EMA qualifies first artificial intelligence tool to diagnose inflammatory liver disease (MASH) in biopsy samples
EMA’s human medicines committee (CHMP) has issued the first Qualification Opinion (QO) on an innovative development methodology based on artificial intelligence (AI). The tool, called AIM-NASH, helps pathologists analyse liver biopsy scans to identify the severity of MASH (metabolic dysfunction associated steatohepatitis; formerly known as non-alcoholic steatohepatitis NASH) in clinical trials.
MASH is a condition where fat builds up in the liver, causing inflammation, irritation and scarring over time, without significant alcohol use or other reasons for liver injury. MASH is linked to obesity, type 2 diabetes, high blood pressure, abnormal cholesterol, and belly fat. If untreated, it can lead to advanced liver disease.
The AIM-NASH tool is expected to enhance the reliability and efficiency of clinical trials for new MASH treatments by reducing variability in measuring disease activity (inflammation and fibrosis).
Following a public consultation, CHMP issued an opinion to qualify this method, which means that the committee can accept evidence generated by the tool as scientifically valid in future applications. CHMP agreed that the tool can increase reproducibility and repeatability in assessments for new MASH treatments. It can help researchers obtain clearer evidence on the benefits of new treatments in clinical trials that include fewer patients. Ultimately, this can bring effective treatments to patients faster.
Testing new MASH treatments often relies on liver biopsies, where small pieces of liver tissue are taken to confirm inflammation and scarring. These biopsies are the gold standard for demonstrating the efficacy of new, investigational medicines. However, high variability in MASH/NASH clinical trials is a challenge, as specialists who review biopsy samples may not always agree on the severity of inflammation or scarring.
The evidence submitted to CHMP shows that AIM-NASH biopsy readings, verified by one expert pathologist, can reliably determine MASH disease activity with less variability than the current standard used in clinical trials, which relies on a consensus by three independent pathologists.
AIM-NASH is an AI-based system that employs a machine learning model trained on more than 100,000 annotations from 59 pathologists who assessed over 5,000 liver biopsies across nine large clinical trials.
The qualified tool is ‘locked,’ which means the machine learning model cannot be modified or replaced. CHMP encourages the optimisation of the model, acknowledging that major changes may require re-qualification of the tool.
All EMA’s activities on AI are coordinated under the multiannual AI workplan by EMA and the Heads of Medicines Agencies, aiming to ensure safe and responsible use of AI across the European medicines regulatory network.