The Bioinformatics and Digital Health team at Kilburn & Strode is producing short updates delving into recent innovations, future directions, and patentability in AI across healthcare, including patient support technologies, telehealth, research, and clinician support technologies. This instalment highlights promising advances in the use of AI in the field of clinical trials.
How can AI help with clinical trials?
New AI methods and applications are revolutionising drug development, a topic we will come back to in this series of articles. However, despite significant AI-driven increases in the speed of identifying new drug candidates, these new candidates must still go through the bottleneck of clinical trials before they reach patients. The clinical trial process can take well over a decade in some cases, and very few candidates make it all the way through to regulatory approval. However, the increasing use of AI in the planning and analysis of clinical trials is set to significantly reduce the time to bring new drugs to the market, and could even improve success rates for drug candidates.
Clinical trials generate a vast amount of data to be used for regulatory approval, for example a Phase III clinical trial, the last step before regulatory approval can be sought, generates on average over 3 million data points. Furthermore, coordinating the trials requires organising drug administration and patient monitoring across thousands of patients. Such a process is an ideal target for applications of AI methods, which can accelerate and optimize the necessary organisation and data analysis, and enable novel insights in some cases. Indeed, a recent study by McKinsey found that the use of AI in clinical development is one of the three most impactful areas of AI use in the pharma industry and is expected to generate value of the same order as the use of AI in research and early discovery on the one hand and commercial operations on the other.
What is the state of the art in this area?
One of the particularly time consuming aspects of clinical trial management is the recruitment and filtering of patients to ensure a high quality data set. Inefficiency at this stage can result in significant delays or, even worse, a patient group that is unsuitable for the drug candidate which can lead to failure of potentially successful drugs. Quantitative measurement of patient metrics and subsequent analysis with AI avoids potentially subjective analysis by investigators, and there are a number of companies working in this area.
For example, Mendel.AI has developed a clinical AI tool capable of understanding context in patient health data and applying reasoning to avoid the errors commonly made by large language models (think ChatGPT-style hallucinations), and this context and reasoning can then be applied to medical records to better inform and simplify patient recruitment. Hallucination errors would be unacceptable in a clinical setting, and the ability of such an AI tool to quickly and accurately identify specific patient populations has the potential to make patient recruitment significantly more efficient.
Another area of clinical trials that is seeing significant AI innovation is the monitoring of patients throughout the process. Monitoring patients is highly time consuming, but crucial for ensuring both that the drug in question is efficacious and that any potential side effects are identified quickly to maintain patient safety. In combination with advances in wearable technology, AI can be used to remotely monitor live physiological data, and identify significant patterns quickly and accurately.
One example of this is the development, by a research group from Turin, Italy, of a wearable device capable of measuring wound healing in clinical research which, when combined with AI data analysis, was able to remotely and accurately classify wounds and quickly identify any deterioration, thereby enabling efficient patient monitoring.
Another example is Virtabot, a recent startup offering an AI-driven patient discovery platform for clinical trials. The technology aims to amplify patient insights with siloed data across the health system to help pharma companies recruit better suited candidates, which can result in lower drop-out rates and significant cost savings while bringing drugs to market faster.
IP strategy for clinical trial AI
As for all technology areas and sectors, it is crucial that your IP strategy is driven by business goals and how you operate. A particular feature of AI driven clinical trial design and support is that much of the processing can happen on proprietary platforms, for example in the cloud, as a service or software as a service model. With that in mind it may well be possible to protect your innovations as trade secrets. However, reasons that might point away from a trade secret strategy, in particular relevant in any area relying on machine learning talent, include a desire or need to publish information about your innovation, the need to disclose technology to partners or regulators, a highly mobile workforce and the like. If these apply to your business, then patents may form an important part of your IP strategy. Contrary to what you may have heard, software can be patented and this is also true for software that uses AI technology for technical purposes.
Further considerations when patenting in this sector relate to the nature of the data obtained in clinical trials and subsequently used to train AI models. Applicants often have to consider how much data they disclose in patent applications to ensure, on the one hand, that competitors cannot access commercially valuable proprietary data, but also, on the other hand, that the application meets patent office requirements of enabling the ‘skilled person’ to work the AI invention and demonstrating that it achieves a technical effect. However, when using patient medical data for clinical trial AI models, there are additional stringent confidentiality requirements that mean you should be doubly cautious about disclosure of data. Innovators should ensure the data they include in applications in this field meet each of these criteria, the combination of which is relatively unique to the application of AI to clinical patient details. Kilburn & Strode’s Bioinformatics and Digital Health team are able to assist with treading this fine line. Further considerations for patenting AI in life sciences will be addressed in the coming articles in this series.
Closing remarks
AI is becoming more embedded into every aspect of everyday life and the healthcare sector is no exception. As AI integration into clinical trials improves everything from patient recruitment to physiological monitoring, innovators in this area should consider the relative merits of patenting as part of their IP strategy, as well as rules around disclosure of data, particularly clinical data, in patent applications. The team at Kilburn & Strode is excited to be supporting innovators in this area, who are pioneering the AI revolution in healthcare. Stay tuned for our next instalment, where we will be discussing applications of AI in drug discovery.
For further information or advice related to patenting inventions in this area, please do not hesitate to contact James Cochrane or your usual Kilburn & Strode advisor.
If you enjoy our series and wish to discuss any of the related content or are interested in protecting your own innovation in these areas, do reach out to our Bioinformatics and Digital Health team.