Clinical trial sites, sponsors and contract research organizations (CROs) are keenly interested in ways to improve the design, operation and results of studies. A number of organizations are exploring artificial intelligence (AI), machine learning (ML), and other advanced technology, and how they can positively impact trials.
Outsourcing-Pharma (OSP) recently talked with Jennifer Bradford (JB) head of data science with Phastar a global CRO offering trial reporting, data management, data science and other services around the globe. She told us now AI and machine learning are making a difference in trials, and how such technologies change the future of clinical research.
OSP: What are some of the key ways in which AI and ML have been impacting the way clinical trials are run?
JB: The most successful and documented uses of AI in clinical trials have been for target identification and drug repurposing, given the large volumes of structured data available in this area. ML-based predictive analytics can also be used in recruitment and retention activities, for example, identifying the right candidates at a faster rate, which can accelerate R&D timelines.
During a trial, applications of AI are also emerging from a clinical operations perspective; supporting data management teams to automatically detect erroneous data. For example, in risk-based management, going beyond a rule-based approach applied to the clinical and meta data to identify problematic sites or even patients based on patterns of behavior, identification of outliers etc.
As clinical studies continue to become more complex, for example by more complex study designs or through the use of remote monitoring technology, it is important that the data generated is used in the optimal way during the trial. Powerful ML technologies have the potential to monitor this data as it is generated; identifying issues and inconsistencies as trials are ongoing.
With remote monitoring we have the potential to continuously monitor many different measurements from a patient as they go about their everyday lives. This in turn will generate large volumes of data, data that would be near impossible for a clinician to monitor and analyze across a number of patients on a regular basis.
ML technologies could be used here to flag to a medical team certain changes, potential issues or anomalies for a particular patient, directing the medical team to take further action if they deem it necessary following review of the data.
Most trials now include data re-use in their consent process, which in turn will expand the volume of data surrounding different treatments and the effects on humans. As the volume of high-quality data continues to increase, and become more accessible, we will likely see an increase in the vital role of AI in evidence-based decision making as a trial is run.
OSP: What sort of learning curve does implementing AI and ML create—both for sites and sponsors, and for patients?
JB: Generally across the industry, AI and ML create a big learning curve. For all consumers of AI technology, whether that be a sponsor, medical teams at a site, or patients, it is important that they understand the limitations of the technology and the context in which it can be used; that AI is only ever as good as the data from which it was built and it may not always have all or even the right answers.
For decision makers, those that may use the output of AI technologies, they of course have the challenge of acting on results. AI after all is a machine, not a human; It has no emotions or empathy and won’t consider the ethical implications of a decision.
These decision makers must learn to use these technologies in the right context; in the way they were intended but always applying their own reasoning and intelligence to augment and critically evaluate the output of the AI.
For sponsors and sites these technologies may have a large impact on existing processes and ways of working and this, of course, must be considered when introducing them. Some individuals may feel that some technologies may replace their own role, others need to understand the potential benefits and have available the appropriate training in order for these technological advancements to be embraced successfully.
This goes for patients, too, as these technologies will increasingly rely on digital technology through the use of smart phones and other devices. Patients will need access to information, education, and support much like employees as disengaged patients can have a big impact on the success of the trial.
I think overall these technologies will be embraced and while we need to ensure that the applications are carefully considered and regulated, there is potential for AI and ML to bring us benefits in ways we cannot yet imagine.
OSP: Could you please tell me about some of the key benefits of using such technology?
JB: There are already some excellent published examples of AI applications and specifically machine learning within clinical trials and across healthcare. We see examples of sophisticated ML applications successfully applied to image analysis, for example classifying tissue images as normal or abnormal, and to drug discovery where these approaches reduce the R&D times, reducing the cost of these processes and also, potentially, getting treatments to patients more quickly.
But more generally these technologies provide us with new opportunities to extract more from the data we have now and will generate in the future. Rather than replace humans, this technology can be used to empower experts, enhance the information available to them to support evidence-based decision making for diagnosis, treatment decisions, and enhancing operational aspects of care.
Unlike a human, a machine can take large volumes of data and sift through it at an alarming rate and without the inherent bias of a human, therefore giving a view of the world based on the data it has seen quickly and efficiently. This is critical for fast decision making and ultimately saving money and possibly even patients’ lives.
OSP: Are there any detractions or concerns for using such technology?
JB: If we think of learning as having different levels starting at the very basic level with memorizing and understanding data as knowledge, both humans and machines have the capacity to do that. As learning becomes more complex; how that knowledge is applied or analyzed in the context of a particular problem, then humans have the advantage.
AI can only ever draw from the data provided to it, where as a human can link it with their external social awareness and apply intuition; therefore, when and where to apply such technology is an important consideration. In healthcare in particular, simple decisions may be handled by ML but more complex tasks that require ethical, emotional or contextual awareness will require human input.
We must also be mindful of how to maintain these technologies. A human continues to learn; if we think of a doctor for instance, they have their initial training and they then continue to learn and expand their knowledge as they perform their job, communicate with colleagues and read the scientific literature.
A machine should be no different, it must continue to learn and adapt. This latter point brings additional challenges from a regulatory perspective; the need for new rules to allow these systems to evolve to a certain extent after initial approval whilst managing the risks of this evolution.
A key consideration is, What happens if a machine gets something wrong? The consequences of this, particularly in healthcare but also in other industries, has the potential to be life-threatening. Who is responsible; is it the person designing the algorithm? The consumer of the algorithm? The manufacturer? These questions need to be thought through and addressed as these technologies become more widely adopted.
Finally, as with anything, such technology can also be abused, and we must be mindful of this when adopting AI or even releasing data for these algorithms. Could the technology be used for harm and how do we protect against that?
OSP: Is there anything you would like to add?
JB: I believe we are now starting to the see the shift from ML and AI being buzzwords to them being a reality in the clinical space. I have been involved for some time now in real projects that apply this technology to clinical datasets to generate new insights and benefits.
In general, AI and ML technologies have huge potential across the pharma industry. We have already seen a number of exciting applications and this will continue to grow, enabling us to speed up drug development timelines, getting the right treatment to the right patient faster.
Ultimately, with the volume of usable data on the increase and with digital technologies becoming more sophisticated and more commonplace on clinical trials, these technologies will be essential if we are to get the most of the growing volume of data and provide earlier evidence-based decision making.