This digest covers key virtual and digital health regulatory and public policy developments during August and early September 2024 from United Kingdom, and European Union.

There have been some important reimbursement updates this month. In France, the French health regulator has issued a negative opinion on the first digital therapeutic that sought reimbursement through the PECAN pathway. This opinion suggests that while PECAN provides a more streamlined process by not requiring formal clinical trials, the regulator still expects robust data to show that the digital therapeutic is worthy of reimbursement. In the UK, the National Institute for Health and Care Excellence has published a position paper that highlights the risks of using AI for generating and reporting evidence for health technology assessments. It sets out that the use of AI in this context should be limited to situations where the value of doing so can be clearly demonstrated, and sets out guidance for companies on how to present such data and to justify the use of AI. Health technology assessment seems to be an area where AI has more readily been accepted, and industry will welcome guidance on the parameters around such use.

Regulatory Updates

EMA and HMA Publish General Principles for Staff Within the European Medicines Regulatory Network Regarding the Use of Large Language Models. The European Medicines Agency (EMA) and the European Heads of Medicines Agencies (HMA) have set out guiding principles and general recommendations for staff within the European medicines regulatory network when using large language models (LLMs). This is focused on the use of LLMs by the authorities, rather than the assessment of data provided by companies, although companies will hope to better understand the EMA and HMA’s approach when considering these principles.

It explains that many of the uses of LLMs, including assistance with writing, translation, and summarizing information, can support work in the medicines regulatory system. However, the paper points out specific risks of using LLMs, including potential hallucinations and inaccuracies. The paper also notes that, due to the nature of the work carried out by regulators, LLMs may not previously have been used to answer specific regulatory and scientific questions. Further, due to the lack of control over the datasets used to train the LLM, additional concerns arise regarding data protection and intellectual property. The principles therefore set out guidance on the use of LLMs while seeking to avoid pitfalls and risks.

Further information can be found in our BioSlice Blog.

Medtech Europe Issues Statement on the EU AI Act. On August 1, 2024, the European trade association for the medical technology industry, MedTech Europe, issued the statement on the industry perspective on the final AI Act. In the statement, MedTech Europe welcomes the AI Act but calls for:

  • Further alignment with the EU MDR/IVDR: MedTech Europe calls for consistency with regard to AI-enabled medical technologies, which are regulated as software under the EU Medical Devices Regulation (MDR)/EU In Vitro Diagnostic Regulation (IVDR) and are now subject to both sets of regulations.
  • A single conformity assessment and technical documentation: MedTech Europe appreciates the effort to align high-risk AI system assessments with existing conformity procedures under MDR/IVDR and emphasizes the importance of maintaining legal certainty. A single set of technical documentation should suffice for compliance with both regulations.
  • Clear pathway for clinical and performance evaluation of medical technologies: The AI Act should not disrupt existing procedures for clinical investigations and performance studies required under MDR/IVDR. MedTech Europe emphasizes that investigational devices and performance study devices should not require AI Act conformity assessment before evaluation, following the MDR/IVDR pre-market testing approach.

Reimbursement Updates

French Health Authority Publishes Negative Opinion on Reimbursement of First Digital Therapeutic Through PECAN. As discussed in our May 2024 digest, in March 2023, the French health regulator (Haute Autorité de Santé, HAS) introduced an accelerated pathway for reimbursement of digital therapeutics (DTx) including health apps, known as PECAN (La prise en charge anticipée numérique).

In an opinion issued on July 23, 2024, HAS set out its view on HelloBetter Insomnia, a prescription-only application that is used as a cognitive behavioral therapy for chronic insomnia and which is the first dossier submitted through the PECAN pathway. The opinion concludes that the available data did not support the app as an effective treatment for the specified indications. The ongoing study, conducted in the general population as opposed to a controlled clinical trial, would not provide sufficient data to justify a favorable opinion within the timeframe given. Further, it was not conducted in relation to the claimed indication (chronic insomnia). The regulator considered that the studies presented did not demonstrate a clinical benefit compared to other available cognitive behavioral therapies. Although this is only the first opinion issued by HAS through PECAN, it suggests that while PECAN provides a more streamlined process by not requiring formal clinical trials, HAS still expects robust data to show that the DTx is worthy of reimbursement. It is worth noting that HelloBetter has secured six approvals through the equivalent German DiGA process.

UK NICE Published Its Position Paper Highlighting Concerns With the Use of AI in Health Technology Assessments. On August 15, 2024, the National Institute for Health and Care Excellence (NICE) set out its position on the use of AI for generating and reporting evidence for health technology assessments. It stresses that the use of AI introduces various risks, such as bias, transparency concerns, and cybersecurity. On the other hand, NICE also recognizes the benefits that AI can bring in this context, including its ability to process and analyze large volumes of data. Therefore, NICE concludes that AI must only be used “where there is demonstrable value from doing so.”

Where an organization makes a submission to NICE, any use of AI in the generation of evidence must be supported by clear justifications and details of the assumptions that have been used. In addition, it is important that organizations contact NICE at an early stage, where possible, to discuss any potential use of AI in evidence generation.

IP Updates

UK Court of Appeal Weighs in on Patentability of Inventions Involving Artificial Neural Networks. In a significant ruling, the UK Court of Appeal handed down its judgment in Comptroller-General of Patents, Designs and Trade Marks v. Emotional Perception AI Ltd [2024] EWCA Civ 825, reversing the High Court’s decision reported in our December 2023 digest.

The key questions before the Court of Appeal were to determine what a computer program is and whether there is a computer program in an Artificial Neural Network (ANN). If an ANN was deemed to be a computer program with no technical contribution, the computer program exclusion to patentability would apply on the basis that the invention was to “a program for a computer … as such” under section 1(2)(c) of the Patents Act 1977.

The Court of Appeal defined a computer as “a machine which processes information” and a computer program as “a set of instructions for a computer to do something” before concluding that an ANN, whether implemented via hardware or software, was a computer. It added that the weights applied to an ANN were a computer program and therefore the computer program exclusion to patentability applied if the invention did not include a technical contribution unrelated to the computer program.

In this instance, the improved file recommendations in Emotional Perception’s invention were semantic in nature and did not provide a technical contribution to make the invention patent eligible.

Following this judgment, the UK Intellectual Property Office (UKIPO) has published statutory guidance titled “Examining patent applications involving artificial neural networks,” making an immediate change to their practice for the examination of ANNs. Readers of the Virtual & Digital Health Digest and virtual and digital health companies considering filing patent applications for inventions involving ANNs should note that the UKIPO confirms that “patent examiners should treat ANN implemented inventions like any other computer implemented invention.” Accordingly, a technical contribution will be required for such inventions to be patentable.

Unified Patent Court’s Paris Local Division Invalidates Dexcom’s Patent in Ongoing Dispute With Abbott in Second Substantive Decision. In the July 2024 digest, we reported on the ongoing global dispute between Abbott and manufacturers and distributors of continuous glucose monitoring (CGM) devices and technology, and summarized decisions relating to the same from the UK Patents Court and The Hague local division of the Unified Patent Court (UPC).

In further infringement and invalidity proceedings between Abbott and Dexcom, before the Paris local division of the UPC, the Court of First Instance held that Dexcom’s patent, EP3435866, was invalid for lack of inventive step. Dexcom’s patent relates to systems and methods of communication between the sensor electronics unit and the display device of an analyte monitoring system (of which CGM systems are an example).

Upon considering if the invention set out in claim 1 of Dexcom’s patent involved an inventive step (a requirement for patent eligibility and one of Abbott’s grounds for the revocation of Dexcom’s patent), the court concluded that it would be obvious to the notional skilled person to use the relevant protocol in claim 1 of the patent, i.e., to use second near-field communication to transmit the second portion of the analyte measurement data resulting in an effect of the well-known advantages of lower power consumption and security due to low range.

Dexcom attempted to avoid revocation of their patent by amending claim 1. Abbott objected to this amendment and the court found it to be not allowable on the grounds of added subject-matter. Further, Dexcom did not submit specific arguments in support of why the remaining patent claims were novel and inventive. Consequently, the court concluded that Dexcom’s patent was invalid, as granted and as amended. Dexcom’s patent was accordingly revoked and the infringement action against Abbott was dismissed.

This decision represents the UPC Paris local division’s first substantive infringement decision on the merits. It was handed down just 24 hours after the first infringement decision on the merits from the Düsseldorf local division. This judgment exemplifies one of the UPC’s key aims — efficiency. The decision was handed down within a year of the proceedings’ commencing, which is in line with the expected timeline provided in the preamble of the Rules of Procedure of the UPC.