The challenge of testing chemicals for potential carcinogenicity using multiple short-term assays: An analysis of a proposed test battery for hair dyes

Recent reports of the association of hair dyes usage with increased bladder cancer risk in women with the slow NAT-2 acetylator phenotype have resulted both in attempts to identify the putative carcinogen as well as in devising batteries of tests that could be used to screen for such putative carcinogens in hair dye formulations, their intermediates and final products. Analytical studies have reported the presence of traces (∼0.5 ppm) of the carcinogen 4-aminobiphenyl in some hair dye preparations. In parallel, SCCNFP (Scientific Committee on Cosmetic and Non-Food Products Intended for Consumers) has suggested the deployment of a battery of six in vitro assays followed by an in vivo assay. The practicality of deploying and interpreting such a battery is analyzed herein as it is expected to result in 64 and 128 possible test results and SCCNFP does not provide detailed guidance of how the test results are to be interpreted. In this study we have applied a previously described Bayesian approach which takes advantage of the known predictive performances of individual assays, to analyze the possible outcomes of the 6–7 test batteries. While the SCCNFP battery is clearly risk-averse, it is shown that performing all of the assays is not always necessary and moreover it does not necessarily improve predictive performance. Finally, based upon the reported mutagenicity of 4-aminobiphenyl, it is doubtful that this “impurity” would be detected by the test battery.