This page separates reported trial results from statistical interpretation. The numerical results on this page are restricted to the ClinicalTrials.gov record in the trial data above.
1. Trial at a Glance
GLISTEN was a completed phase 3, randomized, double-blind, parallel-design clinical trial evaluating linerixibat versus placebo in a trial concerning pruritus associated with primary biliary cholangitis. The registry reports 238 enrolled participants, six arms, one registered primary endpoint, 11 posted outcome measures, and 14 posted statistical analyses.
| Feature | GLISTEN |
|---|---|
| Trial name | Global Linerixibat Itch Study of Efficacy and Safety in Primary Biliary Cholangitis (PBC) (GLISTEN) |
| Phase | Phase 3 |
| Status | Completed |
| Condition | Pruritus |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 238 |
| Interventions | Linerixibat; placebo |
| Primary endpoint type | Binary in the registered endpoint metadata; continuous in the posted formal statistical analysis of mean change |
| Hypothesis type | Superiority |
| Lead sponsor | GlaxoSmithKline |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT04950127 |
2. Clinical Question
The central statistical question was whether linerixibat produced a greater improvement in itch scores than placebo over the prespecified assessment period in Part A of GLISTEN.
Population
The registry identifies pruritus as the condition and the trial title specifies primary biliary cholangitis (PBC).
Intervention
Linerixibat, with the primary Part A comparison specifically reported as linerixibat 40 mg versus placebo.
Comparator
Placebo. The primary statistical comparison is Part A linerixibat 40 mg versus Part A placebo.
Primary question
Does linerixibat produce a greater mean change from baseline in monthly itch scores over 24 weeks than placebo?
3. Trial Design
Linerixibat 40 mg
- Primary analyses compare Part A linerixibat 40 mg with Part A placebo.
- The primary efficacy analysis used the ITT set.
- Participants were classified according to randomized treatment assignment.
Placebo
- Placebo was the comparator for the principal Part A statistical analyses.
- The ITT analysis retained randomized treatment assignment.
- Additional Part B arms are represented in the registry safety data.
The registry also reports Part B safety groups, indicating that the overall six-arm design extended beyond the two groups used for the primary Part A efficacy comparison. The ClinicalTrials.gov record does not provide a complete description of every arm's intervention schedule, so this page does not infer those details.
4. Primary Endpoint
| Endpoint | Registered definition / time frame | Formal analysis |
|---|---|---|
| Part A: Mean Change From Baseline in Monthly Itch Scores Over 24 Weeks Using Numerical Rating Scale (NRS) | Baseline and up to Week 24. Itch scores were assessed using a NRS twice daily, ranging from 0 to 10, where 0 represents no itching and 10 the worst imaginable itching. The worst daily itch score was the worst of the two scores recorded daily; the weekly itch score was the average of the worst daily itch scores in one week; the monthly itch score was the worst weekly itch score for the month. | Mixed Models Analysis / mixed-effects model |
The registry metadata labels the registered endpoint as binary, but the posted formal statistical analysis treats the endpoint as continuous mean change and reports a mean difference with a 95% confidence interval. The formal analysis is therefore the appropriate basis for interpreting the reported primary result.
5. How the Itch Score Was Constructed
The primary endpoint is not simply a single daily NRS measurement. The registry describes a hierarchy of aggregation that moves from twice-daily observations to daily, weekly, and finally monthly scores.
This construction has an important statistical consequence: the primary endpoint is designed to summarize a patient's recurring itch experience rather than treating one isolated daily rating as the definitive measurement. Because the monthly score is itself derived from weekly summaries, the statistical analysis operates on a structured longitudinal outcome.
6. Primary Result
Mean Change From Baseline in Monthly Itch Scores
Net mean difference
95% CI: -1.15 to -0.28 · P = 0.001
Part A: linerixibat 40 mg versus placebo; baseline and up to Week 24.
The analysis was performed on the Intent to Treat (ITT) set, which included all randomized participants. Participants were classified according to the treatment to which they were randomized.
| Primary endpoint | Comparison | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|---|
| Mean change from baseline in monthly itch scores over 24 weeks | Linerixibat 40 mg vs placebo | Mixed-effects model | -0.72 | -1.15 to -0.28 | 0.001 |
The estimated net mean difference was -0.72. Because lower values on the NRS represent less itching, this means the modeled change in monthly itch score was 0.72 points more favorable for linerixibat than for placebo, expressed as the difference between the two randomized groups' estimated mean changes.
The estimate is a between-group mean difference; it does not mean that every participant experienced a 0.72-point reduction, nor does it mean that each individual patient had the same treatment effect. Individual responses can vary substantially around a group-level estimate.
The 95% confidence interval of -1.15 to -0.28 describes uncertainty around the estimated treatment difference under the analysis model and sampling framework. It is not a prediction interval for individual patients and does not tell us where most individual treatment effects must fall.
The P-value of 0.001 measures the compatibility of the observed result with the statistical null hypothesis under the specified analysis framework. It does not measure the magnitude, clinical importance, or probability that linerixibat is effective. The magnitude is conveyed by the estimated mean difference and its confidence interval.
Because the analysis is longitudinal, the interpretation also depends on the mixed-effects model assumptions and on how repeated observations, baseline adjustment, concomitant itch medication, and missing observations are handled. The registry identifies these model components, but the ClinicalTrials.gov record does not provide the complete statistical analysis plan.
7. Secondary Efficacy Results
The registry reports 13 secondary statistical analyses in addition to the primary analysis. These cover an early itch-score assessment, monthly itch responder thresholds, sleep, PBC-40 domains, and patient global impression measures.
| Secondary endpoint | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Mean change from baseline in weekly itch score at Week 2 | Mixed-effects model | -0.71 mean difference | -1.07 to -0.34 | <0.001 |
| Mean change from baseline in monthly sleep score over 24 weeks | Mixed-effects model | -0.53 mean difference | -0.98 to -0.07 | 0.024 |
| Responders with ≥2-point reduction in monthly itch score at Week 24 | Cochran-Mantel-Haenszel | 4.0 risk difference | -9.0 to 17.0 | 0.539 |
| Responders with ≥3-point reduction in monthly itch score at Week 24 | Cochran-Mantel-Haenszel | 13.0 risk difference | 0.0 to 27.0 | 0.043 |
| Responders with ≥4-point reduction in monthly itch score at Week 24 | Cochran-Mantel-Haenszel | 12.0 risk difference | -0.0 to 24.0 | 0.058 |
| PBC-40 Cognitive domain at Week 24 | Mixed-effects model | 0.76 mean difference | -0.34 to 1.86 | 0.176 |
| PBC-40 Emotional domain at Week 24 | Mixed-effects model | 0.27 mean difference | -0.36 to 0.89 | 0.403 |
| PBC-40 Fatigue domain at Week 24 | Mixed-effects model | 1.59 mean difference | -0.48 to 3.67 | 0.132 |
| PBC-40 Itch domain at Week 24 | Mixed-effects model | -0.58 mean difference | -1.34 to 0.18 | 0.132 |
| PBC-40 Social domain at Week 24 | Mixed-effects model | -0.15 mean difference | -1.57 to 1.27 | 0.836 |
| PBC-40 Symptoms domain at Week 24 | Mixed-effects model | 0.45 mean difference | -0.44 to 1.34 | 0.318 |
| Mean change from baseline in PGI-S over 24 weeks | Mixed-effects model | -0.37 mean difference | -0.55 to -0.20 | <0.001 |
| PGI-C scores over 24 weeks | Mixed-effects model | -0.49 mean difference | -0.76 to -0.21 | <0.001 |
8. Early Itch Response at Week 2
Mean difference at Week 2
95% CI: -1.07 to -0.34 · P < 0.001
Mean change from baseline in weekly itch score; Part A linerixibat 40 mg versus placebo.
The Week 2 endpoint uses the same general NRS framework but evaluates mean change in weekly itch score at an early time point. The registry reports a mixed-model analysis using treatment group, week, the week-by-treatment interaction, baseline weekly itch score, the visit-by-baseline interaction, and baseline concomitant itch medication.
The negative estimate indicates a more negative modeled change in the linerixibat group than in the placebo group. Since the NRS assigns larger values to worse itching, the direction is consistent with greater reduction in itch score for linerixibat at Week 2.
The confidence interval remains entirely below zero, from -1.07 to -0.34. This expresses uncertainty around the between-group mean difference rather than uncertainty about whether any individual participant improved.
The Week 2 result is also statistically distinct from the primary 24-week endpoint. An early time point can provide information about the timing of an effect, but it does not replace the prespecified primary endpoint.
9. Responder Analyses
Three secondary endpoints classified participants according to whether they achieved progressively larger reductions from baseline in monthly itch score at Week 24. The registry analyzed these binary outcomes using a Cochran-Mantel-Haenszel approach.
| Responder definition | Risk difference | 95% CI | P-value |
|---|---|---|---|
| ≥2-point reduction | 4.0 | -9.0 to 17.0 | 0.539 |
| ≥3-point reduction | 13.0 | 0.0 to 27.0 | 0.043 |
| ≥4-point reduction | 12.0 | -0.0 to 24.0 | 0.058 |
These are risk differences, expressed as percentage-point differences between the responder proportions after the specified stratified analysis. A value of 13.0 means the estimated responder proportion differed by 13.0 percentage points between the treatment groups under the reported analysis; it does not mean that 13.0% of every individual patient's itch score was reduced.
The continuous primary endpoint asks about the average difference in change. A responder endpoint asks a different question: how the treatment groups compare after imposing a clinically defined threshold for individual response.
The three thresholds produce different estimates and confidence intervals. The ≥2-point analysis has a risk difference of 4.0 with a 95% CI of -9.0 to 17.0. The ≥3-point analysis has a risk difference of 13.0 with a 95% CI of 0.0 to 27.0. The ≥4-point analysis has a risk difference of 12.0 with a 95% CI of -0.0 to 24.0.
The P-values of 0.539, 0.043, and 0.058 describe the separate reported tests. They do not quantify the size of the responder difference and should not be used as a substitute for the risk difference and its confidence interval.
10. Stratified Cochran-Mantel-Haenszel Analysis
The responder analyses used a Cochran-Mantel-Haenszel (CMH) stratified analysis. The registry text identifies two baseline factors for adjustment:
Baseline itch severity
Moderate: ≥4 and <7; severe: ≥7.
Concomitant cholestatic pruritus treatment
Regimen contains bile acid binding resin (BABR); regimen does not contain BABR; or no defined treatment.
The purpose of the stratified analysis is to compare treatment groups while accounting for the prespecified baseline strata. Instead of collapsing all participants into one unadjusted 2-by-2 table, the CMH framework evaluates the treatment association across the defined strata and produces a common adjusted comparison under the model.
The reported analysis also used multiple imputation of missing monthly itch scores before deriving responder definitions. The ClinicalTrials.gov record indicates that the imputed datasets were analyzed, but do not provide the complete imputation model or all implementation details.
The reported CMH estimate is presented as a percentage difference and normalized here as a risk difference. Positive values indicate a higher estimated responder proportion in the linerixibat group under the reported comparison.
11. Patient-Reported Sleep and Global Impression Outcomes
Monthly Sleep Score
Net mean difference
95% CI: -0.98 to -0.07 · P = 0.024
Mean change from baseline in monthly sleep score over 24 weeks.
The sleep endpoint was analyzed using the same broad longitudinal mixed-model framework, including treatment group, visit, their interaction, baseline monthly sleep score, its interaction with visit, and baseline concomitant itch medication.
Patient's Global Impression of Severity
Net mean difference
95% CI: -0.55 to -0.20 · P < 0.001
Mean change from baseline in PGI-S over 24 weeks.
Patient's Global Impression of Change
Net mean difference
95% CI: -0.76 to -0.21 · P < 0.001
PGI-C scores over 24 weeks, with assessment beginning at Week 4.
The itch NRS, sleep score, PGI-S, and PGI-C are related but are not interchangeable endpoints. The itch score directly quantifies reported itching using the specified NRS construction, whereas the global impression measures capture broader patient-reported perceptions.
Consistent directional findings across related outcomes can provide contextual support for the primary endpoint, but each estimate still answers its own statistical question. The presence of several favorable nominal P-values does not convert the secondary endpoints into a single composite result.
12. PBC-40 Domain Results
The registry reports six PBC-40 domain analyses at Week 24. All used mixed-effects models with treatment group, visit, visit-by-treatment interaction, baseline PBC-40 domain score, its interaction with visit, and baseline concomitant itch medication.
| PBC-40 domain | Mean difference | 95% CI | P-value |
|---|---|---|---|
| Cognitive | 0.76 | -0.34 to 1.86 | 0.176 |
| Emotional | 0.27 | -0.36 to 0.89 | 0.403 |
| Fatigue | 1.59 | -0.48 to 3.67 | 0.132 |
| Itch | -0.58 | -1.34 to 0.18 | 0.132 |
| Social | -0.15 | -1.57 to 1.27 | 0.836 |
| Symptoms | 0.45 | -0.44 to 1.34 | 0.318 |
All six confidence intervals include zero. Consequently, the ClinicalTrials.gov record does not establish a statistically distinguishable between-group mean difference for any of these individual PBC-40 domain analyses at the reported nominal significance level.
13. Statistical Methodology
Mixed-effects model for repeated measurements
The principal statistical method was a mixed-model repeated-measures analysis. For the primary endpoint, the model included treatment group, visit, the visit-by-treatment interaction, baseline monthly itch score, the visit-by-baseline monthly itch score interaction, and baseline concomitant itch medication.
This representation is conceptual rather than a reconstruction of the exact statistical software specification. Its purpose is to show why the analysis is more informative than comparing two isolated means at Week 24.
A repeated-measures model uses the longitudinal structure of the observations. Instead of discarding earlier measurements, it models outcomes across visits and estimates how the treatment groups differ in their trajectories after accounting for baseline covariates.
Why baseline adjustment was included
The primary model included baseline monthly itch score and its interaction with visit. Baseline adjustment can improve precision when baseline outcome values explain part of the variability in follow-up measurements. The interaction allows the relationship between baseline score and outcome to vary across visits within the model.
The same general principle was applied to the secondary longitudinal outcomes: the Week 2 weekly itch model included baseline weekly itch score, while the sleep and PBC-40 models included their corresponding baseline measures.
Intention-to-treat analysis
The primary efficacy analysis used the ITT set, defined in the registry analysis as including all randomized participants. Participants remained classified according to the treatment to which they were randomized.
This preserves the treatment comparison created by randomization. It also means that the efficacy estimate is not restricted only to participants who completed treatment or adhered perfectly to the assigned intervention.
Cochran-Mantel-Haenszel test
The CMH method was used for the binary responder endpoints. It is particularly useful when a categorical treatment comparison needs to account for stratification factors rather than relying solely on an unadjusted comparison.
In GLISTEN, the registry-reported analysis text identifies baseline itch severity and concomitant cholestatic pruritus treatment regimen as the adjustment factors. The resulting effect measure was reported as a percentage difference and is normalized here as a risk difference.
Multiple imputation and missing data
The responder analyses incorporated multiple imputation of missing monthly itch scores before responder definitions were derived. This is an important methodological detail because converting a continuous outcome into a binary responder endpoint can otherwise make missing observations directly affect who is classified as a responder.
Multiple imputation replaces missing values with draws or predictions under an imputation model and then combines results across completed datasets. The ClinicalTrials.gov record does not specify the complete imputation model, number of imputations, or every diagnostic procedure, so those details are not inferred here.
14. Statistical Methods Explained
Why was a mixed-effects model used?
The primary outcome was measured longitudinally through Week 24 rather than only once. A mixed-effects repeated-measures framework can model the treatment groups across visits and incorporate baseline covariates and treatment-by-visit differences. This makes the analysis a longitudinal comparison rather than a simple single-time-point test.
What does a mean difference of -0.72 mean?
It is the estimated difference between the treatment groups in mean change from baseline in monthly itch score. The negative sign indicates a more favorable direction for linerixibat because the NRS defines 0 as no itching and 10 as the worst imaginable itching. It does not mean every participant improved by exactly 0.72 points.
Why use a Cochran-Mantel-Haenszel test for responders?
The responder outcomes are categorical: participants either meet the specified reduction threshold or they do not. The CMH method provides a stratified comparison while accounting for baseline itch severity and concomitant cholestatic pruritus treatment regimen.
What does a risk difference of 13.0 mean?
For the ≥3-point responder endpoint, the reported risk difference was 13.0. This represents a 13.0-percentage-point difference in responder proportions between the randomized groups under the reported stratified analysis. It is an absolute effect measure, not a relative risk or odds ratio.
Why is multiple imputation important?
Missing monthly itch scores affect whether a participant can be classified using a response threshold. The registry states that missing monthly itch scores were multiply imputed before responder definitions were derived. This means the reported categorical comparison incorporates an explicit strategy for missing outcome information rather than simply ignoring every participant with a missing value.
Why does the P-value not measure effect size?
The P-value quantifies evidence against the statistical null hypothesis under the specified test and model. It is influenced by the magnitude of the estimated effect, variability, and information in the analysis. The estimated difference and confidence interval are therefore necessary to understand how large the observed treatment difference is and how precisely it was estimated.
What does a confidence interval tell us here?
A 95% confidence interval describes statistical uncertainty around the estimated treatment effect under the specified analysis framework. For the primary mean difference, the interval is -1.15 to -0.28. It should not be interpreted as saying that 95% of individual patients will experience treatment effects within that range.
15. Safety Results
The ClinicalTrials.gov record reports serious adverse events by arm for both Part A and Part B. Because the registry data provide affected participants over participants at risk, the figures can be reported directly without calculating additional rates.
| Part / treatment group | Serious adverse events affected / at risk |
|---|---|
| Part A: Placebo | 4/118 |
| Part A: Linerixibat 40 mg | 14/119 |
| Part B: Placebo in Part A and Part B | 0/53 |
| Part B: Placebo in Part A and Linerixiba | 0/52 |
| Part B: Linerixibat 40 mg in Part A and | 2/46 |
| Part B: Linerixibat 40 mg in Part A and | 0/45 |
For the primary Part A comparison, the ClinicalTrials.gov record reports serious adverse events in 4/118 placebo participants and 14/119 linerixibat 40 mg participants. These are descriptive safety counts. The ClinicalTrials.gov record does not provide a formal hypothesis test, confidence interval, adverse-event adjudication details, or complete safety endpoint hierarchy, so none is inferred.
16. Design Features That Affect Statistical Interpretation
Randomization
Randomized allocation provides the basis for comparing outcomes according to assigned treatment in the ITT population.
Double blinding
The registry identifies the trial as double masked, reducing opportunities for knowledge of treatment assignment to influence outcome assessment or participant reporting.
Parallel design
The registered design model is parallel, so the principal treatment comparison is between randomized groups rather than within-person crossover periods.
Superiority hypothesis
The reported hypothesis type is superiority. The primary result therefore concerns whether the treatment groups differ in the prespecified direction rather than whether linerixibat meets a non-inferiority margin.
The ClinicalTrials.gov record does not report a non-inferiority margin, crossover design, factorial structure, Bayesian analysis, or interim-analysis procedure. Those features are therefore not presented as components of GLISTEN.
17. Analysis Populations and Estimands
The primary analysis is explicitly tied to the ITT set. That distinction is important because the estimand represented by the reported comparison is anchored to randomized treatment assignment rather than to treatment actually received.
| Element | Reported information | Statistical implication |
|---|---|---|
| Population | ITT set including all randomized participants | Preserves randomized-group comparison for primary efficacy analysis |
| Treatment classification | According to treatment as randomized | Does not reclassify participants according to subsequent exposure |
| Primary outcome | Mean change in monthly itch score over 24 weeks | Longitudinal continuous endpoint |
| Primary effect measure | Mean Difference (Net) | Absolute difference in modeled mean change |
| Hypothesis | Superiority | Tests whether the randomized treatment groups differ under the prespecified superiority framework |
One advantage of stating the analysis population explicitly is that it prevents an apparently favorable treatment effect from being mistaken for a result limited to participants who completed all scheduled assessments. Conversely, ITT does not eliminate missing-data assumptions: the mixed-model analysis still depends on how incomplete longitudinal information is handled by the model.
18. Confidence Intervals Across the Reported Results
The confidence intervals illustrate how statistical uncertainty varies across endpoints. The primary mean difference has a 95% interval of -1.15 to -0.28, while the secondary responder estimates have considerably different ranges depending on the response threshold.
| Endpoint | Estimate | 95% CI | What the interval contributes |
|---|---|---|---|
| Primary monthly itch score | -0.72 | -1.15 to -0.28 | Quantifies uncertainty around the modeled mean treatment difference |
| Week 2 weekly itch score | -0.71 | -1.07 to -0.34 | Quantifies uncertainty around the early longitudinal difference |
| ≥2-point responder | 4.0 | -9.0 to 17.0 | Shows substantial uncertainty around the absolute responder difference |
| ≥3-point responder | 13.0 | 0.0 to 27.0 | Provides an absolute range around the estimated responder difference |
| ≥4-point responder | 12.0 | -0.0 to 24.0 | Shows uncertainty extending to values close to no difference |
The primary confidence interval does not cross zero: its reported limits are -1.15 and -0.28. That is consistent with the reported P-value of 0.001 under the two-sided analysis.
But statistical compatibility with zero is not the same as clinical importance. The interval gives a range of plausible values for the modeled group difference under the analysis framework; deciding whether a difference of this magnitude is clinically meaningful requires a clinical threshold or other external interpretive standard, which is not reported in the ClinicalTrials.gov record.
19. Understanding the Difference Between Continuous and Responder Endpoints
GLISTEN provides a useful statistical comparison because the same underlying itch experience is represented in two different ways.
Continuous endpoint
The primary analysis estimates the mean difference in change in monthly itch score. It retains the numerical magnitude of each participant's outcome rather than reducing it to a yes/no classification.
Responder endpoint
The secondary analyses classify participants according to whether they reach predefined reductions of ≥2, ≥3, or ≥4 points.
A continuous analysis can be statistically efficient because it retains more information about the outcome. A responder analysis can be easier to communicate because it expresses results in terms of the proportion of patients reaching a threshold. Neither is universally superior; they answer different questions.
The GLISTEN data show why both views can matter. The primary mean difference was -0.72, while the reported risk differences for the three responder definitions were 4.0, 13.0, and 12.0. These quantities cannot be substituted for one another.
20. What the Primary Estimate Does — and Does Not — Mean
The primary estimate of -0.72 is a model-based comparison of mean change from baseline in monthly itch scores between randomized treatment groups. It is not the percentage of participants who improved, the probability of response, or the average percentage reduction in itching.
The 95% CI of -1.15 to -0.28 describes uncertainty around the estimated mean difference. It does not describe the distribution of individual patient responses or guarantee that the true effect for every clinically relevant subgroup lies within that interval.
The P-value of 0.001 provides evidence against the null hypothesis under the specified model and test. It is not the probability that the null hypothesis is true, and it is not a measure of treatment benefit magnitude.
Because the estimate comes from a mixed-effects repeated-measures model, interpretation depends on the model's treatment-by-visit specification, covariance structure, baseline adjustment, and assumptions concerning incomplete longitudinal data. The ClinicalTrials.gov record does not provide enough information to independently verify every model assumption.
21. Important Limitations and Interpretation Issues
- Registry-derived scope: this analysis is restricted to the ClinicalTrials.gov record. It does not import numerical results from external publications or other sources.
- Endpoint metadata inconsistency: the registered primary endpoint is labeled binary in the registry-reported metadata, while its posted formal statistical analysis is continuous and reports a mean difference. The formal analysis is described explicitly rather than silently resolving the discrepancy.
- Multiple secondary analyses: the registry reports numerous secondary endpoints and statistical tests. The ClinicalTrials.gov record does not provide a complete multiplicity strategy, so nominal P-values should not automatically be interpreted as independently confirmatory.
- Missing data: multiple imputation was used for missing monthly itch scores in responder analyses. The ClinicalTrials.gov record does not include the complete imputation model or sensitivity-analysis framework.
- Longitudinal model assumptions: mixed-effects models require assumptions about the relationship between repeated measurements and their covariance structure. The registry summary reported here does not provide enough information to evaluate all of them.
- Responder thresholds: changing the threshold from ≥2 to ≥3 or ≥4 points changes the statistical question and can materially change the estimated treatment difference.
- PBC-40 interpretation: the ClinicalTrials.gov record provides domain-specific numerical estimates but do not provide complete scoring-direction information for each domain. The signs should therefore not be assigned clinical meaning beyond what is explicitly documented.
- Generalizability: the trial is a controlled phase 3 randomized study; the ClinicalTrials.gov record does not provide enough information to characterize every feature of the population needed for broader external-validity judgments.
22. Why This Trial Matters Statistically
GLISTEN is a useful teaching example because it combines a longitudinal continuous primary endpoint with categorical responder analyses and patient-reported secondary outcomes. It also illustrates how baseline adjustment, stratification, ITT analysis, and missing-data methods can all contribute to a modern clinical-trial analysis.
| Concept | How it appears in GLISTEN |
|---|---|
| Randomization | Participants were randomized in a parallel-design phase 3 trial. |
| Blinding | The registry identifies the trial as double masked. |
| ITT analysis | The primary analysis used all randomized participants classified according to randomized treatment. |
| Mixed-effects model | Used for the primary longitudinal itch endpoint and multiple secondary continuous outcomes. |
| Covariate adjustment | Baseline outcome measures and baseline concomitant itch medication were incorporated into longitudinal models. |
| Stratified analysis | CMH analyses adjusted for baseline itch severity and concomitant cholestatic pruritus treatment regimen. |
| Risk difference | Responder endpoints reported percentage differences normalized as risk differences. |
| Multiple imputation | Missing monthly itch scores were imputed before deriving responder definitions. |
| Confidence intervals | 95% intervals accompany the primary and reported secondary effect estimates. |
| P-values | Reported for the formal primary and secondary analyses. |
| Multiple endpoints | One primary endpoint and multiple secondary outcome analyses were posted. |
The statistical lesson is broader than any single P-value. A well-specified treatment comparison requires attention to the endpoint definition, the analysis population, repeated measurements, baseline covariates, missing observations, stratification, effect measures, and uncertainty. GLISTEN contains examples of each.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Statistical Calculators
25. Sources
- ClinicalTrials.gov: NCT04950127 — GLISTEN.
- Linked publication: PubMed record for PMID 37953500.
Continue through Clinical Biostats
Use the related tutorials and statistical calculators to explore the methods underlying randomized clinical-trial analysis, longitudinal models, stratified categorical comparisons, confidence intervals, and effect measures.
26. Record Summary
GLISTEN provides a compact example of how several statistical approaches can be combined within one randomized phase 3 trial. Its primary analysis uses an ITT population and a longitudinal mixed-effects model to estimate the mean difference in monthly itch-score change through Week 24. Secondary analyses extend the framework to an early weekly itch endpoint, sleep, patient global impressions, six PBC-40 domains, and three binary responder definitions analyzed using a stratified Cochran-Mantel-Haenszel method.
The primary reported estimate was a -0.72 net mean difference in monthly itch-score change, with a 95% CI of -1.15 to -0.28 and P = 0.001. The secondary responder analyses illustrate why continuous and categorical effect measures should be interpreted separately: the reported risk differences were 4.0, 13.0, and 12.0 for the ≥2-, ≥3-, and ≥4-point reduction thresholds, respectively.
The most important statistical interpretation is therefore not simply whether an individual P-value crosses a conventional threshold. The full analysis requires understanding what was measured, when it was measured, which population was analyzed, how repeated observations were modeled, how baseline factors were incorporated, how missing values were handled, and whether the effect measure corresponds to the clinical question being asked.