Drug development for non-alcoholic fatty liver disease (NAFLD) is frequently hampered by the poor translation of preclinical findings into clinical efficacy. To address this critical challenge, we developed a quantitative cross-species model designed to predict human clinical outcomes from efficacy data in mouse models.

We performed a model-based meta-analysis (MBMA) of 18 NAFLD drugs, integrating data from published clinical trials with corresponding preclinical mouse studies identified through a systematic search of the Embase and PubMed databases. Using the change in alanine aminotransferase (ΔALT) as the primary biomarker, we constructed an exponential model to define the relationship between ALT reduction in mice and the placebo-corrected response in humans (ΔΔALT). The model's predictive performance was then externally validated using an independent dataset from a study of Linggui Zhugan Tang (LGZGT).

The analysis yielded a robust exponential model, which revealed that a reduction in mouse ΔALT of at least 53.3 U/L is required for a drug to show superiority over placebo in human trials. A more substantial decrease of 128.3 U/L in mice predicted a clinical efficacy exceeding that of Resmetirom, the first FDA-approved therapy for this condition. The model's predictive power was successfully confirmed through external validation with the LGZGT data.

This study developed a cross-species efficacy model from NAFLD clinical and mouse data, revealing an exponential relationship between human and mouse ALT levels. This provides quantitative thresholds for preclinical screening to improve drug development success rates.