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
AURORA was a randomized, double-blind, phase 3 parallel-group trial evaluating cenicriviroc for liver fibrosis in adults with nonalcoholic steatohepatitis. The registry identifies two primary endpoints: histologic improvement in fibrosis at Month 12 and time to first occurrence of adjudicated clinical events in the full study cohort.
| Feature | AURORA |
|---|---|
| Trial name | AURORA |
| Phase | Phase 3 |
| Population | Adults with nonalcoholic steatohepatitis |
| Design | Randomized, double-blind, parallel-group |
| Primary purpose | Treatment |
| Interventions | Placebo; Cenicriviroc |
| Primary endpoints | Binary fibrosis response at Month 12; time to first occurrence of adjudicated events |
| Enrollment | 1778 |
| Study status | Terminated |
| Start | April 5, 2017 |
| Primary completion | January 12, 2021 |
| Lead sponsor | Tobira Therapeutics, Inc. |
2. Clinical Question
The central question was whether treatment with cenicriviroc 150 mg improved liver fibrosis-related outcomes compared with placebo in adults with nonalcoholic steatohepatitis.
Population
Adults with nonalcoholic steatohepatitis.
Intervention
Cenicriviroc 150 mg.
Comparator
Placebo.
Primary question
Does cenicriviroc improve fibrosis-related histologic outcomes and reduce the occurrence of clinically important adjudicated events?
3. Trial Design
Placebo
- Placebo
- Randomized parallel-group assignment
- Serious adverse events: 70/589
Cenicriviroc 150 mg
- Cenicriviroc 150 mg
- Randomized parallel-group assignment
- Serious adverse events: 159/1180
The trial was randomized and double-blind. The registry identifies a parallel design and treatment as the primary purpose. The posted binary endpoint analyses compare cenicriviroc 150 mg with placebo.
4. Endpoints
Primary Endpoint 1: Fibrosis improvement at Month 12
The first primary endpoint was the percentage of participants with improvement in fibrosis by at least 1 stage and no worsening of steatohepatitis on liver histology at Month 12.
| Element | Registry definition |
|---|---|
| Time frame | Month 12 |
| Endpoint type | Binary |
| Fibrosis system | NASH Clinical Research Network Fibrosis Staging System |
| Fibrosis improvement | A decrease in the NASH CRN fibrosis stage |
| Required histologic outcome | Improvement in fibrosis by at least 1 stage with no worsening of lobular inflammation or hepatocellular ballooning |
The fibrosis staging system defines stages from 0 through 4, with stage 0 representing no fibrosis and stage 4 representing cirrhosis. Intermediate categories include perisinusoidal, periportal, bridging, and cirrhotic patterns.
The registry defines the nonalcoholic fatty liver disease activity score as a semiquantitative score based on steatosis, lobular inflammation, and hepatocellular ballooning. No worsening of steatohepatitis was defined as no worsening of lobular inflammation or hepatocellular ballooning grade.
Primary Endpoint 2: Time to first adjudicated event
The second primary endpoint was time to first occurrence of adjudicated events in the full study cohort.
| Element | Registry definition |
|---|---|
| Time frame | From first dose of study drug to onset of first occurrence of the event, up to approximately 42 months |
| Endpoint type | Time-to-event |
| Population | Full study cohort |
| Event components | Death from any cause; histopathologic progression to cirrhosis; liver transplant; MELD score ≥15; ascites; hospitalization for variceal bleed, hepatic encephalopathy, or spontaneous bacterial peritonitis |
The registry defines the time origin as the first dose of randomized investigational product. The composite endpoint therefore incorporates both mortality and clinically important liver-related disease progression.
5. Statistical Methodology
The posted statistical analyses for the Month 12 histologic endpoints used the Cochran-Mantel-Haenszel test. The effect measures were percentage difference, represented here as risk difference, and odds ratio.
A negative risk difference means that the observed percentage meeting the endpoint was lower in the cenicriviroc group than in the placebo group. The magnitude is expressed in percentage points.
An odds ratio below 1 indicates lower odds of the endpoint in the cenicriviroc group relative to placebo. Odds ratios and risk differences describe different aspects of the same binary comparison.
For the odds-ratio analyses, the registry states that Mantel-Haenszel estimates compared cenicriviroc with placebo while controlling for the randomization strata: fibrosis stage [2 vs 3] and presence or absence of type 2 diabetes mellitus at baseline.
The primary fibrosis analysis used the mITT population for Part 1, defined as participants randomly assigned to a treatment group who received at least one dose of study drug in Part 1.
What the Cochran-Mantel-Haenszel approach contributes
A simple unadjusted comparison can ignore important stratification factors used during randomization. The Cochran-Mantel-Haenszel approach produces a stratified comparison that accounts for the prespecified strata rather than treating all observations as if they came from a single unstructured 2 × 2 table.
For this trial, the reported odds ratios were based on Mantel-Haenszel estimates controlling for fibrosis stage and baseline type 2 diabetes status. This is particularly relevant because those factors defined the randomization strata for the posted analyses.
Time-to-event endpoint
The second primary endpoint is a time-to-event composite. For an endpoint of this type, the usual statistical framework would use time-to-event methods that account for censoring, commonly Kaplan-Meier estimation and a comparison such as a log-rank test, with a Cox proportional-hazards model commonly used to estimate a hazard ratio. The registry's posted statistical-analysis records summarized here do not provide an estimate, confidence interval, or p-value for this primary endpoint.
6. Results
Primary Endpoint: Improvement in Fibrosis by at Least 1 Stage With No Worsening of Steatohepatitis
Risk difference: Cenicriviroc vs Placebo
95% CI: −8.2 to 1.9 · P = 0.2067
Cochran-Mantel-Haenszel analysis in the Part 1 mITT population.
The reported risk difference of −3.2 percentage points means that the percentage of participants meeting the fibrosis-improvement endpoint was estimated to be 3.2 percentage points lower with cenicriviroc than with placebo in the analyzed comparison.
The estimate does not mean that cenicriviroc caused a 3.2-percentage-point decline in every patient, nor does it describe an individual patient's probability of benefit. It is a group-level difference between the two randomized treatment groups.
The 95% confidence interval extends from −8.2 to 1.9 percentage points. Because the interval crosses 0, the data are compatible with both a lower and a higher percentage meeting the endpoint with cenicriviroc under the statistical framework used for the analysis.
The p-value of 0.2067 is evidence used for the superiority hypothesis test; it is not a measure of effect size. The size and direction of the estimated risk difference, together with its confidence interval, provide the information about magnitude and precision.
Primary Endpoint: Odds Ratio for Fibrosis Improvement
Mantel-Haenszel odds ratio
95% CI: 0.6341–1.1044
Cenicriviroc vs placebo; two-sided 95% confidence interval.
An odds ratio of 0.8369 indicates that the estimated odds of meeting the fibrosis-improvement endpoint were lower in the cenicriviroc group than in the placebo group, after accounting for the stated randomization strata.
The odds ratio is not a risk ratio and should not be translated directly into a percentage reduction in probability. When event probabilities are not rare, odds and risks can differ substantially.
The 95% confidence interval of 0.6341 to 1.1044 includes 1.00. Thus, the interval includes values compatible with lower odds as well as values compatible with higher odds under the model and sampling framework.
No p-value is posted with this odds-ratio analysis in the statistical-analysis record. The corresponding superiority analysis with the risk-difference effect measure reports P = 0.2067. A p-value should not be treated as a substitute for the effect estimate or its confidence interval.
Secondary Endpoint: Improvement in Fibrosis by at Least 2 Stages With No Worsening of Steatohepatitis
Risk difference
95% CI: −4.8 to 1.5 · P = 0.2827
Cochran-Mantel-Haenszel analysis in the Part 1 mITT population.
The estimated difference was −1.7 percentage points, indicating a lower observed percentage meeting the at-least-2-stage fibrosis-improvement endpoint in the cenicriviroc group than in the placebo group.
The 95% CI of −4.8 to 1.5 percentage points crosses 0. The interval therefore includes the possibility of a modest lower percentage as well as a modest higher percentage in the cenicriviroc group.
The p-value of 0.2827 addresses the superiority hypothesis under the posted analysis. It does not quantify the clinical importance of the observed difference.
Secondary Endpoint: Odds Ratio for Improvement in Fibrosis by at Least 2 Stages
Mantel-Haenszel odds ratio
95% CI: 0.5032–1.2229
Cenicriviroc vs placebo; adjusted for the stated randomization strata.
The estimated odds ratio of 0.7844 is below 1, corresponding to lower estimated odds of the endpoint in the cenicriviroc group. The 95% CI extends from 0.5032 to 1.2229, crossing 1.00 and therefore including both lower and higher odds relative to placebo.
This odds-ratio result should be read together with the corresponding risk difference rather than as an alternative measure of the same quantity. The risk difference is directly expressed in percentage points; the odds ratio is a relative comparison of odds.
Secondary Endpoint: Improvement in Fibrosis by at Least 1 Stage Regardless of Effect on Steatohepatitis
Risk difference
95% CI: −8.7 to 3.3 · P = 0.4054
Cochran-Mantel-Haenszel analysis in the Part 1 mITT population.
The estimated risk difference was −2.7 percentage points. The confidence interval of −8.7 to 3.3 percentage points spans zero, indicating substantial uncertainty around the direction and magnitude of the between-group difference.
The p-value of 0.4054 is a hypothesis-test result and does not indicate that the treatment groups are biologically identical. A nonsignificant p-value should not be interpreted as proof of equivalence or absence of any treatment effect.
Secondary Endpoint: Odds Ratio for Improvement in Fibrosis by at Least 1 Stage Regardless of Effect on Steatohepatitis
Mantel-Haenszel odds ratio
95% CI: 0.6709–1.1744
Cenicriviroc vs placebo; adjusted for fibrosis stage and baseline type 2 diabetes status.
The odds ratio of 0.8877 is below 1, but its 95% confidence interval, 0.6709 to 1.1744, includes 1.00. The estimate therefore indicates a direction of lower odds in the cenicriviroc group, while the interval reflects uncertainty that includes higher odds as well.
The endpoint differs conceptually from the primary endpoint because it evaluates improvement in fibrosis regardless of the effect on steatohepatitis. That distinction changes the clinical meaning of the binary outcome even though the same general statistical method was used.
Second Primary Endpoint: Time to First Adjudicated Event
The endpoint is a composite time-to-event measure beginning at the first dose of randomized investigational product and extending up to approximately 42 months. Its components include death, histopathologic progression to cirrhosis, liver transplant, MELD score ≥15, ascites, and hospitalization for variceal bleed, hepatic encephalopathy, or spontaneous bacterial peritonitis.
For a time-to-first-event endpoint, a standard analysis would account for the exact follow-up time contributed by each participant and for censoring among participants who do not experience an event during observed follow-up. Kaplan-Meier methods can describe the event-free distribution, while a log-rank comparison and Cox proportional-hazards model are commonly used for randomized-group comparisons. A hazard ratio would describe a relative event rate over time rather than an absolute difference in cumulative incidence.
7. Secondary Results in Context
| Endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| ≥1-stage fibrosis improvement + no worsening of steatohepatitis | Risk difference | −3.2 | −8.2 to 1.9 | 0.2067 |
| Same endpoint | Odds ratio | 0.8369 | 0.6341–1.1044 | — |
| ≥2-stage fibrosis improvement + no worsening of steatohepatitis | Risk difference | −1.7 | −4.8 to 1.5 | 0.2827 |
| Same endpoint | Odds ratio | 0.7844 | 0.5032–1.2229 | — |
| ≥1-stage fibrosis improvement regardless of effect on steatohepatitis | Risk difference | −2.7 | −8.7 to 3.3 | 0.4054 |
| Same endpoint | Odds ratio | 0.8877 | 0.6709–1.1744 | — |
The six posted statistical analyses therefore provide two complementary effect measures for each of three binary histologic outcomes. The risk differences express absolute separation in percentage points, while the odds ratios express relative separation on the odds scale.
An important statistical distinction is that the risk difference and odds ratio are not interchangeable. A risk difference near zero can coexist with an odds ratio that is meaningfully above or below 1 depending on the underlying event rates. Interpretation should therefore begin with the effect measure actually reported rather than translating one measure into another without the underlying group-level event proportions.
8. Statistical Methods Explained
Why was the Cochran-Mantel-Haenszel test used?
The posted binary analyses used a Cochran-Mantel-Haenszel framework because the treatment comparison was conducted across prespecified randomization strata. In AURORA, the odds-ratio analyses explicitly controlled for fibrosis stage [2 vs 3] and presence or absence of type 2 diabetes mellitus at baseline.
Stratification can improve alignment between the analysis and the randomization scheme. Instead of simply pooling all participants into one unadjusted comparison, the Mantel-Haenszel approach combines information across strata while accounting for their structure.
What does a risk difference of −3.2 mean?
A risk difference of −3.2 means the percentage meeting the endpoint was estimated to be 3.2 percentage points lower in the cenicriviroc group than in the placebo group. The unit is percentage points, not a relative percent change.
For example, a risk difference of −3.2 would correspond to 3.2 fewer participants meeting the endpoint per 100 participants if the underlying percentages differed by exactly that amount. It does not mean that the risk was reduced by 3.2% relative to placebo.
What does an odds ratio of 0.8369 mean?
An odds ratio of 0.8369 means the estimated odds of meeting the endpoint were 0.8369 times the odds in the placebo group, after accounting for the stated stratification factors. Equivalently, the estimated odds were lower in the cenicriviroc group.
Because an odds ratio compares odds rather than probabilities, it should not be interpreted as saying that the probability was 16.31% lower. That conversion would require the underlying event probabilities.
Why does the confidence interval matter?
The confidence interval communicates the precision of the estimated treatment effect. For the primary risk difference, the interval ranges from −8.2 to 1.9 percentage points. That interval is considerably more informative than the point estimate alone because it shows the range of effects compatible with the statistical model and sampling uncertainty.
For odds ratios, the neutral value is 1.00. For risk differences, the neutral value is 0. A confidence interval crossing its relevant neutral value indicates that the analysis does not exclude the null value at the corresponding two-sided 95% confidence level.
Why is the p-value not an effect-size measure?
A p-value quantifies the evidence against a specified null hypothesis under the statistical model. It does not tell the reader how large or clinically important the treatment difference is. Effect size and uncertainty are described by the risk difference or odds ratio together with its confidence interval.
A very small p-value can accompany a small effect in a sufficiently large study, while a clinically meaningful effect estimate can have a wide confidence interval when information is limited. Reading the p-value separately from the effect estimate can therefore obscure the actual magnitude of the observed comparison.
Why does the mITT population matter?
The posted Part 1 analysis population included participants who were randomized and received at least one dose of study drug. This differs from a strict all-randomized intention-to-treat population because eligibility for the mITT analysis includes treatment exposure.
The distinction matters because excluding randomized participants after randomization can alter the population represented by the analysis. The exact population definition should always accompany the treatment-effect estimate so that the estimand and denominator are understood.
How should the time-to-event endpoint be interpreted?
The second primary endpoint is defined by the time from first dose to the first adjudicated clinical event. Time-to-event analysis differs from a simple proportion because it incorporates when events occur and how long participants remain under observation without an event.
The composite endpoint also requires careful clinical interpretation because its components range from death and transplant to disease progression and hospitalization. A time-to-first-event result would describe the composite endpoint as defined, not necessarily each individual component separately.
9. Randomization and Stratification
Randomization is central to the causal interpretation of a controlled clinical trial because treatment assignment is determined independently of participants' observed and unobserved baseline characteristics, subject to the randomization procedure.
For the posted odds-ratio analyses, the registry specifies two randomization strata: fibrosis stage [2 vs 3] and presence or absence of type 2 diabetes mellitus at baseline.
Stage 2 vs Stage 3
Present vs absent
The value of stratified analysis is that it respects these prespecified factors rather than ignoring them when estimating the treatment contrast. The registry's odds-ratio analyses explicitly state that the Mantel-Haenszel estimates controlled for these factors.
10. Multiplicity and Multiple Endpoints
AURORA has two registered primary endpoints and multiple secondary endpoints. The posted statistical analyses also evaluate two different effect measures for each of the three reported binary histologic outcomes.
Multiple endpoint analyses create an important distinction between an individual nominal p-value and a complete confirmatory testing strategy. The interpretation of a p-value depends on the hypothesis being tested and on whether the broader family of hypotheses has been accounted for by the trial's prespecified multiplicity strategy.
| Analysis component | Role | Statistical consideration |
|---|---|---|
| Month 12 ≥1-stage fibrosis improvement + no worsening | Primary endpoint | Cochran-Mantel-Haenszel; risk difference and odds ratio |
| Time to first adjudicated event | Primary endpoint | Time-to-event endpoint; posted estimate not contained in the statistical-analysis records summarized here |
| Month 12 ≥2-stage fibrosis improvement + no worsening | Secondary endpoint | Cochran-Mantel-Haenszel; risk difference and odds ratio |
| Month 12 ≥1-stage fibrosis improvement regardless of steatohepatitis | Secondary endpoint | Cochran-Mantel-Haenszel; risk difference and odds ratio |
Because the registry contains multiple analyses, a reader should avoid treating each p-value as though it were an isolated hypothesis test unrelated to the rest of the analysis plan. The hierarchy and multiplicity strategy determine how individual findings fit into the overall confirmatory evidence.
11. Safety Results
The registry reports serious adverse events by arm as follows:
| Safety measure | Placebo | Cenicriviroc 150 mg |
|---|---|---|
| Serious adverse events | 70/589 | 159/1180 |
The registry reports the affected and at-risk counts rather than a posted comparative effect estimate for serious adverse events. The denominators also differ between the two arms, so the raw counts alone should not be interpreted as evidence of a treatment-group difference.
Safety and efficacy answer different statistical questions. Efficacy evaluates whether the prespecified clinical or histologic outcome differs between randomized groups; safety describes adverse events occurring during treatment exposure. Both should be examined independently before considering the overall evidence from a trial.
12. Limitations and Interpretation Issues
- Composite primary endpoint: the time-to-first-event endpoint combines several clinically distinct outcomes, including death, cirrhosis progression, transplant, MELD score ≥15, ascites, and hospitalization for serious liver-related complications.
- Component interpretation: a composite result describes the first occurrence of any component and does not necessarily imply that every component behaves similarly.
- mITT analysis population: the posted Part 1 analyses use a population defined by randomization plus receipt of at least one dose, rather than all randomized participants.
- Confidence intervals: all six posted binary analyses have confidence intervals that include their respective neutral values: 0 for risk differences and 1 for odds ratios.
- Multiple analyses: the registry contains primary and secondary endpoints and more than one effect measure. Individual p-values therefore need to be interpreted within the overall analysis strategy.
- Odds-ratio interpretation: odds ratios are not risk ratios and should not be described as percentage changes in probability without the underlying event rates.
- Time-to-event analysis: interpretation of the second primary endpoint requires attention to censoring, follow-up duration, and the composition of the adjudicated event definition.
- Safety denominators: serious adverse-event counts are reported with different numbers at risk across treatment groups, so counts alone are insufficient for a direct rate comparison.
13. Why This Trial Matters Statistically
AURORA is a useful statistical teaching case because its registered endpoints span two fundamentally different outcome structures: a binary histologic response assessed at a fixed time point and a composite time-to-event endpoint followed over an extended period.
| Concept | How it appears in AURORA |
|---|---|
| Randomization | Randomized allocation to placebo or cenicriviroc |
| Blinding | Double-blind trial design |
| Parallel groups | Two-arm parallel design |
| Binary endpoint analysis | Month 12 fibrosis-improvement outcomes |
| Cochran-Mantel-Haenszel test | Posted analyses of the binary histologic outcomes |
| Risk difference | Absolute difference in percentages between treatment groups |
| Odds ratio | Relative comparison of odds using Mantel-Haenszel estimates |
| Stratification | Fibrosis stage and baseline type 2 diabetes status |
| Confidence intervals | Uncertainty around each posted treatment-effect estimate |
| P-values | Superiority hypothesis testing for the reported risk-difference analyses |
| Time-to-event analysis | Second primary endpoint based on time to first adjudicated event |
| Composite endpoint | Multiple clinically important adjudicated event types combined into one time-to-first-event outcome |
The trial also illustrates why the choice of estimand and effect measure matters. A binary fibrosis endpoint asks whether a participant meets a prespecified histologic definition by Month 12. A time-to-event endpoint asks when the first clinically important event occurs. These are related clinical questions, but they require different statistical representations.
14. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
15. Related Statistical Calculators
16. Sources
- ClinicalTrials.gov: AURORA, NCT03028740.
- Linked publication: PubMed record for PMID 37061109.
Continue through Clinical Biostats
Connect this trial's statistical concepts with deeper biostatistics tutorials and practical statistical calculators.
17. Record Summary
AURORA provides a useful example of how a phase 3 randomized trial can combine histologic and clinical endpoints within one statistical program. The Month 12 binary analyses used Cochran-Mantel-Haenszel methods with both risk differences and Mantel-Haenszel odds ratios, while the second primary endpoint was defined as time to first occurrence of an adjudicated clinical event.
The primary fibrosis-improvement risk difference was −3.2 percentage points with a 95% CI of −8.2 to 1.9 and P = 0.2067. The corresponding odds ratio was 0.8369 with a 95% CI of 0.6341–1.1044. The secondary binary analyses showed the same general statistical pattern: risk-difference confidence intervals crossed zero and odds-ratio confidence intervals crossed 1.
The most important statistical lesson is that these estimates should be interpreted jointly with their endpoint definitions, analysis population, stratification factors, confidence intervals, and multiplicity context. The binary fibrosis endpoint and the time-to-first-event endpoint address different aspects of disease progression and require different statistical reasoning.