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
ENIGMA was a randomized, parallel-group, open-label phase 3 trial evaluating Bacillus Clausii in the management of diarrhea in children with gastroenteritis. The registry reports 264 enrolled participants and two intervention arms: Bacillus Clausii and oral rehydration therapy.
| Feature | ENIGMA |
|---|---|
| Trial acronym | ENIGMA |
| Phase | Phase 3 |
| Condition | Gastroenteritis |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 264 |
| Number of arms | 2 |
| Lead sponsor | Sanofi |
| Sponsor type | Industry |
| Status | Completed |
| ClinicalTrials.gov | NCT00457353 |
2. Clinical Question
The clinical question is whether treatment involving Bacillus Clausii, in the setting of childhood gastroenteritis, is associated with a reduction in the duration of diarrhea compared with the other intervention specified in the randomized trial, while also evaluating the incidence of adverse events.
Population
Children with gastroenteritis enrolled in the ENIGMA phase 3 randomized trial.
Intervention
Bacillus Clausii, identified in the registry as a drug intervention.
Comparator
Oral rehydration therapy, identified in the registry as an other intervention.
Primary questions
Does the randomized treatment comparison affect the duration of diarrhea, and what is the incidence of adverse events during the study period?
3. Trial Design
Bacillus Clausii
- Bacillus Clausii
- Registry intervention type: drug
- Randomized assignment
Oral rehydration therapy
- Oral rehydration therapy
- Registry intervention type: other
- Randomized assignment
The registry describes ENIGMA as randomized and parallel-group, with no masking. These design features determine how the statistical comparison should be interpreted. Randomization provides the basis for comparing the assigned treatment groups, while the absence of masking means treatment assignment was not concealed through a blinded design.
4. Study Timeline
Study start
The ENIGMA trial began in 2007-03 according to the registry.
Randomized treatment study
The phase 3 trial evaluated two parallel intervention groups in children with gastroenteritis.
Primary completion
The registry lists 2007-12 as the primary completion date.
Registry status
The study is listed as completed.
5. Endpoints
| Endpoint | Registry time frame | Statistical role |
|---|---|---|
| Reduction in duration of diarrhea | Throughout the treatment period | Primary efficacy endpoint |
| Incidence of adverse events | Throughout the study period | Primary safety endpoint |
Endpoint 1: Reduction in duration of diarrhea
The first registered primary endpoint is reduction in duration of diarrhea, assessed throughout the treatment period. Statistically, this wording identifies duration as a time-based clinical outcome, but the registry information available here does not specify the precise operational definition, measurement scale, or analysis population.
Endpoint 2: Incidence of adverse events
The second registered primary endpoint is the incidence of adverse events, assessed throughout the study period. Unlike diarrhea duration, adverse-event incidence is fundamentally a count or proportion outcome unless the registry specifies additional time-to-event methodology.
6. Planned Analysis
The registry lists two primary endpoints but does not post statistical analyses or numerical outcome results. The appropriate statistical framework can therefore be described from the endpoint types without assigning an observed treatment effect.
Primary efficacy endpoint: duration of diarrhea
Duration of diarrhea is a continuous or time-to-event outcome depending on how the endpoint was operationalized. If every participant had a clearly defined duration measured on a common scale and the distribution were suitable, a comparison of the two treatment groups could use a two-sample method such as a t-test or a nonparametric alternative. If the endpoint were defined as time from a prespecified starting point until resolution and some participants were censored before resolution, a survival-analysis framework such as Kaplan-Meier estimation and a group comparison using a log-rank test would be more appropriate.
The registry wording alone does not establish which of these structures was used. The statistical interpretation should therefore follow the actual endpoint definition and analysis plan rather than assuming that every duration variable is automatically a conventional continuous outcome.
Primary safety endpoint: incidence of adverse events
For adverse-event incidence, the simplest participant-level analysis compares the proportion of participants experiencing at least one adverse event in each randomized group. A risk ratio, risk difference, or odds ratio can describe the treatment comparison, with an appropriate confidence interval. If adverse events were analyzed according to their timing, an event-rate or time-to-first-event analysis could instead be relevant.
The registry endpoint title does not specify which event-level definition or statistical measure was used. Consequently, incidence should be distinguished from the total number of adverse-event episodes: one participant can experience more than one event.
The correct effect measure depends on whether the endpoint is a continuous duration, a time-to-event variable, a binary incidence outcome, or an event count.
7. Statistical Methodology
Randomized comparison
The central statistical feature of ENIGMA is randomized allocation. With 264 enrolled participants assigned across two parallel arms, the treatment comparison is intended to use the randomized groups as the primary basis for inference.
Randomization is important because it creates a design in which, in expectation, characteristics affecting the outcome are balanced between groups. It does not guarantee identical groups in every realized sample, and it does not eliminate measurement error or problems arising from missing observations.
Analysis of a continuous duration outcome
If diarrhea duration was recorded for every participant as a quantitative measurement with no censoring, the analysis could compare the two group means or medians. The mean difference would be particularly straightforward to interpret when the measurement scale is clinically meaningful.
A negative value would indicate a shorter average duration in the Bacillus Clausii group under this definition of the outcome; a positive value would indicate a longer average duration.
Analysis of a time-to-resolution endpoint
If duration of diarrhea was treated as time until resolution, Kaplan-Meier estimation could be used when some observations were censored. The treatment groups could then be compared using a log-rank test, and a Cox proportional-hazards model could provide a hazard ratio for resolution.
For a resolution endpoint, the direction of the hazard ratio differs from a disease-event endpoint: a larger hazard of resolution would generally correspond to faster resolution, assuming the event is defined as resolution.
Analysis of adverse-event incidence
If the primary safety endpoint is defined as the proportion of participants with at least one adverse event, the basic comparison is a two-group comparison of proportions.
A risk difference of zero indicates equal observed event proportions. A negative value would indicate a lower event proportion in the Bacillus Clausii group under the specified definition.
Confidence intervals
An estimated treatment effect should ordinarily be accompanied by a confidence interval. The interval describes statistical uncertainty around the estimated effect under the chosen model and sampling framework. It does not describe the range of individual patient outcomes.
P-values
A p-value addresses the compatibility of the observed data with a specified null hypothesis under the statistical model. It is not a measure of the magnitude or clinical importance of a treatment effect. For that reason, an effect estimate and confidence interval are more informative than a p-value considered by itself.
8. Statistical Methods Explained
Why does randomization matter?
Randomization determines treatment assignment using a chance-based allocation process. This makes the treatment groups comparable in expectation and supports a causal interpretation of a between-group difference, assuming the trial is conducted and analyzed appropriately.
Why is diarrhea duration different from adverse-event incidence?
Duration is a quantitative time-based outcome, whereas incidence asks whether an adverse event occurred. These endpoints therefore live on different statistical scales. A method appropriate for a duration measurement is not automatically appropriate for a binary adverse-event indicator.
When would Kaplan-Meier analysis be appropriate?
Kaplan-Meier estimation is appropriate when the endpoint is a time until an event and some observations may be right-censored. For example, if diarrhea resolution were the event and a participant's resolution status were not observed through the relevant follow-up, censoring could be incorporated into a survival analysis.
What would a hazard ratio mean if diarrhea resolution were the event?
A hazard ratio above 1 would represent a higher estimated instantaneous rate of reaching resolution in the Bacillus Clausii group relative to the oral rehydration therapy group, assuming that resolution is the event and the model is specified accordingly. It would not mean that the probability of resolution is multiplied by the hazard ratio at every time point.
Why report a confidence interval?
An estimate without an uncertainty interval gives an incomplete picture. A confidence interval shows how precisely the treatment effect has been estimated under the selected statistical framework. Wider intervals indicate greater statistical uncertainty, while narrower intervals indicate greater precision.
Why is adverse-event incidence not the same as adverse-event burden?
Incidence can describe the proportion of participants experiencing at least one event, but it does not necessarily describe how many events each participant experienced, how severe those events were, or how long they lasted. Those are separate safety questions and require appropriately defined endpoints.
9. Interpreting the Trial's Design
10. Analysis Population and Missing Data
The registry information identifies enrollment and randomized allocation but does not provide a detailed analysis-population definition or a posted missing-data strategy.
For a randomized trial, the primary efficacy analysis would ordinarily preserve the randomized comparison as far as possible. The exact handling of participants who discontinue treatment, have incomplete diarrhea-duration measurements, or otherwise lack endpoint data should be prespecified because different assumptions about missingness can change the estimated treatment effect.
For a duration endpoint, missing follow-up can be particularly important. Treating every missing observation as if it represented the same clinical outcome can introduce bias. If the endpoint is a time-to-event measure, censoring rules need to be defined before analysis.
11. Safety Analysis Framework
Incidence of adverse events is registered as a primary endpoint throughout the study period. A statistically interpretable safety analysis should first define the unit of analysis.
| Safety question | Possible statistical representation |
|---|---|
| Did a participant experience at least one adverse event? | Binary outcome; compare participant-level incidence between randomized groups. |
| How many adverse-event episodes occurred? | Event-count outcome; requires methods that account for repeated events within participants. |
| How quickly did the first adverse event occur? | Time-to-event outcome; survival methods may be appropriate. |
| How severe were the adverse events? | Categorical outcome requiring severity definitions and an appropriate comparison. |
The registry endpoint is specifically incidence of adverse events. It does not, in the information available here, provide event counts, severity categories, serious-adverse-event rates, or treatment-specific safety estimates. Those quantities should therefore not be substituted for the registered endpoint.
12. Statistical Interpretation of Duration Outcomes
If the endpoint were analyzed as a mean difference, the estimate would quantify the difference in average diarrhea duration between the randomized groups. If analyzed as a median difference, it would compare central points of the two distributions. If analyzed as a hazard ratio for resolution, it would compare the instantaneous rates of resolution over time.
A treatment-group difference in average duration would not mean that every child experienced exactly that amount of shortening. Likewise, a hazard ratio would not represent a fixed percentage reduction in each participant's duration.
The same clinical concept can produce different statistical estimands depending on whether duration is summarized by a mean, median, or time-to-event model. The interpretation should therefore be tied to the actual analysis definition rather than to the endpoint name alone.
13. Statistical Interpretation of Adverse-Event Incidence
A risk difference compares the proportion of participants experiencing an adverse event between the two randomized groups. It is expressed directly on the probability scale and can therefore be useful for communicating absolute differences.
A risk ratio divides the event probability in one group by the event probability in the other. A value of 1 represents equal risks. Values above or below 1 indicate relative differences, but they do not directly state the absolute difference in event probability.
A small absolute difference can correspond to a substantial relative ratio when the underlying event probability is low, while a large absolute difference can coexist with a more modest relative ratio when baseline risk is high. Both scales can therefore contribute to interpretation.
14. Statistical Assumptions and Diagnostics
If duration is analyzed with a t-test or linear model
The distribution of duration and the adequacy of the mean-based model should be considered. Strong skewness, outliers, unequal variances, or other departures from model assumptions can affect inference. Transformations or nonparametric methods may be considered when justified by the data and prespecified analysis plan.
If duration is analyzed with a Cox model
A Cox proportional-hazards analysis assumes that the relative hazard between groups is adequately represented by the model over time. This assumption should be assessed rather than treated as automatic. A hazard ratio is most useful when its underlying model describes the observed event process adequately.
If adverse-event incidence is analyzed as a binary endpoint
The analysis should account for the randomized group sizes and the observed number of participants with and without events. When event counts are small, asymptotic methods may be unreliable and exact or other small-sample approaches may be considered.
15. Multiplicity and Primary Endpoints
ENIGMA has two registered primary endpoints: reduction in duration of diarrhea and incidence of adverse events. When a study has multiple primary endpoints, the statistical interpretation depends on the prespecified testing strategy.
| Endpoint | Role | Key statistical issue |
|---|---|---|
| Reduction in duration of diarrhea | Primary efficacy endpoint | Requires an analysis appropriate to the endpoint's actual measurement and censoring structure. |
| Incidence of adverse events | Primary safety endpoint | Requires a clearly defined participant-level or event-level safety estimand. |
If both endpoints are intended to support a formal confirmatory claim, the protocol would ordinarily specify how type I error is controlled across the primary hypotheses. Without a posted multiplicity procedure, the two endpoints should not automatically be interpreted as though each had an independent full type I error allowance.
16. What Is Known and What Is Not Reported
| Element | Registry information |
|---|---|
| Trial phase | Phase 3 |
| Enrollment | 264 |
| Number of arms | 2 |
| Allocation | Randomized |
| Design | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Primary efficacy endpoint | Reduction in duration of diarrhea, throughout the treatment period |
| Primary safety endpoint | Incidence of adverse events, throughout the study period |
| Posted statistical analyses | None |
The ClinicalTrials.gov record therefore establishes the trial's design, interventions, enrollment, and registered primary endpoints, but it does not provide numerical efficacy or safety estimates in the posted statistical analyses.
17. Limitations
- No posted statistical analyses: the registry does not provide numerical treatment-effect estimates, confidence intervals, or p-values for the registered primary endpoints.
- Endpoint specification: the registry endpoint name does not fully define how diarrhea duration was measured or analyzed.
- Open-label design: masking is listed as none, which can be relevant to outcomes influenced by participant reporting or clinical assessment.
- Safety endpoint definition: incidence of adverse events does not by itself specify severity, seriousness, recurrence, or the number of events per participant.
- Missing-data strategy: the registry information does not specify how incomplete primary-endpoint observations were handled.
- Multiplicity: two primary endpoints are registered, but a detailed type I error-control strategy is not reported in the available registry information.
- Analysis population: the available registry information does not provide a detailed definition of the population used for each endpoint.
- Statistical model: the registry information does not identify whether duration was analyzed as a continuous outcome, time-to-event outcome, or another endpoint structure.
18. Why This Trial Matters Statistically
ENIGMA is a useful teaching example because its two primary endpoints require different statistical thinking even though they arise from the same randomized clinical trial. The first endpoint concerns the duration of a clinical condition; the second concerns adverse-event incidence. Each requires a clearly defined estimand before a statistical test can be selected.
| Concept | How it appears in ENIGMA |
|---|---|
| Randomization | Participants were randomized to two intervention arms. |
| Parallel-group design | The registry identifies the design model as parallel. |
| Open-label treatment | Masking is listed as none. |
| Duration endpoint | Reduction in duration of diarrhea is a registered primary endpoint. |
| Safety endpoint | Incidence of adverse events is a registered primary endpoint. |
| Time frame | Diarrhea duration is assessed throughout the treatment period; adverse events are assessed throughout the study period. |
| Estimand selection | The appropriate effect measure depends on the exact operational definition of each endpoint. |
| Multiplicity | Two primary endpoints require attention to the prespecified testing strategy. |
| Missing data | Incomplete duration measurements can affect the analysis depending on the endpoint definition and missingness mechanism. |
19. A Practical Statistical Analysis Framework
A rigorous analysis of ENIGMA would proceed from the clinical estimand rather than selecting a statistical test first.
For diarrhea duration, the key decision is whether the endpoint is a fully observed quantitative duration or a time-to-resolution endpoint with censoring. For adverse events, the key decision is whether the estimand is participant-level incidence, event frequency, or time to first event. Those distinctions determine the appropriate statistical model and the meaning of its resulting estimate.
For diarrhea duration
Define the start and end of the measurement, examine the distribution, determine whether censoring exists, then select a continuous or time-to-event analysis accordingly.
For adverse events
Define whether the endpoint is at least one event per participant, total event count, or time to first event before selecting the corresponding statistical method.
For inference
Report an effect estimate with a confidence interval rather than relying on a p-value alone.
For interpretation
Keep efficacy and safety endpoints separate because they measure different dimensions of treatment effect.
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registry describes a randomized comparison with two primary endpoints, but no formal statistical analyses or numerical results are posted in the registry record. The appropriate analysis depends on the detailed definitions of the registered endpoints.
Clinical interpretation
The trial was designed to evaluate Bacillus Clausii in the management of diarrhea in children with gastroenteritis while monitoring adverse events throughout the study period. The registry record does not provide outcome estimates from which a treatment effect can be described.
21. Key Statistical Lessons
The ENIGMA design illustrates several general principles that apply well beyond this particular trial.
| Lesson | Statistical implication |
|---|---|
| Define the estimand first | The same clinical concept can require different methods depending on how the endpoint is operationalized. |
| Match the model to the endpoint | Continuous duration, time-to-event resolution, binary incidence, and event counts have different statistical structures. |
| Use randomized assignment correctly | The randomized groups form the principal basis for the treatment comparison. |
| Separate effect size from significance | A p-value does not quantify the magnitude of a treatment effect. |
| Report uncertainty | Confidence intervals communicate precision around an estimated effect. |
| Define safety outcomes carefully | Participant incidence and the number of adverse-event episodes are not interchangeable. |
| Consider missing observations | Incomplete endpoint data can affect estimates and should be handled according to a prespecified strategy. |
| Account for multiple primary endpoints | The interpretation of confirmatory evidence depends on the prespecified multiplicity strategy. |
22. Sources
- ClinicalTrials.gov: NCT00457353 — ENIGMA.
Continue through the Clinical Biostats knowledge graph
Explore the broader Clinical Biostats collection of statistical tutorials, calculators, and clinical-trial analyses.
23. Record Summary
ENIGMA is a completed phase 3 randomized, parallel-group trial enrolling 264 participants with gastroenteritis. The registry identifies two interventions, Bacillus Clausii and oral rehydration therapy, and two primary endpoints: reduction in duration of diarrhea throughout the treatment period and incidence of adverse events throughout the study period. The trial is listed as unmasked and treatment-focused. No formal statistical analyses are posted in the ClinicalTrials.gov record, so the principal statistical lesson is methodological: duration and adverse-event incidence must be defined precisely before the appropriate estimand, effect measure, and statistical model can be selected.