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
endTB is a completed phase 3 randomized trial evaluating newly approved drugs for multidrug-resistant tuberculosis. The registry describes a six-arm, parallel-group design with 754 enrolled participants and no masking.
| Feature | endTB |
|---|---|
| Official title | Evaluating Newly Approved Drugs for Multidrug-resistant TB |
| Phase | Phase 3 |
| Status | COMPLETED |
| Population | Tuberculosis, Multidrug-Resistant; Infection, Bacterial; Pulmonary Tuberculoses |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 754 |
| Number of arms | 6 |
| Lead sponsor | Médecins Sans Frontières, France |
| Start date | 2016-12 |
| Primary completion date | 2023-06 |
| ClinicalTrials.gov | NCT02754765 |
2. Clinical Question
The central statistical question is whether treatment strategies incorporating newly approved drugs can be evaluated against the clinical and bacteriological criteria that define a favorable outcome at week 73 after randomization in participants with multidrug-resistant tuberculosis.
Population
The registry identifies the conditions as multidrug-resistant tuberculosis, bacterial infection, and pulmonary tuberculosis.
Interventions
The registry lists bedaquiline, delamanid, clofazimine, levofloxacin, moxifloxacin, linezolid, and pyrazinamide among the study drugs.
Comparator
The registry lists a control arm consisting of an MDR-TB regimen consistent with WHO guidelines.
Primary question
What proportion of participants have a favorable outcome at week 73 after randomization?
3. Trial Design
Study interventions
Newly approved drugs
- Bedaquiline
- Delamanid
- Clofazimine
- Levofloxacin
- Moxifloxacin
- Linezolid
- Pyrazinamide
MDR-TB regimen
- Control arm MDR-TB regimen
- Regimen consistent with WHO guidelines
4. Endpoints
Primary endpoint: Week 73 Efficacy
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Week 73 Efficacy | Proportion of participants with favorable outcome at week 73. A participant's outcome is classified as favorable if the outcome is not classified as unfavorable and one of the prespecified bacteriological, radiological, and clinical criteria is satisfied. | Week 73 after randomization |
What constitutes a favorable outcome?
The registry specifies several routes by which a participant can meet the favorable-outcome definition. The definition is therefore not simply a single positive culture result at one visit. It incorporates the pattern and timing of culture results together with bacteriological, radiological, and clinical evolution.
Two recent negative cultures
The last two culture results are negative, with the two cultures obtained from sputum samples collected on separate visits and the latest collected between weeks 65 and 73.
Single recent negative culture
The last culture result from a sputum sample collected between weeks 65 and 73 is negative, with either no other post-baseline culture result or a penultimate positive culture attributable to laboratory cross contamination, together with favorable bacteriological, radiological, and clinical evolution.
No usable late culture
There is no culture result from a sputum sample collected between weeks 65 and 73, or that culture is positive because of laboratory cross contamination, while the most recent culture result is negative and bacteriological, radiological, and clinical evolution is favorable.
Unfavorable outcomes
The favorable definition applies only when the outcome is not classified as unfavorable and one of the specified favorable pathways is satisfied.
Statistically, this creates a binary participant-level endpoint: each participant ultimately contributes a favorable or unfavorable classification at the prespecified week 73 assessment, according to the registry definition.
5. Statistical Methodology
The registry identifies the primary endpoint as a proportion of participants with favorable outcome at week 73. For a binary endpoint of this form, the central analysis is a comparison of favorable-outcome proportions between randomized treatment groups or treatment strategies, using an analysis appropriate to the prespecified multi-arm design.
The resulting proportion summarizes the frequency of favorable outcomes at the specified week 73 assessment. A treatment comparison would then quantify the difference between the corresponding proportions, using the analysis population and comparison structure prespecified for the trial.
Why the endpoint is more than a simple culture rate
The week 73 endpoint incorporates longitudinal microbiological information. The definition requires particular attention to the most recent cultures, the timing of sputum collection, separate visits for paired cultures, and the possibility of laboratory cross contamination. It also incorporates bacteriological, radiological, and clinical evolution in specified circumstances.
This matters statistically because the endpoint is constructed from a sequence of observations rather than from a single measurement. A participant's final classification depends on the available culture history and the prespecified decision rules.
Analysis of a proportion
A favorable-outcome proportion can be summarized with a confidence interval to quantify sampling uncertainty. When two treatment groups are compared, common estimands include an absolute risk difference, a relative risk, or an odds ratio. The appropriate measure depends on the prespecified statistical analysis and the scientific question.
| Quantity | Interpretation |
|---|---|
| Favorable-outcome proportion | Absolute frequency of favorable outcomes at week 73 within a treatment group. |
| Risk difference | Difference in favorable-outcome proportions between two groups. |
| Relative risk | Ratio of favorable-outcome probabilities between two groups. |
| Odds ratio | Ratio of the odds of a favorable outcome between two groups. |
| Confidence interval | Quantifies uncertainty around an estimated treatment effect under the specified statistical framework. |
Multi-arm considerations
Because the registry describes six arms, the statistical interpretation cannot automatically be reduced to a single two-group comparison. A multi-arm trial can involve several treatment-versus-control comparisons, direct comparisons among active regimens, or a prespecified global hypothesis followed by individual comparisons.
The choice affects multiplicity. If several confirmatory comparisons are performed, the type I error associated with the family of tests generally requires explicit control. The registry information available here does not specify a multiplicity strategy, so no particular adjustment method is attributed to the trial.
Analysis population
For a randomized efficacy endpoint, an intention-to-treat framework is commonly used because participants are analyzed according to their randomized assignment. However, the registry information available here does not specify the precise analysis population or handling of every participant with missing week 73 information. Those details are therefore not assigned to the trial on this page.
6. Planned Analysis
The registry states that the primary endpoint will measure the proportion of participants with favorable outcome at week 73 after randomization. The favorable classification follows the prespecified culture-based and clinical criteria described above.
Primary statistical estimand
Participant-level binary efficacy endpoint assessed 73 weeks after randomization.
For this type of endpoint, the statistical analysis would normally begin by estimating the favorable-outcome proportion within each randomized treatment group. Treatment comparisons would then be expressed using an appropriate binary-outcome effect measure, accompanied by confidence intervals and, where prespecified, hypothesis tests.
The important quantity is the probability of being classified as favorable at the specified week 73 assessment. A difference in proportions describes an absolute difference in the frequency of favorable outcomes. A relative measure describes the same comparison on a multiplicative scale.
A confidence interval describes statistical uncertainty around the estimated comparison; it does not describe the range of outcomes that individual participants will experience.
A p-value, if used in the prespecified analysis, addresses compatibility with the null hypothesis under the statistical model and testing framework. It is not a measure of effect size or clinical importance.
7. Statistical Methods Explained
Why is the primary endpoint a proportion?
The endpoint classifies each participant at a fixed time point according to whether the participant has a favorable outcome. This makes the primary measure a binary outcome rather than a continuous measurement or a time-to-event endpoint.
Why does the week 73 definition use multiple cultures?
Repeated culture results provide a way to incorporate the longitudinal bacteriological course into the final classification. Requiring recent cultures from separate visits can distinguish a sustained negative pattern from an isolated observation.
Why does laboratory cross contamination appear in the endpoint definition?
The registry explicitly accounts for culture results that are positive because of laboratory cross contamination. This prevents a specified laboratory artifact from automatically determining an unfavorable classification when the other criteria support a favorable outcome.
Why are radiological and clinical findings included?
In the specified circumstances, the registry requires favorable bacteriological, radiological, and clinical evolution. The endpoint therefore integrates microbiological evidence with broader evidence of disease course rather than treating culture status as the only relevant component.
How would a treatment difference in proportions be interpreted?
Suppose two randomized groups had favorable-outcome proportions of \(p_1\) and \(p_2\). The absolute difference would be \(p_1-p_2\). A positive difference would indicate a higher observed proportion of favorable outcomes in the first group; a negative difference would indicate a lower proportion. The numerical values themselves would need to come from the trial results.
This scale is especially direct for a binary endpoint because it describes the difference in the probability of a favorable outcome at the prespecified assessment time.
Why is multiplicity important in a six-arm trial?
With six randomized arms, multiple treatment comparisons may be possible. If several hypotheses are tested confirmatorily, repeatedly using an unadjusted significance threshold can increase the familywise probability of a false-positive conclusion. A prespecified multiplicity strategy is therefore important when several comparisons share a common type I error budget.
What does the absence of a posted estimate mean?
It means that the ClinicalTrials.gov record available for this analysis does not contain a numerical statistical result for the primary endpoint. The study's endpoint definition and design can still be examined statistically, but an observed treatment effect requires an actual posted analysis or other permitted reported result.
8. Endpoint Construction and Statistical Interpretation
The endTB primary endpoint is particularly instructive because it combines a fixed assessment time with rules based on observations collected over a period of follow-up. The statistical endpoint is therefore best understood as a derived binary outcome.
| Component | Statistical role |
|---|---|
| Randomization | Defines the treatment assignment used for the randomized comparison. |
| Week 65–73 cultures | Provides the principal late-follow-up microbiological evidence incorporated into the favorable definition. |
| Separate visits | Required for the two-negative-culture pathway. |
| Cross contamination rule | Provides a prespecified way to handle specified positive culture results attributable to laboratory cross contamination. |
| Most recent culture | Provides microbiological evidence when the specified late-culture conditions are met. |
| Radiological evolution | Contributes to favorable classification in the specified endpoint pathways. |
| Clinical evolution | Contributes to favorable classification in the specified endpoint pathways. |
| Week 73 classification | Produces the final favorable/unfavorable efficacy endpoint for the primary analysis. |
Why the denominator matters
For a proportion, the denominator determines what population the percentage describes. In a randomized trial, the interpretation depends on whether the analysis includes all randomized participants, an evaluable population, or another prespecified analysis set. The available registry information does not specify the complete denominator-handling rules for the posted analysis.
This is particularly important for a complex endpoint. Missing culture information is not necessarily equivalent to a negative culture, and the registry itself distinguishes the absence of a late culture from several other culture-result patterns. The statistical analysis must therefore follow the prespecified endpoint and missing-data rules rather than treating unavailable measurements as automatically favorable.
9. Design Features That Matter Statistically
Randomization
Randomized allocation provides the fundamental basis for comparing outcomes among the six treatment arms while reducing systematic differences in treatment assignment.
Parallel design
Participants remain associated with their randomized treatment strategy rather than moving through a crossover sequence. The registry identifies the design model as parallel.
No masking
The registry describes the trial as having no masking. This is relevant when considering the possibility of treatment knowledge influencing clinical management or assessment, particularly for components that involve clinical or radiological evolution.
Fixed assessment time
The primary endpoint is defined at week 73 after randomization, providing a common time point for the primary efficacy assessment.
Why randomization does not remove every source of uncertainty
Randomization addresses systematic allocation of participants to treatment groups, but statistical uncertainty remains because the trial studies a finite sample. Confidence intervals quantify this sampling uncertainty. The interpretation of the final estimate also depends on the endpoint definition, analysis population, missing-data handling, and the number of comparisons performed.
10. Missing Data and Endpoint Classification
The registry's favorable-outcome definition explicitly addresses situations in which a sputum culture from weeks 65 to 73 is unavailable. This is important because a missing observation can otherwise create ambiguity about how the primary endpoint should be classified.
| Situation described by the registry | Role in favorable classification |
|---|---|
| Two negative cultures from separate visits, latest between weeks 65 and 73 | Can satisfy the favorable definition. |
| One negative culture from weeks 65 to 73, with specified conditions on other cultures | Can satisfy the favorable definition when bacteriological, radiological, and clinical evolution is favorable. |
| No culture from weeks 65 to 73, with the specified most recent culture negative | Can satisfy the favorable definition when bacteriological, radiological, and clinical evolution is favorable. |
| Positive culture attributable to laboratory cross contamination | Handled explicitly within the favorable-outcome definition. |
The statistical lesson is that endpoint-specific missing-data rules can be part of the endpoint definition itself. They should not be replaced with generic assumptions about missing observations.
11. Six-Arm Randomization and Comparisons
A six-arm randomized trial contains more information than a conventional two-arm comparison, but it also creates a more complex inferential structure. The scientific question may involve one or more comparisons among the active regimens and the control regimen.
A global hypothesis can, in general, ask whether favorable-outcome probabilities are equal across all randomized groups. Specific pairwise hypotheses can then address particular treatment comparisons. The registry information available here does not specify that this exact global test was prespecified for endTB.
The important point is conceptual: the number of arms determines the number of possible comparisons, while the primary statistical interpretation depends on which comparisons were prespecified as confirmatory and how multiplicity was controlled.
12. What a Confidence Interval Would Tell Us
A confidence interval around a treatment-effect estimate describes uncertainty under the specified statistical framework. Narrower intervals generally indicate greater statistical precision, while wider intervals indicate greater uncertainty.
A confidence interval around a group-level treatment effect does not describe the range of outcomes that an individual participant might experience.
A p-value addresses a hypothesis-testing question. It does not quantify how large or clinically important a treatment difference is. The estimated effect and its confidence interval are therefore essential parts of the statistical interpretation.
13. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary endpoint is a binary week 73 efficacy outcome derived from prespecified culture, bacteriological, radiological, and clinical criteria. A formal treatment comparison would be based on favorable-outcome proportions across the randomized groups.
Clinical interpretation
The endpoint is designed to summarize whether participants have achieved a favorable overall outcome at a defined point after randomization, incorporating microbiological evidence together with specified disease-course information.
14. Important Limitations and Interpretation Issues
- No posted numerical efficacy analysis: the ClinicalTrials.gov record contains the primary endpoint definition but no posted statistical analysis with an effect estimate, confidence interval, or p-value.
- Multi-arm complexity: six randomized arms create multiple possible treatment comparisons. The available registry information does not specify the complete confirmatory comparison hierarchy.
- Multiplicity: when multiple confirmatory hypotheses are tested, interpretation depends on the prespecified control of the overall type I error rate. The available record does not specify a multiplicity procedure.
- Endpoint derivation: favorable outcome is determined using prespecified rules involving multiple culture observations and, in specified circumstances, bacteriological, radiological, and clinical evolution.
- Missing culture information: the endpoint definition contains explicit rules for some situations without a late culture, but the complete statistical handling of all missing data is not specified in the available registry information.
- No masking: the registry describes the trial as unmasked. This is relevant to interpretation because some endpoint components involve clinical and radiological evolution.
- Analysis population: the available registry information does not provide a complete definition of the population used for a final statistical analysis.
- Arm-specific sample sizes: the record reports total enrollment of 754 and six arms but does not provide arm-specific enrollment counts in the information used here.
15. Why This Trial Matters Statistically
endTB is a useful teaching example because its primary efficacy endpoint illustrates how a modern clinical trial can transform longitudinal clinical and microbiological observations into a single prespecified binary outcome for randomized comparison.
| Concept | How it appears in endTB |
|---|---|
| Randomization | Participants were randomized among six treatment arms. |
| Parallel design | The registry identifies the study as a parallel-group design. |
| Binary endpoint | The primary endpoint classifies participants according to favorable versus unfavorable outcome at week 73. |
| Fixed-time assessment | The primary efficacy time frame is week 73 after randomization. |
| Composite decision rules | Favorable classification incorporates specified culture patterns and, in relevant pathways, bacteriological, radiological, and clinical evolution. |
| Repeated measurements | Culture results collected at different visits contribute to endpoint classification. |
| Missing-data considerations | The endpoint explicitly addresses circumstances in which a late culture is unavailable. |
| Multi-arm inference | Six randomized arms create multiple possible treatment comparisons. |
| Multiplicity | Multiple confirmatory comparisons would require an explicit inferential strategy. |
| Unmasked design | The registry describes masking as none. |
16. Statistical Concepts in This Trial
The endTB design provides a practical framework for understanding several core clinical-trial statistical concepts.
Binary efficacy analysis
The primary outcome is expressed as the proportion of participants meeting a favorable-outcome definition at a fixed time.
Absolute treatment effect
A risk difference directly describes the change in favorable-outcome probability between randomized groups.
Relative treatment effect
A relative risk expresses the favorable-outcome probability in one group relative to another.
Confidence intervals
Interval estimates quantify uncertainty around treatment-effect estimates and should be considered alongside point estimates.
Multiplicity
Multiple arms can produce multiple hypotheses, making the prespecified inferential hierarchy important.
Endpoint construction
Rules for repeated cultures and clinical evolution determine the participant-level binary outcome before group-level statistical analysis.
17. Sources
- ClinicalTrials.gov: NCT02754765 — endTB.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
Continue through the Clinical Biostats knowledge graph
Clinical trial pages connect study design, endpoints, and statistical reasoning so that readers can move from an individual trial to the broader principles used to analyze randomized evidence.
18. Record Summary
endTB is a completed phase 3 randomized, parallel-group trial with six arms and 754 enrolled participants, evaluating newly approved drugs for multidrug-resistant tuberculosis. Its primary endpoint is the proportion of participants with favorable outcome at week 73 after randomization, using a detailed definition based on recent culture results and, in specified circumstances, bacteriological, radiological, and clinical evolution.
From a statistical perspective, the trial is especially useful for understanding how a complex longitudinal clinical definition becomes a binary efficacy endpoint, how favorable-outcome proportions can be compared across randomized groups, and why a six-arm design raises important questions about the comparison hierarchy and multiplicity. The ClinicalTrials.gov record does not contain posted formal statistical analyses for the primary endpoint, so no numerical treatment-effect estimate is reported here.