The publication of Proctor et al. (2017) about prediction of drug-induced liver injury by human hepatocyte spheroid cultures belongs to the most cited studies published in the Archives of Toxicology in 2017. The authors tested cytotoxicity of 110 drugs (hepatotoxic and non-hepatotoxic) and compared EC50 values to the Cmax of the corresponding compounds in human plasma. The margin of safety concept (MOS) defines a compound as positive (toxic) in vitro if EC50 < Cmax × T, whereby T represents an arbitrarily set threshold that typically ranges between 10 and 100. With a threshold of 10, the authors obtained a sensitivity of 36.2% and a specificity of 97.6% (Proctor et al. 2017). Using a threshold of 100, sensitivity and specificity were 59.4 and 80.5%, respectively. Although the study represents an important milestone, these numbers also illustrate a major remaining challenge; even with the very high threshold of 100 a relatively high fraction of in vivo hepatotoxic compounds is predicted as negative by the in vitro test. Several other studies used similar approaches and reported similar limitations (Xu et al. 2008; Khetani et al. 2013; Albrecht et al. 2019; Gu et al. 2018; Frey et al. 2014; Hewitt et al. 2007).

Mechanisms leading to hepatotoxicity are complex (Zárybnický et al. 2018; Hammad et al. 2017; Hoehme et al. 2010); often hepatocyte death is a primary key event (Ghallab et al. 2019; Leist et al. 2017; Sachinidis et al. 2019). However, also toxic effects to cholangiocytes endothelial cells or immune cells may represent initiating events (Godoy et al. 2013; Connolly et al. 2011; Stout-Delgado et al. 2007; Stachlewitz et al. 1999). It has become clear that cytotoxicity represents a highly relevant readout of in vitro tests with cultivated hepatocytes that is sufficient to identify at least 50% of all hepatotoxic compounds. However, for identification of a relatively large fraction of hepatotoxic compounds, cytotoxicity alone is inadequate. In future, additional readouts should be determined in test batteries to study whether they improve sensitivity without reducing specificity. Possible candidates of additional readouts are inhibition of bile salt export (Godoy et al. 2013; Jansen et al. 2017), altered gene expression (Grinberg et al. 2014; Godoy et al. 2016), or secretion of cytokines that may activate immune cells (Wewering et al. 2017). Systematic studies analyzing these additional readouts with respect to sensitivity and specificity are not yet available. However, the already existing cytotoxicity-based studies offer a good basis, since they have identified a relatively large set of compounds, whose hepatotoxicity cannot be sufficiently evaluated by cytotoxicity alone.