AI Tool Accurately Predicts Barrett’s Esophagus Recurrence

A novel artificial intelligence (AI) tool demonstrates significant potential in enhancing surveillance for patients who have undergone endoscopic eradication therapy for Barrett’s esophagus (BE) linked to dysplasia and early esophageal adenocarcinoma. BE is the sole known precursor to esophageal adenocarcinoma, a highly aggressive cancer associated with considerable mortality rates.

Developed and validated by researchers in the United States, this AI model boasts over 90% accuracy in predicting which patients are likely to experience a recurrence of BE post-therapy and estimating when this is likely to happen.

The results of this research were published in ‘Clinical Gastroenterology and Hepatology.’

“Early detection of Barrett’s esophagus-related dysplasia and associated esophageal adenocarcinoma can save lives. Identifying recurrence in the form of BE, BE-related dysplasia, and BE-related esophageal adenocarcinoma sooner—especially in high-risk patients who have undergone endoscopic eradication therapy—creates opportunities for timely intervention before cancer develops or advances.”

Sachin Wani, MD, study’s senior author and executive director of the University of Colorado Anschutz Cancer Center’s Rady Esophageal and Gastric Center of Excellence

Endoscopic eradication therapy (EET) effectively treats BE-related dysplasia and early esophageal adenocarcinoma by eliminating abnormal Barrett’s tissue and significantly lowering the risk of cancer progression.

“The challenge lies in the fact that recurrence of Barrett’s esophagus can still happen even after successful therapy. Current surveillance strategies do not differentiate between high and low-risk patients; everyone is typically monitored using the same schedule,” stated Wani.

Utilizing artificial intelligence and data from over 2,500 patients, Wani and a team of distinguished experts from various institutions developed a machine-learning tool. They analyzed comprehensive clinical data from patients who received EET and were followed for a duration to monitor whether and when BE and its related conditions re-emerged. This investigation revealed that nearly 30% of patients experienced recurrence post-treatment, with recurrence occurring approximately two years after therapy on average.

The AI tool was trained to assess multiple patient factors simultaneously, such as age, body weight, disease severity, and treatment specifics. It identified patterns that may not be visible to human analysis, particularly how various combinations of factors influence risk. They found that recurrence was more common among patients exhibiting:

• A larger area of Barrett’s tissue
• Higher body weight
• Older age
• The need for more treatment sessions to eliminate abnormal tissue
• More advanced cellular changes at the time of diagnosis

The model underwent testing in two phases: evaluating its effectiveness on patients resembling those in the training dataset and assessing its performance on different patient groups from other datasets. The tool proved accurate for both cohorts.

This innovative tool could assist healthcare providers in personalizing follow-up care after treatment, allowing for tailored monitoring schedules. Patients deemed at higher risk for recurrence could receive more frequent evaluations, while those at lower risk could require fewer follow-up procedures. This strategy could minimize unnecessary testing, alleviate anxiety for patients, and optimize healthcare resource allocation.

“This project is the result of several years of joint efforts across various institutions. It would not have been feasible without the collaboration of colleagues who contributed their data and expertise,” said Wani.

Collaborators include experts from Johns Hopkins University, Mayo Clinic, UZ Leuven, University of North Carolina at Chapel Hill, Washington University School of Medicine, Cleveland Clinic London, Northwestern Feinberg School of Medicine, University College London, University of California Los Angeles, University of Kansas, and Hirlanden Clinic Zurich.

The next step involves further validating the model using international datasets through partnerships in the Netherlands, the United Kingdom, Belgium, and Switzerland. The ultimate aim is to establish the tool for widespread application and ensure its use as a reliable resource in clinical settings.