Toxicity data, correlation with mechanistic

Toxicity data derived from single laboratories on selected animal subjects are known to correlate well with mechanistic descriptors such as logP. Examples are reported for barbiturates on rats, mice and rabbits (JJ, bicyclophosphates on mice (2, 2,6-dialkylanilines on rats (2.), and phenols on mice ( ). Optimal dependance on logP is frequently observed. 

Supporting data from studies in isolated rodent tissues and the standard geno-toxicity assays in Salmonella and other species can help inform the mechanistic interpretations. However, a risk assessor must evaluate these data with caution. Are the results from liver (where regenerative hyperplasia is most often observed) relevant to tumors that may be observed in other organs If a chemical caused tumors in liver, bladder, and lung, for example, but correlative cytotoxicity was only documented in liver, what should be concluded from the liver cytotoxicity data The data may seem solid for a cytotoxic threshold for liver tumors in rats, while kidney tumors are observed without notable renal cytotoxicity in mice, and a genotoxic metabolite is detected. In this case, a cautious risk assessor would not likely conclude that a nonlinear extrapolation is appropriate for determination of a regulatory standard, based on the mouse data. 

Toxic response of animals to important chemical classes can be analyzed in terms of mechanistic descriptors, even with data from many different laboratories. We estimate that 8-10 data points are needed to support each descriptor with inter-laboratory data, a doubling of the requirement for single labs. Given the large data sets available from toxicity compilations like Sax and RTECS, we were able to achieve many successful correlations. Differences among animals or between different routes of administration were easily perceived in mechanistic terms. Important conclusions can be drawn in a three-way comparison of chemical class, animals and routes by examining the form of correlation and the mean level of toxicity. 

---