How AI and Computer Vision Can Solve Challenges in IBD Drug Development
Contributed Commentary by Chris Fourment, MD
June 24, 2022 | Inflammatory bowel disease (IBD) is a chronic medical condition affecting millions worldwide, including 3 million in the US alone. The two disease states making up IBD, Crohn’s disease (CD) and ulcerative colitis (UC), represent an enormous market opportunity for the pharmaceutical industry, which spends billions of dollars a year on new drug development. Some of the world’s best-selling drugs are targeted at IBD. Yet, clinical trials can be costly and often fail to finish on time, causing continued suffering for patients who desperately need new therapeutic advancements. Furthermore, CD and UC can be challenging conditions to treat.
Advances in artificial intelligence (AI)-driven algorithms and computer vision can address bottlenecks that sponsors and physician investigators face in developing therapeutic options and matching IBD patients with a clinical trial. Computational solutions also have the exciting long-term potential to disrupt the current paradigm for diagnosis and therapeutic monitoring, which relies heavily on manual interpretation of endoscopy.
AI Solutions Can Improve Clinical Trial Recruitment
Problems with IBD drug development are partially due to challenges inherent in the current diagnostic workups and therapeutic monitoring standards. Such standards are mainly based on clinical and endoscopic assessments, with endoscopy necessary for evaluating disease activity and mucosal healing.
Disease severity endoscopy-based scoring is an important criterion for determining patient eligibility for clinical trials and a key endpoint for evaluating therapeutic response in clinical research and care. Physicians use various scoring methods to make this assessment, such as the Mayo Endoscopic Score for UC (MES) and the Simple Endoscopic Score for Crohn’s Disease (SES-CD). MES is a four-point scoring system in which patients are identified as having normal, inactive, mild, moderate, or severe levels of disease severity ranging from 0 to 3, respectively. The SES-CD is a similar scale tailored to the unique characteristics of Crohn’s Disease.
However, endoscopic readings today are fairly subjective and highly dependent on practitioner experience and training. The variability of endoscopic interpretation of disease severity makes clinical trial recruitment a time-consuming challenge for referring clinicians and pharmaceutical companies, adding to the complexity and timelines needed to determine results/endpoints in clinical trial research.
Today, investigators address this variability by requiring multiple IBD experts to score disease severity, but the time involved may delay efficient interpretation and consistency due to inter-rater reliability. Moreover, a critical need exists to improve the correlation between existing endoscopic and clinical assessment scales and individual prognosis and outcomes.
AI solutions can automate several challenging and cumbersome steps in endoscopy readings, leading to improved processes for patient enrollment in IBD clinical trials and more consistent endpoints for clinical research. For example, software can automate the calculation of a patient’s minimal threshold (minimum) score of disease severity, as measured by the MES or SES-CD, as a clinical research aid in determining pre-screening qualification for IBD clinical trials. This approach enhances the likelihood of patients meeting the sponsor’s disease severity inclusion and exclusion criteria when referred for a clinical trial.
The ability to determine patient qualification through prescreening may improve the patient referral process, ensuring all eligible patients are considered for participation. Only eligible IBD patients start the clinical trial enrollment process as defined by the MES or SES-CD.
Further downstream in the clinical trial process, computational solutions have the potential to refine diagnostic and monitoring inputs, shortening the development path for IBD drugs to market. Novel endoscopic scoring modalities developed from AI innovations could be designed to be more consistent, accurate, and reflective of the individual nature of GI diseases.
More Meaningful Assessment of IBD in Clinical Care
In the clinical setting, AI-driven solutions, developed in collaboration with leading data scientists and GI experts, can potentially disrupt the current paradigm for measuring treatment response. By identifying new predictive endpoints, gastroenterologists could better assess therapeutic response and disease progression over time.
Successful applications of AI-driven technologies have the potential to bring precision medicine to gastroenterology. This specialty has not benefited from these advances as much as other specialties like oncology and infectious disease. The novel insights that can come from AI applications have enormous potential to drive improvements in diagnosis, treatment, and access to high-quality care for millions living with these complex and chronic serious illnesses.