Improved in silico prediction of carcinogenic potency (TD50) and the risk specific dose (RSD) adjusted Threshold of Toxicological Concern (TTC) for genotoxic chemicals and pharmaceutical impurities

The Threshold of Toxicological Concern (TTC) is a level of exposure to a genotoxic impurity that is considered to represent a negligible risk to humans. The TTC was derived from the results of rodent carcinogenicity TD50 values that are a measure of carcinogenic potency. The TTC currently sets a default limit of 1.5 μg/day in food contact substances and pharmaceuticals for all genotoxic impurities without carcinogenicity data. Bercu et al. (2010) used the QSAR predicted TD50 to calculate a risk specific dose (RSD) which is a carcinogenic potency adjusted TTC for genotoxic impurities. This promising approach is currently limited by the software used, a combination of MC4PC (www.multicase.com) and a Lilly Inc. in-house software (VISDOM) that is not available to the public. In this report the TD50 and RSD were predicted using a commercially available software, SciQSAR (formally MDL-QSAR, www.scimatics.com) employing the same TD50 training data set and external validation test set that was used by Bercu et al. (2010). The results demonstrate the general applicability of QSAR predicted TD50 values to determine the RSDs for genotoxic impurities and the improved performance of SciQSAR for predicting TD50 values.     

Mutagenic impurities in pharmaceuticals: A critique of the derivation of the cancer TTC (Threshold of Toxicological Concern) and recommendations for structural-class-based limits

The cancer TTC (Threshold of Toxicological Concern) concept is currently employed as an aid to risk assessment of potentially mutagenic impurities (PMIs) in food, cosmetics and other sectors. Within the pharmaceutical industry the use of one default cancer TTC limit of 1.5 μg/day for PMIs is being increasingly questioned. Its derivation, originally in the context of foodstuffs, can be broken down into five key elements: dataset composition; determination of carcinogenicity/mutagenicity status and carcinogenic potency (based on TD₅₀s) of compounds in the dataset; linear extrapolation of carcinogenic potencies; evaluation of the more potent compounds in each structural category, and presence of representative structural alerts amongst the more potent compounds. A detailed evaluation reveals that the derivation process is distorted by the use of the lowest statistically significant TD₅₀s (which can produce a false-carcinogen phenomenon) and by employing linear extrapolation for non-mutagenic carcinogens. By correcting for these two factors, it is concluded that only around 50% of conventional structural-alert categories were adequately addressed and that limits higher than the default value appear to be justified in many cases. Using similar criteria for PMIs in pharmaceuticals, four distinct potency categories of conventional structural alerts can be distinguished, ranging from alerts with questionable validity to those with high potency, which are considered to provide a range of flexible and pragmatic limits for such impurities.