This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
TBTC Study 26 was a randomized phase 3 parallel-group trial in tuberculosis prevention. It enrolled 8053 participants and compared a three-month regimen of once-weekly rifapentine plus isoniazid with a nine-month regimen of daily isoniazid.
| Feature | TBTC Study 26 |
|---|---|
| Trial name | TBTC Study 26 |
| Brief title | Three Months of Weekly Rifapentine and Isoniazid for M. Tuberculosis Infection |
| Phase | Phase 3 |
| Condition | Tuberculosis |
| Purpose | Prevention |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Enrollment | 8053 |
| Lead sponsor | Centers for Disease Control and Prevention |
| Status | Completed |
| Start | 2001-06 |
| Primary completion | 2010-10 |
| ClinicalTrials.gov | NCT00023452 |
2. Clinical Question
The central statistical question was whether the cumulative rate of tuberculosis disease through Month 33 with three months of once-weekly rifapentine plus isoniazid was non-inferior to the cumulative tuberculosis disease rate with nine months of daily isoniazid.
Population
Participants enrolled in TBTC Study 26 for prevention of tuberculosis, with the primary analysis conducted in the modified intention-to-treat population of enrolled participants who were eligible.
Intervention
Rifapentine plus isoniazid once weekly for three months, given by directly observed therapy.
Comparator
Isoniazid daily for nine months.
Primary question
Is the cumulative tuberculosis disease rate with 3RPT/INH sufficiently close to the 9INH rate to satisfy the prespecified non-inferiority criterion?
3. Trial Design
Three-month regimen
- Rifapentine plus isoniazid
- Once weekly
- Three months
- Given by directly observed therapy
Nine-month regimen
- Isoniazid
- Daily
- Nine months
The registry describes the study as randomized, parallel-group, and unmasked. The primary purpose was prevention rather than treatment of established tuberculosis disease.
4. Analysis Populations
The registry reports different analysis populations for different questions. This distinction is important because a non-inferiority efficacy analysis and a treatment-exposure safety analysis answer different statistical questions.
| Population | Definition / role |
|---|---|
| Modified intention-to-treat (MITT) | All participants who enrolled in the study and were eligible. The registry identifies source tuberculosis cases resistant to isoniazid or rifampin and source cases culture-negative for Mycobacterium tuberculosis among the reasons for ineligibility. |
| Per-protocol | Eligible enrolled participants in the MITT population who completed study drug within the targeted treatment period. The registry describes targeted completion as 11–12 3RPT/INH doses within 10–16 weeks or the specified isoniazid dosing period. |
| Safety population | Participants who enrolled and took at least one dose of study drug. |
5. Primary Endpoint
| Endpoint | Registry definition / time frame | Analysis |
|---|---|---|
| Cumulative Rate of Culture-Confirmed TB Disease in Participants ≥18 Years of Age AND Culture-Confirmed or Probable (Clinical) TB Disease in Participants Less Than [<]18 Years of Age at 33 Months After Enrollment | Baseline up to Month 33 | Modified intention-to-treat population; difference in cumulative tuberculosis disease rate |
The registry defines cumulative tuberculosis disease using culture-confirmed tuberculosis disease for participants aged 18 years or older and culture-confirmed or probable clinical tuberculosis disease for participants younger than 18 years. The registry definition describes probable clinical disease using objective clinical evidence such as cough, fever, night sweats, weight loss, or hemoptysis together with the specified clinical assessment.
6. Secondary Endpoints
| Endpoint | Time frame | Population / method |
|---|---|---|
| Cumulative Rate of Culture-Confirmed TB Disease in Participants ≥18 Years of Age AND Culture-Confirmed or Probable (Clinical) TB Disease in Participants <18 Years of Age at 24 Months Following Completion of Study Therapy | Baseline up to Month 27 (3RPT/INH) or Month 33 (9INH) | MITT; difference in cumulative TB rate; non-inferiority or equivalence |
| Cumulative Rate of Culture-Confirmed or Probable (Clinical) TB Disease (Regardless of Age) At 33 Months After Enrollment | Baseline up to 33 Months | Difference in cumulative TB disease rate; non-inferiority or equivalence |
| Cumulative Rate of Culture-Confirmed TB Disease in Participants ≥18 Years of Age AND Culture Confirmed or Probable (Clinical) TB Disease Among Participants <18 Years of Age Who Completed Study Phase Therapy Within 33 Months of Enrollment | Baseline up to Month 33 | Per-protocol population; difference in cumulative TB disease rate; non-inferiority or equivalence |
| Percentage of Participants With Drug Discontinuation Due to Adverse Drug Reactions Associated With 3RPT/INH or 9INH | Baseline up to 60 days after the last dose of study drug | Safety population; chi-squared test; superiority hypothesis |
| Percentage of Patients With Grade 3 or 4 Drug Toxicities Associated With 3RPT/INH or 9INH | Baseline up to 60 days after the last dose of study drug | Safety population; chi-squared test; superiority hypothesis |
| Percentage of Participants With Death Due to Any Cause | Baseline up to Month 35 | Safety population; chi-squared test; superiority hypothesis |
| Percentage of Participants With Drug Discontinuation for Any Reason Associated With 3RPT/INH or 9INH | Baseline up to Month 3 (3RPT/INH) or Month 9 (9INH) | MITT; chi-squared test; superiority hypothesis |
| Percentage of Participants Who Completed the Treatment Regimen | Baseline up to Month 3 (3RPT/INH) or Month 9 (9INH) | MITT; chi-squared test; superiority hypothesis |
7. Primary Result: Cumulative Tuberculosis Disease Rate
Difference in cumulative TB disease rate
95% CI: upper bound 0.01 · one-sided registry classification
3RPT/INH minus 9INH · Modified intention-to-treat population
The primary effect measure is explicitly defined in the registry as the cumulative tuberculosis disease rate in the 3RPT/INH arm minus the rate in the 9INH arm. Therefore, the negative estimate of -0.24 indicates that the estimated cumulative disease rate was lower in the 3RPT/INH group than in the 9INH group under this analysis.
What the estimate means: the reported difference is directional: negative values favor a lower cumulative tuberculosis disease rate in the 3RPT/INH arm because the registry defines the contrast as 3RPT/INH minus 9INH.
What it does not mean: the value -0.24 is not a hazard ratio, relative risk, odds ratio, or probability that an individual participant will remain free of tuberculosis. It is a difference in cumulative disease rates as defined by the registry.
What the confidence interval says: the registry reports a 95% confidence interval with an upper bound of 0.01. The registry-reported analysis description states that the non-inferiority test statistic was based on the difference between nonparametric Kaplan-Meier estimators, with asymptotic variance obtained using Greenwood's formula.
Why the p-value is different from effect size: no p-value is reported for this primary analysis in the ClinicalTrials.gov record. A non-inferiority decision is instead tied to the confidence-bound relationship to the prespecified non-inferiority margin.
Important caution: the registry analysis text describes a two-sided 95% confidence interval for the difference, while the statistical-analysis record labels the CI as one-sided. This distinction should be retained rather than silently harmonized. The ClinicalTrials.gov record is sufficient to report the upper bound, but not to recover the omitted non-inferiority margin.
Why a difference in cumulative rates can be negative
The sign is determined entirely by the direction of the comparison. Because the registry defines the difference as 3RPT/INH minus 9INH, a negative estimate means the cumulative rate in the 3RPT/INH arm is estimated to be lower than the corresponding rate in the 9INH arm. Reversing the subtraction order would reverse the sign without changing the underlying comparison.
The reported estimate is -0.24. The registry's analysis notes explicitly define the direction of this difference.
8. How the Primary Non-Inferiority Analysis Works
The registry-reported analysis description states that the non-inferiority test statistic was based on the difference of the nonparametric Kaplan-Meier estimators. Its asymptotic variance was obtained through Greenwood's formula, after which a confidence interval for the difference was constructed.
This is an important methodological point. Although the registry's normalized endpoint type is listed as binary, the primary statistical analysis uses cumulative disease rates derived from a time-to-event framework. That approach allows participants who are censored before Month 33 to contribute information up to their available follow-up rather than requiring every participant to have complete observation through the endpoint time.
Kaplan-Meier estimator
Estimates the event-free survival function while accounting for right censoring. The cumulative event probability can then be derived from the estimated survival function.
Greenwood's formula
Provides an estimated variance for the Kaplan-Meier survival estimate, which can be used to quantify uncertainty in the estimated treatment-group difference.
Non-inferiority logic
The question is not simply whether the observed difference is statistically different from zero. The treatment must remain within a prespecified clinically acceptable loss relative to the comparator.
Missing margin in the ClinicalTrials.gov record
9. Secondary Efficacy Results
TB disease through the post-treatment follow-up framework
24 months following completion of study therapy
95% CI: upper bound 0.04 · one-sided registry classification
Baseline up to Month 27 for 3RPT/INH or Month 33 for 9INH
This secondary endpoint was analyzed in the MITT population. The registry again defines the effect as the cumulative TB rate in the 3RPT/INH arm minus the cumulative rate in the 9INH arm.
The estimate of -0.21 has the same directional interpretation as the primary endpoint: the estimated cumulative TB disease rate was lower in the 3RPT/INH group under the reported contrast. The upper confidence bound was 0.04. No p-value is reported for this non-inferiority analysis in the ClinicalTrials.gov record, and the numerical non-inferiority margin is not provided.
TB disease regardless of age through Month 33
Cumulative TB disease rate difference
95% CI: upper bound 0.03 · one-sided registry classification
Baseline up to 33 Months
This endpoint broadened the disease definition to culture-confirmed or probable clinical tuberculosis disease regardless of age. The effect measure again was the difference between the 3RPT/INH and 9INH cumulative rates.
The estimate of -0.25 indicates a lower estimated cumulative TB disease rate in the 3RPT/INH arm under the registry's direction of subtraction. The reported upper confidence bound was 0.03. As with the primary endpoint, the absence of the numerical non-inferiority margin in the ClinicalTrials.gov record prevents a complete reconstruction of the formal margin-based decision rule.
Per-protocol tuberculosis disease analysis
Cumulative TB disease rate difference
95% CI: upper bound 0.06 · one-sided registry classification
Per-protocol population · Baseline up to Month 33
The per-protocol population consisted of eligible enrolled participants who completed study drug within the targeted treatment period. This provides a complementary perspective to the MITT analysis, particularly in a non-inferiority setting where treatment adherence and protocol deviations can affect interpretation.
The estimate of -0.19 again favors a lower cumulative TB disease rate in the 3RPT/INH arm under the stated direction of the difference. The upper confidence bound was 0.06. The per-protocol analysis should not be treated as interchangeable with the MITT analysis: it conditions on treatment completion and therefore represents a different analysis population.
10. Safety and Treatment-Exposure Results
Drug discontinuation due to adverse drug reactions
The registry reports a superiority hypothesis for the percentage of participants who discontinued study drug because of adverse drug reactions. The comparison used a chi-squared test in the safety population.
Chi-squared comparison
95% CI reported in the registry analysis record · Safety population
Baseline through 60 days after the last dose of study drug
The relevant treatment windows were Month 5 for 3RPT/INH and Month 11 for 9INH. The ClinicalTrials.gov record does not include the arm-specific percentages, so the result is reported as the registry-posted statistical comparison rather than as reconstructed event rates.
The P = 0.02 indicates evidence against the null hypothesis used for this superiority comparison under the registry's reported analysis. It does not quantify the magnitude of the difference, and it does not tell us how clinically important that difference is. Without the arm-specific percentages, the direction and absolute size of the discontinuation difference cannot be reconstructed from the ClinicalTrials.gov record alone.
Grade 3 or 4 drug toxicities
| Analysis | Result | Population |
|---|---|---|
| Grade 3 drug toxicity | P = 0.24 | Safety population |
| Grade 4 drug toxicity | P = 0.59 | Safety population |
Both comparisons used chi-squared tests and were specified as superiority hypotheses. The registry provides a 95% CI field for these analyses, but the ClinicalTrials.gov record does not provide numerical confidence-limit values.
The two p-values answer whether the observed categorical distributions provide evidence against their respective null hypotheses under the posted chi-squared analyses. They do not establish that the treatment groups have identical toxicity profiles, nor do they measure the size of any difference. Precision and clinical relevance require the underlying arm-specific event percentages and confidence intervals, which are not included in the ClinicalTrials.gov record.
Death due to any cause
Superiority comparison
Baseline up to Month 35 · Safety population · Chi-squared test
This registry-posted comparison concerns the percentage of participants with death due to any cause through Month 35. The ClinicalTrials.gov record does not contain the arm-specific percentages or event counts for this endpoint.
Discontinuation for any reason
Treatment discontinuation comparison
Baseline through Month 3 for 3RPT/INH or Month 9 for 9INH · MITT population
This endpoint was analyzed with a chi-squared test under a superiority hypothesis. The registry reports a highly small p-value, but the ClinicalTrials.gov record does not include the arm-specific discontinuation percentages or the corresponding confidence limits.
Treatment completion
Completion of treatment regimen
Baseline through Month 3 for 3RPT/INH or Month 9 for 9INH · MITT population
The treatment-completion endpoint used a chi-squared test under a superiority hypothesis. Again, the p-value is evidence about the statistical comparison, not a direct measure of the magnitude of the difference in completion rates.
11. Serious Adverse Events
The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants divided by participants at risk.
| Arm | Serious adverse events | Affected / at risk |
|---|---|---|
| 9INH | Serious adverse events | 102 / 3759 |
| 3RPT/INH | Serious adverse events | 60 / 4040 |
These counts should be kept separate from the chi-squared analyses of drug discontinuation, grade 3 toxicity, grade 4 toxicity, and death because those endpoints have their own registry definitions and analysis records. The ClinicalTrials.gov record does not provide a p-value or confidence interval specifically for the serious-adverse-event counts shown here.
12. Statistical Methodology
Kaplan-Meier estimation
The primary non-inferiority analysis was based on differences between nonparametric Kaplan-Meier estimators. This is appropriate for a time-to-event endpoint because participants may have different lengths of observed follow-up and some may be censored before the end of the study window.
Here di represents events at an event time and ni represents participants at risk immediately before that time. The corresponding cumulative event probability is related to the complement of the survival estimate.
Greenwood's variance estimate
The primary analysis description states that the asymptotic variance of the non-inferiority test statistic was obtained through Greenwood's formula. In practical terms, this provides a way to quantify uncertainty around Kaplan-Meier-based estimates and their comparison.
This formula is shown as an educational representation of the variance calculation referenced by the registry analysis. The trial-specific implementation should be understood from the posted statistical-analysis description rather than reconstructed beyond the ClinicalTrials.gov record.
Chi-squared tests
The registry explicitly reports chi-squared testing for several binary secondary outcomes, including drug discontinuation due to adverse drug reactions, grade 3 or 4 drug toxicities, death due to any cause, discontinuation for any reason, and treatment completion.
A chi-squared test compares observed categorical counts with what would be expected under a null hypothesis of no association between treatment group and the categorical outcome. Its p-value does not itself describe the magnitude of the treatment difference.
Intention-to-treat analysis
The primary endpoint was analyzed in a modified intention-to-treat population. The registry description defines this population as enrolled participants who were eligible after the specified eligibility exclusions. The MITT framework retains the randomized treatment comparison while excluding participants whom the registry identifies as ineligible.
Per-protocol analysis
A secondary tuberculosis disease endpoint used a per-protocol population. In this trial, that population required completion of study drug within the targeted treatment period. Per-protocol analysis can provide a complementary view of treatment efficacy among participants who adhered sufficiently to the protocol, but it is analytically different from the randomized MITT comparison.
Non-inferiority analysis
Non-inferiority trials ask whether the new or shorter regimen performs no worse than the comparator by more than a prespecified clinically acceptable margin. The direction of the confidence bound is therefore central. A p-value for superiority against zero is not the defining criterion for non-inferiority.
13. Statistical Methods Explained
Why was a Kaplan-Meier framework used for the primary endpoint?
The primary endpoint is defined over a follow-up period extending to Month 33, so participants can contribute information for different lengths of time. Kaplan-Meier estimation accounts for censoring and uses the timing of observed events rather than requiring a simple complete-case proportion at the endpoint.
Why does the primary analysis use a difference rather than a hazard ratio?
The registry reports the effect measure as the difference in cumulative TB disease rate. The analysis description specifically says that the non-inferiority test statistic was based on differences between Kaplan-Meier estimators. Therefore, a hazard ratio should not be substituted for the reported effect measure.
Why is the sign of -0.24 important?
The sign follows from the prespecified direction of the subtraction: 3RPT/INH minus 9INH. A negative value therefore represents a lower estimated cumulative disease rate in the 3RPT/INH arm. If the comparison were written in the opposite direction, the same underlying data would produce a positive value.
Why does non-inferiority depend on a margin?
A non-inferiority trial is designed around the largest clinically acceptable loss that would still make the new regimen acceptable for the question being studied. Consequently, the confidence interval is interpreted against that margin rather than merely against zero. The numerical margin is not included in the ClinicalTrials.gov record, so the exact decision threshold cannot be reconstructed here.
Why are both MITT and per-protocol analyses informative?
They address somewhat different threats to interpretation. The MITT analysis preserves the randomized comparison within the eligible population, while the per-protocol analysis focuses on participants who completed therapy within the specified treatment window. Agreement between these approaches can be informative in non-inferiority settings, although the ClinicalTrials.gov record does not provide a single combined decision rule beyond the individual posted analyses.
Why are chi-squared p-values not enough to understand safety?
A p-value indicates how compatible the observed categorical comparison is with its null hypothesis. It does not report the absolute rates, the size of the difference, or its clinical importance. For safety outcomes, the event rates, confidence intervals, severity definitions, treatment exposure, and denominator should all be considered alongside the p-value.
14. Non-Inferiority Logic and the Confidence Interval
The primary analysis is a useful example of why non-inferiority interpretation differs from a conventional superiority test. The registry reports an estimated difference of -0.24 with an upper confidence bound of 0.01. The analysis description says that the non-inferiority test statistic was based on Kaplan-Meier estimators and that the decision involved the upper confidence bound relative to a prespecified margin.
Superiority question
Is there evidence that the two treatment groups differ, typically using a null hypothesis centered on zero for an appropriate effect measure?
Non-inferiority question
Is the new regimen's effect sufficiently close to the comparator that the loss does not exceed a prespecified acceptable margin?
Confidence-bound focus
The unfavorable side of the confidence interval is particularly important because it represents the greatest plausible loss consistent with the statistical uncertainty.
Why the margin matters
Without the numerical margin, one cannot independently reproduce the formal non-inferiority decision from the estimate and confidence bound alone.
The primary estimate being negative does not, by itself, prove non-inferiority. Likewise, a confidence interval that includes zero does not automatically rule out non-inferiority. The correct decision depends on where the relevant confidence bound lies relative to the prespecified non-inferiority margin.
15. Results Summary
| Endpoint | Population | Effect / result |
|---|---|---|
| Primary cumulative TB disease rate at Month 33 | MITT | Difference -0.24; 95% CI upper bound 0.01 |
| TB disease at 24 months following completion | MITT | Difference -0.21; 95% CI upper bound 0.04 |
| TB disease regardless of age at Month 33 | Not specified in the ClinicalTrials.gov record | Difference -0.25; 95% CI upper bound 0.03 |
| TB disease in per-protocol population | Per protocol | Difference -0.19; 95% CI upper bound 0.06 |
| Drug discontinuation due to adverse drug reactions | Safety | P = 0.02; chi-squared |
| Grade 3 drug toxicity | Safety | P = 0.24; chi-squared |
| Grade 4 drug toxicity | Safety | P = 0.59; chi-squared |
| Death due to any cause | Safety | P = 0.22; chi-squared |
| Discontinuation for any reason | MITT | P < 0.001; chi-squared |
| Treatment completion | MITT | P < 0.0001; chi-squared |
The table intentionally separates non-inferiority efficacy estimates from superiority-oriented categorical safety and treatment-exposure analyses. These are different statistical questions and should not be collapsed into a single overall test.
16. Crossover, Factorial Design, Interim Analysis, and Bayesian Methods
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Crossover | No crossover is reported in the ClinicalTrials.gov record. |
| Factorial design | No factorial design is reported. The registry describes a two-arm parallel-group study. |
| Interim analysis | No interim analysis is reported in the ClinicalTrials.gov record. |
| Bayesian methods | No Bayesian analysis is reported in the ClinicalTrials.gov record. |
| Stratification | No stratification variables are provided in the ClinicalTrials.gov record. |
| Missing-data imputation | No missing-data imputation method is reported in the ClinicalTrials.gov record. |
| Multiplicity | The ClinicalTrials.gov record identifies multiple posted analyses and distinguish non-inferiority/equivalence from superiority hypotheses, but does not provide a multiplicity-adjustment procedure. |
17. Understanding the Secondary Chi-Squared Analyses
Several secondary outcomes were binary categorical endpoints. The registry explicitly identifies the chi-squared test as the analysis method. Conceptually, the test compares the observed outcome distribution between the two treatment groups.
For example, participants can be classified according to whether they discontinued treatment or completed treatment. The chi-squared statistic evaluates the compatibility of the observed counts with a null hypothesis of no association between treatment assignment and the categorical outcome.
For this trial, the p-values should therefore be read as evidence about the corresponding categorical comparison, not as direct estimates of the treatment effect. The size and direction of an effect require the underlying event rates or an explicitly reported effect measure.
18. Interpreting the Primary Estimate Without Overinterpreting It
The primary estimate of -0.24 means the reported difference in cumulative tuberculosis disease rate, defined as 3RPT/INH minus 9INH, was negative. That is evidence about the direction of the estimated difference, not a statement that every participant experienced a benefit.
The reported 95% confidence interval has an upper bound of 0.01. A confidence interval quantifies uncertainty around the estimated treatment comparison under the specified statistical framework; it does not describe the distribution of individual participant outcomes.
No p-value is posted on ClinicalTrials.gov for the primary non-inferiority comparison. This is important because non-inferiority is not properly summarized by asking only whether the treatment difference differs from zero. The prespecified non-inferiority margin is the relevant benchmark.
The primary analysis uses the modified intention-to-treat population. The per-protocol result is complementary but should not be substituted for the primary analysis population.
19. Limitations
- Incomplete arm-specific secondary results: several chi-squared analyses provide p-values but not the corresponding arm-level percentages in the ClinicalTrials.gov record. Those p-values cannot be converted into treatment-effect sizes without additional information.
- CI reporting discrepancy: the analysis record labels the confidence interval as one-sided, while the analysis description says that a two-sided 95% confidence interval was constructed. The discrepancy is reported rather than silently resolved.
- Different analysis populations: MITT, per-protocol, and safety analyses address different questions and should not be pooled into one estimate.
- Time-frame differences: the secondary endpoints use different follow-up definitions, including Month 27 for 3RPT/INH versus Month 33 for 9INH for one endpoint. These definitions should be preserved when interpreting the results.
- Endpoint-definition complexity: the primary endpoint combines culture-confirmed TB disease for participants aged 18 years or older with culture-confirmed or probable clinical TB disease for participants younger than 18 years.
- Limited methodological detail in the ClinicalTrials.gov record: the data do not provide a full statistical analysis plan, including stratification variables, missing-data procedures, multiplicity adjustment, or interim-monitoring procedures.
- Safety denominator differences: safety outcomes use a safety population rather than the same analysis population used for the primary efficacy endpoint.
20. Why This Trial Matters Statistically
TBTC Study 26 is a useful teaching example because the statistical question is not simply whether two treatment groups have different observed event rates. The trial illustrates how a randomized prevention study can combine time-to-event estimation, non-inferiority logic, modified intention-to-treat analysis, per-protocol analysis, and categorical safety comparisons.
| Concept | How it appears in TBTC Study 26 |
|---|---|
| Randomization | Participants were randomized to two parallel treatment arms. |
| Modified intention-to-treat analysis | The primary endpoint was analyzed in the MITT population. |
| Per-protocol analysis | A secondary tuberculosis endpoint was evaluated among participants meeting the specified treatment-completion criteria. |
| Kaplan-Meier estimation | The primary non-inferiority test statistic was based on differences between nonparametric Kaplan-Meier estimators. |
| Greenwood's formula | Used to obtain the asymptotic variance of the primary test statistic. |
| Non-inferiority design | The primary and several secondary TB disease analyses were identified as non-inferiority or equivalence hypotheses. |
| Confidence intervals | The primary and secondary disease-rate comparisons report 95% confidence information and upper bounds. |
| Chi-squared testing | Used for multiple binary safety, discontinuation, death, and treatment-completion outcomes. |
| Analysis populations | MITT, per-protocol, and safety populations were used for different questions. |
| Endpoint timing | TB disease was assessed at specified follow-up windows, including Month 33 and time frames following completion of therapy. |
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: NCT00023452 — TBTC Study 26.
- Linked PubMed record: PMID 22150035.
- Linked PubMed record: PMID 38640958.
- Linked PubMed record: PMID 31690107.
- Linked PubMed record: PMID 29393655.
- Linked PubMed record: PMID 28087928.
Continue through the Clinical Biostats statistical pathway
Explore the trial-design and statistical concepts that connect directly to the analysis of randomized prevention trials.
24. Record Summary
TBTC Study 26 provides a clear example of a randomized phase 3 non-inferiority question in tuberculosis prevention. The primary comparison used the difference in cumulative tuberculosis disease rates between three months of once-weekly rifapentine plus isoniazid and nine months of daily isoniazid, with the primary analysis conducted in a modified intention-to-treat population and the test statistic based on nonparametric Kaplan-Meier estimators.
The reported primary difference was -0.24, with a 95% confidence interval upper bound of 0.01. Secondary tuberculosis analyses reported differences of -0.21, -0.25, and -0.19 under their respective analysis populations and time frames. Several categorical secondary outcomes were evaluated using chi-squared tests, including drug discontinuation due to adverse drug reactions, grade 3 or 4 toxicities, death due to any cause, discontinuation for any reason, and treatment completion.
The most important statistical lesson is that the primary non-inferiority conclusion depends on the relationship between the confidence bound and the prespecified non-inferiority margin. Because the numerical margin is not contained in the ClinicalTrials.gov record, the page does not infer it. That limitation is preferable to presenting an apparently precise non-inferiority decision that cannot be independently reconstructed from the provided record.