PRIMA implant: a chip under the retina restores reading in AMD, but not sight

On 22 July 2026, Science Corporation announced CE marking for PRIMA, a 2-millimetre chip implanted beneath the retina that allows patients with the most advanced form of dry AMD (age-related macular degeneration) to read letters again. It is the first time a device has restored a form of central vision in geographic atrophy, a disease for which medicine had until now had nothing to offer. The published results are real and solid. The figure circulating most widely in the press, however, does not say what it is being made to say.

THE ANNOUNCEMENT

What was announced on 22 July 2026

Science Corporation, the American company developing the implant, announced on 22 July 2026 that it had obtained CE marking for PRIMA under the European Medical Device Regulation. The certification was issued by the notified body DEKRA. It authorises marketing of the system in thirty European countries.

One detail deserves to be pointed out, because several articles got it wrong: the date on which the marking was actually granted has not been made public. 22 July is the date of the press release, not that of the decision. This may seem pedantic; it is the kind of imprecision that, repeated often enough, ends up turning an industrial announcement into a dated regulatory event.

A CE marking is not the same as availability

This is the most frequent confusion, and it has very concrete consequences for the patients who ring a consulting room after reading the news. CE marking certifies that a device meets European safety and performance requirements. It says nothing about price, nothing about reimbursement, and nothing about which centres are able to fit it.

The manufacturer’s press release is explicit on this point: country-specific reimbursement applications and the activation of clinical centres are “under way” across Europe. The first commercial implantation is expected shortly in Germany. No other country is named for what follows. In other words, between the announcement of July 2026 and the moment when a French patient will be able to benefit from it, an entire administrative road remains to be travelled.

THE DISEASE

Geographic atrophy, or why there was nothing to offer

To understand why this announcement matters, you have to know exactly which disease is being talked about. Geographic atrophy is the advanced stage of dry AMD. The cells of the retinal pigment epithelium disappear progressively at the centre of the retina, taking with them the photoreceptors they nourish. On fundus images, these dead zones trace patches with sharp outlines, like continents on a map; hence the name.

The patient does not become blind in the everyday sense of the word. Peripheral vision remains: they continue to move about, to perceive obstacles, to make out a silhouette entering a room. What is lost is the centre of the visual field, that is to say precisely the zone used for reading, for writing, for recognising a face, for dialling a telephone number. A blind spot settles in exactly where the gaze lands. It is a disability of a particular brutality, because it attacks the gestures of independence without ever preventing someone from walking.

The New England Journal of Medicine paper recalls that geographic atrophy affects more than five million people worldwide, and that no vision-restoring treatment existed until now. That last wording matters: it does not say that nothing exists, it says that nothing exists that gives back vision. Follow-up, monitoring for conversion to the wet form, lifestyle adaptation and low-vision rehabilitation all remain useful; but none of them brings back what has gone. That is the wall the implant is beginning to breach. We set out current management in our article on monitoring dry AMD.

THE DEVICE

How the PRIMA implant works

The implant is a square chip 2 millimetres across and 30 micrometres thick, three times finer than a hair. It carries 378 electrodes and is slid surgically beneath the retina, at the level of the atrophic zone. It is photovoltaic, which means that it contains neither battery nor wire: it converts the light it receives directly into electrical current.

The patient then wears glasses fitted with a camera. The camera films the scene, a processor handles it, and the image is projected onto the chip. The electrodes convert that light into electrical pulses which stimulate the retinal neurons still alive above the destroyed area. The signal then travels on through the optic nerve to the brain. The natural visual chain is therefore short-circuited at the exact point where it is broken, and only there.

The system also includes a zoom, adjustable up to twelve times magnification. This is an important part of the real benefit: enlarging a character makes it fall on more electrodes, and therefore makes it legible. It also means that the reading achieved is assisted reading, slow, with a device worn on the face; not a return to reading a newspaper spontaneously.

A technology born in Paris

The lineage is worth recalling, because it is French. The proof of concept came from the Institut de la Vision in Paris, and gave rise in 2011 to the spin-off Pixium Vision, whose assets are today held by Science Corporation. Development was co-led by José-Alain Sahel, between Pittsburgh and the Institut de la Vision, Daniel Palanker at Stanford, and Frank Holz in Bonn. The twelve-month results were presented at the EURETINA congress in Paris, from 4 to 7 September 2025, by Frank G. Holz, before their publication in the New England Journal of Medicine.

THE RESULTS

What the PRIMAvera trial actually showed

The trial is called PRIMAvera. It is an open-label, multicentre, prospective, single-group study in which each patient serves as their own control: vision after implantation is compared with vision beforehand. Seventeen centres across five countries took part: France, Germany, Italy, the Netherlands and the United Kingdom. Participants were aged 60 and over, with a mean age of 78.9 years, and visual acuity already very low at inclusion: 1.2 logMAR or worse, the logMAR scale rising as vision deteriorates.

Thirty-eight patients were implanted. Thirty-two could be assessed at twelve months. The other six were not assessed, for reasons that should be stated without euphemism: three deaths, one withdrawal of consent, two people unavailable. In a population whose mean age approaches 79, one-year mortality is in no way unexpected, but it is a reminder of how frail the ground is on which this surgery is performed.

The main result

The primary endpoint was the number of patients achieving a clinically meaningful improvement in visual acuity, defined as a gain of at least 0.2 logMAR, that is around ten letters on a standard chart. The result: 26 patients out of 32, or 81%, with a 95% confidence interval of 64 to 93%, and p below 0.001. Taking statistical account of the participants who were not assessed, the figure remains at 80%, with an interval of 66 to 94%.

The Inserm press release specifies that 78% of patients exceeded the higher threshold of 0.3 logMAR, that is fifteen letters, and that the largest gain observed reached 1.18 logMAR in one patient, the equivalent of 59 letters. At twelve months, 84.4% of participants were able to read letters, numbers and words at home with the glasses, which is perhaps the most telling result, because it is measured outside the laboratory.

One technical point, finally, often passed over in silence and yet reassuring: mean natural peripheral acuity after implantation remained equivalent to that measured before the operation. Placing a chip beneath the macula did not therefore cost patients the peripheral vision they depend on to get about.

The figure of 25 letters calls for a clarification

Since the CE marking announcement, one figure has been circulating everywhere: patients are said to have gained an average of 25.5 letters, that is five lines on a chart. It is picked up as it stands as the result of the study. It is not.

That figure appears neither in the abstract of the New England Journal of Medicine paper, nor in the Inserm press release, nor on the Institut de la Vision website. It comes from the manufacturer’s press release. The University of Pittsburgh, where José-Alain Sahel works, phrases it in a revealing way: the sentence about the 25 letters comes immediately after the one describing the 26 patients out of 32 who achieved a clinically meaningful improvement. It therefore describes the mean among the patients who responded to the treatment, not the mean across all participants.

The difference is not rhetorical. Saying “patients gain 25 letters” implies that a patient who undergoes implantation can expect that gain. Saying “81% of patients gain at least 10 letters, and among those who respond the mean gain is around 25 letters” describes the same study, honestly, and leaves the patient room to understand that roughly one patient in five did not obtain a clinically meaningful improvement. That is the version a patient is entitled to hear before consenting to subretinal surgery.

It should be added that the trial was funded by Science Corporation, that is by the manufacturer of the device, with co-funding from the NIHR biomedical research centre at Moorfields. This in no way invalidates results published in one of the most demanding peer-reviewed journals in the world. But it does justify carefully distinguishing what comes from the paper and what comes from the press office.

THE VISION ACHIEVED

What prosthetic vision looks like

This is the question patients ask from the second minute of the consultation, and the answer is rarely in the press coverage. In 2026 a team published in the Journal of NeuroEngineering and Rehabilitation a simulation of the vision obtained with the PRIMA system, precisely in order to characterise it.

First point: the vision restored is monochrome. The simulation algorithm includes a greyscale filter, because the device does not transmit colour information. An implanted patient does not recover the colour of their grandchildren’s faces; they recover a contrast.

Second point: the resolution corresponds to the pixel size of the chip, that is 100 micrometres. This is what caps the acuity obtained and explains why the zoom is indispensable. The authors also note reduced contrast sensitivity compared with natural vision.

Third point, and this is the most important in human terms: patients report difficulty perceiving faces. They read, they write, they decipher a label or a number; but recognising someone remains difficult. The paper in fact works specifically on improving the representation of faces, which says a great deal about the scale of the problem. A counter-intuitive detail: patients describe perceiving smooth patterns rather than a mosaic of dots, which suggests that the brain reconstructs part of what it receives.

José-Alain Sahel himself states the limit plainly: “I don’t think we’ll ever be able to restore full 20/20 vision with the implant alone.” Coming from the researcher who has carried the project from the outset, it is the most useful phrasing there is. To get a sense of what degraded vision means, our vision simulator makes it possible to handle these notions interactively.

THE RISKS

The complications observed in the trial

Placing an implant beneath the retina requires major vitreoretinal surgery, with a retinotomy, a deliberate opening of the retina, to slide the chip into the subretinal space. The trial recorded 26 serious adverse events in 19 participants, that is half of the people implanted.

Two figures qualify this picture. First, 21 of these 26 events, or 81%, occurred within two months of surgery: these are perioperative complications, not a toxicity that would set in over time. Second, 20 of these 21 events, or 95%, resolved within two months of onset.

The Inserm press release details their nature: raised intraocular pressure is the most frequent, ahead of retinal detachments, macular holes and subretinal haemorrhages. These are the classic complications of retinal surgery, known and managed, but they are a reminder that we are talking about a procedure reserved for specialist teams. We describe elsewhere the signs that should prompt an emergency consultation in cases of retinal detachment.

José-Alain Sahel sums up the balance as follows: “The benefit proved far greater than the adverse effects.” (translated from the French) This is a clinician’s judgement on a population that had no alternative, and it has to be read in that precise context: in a patient whose central vision is already destroyed, the benefit-risk ratio bears no relation to that of elective surgery.

What was learnt from examining the eyes under the microscope

A 2026 publication in Ophthalmology Retina sheds a kind of light that few devices ever receive: histological analysis of four eyes from two trial participants who had died, one at seven weeks after implantation, the other at eighteen months.

The conclusions are broadly favourable. The implant sat correctly at the level of the outer plexiform layer, as intended by the design. Biocompatibility is judged good: no significant encapsulation of the chip, no notable inflammatory reaction around it. That is far from a given for a foreign body left in contact with nerve tissue.

At eighteen months, the authors describe, by contrast, a small break in Bruch’s membrane accompanied by localised subretinal fibrosis, a full-thickness scar at the retinotomy site, and limited focal atrophy of the inner retina. Nothing catastrophic, but structural changes that are quite real, and whose significance at five or ten years remains unknown. Trial follow-up is planned to 36 months: still short for a device meant to stay in place for life.

IN FRANCE

Can it be obtained in France today?

No. CE marking opens up commercialisation in thirty European countries, but the first commercial implantation is expected in Germany, and the manufacturer indicates that country-by-country reimbursement applications are still under way. At this stage, the PRIMA implant is neither available nor reimbursed in France outside the research setting.

This is not for want of national expertise. The affiliations on the New England Journal of Medicine paper include the Fondation Adolphe de Rothschild in Paris, the Croix-Rousse hospital in Lyon, the CHI de Créteil, the Henri-Mondor hospital, the university hospitals of Nantes and Bordeaux, the Hôpital National des Quinze-Vingts and the Institut de la Vision in Paris, as well as Sorbonne Université. France was one of the main contributors to the trial. Moving from clinical trial to routine care is now a matter of assessment and pricing procedures, not of technical feasibility.

What to do in the meantime

Three things, none of them spectacular, but with the advantage of being available now.

• Have the diagnosis confirmed. Not every loss of central vision in a person over 60 is geographic atrophy. An epiretinal membrane, macular oedema or an associated cataract may account for part of the difficulty; and those can be treated.
• Watch for conversion to the wet form. Dry AMD can be complicated by neovascularisation, and there an effective treatment does exist, provided it is started quickly: intravitreal anti-VEGF injections. Sudden distortion of straight lines calls for a prompt opinion.
• Do not wait until the end stage to organise low-vision support. Optical aids, lighting, orthoptics, adapting the home: these measures change daily life and are put in place well before an implant might eventually become accessible.

WHAT TO TAKE AWAY

A real step, kept in proportion

What has just happened should not be understated. For the first time, a device is enabling people whose central retina is destroyed to read words again, in a multicentre trial published in the New England Journal of Medicine, with the primary endpoint met and a safety profile documented down to histology. In a disease where monitoring was until now all that could be offered, that is a shift.

Nor should it be overstated. The vision obtained is monochrome, assisted by glasses, insufficient for recognising a face; implantation carries real surgical risks; follow-up is twelve months, with monitoring planned to three years; and access in France remains to be built. A patient who reads “sight restored” and comes to the consultation thinking the problem is solved will leave disappointed. A patient to whom what has actually been demonstrated is explained will, by contrast, be able to take a position in full knowledge of the facts on the day the implant becomes accessible.

This development belongs to a wider sequence. A few days after the CE marking announcement, a team in Turin carried out the first macular autograft in the world, following the opposite logic: repairing with living tissue rather than substituting with electronics. Two very different routes, but the same starting observation: the central retina, long considered definitively lost, is no longer quite that.

FAQ

Frequently asked questions

Sources

1. Holz FG, Le Mer Y, Muqit MMK, Hattenbach LO, Cusumano A, Grisanti S, et al. Subretinal Photovoltaic Implant to Restore Vision in Geographic Atrophy Due to AMD. N Engl J Med. 2026;394(3):232-242. PMID: 41124203
2. Inserm. Un implant sous-rétinien restaure partiellement la vision de personnes atteintes de DMLA. Communiqué du 20 octobre 2025. presse.inserm.fr
3. Science Corporation. PRIMA receives CE mark. Press release, 22 July 2026. science.xyz
4. Simulation of prosthetic vision with the PRIMA system and enhancement of face representation. J Neuroeng Rehabil. 2026. PMID: 41896965
5. Histologic Findings after Subretinal Implantation of the PRIMA Photovoltaic Array. Ophthalmol Retina. 2026. PMID: 42250602

Further reading

• AMD: diagnosis and treatment
• Intravitreal injections: indications and protocol
• Dr Moïse Tourabaly — Ophthalmologist

Cachan practice · Tel. +33 1 45 47 08 11

Disclaimer

This article is for information purposes. A personalised ophthalmological opinion remains essential for any treatment decision.

This article describes the state of published knowledge as at 3 August 2026 and constitutes neither a treatment proposal nor a surgical indication. The PRIMA implant is not available in France outside research. The results presented come from a single-group trial of 38 patients followed for twelve months and do not predict any individual outcome. Any decision concerning AMD is a matter for an ophthalmological opinion after examination.