Gastroenterology Advisor
Jessica Nye, PhD
January 25, 2024
Key Point: Although a proof-of-concept study, the findings illustrate AI’s potential for creating phenotypic representations of real-world data that can help identify rapid fibrosis progression in MASH.
Detailed real-world data from organ systems may be used to stratify patient risk for rapid fibrosis progression using an artificial intelligence (AI) approach, according to study results presented at the Therapeutic Agents for Nonalcoholic Steatohepatitis and Liver Fibrosis (NASH-TAG) conference, held in Park City, Utah, from January 4 to 6, 2024.
Patients with metabolic dysfunction-associated steatohepatitis (MASH; previously known as nonalcoholic steatohepatitis [NASH]) who have rapid progression of fibrosis are at increased risk for liver– and disease-specific mortality. However, given the multifactorial nature of MASH, risk stratification is complicated.
Researchers conducted a retrospective cohort study to identify clinical phenotypes associated with rapid fibrosis progression. Data were sourced from the OM1 Real-World Data Cloud, which is a dataset comprising electronic medical records from more than 300 million individuals residing in the United States. Patients (N=1795) diagnosed with MASH between 2015 and 2022 with a fibrosis-4 (FIB-4) assessment within 90 days of diagnosis and at least 2 FIB-4 scores measured at follow-up were assessed for phenotypes of rapid fibrosis progression using the AI-based PhenOMTM platform. Rapid progression was defined by FIB-4 score trajectories from baseline to follow-up, in which low was defined as below 1.30, indeterminate as 1.30 to 2.67, and high as above 2.67.
A total of 175 patients were categorized as rapid progressors (FIB-4 scores of low-high-high: n=6; low-indeterminant-high: n=44; indeterminant-high-high: n=125) and 1620 as nonprogressors (FIB-4 scores of low-low-low). The rapid progressors vs nonprogressors were older (mean [SD] age, 66.3[7.2], 63.5[9.4], 67.3[9.1] vs 54.6[9.1] years, respectively) and more were women (66.7%, 63.3%, 65.6% vs 62.3%, respectively).
“The emergence of CV issues in both absolute and relative signal dimensions demonstrates the potential importance of using these features to distinguish patients with rapid fibrosis progression from others.”
The phenotypes that emerged as different between the rapid progressors and nonprogressors were older age, history of anemia, thrombocytopenia, heart failure, prothrombin time, electrocardiogram, emergency room visits, troponin, infection, sepsis, kidney-associated hospitalization, and cardiovascular conditions.