolute and relative measures can prevent a relative effect from being interpreted without its underlying event probabilities.
Why does the Day 1 to day 28 window matter?
The time window defines the period over which the primary outcome is evaluated. An event occurring outside that specified window is not equivalent to an event occurring within it for purposes of the registered primary endpoint. Fixing the observation period also makes the treatment comparison more precisely defined.
Why is randomization important for the statistical analysis?
Randomization determines treatment assignment independently of the participant's underlying prognosis, subject to chance imbalances. This allows differences in outcomes between the randomized groups to be interpreted as treatment-group differences rather than simply differences between groups selected by clinical characteristics or treatment preference.
Why does the absence of masking matter?
The registry identifies STOIC as having no masking. For an endpoint involving emergency department attendance or hospitalisation, lack of masking can matter because knowledge of treatment assignment can potentially influence aspects of participant behaviour, clinical management, or outcome assessment. The direction and magnitude of any such influence cannot be established from the design fields alone.
8. Interpreting the Primary Endpoint
The primary endpoint is a composite event definition: emergency department attendance or hospitalisation related to COVID-19 during Day 1 to day 28. A participant meeting either component would satisfy the overall endpoint as written.
The key comparison is therefore the frequency of participants experiencing the registered event in each randomized treatment group. The statistical effect should be described using an explicitly defined effect measure rather than relying on an isolated p-value.
A confidence interval would describe uncertainty around the estimated difference or ratio between treatment groups. A narrow interval indicates greater statistical precision, whereas a wide interval indicates that the observed data permit a broader range of plausible underlying effects under the chosen statistical model.
A p-value addresses evidence against a specified null hypothesis; it does not quantify the clinical magnitude of the treatment effect. Two analyses can have similar effect estimates but different p-values because of differences in sample size or variability. For STOIC's primary endpoint, the event proportions and an interpretable effect measure would remain essential even if a hypothesis test were performed.
9. Composite Endpoint Considerations
The registered primary endpoint combines two clinically related outcomes: emergency department attendance and hospitalisation related to COVID-19. Composite endpoints can increase the number of participants experiencing the overall endpoint, which may provide more information for statistical comparison than either component considered alone.
What the composite answers
The primary endpoint addresses whether a participant experienced either of the specified components during Day 1 to day 28.
What it does not establish
A treatment effect on the composite does not automatically establish an identical effect on each individual component. Component-specific frequencies and effects would be needed to evaluate that question.
This distinction is especially important when interpreting a composite outcome. The two components can differ in clinical severity, frequency, and susceptibility to ascertainment or treatment effects. The overall endpoint should therefore not be mentally substituted for either component alone.
10. Analysis Population and Randomization
The registry identifies STOIC as randomized with two arms and an enrollment of 146 participants. These design features establish the basic framework for comparing treatment groups, but the record summarized here does not specify detailed analysis-population definitions or a posted statistical analysis plan.
| Design feature | Registry information | Statistical implication |
|---|---|---|
| Allocation | Randomized | Supports comparison according to randomized treatment assignment. |
| Design model | Parallel | Participants are compared across concurrent randomized treatment groups. |
| Masking | None | Outcome interpretation should consider the potential consequences of an open-label design. |
| Number of arms | 2 | The primary treatment comparison is between two randomized groups. |
| Enrollment | 146 | Determines the overall amount of participant-level information available for the study comparison. |
11. Planned Analysis
The ClinicalTrials.gov record identifies emergency department attendance or hospitalisation related to COVID-19 as the primary endpoint, evaluated from Day 1 to day 28. No results or formal statistical analyses are posted in the record.
For this type of fixed-window binary endpoint, the primary analysis would typically compare the proportion of participants experiencing the endpoint in the two randomized arms. Appropriate reporting would include the event count and denominator in each arm, the absolute risk difference, and a relative measure such as the risk ratio, each accompanied by an appropriate confidence interval.
| Statistical quantity | Purpose |
|---|---|
| Event count | Shows how many participants experienced emergency department attendance or hospitalisation related to COVID-19. |
| Event proportion | Describes the observed probability of the endpoint in each randomized arm. |
| Risk difference | Quantifies the absolute difference in endpoint probability between arms. |
| Risk ratio | Quantifies the relative endpoint probability between arms. |
| Confidence interval | Describes statistical uncertainty around the estimated treatment effect. |
| Hypothesis test | Assesses the compatibility of the observed data with a specified null hypothesis. |
12. What the Primary Endpoint Would Mean Statistically
The event probability answers: what proportion of participants experienced emergency department attendance or hospitalisation related to COVID-19 during Day 1 to day 28? This is the most direct statistical description of the registered primary outcome.
The risk difference answers how many more or fewer participants per unit of population experienced the endpoint in one randomized arm than the other, based on the observed proportions.
The risk ratio answers how the probability of the endpoint in one arm compares proportionally with the probability in the other arm. It does not indicate how many individual participants benefited, nor does it replace the underlying absolute event rates.
13. Limitations and Interpretation Issues
- No posted numerical results: The ClinicalTrials.gov record does not contain formal statistical analyses or numerical outcome estimates for the registered primary endpoint.
- Composite endpoint: Emergency department attendance and hospitalisation are combined into one primary outcome, so an overall composite effect would not by itself establish the treatment effect for each component separately.
- Open-label design: The registry identifies the study as having no masking, which can introduce potential differences in behaviour, clinical management, or outcome assessment.
- Fixed follow-up window: The primary endpoint is specifically defined from Day 1 to day 28. Interpretation should remain tied to that registered period.
- Analysis-population detail: The registry information summarized here does not specify detailed efficacy analysis-population rules or a posted statistical analysis plan.
- Comparator specification: The registry fields identify two randomized arms and the budesonide dry powder inhaler intervention, but the summarized record does not specify the other arm.
- Event frequency: The statistical precision of a binary endpoint depends strongly on the number of participants experiencing the event. Sparse events can produce unstable relative estimates and wide confidence intervals.
- Randomized comparison versus individual outcome: A randomized trial estimates differences between treatment groups at the population level. It does not establish that every participant would experience the same treatment effect.
14. Why This Trial Matters Statistically
STOIC provides a useful teaching example because its primary endpoint illustrates how a randomized treatment question can be translated into a clearly defined binary clinical outcome. The trial also demonstrates the importance of distinguishing the registered endpoint from the statistical quantities used to analyse it.
| Concept | How it appears in STOIC |
|---|---|
| Randomization | Participants were allocated randomly to two study arms. |
| Parallel design | The registry identifies a parallel trial model rather than a crossover design. |
| Binary endpoint | The primary outcome can be represented by whether the specified event occurred during Day 1 to day 28. |
| Risk difference | Provides an absolute comparison of event probability between randomized groups. |
| Risk ratio | Provides a relative comparison of event probability between randomized groups. |
| Confidence interval | Quantifies uncertainty around a treatment-effect estimate. |
| Composite endpoint | Emergency department attendance and hospitalisation are combined in the primary endpoint. |
| Open-label design | The registry identifies the study as having no masking. |
| Fixed follow-up | The primary endpoint is assessed from Day 1 to day 28. |
| Phase 2 trial | The study is registered as phase 2 with treatment as its primary purpose. |
15. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The registered primary endpoint is a fixed-window binary event outcome. A statistical comparison would focus on the event proportions between the two randomized arms, together with an absolute and/or relative effect estimate and its uncertainty.
Clinical interpretation
The endpoint is intended to capture emergency department attendance or hospitalisation related to COVID-19 during the specified 28-day period. Its clinical interpretation depends on the observed frequency and component-specific pattern of those events.
16. Understanding What a Future Result Would and Would Not Mean
If numerical results were available, the first step would be to identify the number of participants experiencing the primary endpoint and the number evaluated in each randomized arm. Those counts would establish the observed event probabilities.
This is a reporting framework rather than a result for STOIC. The registry record does not provide the numerical quantities represented by x, y, n, RD, or RR.
The distinction between absolute and relative measures is important. A relative reduction can sound substantial while corresponding to a small absolute change when the underlying event is uncommon. Conversely, a modest relative effect can correspond to a meaningful absolute difference when the baseline event probability is higher.
17. Composite Endpoints and Component Effects
Because the registered endpoint combines emergency department attendance with hospitalisation, an eventual statistical analysis should distinguish the overall composite from its individual components.
Overall composite
The primary endpoint is satisfied when emergency department attendance or hospitalisation related to COVID-19 occurs during Day 1 to day 28.
Component interpretation
Separate component frequencies would be needed to determine whether an observed composite difference was driven mainly by emergency department attendance, hospitalisation, or both.
This is not merely a reporting detail. When components have different clinical meanings, the overall composite can conceal important differences in how a treatment affects the individual outcomes. A complete statistical interpretation should therefore retain the registered composite definition while examining its components when those data are available.
18. Statistical Precision in a Two-Arm Trial
The total enrollment of 146 provides the overall size of the randomized study population. Statistical precision for the primary binary endpoint, however, depends not only on total enrollment but also on how participants are distributed between arms and how many experience the event.
For binary outcomes, the amount of information available for estimating a treatment effect is strongly influenced by the number of observed events as well as the number of participants.
This is one reason a point estimate should never be interpreted without its confidence interval. A treatment-effect estimate based on relatively few events may be numerically large or small while still being statistically imprecise.
19. Randomization and Causal Interpretation
Randomization is central to the causal interpretation of a clinical trial. Because assignment is randomized rather than chosen according to participant characteristics or clinician preference, systematic differences in prognosis are less likely to explain the observed between-arm comparison at the time of assignment.
Randomization does not eliminate every source of uncertainty. Chance imbalances can still occur, and subsequent treatment exposure, adherence, withdrawals, missing observations, or outcome ascertainment can influence the observed data. The appropriate analysis therefore preserves the distinction between the randomized treatment comparison and what happens to individual participants after assignment.
20. Open-Label Design and Outcome Assessment
The registry classifies STOIC as having no masking. This is an important design characteristic because participants and investigators may know treatment assignment during the study.
For an endpoint involving emergency department attendance or hospitalisation, lack of masking does not automatically invalidate the endpoint. However, it creates a potential pathway through which treatment knowledge could influence decisions about seeking care, clinical management, or documentation. The extent of any such influence would require empirical information beyond the basic registry design fields.
21. Trial Status and Statistical Interpretation
The ClinicalTrials.gov record identifies STOIC as terminated. It also records a primary completion date of 2021-01-12.
Trial status is distinct from treatment effect. A terminated status does not, by itself, establish whether the intervention was beneficial, ineffective, harmful, or otherwise associated with the primary outcome. Those conclusions require the underlying outcome data and an appropriate statistical analysis.
22. A Practical Analysis Workflow for This Endpoint
This workflow illustrates the statistical logic from endpoint definition to interpretation. The most important step is to preserve the original endpoint definition: the analysis concerns emergency department attendance or hospitalisation related to COVID-19 during Day 1 to day 28, rather than an unrestricted measure of all subsequent clinical outcomes.
23. What Is Known From the Registry
| Question | Registry information |
|---|---|
| What was studied? | Budesonide dry powder inhaler in coronavirus infection. |
| What phase? | Phase 2. |
| How were participants assigned? | Randomized allocation. |
| What was the design? | Parallel. |
| Was the study masked? | No. |
| How many participants? | 146. |
| How many arms? | 2. |
| What was the primary endpoint? | Emergency department attendance of hospitalisation related to COVID-19. |
| When was it assessed? | Day 1 to day 28. |
| Were numerical statistical analyses posted? | No. |
24. Summary
STOIC is a randomized phase 2, two-arm, parallel trial of budesonide dry powder inhaler in coronavirus infection. Its registered primary endpoint is emergency department attendance or hospitalisation related to COVID-19 from Day 1 to day 28. The registry identifies the study as unmasked and lists an enrollment of 146 participants.
Statistically, the primary endpoint is naturally expressed as a fixed-window binary outcome. The core analysis would compare event probabilities between the randomized arms, with risk difference and risk ratio providing complementary absolute and relative descriptions. Confidence intervals would quantify precision, while interpretation of the composite endpoint should distinguish the combined outcome from its individual components.
The ClinicalTrials.gov record does not post formal statistical analyses or numerical results for the primary endpoint. Consequently, the central statistical lesson from the registry record is the relationship between randomization, endpoint definition, fixed follow-up, binary-event analysis, and uncertainty.
Continue through the Clinical Biostats knowledge graph
Explore additional clinical-trial analyses and statistical methods across Clinical Biostats.
25. Sources
- ClinicalTrials.gov: STOIC (NCT04416399).