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
MAGELLAN-AMD was a randomized, parallel-group, quadruple-masked phase 3 trial designed to show that the proposed biosimilar FYB203 and the reference product Eylea® produce equivalent changes in visual acuity in patients with neovascular age-related macular degeneration (nAMD).
| Feature | MAGELLAN-AMD |
|---|---|
| Phase | Phase 3 |
| Condition | Neovascular age-related macular degeneration |
| Design | Randomized, parallel-group, two arms |
| Masking | Quadruple |
| Interventions | FYB203 (proposed aflibercept biosimilar); Eylea® (aflibercept) |
| Hypothesis | Equivalence |
| Primary endpoint | Change in BCVA (ETDRS letters) from baseline to Week 8 |
| Equivalence margin | ]−3.5 letters; 3.5 letters[ |
| Enrollment | 434 |
| ClinicalTrials.gov | NCT04522167 |
| Lead sponsor | Bioeq GmbH |
2. Clinical Question
A biosimilar trial asks a different question from a conventional efficacy trial. The goal is not to show that FYB203 is better than Eylea®, but that any difference in effect on visual acuity is small enough to be clinically unimportant in either direction.
Population
Patients with neovascular age-related macular degeneration.
Intervention
FYB203, a proposed aflibercept biosimilar.
Comparator
Eylea® (aflibercept), the reference product.
Primary question
Is the difference in mean change in BCVA from baseline to Week 8 between FYB203 and Eylea® contained within ±3.5 ETDRS letters?
3. Trial Design
FYB203
- Proposed aflibercept biosimilar
- Test product in the equivalence comparison
Eylea® (aflibercept)
- Reference product
- Comparator in the equivalence comparison
4. Endpoints
| Role | Endpoint | Registry description | Time frame |
|---|---|---|---|
| Primary | Functional changes in Best Corrected Visual Acuity (BCVA) by ETDRS letters at Week 8 of treatment with FYB203 or Eylea compared to baseline | Change in BCVA by ETDRS letters from the Baseline Visit (Visit 1) to Week 8 (Visit 3): the number of letters a participant could correctly read with the study eye on the ETDRS chart. An increase from baseline signifies improvement. Change = observed post-baseline value minus baseline value. | Week 8 |
BCVA measured in ETDRS letters is a continuous, clinically interpretable scale: each line on the chart contains five letters, so a 3.5-letter margin corresponds to less than one line of visual acuity. A Week 8 primary time point is early in treatment; for biosimilar comparisons an earlier time point is often chosen because it lies on the steep part of the response curve, where differences between products would be most detectable.
5. Analysis Population
The primary analysis was performed in the Full Analysis Set (FAS). The registry does not give a more detailed definition of the FAS on the posted analysis.
6. Primary Endpoint Result: BCVA Change at Week 8
LS mean difference in BCVA change, FYB203 vs Eylea®
90.4% CI: −0.3 to 2.2 (two-sided)
Equivalence margin: ]−3.5; 3.5[ letters · Full Analysis Set
90.4% CI: [−0.3 ─── 1.0 ─── 2.2]
The interval −0.3 to 2.2 lies entirely inside −3.5 to +3.5, which meets the prespecified equivalence criterion stated in the registry.
What the estimate means. The model-adjusted mean change in BCVA from baseline to Week 8 was estimated to be 1.0 letter higher with FYB203 than with Eylea®, in the order the comparison is listed. One letter is a fraction of a single chart line.
What it does not mean. A positive point estimate does not suggest FYB203 is better; the trial was not designed or tested for superiority, and the interval includes zero. Equally, equivalence does not mean the products are identical — it means the plausible difference is within the prespecified range judged clinically unimportant.
Precision. The 90.4% CI of −0.3 to 2.2 letters spans 2.5 letters. Its lower limit is 3.2 letters away from −3.5 and its upper limit is 1.3 letters away from +3.5, so the upper side is where the result sits closest to the margin.
No p-value is posted. In an equivalence trial, the conclusion comes from comparing the confidence interval with the margin, not from a p-value against zero. A non-significant test of "no difference" would not show equivalence, and a significant one would not rule it out; only the position of the interval relative to ±3.5 answers the question.
Cautions. The result is from the FAS, the margin choice drives the conclusion, and the 90.4% confidence level reflects a slightly different alpha allocation from the conventional 90% (its rationale is not described in the registry).
7. Safety: Serious Adverse Events
| Category | FYB203 (affected / at risk) | Eylea® (affected / at risk) |
|---|---|---|
| Systemic serious adverse events | 17 / 215 | 23 / 218 |
| Ocular serious adverse events, study eye | 2 / 215 | 2 / 218 |
| Ocular serious adverse events, fellow eye | 0 / 215 | 3 / 218 |
The registry reports serious adverse events separately for systemic events and for ocular events in the study eye and the fellow eye. Counts were similar or lower in the FYB203 arm. Ocular events were rare in both arms, which means the trial has little ability to detect differences in uncommon but important ocular safety events; biosimilar safety comparisons rely on consistency of the overall profile rather than formal hypothesis tests. No statistical comparison of SAEs is posted.
8. Statistical Methodology
Equivalence hypotheses
An equivalence trial reverses the usual null hypothesis. The null states that the products differ by at least the margin; equivalence is concluded only if the data reject both one-sided possibilities.
H02: Δ ≥ +3.5 vs H12: Δ < +3.5
Δ is the true difference in mean BCVA change (FYB203 minus Eylea®). Rejecting both nulls at level α is equivalent to the two-sided 100(1 − 2α)% confidence interval lying entirely inside the margin.
Why a 90.4% confidence interval?
Two one-sided tests each at α = 0.05 correspond to a two-sided 90% interval. The reported 90.4% level corresponds to each one-sided test being run at α = 0.048, a slight adjustment of the error rate. Adjustments of this size typically arise from spending a small part of the alpha elsewhere in the design; the registry does not state the reason.
Least squares mean difference
The effect is reported as an LS mean difference, meaning it comes from a regression-type model (such as ANCOVA or a repeated-measures model) with covariates. Adjusting for baseline BCVA is standard in this setting because baseline acuity strongly predicts subsequent change. The registry does not list the model covariates.
Choice of the margin
The ±3.5-letter margin is a prespecified clinical judgment: the largest difference in mean BCVA change considered unimportant. Margins in biosimilar trials are usually derived from the size of the reference product's effect and from what clinicians regard as a meaningful change in vision. The registry reports the margin but not its derivation.
9. Statistical Methods Explained
Why is equivalence judged against the margin rather than a p-value?
A conventional p-value tests whether the difference is zero. Failing to reject that null only shows the data are compatible with zero, not that the difference is small. Equivalence requires positive evidence that the true difference lies within ±3.5 letters, which is what the confidence-interval comparison provides.
What does an LS mean difference of 1.0 letter mean?
After model adjustment, the average change in letters read from baseline to Week 8 was 1.0 letter larger in the FYB203 arm than in the Eylea® arm. It is a between-group difference in average change, not the change any patient experienced.
Why is a two-sided margin used?
For a biosimilar, being meaningfully better would be as unexpected as being meaningfully worse; either would suggest the products are not highly similar. The margin therefore bounds the difference on both sides.
Why a 90.4% interval instead of 95%?
Equivalence is two one-sided tests. A two-sided 95% interval would correspond to testing each side at 2.5%, which is stricter than the usual 5%-per-side convention for TOST. The 90.4% level reflects that convention with a small adjustment.
Why is the full analysis set not automatically conservative here?
Non-adherence and protocol deviations tend to pull the two arms together. In superiority testing that biases toward "no difference" and is conservative; in equivalence testing it biases toward the conclusion being sought.
Why measure the primary endpoint at Week 8?
Differences between two versions of an active drug are most detectable while patients are responding rapidly. A later time point, when both arms have plateaued, may be less sensitive to a real difference in activity.
10. Limitations and Interpretation Issues
- Margin dependence: the conclusion of equivalence is only as meaningful as the ±3.5-letter margin, whose derivation is not reported in the registry.
- Single analysis population: only the FAS result is posted; per-protocol consistency is not reported.
- Early time point: Week 8 addresses sensitivity to early response, not long-term comparability.
- Unexplained confidence level: the 90.4% level indicates an alpha adjustment whose rationale is not stated.
- Model details: covariates used in the LS mean model are not listed.
- Rare safety events: ocular serious adverse events were few, limiting comparison of uncommon risks.
- Surrogate of long-term function: an 8-week acuity change does not capture durability, dosing interval effects or anatomic outcomes.
11. Why This Trial Matters Statistically
MAGELLAN-AMD illustrates the logic of biosimilar equivalence trials, where the question, the null hypothesis and the role of the confidence interval all differ from a conventional superiority trial.
| Concept | How it appears in MAGELLAN-AMD |
|---|---|
| Equivalence trial design | Two-sided margin of ±3.5 ETDRS letters |
| Confidence intervals | 90.4% CI −0.3 to 2.2 compared with the margin |
| Two one-sided tests | Equivalence concluded if both one-sided nulls are rejected |
| Intention-to-treat analysis | FAS used for the primary analysis, with the equivalence-specific caveat |
| Adjusted mean difference | LS mean difference of 1.0 letter |
| Sensitive time point | Week 8 primary endpoint |
12. Related Tutorials
Learn more about the methods used in this trial:
13. Related Calculators
14. Sources
- ClinicalTrials.gov registry record: NCT04522167 — Efficacy and Safety of the Aflibercept FYB203 Biosimilar in Comparison to Eylea® in Patients With Neovascular Age-Related Macular Degeneration.
- PubMed: PubMed record 41494750.
Continue exploring clinical trial statistics
Connect this trial's endpoints and methods to in-depth tutorials, then to a calculator for your own analysis.
15. Summary
MAGELLAN-AMD enrolled 434 patients with neovascular AMD and compared FYB203 with Eylea® under an equivalence hypothesis. The LS mean difference in BCVA change at Week 8 was 1.0 letter with a 90.4% CI of −0.3 to 2.2, entirely within the prespecified ±3.5-letter margin. Interpreting the result rests on the margin, the position of the interval relative to it, the analysis population, and the early time point, rather than on any test of a zero difference.