AI powers prediction tool for condition associated with bowel disease

Using machine learning and a large clinical dataset, an international research team led by the Mayo Clinic has developed a tool that predicts outcomes in patients with primary sclerosing cholangitis.

The disease of the bile ducts, which is commonly associated with inflammatory bowel disease and ulcerative colitis, can lead to serious liver damage.

The problem, according to Mayo Clinic gastroenterologist John Eaton, MD, is that primary sclerosing cholangitis is a “progressive liver condition, and researchers as well as clinicians need better prognostic tools to assess patients’ outcomes.”

Also See: Mayo expands use of image-guided liver treatment platform

To address the shortcoming, the Mayo Clinic—along with researchers from the Norwegian Primary Sclerosing Cholangitis Research Center—leveraged machine learning to analyze the available clinical data regarding patients with the disease and, in the process, discovered and validated a tool that predicts outcomes at five years.

The Primary Sclerosing Cholangitis Risk Estimate Tool (PREsTo) that researchers developed has nine variables: bilirubin, albumin, serum alkaline phosphatase times the upper limit of normal, platelets, aspartate aminotransferase, hemoglobin, sodium, patient age and the number of years since primary sclerosing cholangitis was diagnosed.

The model, created through a machine learning technique called gradient boosting, was derived using 509 patients from a multicenter North American cohort and validated in an international multicenter cohort of 278 patients. A study, published in the journal Hepatology, had an end point of hepatic decompensation—a condition that occurs when there is a deterioration in liver function.

“We found that PREsTo accurately predicts hepatic decompensation in primary sclerosing cholangitis patients and exceeds the performance of other widely available, noninvasive prognostic scoring systems for this disease,” says Eaton, who is the study’s first author. “We are proud to have developed an objective tool that can be utilized in clinical research and to provide more accurate estimates of liver decompensation in five years for patients with primary sclerosing cholangitis.”