Biologically based dose response model for hepatic toxicity: a mechanistically based replacement for traditional estimates of noncancer risk

Abstract

Uncertainty in risk assessment can be reduced by increasing the use of relevant data specific to the particular xenobiotic and exposed organism. We describe the development of a preliminary, mechanism-based exposure response model for chloroform hepatotoxicity consisting of toxicokinetic (TK) and toxicodynamic (TD) submodels. The TK submodel is based on an existing physiologically based toxicokinetic (PBTK) model for chloroform. The TD submodel consists of an empirical function linking tissue dose, defined by the PBTK model, with hepatocyte killing and subsequent regenerative cellular replication. Chloroform-induced cell killing was inferred quantitatively from dose-response hepatic labelling index studies conducted in female B6C3F1 mice and male F344 rats. The overall model was scaled to humans by conventional scaling of the TK submodel and by using the TD submodel as is i.e. as developed from the rodent data. The resulting human model was used to analyze a case of human poisoning which developed after repeated ingestion of large doses of cough syrup containing chloroform and alcohol. The model predicted the observed toxic response after the capacity for chloroform metabolism was increased by a factor of 3 from the value estimated using human liver microsomes. This is an acceptable adjustment of this parameter, given the uncertainty associated with the extrapolation from microsomes and the coexposure to alcohol. This preliminary result is encouraging, suggesting that the model, at its current stage of development, is able to approximate actual human risks of hepatotoxicity from chloroform exposure. The extensive use of data on chloroform TK and cytolethality-induced regenerative cellular replication for model development suggests that the model has reduced uncertainty relative to the current U.S. EPA oral reference dose (RfD) calculation for chloroform, which does not use any mechanistic or dose-response data.