How Noninvasive Testing Could Change MASH Clinical Trials

Dr. Winston Dunn discusses how noninvasive testing, fibrosis-focused approaches and machine learning could shape patient screening in MASH trials, along with the growing need to match treatments to different patient profiles.

Metabolic liver disease research is changing how clinical trials find and evaluate patients. In metabolic dysfunction-associated steatohepatitis (MASH), this includes questions around who qualifies for a study, how fibrosis can be identified without a biopsy and how treatment choices may differ from one patient to the next. Noninvasive testing (NIT) and machine learning are becoming increasingly important in those discussions.

The treatment landscape has changed too. The first FDA-approved treatment for MASH arrived in March 2024, followed by a second accelerated approval in August 2025.

MASH is a more advanced form of metabolic dysfunction-associated steatotic liver disease (MASLD), in which excess liver fat is accompanied by inflammation and liver injury and can lead to fibrosis, or liver scarring.

In this Xtalks Spotlight feature, Winston Dunn, MD, FAASLD, Professor of Medicine in the Division of Gastroenterology at the University of Kansas Medical Center, Summit Clinical Research Solutions and Pinnacle Clinical Research – San Antonio, discussed the gap between biopsy-selected clinical trial populations and patients treated in routine care, the potential for fibrosis-focused and noninvasive trial approaches and how machine learning could improve liver disease risk stratification.

Dr. Dunn also shared how the treatment conversation is shifting as new options become available.

Why Real-World Evidence Has a Role After MASH Clinical Trials

MASH clinical trials have commonly used liver biopsy as part of their inclusion criteria. According to Dr. Dunn, this can produce high screen-failure rates.

“That means 80 to 90% of the patients are not represented,” he stated.

Patient identification can change once a drug receives FDA approval. Dr. Dunn noted that treatment in clinical practice is almost always guided by NITs instead of liver biopsy.

“The cohort is quite different from the original cohort that is based on liver biopsy,” he explained.

Dr. Dunn said this difference between trial and real-world patient populations emphasizes the need for real-world evidence to validate clinical trial findings after a treatment enters practice.

Could Fibrosis and Noninvasive Tests Play a Larger Role in MASH Trials?

Dr. Dunn pointed to research presented by Dr. Zobair Younossi at the European Association for the Study of the Liver (EASL) meeting, in which Dr. Younossi described liver fibrosis as the only predictor of liver-related outcomes.

He also linked some trial screen failures to assessments of MASH and hepatocellular ballooning, citing interobserver and intraobserver variability.

“I am hoping to see more clinical trials that are based on [biopsy-confirmed] fibrosis only and less reliability on MASH in the clinical trial or having a cohort two that is based on fibrosis only,” Dr. Dunn said.

He also hopes to see more studies that use noninvasive measurements without requiring liver biopsy. Dr. Dunn referenced Eli Lilly’s major adverse liver outcomes (MALO) trial and its use of liver stiffness measurement (LSM) and the Enhanced Liver Fibrosis (ELF) test.

Screening tools can also affect how many patients ultimately qualify for a trial. Dr. Dunn cited another study in which using aspartate aminotransferase (AST) alone resulted in a 60% screen-failure rate.

He said the field should move toward tools such as ELF, LSM and machine-learning approaches and “replace the more blunt instrument, such as AST alone, that results in a lot of screen failures.”

Machine Learning Could Change How Liver Disease Risk Is Assessed

Dr. Dunn contrasted newer machine-learning approaches with risk-stratification tools developed when clinical calculations needed to stay relatively simple.

The Fibrosis-4 (FIB-4) index is a widely used score that helps estimate the likelihood of liver fibrosis using routinely available clinical information. Dr. Dunn shared that FIB-4 was developed to be simple enough for clinicians to calculate easily.

Newer computational approaches can use more complex combinations of clinical information without asking clinicians to perform those calculations themselves. Machine-learning tools such as ALADDIN and LiverPRO can be accessed through web-based calculators, Dr. Dunn noted.

“More complex and accurate tools are based on machine learning tree-based techniques that are far more accurate,” he said.

Dr. Dunn also briefly discussed AI tools such as ChatGPT and OpenEvidence, which he said can help clinicians access information and support timely diagnosis and decision-making in practice.

As MASH Treatments Expand, Patient Needs Are Coming Into Focus

Dr. Dunn also sees important unmet needs in MASH treatment, particularly because patients can have different metabolic and clinical profiles.

Some patients may need treatment that addresses lipid levels, cardiometabolic risk factors, obesity and MASH at the same time. Other considerations include tolerability, preserving muscle in patients with obesity and access to both oral and injectable options.

“Most [MASH] patients eventually die of cardiometabolic complications,” Dr. Dunn said.

He also identified existing cirrhosis, alcohol-associated liver disease and metabolic dysfunction and alcohol-associated liver disease (MetALD) as areas with ongoing treatment needs. Dr. Dunn emphasized the need for tailored treatments that can meet the individual patient’s needs.

“A few years ago, the question was if we will ever have a treatment for MASH, and then we have two accelerated approvals for MASH, and then the question has changed.”

“Now, our question is, ‘What is going to be the ideal treatment for MASH? What would be the ideal treatment for each patient with MASH, and how do we optimize the treatment for each patient with a very different need?’” he concluded.

This article was created in collaboration with the sponsoring company and the Xtalks Editorial team.