A large number of drugs and compounds produced by the chemical and agrochemical industry, often referred to as 'non-genotoxic carcinogens' (NGC), score as tumour promotors in rodent models. It is unclear whether these compounds act similarly in humans. The most extensively investigated compounds have been the anti-convulsive drugs, phenobarbital (PB), and phenytoin. Liver tumours induced by PB are dependent upon the activation of the constitutive androstane receptor (CAR). However, marked species differences in CAR activation by exogenous chemicals exist with some being much more potent activators of human CAR, e.g., 6-(4-chlorophenyl)imidazo[2,1-β][1,3]thiazole-5-carbaldehyde-O-(3,4-dichlorobenzyl)oxime (CITCO). We have compared CITCO-induced tumour formation in the livers of mice in which murine CAR has been replaced with its human counterpart. Our findings reveal that CITCO-dependent liver tumours are only formed in mice-expressing human CAR and not in wild-type animals. In addition, contrary to one of the proposed mechanisms of NGC carcinogenicity, we show that CITCO did not induce a hyperplastic response in the liver of the humanised mice. These data raise some key questions about the mechanism of action of NGCs and identify the limitations of current rodent carcinogenicity test systems in relation to risk assessment.