After CE marking: what a digital medical device still has to prove in France

People with chronic macular disease used to be given a paper grid to stick on the fridge. If the lines started to look wavy, they were supposed to call their ophthalmologist. Tilak Healthcare set out to replace the grid with Odysight, a CE-marked smartphone app in which patients test their own vision at home through short games, and which alerts the treating ophthalmologist when vision drops. Édouard Gasser, the company's chief executive, came from video games, where downloads could reach a million in a day. In its early years, he says, Odysight counted 200 patients and 15 doctors.

That gap between a product that works on paper and a product people use is the subject of a three-part French-language Tech Anatomy series. Gasser speaks alongside Laurence Comte-Arassus, president since 2023 of the Snitem, the French medical technology industry association, and a GE Healthcare executive, with Guillaume Laguette, chief revenue officer of the French handheld ultrasound company echOpen, as co-host. Their shared starting point is that CE marking for a digital medical device (dispositif médical numérique, or DMN) is, in Gasser's words, "indispensable, necessary, but not sufficient" (quotations translated from French).

Why approval comes first, and why it settles little

CE marking certifies that a device meets European safety and performance requirements. Since the EU Medical Device Regulation (MDR) took full effect in 2021, that certification goes through notified bodies, the private organisations designated to audit manufacturers. Comte-Arassus describes the transition as a bottleneck: products already on the market had to be re-certified, and "we prevented new innovations from arriving, because we had to deal with the past." In December 2025 the European Commission proposed a targeted revision of the MDR, which Comte-Arassus says in the third episode should be implemented quickly rather than in two years.

For Gasser, the real marker of success is use, by patients first and then by doctors. Yet ophthalmologists, of whom France had around 5,700 at the start of 2023 according to the health ministry's statistics office, had no time to explain the app or follow patients remotely. Tilak learned that the decision was not the doctor's alone: orthoptists deliver much of the patient education, secretaries handle triage, and the whole centre had to be convinced. Comte-Arassus calls this the stakeholder map: "we forget one. That's the one that stops us."

The French route from experiment to reimbursement

Odysight entered an Article 51 experiment in February 2022. Article 51, created by the 2018 social security budget, lets the health system fund new care organisations on a trial basis. Then came PECAN (prise en charge anticipée numérique), an early-coverage route in force since 2023 that pays for one non-renewable year for a digital therapeutic or telemonitoring device presumed innovative, while the manufacturer applies for standard reimbursement. A PECAN file needs CE marking and a certificate of conformity with the interoperability and security standards of the ANS, the state digital health agency. The standard route for telemonitoring, the list of telemonitoring activities (LATM), has existed in common law since July 2023, after the ETAPES pilot programme ended.

In March 2025, the CNEDiMTS, the HAS committee that assesses medical devices, concluded that Odysight's benefit over conventional follow-up "cannot be established" and gave an unfavourable opinion on listing. Its report notes that the studies were retrospective and single-arm, that the Article 51 results were descriptive, and that in one study the average delay between an alert and a clinic visit was 33 days. Gasser says its Article 51 data were "judged not robust enough." Doctors who had prescribed the app for three or four years, he says, "told themselves, this is the end of the story." In October 2025, Tilak announced a favourable opinion under PECAN, the first for ophthalmology telemonitoring according to the company, which reported more than 30,000 patients and 1,200 prescribing doctors in total.

Real-world data and the cost of a trial

The second episode turns to evidence. Comte-Arassus puts the stakes plainly: "An innovation that does not reach patients is not an innovation, it is a trial." Gasser offers an image: clinical data is what arrives on the plate in a three-star restaurant, while real-world data is the back kitchen. His example is a dashboard. Asked whether their patients used the app correctly, doctors answered that theirs were exemplary; the usage data showed patients "were not doing at all" what had been recommended.

Both describe trial design as the riskiest step for a small company. Comte-Arassus estimates that a start-up spends at least €200,000 to €300,000 on a study whose result is never guaranteed. Gasser's advice is to be conservative in endpoints and to build a scientific board early, with one traditional methodologist who thinks like the authorities and one more modern methodologist looking for cheaper, faster designs. On PECAN itself, she suggests that "when companies don't apply, it's certainly because it's too structured."

The episode compares France with Germany's DiGA scheme, legislated in late 2019, which lets doctors prescribe listed apps paid for by statutory insurers. Gasser recalls weak early uptake and prices roughly halved after negotiation. The insurers' 2025 report gives a fuller picture: 1.9 million DiGA prescribed or approved by the end of 2025, 74 apps listed since 2020 of which 58 remained, and negotiated prices for permanently listed apps such as HelloBetter and Kranus around 60 percent below the launch price. France, in Gasser's reading, leads on telemonitoring and lags on digital therapeutics, while Germany shows the reverse.

Patient pull or doctor push

Two explanations of adoption sit side by side. The first, Gasser's, is that patients lead: once they use a tool, doctors follow. Comte-Arassus supports it with a precedent from diabetes. Abbott's FreeStyle Libre glucose sensor arrived in France in 2014 and was reimbursed nationally from June 2017. In her account, patients began paying for it themselves before coverage, at a time when industry assumed that in a system where nearly everything is reimbursed, coverage had to come first.

The second explanation puts the professional system at the centre. Doctors, Gasser says, hesitate to stake their medico-legal credibility on a product that may not last. Comte-Arassus describes a hospital that reorganised around an innovation only for the supplier to go bankrupt. Distribution enters the same reading. Tilak worked with Novartis and later with Laboratoires Théa, and Gasser reports that results varied with the sales network used, since sales forces used to selling drugs that do not change practice struggle to promote a device that does. Both readings fit Tilak's history, and each points to different spending: on patient experience, or on clinicians, care teams and partners.

What investors are asked to accept

The third and final part, published under the same title as the second, is addressed to investors. Gasser contrasts a typical fund horizon of around seven years with the ten years he believes digital health needs to mature; Comte-Arassus puts it closer to twelve or thirteen. Gasser describes retention as the number funds now scrutinise and says that after a year about half of Odysight patients remain active, in a population with an average age of 72. He estimates that entering Germany would cost a company about €2 million in new studies, registration and deployment, and calls for a common European base for market access. Comte-Arassus argues that "a French start-up that started with the US will stay in the US."

What the series does not establish

The guests are an industry representative and a founder whose product is under assessment, and the series reflects their experience rather than a review of adoption data. Tilak's retention figure has not been independently verified, and the HAS report on the Article 51 cohort gives a lower one: 37 percent at twelve months. The trial costs, the time-to-market estimates and the €2 million for Germany are the guests' figures. Prescription counts, in France as in Germany, measure uptake rather than benefit to patients.

What to watch

The indicators that would separate the two readings are concrete: whether Odysight, at the end of its PECAN year, obtains a listing on the LATM and on what evidence; how many PECAN files the HAS receives and approves; whether the MDR revision proposed in December 2025 is adopted and changes timelines for small manufacturers; and whether a digital device validated in one EU country can enter another without repeating its studies. If reimbursed devices still struggle to be used, the case for patient pull and distribution will gain weight over the case for payment.

Listen to the episode

La vraie valeur d’un DMN avec Laurence Comte-Arassus & Édouard Gasser (ft. Guillaume Laguette)

Listen · 25 min