Wet AMD durability race 2026: $OCUL, $EYPT, $RGNX, $FDMT and the readouts that settle it

The problem nobody has solved

Anti-VEGF therapy is one of the genuine triumphs of modern medicine. Before ranibizumab, neovascular age-related macular degeneration meant a near-certain slide into legal blindness within two years. After it, the average patient in a registration trial gained letters. Not held steady, gained. Two decades later the drugs are better still, and the disease is, in the controlled setting of a clinical trial, close to managed.

The controlled setting of a clinical trial is where the achievement ends.

In the pivotal studies, patients received an injection into the eye on a fixed monthly schedule, chased by coordinators, transported when they could not drive, and dropped from the analysis if they failed to show up. Outside those studies, in ordinary retina practice, none of that happens. An analysis of 459,237 Medicare beneficiaries found an average of 4.3 injections in the first year of treatment. An IRIS Registry analysis of 13,859 eyes found 6.4 injections for ranibizumab, 6.2 for aflibercept, 5.9 for bevacizumab. A protocol that permits thirteen doses in fifty-two weeks is being delivered at roughly a third to a half of that rate.

The consequences are measurable and they are not small. In the same Medicare cohort, 57 per cent of patients had discontinued treatment by twelve months and 71 per cent by twenty-four. Among patients who received seven or more injections in year one, only 31 per cent maintained a comparable level in year two. Registry work from Fight Retinal Blindness! quantifies the price: eyes with a treatment gap of six to twelve months lost 4.9 letters, and eyes stretched beyond a twelve-week interval lost 4.7 letters against a gain of 0.4 letters for eyes kept at six weeks or under, with roughly triple the risk of losing fifteen letters or more.

This is the entire commercial thesis of every company in the race. Nobody is trying to beat aflibercept on efficacy in a trial. They are trying to beat aflibercept in a world where the patient is eighty-four, does not drive, depends on a daughter who works, and simply stops coming.

Wave one already won, and it was not an implant

The challengers cannot be assessed without an honest account of the incumbents, because the most common analytical error in this space is to model a market that no longer exists.

Regeneron’s EYLEA HD, the 8mg aflibercept formulation, generated $596.3 million in US net sales in the second quarter of 2026 against $393.2 million a year earlier, growth of 51.6 per cent. Over the same period original 2mg EYLEA fell from $754.3 million to $412.2 million, a decline of 45.3 per cent. The high-dose product has passed the product it was designed to replace. Combined US franchise sales were $1.01 billion for the quarter, down 12.1 per cent year on year, which tells its own story about biosimilar and competitive pressure, but the internal transition is unmistakable.

Roche’s VABYSMO, the bispecific targeting both VEGF-A and angiopoietin-2, recorded CHF 2.06 billion in first-half 2026 sales, up 8 per cent at constant exchange rates, with international markets growing 65 per cent and Japan 26 per cent.

What both products sell is not superior vision. It is interval. EYLEA HD and VABYSMO are approved for dosing schedules that stretch to twelve and sixteen weeks in appropriately selected patients. Every quarter that these products push the practical treatment interval outward, the incremental value of a six-month implant narrows.

The implant companies have an answer, and it is a reasonable one: an approved sixteen-week interval and a delivered sixteen-week interval are different objects. The registry data above shows exactly how different. A depot that physically cannot be skipped removes the adherence variable rather than negotiating with it. Whether payers, physicians and patients agree that this distinction is worth a new procedure and a new reimbursement pathway is the question the next two years will answer.

Wave two: the six-month implant, and two companies that need it to work

Two companies have taken the same insight to Phase 3 with the same class of molecule. Both deliver a tyrosine kinase inhibitor, a small molecule that blocks VEGF receptor signalling inside the cell rather than mopping up the ligand outside it, from a bioerodible depot placed in the eye. Both are aiming at a six-month redosing interval. Neither has a marketed product of any consequence to fall back on.

What SOL-1 established, and what it left open

Ocular Therapeutix’s AXPAXLI is an intravitreal implant releasing axitinib. Its Phase 3 SOL-1 study, NCT06223958, randomised 344 treatment-naive patients to a single 0.45mg AXPAXLI implant or to aflibercept 2mg, and it was designed as a superiority trial rather than the customary non-inferiority design.

It met that bar. The company reported that 74.1 per cent of patients in the AXPAXLI arm maintained vision at week 36, defined as losing fewer than fifteen ETDRS letters, a risk difference of 17.5 per cent with p=0.0006. The week 52 secondary endpoint gave 65.9 per cent maintenance and a risk difference of 21.1 per cent with p<0.0001. Ocular Therapeutix has since stated FDA alignment on filing a new drug application on the strength of the week 52 data, with submission planned for the fourth quarter of 2026.

Two features of that design carry more weight than the headline percentages. The endpoint is maintenance of vision, not mean change in visual acuity, and the comparison ran against aflibercept 2mg on a schedule that permitted the control arm to be undertreated relative to modern practice. A trial that asks whether patients avoid catastrophic loss is asking something real and something clinically defensible. It is not asking the question that EYLEA HD and VABYSMO answer, which is how many letters a patient gains and holds.

The second Ocular Therapeutix study, SOL-R, NCT06495918, is the repeat-dosing trial and the one that speaks to the commercial proposition. It randomised 555 patients across AXPAXLI redosing, aflibercept 2mg and aflibercept 8mg, on a non-inferiority design with a 4.5 letter margin. Its topline has moved twice and now sits in the first quarter of 2028, with masking held to week 96 to support additional secondary endpoints. That timeline matters: the NDA, if it proceeds on SOL-1, would reach the agency well before the repeat-dosing evidence is public.

What LUGANO is designed to prove

EyePoint’s DURAVYU delivers vorolanib, a different tyrosine kinase inhibitor, from the company’s bioerodible Durasert E insert. The Phase 3 programme is two identically designed studies, LUGANO (NCT06668064) and LUCIA (NCT06683742), together enrolling more than 900 patients, each randomising one to one against on-label aflibercept.

The design differs from SOL-1 in three ways that matter. The primary endpoint is average change in best corrected visual acuity across weeks 52 and 56, the conventional measure. The design is non-inferiority, the conventional bar. And the population includes both treatment-naive and previously treated patients, which is the population a commercial product actually meets.

LUGANO’s estimated primary completion in the registry is August 2026 and the company has guided topline to the same window. As of the most recent filings, no readout has been published. The independent data safety monitoring committee has issued three consecutive recommendations to continue without protocol modification, most recently in May 2026, with all active treatment-arm patients having reached a second dose and more than 35 per cent a third.

The two trials are not measuring the same thing

A superiority result on vision maintenance in naive patients and a non-inferiority result on mean letter change in a mixed population are different claims about different populations against different comparators. They will be reported within months of each other and they will be compared anyway, by headline writers and by traders, as though one were a read-across for the other.

The read-across that is defensible is narrow and it concerns the class, not the product. If a sustained-release tyrosine kinase inhibitor at a six-month interval holds visual acuity within 4.5 letters of on-label aflibercept in a real-world population, the mechanism works and the delivery format works. If it does not, both companies are explaining the same disappointment with different molecules. The read-across that is not defensible is the numeric one: a percentage from SOL-1 and a letter score from LUGANO cannot be laid side by side.

The balance sheets behind the science

Clinical arguments in this sector are settled by data. Corporate outcomes are settled by whether a company can still write cheques when the data arrives, and on that measure the two lead companies are in very different positions.

Ocular Therapeutix closed the second quarter of 2026 with $598.6 million in cash and cash equivalents, against $737.1 million at the end of 2025. It carries $82.5 million of principal under a Barings credit facility drawn in August 2023, maturing in 2029, priced at SOFR plus 6.75 per cent with a 1.50 per cent floor. Net loss for the quarter was $78.8 million, with research and development at $54.1 million. It also has something none of its peers has: a marketed product. DEXTENZA generated $13.5 million in the quarter and $24.3 million across the half. The company guides to cash sufficiency into 2028 and drew nothing from its at-the-market facility in the first half.

EyePoint closed the same quarter with $110.5 million in cash and equivalents plus $70.0 million in marketable securities, $180.5 million in total, down from $306.1 million at the end of 2025. Net loss was $94.5 million, with research and development at $83.6 million against $55.5 million a year earlier. Revenue was $0.5 million. The company carries no debt and guides to funding operations into the fourth quarter of 2027. It sold 1,429,047 shares through its Cantor at-the-market facility in the first half at an average of $14.16 for roughly $20.2 million gross, and a further 1,208,718 shares in July at $14.44 for roughly $17.5 million gross.

Set those figures against each other and a structural asymmetry appears. Ocular Therapeutix holds more cash than EyePoint, Kodiak Sciences and REGENXBIO combined, carries a modest debt load it can service, and books modest but real product revenue. EyePoint spends more per quarter than Ocular Therapeutix while holding less than a third of the cash, has no product revenue, and is tapping the equity market in the months immediately before its defining readout.

Neither position is fatal and neither is decisive. EyePoint’s stated runway is a company assertion supported by planning it can see and outsiders cannot, and a positive LUGANO would open financing at terms unavailable today. The sequencing, however, is unambiguous: EyePoint needs LUGANO to work partly because of what LUGANO is, and partly because of what happens to its cost of capital if it does not.

Share counts sit at approximately 219.6 million for Ocular Therapeutix from the 10-Q cover page, though the company’s own earnings release cites approximately 225.0 million as of the same date, a discrepancy that remains unreconciled rather than resolved by guesswork, and 86.2 million for EyePoint.

Wave three: the one-shot problem

Behind the implants sits a class of therapy that, if it works, makes the six-month interval look like an interim measure. Gene therapy delivers a construct that instructs the retina to manufacture its own anti-VEGF protein continuously, from a single administration. There is no redosing interval because in the intended case there is no redosing.

REGENXBIO’s ABBV-RGX-314, partnered with AbbVie, is the furthest advanced. The subretinal programme comprises ATMOSPHERE (NCT04704921, 671 patients, primary endpoint mean BCVA change at week 54 against monthly ranibizumab) and ASCENT (NCT05407636, 735 patients, against aflibercept every eight weeks), both non-inferiority, both fully enrolled, with topline guided to the fourth quarter of 2026. The AbbVie collaboration carried a $370.0 million upfront, up to $1.38 billion in milestones, a fifty-fifty US profit split and tiered ex-US royalties. REGENXBIO received a $100.0 million milestone in July 2026 on first patient dosing in NAAVIGATE, and raised $107.8 million net in the same month, taking a $105.5 million June cash balance to a funded position into the fourth quarter of 2027.

4D Molecular Therapeutics takes the intravitreal route with 4D-150, which avoids the vitrectomy-adjacent surgical complexity of subretinal delivery. 4FRONT-1 (NCT06864988) completed enrolment of more than 500 patients in roughly eleven months and reads out in the first half of 2027; 4FRONT-2 (NCT07064759) is recruiting toward a 2027 second-half readout. The company held $457.6 million at the end of March 2026, the deepest cash position among the challengers.

The third intravitreal programme has changed hands. Adverum’s Ixo-vec, in Phase 3 ARTEMIS (NCT06856577, 311 patients, non-inferiority with a 4.5 letter margin, primary completion December 2026), no longer belongs to a small biotech. Eli Lilly agreed to acquire Adverum in October 2025 at $3.56 per share in cash plus a non-tradeable contingent value right worth up to $1.78 on US approval and up to $7.13 on annual worldwide net sales exceeding $1.0 billion. Adverum deregistered in December 2025.

That transaction is the single most underrated fact in this competitive set. A programme that a capital-constrained small cap was struggling to fund now sits inside a company with the balance sheet to run it to approval and the commercial infrastructure to launch it. The competitive question for the implant companies is no longer whether Adverum can afford Phase 3. It is what Lilly does with a retinal gene therapy once it has one.

Three ways to die in wet AMD

The durability race has produced casualties, and each failed differently. Taken together they map the ways a company loses in this indication.

Death by endpoint. Opthea’s sozinibercept, a VEGF-C and VEGF-D inhibitor given in combination with a standard anti-VEGF, failed the Phase 3 COAST study in March 2025: combination arms produced mean BCVA improvements of 13.5 and 12.8 letters at week 52 against 13.7 letters for aflibercept monotherapy. The accelerated ShORe readout failed as well, against ranibizumab. Opthea discontinued the programme. The lesson is uncomfortable for everyone still running: the control arm in modern wet AMD is very good, and beating it requires a margin that biology may not supply.

Death by balance sheet. Clearside Biomedical was developing CLS-AX, suprachoroidal axitinib. The same molecule as AXPAXLI, delivered by a different route. Its Phase 2b ODYSSEY study (NCT05891548, 60 patients) reported stable visual acuity and central retinal thickness to week 36 against aflibercept, with 67 per cent of patients requiring no additional treatment before mandatory redosing at twenty-four weeks and an 84 per cent reduction in injection frequency against pre-screening rates. It secured an end-of-Phase 2 FDA meeting and alignment on Phase 3 plans. It then ran out of money. Clearside filed a voluntary Chapter 11 petition on 25 November 2025 in the District of Delaware, case 25-12109, designated Health Ocean Limited as stalking horse bidder for substantially all assets on 17 December 2025, was approved for auction on 19 December, was delisted from Nasdaq, and deregistered its shares in January 2026. Its stock trades over the counter as CLSDQ. The company held $6.8 million in cash at 30 September 2025 with substantial doubt about going concern disclosed. Positive Phase 2b data and FDA alignment on a Phase 3 design were not sufficient to survive to Phase 3.

Death by clock. Kodiak Sciences remains in the race with KSI-501 in Phase 3 DAYBREAK (NCT06556368, 675 patients across tarcocimab tedromer, tabirafusp tedromer and aflibercept control), with the one-year primary endpoint reading out in the third quarter of 2026. It held $169.5 million at the end of March 2026 and disclosed substantial doubt about its ability to continue as a going concern, stating that existing cash was insufficient for the twelve months following the filing. A company can be running a fully enrolled Phase 3 in a large indication and still be racing its own auditors.

Clearside is the case that should concentrate attention, because it is the closest analogue. The molecule inside AXPAXLI and the molecule inside CLS-AX are the same compound. The difference in outcome was not pharmacology. It was $598.6 million against $6.8 million.

What each side has to fear

The two lead companies face overlapping but distinctly weighted risks, and the differences are more instructive than the similarities.

The obstacles in front of $OCUL

An approval built on one trial and one population. SOL-1 enrolled treatment-naive patients and measured maintenance of vision. If the FDA grants approval on that basis, the label will describe that population and that claim, while the commercial opportunity sits substantially in previously treated patients on established anti-VEGF regimens. SOL-R, which addresses repeat dosing and includes an aflibercept 8mg comparator, does not read out until the first quarter of 2028. For roughly a year and a half, a marketed product would be competing on data that does not directly answer the question retina specialists will ask, which is what happens at the second and third implant.

A timeline that has already moved. SOL-R topline was guided to the first quarter of 2027 and now sits in the first quarter of 2028, following a decision to hold masking to week 96 for additional secondary endpoints. There is a coherent scientific rationale for that choice. There is also a pattern, and readouts that move once have a habit of moving again.

The comparator is a moving target. SOL-1 ran against aflibercept 2mg. By the time an approved product reaches the market, the practical standard of care will be aflibercept 8mg and faricimab at extended intervals. A superiority result against the old comparator does not automatically survive translation to the new one, and payers will ask.

Registry-grade real-world risk. The commercial case rests on the claim that a depot removes adherence risk. A patient who stops attending appointments also stops receiving implants. The depot converts a monthly adherence problem into a semi-annual one, which is a large improvement and not an elimination.

The obstacles in front of $EYPT

Concentration. DURAVYU is effectively the company. Revenue was $0.5 million in the quarter. There is no marketed product generating meaningful cash, no second franchise, and no partner absorbing Phase 3 cost. LUGANO and LUCIA are not important readouts for EyePoint, they are the company’s valuation.

The financing sequence. A $94.5 million quarterly net loss against $180.5 million of cash is a ratio under two quarters, and the company has been selling stock through its at-the-market facility in the months immediately preceding the readout, including in July 2026. Management guides to the fourth quarter of 2027 and has visibility into a spending profile that steps down once Phase 3 enrolment costs roll off. Both statements are true simultaneously. What they mean together is that the terms on which EyePoint next raises capital are set by LUGANO, not chosen by EyePoint.

Non-inferiority cuts both ways. Meeting a non-inferiority margin against aflibercept establishes that DURAVYU is not worse. It does not establish that a physician should adopt a new procedure, a new product and a new reimbursement pathway. The commercial argument then rests entirely on the durability claim, which means the injection-frequency and rescue-treatment secondary endpoints will matter as much as the primary.

A second readout that can undo the first. LUCIA follows LUGANO. Two identically designed studies exist because regulators want replication, and replication can fail. A positive LUGANO followed by a missed LUCIA is a specific, structurally possible outcome that the market rarely prices in advance.

What both have to fear, and it is the same thing

Gene therapy. If ABBV-RGX-314 reads out well in the fourth quarter of 2026, the conversation about retinal durability changes from how many months to how many administrations, and the answer becomes one. AbbVie has the commercial reach to make that case, Lilly now has a competing intravitreal asset, and 4D Molecular has $457.6 million and a delivery route that avoids surgery. An implant that must be replaced twice a year is a superior product to an injection given eight times a year and an inferior one to a treatment given once, and the six-month depot occupies the ground in between.

The counter-argument is real and should not be dismissed. Ocular gene therapy carries an immune-response and intraocular-inflammation risk profile that implants do not, subretinal delivery requires surgical expertise that does not exist at every site, durability beyond the first years is still being established, and the pricing and reimbursement model for a one-time retinal therapy in a Medicare population is unsolved. A depot is a known quantity administered by a physician who already performs the procedure. That familiarity has commercial value that spreadsheets systematically underweight.

The eighteen months that decide it

Five readouts land between now and the middle of 2027, and they arrive in an order that compounds. The class question is answered first, by EyePoint. The gene therapy question is answered second, by REGENXBIO and AbbVie. Ocular Therapeutix files in between and defends its repeat-dosing case last.

Dates for trials that have not read out are company guidance or registry estimates, not commitments. Three of the eight have already moved once.

Five questions to ask of any durability readout

The releases that follow these trials are written to be quoted, and the quotable sentence is rarely the informative one. Five questions separate a result that changes a market from a result that produces a headline. They apply to every readout in the table above and to the ones that follow.

One: what was the comparator, and was it dosed the way a good clinic doses it? A trial against aflibercept 2mg every eight weeks is a different test from a trial against aflibercept 8mg at extended intervals. Both are legitimate. Only one describes the market a product will actually enter. When a release names the comparator without naming the schedule, the schedule is the thing to go and find.

Two: superiority or non-inferiority, and against what margin? Non-inferiority with a 4.5 letter margin permits a product to be up to 4.5 letters worse and still succeed. That is the correct design for a therapy selling convenience rather than efficacy, and it is not a claim of equivalence. A superiority claim on a maintenance endpoint is a third distinct thing. These three results are routinely reported in the same language and they are not the same finding.

Three: how many patients needed rescue, and when? For a durability product this is the commercial endpoint wearing a secondary-endpoint costume. Supplemental or rescue treatment rates, and the distribution of time to first rescue, tell you what the real dosing interval will be in practice. A product that meets its primary endpoint while a third of patients receive rescue injections before month six has not delivered a six-month product.

Four: naive, previously treated, or both? Treatment-naive eyes respond differently from eyes with years of anti-VEGF exposure and established fibrosis or atrophy. A result in naive patients does not transfer to a switch population, and the switch population is where the prescriptions are.

Five: what does the balance sheet look like on the day the data lands? Clearside had positive Phase 2b data, FDA alignment on a Phase 3 design, and $6.8 million. The programme is now an asset in a bankruptcy auction. Data quality and corporate survival are separate variables, and the market prices them together only after it has been reminded that they are not the same.

How the field looks under each outcome

Four broad configurations follow from the next two readouts. They are descriptions of consequence, not predictions of probability, and none of them is a view on any security.

LUGANO succeeds and the gene therapy readouts succeed. The durability thesis is validated twice over and the market fragments by patient rather than by product: depots for patients who want a procedure they understand from a physician who already performs it, one-time therapy for patients and systems willing to trade an upfront cost and an immune-risk profile for the elimination of the treatment schedule. Both implant companies have a business. Neither has the market.

LUGANO succeeds and gene therapy disappoints. The six-month depot becomes the destination rather than the waypoint, and the two implant companies own a category with the incumbents forced to respond through pricing and interval extension rather than through mechanism. This is the configuration in which the class is worth most.

LUGANO misses and gene therapy succeeds. The tyrosine kinase inhibitor depot approach absorbs damage as a class, including a product with a positive Phase 3 already in hand, because the market will read a failure as a statement about sustained intracellular VEGF blockade rather than about vorolanib. Attention and capital move to the one-time therapies and to whichever company can fund the wait.

Both disappoint. The incumbents keep the market by default, EYLEA HD and VABYSMO continue extending intervals incrementally, and the undertreatment gap documented in the registry literature stays open. Commercially this is the quietest outcome. Clinically it is the worst one, because the patients described at the outset are still losing 4.9 letters to a treatment gap that nobody has closed.

Across all four, one fact does not change: 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.