This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. View the ClinicalTrials.gov record.
1. Trial at a Glance
TBTC Study 22 was a randomized, parallel-group, phase 3 treatment trial in pulmonary tuberculosis. The registry reports an enrollment of 1000 participants and identifies rifapentine and isoniazid as the study interventions.
| Feature | TBTC Study 22 |
|---|---|
| Official title | TBTC Study 22: Efficacy of Once-Weekly Rifapentine and Isoniazid in Treatment of Tuberculosis |
| ClinicalTrials.gov identifier | NCT00023335 |
| Phase | Phase 3 |
| Status | Completed |
| Condition | Pulmonary Tuberculosis |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 1000 |
| Lead sponsor | Centers for Disease Control and Prevention |
| Sponsor type | Federal |
| Start date | April 1995 |
| Interventions | Rifapentine; Isoniazid |
2. Clinical Question
The registry title frames the study around the efficacy of once-weekly rifapentine and isoniazid in the treatment of tuberculosis. The registered condition is pulmonary tuberculosis, and the study's primary purpose is treatment.
Population
Participants with pulmonary tuberculosis, as identified by the registry condition.
Interventions
Rifapentine and isoniazid.
Primary purpose
Treatment.
Primary statistical question
The registry title identifies efficacy of once-weekly rifapentine and isoniazid as the central clinical question.
The registry does not list a registered primary endpoint. Consequently, the ClinicalTrials.gov record does not specify a primary outcome definition, assessment time frame, or endpoint-specific estimand from which a formal primary efficacy comparison can be described.
3. Trial Design
Why randomization matters
Randomization is the central design feature supporting a causal comparison between treatment groups. When participants are assigned randomly, measured and unmeasured baseline characteristics can, in expectation, be balanced between groups. That balance allows differences observed after treatment to be interpreted as differences associated with assignment, subject to adherence, follow-up, missing data, and other features of the trial.
The registry identifies the study as randomized and parallel but does not provide additional allocation details such as an allocation ratio or stratification factors. Those design elements therefore cannot be characterized from the registry record.
What the absence of masking means
The registry classifies masking as none. In an unmasked trial, participants and study personnel are not described as being blinded to treatment assignment. The statistical implications depend on the endpoint: subjective outcomes can be more vulnerable to knowledge of treatment assignment, whereas objectively defined outcomes may be less susceptible to this particular source of bias.
4. Endpoints
The registry therefore does not provide a primary endpoint definition or time frame for a formal endpoint-specific statistical analysis. It also does not provide a registered secondary endpoint set in the record represented here.
| Endpoint information | Registry record |
|---|---|
| Registered primary endpoints | None listed |
| Primary endpoint definition | Not listed |
| Primary endpoint time frame | Not listed |
| Registered secondary endpoints | None listed in the record |
Why endpoint definitions matter
A clinical-trial result is meaningful only in relation to a clearly defined endpoint. For example, an efficacy endpoint might be binary, continuous, or time-to-event. Each structure leads to a different statistical estimand and analysis. A binary endpoint can commonly be summarized using a risk difference, risk ratio, or odds ratio; a continuous endpoint can be analyzed using a difference in means or another appropriate model; and a time-to-event endpoint generally requires explicit handling of censoring.
Because the registry does not list the primary endpoint for this study, the statistical method cannot be tied to a specific registered outcome definition or assessment time point.
5. Statistical Methodology
Randomized comparison
The principal statistical structure is a randomized comparison in a parallel-group phase 3 treatment trial. The fundamental analysis would compare outcomes according to randomized treatment assignment, preserving the treatment contrast created by randomization.
Analysis depends on endpoint type
The appropriate statistical method is determined by the endpoint's scale and timing. Without a registered primary endpoint definition, it is not possible to identify one endpoint-specific model as the definitive primary analysis from the ClinicalTrials.gov record.
The endpoint determines whether the analysis should focus on proportions, means, rates, or time-to-event distributions. Follow-up and censoring requirements further determine the appropriate model.
Binary efficacy endpoint
If efficacy were defined as a binary outcome, a standard randomized comparison could report the event proportion in each group together with an absolute risk difference and a relative effect measure. Confidence intervals would quantify uncertainty around the estimated treatment contrast.
Continuous efficacy endpoint
If efficacy were defined using a continuous measurement, the analysis could compare treatment-group means or another prespecified summary. Regression methods could be used when baseline covariates or other design features need to be incorporated.
Time-to-event efficacy endpoint
If the efficacy outcome were defined as time to an event, Kaplan-Meier estimation could describe the event-time distribution and a log-rank or regression-based comparison could be used depending on the prespecified analysis plan. A Cox model is one common approach for estimating a hazard ratio, but it should only be presented as the primary model when supported by the endpoint definition and statistical analysis plan.
Intention-to-treat principle
For a randomized efficacy comparison, an intention-to-treat analysis generally evaluates participants according to the treatment assignment generated by randomization. This preserves the original randomized contrast even when treatment exposure or adherence differs after assignment.
6. Statistical Methods Explained
Why does randomization matter statistically?
Randomization creates the basis for comparing treatment groups without relying solely on statistical adjustment for measured baseline characteristics. With a sufficiently large randomized sample, chance rather than investigator selection determines treatment assignment, reducing systematic confounding at baseline.
Why is the endpoint definition essential?
The endpoint determines what quantity is being estimated. A proportion of participants with treatment success is fundamentally different from a change in a continuous measurement or time until recurrence. Statistical methods are designed around these different data structures.
Why would an unmasked design matter?
Knowledge of treatment assignment can influence behavior, reporting, clinical management, or assessment when outcomes are subjective. The importance of this issue depends on how the outcome is defined and measured.
What does an efficacy analysis estimate?
An efficacy analysis estimates the difference or relative contrast between randomized treatment groups for a prespecified outcome. The estimate should be accompanied by an uncertainty measure, such as a confidence interval, so readers can distinguish the observed treatment contrast from the precision of that estimate.
Why are confidence intervals useful?
A confidence interval communicates statistical precision around an estimate. A narrow interval indicates greater precision than a wide interval under the same statistical framework. The interval should be interpreted in relation to the clinical question and the estimand rather than treated as a range of individual patient outcomes.
Why can the analysis population matter?
Different analysis populations answer different questions. An intention-to-treat analysis maintains the randomized comparison, while a population defined by treatment exposure or protocol adherence can answer a different question. The choice should be prespecified because changing the analysis population after seeing outcomes can affect interpretation.
7. Planned Analysis
The ClinicalTrials.gov record identifies TBTC Study 22 as a randomized, parallel-group phase 3 treatment study evaluating the efficacy of once-weekly rifapentine and isoniazid in pulmonary tuberculosis. However, the record does not list a registered primary endpoint or endpoint-specific time frame.
What would normally be analyzed?
The prespecified efficacy endpoint would normally be compared between randomized treatment groups using a method appropriate to its statistical scale and follow-up structure.
What determines the method?
The endpoint definition, timing of assessment, censoring or missingness, and prespecified estimand determine the appropriate statistical analysis.
What would accompany the estimate?
A formal treatment comparison would normally include an effect estimate together with an appropriate measure of statistical uncertainty.
Registry results status
No formal statistical analyses are posted to ClinicalTrials.gov for this record.
The absence of a registered primary endpoint means that the registry alone does not establish which outcome was intended to serve as the primary confirmatory efficacy endpoint or when it was to be assessed. Results have not been posted on ClinicalTrials.gov.
How a primary efficacy analysis would ordinarily be structured
| Step | Statistical purpose |
|---|---|
| 1. Define the estimand | Specify the population, treatment contrast, outcome, time frame, and handling of intercurrent events. |
| 2. Define the endpoint | Specify exactly what constitutes treatment success, failure, progression, recurrence, or another outcome. |
| 3. Specify the analysis population | Establish whether the primary analysis follows randomized assignment or another prespecified population. |
| 4. Select the model | Choose a binary, continuous, count/rate, or time-to-event method according to the endpoint. |
| 5. Estimate the treatment effect | Report the prespecified contrast between treatment groups. |
| 6. Quantify uncertainty | Provide a confidence interval and, when specified, a hypothesis-test result. |
8. Statistical Interpretation of a Randomized Phase 3 Trial
Randomization establishes the structure of the comparison; it does not guarantee a statistically significant treatment difference. The strength of evidence comes from the observed effect estimate, its uncertainty, and the prespecified inferential framework.
An effect estimate describes the magnitude and direction of the observed treatment contrast. A p-value, when used, addresses compatibility of the observed data with a specified null hypothesis under the statistical model. A p-value is therefore not a measure of the clinical magnitude of an effect.
A treatment effect without an uncertainty interval can be difficult to interpret. Confidence intervals show how precisely the treatment contrast was estimated under the selected statistical framework and should be considered alongside the effect estimate.
A statistically detectable difference does not automatically establish that the magnitude is clinically important. Conversely, an apparently meaningful point estimate can remain uncertain when the sample size or number of outcome events is limited.
9. Sample Size and Information
Registered enrollment
Participants were registered for enrollment in the phase 3 study.
The enrollment of 1000 provides the scale of the randomized clinical investigation. Sample size is relevant to statistical precision because, all else equal, larger samples provide more information for estimating treatment effects. However, enrollment alone does not determine the precision of an efficacy estimate: the endpoint type, number of observed events, amount of follow-up, variability, missingness, and analysis method also contribute.
The registry record does not provide a registered primary endpoint, power calculation, effect-size assumption, allocation ratio, or event target. Those quantities therefore cannot be used to reconstruct the statistical design from the registry record.
10. Missing Data and Follow-Up
Missing-data handling is an important component of any clinical-trial analysis because incomplete outcome information can affect both the estimated treatment effect and its uncertainty.
Binary outcomes
Missing outcome status requires a prespecified strategy because different assumptions about missing participants can produce different treatment-effect estimates.
Continuous outcomes
Longitudinal measurements may require methods that account for repeated observations and assumptions about the missing-data mechanism.
Time-to-event outcomes
Patients without an observed event can contribute information until censoring, but the validity of the analysis depends on the censoring assumptions and endpoint definition.
Prespecification
The preferred missing-data strategy should be established before outcome results are known and should be aligned with the estimand.
The ClinicalTrials.gov record does not specify a missing-data or imputation method for a registered primary endpoint.
11. Multiplicity and Interim Analysis
Multiplicity becomes important when a trial formally tests multiple endpoints, multiple treatment comparisons, multiple time points, or multiple interim analyses. Repeated hypothesis testing can increase the probability of incorrectly rejecting at least one true null hypothesis unless the design includes appropriate error control.
| Design feature | Registry information |
|---|---|
| Multiple registered primary endpoints | No registered primary endpoints listed |
| Multiplicity procedure | Not specified in the registry record |
| Interim analysis | Not specified in the registry record |
| Alpha-spending procedure | Not specified in the registry record |
Without a registered primary endpoint hierarchy or interim-monitoring plan, no particular multiplicity adjustment or alpha-spending framework can be attributed to this trial from the ClinicalTrials.gov record.
12. Stratification and Covariate Adjustment
Stratification can improve the precision of a randomized comparison when important prognostic factors are known before treatment assignment. In a stratified analysis, treatment effects can be estimated while accounting for prespecified strata.
The registry identifies randomization but does not report stratification factors or a covariate-adjustment strategy. Consequently, no particular stratified statistical model should be attributed to TBTC Study 22 based solely on the registry record.
13. Non-Inferiority, Superiority, and the Choice of Hypothesis
The registry record describes the study as an efficacy study of once-weekly rifapentine and isoniazid, but it does not provide a registered primary endpoint or a statistical hypothesis framework such as a superiority or non-inferiority margin.
A non-inferiority trial requires a prespecified margin defining the largest clinically acceptable loss of efficacy. The confidence interval for the treatment contrast is then interpreted in relation to that margin. A superiority trial instead evaluates whether the evidence is sufficiently inconsistent with a null hypothesis of no treatment difference under the prespecified testing framework.
Because no such margin is reported in the registry record, a non-inferiority analysis cannot be attributed to the study from this record.
14. Bayesian Methods
Bayesian analysis incorporates prior information into a probability model and produces a posterior distribution for the parameter of interest. This differs from the conventional frequentist framework in both interpretation and inferential machinery.
The ClinicalTrials.gov record does not identify a Bayesian statistical method for TBTC Study 22. No Bayesian prior, posterior probability, or Bayesian decision rule is reported.
15. Trial History
Study start
The registry lists the study start date as April 1995.
Randomized treatment study
The study is classified as a randomized phase 3 trial with a parallel design and no masking.
Study status
The registry lists TBTC Study 22 as completed.
16. What Can Be Learned From the Design
Even without a posted endpoint analysis, the study illustrates several core principles of clinical-trial statistics.
| Concept | How it appears in TBTC Study 22 |
|---|---|
| Randomization | The registry classifies allocation as randomized. |
| Parallel-group design | The registry identifies the design model as parallel. |
| Phase 3 development | The study is classified as phase 3. |
| Treatment purpose | The primary purpose is listed as treatment. |
| Unmasked design | The registry classifies masking as none. |
| Large trial scale | Enrollment is 1000 participants. |
| Endpoint specification | No registered primary endpoints are listed. |
| Statistical analysis availability | No posted statistical analyses are listed. |
17. Limitations
- No registered primary endpoint: the ClinicalTrials.gov record does not identify a primary outcome definition or time frame.
- No posted statistical analyses: the registry does not contain formal statistical analyses for the study.
- Limited design detail: the registry identifies randomization and a parallel design but does not provide an allocation ratio or stratification factors.
- No registered statistical testing framework: the record does not specify a primary hypothesis, multiplicity strategy, or interim-analysis procedure.
- No reported missing-data strategy: the record does not identify an imputation method or other prespecified approach to incomplete endpoint information.
- No endpoint-specific estimand: without a registered primary endpoint, the population, treatment contrast, outcome, and time frame of the primary efficacy estimand cannot be fully specified from the record.
- Unmasked design: masking is listed as none, which can be relevant for subjective outcomes and assessment processes.
- Registry versus full statistical documentation: a registry record provides important study-design information but does not necessarily contain the complete statistical analysis plan.
18. Why This Trial Matters Statistically
TBTC Study 22 is a useful teaching example because it demonstrates an important distinction between trial design information and reported statistical results. The registry clearly establishes a randomized phase 3 parallel-group treatment study with 1000 participants, but the record represented here does not provide a registered primary endpoint or formal posted statistical analysis.
That distinction matters because statistical interpretation begins with the estimand. A reader cannot determine whether the primary comparison should use a risk ratio, mean difference, rate ratio, hazard ratio, or another effect measure until the endpoint and its timing have been defined.
The study also illustrates why registry completeness matters for statistical transparency. A trial can have a clearly described clinical objective while still requiring additional statistical documentation to determine exactly how efficacy was intended to be measured and tested.
Design lesson
Randomization and parallel allocation establish a strong framework for comparing treatment groups.
Endpoint lesson
A clinical question must be translated into a precisely defined endpoint before a statistical model can be selected.
Inference lesson
Effect estimates, confidence intervals, and hypothesis tests are meaningful only when linked to a prespecified estimand and analysis framework.
Transparency lesson
Registry information and formal statistical-analysis documentation serve complementary roles in understanding a clinical trial.
19. Statistical Methods Explained: A Deeper View
Why is a randomized design especially important in a phase 3 efficacy study?
A phase 3 treatment comparison generally seeks an estimate of the difference attributable to assigning one intervention rather than another. Randomization provides the foundation for that causal comparison by making treatment assignment independent of baseline characteristics in expectation.
Why isn't enrollment alone enough to judge statistical precision?
The enrollment of 1000 establishes the number of participants entered into the study, but precision depends on the outcome being measured. For a time-to-event endpoint, the number and timing of events are especially important. For a continuous endpoint, outcome variability matters. For a binary endpoint, the event rate also affects information.
Why does the lack of a primary endpoint matter?
A primary endpoint identifies the outcome around which the principal confirmatory analysis is organized. It also determines the effect measure, statistical model, timing of analysis, and interpretation of the trial's main hypothesis. Without that definition, the primary statistical estimand cannot be reconstructed from the registry record.
What is the difference between an effect estimate and a p-value?
The effect estimate describes the observed treatment contrast. A p-value evaluates the observed data relative to a specified null hypothesis under the assumed statistical framework. A very small p-value can accompany a small effect in a large study, while a potentially important effect can remain statistically uncertain in a smaller or less informative study.
Why does masking matter differently for different endpoints?
If an endpoint requires subjective judgment, awareness of treatment assignment can influence assessment. Objective outcomes are generally less susceptible to this mechanism, although knowledge of treatment can still affect other aspects of trial conduct. The statistical consequence therefore depends on how the endpoint is defined and measured.
Why should an analysis plan be prespecified?
Prespecification limits the opportunity to select endpoints, models, subgroups, or statistical thresholds after seeing the results. It also gives readers a framework for distinguishing confirmatory analyses from exploratory analyses.
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registry establishes a randomized phase 3 parallel-group study with 1000 participants, but does not provide a registered primary endpoint or posted statistical analysis from which a primary treatment-effect estimate can be evaluated.
Clinical interpretation
The study was designed around the efficacy of once-weekly rifapentine and isoniazid in pulmonary tuberculosis. The registry identifies the clinical objective and interventions but does not provide an endpoint-specific result in the record represented here.
Keeping these two levels separate is important. The design can be described from the registry even when a quantitative efficacy conclusion cannot be drawn from the registry's posted statistical fields.
21. Important Questions for Interpreting the Trial
| Question | Statistical significance |
|---|---|
| What was randomized? | The study used randomized allocation in a parallel design. |
| How large was the study? | Enrollment was 1000. |
| What was the clinical condition? | Pulmonary tuberculosis. |
| What interventions were registered? | Rifapentine and isoniazid. |
| Was the study masked? | No; masking was listed as none. |
| What was the registered primary endpoint? | No registered primary endpoint is listed. |
| Were formal statistical analyses posted? | No. |
| Can a primary hazard ratio, risk ratio, mean difference, or p-value be identified from the registry? | No endpoint-specific estimate is posted in the registry record. |
22. Sources
- ClinicalTrials.gov: TBTC Study 22 — NCT00023335.
Continue through the Clinical Biostats trial library
Explore additional clinical-trial analyses and statistical methods across randomized studies and therapeutic areas.
23. Record Summary
TBTC Study 22 is registered as a completed phase 3 randomized parallel-group treatment study in pulmonary tuberculosis, with an enrollment of 1000 participants. The registry identifies rifapentine and isoniazid as the interventions, classifies masking as none, and lists the Centers for Disease Control and Prevention as the lead sponsor.
From a statistical perspective, the most important feature of the available registry record is the distinction between a clearly defined randomized trial design and the absence of a registered primary endpoint or posted statistical analysis. A rigorous efficacy analysis requires an explicit endpoint, time frame, estimand, analysis population, statistical model, and inferential framework. Those elements are not specified sufficiently in the ClinicalTrials.gov record to support a primary quantitative treatment-effect analysis.