Date: 27 December 2023

Dean Tai, Chief Scientific Officer at HistoIndex, joins Jörn Schattenberg, Louise Campbell and Roger Green to discuss how advances in MASLD artificial intelligence will spur knowledge, insight and innovation in drug development and NITs.

About the Episode

Surfing the MASH Tsunami continues its 2023 wrap-up conversations with HistoIndex Chief Scientific Officer Dean Tai, along with co-hosts Jörn Schattenberg, Louise Campbell and Roger Green. The conversation focuses on growth in the use of AI in Steatotic Liver Disease and some of the insights the profession is developing as a result.

The conversation starts with Dean Tai and Jörn Schattenberg discussing the many promising advances in artificial intelligence over the past year. The two leaders agree both that the advances are providing exceptional new insights already, and will provide even greater insight when we can link AI to outcomes. Roger Green also notes how important it is that we improve the insights we develop from biopsy and, at the same time, the quality of non-invasive tests and suggests that AI has the ability to help on both issues. Finally, from the patient perspective, Louise Campbell suggests that anything that reduces the number of patients we biopsy or the number of biopsies per patient is a significant advance.

Becoming more specific, Dean points out that some of the newer clinical trials are yielding decreases of 30-70% in liver fat density, along with significant decreases in liver volume. Since these measures reduce faster than fibrosis, AI is now giving us the ability to learn more about the process of fibrosis reduction in the aftermath of density and volume declines. Jörn notes that one important issue here involves differences in how pathologists and AI read changes in fibrosis. Specifically, we sometimes overread fibrosis levels in the presence of fast liver fat reductions. He also notes that we can combine these findings with direct fibrosis measures to learn even more and faster. Louise anticipates that this will provide additional benefits if we can fit patient quality-of-life metrics with this data to pinpoint when in the disease treatment process patients begin actually to feel better.

Dean goes on to a second point. AI allows researchers to explore reductions in different regions of the liver in terms of how different regions relate to outcomes. He also points out that in Steatotic Liver Diseases, we see that fibrosis continues in parallel with fibrolysis, which researchers now need to consider in the context of overall disease. Roger mentions two items: that Dean has said in private conversation that in AI-based studies, placebo response usually occurs in ~1/3 of patients and, separately, that bariatric studies suggest that one level of regression might require five years. Dean responds by saying that AI allows us to determine when underlying disease is resolving even if the pathologist considers the patient as presenting with regressive disease.

Roger asks whether AI can be used to assure patients how thoroughly we have studied these drugs and how much we have learned and, separately, to demonstrate more value to payers than we believe using “naked eye” data. Dean suggests the keys will be to simplify the data we present and also to generate a joint statement from the entire community on these issues.

Looking at 2024, Dean anticipates more and richer data than we have. He also cautions that we should shift from biopsy to NITs in clinical trials before we have developed deeper knowledge on how the liver works and what NITs must capture. In this process, he envisions 2024 as a “proof-of-concept” year before we can move to totally non-invasive monitoring in 2025 or 2026. Jörn, who had dropped off the conversation for a few moments, suggested that AI will improve clinical practice over time as well.

In closing, Dean states his concern for 2024 is that we shift too quickly from biopsy to NITs. Louise comments that we cannot discredit biopsy (at least, not yet) and that providers can explain to patients why it is necessary.