Characterizing the histologic implications of resmetirom-induced liver volume reduction using artificial intelligence-powered digital pathology

In nonalcoholic steatohepatitis (NASH) clinical trials, liver biopsies are evaluated for histologic evidence of 1) NASH and fibrosis at screening, and 2) drug effect at study completion. Liver volume is typically increased in patients with NASH, and an emerging hypothesis suggests that drug-induced fat and potentially related liver volume (LV) reduction may complicate histologic interpretation of liver biopsies by altering the apparent abundance of cardinal histologic biomarkers of disease progression and regression.^1,2,3 It is unclear how this potential effect may impact NASH clinical trial outcomes; in particular, accurate measurement of histologic endpoints could be difficult without an improved understanding of drug-inducedglobal changes in the liver. Here, we used artificial intelligence (AI) models, trained to characterize and quantify cell types and tissues present in NASH biopsies, to investigate whether histological changes resulting from LV reduction could be detected in biopsies from the Phase 2 trial of resmetirom (MGL-3196) for treatment of NASH in patients with NASH CRN Fibrosis Stage 1-3 and NAFLD Activity Score ≥4 with at least a grade of 1 for each component feature. The impact of drug-induced fat loss on cell and tissues, as well as on fibrosis staging, was evaluated.

¹ Patel NS et al., Clin Gastroenterol Hepatol. 2015;13(3):561-568.
² Kromrey ML., et al. Eur J Radiol. 2018;106: 32-37
³ Tai et al., J. Hepatology. 2022;77: S32