A Study to Evaluate the Efficacy and Safety of Faricimab (RO6867461) in Participants With Diabetic Macular Edema (YOSEMITE): 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 (YOSEMITE) (NCT03622580), 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

3. Trial Design

YOSEMITE is described as a phase III, multicenter, randomized, active comparator-controlled study in participants with diabetic macular edema. The registry record lists a parallel-group model and enrollment of 940. The study status is completed; the recorded start date is 2018-09-05 and the primary completion date is 2020-10-20.

The randomized groups received faricimab on a fixed every-8-week schedule after initial dosing, faricimab on a personalized treatment interval, or aflibercept on an every-8-week schedule after initial dosing. Initial treatment frequency differed between the arms. For the personalized faricimab schedule, the record lists intervals of 4, 8, 12, or 16 weeks after the initial treatment period. The duration of that schedule is not reported in the registry record.

The official title describes the study as double-masked, while the registry masking classification is triple. The record does not explain how these descriptions relate to each other. It also does not report the procedures used to maintain masking or which study roles were masked. The enrollment figure is the registry enrollment, not a statement about how many participants contributed to each particular outcome analysis.

The brief study summary says the study evaluated efficacy, safety, and pharmacokinetics. The listed primary outcome concerns visual acuity, and the record also lists several diabetic retinopathy and visual-acuity secondary outcomes. Numerical primary outcome results are not reported in the registry record. The design and outcome descriptions therefore establish what the comparison was intended to address, but do not by themselves establish the magnitude or precision of any treatment difference.

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 (940) 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 was change from baseline in best corrected visual acuity in the study eye, measured in ETDRS letters and averaged over Weeks 48, 52, and 56. The registry describes the ETDRS score as ranging from 0 to 100, with a gain from baseline indicating an improvement in visual acuity. The score was measured using the ETDRS chart at a starting distance of 4 meters.

Averaging observations over the specified visits makes the endpoint reflect performance across a period of follow-up rather than relying on a single visit. In general, repeated measurements can help describe the course of an outcome and reduce the influence of a fluctuation at an individual measurement occasion. That general rationale does not establish how much averaging changed the estimate in this trial.

The endpoint was analyzed in the intention-to-treat and treatment-naive populations. Its unit is ETDRS letters, and its reported dispersion was a 97.5% confidence interval; the registry notes that this rounds from 97.52%. The registry record does not report the numerical endpoint estimates, confidence limits, or a contrast between treatment groups. The clinical magnitude and precision of the observed differences are therefore not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

How should the repeated-measures model be understood?

The registry describes a mixed model for repeated measures, abbreviated MMRM. In general, this type of model analyzes observations collected from the same participant over multiple visits while accounting for the fact that measurements from the same person may be related. It estimates outcome patterns over time and can compare treatment groups while incorporating the repeated observations rather than treating every visit as independent.

The listed model terms were treatment arm, visit, the interaction between visit and treatment arm, baseline BCVA as a continuous measure, baseline BCVA category (<64 versus ≥64 letters), prior intravitreal anti-VEGF therapy (yes versus no), and region of enrollment. In general, adjustment for baseline measurements and other listed factors can account for differences in those variables when estimating treatment effects. The visit-by-treatment interaction allows the modeled group differences to vary across visits. The registry record does not report the fitted coefficients or numerical adjusted estimates.

What does an unstructured covariance mean?

The record says the MMRM used an unstructured covariance structure. In general, a covariance structure represents the relationships among repeated measurements. An unstructured choice does not impose a simple pattern, such as correlations declining regularly with time; instead, it allows the relationships among the modeled visits to be estimated flexibly. This flexibility can be useful when the pattern of within-participant correlation is not assumed in advance. It can also require the data to support estimation of those relationships. The registry record does not provide model diagnostics or details about whether the chosen structure fit well.

How should the confidence interval be read?

The endpoint’s dispersion is described as a 97.5% confidence interval, which the registry says is a rounding of 97.52%. In general, a confidence interval gives a range of values compatible with the data and the statistical model at the stated confidence level. Its width conveys precision: a wider interval indicates more uncertainty in the estimate than a narrower interval. The interval does not give the probability that a particular treatment effect is true. Because the numerical estimates and interval limits are not reported in the registry record, the precision of the trial’s primary result cannot be assessed.

How were missing observations handled?

The registry says missing data were implicitly imputed by MMRM. In general, this means the model can use the available repeated outcome measurements without filling in every missing value through a separate, explicit replacement step. Such analyses rely on assumptions about the relationship between observed and missing data, together with the model specification. The record does not report sensitivity analyses that would show how conclusions might change under different assumptions about missingness.

The description also distinguishes strategies for intercurrent events. A treatment policy strategy was applied to non-COVID-19-related intercurrent events, using all observed values. A hypothetical strategy was applied to COVID-19-related intercurrent events, censoring values after the event. In general, an intercurrent event is an event after treatment begins that affects how an outcome is interpreted or collected. These strategies define which observations contribute to the estimand and analysis; they are not themselves findings about the outcome.

Invalid BCVA values were excluded. The registry does not describe the criteria for identifying invalid values or report how many observations were excluded. Those details are not reported in the registry record.

What do the planning assumptions mean?

The registry describes sample-size calculations as planning illustrations based on the observed effect, not as the trial’s own design justification. The listed assumptions are a two-sided significance level of 0.05, equal allocation between two arms, a normal approximation, and an effect to detect expressed as a stated fraction of the observed effect. In general, such calculations illustrate the sample size associated with chosen assumptions; they do not measure the probability that the completed trial detected an effect. The numerical planning outputs and the effect fractions are not reported in the registry record. They should not be interpreted as observed power.

The registry distinguishes time-to-event planning calculations, which use a total number of events, from other planning calculations, which use participants per arm. That distinction concerns the calculation scale, not the primary visual-acuity endpoint. The assumptions do not replace a protocol-based account of the original design. The sample-size justification and any protocol-specific planning rationale are not reported in the registry record.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists a primary outcome assessed in the intention-to-treat and treatment-naive populations, along with secondary outcomes covering diabetic retinopathy severity and visual-acuity change or letter-gain thresholds. It also records 60 outcome measures and 68 analyses. These features mean that the record describes multiple analyses, but the counts alone do not reveal which comparisons were confirmatory or how the family of outcomes was organized.

An adjustment for multiplicity can help control the chance of false-positive findings when several hypotheses are tested. Whether such an adjustment was used, and which outcomes or populations it covered, is not reported in the registry record. A prespecified testing hierarchy or other multiplicity strategy is also not reported in the registry record. Therefore, the listed outcomes should not be assumed to share a particular inferential status.

The primary analysis used MMRM, and the record describes how missing values and COVID-19-related and non-COVID-19-related intercurrent events were handled. It does not report sensitivity analyses for missing-data assumptions or numerical results for the primary endpoint. Details of outcome-specific analysis populations and the definitions of those populations are not reported in the registry record.

These absences limit interpretation of both statistical significance and robustness. The modeling description explains the intended analysis framework, but without estimates, interval limits, and multiplicity information, the evidence for any particular comparison cannot be evaluated from the registry record.

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: covid-19 (3.2%), pneumonia (2.6%), osteomyelitis (2.3%), sepsis (2.3%), cardiac failure congestive (2.2%).

The registry lists serious adverse events and other adverse events by treatment group, with the number of participants at risk and the number affected. For serious adverse events, the listed figures are 111 of 313 for faricimab 6 mg Q8W, 117 of 313 for faricimab 6 mg PTI, and 93 of 311 for aflibercept 2 mg Q8W. For other adverse events, the figures are 161 of 313, 163 of 313, and 163 of 311, respectively. These are the registry’s counts; no rates or comparative estimates are reported.

The listed deaths are 16, 21, and 13 in the same arm order. The registry also lists serious adverse-event terms and percentages: COVID-19, 3.2%; pneumonia, 2.6%; osteomyelitis, 2.3%; sepsis, 2.3%; cardiac failure congestive, 2.2%; and myocardial infarction, 1.6%. The denominator and specific analysis basis for these term percentages are not reported in the registry record.

In general, adverse-event counts describe events recorded under the study’s reporting definitions; they do not by themselves establish that treatment caused an event. The registry record does not provide adjusted safety comparisons, exposure-adjusted rates, or detailed event timing in the registry record.

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

The registry record establishes that YOSEMITE was a randomized phase III comparison of faricimab schedules with aflibercept in participants with diabetic macular edema. It identifies change in ETDRS letter score averaged over Weeks 48, 52, and 56 as the primary outcome, and describes an adjusted repeated-measures analysis. Those details clarify the question and the planned statistical approach.

They do not establish the comparative result. Numerical estimates, confidence limits, and a treatment contrast for the primary outcome are not reported in the registry record. It is therefore not possible to determine from these registry entries which group had the larger estimated change, how precise any difference was, or whether the result met a statistical criterion.

The outcome is a measure of visual acuity, not a complete account of every effect relevant to participants. The study summary also refers to safety and pharmacokinetics, while the listed secondary outcomes include diabetic retinopathy severity and other visual-acuity summaries. The registry entries do not provide numerical results for those secondary outcomes or pharmacokinetic findings. Nor do they provide enough information to assess the clinical importance of a primary outcome difference.

Safety counts provide descriptive information about adverse events and deaths as recorded in the registry. They do not, without suitable comparative analyses and context, identify causes or quantify a causal treatment effect. Likewise, the completed status indicates the trial is recorded as completed; it does not resolve the missing numerical efficacy results.

In general, a sound interpretation combines the outcome estimate, its uncertainty, the analysis population, the missing-data approach, and the prespecified testing strategy. Several of those elements are described, but the result estimates and multiplicity approach are not reported in the registry record. Conclusions should therefore remain limited to the design and recorded counts rather than extending to claims of benefit, lack of benefit, equivalence, or safety differences.

13. Why This Trial Matters Statistically

YOSEMITE is statistically notable in the way its primary endpoint combines repeated visual-acuity measurements over a defined follow-up period and uses a repeated-measures model. In general, repeated-measures methods can use the pattern of observations over time while accounting for within-participant correlation. The model description also identifies baseline factors and enrollment region used for adjustment, making the intended analysis more explicit than a simple comparison of unadjusted group averages.

The comparison involves distinct faricimab dosing strategies and an active aflibercept comparator. That structure makes the treatment schedule part of the question: a fixed interval and a personalized interval are both represented. The registry record does not provide outcome estimates that would show how the strategies compared, so the design’s statistical interest should not be mistaken for evidence of a particular result.

The endpoint is assessed in both the intention-to-treat and treatment-naive populations, and the record lists multiple secondary analyses. This creates an important interpretive issue: different populations and outcome summaries can address related but nonidentical questions. The record does not report the multiplicity plan or define all analysis populations, limiting assessment of how these analyses were intended to support inference.

The registry also specifies different approaches to COVID-19-related and other intercurrent events, as well as implicit handling of missing data by MMRM. These choices matter because the estimand and its interpretation depend on what happens to observations after such events. Their presence in the analysis description is informative, while the absence of numerical results and sensitivity analyses limits evaluation of their practical impact.

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.

The main limitation for statistical interpretation is that numerical primary outcome results are not reported in the registry record. Group estimates, a treatment contrast, confidence interval limits, and significance results are absent in the registry record for interpretation. The endpoint’s direction, magnitude, and precision therefore cannot be assessed.

Additional analysis details are also unavailable. Multiplicity procedures, detailed definitions of the intention-to-treat and treatment-naive populations, and sensitivity analyses for missing data are not reported in the registry record. The MMRM description identifies its covariates and covariance structure, but does not provide diagnostics or enough information to assess model fit.

The masking descriptions are not fully aligned: the official title says double-masked, whereas the registry classification says triple. How these descriptions are reconciled is not reported in the registry record. The personalized faricimab arm description ends without specifying the end of its dosing period or final visit, and the role of the listed sham procedure is not reported in the registry record.

The safety listing gives counts and selected serious adverse-event terms, but does not provide detailed event timing, exposure-adjusted comparisons, or the denominator basis for the term percentages. The registry record also does not provide numerical results for the listed secondary outcomes or pharmacokinetic findings. These limitations restrict what can be concluded beyond describing the registered design, endpoint, analysis approach, and adverse-event counts.

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 (YOSEMITE) — NCT03622580.
• 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

NCT03622580 is the registry record for YOSEMITE, a completed phase III randomized, parallel-group study of faricimab and aflibercept in diabetic macular edema. The record gives enrollment as 940, a start date of 2018-09-05, and a primary completion date of 2020-10-20. Its official title describes a multicenter, double-masked, active comparator-controlled study, while the registry masking classification is triple; the record does not explain the difference.

The primary outcome was change from baseline in BCVA in the study eye, averaged over Weeks 48, 52, and 56, assessed in the intention-to-treat and treatment-naive populations. The analysis description specifies MMRM and a 97.5% confidence interval. Numerical primary outcome results and the multiplicity strategy are not reported in the registry record. Safety counts and selected serious adverse-event terms are listed, but they do not provide a causal comparison.