A Study to Evaluate the Efficacy and Safety of Faricimab (RO6867461) in Participants With Diabetic Macular Edema: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Evaluate the Efficacy and Safety of Faricimab (RO6867461) in Participants With Diabetic Macular Edema (NCT03622593), a randomized phase 3 study. The primary outcome is Change From Baseline in BCVA in the Study Eye Averaged Over Weeks 48, 52, and 56, ITT and Treatment-Naive Populations.

2. Clinical Question

In general, a randomized active-comparator trial can estimate how outcomes differ between assigned treatment strategies under the study conditions. The clinically relevant statistical question is not only whether average BCVA changed, but how the estimated changes compare across the randomized groups and how uncertain those estimates are. The registry record describes the outcome and analysis approach, but does not report the primary estimates needed to answer that comparison quantitatively.

3. Trial Design

RHINE is described as a Phase III, multicenter, randomized, active comparator-controlled study in participants with diabetic macular edema. The registry classifies its design as parallel and its masking as triple; the official title describes the study as double-masked. The identity of the people masked under either description is not reported in the registry record, so the difference in wording cannot be resolved from the record.

The three treatment arms compare two faricimab dosing strategies with aflibercept given on a fixed schedule. All regimens begin with more frequent intravitreal injections, followed by the assigned schedule or, for the personalized treatment interval group, dosing adjusted according to the protocol. Treatment continued through Week 96, with the fixed-schedule arms describing a final study visit at Week 100. The final visit timing for the personalized treatment interval arm is not reported in the registry record.

The primary endpoint is BCVA change from baseline averaged over Weeks 48, 52, and 56, with results intended for the ITT and treatment-naive populations. The endpoint timeframe is baseline through Week 56. The registry describes 60 outcome measures and 68 analyses, indicating that multiple outcomes and analyses were registered; it does not provide enough outcome results in this record to reconstruct all of them.

Recruitment began on 2018-10-09, primary completion was recorded as 2020-10-19, and the study status is completed. The sponsor is Hoffmann-La Roche. Although total enrollment is reported as 951, the allocation count for each randomized group is not reported as a design detail in the registry record. The at-risk counts included in the safety summaries are reported separately and should not automatically be treated as the analysis population sizes for every endpoint.

Details such as the randomization procedure, allocation concealment, protocol-specific eligibility criteria, and the rationale for the planned enrollment are not reported in the registry record. These omissions limit the ability to evaluate those design features from this record alone.

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.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: ITT Population and Treatment-Naive Population

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: The overall number of baseline participants (951) comprises the Intent-to-Treat (ITT) Population, which includes all global participants who were randomized in the study, grouped according to the treatment assigned at randomization.

6. Primary Endpoint

The registry lists the primary outcome as Change From Baseline in BCVA in the Study Eye Averaged Over Weeks 48, 52, and 56, ITT and Treatment-Naive Populations, measured over From Baseline through Week 56.

The primary endpoint is change from baseline in BCVA in the study eye, averaged over Weeks 48, 52, and 56, and assessed in the ITT and treatment-naive populations. BCVA was measured using the Early Treatment Diabetic Retinopathy Study (ETDRS) chart at a starting distance of 4 meters. Its letter score ranges from 0 to 100, with 100 identified as the best score. A gain in letters from baseline indicates improved visual acuity as measured by this scale.

Using an average across several scheduled visits makes the endpoint reflect measurements across the specified follow-up period rather than relying on only one of those visits. In general, averaging repeated measurements can reduce the influence of visit-specific variation, although the registry does not report the treatment estimates needed to assess how this endpoint behaved in RHINE. The outcome is a continuous measure, so the principal treatment contrast would ordinarily be expressed as a difference in mean change, with an interval describing uncertainty. Such estimates and intervals are not reported in the registry record.

The analysis used a mixed model for repeated measures (MMRM), adjusted for treatment arm, visit, their interaction, baseline BCVA as a continuous measure, baseline BCVA category (<64 vs. ≥64 letters), prior intravitreal anti-VEGF therapy (yes vs. no), and region of enrollment. The registry describes a 97.5% confidence interval and notes that this is a rounding of a 97.52% confidence interval. The reason for selecting that confidence level is not reported in the registry record.

ETDRS letter scores provide a standardized numerical scale for describing visual acuity change. The endpoint is a measure of visual function on that scale; it is not, by itself, a complete account of every outcome relevant to participants.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome analysis is described as a mixed model for repeated measures (MMRM). In general, this type of model analyzes repeated outcome measurements while accounting for the fact that measurements from the same participant are related. It estimates average outcome patterns over visits and can use observed measurements from participants who do not have every scheduled measurement, subject to the model and missing-data assumptions. The registry states that missing data were implicitly imputed by MMRM, but does not report further diagnostics or sensitivity analyses for those assumptions.

The model included treatment arm, visit, and the visit-by-treatment-arm interaction. In general, including visit allows the mean outcome to vary over time, while the interaction allows the treatment-group pattern to differ across visits. Treatment arm represents the assigned group. The listed model also adjusted for baseline BCVA as a continuous variable, baseline BCVA category (<64 vs. ≥64 letters), prior intravitreal anti-VEGF therapy (yes vs. no), and region of enrollment. The record does not provide coefficient estimates, model fit information, or the exact contrast used to produce each primary comparison.

Baseline adjustment is commonly used to account for starting outcome values and other prespecified characteristics in an outcome model. In general, adjustment can improve precision when the covariates are informative, but the adjusted estimates still depend on model specification and available data. Because the registry does not report the primary estimates or their uncertainty, the contribution of adjustment to the trial's results cannot be evaluated.

The covariance structure was unstructured. In general, a covariance structure describes how repeated measurements at different visits vary together. An unstructured choice allows each modeled variance and covariance to be estimated without imposing a simpler pattern across visits. This flexibility can represent varied relationships among repeated measurements, but it also relies on the data containing enough information to estimate those relationships. The record does not report model convergence, covariance estimates, or any alternative structure analyses.

The endpoint description distinguishes strategies for intercurrent events. A treatment policy strategy was used for non-COVID-19 related intercurrent events, meaning all observed values were used. A hypothetical strategy was used for COVID-19 related intercurrent events, with values censored after the event occurred. In general, these strategies define which outcome data contribute to the estimand after events that may affect interpretation of treatment. A treatment policy approach retains observed post-event data; a hypothetical approach targets an outcome under a specified scenario that excludes data after the event. The registry does not provide event counts for these categories or a fuller description of how the hypothetical values were handled beyond the stated censoring approach and model-based missing-data handling.

The outcome was measured in ETDRS letters, with a score range of 0 to 100 and higher scores described as better visual acuity. Invalid BCVA values were excluded. The record does not define the invalid-value criteria, report how many measurements were excluded, or describe sensitivity analyses for those exclusions.

The reported confidence interval level is 97.5%, which the registry identifies as a rounding of 97.52%. In general, a confidence interval provides a range of values compatible with the estimate under a specified procedure and its assumptions. A confidence interval is not a guarantee that the true effect lies within the particular calculated interval. The numerical bounds are not reported in the registry record, so the interval's width and implications cannot be assessed.

For the posted sample-size illustrations, the stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The illustration is based on the observed effect, with the effect to detect set as a stated fraction of that observed effect. These are planning illustrations, not observed power and not the trial's own design justification. The record does not report the trial's sample-size justification. The primary outcome was not reduced to one comparison because it has several categories or time points; no single comparison-specific planning interpretation can be inferred from that information.

In general, a model-based analysis does not eliminate the need to inspect the data, assumptions, and sensitivity of results to analysis choices. The registry description gives several important specifications, including covariate adjustment, covariance structure, and intercurrent-event strategies. Details such as the handling of protocol deviations, model diagnostics, robustness analyses, and full analysis code are not reported in the registry record.

10. Multiplicity, Missing Data and Analysis Populations

The primary outcome is described for two populations, ITT and treatment-naive, and the registry lists many outcome measures and analyses. In general, conducting multiple statistical tests raises the chance that at least one result may appear notable by chance if all tests are interpreted without accounting for multiplicity. A prespecified hierarchy, adjustment procedure, or other multiplicity strategy can clarify which comparisons support confirmatory claims. The registry record does not report a multiplicity adjustment plan or testing hierarchy, so it is not possible to determine how multiple outcomes and populations were handled for inferential claims.

Missing-data handling is described through the MMRM, with missing values implicitly imputed by the model. In general, a repeated-measures model can use observed outcomes without filling in every missing value as a separate, explicitly generated score. Its interpretation depends on assumptions about how missingness relates to observed and unobserved outcomes. The registry does not report the amount or pattern of missing primary endpoint data, nor does it describe sensitivity analyses addressing alternative missing-data assumptions.

The record specifies different approaches to intercurrent events: observed values were used under a treatment policy strategy for non-COVID-19 related events, while values after COVID-19 related events were censored under a hypothetical strategy. These are distinct estimand choices, not interchangeable ways of describing the same analysis. The counts of participants affected by these events and the resulting impact on estimates are not reported in the registry record.

The ITT and treatment-naive populations are both named, but population sizes, detailed definitions, and results are not reported in the registry record. Accordingly, the available information does not establish how many participants contributed to each endpoint analysis or whether conclusions were consistent across the populations.

11. 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: cellulitis (2.5%), cardiac failure congestive (2.5%), cataract (2.2%), diabetic retinal oedema (2.2%), pneumonia (2.2%).

The adverse-event summaries give, for each arm, the number of participants at risk and the number affected by serious adverse events, other adverse events, and deaths. The serious-event figures are 97 of 317 for faricimab 6 mg Q8W, 82 of 319 for faricimab 6 mg PTI, and 100 of 314 for aflibercept 2 mg Q8W. The other adverse-event figures are 195 of 317, 169 of 319, and 178 of 314, respectively. Deaths are reported as 12, 9, and 10 in the same arm order.

These are descriptive counts. In general, a count alone does not show whether an event was caused by treatment, and a comparison of counts requires attention to group size, follow-up, event definitions, and statistical uncertainty. The registry record does not report adjusted comparisons, confidence intervals, or attribution assessments for these safety summaries.

The registry also lists serious-event terms with percentages: Cellulitis, 2.5%; Cardiac failure congestive, 2.5%; Cataract, 2.2%; Diabetic retinal oedema, 2.2%; Pneumonia, 2.2%; and COVID-19, 2.2%. The record does not assign these listed percentages to specific arms in the information. They should therefore be read as listed terms and values, not as arm-specific rates or evidence of a treatment relationship.

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

The result that can be described from the registry record is the planned and reported structure of the analysis, not the magnitude of the primary treatment effect. RHINE compared two faricimab schedules with an aflibercept schedule, and its primary endpoint was change in BCVA averaged over Weeks 48, 52, and 56. The endpoint was modeled with repeated-measures methods and reported for ITT and treatment-naive populations. Numerical estimates and confidence intervals for that endpoint are not reported in the registry record.

In general, an effect estimate answers how large an observed difference is on the outcome scale; a confidence interval communicates how precisely that estimate is measured under the analysis. A p-value, when reported, addresses compatibility with a specified null model and does not measure clinical importance or the probability that a treatment effect is true. None of these primary numerical results are reported in the record, so the statistical size, precision, and strength of evidence for the comparison cannot be evaluated from it.

The existence of a randomized design does not make missing outcome results available. Nor do the treatment schedules or model specification reveal what the results were. The registry's safety counts provide descriptive information, but do not by themselves establish whether an event was treatment-related or whether risk differed between arms.

These limitations do not establish that there was no difference between groups, and they do not support a claim that one strategy was better or worse. They indicate that the posted registry information is insufficient for a quantitative interpretation of the primary efficacy comparison. The results of secondary outcomes and analyses across visits are also not reported in the registry record, so they cannot fill that gap.

In general, a clinical trial result should be read in light of its estimand, analysis population, missing-data assumptions, and uncertainty. RHINE's record names relevant features of these methods, including the ITT and treatment-naive populations, the handling of intercurrent events, and the MMRM covariance structure. Without the corresponding numerical results and fuller analysis details, the record supports a methodological description rather than a conclusion about comparative efficacy.

13. Why This Trial Matters Statistically

RHINE is statistically informative as an example of a trial where the endpoint and model are described in more detail than the numerical results available in the registry record. The record identifies a repeated-measures primary analysis, baseline and treatment-history covariates, an unstructured covariance structure, and different strategies for non-COVID-19 and COVID-19 related intercurrent events. Each of those decisions affects how an estimated treatment effect would be defined and interpreted.

The repeated-measures approach recognizes that BCVA was assessed over time and that measurements from one participant are related. The primary outcome averages results over Weeks 48, 52, and 56, while the broader outcome listing includes change over time and letter-gain thresholds. These choices illustrate how one clinical trial can address several outcome questions, each with a different scale and interpretation.

The two analysis populations also matter. ITT and treatment-naive results can address different questions about the study population, but comparisons between them require their definitions, sizes, estimates, and uncertainty. Those details are not reported in the registry record. Likewise, multiple outcomes and populations make a prespecified approach to multiplicity important for interpreting a collection of tests; such an approach is not reported.

The safety information illustrates another distinction: descriptive event counts are useful for understanding what was recorded, but they are not substitutes for comparative analyses. Counts, denominators, follow-up, event classification, and uncertainty all contribute to interpretation.

Finally, the registry's planning-illustration assumptions should not be confused with the trial's original design justification or with observed power. The actual sample-size rationale is not reported in the registry record. Taken together, these points show why methods and numerical results must be read together: a model description clarifies the intended analysis, but without estimates and intervals it cannot reveal the observed comparative effect.

14. 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.

A major trial-specific limitation for this statistical reading is the lack of numerical primary outcome results in the registry record. No group estimates, between-group contrasts, confidence interval bounds, or p-values are included in the information. Therefore, the primary outcome's direction, magnitude, and precision cannot be evaluated.

The registry names both ITT and treatment-naive populations but does not report their sizes or detailed definitions. It also does not report population-specific estimates. This prevents assessment of whether the primary findings were similar across those populations and whether the analysis populations align with the intended clinical questions.

The MMRM description is informative but incomplete for independent assessment. Missing data were implicitly imputed by the model, yet the record does not report missing-data amounts or patterns, diagnostic checks, or sensitivity analyses. It describes a hypothetical strategy for COVID-19 related intercurrent events and a treatment policy strategy for non-COVID-19 related events, but does not report the frequency of those events or their influence on the estimates.

The registry says invalid BCVA values were excluded but does not define invalidity or give the number of excluded measurements. Model coefficients, covariance estimates, and model diagnostics are also not reported. The record lists many outcomes and analyses, but a multiplicity strategy is not reported, limiting interpretation of a family of potential statistical findings.

Some design details are also absent. The official title calls the trial double-masked, while the registry's masking classification is triple; the people masked are not identified in the record. The role of the listed sham procedure is not described in the arm information. The sample-size justification and randomization implementation details are not reported in the registry record. These gaps should be kept distinct from the trial's stated methods and should not be filled by assumption.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: A Study to Evaluate the Efficacy and Safety of Faricimab (RO6867461) in Participants With Diabetic Macular Edema — NCT03622593.
• Registry record last updated 2025-07-11; status last verified 2025-07.
• Results first posted 2022-03-22.
• 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.

18. Record Summary

RHINE (NCT03622593) is a completed Phase III, multicenter, randomized, parallel, active comparator-controlled study of faricimab and aflibercept in participants with diabetic macular edema. Total enrollment was 951. The registry records a start date of 2018-10-09 and primary completion on 2020-10-19.

The three described arms were faricimab 6 mg Q8W, faricimab 6 mg with personalized treatment intervals, and aflibercept 2 mg Q8W. The primary endpoint was change from baseline in study-eye BCVA averaged over Weeks 48, 52, and 56, assessed in ITT and treatment-naive populations. BCVA was measured using the ETDRS chart, and the MMRM included listed treatment, visit, baseline, prior-therapy, and region terms.

The registry includes descriptive adverse-event and death counts by arm. Numerical primary efficacy estimates and the trial's sample-size justification are not reported in the registry record. Consequently, the available record supports a description of the design and statistical approach, but not a quantitative conclusion about the primary treatment comparison.