A quantitative structure-activity relationship (QSAR) investigation of a Draize eye irritation database

National and international regulations require that the ocular safety of new chemicals is established and demand that, unless the substance can be predicted to be harmful to the eye, this will be established by means of the Draize rabbit eye test. To date, no alternative in vitro methods are accepted by the regulatory authorities. Investigation of correlations between physicochemical properties of a chemical and its eye irritation potential has received scant attention. A 'gold standard' dataset of eye irritation tests collated by the European Centre for Ecotoxicology and Toxicology of Chemicals has been supplemented by other published data to form a consistent set of biological activities for QSAR modelling. Using this dataset, the applicability of QSAR methods for modelling the eye irritation responses has been assessed. The possibilities, but also clear limitations, of a QSAR approach to the evaluation of eye irritation potential have been demonstrated. The dataset is the best available, but nevertheless presents several problems: the number of irritants is small compared with the number of non-irritants and the whole dataset is chemically heterogeneous, with the implication that multiple mechanisms of eye irritation may be involved; the data come from several laboratories; the numerical Draize score is a pseudo-quantitative parameter based on a summation of manipulated subjected scores. A molar adjusted eye irritation score proved to be at least partially useful in demonstrating and understanding the structure-activity relationships involved with such an endpoint. A more promising approach proved to be pattern recognition techniques using principal component analysis and relying on the overall classification of eye irritation potential. This has led to a qualitative and, importantly, testable, relationship between eye irritation potential and physicochemical properties, particularly those describing hydrophobicity. The availability of quality eye irritation data on more chemicals would allow refinement and further testing of this hypothesis.