Current limitations and future opportunities for prediction of DILI from in vitro.

Drug-induced liver injury (DILI) is a major concern for drug developers, regulators and clinicians. It is triggered by drug and xenobiotic insults leading to liver impairment or damage, in the worst-case liver failure. In contrast to acute "intrinsic" hepatotoxicity, DILI typically manifests in a very small subset of the population under treatment with no clear dose relationship and inconsistent temporal patterns and is therefore termed an idiosyncratic event. Involved are multifactorial, compound-dependent mechanisms and host-specific factors, making the prediction in preclinical test systems very challenging. While preclinical safety studies in animals usually are able to capture direct, acute liver toxicities, they are less predictive for human DILI, where specific, human-derived in vitro models can potentially close the gap. On one hand, mechanistic approaches addressing key mechanisms involved in DILI in well-characterized and standardized in vitro test systems have been developed. On the other hand, co-cultures of different cell types, including patient- and/or stem cell-derived cells, in a three-dimensional setup allow for prolonged incubations and multiplexed readouts. Such complex setups might better reflect multifactorial human DILI. One major challenge is that for many compounds with human DILI the underlying mechanisms are not yet fully understood, complicating establishment and validation of predictive cellular tools. A tiered approach including rapid mechanism-based in vitro screens followed by confirmatory tests in more physiologically relevant models might allow minimizing DILI risk early on in vitro. Such complex, integrated approaches will gain from larger collaborations in multidisciplinary groups bringing existing knowledge and state-of-the-art technology together.