While clinical oncology trials were maintained during the COVID-19 pandemic, researchers turned to a wealth of data to find patients across the country who would qualify for the trials, even if they were not physically there.
Artificial intelligence has made this process possible and may have moved towards decentralized trials that could potentially last long after the pandemic ends.
Jeff Elton is the CEO of ConcertAI, which works with some of the largest oncology pharmaceutical companies and research organizations. IT health news he interviewed Elton to outline his thoughts on this shift and what it means for both treatment and patient outcomes.
Q: While the trials are pending, the researchers worked on all of this data to find patients who would qualify for the trials, even if they were not physically there. How did artificial intelligence technology make this possible?
A: By releasing data into cancer centers to work. We process structured and unstructured data – combing EHRs as well as other sources of patient information that EHRs may not include. Natural language processors and other tools that are an integral part of the workflow are critical here.
Clinical environments have a wealth of data. When participation in trials declined, they had to use all the data at their fingertips quickly and efficiently to find as many potential eligible patients as possible. People who work manually would take too long and might overlook something. AI could do that. AI improves the ability to identify patients who qualify for clinical trials.
It is a complex process. We need to eliminate false negatives, which means that if a patient potentially qualifies for a clinical trial, we identify them. We also ensure that we do not have too many false positive results. Otherwise, we are just creating work.
We also use AI tools to ensure we see what we expect and need in clinical settings data – tools for detecting and reporting exceptions and anomalies are key to recognizing and understanding accurate data.
It is crucial to understand that if there is no data there is no AI either. Significant AI and machine learning capabilities require broad data access, the ability to prepare data for specific AI methods and tools, and reserved data for independent verification. Of course, we must also be vigilant in basically health and biological trends for retraining or re-specifying AI models.
We can also generate evidence from complementary data from retrospective sources for prospective studies – and sometimes retrospective data only for label extensions.
The FDA is increasingly accepting studies with retrospective data that have been replaced for pre-recruited patients in standard care controls as “external control arms”. This change is in the best interest of patients and allows for more efficient study, as patients can be recruited exclusively into the therapeutic arm with a new therapeutic agent.
Q: Has AI initiated a shift toward decentralized clinical trials, a move that could potentially be sustained long after the end of the pandemic?
A: We are not going backwards. Decentralized trials have taken place over the last few years. COVID-19 was an event of stumble or shock that accelerated the trend.
Decentralized testing, by the way, does not require artificial intelligence at all, but it can take advantage of artificial intelligence given that workflows are all digital and most data is machine readable. We will enter a period in which decentralized trials are extensive, in parallel with inherited approaches.
But it will only exist for a temporary period – possibly only in digital form – with a deeply embedded AI … the only approach. I use the term “integrated digital testing” to describe what follows.
In addition to integrated digital trials, clinical studies are an integral part of the nursing process itself, as opposed to being imposed on them. Tests should not burden service providers and patients from standard care.
This thing is incredibly important. Reducing the burden that trials place on patients and service providers allows us to move clinical trials to a community where 80% of patients receive care. It is both democratization and the ubiquity of clinical trials.
Q: What does this shift mean for both treatment and patient outcomes?
A: This is all good. This is good for patients, first of all, because they can participate in trials in a wider range of treatment settings. This is good for treatment innovation because more alternatives are available for studies in multiple environments with lower barriers to participation.
Standard treatment for a new therapy relative to a separate clinical trial should increase the likelihood of a positive clinical outcome. We want to provide patients with more potentially useful options, faster and with greater precision.
Q: Share an anecdote of your work over the past year with pharmaceutical companies and research organizations about how AI has improved or enhanced oncology clinical trials.
A: One of our partners had a study that could not collect patients. The trial sponsor wanted our tools, clinical sites, and data to solve their problem. Yes, but it turned out that the problem was a trial design that could not be executed. Our study design solution optimized for intelligence found a problem. No insight was expected, but it was worth it nonetheless.
More importantly, we and our sponsoring partners over the past year have reaffirmed our commitment to addressing research differences that are sometimes at the root of health and other inequalities.
We have successfully combined our combination of rich clinical data and AI optimizations to review clinical trial design to ensure diversity, avoid unintentional exclusions, and identify locations and investigators that can ensure study success and timeliness of completion.