Wood discusses how AI is being used to improve the process of finding patients for clinical trials.
Susan Wood, CEO of VIDA, an imaging intelligence company, spoke with MD&T about her company’s work with AI and using it to develop solutions for clinical problems. VIDA has recently been on clinical trials and discovering ways to improve equality inefficiencies.
(MDT:) What sort of issues with clinical trials are you experiencing when it comes to diversity?
Wood: The diversity piece of it is, for good reason, had heightened awareness. Part of that is because of the COVID-19 pandemic which laid bare some of those inequities in corporations and clinical trials. Skewed data has the potential to represent skewed results. If you don’t have a more generalized database of populations included into a clinical trial, you are at risk of having the outcomes of those represented groups being skewed. We focus a lot on using our large database of disease specific evidence and being able to mine that data for more generalized subject and subject classes to train the algorithms in a more generalized way.
To have more generalization and diversity in clinical trials is a larger problem of communication and awareness. Just being aware of participation in clinical trials can help. I’ve been to the doctor for decades and no one has ever asked me to participate in a clinical trial, either as a control or for having a disease. Having awareness of trials and the opportunities that are out there is part of the issue. Creating a network of clinical trial sites in underrepresented communities, because these groups tend have a greater prevalence of disease, is important.
Our approach is to have a more sustainable platform for support of clinical trials and building a data infrastructure so we can manage and find real world data on the populations that could and should be included in clinical trials. We pull from normal clinical workflows to mine them for prospective subjects that match into clinical trials.
(MDT:) What issues have you face building an expansive network of trial sites?
Wood: The trial site issue has been a big one. COVID and post-COVID has transformed the market and the industry, but the resources at these sites were always, and especially during COVID, were stretched very thin. It’s been key to make everything as easy as possible to implement clinical trials for these valued, but sparse, resources. We used this opportunity to build a trial platform to accommodate the decentralization of clinical trials. This takes trials out of the major academic centers and puts them in decentralized areas. To accommodate that, we need better levels of technology.
We train the decentralized sites or imaging centers to do imaging for clinical trials. We train them with E-learning in the cloud so that it’s on their schedule. Those are some of the kinds of challenges we’ve faced with trial sites. It’s been a reaction the market and how it’s changed for decentralization.
Some of the more local challenges have been to implement imaging in a clinical trial, which we want to make as simple and easy as possible. They have to adhere to a protocol that’s fundamental to high-quality imaging results. We train the sites to do that and maintain consistency. AI is used to normalized longitudinal data.
We’ve faced challenges, but nothing that can’t overcome.
(MDT:) In regards to organizing clinical trials, how big of a role will AI play in pharma’s future?
Wood: The development or change in pharmaceutical’s future is that there’s $2.8 billion per drug on average, with a 90% failure rate. It is a process that is ripe for change and greater efficiency. One revolution is using AI to develop the compound in general. They mine through troves of data on the human body and how it will react to certain molecules, and AI can be used to develop the compound. However, AI is also useful for better modeling of the trials and the outcome measures: who should be included, how it should measured, where the trial should be executed, how are representative populations found, and much more. AI enabled precision biomarkers can improve the timeline by finding disease indications and responses much earlier in the process. It can also provide greater efficiency, improved patient outcomes, and more. The biobank that we have has been collecting information for decades and be used to have a digital control.
After COVID, there’s just not as many patients out there that we can find. Instead, we can use data to be a digital control arm and have an understand a patient’s history of disease and how they’ll respond to different therapeutics.
In general, as a broad-based response, AI brings pharmaceutical companies closer to their patients. We’ve got better data and better information on patients and how they may respond.
(MDT:) How do you approach improving direct patient-to-trial matching?
Wood: One of the big pain points in pharmaceutical trials is just finding the right patients. We have a biobank of information that is a highly curated dataset of disease information. We’ve been collecting over time and that information is robust. It helps us to better understand what the requirements for inclusion in the trial should be, the best metrics for evaluating patients, and who should be participating. Understanding this historical information is a key component of trial-matching.
We have better data on how participants can be found, both prospectively and retrospectively. We can use AI and quantitative imaging to find patients that have certain attributes that we know should be included in the trial.