Analysis: an ad hoc exclusion is not the endpoint

The distinction that matters here is between what was pre-specified and what was not. The primary endpoint was pre-specified, and it was not met. The analysis showing non-inferiority at a nominal p-value of 0.0096 is described by the company itself as ad hoc, meaning the exclusion of the nine patients was decided after the data were seen. Regulators treat those two categories very differently, and the company’s own p-values are labelled nominal throughout, which signals they are not adjusted for multiplicity. The secondary endpoints on treatment burden and supplement-free rates were pre-specified and are the stronger part of the release, but under a non-inferiority design built on visual acuity they support a label argument rather than replace the endpoint.

The company’s explanation is testable in principle. If roughly 3 to 5 percent of aflibercept patients lost 15 letters or more in comparable prior trials and none did so in LUGANO’s control arm, the control arm was unusually favorable. That is a real observation about a single trial, not a defect in DURAVYU. It is also unverifiable from a topline release: the subgroup analyses promised for the Retina Society meeting are where a reader can see whether the nine excluded patients differ systematically from the rest. Until then the record is that the pre-specified primary endpoint was not met in the full dataset and that the topline release does not contain the patient-level detail a reader would need to test the explanation.

LUCIA is now the whole program. It is described as an identical trial, which cuts both ways: identical design means the same control arm behavior is unlikely to repeat by construction, but it also means a second miss would not be explicable the same way. The registry lists LUGANO’s estimated primary completion as August 2026 and completion as August 2027, so the two-year data is still ahead even for the trial that has read out. A filing in the first half of 2027 on the strength of one clean trial and one missed primary endpoint is a decision that will rest on how the agency views the supplement-free and treatment-burden data.

The financial clock is the other constraint. EyePoint reported research and development expenses of $83,614 thousand in the three months to June 30, 2026 against $55,498 thousand a year earlier, with direct DURAVYU spending of $57,985 thousand against $37,349 thousand and personnel costs of $20,832 thousand against $12,703 thousand. Total operating expenses were $97,887 thousand and the net loss was $94,470 thousand. Against that, cash, cash equivalents and investments in marketable securities were $180.5 million at June 30, 2026, which the company says will fund current and planned operations for at least the next twelve months. It also states it expects to continue funding through licensing transactions, further equity raises and other arrangements.

On the disclosed figures, quarterly operating expenses of $97,887 thousand run against $180.5 million of cash, cash equivalents and investments, and the LUCIA topline is due in the fourth quarter of 2026. Separately, the filings note a warning letter received from the FDA in July 2024 concerning YUTIQ manufacturing at the Watertown facility, following a February 2024 inspection, and the company’s implementation of corrective and preventive actions. That is a manufacturing matter rather than a clinical one, but it sits on the same site and the same approval pathway.

What the documents say

EyePoint, Inc. (Nasdaq: EYPT) reported on August 17, 2026 that LUGANO, the first of two pivotal Phase 3 trials of DURAVYU 2.7mg in wet age-related macular degeneration, did not achieve its primary endpoint in the full dataset. The endpoint was non-inferiority in the average change in best corrected visual acuity at weeks 52 and 56 against a 2 mg aflibercept on-label control. The Watertown, Massachusetts company furnished the announcement and a slide presentation as exhibits to a Form 8-K.

What the company reported

EyePoint attributes the miss to what it calls an asymmetric cohort. It says 9 of 211 patients, 4 percent, experienced vision loss of 15 letters or more unrelated to wet AMD, while no patients in the aflibercept arm did. In an ad hoc analysis excluding that cohort, DURAVYU was non-inferior to the control with a nominal p-value of 0.0096. The company also argues that the control arm overperformed, citing prior pivotal trials at similar scale in which approximately 3 to 5 percent of aflibercept patients lost 15 letters or more, and referencing the VIEW, HAWK and HARRIER, TENAYA and LUCERNE datasets and an FDA statistical review.

On the secondary endpoints the company reports a 42 percent reduction in treatment burden against a maximum possible reduction of 60 percent, with a nominal p-value below 0.0001, which it says translates to two fewer injections on average up to Week 56. It reports that 76 percent of DURAVYU patients were supplement-free up to Week 32 and 94 percent received zero or one supplement over that period, falling to 54 percent supplement-free and 79 percent with zero or one supplement up to Week 56. A pre-specified analysis of change in best corrected visual acuity in supplement-free patients showed non-inferiority against supplement-free aflibercept with a nominal p-value of 0.0035. On anatomy, the company reports a mean difference of 4 microns in central subfield thickness against the control at Week 56, and 3 microns among the supplement-free group.

On safety, EyePoint says DURAVYU was observed to be safe and well tolerated with repeat dosing, with no difference in cataracts, raised intraocular pressure or intraocular inflammation against control, and no observed insert migration, anterior chamber opacities, free-floating drug particles, retinal vasculitis or severe intraocular inflammation.

Chief executive Jay S. Duker said the primary endpoint result for the full dataset was unexpected, and that the pre-specified secondary results and the ad hoc primary analysis present a case for DURAVYU as a new potential option. The company plans to present subgroup analyses starting at the Retina Society 59th Annual Scientific Meeting from September 23 to 26, 2026.

The trial design and what comes next

The registry entry for LUGANO, NCT06668064, describes a two-year Phase 3 study of EYP-1901 compared with aflibercept in wet AMD, randomized, parallel-group, with quadruple masking covering participant, care provider, investigator and outcomes assessor. It records an estimated enrollment of 400, a start date of October 22, 2024, an estimated primary completion in August 2026 and estimated completion in August 2027, with the study listed as active and not recruiting. The comparator arm is aflibercept at 2.0 mg. The registered primary outcome is the average change in best corrected visual acuity; registered secondary outcomes include the rate of injection burden.

EyePoint says LUGANO and LUCIA, NCT06683742, are identical trials with over 900 patients enrolled across both, randomized 1:1 to DURAVYU every six months or on-label aflibercept, and that all active patients in the treatment arm have reached the Week 32 visit and received a second dose. LUCIA topline data is expected in the fourth quarter of 2026, with a possible New Drug Application in the first half of 2027. Two Phase 3 trials in diabetic macular edema, COMO and CAPRI, are fully enrolled with topline data anticipated in the fourth quarter of 2027.