$EYPT, $OCUL, $FDMT, $RGNX: what LUGANO actually cost, and what it did not

What the release says, in the order it says it

EyePoint published topline results from LUGANO at seven in the morning New York time on Monday 17 August 2026, alongside a thirty-slide data presentation filed as an exhibit to its Form 8-K. The headline sentence and the story are not the same thing, and the gap between them is why the stock behaved the way it did.

LUGANO enrolled 432 patients randomised one to one. The baseline tables report 211 patients on DURAVYU 2.7mg dosed every six months and 216 on aflibercept 2mg dosed according to its label, in a mixed population that was 75 per cent treatment-naive and 25 per cent previously treated. The primary endpoint was the mean change in best corrected visual acuity from day one to a blend of weeks 52 and 56, against the control, with a pre-specified non-inferiority margin of 4.5 letters. That endpoint was not achieved on the full dataset. The company states it in the third line of its own headline.

The argument that follows runs like this. Nine patients out of 211 in the DURAVYU arm lost fifteen or more letters of vision from causes other than wet age-related macular degeneration. Remove those nine and the arm gained 3.5 letters against 5.9 for the control, a difference of 2.4 letters, comfortably inside the 4.5-letter margin, with a nominal p-value of 0.0096. In the aflibercept arm, one patient in 216 lost fifteen or more letters, a rate of 0.5 per cent, and none from a non-wet-AMD cause.

Then come the secondary endpoints, and they are strong. Treatment burden fell 42 per cent against a design ceiling of 60 per cent, a superiority result with a nominal p-value below 0.0001: 3.1 injections after the loading phase against 5.3 for the control, or roughly two fewer per patient through week 56. Fifty-four per cent of DURAVYU patients required no supplemental injection at all through week 56, 79 per cent needed zero or one, and 88 per cent needed two or fewer. Through week 32 the supplement-free rate was 76 per cent. Among the 113 patients who never needed a top-up, visual acuity was non-inferior to supplement-free control patients at a nominal p of 0.0035, and central subfield thickness sat three microns apart. Across the whole arm the anatomical difference was four microns at week 56.

The safety profile is clean, and it matters more than a clean safety table usually does. Cataract 4.3 per cent against 5.4, raised intraocular pressure 3.8 against 3.2, intraocular inflammation 0.5 per cent in both arms. No insert migration, no anterior chamber opacities, no free-floating drug particles, no retinal vasculitis, no severe intraocular inflammation. Discontinuation ran at 6 per cent in each arm and none was related to DURAVYU. Sustained-release intravitreal products have died on precisely these findings, and this one did not.

Two dates close the file. LUCIA, the identically designed second pivotal trial with 475 patients, reads out in the fourth quarter of 2026. A new drug application follows in the first half of 2027, and the company writes it as conditional: engage the FDA on the data package, then submit LUGANO and LUCIA together, including pooled data. The full LUGANO dataset with subgroup analyses appears publicly for the first time at the Retina Society annual meeting between 23 and 26 September 2026.

The most important word in the release is “nominal”

Every favourable p-value EyePoint published on 17 August carries the same qualifier. Non-inferiority in the ad hoc analysis: nominal. Superiority on treatment burden: nominal. Non-inferiority among supplement-free patients: nominal. The word appears four times and it is not decoration.

A confirmatory trial allocates a fixed budget of type I error, conventionally five per cent, across its hypotheses in a pre-specified order. The primary endpoint is tested first. If it succeeds, the alpha flows down a testing hierarchy to the secondary endpoints in the sequence agreed with the regulator, and those secondary results can then support label claims. If the primary fails, the hierarchy stops there. Everything below it is descriptive. The arithmetic still runs and the p-value still prints, but it is no longer controlled for multiplicity, and a regulator will not treat it as evidence of effect.

This is why a release can contain a p-value of 0.0001 and a share price can fall seventy per cent on the same morning without either being irrational. A nominal p of 0.0001 on treatment burden says the observed difference in injection counts is very unlikely to be noise. It does not say the FDA will let the company print that difference on a label, because the gate that would have permitted it did not open.

The same logic applies with more force to the exclusion of the nine patients. An analysis that removes subjects from the primary analysis set after the data are unblinded, on the basis of an investigator judgment about causality, is by construction not the pre-specified primary analysis. It can be persuasive. It can be scientifically correct. It is not the test the trial was designed to run, and the difference between those two statements is the entire regulatory question.

Nine patients out of 211, and 258 letters

The number that decided LUGANO fits in a single clinic’s waiting room. Two hundred and eleven patients received DURAVYU. Nine lost fifteen or more letters from causes the company attributes to something other than the disease under study: geographic atrophy in six, glaucoma in two, retinal detachment in one, the last of them also carrying cataract progression and an epiretinal membrane. Between them those nine patients lost 258 letters, a mean of nearly twenty-nine letters each.

Four per cent of an arm is a rounding error in most therapeutic areas. In a non-inferiority trial powered on a continuous letter score it is not, and the arithmetic shows why. The remaining 202 patients averaged a 3.5-letter gain. Put the 258 lost letters back and the whole arm averages about 2.1 letters, against 5.9 for the control. That is a difference of roughly 3.8 letters against a margin of 4.5.

That produces a more precise reading of the miss than either the release or the coverage offered: on the full dataset the point estimate still sits inside the non-inferiority margin. The trial failed on the confidence interval, not on the difference. The reconstruction is approximate, because the trial reports least-squares means from a mixed model and the calculation above uses simple arithmetic, and EyePoint has not published the full-dataset point estimate or its interval. The company can settle the question in one slide at Retina Society in September, and until it does, the size of the miss is the largest undisclosed number in the file.

A second detail sits in the baseline table. Medical history of glaucoma in the study eye ran at 11.4 per cent in the DURAVYU arm against 5.9 per cent in the control, close to double. Two of the nine excluded patients lost vision to glaucoma. Total choroidal neovascular area was also larger at baseline in the treatment arm, 5.69 against 4.60 square millimetres. Randomisation does not guarantee balance on every covariate in a 432-patient trial, and neither imbalance is large enough on its own to explain the outcome. Both are the kind of thing a reviewer will notice.

The question that carries into LUCIA is whether the imbalance was chance or structure. If nine against zero is chance, the second trial will not reproduce it. If something about the insert procedure, the protocol or the enrolled population generates non-disease vision loss at a rate the control does not, LUCIA will reproduce it, and a pattern that looked like misfortune becomes a finding.

What the presentation rules out, and what it leaves open

The data deck filed with the 8-K does something topline decks rarely do: it goes looking for the explanations that would be worst for the company and reports what it found. Three are addressed directly.

Did adverse events drive the result? The company says no, and the table supports it on the measures that would matter most. Cataract, raised intraocular pressure, intraocular inflammation and retinal detachment all ran at effectively identical rates between arms. The single retinal detachment patient in the DURAVYU arm lost forty letters against one gained in the control. That is one patient in each arm, and one patient carries no weight on its own.

The denominators move between tables. The company states 432 patients enrolled, the baseline table reports 211 and 216 across the two arms, which is 427, and the adverse event tables use 221 for the control. The gaps are small and probably reflect a safety population and an efficacy population that differ from total enrolment, which is routine. The company does not say so anywhere in either document, and percentages calculated on different denominators are not the same percentages.

Did the supplementation criteria fail? Also no, and this is the strongest piece of evidence the company has. Among the 92 supplemented DURAVYU patients outside the excluded cohort, visual acuity was flat across the supplement visit, plus 0.1 letters. Among the six excluded patients who received supplemental aflibercept, acuity fell 6.4 letters and did not recover. Rescue worked when the problem was wet AMD and did nothing when the problem was something else, which is close to a definition of the cohort being genuinely different.

Did geographic atrophy drive it? The company reports no trend of worsening macular atrophy on fundus autofluorescence against the control between baseline and week 56, on 180 and 168 patients respectively. Six of the nine lost vision to geographic atrophy, so the arm-level absence of a trend and the individual outcomes have to be held together rather than one substituting for the other.

What the deck leaves open is narrower and more awkward. Neovascular age-related macular degeneration appears as an adverse event in 7.1 per cent of the DURAVYU arm against 1.4 per cent of the control. Retinal haemorrhage runs 4.7 against 1.8 per cent, retinal oedema 2.8 against 0.9, and visual impairment 2.4 against zero. Conjunctival haemorrhage at 11.4 against 5.0 per cent and vitreous floaters at 9.0 against 3.6 are procedural and expected from an insert. The disease-activity terms are not procedural, and they sit uncomfortably beside an anatomical difference of four microns. Both readings can be true, because an adverse-event term records a physician’s judgment of an event while central subfield thickness records a machine measurement of a structure, and the two do not have to agree. Reconciling them is a subgroup question, and September is where it gets answered.

The overperforming control: does the claim hold up?

EyePoint’s second line of defence is that the comparator was unusually good, and it makes the case with a table rather than an assertion. Across seven prior registration trials with an aflibercept 2mg control arm, the proportion of patients losing fifteen or more letters ran from 2.7 per cent in LUCERNE to 5.9 per cent in TENAYA, with VIEW 1 at 4.9, VIEW 2 at 4.4, HAWK at 5.5, HARRIER at 4.8 and PULSAR at 3.3. EyePoint’s own Phase 2 DAVIO 2 produced 7.7 per cent in a small control arm. LUGANO produced 0.5 per cent.

The benchmark holds. A control arm that delivers one severe vision loss in 216 patients is running at roughly a fifth of the lowest rate any comparable trial has published, and well below a decade and a half of registration data. Sponsors routinely make claims of this kind that collapse on contact with the literature. This one does not, and the sources are the published record rather than internal data.

What the check does not do is repair the trial. A comparator performing at the favourable tail of its historical distribution is not a protocol violation, a randomisation failure or a data integrity problem. It is variance, and a non-inferiority design accepts in advance that the control will land somewhere inside a distribution. There is no regulatory mechanism that discounts a control arm for having done too well, and there should not be, because the mechanism that adjusted for a strong control could be used to adjust for a weak one.

The two defences also sit awkwardly together. If the control arm is unusually good and the treatment arm is unusually unlucky, the sponsor is asking a regulator to adjust both tails of the same comparison in the same direction. Each adjustment is individually arguable. Requesting both at once is what makes a topline release read as a defence rather than as a result, and it is the reason a company can publish a technically accurate document and still lose seventy per cent of its market value by lunchtime.

What LUGANO did prove, and it is not nothing

Strip out the endpoint dispute and a different trial appears underneath, one that answered the question this product exists to answer.

DURAVYU is not sold on letters. Aflibercept already delivers the letters. The 13 August map of this field set out the reason durability matters at all: registry data puts real-world delivery at between 4.3 and 6.4 injections in the first year against a protocol ceiling of thirteen, and the vision lost in routine practice is lost to appointments that do not happen rather than to a drug that does not work. A six-month depot is a bet that removing several visits a year from a patient’s life is worth more than a marginal difference in acuity measured inside a trial where attendance is enforced.

On that measure LUGANO delivered. Fifty-four per cent of patients crossed the full 56 weeks without a single supplemental injection. Seventy-nine per cent needed at most one, 88 per cent at most two, and the average patient took 1.1 supplemental injections across the year. Total injections after the loading phase were 3.1 against 5.3 for on-label aflibercept, a 42 per cent reduction against a theoretical ceiling of 60 per cent, meaning the product captured roughly seven tenths of the durability available to it by design.

The anatomical result is the part that should travel furthest. Four microns of central subfield thickness separated the arms at week 56 across the whole population, three microns among the supplement-free patients. That is not a product limping along on a technicality. It is a depot holding the retina dry at six-month intervals against an injection given every eight weeks, and it is the finding that makes the nine-patient explanation plausible rather than convenient: all nine had good anatomic control of their wet AMD while losing vision to something else.

The two documents also disagree with each other. The press release states that 94 per cent of DURAVYU patients received zero or one supplement through week 32. The data presentation filed the same morning states 96 per cent. The two documents were published within minutes of each other and disagree by two points on a secondary endpoint, which is small, checkable and the sort of thing that ought to be right.

The synthesis is uncomfortable. DURAVYU appears to work as a durability product and failed as a non-inferiority product, and it failed on a measure the durability case never depended on. Whether that distinction survives contact with a regulator is what the next two quarters decide.

One trial down, one to go, and what a filing actually needs

The conventional standard for approval is two adequate and well-controlled trials. That is why LUGANO and LUCIA exist as an identical pair rather than as one larger study. Replication is the point, and EyePoint has been explicit that the plan is to submit the two together, including pooled data.

Pooling is the tell. A pooled analysis across two identically designed trials is a legitimate and common component of a wet AMD submission, and it is also the mechanism by which a narrow miss in one trial can be absorbed by a clear win in the other. It works when the second trial is strong. It does not work when the second trial is marginal, because pooling a miss with a near-miss produces a pooled near-miss. Three broad shapes follow.

LUCIA hits the primary endpoint on its full dataset. This keeps the programme alive in something close to its original form. A sponsor with one clean win and one narrow miss attributable to an identified imbalance has a case to make, the FDA has approved products on one positive pivotal trial plus confirmatory evidence before, and LUCIA is larger at 475 patients. The label negotiation would be harder than two wins would have made it, and the durability claims carrying the commercial case would be argued rather than granted. But there would be a filing.

LUCIA misses in the same way. Two narrow misses with consistent secondary endpoints is a worse conversation, because a pattern that looked like misfortune in one trial looks systematic in two. The programme would not necessarily end, but the path would move from a 2027 filing to a discussion about what additional trial is required, and that discussion happens against the balance sheet described below.

LUCIA misses for a different reason. The least discussed and the worst. Two trials failing in two unrelated ways removes the explanation that holds the LUGANO story together, and there is no version of that outcome in which the nine patients still matter.

One further item belongs in the filing calculus and appears in EyePoint’s own risk factors rather than in any headline: an FDA warning letter concerning the company’s Watertown, Massachusetts facility. A new drug application is a manufacturing submission as much as a clinical one, and unresolved facility findings are a known source of delay independent of how the data reads. The company also runs a commercial manufacturing site in Northbridge that it has been scaling through 2026 and describes as in place for a US launch.

The balance sheet is now the story

The 13 August analysis named the financing sequence as the specific structural risk in front of EyePoint, on the reasoning that the terms of the company’s next capital raise would be set by LUGANO rather than chosen by EyePoint. That is what happened.

At 30 June 2026 EyePoint held $110.5 million in cash and equivalents plus $70.0 million in marketable securities, $180.5 million in total, against $306.1 million at the end of 2025. The second-quarter net loss was $94.5 million, of which research and development was $83.6 million. Revenue was $0.5 million. There is no debt. Management guides to funding operations into the fourth quarter of 2027, on the basis that Phase 3 enrolment costs roll off.

There were 86,212,138 shares outstanding as of 31 July 2026, against 300 million authorised. At Friday’s close of $14.75 that made the company worth roughly $1.27 billion. At $4.48 on Monday morning it is worth roughly $386 million, of which $180.5 million was cash six weeks earlier. The market is assigning something in the region of $200 million to a Phase 3 asset in the two largest retinal indications, a commercial manufacturing facility, and a diabetic macular edema programme that enrolled more than 480 patients in five months and reads out in the fourth quarter of 2027.

The dilution arithmetic is the part that compounds. EyePoint sold 1,429,047 shares through its at-the-market facility in the first half of 2026 at an average of $14.16, and a further 1,208,718 shares in July at $14.44. Raising $200 million at $14.44 would have required about 13.9 million shares, sixteen per cent of the count. Raising the same $200 million at $4.48 requires about 44.6 million shares, fifty-two per cent of the count. That illustration is arithmetic rather than a forecast, and the company has announced no raise, but it describes the mechanism by which a clinical result becomes a capital structure.

The runway guidance now rests on assumptions written before the readout. It was issued on 5 August, twelve days before LUGANO, on a plan that presumed a particular sequence of events afterwards. Companies do not usually restate runway on the morning of a topline, and the number that answers the question is not the guidance but the next 10-Q.

The read-across that did not happen

The 13 August analysis of this field laid out four configurations for how the durability race would look after LUGANO and the gene therapy readouts. One of them described a LUGANO miss, and it said this: the tyrosine kinase inhibitor depot approach would absorb damage as a class, including a product with a positive Phase 3 already in hand, because the market would read a failure as a statement about sustained intracellular VEGF blockade rather than about vorolanib.

Half of that was right and half of it was wrong, and the wrong half is the more interesting one.

The capital rotation happened exactly as described. Attention and money moved toward the one-time therapies, with 4D Molecular up roughly twenty-one per cent on the day. What did not happen is the class damage. Ocular Therapeutix, the other six-month tyrosine kinase inhibitor implant, rose. Kodiak rose. REGENXBIO rose. On the morning its closest analogue lost seventy per cent, the company running the same therapeutic concept with a different molecule was worth more than it had been on Friday.

The explanation is in the secondary endpoints, and it is why the distinction between a mechanism failure and an endpoint failure is not pedantry. If DURAVYU had failed to hold the retina dry, or if patients had needed rescue injections at month three, the read-across would have been immediate and correct: sustained intracellular VEGF blockade at a six-month interval would have been shown not to work, and every implant in the field would have carried the same wound. That is not what the data showed. Seventy-nine per cent of patients on one implant plus at most one top-up, and three to four microns of anatomical difference, is the class thesis working. The trial failed on a mean letter count distorted by nine patients, which is a statement about this trial, not about this class.

Markets are frequently accused of reading topline releases at the level of the headline. On 17 August this one read past the headline within a morning, and the price action across five tickers is the evidence.

The second lesson concerns the other three configurations. All four scenarios in that piece assumed a binary: LUGANO succeeds or LUGANO misses. What arrived is neither, and the shape that arrived is the most common shape in late-stage clinical development. A primary endpoint missed by a margin small enough to be attributed to identifiable patients, with secondary endpoints that support the commercial thesis, is not a fifth branch nobody considered. It is the branch that scenario frameworks systematically underweight because it does not resolve anything.

What changes for everyone else

Ocular Therapeutix ($OCUL)

The nearest competitor gains on two fronts and loses on one. It gains time: the company with the fully enrolled second Phase 3 and the earlier readout has just used its shot, and whatever DURAVYU’s label eventually says, it will be argued rather than assumed. It gains a comparison: AXPAXLI’s SOL-1 met its endpoint, and a field in which one implant hit and the other missed is a better field to be the one that hit.

What it does not gain is a change in its own risk profile, and the risks named in the 13 August analysis are unchanged by anything that happened on Monday. SOL-1 measured maintenance of vision in treatment-naive patients. SOL-R, which addresses repeat dosing and includes an aflibercept 8mg comparator, does not read out until the first quarter of 2028. The NDA submission is guided to the fourth quarter of 2026. Ocular Therapeutix closed the second quarter with $598.6 million in cash, more than EyePoint, Kodiak and REGENXBIO combined, and it has a marketed product in DEXTENZA generating $13.5 million in the quarter.

There is one indirect effect that cuts against it. Regulators reading two Phase 3 programmes in the same class within a year of each other will read them together. A reviewer who has just examined a narrowly missed non-inferiority trial in a mixed population may bring sharper questions to an application built on a maintenance endpoint in naive patients.

REGENXBIO ($RGNX) and AbbVie

ATMOSPHERE and ASCENT read out in the fourth quarter of 2026, the same quarter as LUCIA. The strategic position improves for a straightforward reason: the argument that a six-month depot is good enough to make one-time therapy unnecessary just lost some of its force. If ABBV-RGX-314 delivers, the question in retinal durability moves from how many months to how many administrations.

4D Molecular Therapeutics ($FDMT)

The largest one-day gainer, and the company with the most distant catalyst. 4FRONT-1 completed enrolment of more than 500 patients and reads out in the first half of 2027; 4FRONT-2 follows in the second half. The company held $457.6 million at the end of March 2026 and takes the intravitreal route, avoiding the surgical complexity of subretinal delivery. A twenty-one per cent move on someone else’s data is a repricing of narrative rather than of evidence, and it should be read as such.

Kodiak Sciences ($KOD)

DAYBREAK’s one-year topline is guided to the third quarter of 2026, which means it may arrive before LUCIA. Kodiak held $169.5 million at the end of March 2026 and disclosed substantial doubt about its ability to continue as a going concern. It is the clearest live illustration of the fifth question below.

The incumbents

Regeneron and Roche are the quiet beneficiaries of every delay in this field. EYLEA HD passed original EYLEA in US sales during the second quarter of 2026, growing 51.6 per cent year on year, and every quarter in which no six-month depot reaches the market is a quarter in which interval extension by the incumbents remains the only durability product a physician can actually prescribe.

The calendar from here

LUGANO removed one entry from the schedule and raised the weight of every entry that remains. Three of the events below land inside the fourth quarter of 2026.

Dates for trials that have not read out are company guidance or registry estimates, not commitments. Several entries on this table have moved once already.

The five questions, applied to LUGANO

The 13 August analysis proposed five questions to separate a durability readout that changes a market from one that produces a headline. LUGANO is the first result to run through them, and it scores unevenly enough to be a useful worked example.

One: what was the comparator, and was it dosed the way a good clinic doses it? On-label aflibercept 2mg. This is the honest comparator and the harder one, and EyePoint chose it over a fixed-interval strawman, which was the correct call. It is also, on this occasion, the comparator that beat the company, because it performed at the top of its historical distribution. Choosing the difficult control is the correct decision and it carries a cost.

Two: superiority or non-inferiority, and against what margin? Non-inferiority on mean letter change at weeks 52 and 56, against a margin of 4.5 letters. The margin is disclosed, in the slides rather than the release. What is disclosed nowhere is the point estimate on the full dataset and its confidence interval. Those are the two numbers that separate a trial missed by a hair from a trial missed by a mile, and until they appear at Retina Society in September nobody outside the company can size the gap precisely.

Three: how many patients needed rescue, and when? This is where LUGANO performed best. Fifty-four per cent supplement-free through week 56, 79 per cent on zero or one supplement, 76 per cent supplement-free through week 32. The formulation there was that a product meeting its primary endpoint while a third of patients receive rescue injections before month six has not delivered a six-month product. LUGANO produced the mirror image: a product that missed its primary endpoint while delivering the six-month interval it was built to deliver.

Four: naive, previously treated, or both? Both, which is the population a commercial product actually meets and the harder test. The release does not break results down by prior treatment status. That breakdown is the single most valuable item likely to appear in September, because a product that holds up in switch patients and stumbles in naive patients, or the reverse, is a different commercial proposition from the pooled number.

Five: what does the balance sheet look like on the day the data lands? $180.5 million of cash at the last balance sheet date against a $94.5 million quarterly loss, no debt, no product revenue, and a share price down roughly seventy per cent on the morning. The answer given there was Clearside Biomedical, which had positive Phase 2b data, FDA alignment on a Phase 3 design and $6.8 million, and is now an asset in a bankruptcy auction. EyePoint is nowhere near that position. It is, however, now in the position where the answer to question five interacts with the answer to question two, and that interaction is the whole of the next two quarters.

How the field looks from here

Four configurations follow from LUCIA and the fourth-quarter gene therapy readouts. They are descriptions of consequence, not predictions of probability, and none of them is a view on any security.

LUCIA hits and the gene therapy readouts hit. EyePoint files in 2027 on a mixed package and negotiates a label in which the durability claims are the argument. The market fragments by patient rather than by product, as it would have under the original positive case, but EyePoint enters it having spent its balance sheet cushion and its credibility premium. Ocular Therapeutix arrives at the same market with more cash and a cleaner file.

LUCIA hits and gene therapy disappoints. The best available outcome for the implant class and the one in which the six-month depot becomes the destination rather than the waypoint. It is also the outcome in which the LUGANO miss costs EyePoint the most in relative terms, because two implants would then be competing for the same category and one of them would carry an asterisk into every payer conversation.

LUCIA misses and gene therapy hits. The configuration in which the read-across that did not happen on 17 August happens after all. Two missed pivotal trials in the same programme is a statement about the molecule and, at that point, plausibly about the class. Capital moves to the one-time therapies and stays there.

Both disappoint. The incumbents keep the market by default. EYLEA HD and VABYSMO continue extending intervals incrementally, and the treatment gap documented in the registry literature stays open, which is the outcome that is quietest commercially and worst clinically.

Across all four, the fact that did not change on 17 August is the one the 13 August analysis ended on: Ocular Therapeutix reaches every branch with more cash than anyone else in the challenger set, and EyePoint reaches every branch having spent more to get there while holding less. What changed is that EyePoint now reaches them holding less than it did on Friday, in a currency worth a third of what it was worth on Friday.

The bottom line

LUGANO produced a product that appears to do what a six-month implant is supposed to do and a trial that did not prove what a registration trial is supposed to prove. Both statements are true, and the market spent Monday morning deciding which of them it was pricing. It priced the second one for EyePoint and the first one for everybody else in the class, which is a more discriminating response than a topline miss usually receives.

Everything now runs through LUCIA in the fourth quarter, and the September subgroup data at Retina Society is the first place where the nine patients acquire names, diagnoses and a distribution. Until then the sizeable unknown is not whether DURAVYU is durable, which LUGANO answered, but by how much it missed, which LUGANO did not disclose.