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
NIX was a completed phase 3, single-group treatment trial enrolling 109 participants with pulmonary tuberculosis. The registry identifies bedaquiline, PA-824, and linezolid as the interventions and defines the primary endpoint as the proportion of treatment failure through 6 months after the end of treatment.
| Feature | NIX |
|---|---|
| Trial acronym | NIX |
| ClinicalTrials.gov identifier | NCT02333799 |
| Phase | Phase 3 |
| Status | Completed |
| Condition | Pulmonary Tuberculosis |
| Enrollment | 109 |
| Allocation | NA |
| Design model | Single group |
| Masking | None |
| Primary purpose | Treatment |
| Number of arms | 1 |
| Lead sponsor | Global Alliance for TB Drug Development |
| Sponsor type | Other |
| Start date | 2015-03 |
| Primary completion date | 2019-01-14 |
| Results posted | Yes |
2. Clinical Question
The clinical question represented by the registry is centered on treatment with a three-drug regimen consisting of bedaquiline, PA-824, and linezolid in participants with drug-resistant tuberculosis. The primary registered outcome is the proportion of participants experiencing an unfavorable outcome defined through a composite of bacteriologic failure, relapse, or clinical failure.
Population
Participants with drug-resistant tuberculosis, with the registered condition identified as pulmonary tuberculosis.
Intervention
Bedaquiline, PA-824, and linezolid.
Comparator
The registry identifies a single-group design with one arm and does not identify a comparator arm.
Primary question
What proportion of participants experience treatment failure, defined as bacteriologic failure, bacteriologic relapse, or clinical failure, through 6 months after the end of treatment?
3. Trial Design
The single-group structure is statistically important. Unlike a parallel randomized trial, there is no concurrent randomized control group against which the regimen's outcome is directly compared. The primary endpoint is therefore naturally expressed as a proportion within the treated study population rather than as a treatment-versus-control effect estimate.
4. Treatment Regimen and Reported Safety Groups
The registry identifies bedaquiline, PA-824, and linezolid as the study interventions. The serious-adverse-event summary is reported using two linezolid dosing groupings, with affected participants and participants at risk provided for each grouping.
Bedaquiline + PA-824 + 600mg BID linezolid
- Serious adverse events: 13 affected
- Participants at risk: 44
- Reported as 13/44
Bedaquiline + PA-824 + 1200mg QD linezolid
- Serious adverse events: 6 affected
- Participants at risk: 65
- Reported as 6/65
The registry also reports a total of 19 serious-adverse-event-affected participants among 109 participants at risk. These safety groupings should be kept conceptually separate from the registry's single-group design description: the study has one registered arm, while the safety result is presented using two linezolid dosing groupings.
5. Primary Endpoint
| Endpoint | Time frame | Registry definition |
|---|---|---|
| Proportion of Treatment Failure (Unfavorable Outcome) | 6 Months post End of Treatment | Bacteriologic failure or relapse or clinical failure (derived). |
Bacteriologic failure
During the treatment period, bacteriologic failure is defined as failure to attain culture conversion to negative.
Bacteriologic relapse
During the follow-up period, bacteriologic relapse is defined as failure to maintain culture conversion to negative status in culture, with conversion to positive status involving a Mycobacterium tuberculosis strain that is genetically identical to the infecting strain at baseline.
Clinical failure
Clinical failure is defined as a change from protocol-specified tuberculosis treatment due to treatment failure, retreatment for tuberculosis during follow-up, or tuberculosis-related death.
6. How the Primary Endpoint Should Be Read
The primary endpoint is a binary composite outcome: each participant can ultimately be classified according to whether an unfavorable treatment outcome occurred under the registry definition during the specified observation window.
where x is the number of participants experiencing the defined unfavorable outcome and n is the population included in the endpoint analysis.
The important statistical feature is the composite definition. A participant does not need to experience every component to meet the endpoint. Bacteriologic failure, bacteriologic relapse, and clinical failure are alternative routes to an unfavorable outcome. This makes the endpoint clinically broader than a single laboratory event.
The six-month time frame also matters. The endpoint is not simply a statement about whether failure ever occurred. It is explicitly defined through 6 months after the end of treatment. The denominator and endpoint status therefore need to be aligned with that observation period.
7. Planned Analysis
The registry reports that results are posted for the primary endpoint, but no formal statistical analyses are posted in the ClinicalTrials.gov record. The registry therefore provides the endpoint definition and time frame without a posted effect estimate, confidence interval, or p-value for the primary endpoint.
For an endpoint defined as the proportion of treatment failure, the natural statistical analysis is a binomial proportion analysis. The principal descriptive quantity is the observed proportion of participants meeting the unfavorable-outcome definition. A confidence interval can then quantify uncertainty around that proportion.
Primary quantity
The treatment-failure proportion is the number of participants meeting the composite unfavorable-outcome definition divided by the relevant endpoint-analysis population.
Precision
A confidence interval around the proportion describes statistical uncertainty arising from the finite study population.
Single-group interpretation
Without a randomized comparator, the primary result is principally an estimate of the observed outcome proportion under the study regimen.
Clinical interpretation
The composite endpoint must be interpreted using its registry definition rather than treating every unfavorable outcome as the same clinical event.
8. Statistical Methods Explained
Why is a proportion the natural summary for the primary endpoint?
The registered primary endpoint asks for the proportion of treatment failure. Each participant is classified according to whether bacteriologic failure, bacteriologic relapse, or clinical failure occurred through the specified six-month follow-up period. A proportion directly summarizes the frequency of that binary outcome.
What does the denominator mean?
The denominator represents the participants belonging to the population used for the endpoint calculation. This is particularly important in clinical trials because the meaning of a reported percentage depends on who was eligible for inclusion in that calculation. The ClinicalTrials.gov record gives the overall enrollment as 109 but does not provide a separate posted endpoint-analysis denominator.
Why use a confidence interval for a single-group proportion?
A sample proportion is an estimate rather than a perfectly known population quantity. A confidence interval provides a measure of statistical precision around the observed proportion. It is especially useful when the study is interpreted as estimating the frequency of an unfavorable outcome under a single treatment regimen.
Why is a single-group design different from a randomized comparison?
In a randomized trial, differences between treatment groups can be compared while randomization provides the basis for balancing measured and unmeasured prognostic factors in expectation. NIX is registered as a single-group study with one arm. Its primary endpoint therefore describes outcomes within the treated population rather than a randomized treatment-control contrast.
Why is the composite endpoint statistically important?
Combining bacteriologic failure, bacteriologic relapse, and clinical failure increases the range of events that qualify as an unfavorable outcome. The resulting proportion can provide a broad measure of treatment success or failure, but it also compresses potentially different clinical processes into one binary endpoint.
Why does the six-month follow-up window matter?
Endpoint definitions in clinical trials are inseparable from their time frames. A failure occurring during treatment and a relapse occurring during post-treatment follow-up are incorporated into the same registered composite, but the endpoint is evaluated through a specified period of 6 months after treatment ends. Changing the observation window would change the estimand being described.
9. Statistical Interpretation of the Primary Endpoint
The primary treatment-failure proportion is intended to quantify the fraction of participants who experience an unfavorable outcome under the registry's composite definition through 6 months after the end of treatment. It is a frequency measure, not a relative treatment effect.
A treatment-failure proportion does not by itself establish how the regimen compares with another treatment. In a single-group design, there is no concurrent randomized control arm in the registry against which the proportion can be directly contrasted.
The confidence interval around a proportion addresses precision of the estimated frequency. A narrower interval indicates greater statistical precision, whereas a wider interval indicates more uncertainty. The interval does not describe the range of outcomes that an individual participant could experience.
If a formal hypothesis test is performed, its p-value addresses evidence against a specified null hypothesis. It does not measure the magnitude of treatment failure, clinical importance, or the probability that the treatment is effective. For a single-group proportion, the null value would need to be defined in advance for such a test to have a clear confirmatory interpretation.
10. Understanding the Composite Outcome
The registry's endpoint is deliberately broader than a simple culture-conversion endpoint. It combines three pathways: failure to achieve culture conversion during treatment, microbiologically defined relapse during follow-up, and specified clinical events including retreatment or tuberculosis-related death.
| Component | When assessed | Statistical role |
|---|---|---|
| Bacteriologic failure | During treatment | One pathway to the unfavorable composite outcome |
| Bacteriologic relapse | During follow-up | One pathway to the unfavorable composite outcome |
| Clinical failure | During treatment or follow-up as defined | One pathway to the unfavorable composite outcome |
The composite definition also illustrates why endpoint adjudication rules matter. Culture conversion is not simply one negative sample: the registry requires at least 2 consecutive culture negative/positive samples at least 7 days apart. The special rule for participants unable to produce sputum further illustrates how a protocol can translate incomplete specimen availability into an endpoint classification when the participant is clinically considered to be responding well.
11. Safety Results
The registry reports serious adverse events by the two linezolid dosing groupings. Overall, 19 of 109 participants are reported as affected by serious adverse events.
| Reported safety grouping | Affected | At risk | Registry result |
|---|---|---|---|
| Bedaquiline + PA-824 + 600mg BID linezolid | 13 | 44 | 13/44 |
| Bedaquiline + PA-824 + 1200mg QD linezolid | 6 | 65 | 6/65 |
| Total | 19 | 109 | 19/109 |
The two reported safety groups contain different numbers of participants at risk. Consequently, comparing the raw affected counts alone would not be an appropriate comparison of safety frequency. The reported denominators are essential to interpreting the safety data.
12. A Statistical Reading of the Safety Data
The safety results illustrate a basic principle of clinical-trial analysis: counts and denominators must be considered together. The first reported grouping has 13 affected participants among 44 at risk, while the second has 6 among 65. The total is 19 among 109.
These data can be summarized descriptively as event proportions. However, the registry information does not provide a formal comparative safety analysis, confidence intervals, adjustment variables, or a hypothesis test comparing the two dosing groupings. It would therefore be inappropriate to attach a formal inferential conclusion to the difference in raw event counts alone.
For a safety proportion, the denominator defines the population in which the frequency is being described. Different denominators can make raw event counts misleading when viewed without context.
13. Analysis Populations and Denominators
Clinical-trial analysis depends on clearly defined populations. The registry establishes an overall enrollment of 109 and provides a total serious-adverse-event denominator of 109. It does not provide a separate posted denominator for the primary treatment-failure analysis in the information summarized here.
| Quantity | Reported population | Statistical implication |
|---|---|---|
| Overall enrollment | 109 | Describes the size of the registered study population. |
| Number of arms | 1 | Supports a single-group interpretation rather than a randomized treatment-control comparison. |
| Total serious-adverse-event population | 109 at risk | Provides the denominator for the reported total safety count. |
| Primary endpoint population | Separate denominator not reported here | The endpoint proportion requires the population used for its formal calculation. |
This distinction is not merely bookkeeping. If participants have different amounts of follow-up or if some participants are not evaluable for a particular endpoint, the analysis population determines what the resulting percentage actually represents.
14. Missing Data and Endpoint Classification
The registry's definition contains an explicit rule for participants who are unable to produce sputum. When such a participant is documented at a visit as unable to produce sputum and is clinically considered to be responding well to treatment, the participant is considered culture negative at that visit.
This is an example of a protocol-defined endpoint classification rule rather than a generic statistical imputation procedure. It is important to distinguish the two concepts. Statistical imputation generally replaces or estimates missing measurements using a specified statistical model or rule. Here, the registry describes how a particular clinical and specimen-availability circumstance is classified for purposes of culture status.
Observed laboratory information
Culture results provide the microbiologic basis for conversion and relapse assessment.
Protocol classification
The registry specifies how certain participants unable to produce sputum are classified when they are clinically considered to be responding well.
Composite endpoint
The final unfavorable-outcome classification can arise from bacteriologic or clinical pathways.
Statistical consequence
Endpoint rules can affect the observed numerator and therefore the estimated treatment-failure proportion.
15. Why a Single-Group Trial Requires Different Statistical Language
Randomized controlled trials often report quantities such as risk ratios, risk differences, odds ratios, or hazard ratios because they compare outcomes between treatment groups. NIX is registered as a single-group trial, so those comparative measures are not the natural primary summaries for the registered treatment-failure endpoint.
The central estimand is instead a within-study outcome proportion: the frequency of an unfavorable outcome under the study regimen during the specified follow-up period. This does not make the statistic less useful; it changes the question that the statistic answers.
| Statistical quantity | What it would answer | Fit with the registered NIX design |
|---|---|---|
| Proportion of treatment failure | How frequently the registered unfavorable outcome occurred in the study population | Directly aligned with the primary endpoint |
| Risk difference | How much outcome risk differs between two groups | Requires a comparison group |
| Risk ratio | How outcome risk compares between two groups | Requires a comparison group |
| Hazard ratio | How instantaneous event rates compare between groups over time | Not the natural primary measure for the registered single-group endpoint |
This distinction is central to responsible interpretation. A well-defined single-group proportion describes what happened among the participants who received the study regimen. It does not, by itself, identify what would have happened to the same participants under another regimen.
16. Interpreting the Study Without a Comparator
In a single-group design, the principal statistical uncertainty concerns estimation of the observed outcome frequency. The inferential problem is therefore different from the one faced by a randomized comparison.
The observed treatment-failure proportion describes the frequency of the registered unfavorable outcome in the study population during the specified endpoint window.
A statement that one treatment is better or worse than another requires a valid comparison. A single-group study does not generate that comparison through randomization because there is no concurrent randomized control arm.
If a single-group proportion were compared with a historical or external benchmark, that comparison would depend on the validity and comparability of the benchmark population, endpoint definitions, follow-up, and treatment context. Such a benchmark is not identified in the registry information presented here.
17. What the Registry Reports — and What It Leaves Unspecified
The ClinicalTrials.gov record establishes the trial's basic design, enrollment, intervention, primary endpoint definition, endpoint time frame, and serious-adverse-event counts by the reported linezolid dosing groupings. It does not provide a posted formal statistical analysis for the primary endpoint in the information summarized here.
| Element | Registry information |
|---|---|
| Primary endpoint | Proportion of treatment failure (unfavorable outcome) |
| Primary endpoint time frame | 6 Months post End of Treatment |
| Results status | Yes |
| Formal posted statistical analyses | None |
| Overall enrollment | 109 |
| Serious adverse events | 19/109 total; 13/44 and 6/65 by reported linezolid dosing grouping |
This distinction is important when reading registry-based evidence. A registry can contain posted results while still providing limited inferential detail for a particular endpoint. An observed number or percentage should not automatically be interpreted as the output of a formal hypothesis test unless the corresponding analysis is documented.
18. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary endpoint is a single-group proportion defined by a composite of bacteriologic failure, bacteriologic relapse, and clinical failure through 6 months after the end of treatment. The appropriate basic inferential framework is estimation of a binomial proportion and its precision.
Clinical interpretation
The endpoint incorporates microbiologic and clinical pathways to treatment failure. Its meaning therefore depends on the complete registry definition rather than on treating every component as interchangeable.
The two perspectives should remain distinct. Statistical analysis describes the frequency and uncertainty of the observed outcome. Clinical interpretation considers what the outcome represents in the context of tuberculosis treatment and the endpoint's component definitions. Neither perspective creates a randomized comparison that is absent from the study design.
19. Important Limitations and Interpretation Issues
- Single-group design: the study has one registered arm, so the primary endpoint does not provide a randomized treatment-versus-control comparison.
- Composite endpoint: treatment failure can arise from bacteriologic failure, bacteriologic relapse, or clinical failure, which are clinically distinct pathways.
- Time-frame dependence: the registered primary endpoint is specifically evaluated through 6 months after the end of treatment.
- Endpoint denominator: the overall enrollment is 109, but a separate posted denominator for the primary endpoint analysis is not identified in the registry information summarized here.
- Culture classification: the protocol definition includes a specific rule for participants unable to produce sputum who are clinically considered to be responding well.
- Safety denominators: the two serious-adverse-event groupings contain different numbers of participants at risk, so raw affected counts should not be compared without their denominators.
- Formal inference: the ClinicalTrials.gov record reports results but does not provide a posted formal statistical analysis for the primary endpoint in the information summarized here.
- Generalizability: the study population and its treatment setting determine the population to which the observed single-group outcome proportion can reasonably be applied.
20. Why This Trial Matters Statistically
NIX is a useful statistical teaching case because it demonstrates that the design of a clinical trial determines the kinds of conclusions that its numerical results can support. The primary endpoint is a binary composite proportion, while the study itself is registered as a single-group phase 3 treatment trial.
| Concept | How it appears in NIX |
|---|---|
| Single-group design | One registered study arm with 109 participants |
| Binary endpoint | Participants are classified according to whether an unfavorable treatment outcome occurs |
| Composite endpoint | Bacteriologic failure, bacteriologic relapse, or clinical failure |
| Time frame | Through 6 months after the end of treatment |
| Endpoint classification | Culture conversion requires at least 2 consecutive culture negative/positive samples at least 7 days apart |
| Clinical endpoint rule | Specified treatment changes, retreatment, or tuberculosis-related death can constitute clinical failure |
| Single-group estimation | The primary statistical quantity is naturally a proportion rather than a treatment-effect ratio |
| Safety analysis | Serious adverse events are reported as affected participants over participants at risk |
| Denominator discipline | Overall enrollment and safety denominators must be distinguished from the primary endpoint analysis population |
21. A Practical Framework for Analyzing the NIX Endpoint
A rigorous analysis of the registered primary endpoint can be organized into four sequential questions.
1. Who is analyzed?
Identify the participants belonging to the prespecified endpoint-analysis population and establish the denominator.
2. Who meets the endpoint?
Apply the bacteriologic failure, relapse, and clinical-failure definitions exactly as specified.
3. What is the observed proportion?
Divide the number of participants meeting the composite endpoint by the endpoint-analysis denominator.
4. How precise is the estimate?
Report an appropriate confidence interval for the single-group proportion and distinguish precision from clinical importance.
This framework prevents several common errors. It avoids confusing enrollment with the endpoint denominator, prevents the composite endpoint from being reduced to one of its individual components, and keeps an observed proportion separate from a comparative treatment effect.
22. Serious Adverse Events and Denominator Effects
The safety data provide a compact example of why denominator-aware reporting matters. The reported serious-adverse-event counts are 13 among 44 participants in one dosing grouping and 6 among 65 in the other, for a total of 19 among 109.
The same count can have different clinical meaning when the number of participants at risk differs. Reporting both numerator and denominator preserves that information.
It is also important not to interpret these two safety groupings as if they were necessarily randomized treatment arms. The registry describes the overall study as a single-group design, while the safety result is presented using the two linezolid dosing descriptions. The available information therefore supports descriptive reporting of the safety counts and denominators, but not a causal comparison between the dosing groups.
23. What a Confidence Interval Would Add
For a single-group proportion, the confidence interval complements the point estimate. Suppose an endpoint analysis produces an observed proportion p. The point estimate tells the reader the observed frequency, while the confidence interval communicates how precisely that frequency has been estimated from the available participants.
A binomial confidence interval is then constructed around p̂ using an appropriate interval method. The choice of interval method affects coverage properties, especially when sample sizes or event counts are small.
A confidence interval should not be mistaken for a prediction interval for future participants. Nor does it quantify variation in treatment response among individuals. It addresses uncertainty in estimating the population proportion represented by the study sample under the stated statistical framework.
24. Primary Statistical Takeaways
Because NIX is registered as a single-group study, the primary treatment-failure endpoint is naturally summarized as a proportion rather than as a randomized treatment effect.
The unfavorable outcome is a composite of bacteriologic failure, bacteriologic relapse, and clinical failure. A valid analysis must preserve those definitions and the six-month post-treatment time frame.
The serious-adverse-event results demonstrate why affected counts should be reported with participants at risk: 13/44, 6/65, and 19/109.
A confidence interval describes uncertainty around an estimate. It does not establish clinical importance and does not turn a single-group study into a comparative trial.
25. Sources
- ClinicalTrials.gov: NCT02333799 — NIX.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical-trial analyses and statistical methods used to interpret study endpoints, safety outcomes, and trial designs.
26. Record Summary
NIX is a phase 3, completed, single-group treatment trial enrolling 109 participants with pulmonary tuberculosis. Its registered primary endpoint is the proportion of treatment failure (unfavorable outcome) through 6 months after the end of treatment, with treatment failure defined through bacteriologic failure, bacteriologic relapse, or clinical failure. The registry's culture-conversion rules are integral to endpoint classification, including the requirement for at least 2 consecutive culture negative/positive samples at least 7 days apart and the specified classification of participants unable to produce sputum who are clinically considered to be responding well.
Statistically, the most important feature of NIX is the relationship between its endpoint and its single-group design. The primary result is fundamentally an estimate of the frequency of a defined unfavorable outcome under the study regimen. Without a concurrent randomized comparator, that proportion should not be transformed conceptually into a treatment effect against an alternative regimen. The safety data reinforce the importance of denominator-aware reporting, with serious adverse events reported as 13/44, 6/65, and 19/109 across the reported dosing groupings and total population.