AI machine learning techniques and synthetic data modeling is causing health care professionals to rethink how clinical data evidence can be produced. (GETTY IMAGES)
By Benjamin Ross, Senior Editorial Assistant, AI Trends
“What I’ve been looking at for the past few years is how things are evolving within the …
Contributed Commentary by James Streeter, Global Vice President Life Sciences Product Strategy, Oracle Health Sciences
Clinical trials have changed significantly over the past several years. As drugs and devices—and the conditions they are trying to impact—have become increasingly more complex, so has the design and structure of clinical trials. But protocols are costly to change and identifying and enrolling the right patient cohorts is also no easy feat—especially when rare diseases are the target. So, how are study teams keeping up with this rapid pace of change?
Pharmaceutical companies, biotechs, and CROs have been incorporating technology at variousstages of the trial process to address these challenges; but, ironically, some of these technologies have introduced new challenges such as the sheer volume of data that are being generated.
Information and imagery from clinical visits, digitized voice notes, and the readings streaming in every second from patient-worn devices create a constant flow of data. Megabyte and gigabytes have given way to terabytes, petabytes, and zettabytes, overwhelming the ability of legacy systems and traditional human effort to manage. While it’s possible to capture and store all that information in data warehouses and data lakes, the question remains: How can this much data be cleaned, processed, managed, and evaluated to extract the insights that it holds?
Artificial Intelligencefor Efficient Design
The answer lies in the application of artificial intelligence (AI) and machine learning (ML). AI can not only process data much faster than traditional methods, itcan change the way the dataare used. Along with ML, which will teach the various systems to interpret and comprehend data so that technologies are continuously learning, AI will make it possible to avoid design errors of the pastand create entirely new trials of the future.
To understand the full potential of AI and ML in clinical trials, firstconsider how they can help with protocol design. Having humans manually review prior studies, draw up designs, and handle endless amendments is a time-consuming anderror-prone process.AI, in combination with ML,can do that work accurately and in far less time. By reviewing allthe available historical data quickly and efficiently, these two technologiescan unearthall the problems created from prior protocols and mid-study changes to fully optimize protocol building—ensuring past problems aren’t repeated.
For example, many protocols define an age group that isn’t wide enough to recruit the number of patients needed for that specific indication. With the use of ML, expanded age groups can be identified to optimize recruitment based on learnings from previous studies. Automating the building of trials in this way can eliminateobstacles like mid-study changes, misinterpretation of protocols, human errors,and other issuesthat ultimately delay getting drugs to market. The result is a more precise and better–designed protocol that will likely require minimal changes, potentially saving millions of dollars in costs previously spent on change requests.
Taming the Data Storm
Once the trial is underway, AI becomes crucial to managing the flow of data.
While traditional trials involve collecting data over a series of in-person clinic visits, these events are no longer the only point at which patient data is gathered in a clinical trial.We’re moving to a world where traditional site visits will be supplemented by data from intelligent devices worn by patients that send readings multiple times a day—or even on a continuous basis. In the future, thatinformation could be augmented with external data such as environmental factors like the weather, air quality,a patient’s location or even theiractivity level at any moment.
The useof real-world data in clinical trials represents a huge paradigm shift in the industry. Each enrolled patient could be creating hundreds or thousands to millions of data points a week—or even per day! That amount of data is far more information than humans can process or manage, and outsourcing or throwing more people at the problem is no longer sustainable or effective. But withAI, massive amounts of data can be analyzed in record time. Each piece of informationin large data sets can be scrutinized tocheck for potential problems and compared to statistical norms to reject outliers, while also identifying missing data points—all while being performed at speeds that are not humanly possible.
Insights for Efficacy and Safety
But AI can do more than that. Combined with ML, it can also interpret and draw insights from clinical data. With information coming in on each patient at all times from all sides, AI will become the foundation of data interpretation. Not only can AIdeal with structured data (such as that gathered in form fields), it can also process and interpret unstructured data, such as free-text, audio and visual information. For example, AI can “listen” to an examining doctor’s notes on a trial patient, or even scan images and recognize them.This opens up whole new sources of insight that can be used to inform study teams and help them make decisions about the trial.
Another important benefit of AI and ML’s ability to analyze and recognizedata trends is that it can result in more robust safety reporting. Safety issues can often be subtle or go undetected until they become serious problems. AI can spot trends early on and allow safety teams to respond to them quickly and efficiently.It’s even possible to gatherrelevant safety data from outside the data collected as a part of the trial. For example, posts within patient discussion groups on social media could indicate a potential adverse event. While such surveillance could be done manually, the rate at which this information grows and spreads makes it extremely difficult for humans to keep pace. And while tracking could be done with people performing basic online searches,this approach increases the human resources needed—which can raise overall costs. AI can be used to automate this surveillance process and the triage of incoming safety cases faster and more effectively than a human attempting to do it manually could ever achieve.
Trials of the Future
AI is already being incorporated into advanced cloud-based Life Sciences technology platforms to support trial design, data monitoring, and safety case management; however, we are just at the beginning of the AI era.Within a few years,AI and ML will be able to do things we can only imagine today.
AI is the key to decentralized trials—trials in whicha portion of patient data will come directly from the patients themselves, instead of through traditional methods during site visits. As patient pools grow smaller with “precision” and personalized medicine, patient recruitment becomes more challenging. However, with the expansion of decentralized trials, people who were previously not able to participate in a trial because of their distance from the site will now beable to participate.
We will also see the advent of “patient-less” trials, where we can use historical data rather than live data from people to conduct a trial. While it’s hard to imagine a trial without any patients, it is possible to imagine a trial in which the placebo arm is run on a virtual “placebo” group that is built on historic patient data. This kind of design would not only lower costs but would also make trials more patient-centric and ethical by ensuring allthe recruited patients receive the proposed treatment.
AI and MLwill only continue to develop and improve, especially when applied to cloud-based platforms that can draw on global data sources at scale. As these technologies become more widespread and further embedded into clinical trial platforms, they will usher in a new era of better, more efficient and effective trials that will reducethe cost of bringing new drugsto market, while speeding the process of development—helping to deliverlifesaving drugs to market to patients who are waiting in need.
James Streeter is currently Global Vice President of Life Sciences Product Strategy within Oracles Health Sciences Global Business Unit. James previously held leadership roles at PPD in both operations, as the Global Head of Global Clinical Technical Operations and EDC and recently in IT, as Global Head of Systems Development, Business Operations Teams, and eClinical Strategy and Innovation. He can be reached at firstname.lastname@example.org.