A Realist Looks At AI In Clinical Trials
January 28, 2019 | When assessing new technologies for clinical trials, one needs to be both visionary and skeptical. That seems to be even more true for AI technologies.
Ronald Dorenbos, Associate Director Materials & Innovation and part of the digital strategies group for neuroscience at Takeda, assesses technologies and innovation that may benefit Takeda’s core areas: Oncology, Neuroscience, and Gastroenterology. In that role, he also follows how Artificial Intelligence and machine learning can be applied in clinical trials. Dorenbos has a background in microbiology and neuroscience and has led projects for some of the world’s top 10 pharmaceutical companies that included programs around strategy, commercialization, and digital health. He now helps Takeda with their innovation and technology strategy. Many current projects are related to digital health and it becomes ever more important to consider involvement of AI, machine learning, and how such novel tools and technologies can be implemented in different areas in pharma and healthcare organizations.
On behalf of Clinical Research News, Marina Filshtinsky spoke with Dorenbos about how AI can be applied in clinical trials, what challenges lie ahead of us, and whether AI is really all it’s sold to be.
Editor’s note: Marina Filshtinsky, Executive Director of Conferences at Cambridge Healthtech Institute, is planning a track dedicated to Artificial Intelligence in Clinical Research at the upcoming Summit for Clinical Ops Executives, SCOPE, in Orlando, February 18-21. Dorenbos will be speaking on the program. Their conversation has been edited for length and clarity.
What are the key areas of artificial intelligence applications in clinical research?
There are quite a number to mention. Recruitment is one thing and several firms are focused on enrolling the right patients for clinical trials by using electronic health records and other sources of patient information together with info about the clinical trial to very quickly pinpoint the right patient population. Where this process used to take sometimes months or even years, there are now firms that promise to do the complete recruitment within a matter of weeks and they may be able to show the right cohorts of patients in a matter of hours. That's a huge advantage as it significantly decreases the time of the whole clinical trial process.
Another important area is adherence. Patients are notorious for not adhering to treatment protocols, and a lot of clinical trials fail because of non-adherence, resulting in huge losses in time and money. AI in different formats, for example with smart phone apps that check whether the patient takes his medication or apps that support patients by telling them to take the medication, can significantly increase adherence rates and in that way the successful outcome of clinical trials.
Other fields where I see AI being applied in clinical trials are in the decision-making process and in diagnostics. In deciding on the right course of action before or during a trial, AI can be helpful as it can consider current and historic information from numerous resources such as monitoring devices, other clinical trials, doctor’s notes, publications etc. For diagnostics there's a whole range of different areas where AI can assist that include things like image analysis, voice analysis, keystroke analysis, analysis of EEG's, ECG's, breathing, heart rate, sleep pattern and movement. Monitoring and analyzing all these functions will be much more efficient with AI and AI will be able to monitor many of the parameters around the clock.
What are three challenges in the way of harnessing the power of artificial intelligence in clinical trials? Is the clinical data science, data management ready for artificial intelligence and machine learning in clinical research?
Many of the data scientists, data institutes and data groups within the pharma and healthcare industry are ready or getting ready because everybody knows this is coming. But there are definitely some challenges.
One of the challenges is the fact that data is coming from a variety of resources. Think for example of the Internet of Things with a lot of different sensors and technologies being used to capture the data. Different locations and geographies may not be equipped with the same kind of instruments and equipment. The result is a huge variety in different data streams that do not necessarily communicate well with each other. All this data needs to be brought together in a format that can be understood by the AI tools that are being used.
Another challenge is the fact that data that comes in is not clean enough to be immediately used by AI, so apart from a need to have it in the right format, you also must make sure you have clean data. A lot of time and effort are currently being spent to clean up the data.
A different challenge comes in the form of skepticism. There is still a lot of reluctance in the clinical world among healthcare providers and other people involved and associated with clinical trials who do not necessarily see benefits of AI, and do not see the efficiency that it can bring to the whole process. There's a need to educate these people as well as the patients to accept involvement of AI in the clinical trial process. It is therefore important to show that AI can help prevent people from becoming a patient or help patients to become better faster to increase acceptance rates for AI-assisted technologies.
Another part is concerning privacy issues. With AI you may want to collect vast amounts of data to do analyses and make predictions. Collecting this data, you need to assure that patients are willing to share their information and have a good understanding of what you are going to do with all the data that you are collecting. Different privacy laws in different geographies like the U.S. and Europe, make this even more challenging.
A last challenge I like to mention is regulation. The regulatory bodies are struggling to keep track of all the developments. A couple of years ago, there were close to 200,000 health care apps and that number has increased dramatically. The IQVIA Institute for Human Data Science recently reported that there are currently over 318,500 mobile health apps with over 200 new apps released every day. As it is difficult for organizations like the FDA to keep track of everything that is going on it leads to a reluctance to (quickly) give approval for developments that are going on in this field. Introduction of novel clinical end points based on combinations of AI and digital health that are technically feasible and that could improve clinical trials, get delayed because the right regulatory frameworks are not yet in place.
How will new technology impact the whole idea of patient-centricity of clinical trials?
I believe the patient should always have a major role in the whole process. During all the work we do in the pharmaceutical industry it is very important to keep the patient in mind and if possible to get the patient involved. I have spoken with some patients about their perspective on AI in clinical trials, and some of the patients were saying that it might have been helpful to find the right specialist quickly and AI might have been able to advise on the best course of action during the trial. But at the same time, you hear some concerns. Some patients translate (incorrectly) the introduction of AI to a robot invasion into the clinical trial and they do not see AI as a substitute for human interaction and the doctor-patient connection. Another concern was around too much information. One of the patients I spoke with, a good friend of mine, was diagnosed with Waldenstrom macroglobulinemia, a very rare disease with about 1,500 cases a year in the US. AI would probably have told her that she would have had no chance of survival, and that kind of information might have resulted in her not participating in the clinical trial that ultimately cured her (she is now happy and well and a celebrated artist in the Boston area). It illustrates that we have to be careful with the kind of AI generated information we give to patients as well as to healthcare providers.
It is crucial to get the patients involved and educate them about what AI is and how it can help us to make the clinical trial process better. When we can show people that AI will prevent people from becoming a patient or get patients better faster the general population will accept the introduction of these new tools.
Is AI a panacea for clinical trials?
Everybody knows that there's quite some hype around AI. Not everything that is being promised and being shown is true or immediately available. We should, however, recognize that AI is making advancements and there are some breakthroughs. At the moment, the field is still in its infancy, but for the next, say two to five years, we will witness some really major improvements, not only in clinical trials, but also in things like drug development, post market research and other areas that are important for the healthcare and pharmaceutical industry. It will be exciting and fascinating to see these developments and to be part of some of the initiatives that will help improving the way we cure people or prevent people from becoming a patient in the first place.