In this conversation, the panel discusses challenges in MASH Drug Development that are centered around efficacy endpoints and NAS scoring.

This conversation starts with Will Alazawi suggesting that the MASH clinical trial field suffers from the previous experience with Hepatitis C, where medications became capable of eradicating disease in a fairly linear fashion. He suggests that MASH trials undervalue the value of simply preventing progression, which leads Sven Francque to note that preventing progression to cirrhosis is now accepted by regulators as an endpoint. 

Roger Green refers to last week's episode (S5 – E11), in which Michael Charlton said he would continue Rezdiffra therapy for any patient exhibiting a lack of progression, and that a clinically valuable efficacy measure must be applicable to true clinical practice. Roger goes on to recall Sven’s earlier comment that having the NAS score as a co-endpoint complicates the challenge of proving efficacy because the scale is relatively blunt and not well suited to the task at hand. He suggests that an activity score focused on inflammation and ballooning might function better than one including steatosis. Jörn Schattenberg suggests that this depends on the drug's Mode of Action. 

Sven reminds the group that Stephen Harrison presented a paper at AASLD suggesting the approach Sven describes: splitting NAS into separate steatosis and activity scores, with the activity score based on lobular inflammation and ballooning. Roger asks whether the widely reported challenges in coding ballooning will render this method less valuable. Sven says it might, and states the challenge stems from the lack of detail in the scoring system. 

Louise Campbell suggests that in addition to stabilizing MASH, endpoints might look at related metabolic diseases that poor liver health can affect. Will agrees, noting that many patients may be on other metabolic agents at the outset of a trial or, more challenging, the definition of "good practice" might change during the trial, which can add variability to the sample. He wonders whether differences in placebo rates can provide insight on this issue. Roger recalls a comment from Dean Tai of HistoIndex (S4 - E50.4) that some HistoIndex AI-driven analyses produce consistent placebo rates of ~33%, with efficacy rates far higher.