Metabolic Interventions to Resolve Non-alcoholic Steatohepatitis (NASH) With Fibrosis (MIRNA): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Metabolic Interventions to Resolve Non-alcoholic Steatohepatitis (NASH) With Fibrosis (MIRNA) (NCT04321031), a randomized phase 2 study. The primary outcome is Mean Proportion of Participants Achieving Resolution of NASH Without Worsening/Improvement of Fibrosis by >=1 Stage Without Worsening of NASH/Both Based on Assessment by Sponsor-Identified Central Pathologist at Week 48:Bayesian Dose Response Model (BDRM).

2. Clinical Question

The study was designed to examine the effects of a range of PF-06865571 doses, alone and in combination with PF-05221304, on histologic outcomes. The primary measure combines clinically distinct biopsy changes: resolution of NASH without worsening fibrosis and improvement in fibrosis without worsening NASH. These outcomes are based on liver biopsy assessment rather than a symptom measure or a laboratory value.

The registry reports posterior mean proportions for placebo and the PF-06865571-alone groups in its primary dose-response results. It also reports pairwise risk differences between placebo and those dose groups. Although the trial included combination-treatment arms, combination-arm results are not included in the posted primary endpoint estimates listed in the record. Whether that reflects the endpoint reporting or the analysis population is not reported in the registry record.

The core statistical question is not only whether estimated proportions differ, but also how uncertain those estimates are. The reported primary risk-difference intervals include values on either side of zero for each posted dose comparison, so the results do not pinpoint a clear difference from placebo at the reported interval level.

3. Trial Design

MIRNA was a randomized, double-blind, double-dummy, placebo-controlled, dose-ranging, parallel-group phase 2 study sponsored by Pfizer. The registry lists quadruple masking. It enrolled 256 participants and is marked completed. The listed study start was 2020-06-15, and the primary completion date was 2024-01-23.

The target population was adults with biopsy-confirmed nonalcoholic steatohepatitis and fibrosis stage 2 or 3. The treatment period was 48 weeks. The listed arms were placebo; PF-06865571 at 25 mg, 75 mg, 150 mg, and 300 mg twice daily; PF-06865571 150 mg twice daily with PF-05221304 5 mg twice daily; and PF-06865571 300 mg twice daily with PF-05221304 10 mg twice daily.

The primary endpoint results include placebo and four PF-06865571-alone groups. Their reported analysis sizes are 34 for placebo, 35 for the 25 mg group, 48 for the 75 mg group, 42 for the 150 mg group, and 31 for the 300 mg group. These endpoint-specific counts should not be confused with the overall enrollment figure. The record describes the full analysis set as randomized participants who took at least one dose and had baseline data for the primary endpoint; further detail about exclusions or missing endpoint data is not reported in the registry record.

The trial’s stated purpose was to evaluate a range of doses of PF-06865571 alone and with PF-05221304, assessing NASH resolution or improvement in liver fibrosis histologically. The posted results include 16 outcome measures and 84 analyses. The primary endpoint uses a Bayesian dose-response model, while several listed secondary outcomes involve pairwise comparisons using logistic regression or analysis of covariance. Further details of the study’s planned statistical procedures, including any interim analyses or sample-size justification, are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: F2: significant stage of fibrosis when scarring had occurred and extended outside liver area and F3: severe stage of fibrosis with spreading and forming bridges with other fibrotic liver areas.

Only the first 6 of 7 groups are shown.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: BKS: scaling for fibrosis (0=none,1=perisinusoidal/ periportal,2=perisinusoidal, portal/ periportal,3=bridging,4=cirrhosis; higher scores=more DA). Full analysis set (FAS):all randomized participants who took at least 1 dose of investigational product who had provided baseline data for primary endpoint. This outcome measure was not planned to be analyzed in combination arms (DGAT2i/PF-06865571 150 mg BID + ACCi/PF-05221304 5 mg BID and DGAT2i/PF-06865571 300 mg BID + ACCi/PF-05221304 10 mg BID).

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: Safety population included all participants who took at least 1 dose of investigational product. Participants were analyzed according to the treatment they actually received.

Only the first 6 of 8 groups are shown.

6. Primary Endpoint

The registry lists the primary outcome as Mean Proportion of Participants Achieving Resolution of NASH Without Worsening/Improvement of Fibrosis by >=1 Stage Without Worsening of NASH/Both Based on Assessment by Sponsor-Identified Central Pathologist at Week 48:Bayesian Dose Response Model (BDRM), measured over Week 48.

The primary endpoint is a mean proportion of participants at week 48 meeting a histology-based response definition. The response includes NASH resolution without worsening of fibrosis, improvement in fibrosis by at least one stage without worsening of NASH, or both, as assessed by sponsor-identified central pathologists. The registry describes NASH resolution through biopsy features including ballooning, lobular inflammation, and steatosis. Fibrosis stage is assessed using the Brunt-Kleiner scale.

In general, a composite endpoint can capture more than one route to a favorable histologic outcome. That makes the outcome relevant to the trial’s stated aims, but it also means that a single proportion combines distinct response patterns. The registry lists a separate secondary outcome for NASH resolution without worsening fibrosis and other outcomes for fibrosis improvement without worsening NASH. These can help distinguish components of the broader primary endpoint.

The primary model characterizes dose response across twice-daily treatment groups and estimates posterior mean proportions with 90% credible intervals. The registry calls these intervals confidence intervals in its description. A credible interval expresses uncertainty about a parameter under the specified Bayesian model and prior assumptions; details of those assumptions are not reported in the registry record.

Biopsy-based assessment is suited to this endpoint because the response definition depends on changes in liver tissue features and fibrosis stage. It also means the result reflects the pathology assessment and the endpoint’s defined thresholds, rather than a direct measure of how a participant feels or functions. The record does not report additional validation or clinical interpretation criteria for the combined response definition.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is analyzed with a Bayesian dose-response model. In general, a Bayesian model combines the observed data with a prior distribution to produce a posterior distribution for quantities of interest. A posterior mean is a summary of that distribution; a credible interval describes a range containing a stated proportion of posterior probability under the model. Here, the registry reports posterior mean proportions and 90% credible intervals, while noting that the intervals are indicated as confidence intervals in the posted results. The prior distributions, model structure beyond its dose-response purpose, and computational details are not reported in the registry record.

The primary results give an estimate for placebo and each PF-06865571-alone dose group. A group-level proportion describes the estimated frequency of meeting the combined histology response definition in that group. It is not itself a treatment-versus-placebo contrast. The risk difference is the contrast between two proportions: it expresses the difference on the absolute proportion scale. A risk difference of zero represents equal proportions. The registry’s reported primary risk differences are accompanied by 90% intervals, which include zero in each listed comparison. That indicates uncertainty about whether the contrast is positive or negative; it does not establish equivalence.

A confidence interval, in general, is a procedure-based interval with a repeated-sampling interpretation. A credible interval, in general, is interpreted through posterior probability conditional on the model and prior. Because the registry’s endpoint description identifies the primary intervals as credible intervals but labels them as confidence intervals in the results, the distinction matters when interpreting the interval. The record does not describe the prior or give enough detail to independently reconstruct the Bayesian analysis.

Several secondary binary outcomes are reported as participant counts, and pairwise comparisons are identified as logistic regression analyses. Logistic regression models the log odds of a binary outcome as a function of explanatory variables. The registry does not report all model covariates or the full specification. The posted arm counts also allow simple unadjusted re-calculations of risk differences. Those re-calculations use the observed counts and group sizes without adjustment, and therefore are not the same estimand or procedure as a logistic regression analysis. Their p-values are unadjusted and should be treated as exploratory descriptive calculations, not as replacements for the sponsor’s analysis.

In general, sparse binary outcomes can produce imprecise estimates because few participants experience the outcome. The fibrosis improvement by at least two stages outcome has small counts across the listed groups, so estimates based on it are especially sensitive to the individual observations. Wide intervals are a natural reflection of limited information. Conversely, a comparatively large count does not by itself guarantee a precise or unbiased estimate: the analysis population, missing observations, outcome definition, and model also matter.

The sample-size figures associated with the unadjusted secondary comparisons are planning illustrations based on the observed effect. They are not observed power and are not the trial’s documented design justification. In general, such illustrations ask how many participants per arm might be required to detect an effect under stated assumptions. The record describes a two-sided significance level of 0.05, equal allocation, and a normal approximation, with alternative planning scenarios based on stated fractions of the observed effect. These are hypothetical planning assumptions; they do not establish that the trial was designed using them or that the observed effects will recur in another study.

Multiplicity is relevant because the record lists many outcome measures and analyses. In general, testing multiple hypotheses increases the chance of at least one apparently unusual result when all null hypotheses are true. Adjustment methods, a prespecified testing hierarchy, and the handling of multiplicity are not reported in the registry record. Thus, unadjusted secondary p-values should not be interpreted as if they were protected by a single prespecified confirmatory test.

The full analysis set is described as all randomized participants who took at least one dose of investigational product and had baseline data for the primary endpoint. The definition does not state how missing week 48 biopsy data were handled, whether additional eligibility criteria were applied, or how participants with incomplete assessments contributed. Those details affect how estimates should be understood and cannot be inferred from the brief population description.

Over-dispersion is a general concern when variation in binary outcomes exceeds what a simple probability model expects. If present, it can make standard errors too small and intervals too narrow unless the model accounts for it. The registry record does not report an assessment or correction for over-dispersion. This is a general statistical consideration, not a claim that over-dispersion occurred in this trial.

10. Analyses Reported by the Sponsor

The sponsor’s posted primary analysis uses a Bayesian dose-response model and reports posterior mean proportions and 90% intervals. It also reports placebo comparisons as risk differences with 90% intervals. Those are the primary estimates to use when describing the sponsor’s analysis. The posted primary comparisons do not include p-values, and the intervals include zero for each PF-06865571-alone dose comparison.

For selected secondary binary outcomes, the registry identifies pairwise comparisons using logistic regression. The separate recalculations use only the posted counts and group sizes to produce simple unadjusted contrasts. They do not reproduce the sponsor’s regression model, and differences between these results can arise because the methods and estimands need not match exactly. The sponsor’s reported secondary risk differences and intervals should therefore be distinguished from the crude re-calculations rather than blended with them.

The record does not mark a discrepancy between sponsor results and recalculations. Even so, a simple unadjusted calculation is not a substitute for the sponsor’s analysis. The recalculated p-values are explicitly unadjusted and do not account for multiple outcomes or comparisons. They can serve as a check on the direction and approximate scale of a contrast from the posted counts, but the registry’s own analysis remains the appropriate reference for what the sponsor reported.

For the primary endpoint, the posted information does not include a corresponding recalculation of the Bayesian model. Reproducing that analysis would require model and prior details not reported in the registry record. No independent primary-model estimate should be inferred from the secondary count-based calculations.

11. Multiplicity, Missing Data and Analysis Populations

The registry record lists 16 outcome measures and 84 analyses, including a primary Bayesian dose-response endpoint and multiple secondary outcomes. It does not report whether a multiplicity adjustment was planned or applied, whether hypotheses were tested in a hierarchy, or which analyses were considered confirmatory. Without that information, nominal intervals or unadjusted p-values across secondary comparisons should not be read as a family of controlled confirmatory tests. In general, the more hypotheses examined, the greater the opportunity for chance findings.

The record describes the primary analysis population as randomized participants who received at least one dose of investigational product and had baseline data for the primary endpoint. It does not report a complete account of exclusions, the number of participants missing week 48 biopsy results, or the methods used to handle missing endpoint data. It is therefore not possible to assess from the registry description whether the primary analysis used imputation, a missing-at-random assumption, a complete-case approach, or another method.

The primary endpoint depends on histologic assessment at week 48. In general, missing biopsies or unavailable pathology assessments can affect the observed response proportion, particularly if missingness is related to treatment, health status, or the probability of a response. The registry does not report sensitivity analyses for missing data or a comparison of participants with and without endpoint assessments.

Additional analysis details are also not reported in the registry record, including a full primary model specification, prior choices, handling of protocol deviations, and whether interim analyses occurred. No sample-size justification is reported. The posted planning illustrations based on observed secondary effects are not evidence of the study’s original design assumptions. These gaps limit assessment of how fully the analysis followed a prespecified statistical plan and how robust the results are to alternative assumptions.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: coronary artery stenosis (3.3%), myocardial infarction (3.3%), constipation (3.2%), drug-induced liver injury (3.2%), hepatic haemorrhage (3.2%).

The safety results report serious adverse events in each arm and separately report participants affected by other adverse events. For serious adverse events, the counts were 1 in placebo, 1 with PF-06865571 25 mg twice daily, 5 with 75 mg twice daily, 1 with 150 mg twice daily, 5 with the 150 mg combination, 4 with 300 mg twice daily, and 2 with the 300 mg combination. The corresponding numbers at risk were 34, 35, 48, 42, 35, 31, and 30. No deaths were reported in any listed arm.

These are counts of affected participants, not measures of whether treatment caused an event. The registry also lists participants affected by other adverse events by arm, but the aggregate count does not describe event severity, duration, recurrence, or attribution. The record lists serious-event terms including coronary artery stenosis, myocardial infarction, constipation, drug-induced liver injury, hepatic haemorrhage, and acute kidney injury, with percentages. The denominator and arm-specific context for those listed percentages are not reported in the registry record.

In general, safety counts are best interpreted alongside the number at risk, duration of observation, event definitions, and clinical review of each event. The arm counts alone do not establish comparative risk or causation, especially when the groups are small. These figures describe reported events in this trial and do not support medical advice.

13. What the Result Does, and Does Not, Mean

The primary results show estimated week 48 histologic response proportions of 0.38 for placebo and 0.45, 0.48, 0.50, and 0.51 across the PF-06865571-alone doses. The associated reported risk differences versus placebo are positive, but each 90% interval includes zero. Taken together, these estimates describe a pattern in which the model’s point estimates are higher in the active dose groups while uncertainty remains substantial about the pairwise differences.

That pattern does not establish that any dose is superior to placebo, nor does it establish that the groups are equivalent. An interval including zero can be compatible with no difference and with effects on either side of zero. The intervals also overlap, so the numerical ordering of posterior means should not be treated as proof that a higher dose produced a larger response. The registry does not report a primary p-value or a formal conclusion about a monotonic dose trend.

The endpoint is based on central pathology assessment of liver biopsy features. It captures the study’s stated histologic response criteria, not every aspect of NASH or fibrosis, and not a direct measure of symptoms, daily functioning, or long-term clinical outcomes. The response definition permits more than one route to meeting the primary endpoint, while the secondary measures examine particular components separately.

The secondary findings and safety counts provide additional descriptive information, but they do not remove uncertainty in the primary estimates. Multiple outcomes, unadjusted secondary comparisons, incomplete details about missing data, and limited information about the statistical plan all constrain interpretation. The results apply to the populations and analyses reported in the registry; they should not be extended beyond those data or treated as a recommendation about care.

14. Why This Trial Matters Statistically

This trial raises several useful statistical issues for clinical researchers. First, its primary endpoint is a composite histologic response, while secondary endpoints separate NASH resolution from fibrosis improvement. When an endpoint combines several pathways, it is important to inspect component outcomes without assuming that each component behaves like the composite. A coherent interpretation depends on both the endpoint definition and the pattern across related outcomes.

Second, the primary analysis uses a Bayesian dose-response model and reports posterior mean proportions with 90% intervals. This invites a distinction between an estimated group response and a treatment contrast. It also highlights the need to identify the kind of interval being reported: posterior credible intervals and frequentist confidence intervals have different interpretations. The registry’s terminology makes that distinction particularly relevant.

Third, the study includes several doses and combination regimens, but the posted primary estimates listed are for placebo and the PF-06865571-alone groups. This makes it important to distinguish the trial’s full design from the subset of results available for a particular endpoint. A design that includes combination arms does not mean every endpoint’s posted analysis necessarily reports those arms.

Finally, the registry records many outcomes and analyses, while not reporting the multiplicity strategy, detailed missing-data methods, or full primary model specification. The case illustrates why statistical interpretation depends on more than point estimates: uncertainty, analysis population, endpoint construction, and reporting completeness all matter. These are general lessons in reading clinical trial results, not claims that any unreported method was or was not used.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The primary endpoint results listed in the record cover placebo and PF-06865571-alone groups, while the randomized design also includes combination groups. The registry does not report why the combination arms are absent from the posted primary estimates. This limits direct interpretation of the primary endpoint across the full set of treatment strategies in the trial.

The primary population description is truncated after stating that participants had baseline data for the primary endpoint. Full details of the analysis population, endpoint availability, missing-data handling, and participant flow are not reported in the registry record. These details are particularly relevant for a week 48 biopsy endpoint.

The primary endpoint combines distinct histologic response routes. Although separate secondary outcomes address components, the record does not report a decomposition of the primary response or a formal relationship between component outcomes and the composite. The primary model’s prior assumptions and full specification are also not reported, limiting independent reproduction or evaluation of sensitivity to modeling choices.

The registry lists many outcomes and analyses, but multiplicity handling and the prespecified testing hierarchy are not reported. Secondary results based on small or sparse counts can be imprecise, and the simple recalculations do not match the sponsor’s stated logistic regression method. Safety summaries are aggregate counts; the context for listed serious-event percentages and event-level attribution is not reported.

These limitations concern what can be established from the registry record. They do not show that particular analyses or procedures were omitted from the full study conduct; unreported details cannot be inferred.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Metabolic Interventions to Resolve Non-alcoholic Steatohepatitis (NASH) With Fibrosis (MIRNA) — NCT04321031.
• Registry record last updated 2025-03-21; status last verified 2025-03.
• Results first posted 2025-03-21.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

MIRNA was a completed randomized, double-blind, double-dummy, placebo-controlled, parallel-group phase 2 study sponsored by Pfizer. It enrolled 256 adults with biopsy-confirmed NASH and fibrosis stage 2 or 3. Participants received placebo, PF-06865571 at one of four twice-daily doses, or one of two PF-06865571 and PF-05221304 combinations for 48 weeks.

The primary week 48 endpoint was a histology-based composite analyzed with a Bayesian dose-response model. Posted posterior mean proportions for placebo and PF-06865571 alone were 0.38, 0.45, 0.48, 0.50, and 0.51. The reported 90% risk-difference intervals versus placebo include zero for each listed dose comparison. Secondary outcomes address separate histologic components, and safety results report serious adverse events and other adverse events by arm. Multiplicity procedures, detailed missing-data handling, and a complete statistical analysis plan are not reported in the registry record.