FDA Qualifies AI Tool to Support MASH Clinical Trials

FDA qualification of an AI pathology tool could streamline MASH trials by improving consistency, hinting at broader acceptance of AI in drug development

12 Dec 2025

A small change at America’s drug regulator hints at a larger shift in how medicines are tested. For the first time, the Food and Drug Administration (FDA) has formally qualified an artificial-intelligence tool to help measure outcomes in a major disease area. It is a narrow decision, but one with wide implications.

The tool, known as AIM-NASH, is designed for clinical trials in metabolic dysfunction-associated steatohepatitis, or MASH, a liver disease closely tied to obesity and type 2 diabetes. By qualifying it, the FDA allows drugmakers to use the system to interpret liver biopsies across trials without having to prove its worth anew each time.

That matters because MASH is notoriously hard to study. Patient numbers are rising fast, but drug development has been slow and costly. Many late-stage trials fail, often not because treatments do nothing, but because proving they work is difficult. Much hinges on how liver damage is measured.

At present that task falls to expert pathologists, who examine biopsy samples under a microscope. The process is slow, expensive and subjective. Two specialists can look at the same slice of tissue and disagree on the severity of disease. To overcome that noise, trials must recruit more patients or run longer, driving up costs.

AIM-NASH promises more consistency. Built by PathAI, a firm specialising in AI-driven diagnostics, the system uses machine learning to analyse digital images of biopsies and apply standard scoring rules in a uniform way. The FDA’s qualification signals that, for this purpose, the tool is accurate and reliable enough to be trusted.

Regulators are keen to play down any talk of automation run amok. The biopsies are still taken, and humans remain in charge. The AI is meant to assist, not replace, existing practice. Still, the decision reflects a cautious willingness to embrace digital tools that reduce variation without weakening standards.

For drug developers working on metabolic diseases, the benefits are prosaic but valuable. Clearer data could mean shorter trials, lower costs and results that are easier to interpret. In a field crowded with failures, even incremental improvements can tilt the balance.

The change is unlikely to grab headlines. But by giving one algorithm a regulatory seal of approval, the FDA has opened the door, just a little, to a different way of running trials.