Will there ever be an in vitro assay that is predictive for human hepatotoxicity?

Advances in novel alternative methods (NAMs)-including in silico modeling, hepatic co-cultures, 3D models such as spheroids and organoids, and microphysiological liver platforms-have greatly improved our ability to study human hepatic biology and reduce reliance on animal testing. However, whether such systems can predict human hepatotoxicity remains doubtful. Low-incidence, idiosyncratic liver injury typically arises from patient-specific factors and complex immune interactions that are not captured in simplified or genetically limited in vitro settings. Temporal constraints, limited biomarker diversity, and inconsistent validation further hinder predictive claims. Even a "perfect" in vitro liver model would still represent the biology of a single donor, comparable to a clinical study with one subject. These systems are invaluable for mechanistic and metabolic investigations but should be regarded as investigative-not predictive-tools for hepatotoxicity risk assessment.