Drug-induced toxicity is of considerable concern in drug discovery and development, placing emphasis on the need for predictive in vitro technologies that identify potential cytotoxic side effects of drugs. A label-free, real-time, multiparametric cytosensor system has therefore been established for in vitro assessment of drug-induced toxicity. The system is based on monitoring cellular oxygen consumption, acidification and impedance of human hepatocarcinoma-derived HepG2 cells. The read-out derived from the multiparametric cytosensor system has been optimised and permits sensitive, reliable, and simultaneous recording of cell physiological signals, such as metabolic activity, cellular respiration and morphological changes and cell adhesion upon exposure to a drug. Analysis of eight prototypic reference drugs revealed distinct patterns of drug-induced physiological signals. Effects proved to be rigidly concentration-dependent. Based on signal patterns and reversibility of the observed effects, compounds could be classified based as triggering mechanisms of respiratory or metabolic stress or conditions leading to cell death (necrosis-like and apoptosis-like). A test-flag-risk mitigation strategy is proposed to address potential risks for drug-induced cytotoxicity.