AI Tool Achieves Over 90% Accuracy in Predicting Barrett's Esophagus Recurrence
A groundbreaking artificial intelligence (AI)-based tool has been developed to significantly enhance surveillance for patients undergoing endoscopic eradication therapies (EET) for Barrett's esophagus (BE)-related dysplasia and early esophageal adenocarcinoma. Barrett's esophagus is a critical precursor to esophageal adenocarcinoma, an aggressive form of cancer.
Developed and validated by U.S. researchers, the AI model demonstrated over 90% accuracy in predicting which patients would experience a recurrence of BE after EET and precisely when such recurrence was likely to occur. These pivotal findings have been published in Clinical Gastroenterology and Hepatology.
"Early detection of BE-related dysplasia and associated esophageal adenocarcinoma can be life-saving. Identifying recurrence earlier, particularly in high-risk patients who have undergone EET, allows for timely treatment before cancer develops or progresses."
— Sachin Wani, MD, Senior Author and Executive Director, University of Colorado Anschutz Cancer Center's Rady Esophageal and Gastric Center of Excellence
The Challenge of Barrett's Esophagus Recurrence
Endoscopic eradication therapy (EET) stands as an effective treatment for BE-related dysplasia and early esophageal adenocarcinoma. Its primary goal is to eliminate abnormal Barrett's tissue, thereby reducing the risk of progression to esophageal cancer. However, despite successful initial treatment, recurrence of Barrett's esophagus can still occur. Current surveillance strategies are not individualized, applying the same follow-up schedule to all patients regardless of their specific risk level.
Introducing the AI Solution
Dr. Wani, alongside a team of experts, meticulously developed this machine-learning tool. They leveraged AI and comprehensive clinical data from over 2,500 patients who had been treated with EET. The research involved tracking if and when BE, BE-related dysplasia, or cancer returned in these patients. This extensive analysis revealed a significant finding: nearly three in 10 patients experienced a recurrence, averaging approximately two years after therapy.
The sophisticated AI tool was trained to simultaneously analyze a multitude of patient factors, including age, body weight, disease severity, and specifics of their treatment.
Key Predictors of Recurrence Identified by AI
The model pinpointed specific patterns indicating that recurrence was more likely in patients with:
- A longer area of Barrett's tissue
- Higher body weight
- Older age
- A need for more treatment sessions to fully remove abnormal tissue
- More advanced cell changes at the time of diagnosis
The model's robust performance was rigorously evaluated by testing its effectiveness on patient groups similar to its training data, as well as on entirely different patient cohorts from other sources. The tool consistently exhibited high accuracy across both sets of patients.
Revolutionizing Patient Care Through Personalization
This innovative AI tool holds immense potential to transform post-treatment follow-up care.
By enabling doctors to personalize follow-up care, patients identified as higher risk for recurrence could receive more frequent and intensive monitoring. Conversely, those at lower risk might require fewer follow-up procedures, leading to a reduction in unnecessary tests, decreased patient stress, and optimized healthcare resource allocation.
Future Directions and Global Validation
The next critical phase for this AI model involves further validation using international datasets. Collaborations are already planned with institutions in the Netherlands, the United Kingdom, Belgium, and Switzerland. The ultimate objective is to validate the tool for broad, global application, ensuring its reliability as an aid in routine clinical care.