Clinical research and drug development professionals are largely aware of artificial intelligence (AI), machine learning (ML), and other advanced analytical tools. However, many are not yet aware of their full potential, or how to best put such tools to work.
Lucas Glass, vice president of the IQVIA Analytics Center of Excellence, spoke with Outsourcing-Pharma about how the adoption of AI/ML is evolving, what people in the field need to understand, and what might lie ahead.
OSP: Could you please tell us what the biggest challenge has been facing professionals in your corner of the life-sciences industry?
LG: The biggest challenge facing the industry is user empathy between the technologists and the clinical trial professionals. Professional training for technologists rarely includes user empathy. Machine learning professionals are by nature disruptors, and clinical trial professionals come from a world driven by measures of accuracy.
Predictive modeling in the clinical trial space is rarely at the point where it can fully automate any of the activities. The current state of machine learning in clinical trials is more successful at augmenting the professionals; therefore, the experts who are building the technology need to have a very good appreciation for how the users will interact with their algorithms.
This user empathy requires a solid understanding of the business, the process, and how and why decisions are made. Technology professionals need to keep in mind that this understanding can take years to establish, in some cases.
OSP: What have been the greatest technological advances in predictive modeling and other areas over the course of 2021?
LG: The greatest technological advance this past year is the capability of AI/ML to optimize all stages of clinical trial operations, including:
- How to determine which trial to run using a trial outcome prediction model
- How to select which site to run the trial using a trial site selection method
- How to find which patients to participate in the trial with a patient-trial matching algorithm
For trial outcome prediction, we can predict Phase III trial success with over 80% accuracy by leveraging large data sets of clinical trial descriptions and historical trial outcomes. For trial site selection, we can find trial sites that improve patient enrollment while enhancing patient diversity in race and ethnicity. For patient trial matching, we can automatically match patients to a trial using a large database of patient records and trial descriptions.
OSP: If you could glimpse into your crystal ball, what do you think will be the most significant trend (or trends) to look out for in 2022?
LG: The most disruptive capability coming out in predictive modeling is the ability to model clinical responses to novel molecules. Given it is the fundamental objective of clinical trials, this core capability has many applications in R&D, including:
- Creating synthetic cohorts
- Reducing trial risk by through more accurate outcome measure predictions
- Helping to repurpose medications that are predicted to be effective in new indications.
- Optimizing novel molecules to ensure they are predicted to have the ideal efficacy and safety profiles
