Our week of “Greatest Hits” episodes from the vault continues with this conversation from the episode discussing systematic failures of NAS scoring due to inconsistencies in how different pathologists interpret slides. This conversation explores the reasons for different histopathologists differing consistently in how they interpret liver slides.

One of the primary drivers of reassessing the role of semi-quantitative histopathology in assessing developmental drugs was the paper “Complexity of ballooned hepatocyte feature recognition: Defining a training atlas for artificial intelligence-based imaging in NAFLD,” e-published in January 2022 in the Journal of Hepatology. This paper demonstrated strong systematic flaws in the reading of ballooned hepatocytes, which is one of the pivotal factors in assessment drug efficacy. This paper showed that ilow levels of inter-reader consistency led to an array of interpretive challenges. In this episode, last co-author Quentin Anstee led the Surfers, including co-author Stephen Harrison, through a review of the paper and a discussion of how to resolve this issue.

This conversation starts with Stephen Harrison noting that the single largest problem facing drugs in development is the inconsistency in efficacy reads (and particularly placebo response rates) at the back end of clinical trials. From here, the group goes on to explore the ways that histopathology training leads to variability in interpretation of slides and inconsistent assessments of drug efficacy.