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
REGENERATE was a randomized, quadruple-masked, parallel-group phase 3 study evaluating obeticholic acid versus placebo in participants with nonalcoholic steatohepatitis (NASH). The primary endpoint was a time-to-event composite of adjudicated liver-related clinical outcomes, with the clinical event distribution estimated using Kaplan-Meier methodology.
| Feature | REGENERATE |
|---|---|
| Phase | Phase 3 |
| Condition | Non Alcoholic Steatohepatitis (NASH) |
| Therapeutic area | Gastroenterology & Hepatology |
| Design | Randomized, parallel-group, quadruple-masked |
| Primary purpose | Treatment |
| Enrollment | 2,477 |
| Interventions | Obeticholic acid and placebo |
| Primary endpoint type | Time-to-event |
| ClinicalTrials.gov | NCT02548351 |
| Lead sponsor | Intercept Pharmaceuticals |
| Sponsor type | Industry |
| Status | Terminated |
2. Clinical Question
The central statistical question was whether treatment with obeticholic acid, compared with placebo in conjunction with established local standard of care, changed the time to the first adjudicated event in a composite of clinically important liver-related outcomes in participants with NASH.
Population
Participants with nonalcoholic steatohepatitis (NASH) included in the randomized phase 3 study.
Intervention
Obeticholic acid at 10 mg or 25 mg.
Comparator
Placebo.
Primary question
Does obeticholic acid alter time to the first adjudicated event in the clinical-outcome composite compared with placebo?
3. Trial Design
OCA 10 Milligrams
- Obeticholic acid
- 10 mg treatment group
- Serious adverse events: 273/825
OCA 25 mg
- Obeticholic acid
- 25 mg treatment group
- Serious adverse events: 276/827
Placebo
- Placebo treatment group
- Comparator arm
- Serious adverse events: 248/825
The three serious-adverse-event denominators sum to the total enrollment of 2,477, providing the arm-level at-risk counts reported in the registry safety analysis.
Study timing
| Milestone | Date |
|---|---|
| Study start | 2015-09-22 |
| Primary completion | 2023-09-15 |
4. Primary Endpoint
The primary endpoint was Time to the First Adjudicated Event for Clinical Outcome Composite Endpoint: Percentage of Participants With an Event, assessed over a time frame of up to 7 years.
The composite endpoint was defined as time to the first occurrence of any of the following adjudicated events:
- Death from any cause.
- Liver transplant.
- Model of End-stage Liver Disease (MELD) score ≥15.
- Hospitalization, defined by a stay of 24 hours or greater, for variceal bleed.
- Hospitalization for hepatic encephalopathy, defined by a West Haven score of ≥2.
- Hospitalization for spontaneous bacterial peritonitis confirmed by diagnostic paracentesis.
- Ascites secondary to cirrhosis requiring medical intervention.
- Histological progression to cirrhosis.
Clinical event distribution was estimated using Kaplan-Meier methodology. The registry reports the primary endpoint as a time-to-event outcome, so the primary treatment comparisons are appropriately interpreted through survival-analysis methods rather than as simple cross-sectional event proportions.
5. Secondary Endpoints
| Secondary endpoint | Time frame | Type | Analysis |
|---|---|---|---|
| Percentage of responders with improvement of fibrosis by at least one stage with no worsening of NASH using consensus read method of scheduled liver biopsies | Up to 7 years | Binary | Cochran-Mantel-Haenszel |
| Percentage of participants who showed improvement in fibrosis by at least 1 stage and/or resolution of NASH without worsening of either using consensus read method | Up to 7 years | Binary | Cochran-Mantel-Haenszel |
Both secondary endpoint families were analyzed in the ITT population. The registry reports treatment differences for these binary outcomes rather than hazard ratios.
6. Statistical Methodology
Intention-to-treat analysis
The primary analyses used an Intent-to-Treat (ITT) Population comprised of all randomized participants with fibrosis stage 2 or stage 3 who received at least one dose of investigational product. The secondary analyses were reported in the ITT population.
The ITT principle keeps participants associated with their randomized treatment group for efficacy comparisons. This is important because the validity of a randomized comparison depends on preserving the treatment assignment established at randomization rather than redefining groups according to subsequent exposure or response.
Log-rank test
The registry reports the log-rank test for both primary comparisons: OCA 10 mg versus placebo and OCA 25 mg versus placebo. The log-rank test is designed for comparing time-to-event distributions while accounting for differing follow-up times and right-censoring.
The log-rank test compares the observed pattern of events over follow-up rather than simply comparing the percentage of participants who have experienced an event at one fixed time point.
Kaplan-Meier estimation
The registry specifies Kaplan-Meier methodology for estimating the clinical-event distribution. Kaplan-Meier estimation accounts for censoring by using the information contributed by each participant up to the time of the event or censoring.
Here, di represents events at time ti, while ni is the number at risk immediately before that time.
Cochran-Mantel-Haenszel analysis
The secondary binary endpoints were analyzed using the Cochran-Mantel-Haenszel test. This framework is used for comparing categorical outcomes while accounting for stratification factors when such strata are part of the analysis.
The reported effect measures for these endpoints are treatment differences in percentages. Unlike a hazard ratio, a treatment difference is an absolute difference between response proportions expressed in percentage points.
Effect measures
| Measure | Used for | Interpretation |
|---|---|---|
| Hazard ratio | Primary time-to-event endpoint | Relative comparison of the estimated instantaneous event rate over follow-up |
| Risk difference / treatment difference | Secondary binary endpoints | Absolute difference between response percentages |
| 95% confidence interval | Primary and secondary estimates | Quantifies uncertainty around the estimated treatment effect |
| P-value | Primary and secondary comparisons | Measures compatibility of the observed data with the statistical null hypothesis under the specified test |
7. Primary Results: OCA 10 mg vs Placebo
For the primary clinical-outcome composite, the registry reports a comparison of OCA 10 mg with placebo using a log-rank test in the ITT population. The hazard ratio was 0.814, with a two-sided 95% confidence interval of 0.635 to 1.043 and a P-value of 0.1028.
OCA 10 mg vs placebo
95% CI: 0.635–1.043 · P = 0.1028
Log-rank test · ITT population · Up to 7 years
A hazard ratio of 0.814 corresponds to an estimated instantaneous event rate approximately 18.6% lower for OCA 10 mg than placebo under the hazard-ratio interpretation. It does not mean that 18.6% fewer participants experienced an event, nor does it mean that every participant experienced the same proportional reduction in event risk.
The two-sided 95% confidence interval of 0.635–1.043 describes the statistical uncertainty around the estimated hazard ratio. Because the interval includes 1, the data are compatible with both a lower and a higher instantaneous event rate under the model. The P-value of 0.1028 is a measure of compatibility with the null hypothesis for the specified log-rank comparison; it is not a measure of the magnitude or clinical importance of the hazard ratio.
Because this is a time-to-event analysis, interpretation also depends on censoring and the assumptions underlying the comparison. A hazard ratio should not be translated automatically into an absolute risk difference without information on the underlying event-time distributions.
8. Primary Results: OCA 25 mg vs Placebo
For the same primary clinical-outcome composite, the registry reports a comparison of OCA 25 mg with placebo using a log-rank test in the ITT population. The hazard ratio was 0.772, with a two-sided 95% confidence interval of 0.600 to 0.994 and a P-value of 0.0444.
OCA 25 mg vs placebo
95% CI: 0.600–0.994 · P = 0.0444
Log-rank test · ITT population · Up to 7 years
A hazard ratio of 0.772 corresponds to an estimated instantaneous event rate approximately 22.8% lower for OCA 25 mg than placebo under the hazard-ratio interpretation. This is a relative time-to-event measure, not a statement that 22.8% of participants avoided an event.
The 95% confidence interval of 0.600–0.994 is relatively close to 1 at its upper boundary. It indicates uncertainty around the estimated treatment effect while remaining below 1 throughout the reported interval. The P-value of 0.0444 describes the result of the specified statistical test and should not be interpreted as the probability that the treatment works or as a measure of effect size.
The interpretation is also conditional on the analysis population, censoring rules, and time-to-event assumptions. In particular, a single hazard ratio summarizes a relative event-rate comparison over follow-up; it does not provide the complete shape of the Kaplan-Meier curves or the absolute probability of an event at a particular time.
| Primary comparison | Method | Analysis population | HR | 95% CI | P-value |
|---|---|---|---|---|---|
| OCA 10 mg vs placebo | Log-rank | ITT | 0.814 | 0.635–1.043 | 0.1028 |
| OCA 25 mg vs placebo | Log-rank | ITT | 0.772 | 0.600–0.994 | 0.0444 |
9. Secondary Results: Fibrosis Improvement Without Worsening of NASH
The first secondary endpoint evaluated the percentage of responders with improvement of fibrosis by at least one stage with no worsening of NASH using the consensus read method of scheduled liver biopsies. The registry reports separate OCA 10 mg and OCA 25 mg comparisons with placebo, both using the Cochran-Mantel-Haenszel test in the ITT population.
OCA 10 mg vs Placebo
Treatment difference
95% CI: 3.6–10.6 · P < 0.0001
Cochran-Mantel-Haenszel test · ITT population
The reported treatment difference of 7.1 represents an absolute difference in the percentage of participants meeting the responder definition, with the OCA 10 mg group having the higher reported response percentage than placebo under the comparison.
The 95% confidence interval of 3.6–10.6 expresses uncertainty around that treatment difference. Unlike the hazard ratio used for the primary time-to-event endpoint, this measure is an absolute percentage-point difference. The P-value of <0.0001 reflects the Cochran-Mantel-Haenszel test under the specified analysis and does not quantify the size or clinical importance of the treatment difference.
OCA 25 mg vs Placebo
Treatment difference
95% CI: 5.8–13.0 · P < 0.0001
Cochran-Mantel-Haenszel test · ITT population
The reported treatment difference of 9.4 represents an absolute percentage-point difference in the responder endpoint between OCA 25 mg and placebo. The 95% confidence interval of 5.8–13.0 describes the uncertainty around that estimate.
The P-value of <0.0001 is evidence from the specified categorical-data test against its null hypothesis. It does not indicate that the treatment effect is 0.0001 in magnitude, and it should not be confused with the confidence interval or the treatment difference itself.
10. Secondary Results: Fibrosis Improvement and/or Resolution of NASH
A second secondary endpoint evaluated the percentage of participants who showed improvement in fibrosis by at least 1 stage and/or resolution of NASH without worsening of either, using the consensus read method. Both dose comparisons used the Cochran-Mantel-Haenszel test in the ITT population.
OCA 10 mg vs Placebo
Treatment difference
95% CI: 3.0–10.4 · P = 0.0004
Cochran-Mantel-Haenszel test · ITT population
The treatment difference of 6.7 is an absolute percentage-point contrast in the proportion meeting the composite histologic response definition. The 95% confidence interval of 3.0–10.4 quantifies uncertainty around that estimate.
The P-value of 0.0004 pertains to the statistical comparison under the Cochran-Mantel-Haenszel test. It does not express the probability that the treatment effect equals 6.7 percentage points and does not establish how large an effect is clinically important.
OCA 25 mg vs Placebo
Treatment difference
95% CI: 5.2–12.8 · P < 0.0001
Cochran-Mantel-Haenszel test · ITT population
The treatment difference of 9.0 is an absolute percentage-point comparison for the fibrosis-improvement and/or NASH-resolution endpoint. The 95% confidence interval of 5.2–12.8 describes uncertainty around the estimate.
The P-value of <0.0001 describes the statistical evidence from the specified categorical-data analysis. It should be considered alongside the effect estimate and confidence interval rather than used as a standalone measure of treatment effect.
| Secondary endpoint | Comparison | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Fibrosis improvement ≥1 stage with no worsening of NASH | OCA 10 mg vs placebo | 7.1 | 3.6–10.6 | <0.0001 |
| Fibrosis improvement ≥1 stage with no worsening of NASH | OCA 25 mg vs placebo | 9.4 | 5.8–13.0 | <0.0001 |
| Fibrosis improvement ≥1 stage and/or resolution of NASH without worsening of either | OCA 10 mg vs placebo | 6.7 | 3.0–10.4 | 0.0004 |
| Fibrosis improvement ≥1 stage and/or resolution of NASH without worsening of either | OCA 25 mg vs placebo | 9.0 | 5.2–12.8 | <0.0001 |
11. Safety: Serious Adverse Events
The registry reports serious adverse events by randomized treatment arm using the following affected-participant and at-risk counts:
| Treatment group | Participants with serious adverse events | Participants at risk | Reported proportion |
|---|---|---|---|
| OCA 10 Milligrams (mg) | 273 | 825 | 273/825 |
| OCA 25 mg | 276 | 827 | 276/827 |
| Placebo | 248 | 825 | 248/825 |
These counts describe the number of participants affected by serious adverse events relative to the number at risk in each arm. They are distinct from the efficacy endpoints and should not be combined with the clinical-outcome composite into a single benefit-risk statistic.
12. Statistical Methods Explained
Why was a log-rank test used for the primary endpoint?
The primary endpoint is explicitly a time-to-event outcome: time to the first adjudicated occurrence of any component of the clinical-outcome composite. Participants can therefore have different follow-up durations and may be censored before experiencing an event. The log-rank test is designed for comparing such survival distributions across randomized groups.
What does a hazard ratio of 0.772 mean?
A hazard ratio of 0.772 means that the estimated instantaneous rate of experiencing the primary event was approximately 22.8% lower for OCA 25 mg than placebo under the hazard-ratio interpretation. It does not mean that 22.8% fewer participants necessarily experienced an event by the end of follow-up.
Why is a hazard ratio different from a risk difference?
A hazard ratio compares event rates over time, while a risk difference compares the proportions experiencing an outcome over a defined period. The primary endpoint uses a hazard ratio because timing and censoring are central to the analysis. The secondary histologic endpoints are binary and are therefore reported using treatment differences in percentages.
Why use the Cochran-Mantel-Haenszel test for the secondary endpoints?
The secondary endpoints classify participants according to whether they meet a binary histologic response definition. The Cochran-Mantel-Haenszel framework provides a method for testing treatment differences in categorical outcomes while accounting for stratification when applicable.
What does the 95% confidence interval tell us?
The confidence interval quantifies statistical uncertainty around an estimated treatment effect. For the OCA 25 mg primary comparison, the interval of 0.600–0.994 indicates uncertainty around the hazard-ratio estimate of 0.772. For the histologic endpoint comparing OCA 25 mg with placebo, the interval of 5.8–13.0 indicates uncertainty around the reported 9.4-percentage-point treatment difference.
Why is the P-value not a measure of effect size?
A P-value summarizes how compatible the observed data are with a specified null hypothesis under the statistical test. It does not tell us how large the treatment effect is. Effect size is described by the hazard ratio or treatment difference, while the confidence interval provides information about precision.
Why does the primary endpoint use a composite?
The registry-defined primary endpoint combines several clinically important events into a single time-to-first-event outcome. A composite endpoint can increase the number of events available for analysis and capture multiple manifestations of clinical progression, but its interpretation depends on the individual components and on which component occurs first. The registry specifies that all potential liver-related clinical outcomes were adjudicated by a blinded, independent Hepatic Outcomes Committee.
13. Interpreting the Two Dose Comparisons
The primary endpoint was evaluated separately for OCA 10 mg versus placebo and OCA 25 mg versus placebo. This creates two primary treatment comparisons against the same placebo group.
| Feature | OCA 10 mg vs placebo | OCA 25 mg vs placebo |
|---|---|---|
| Hazard ratio | 0.814 | 0.772 |
| 95% CI | 0.635–1.043 | 0.600–0.994 |
| P-value | 0.1028 | 0.0444 |
| Analysis | Log-rank | Log-rank |
| Population | ITT | ITT |
The numerical estimates differ between the two dose comparisons, but a difference between two estimated hazard ratios should not automatically be interpreted as evidence of a dose-response relationship. Demonstrating a formal dose-response interaction would require an analysis designed to test that question directly.
The 25 mg comparison has a hazard-ratio estimate of 0.772, whereas the 10 mg comparison has an estimate of 0.814. These estimates are descriptive treatment-effect estimates relative to the common placebo comparator. The confidence intervals are important because they show that the precision of each estimate is limited; the intervals also overlap substantially.
14. Time-to-Event Interpretation
The primary endpoint is not simply the proportion of participants who eventually experience an event. It is explicitly defined as time to the first adjudicated event. That distinction changes how the statistical analysis should be understood.
Event timing matters
A participant experiencing an event early contributes a different event-time history from a participant who remains event-free for years before censoring.
Censoring matters
Participants who do not experience the event during observed follow-up can contribute information through their censoring time without being treated as if they experienced the event.
Composite structure matters
The first qualifying event determines the primary time-to-event endpoint, even though the composite contains several possible clinical outcomes.
Adjudication matters
The registry specifies blinded independent adjudication through the Hepatic Outcomes Committee before clinical outcomes were used in the endpoint assessment.
15. Confidence Intervals and Statistical Precision
The primary hazard-ratio confidence intervals provide a useful way to assess the precision of the two dose-specific estimates.
For OCA 10 mg, the 95% confidence interval extends from 0.635 to 1.043. For OCA 25 mg, it extends from 0.600 to 0.994. The point estimates alone therefore provide an incomplete picture: the confidence intervals show the range of treatment-effect values compatible with the statistical uncertainty represented by the analysis.
16. Primary and Secondary Endpoint Structure
| Endpoint family | Role | Outcome type | Effect measure | Statistical method |
|---|---|---|---|---|
| Clinical-outcome composite | Primary | Time-to-event | Hazard ratio | Log-rank |
| Fibrosis improvement ≥1 stage with no worsening of NASH | Secondary | Binary | Treatment difference / risk difference | Cochran-Mantel-Haenszel |
| Fibrosis improvement ≥1 stage and/or resolution of NASH without worsening of either | Secondary | Binary | Treatment difference | Cochran-Mantel-Haenszel |
This structure illustrates an important statistical principle: the endpoint determines the appropriate effect measure. A time-to-event endpoint naturally leads to survival methods and a hazard ratio, while a binary response endpoint can be summarized with an absolute treatment difference.
17. Limitations
- Composite endpoint: the primary outcome combines multiple clinical events with potentially different clinical meanings and frequencies. A composite estimate should therefore be interpreted as the effect on the combined first-event endpoint rather than as a separate effect on every component.
- First-event structure: once a participant experiences the first qualifying event, later events do not contribute additional events to the primary time-to-first-event analysis.
- Hazard-ratio interpretation: a hazard ratio is a relative time-to-event measure, not an absolute probability and not a direct measure of the proportion of participants benefiting.
- Confidence-interval precision: the OCA 10 mg confidence interval extends above 1, while the OCA 25 mg interval ends at 0.994. The estimates should therefore be interpreted together with their uncertainty intervals rather than as point estimates alone.
- Multiple treatment comparisons: the primary endpoint was analyzed for both OCA doses against placebo. The registry data do not state a specific multiplicity-adjustment procedure for these two primary comparisons.
- Secondary endpoints: the histologic outcomes use a different analysis framework from the primary time-to-event endpoint and should not be interpreted as interchangeable measures.
- Safety interpretation: serious-adverse-event counts are reported as affected participants over participants at risk. They do not provide the event-time information required for a time-to-first-serious-adverse-event analysis.
- Generalizability: the trial population is defined by the study's eligibility and enrollment framework, so the results describe the randomized study population rather than every person with NASH.
18. Why This Trial Matters Statistically
REGENERATE is a useful statistical teaching example because it combines a clinically complex composite time-to-event endpoint with binary histologic outcomes and multiple randomized treatment groups. The resulting analysis illustrates how endpoint structure drives the choice of statistical method.
| Concept | How it appears in REGENERATE |
|---|---|
| Randomization | 2,477 participants were enrolled into a randomized three-arm phase 3 study. |
| Blinding | The study used quadruple masking, and potential clinical outcomes were adjudicated by a blinded independent Hepatic Outcomes Committee. |
| ITT analysis | The primary analyses used the ITT population defined in the registry. |
| Time-to-event endpoint | The primary outcome measured time to the first adjudicated clinical event in a composite endpoint. |
| Kaplan-Meier estimation | The registry specifies Kaplan-Meier methodology for the clinical event distribution. |
| Log-rank test | The two primary OCA-versus-placebo comparisons used the log-rank test. |
| Hazard ratio | The primary treatment effects were reported as hazard ratios with 95% confidence intervals. |
| Binary endpoint analysis | Histologic responder endpoints were analyzed using the Cochran-Mantel-Haenszel test. |
| Risk difference | Secondary treatment effects were reported as differences in percentages. |
| Composite endpoint | Several adjudicated clinical outcomes were combined into a time-to-first-event endpoint. |
19. Overall Statistical Interpretation
The REGENERATE statistical results contain two distinct layers of evidence. The primary layer is the time-to-event comparison of OCA with placebo for the adjudicated clinical-outcome composite. The OCA 10 mg comparison produced a hazard ratio of 0.814 with a 95% confidence interval of 0.635–1.043 and P = 0.1028. The OCA 25 mg comparison produced a hazard ratio of 0.772 with a 95% confidence interval of 0.600–0.994 and P = 0.0444.
The second layer consists of the histologic secondary endpoints. The reported treatment differences were 7.1 and 9.4 percentage points for fibrosis improvement by at least one stage without worsening of NASH, and 6.7 and 9.0 percentage points for fibrosis improvement and/or resolution of NASH without worsening of either, for OCA 10 mg and OCA 25 mg respectively.
These estimates answer different statistical questions. The primary hazard ratios describe relative time-to-event effects, whereas the secondary treatment differences describe absolute differences in binary responder outcomes. Reading the trial statistically therefore requires keeping the endpoint definitions, analysis populations, effect measures, confidence intervals, and P-values aligned with the analysis from which they originated.
The most important distinction is between what was measured and how it was analyzed. A time-to-first-event endpoint should not be reduced to a simple percentage comparison, while a binary histologic endpoint should not be interpreted through a hazard ratio. Matching the effect measure to the endpoint is fundamental to correct clinical-trial interpretation.
20. Related Tutorials
Learn more about the methods used in this trial:
21. Related Calculators
22. Sources
- ClinicalTrials.gov: REGENERATE, NCT02548351.
- PubMed: PMID 38126958.
- PubMed: PMID 34793868.
- PubMed: PMID 34274514.
- PubMed: PMID 31813633.
- PubMed: PMID 31260793.
Continue through the Clinical Biostats knowledge graph
Connect the endpoints and statistical methods used in REGENERATE with deeper tutorials and practical statistical tools.
23. Record Summary
REGENERATE illustrates the statistical distinction between time-to-event and binary clinical-trial endpoints. Its primary analysis used log-rank testing and hazard ratios for time to the first adjudicated event in a clinically defined composite, while secondary histologic outcomes used Cochran-Mantel-Haenszel testing and treatment differences in percentages. The reported results therefore require interpretation on their respective statistical scales: hazard ratios for the primary time-to-event comparison and absolute percentage-point differences for the binary histologic outcomes.