A number of low molecular weight (LMW) organic chemicals are known
to cause occupational respiratory or skin sensitisation. A set of 200
LMW organic skin and approximately 75 respiratory sensitisers were
identified by critical appraisal of published case literature. The respiratory sensitisers (asthmagens) were systematically compared in turn
with suitable control chemicals and the skin sensitisers using a case-control methodology. The control chemicals were selected from known
hazardous LMW organic chemicals for which no reports of respiratory
sensitisation could be found.
Several potential methods of differentiating between asthmagens and
controls or asthmagens and skin sensitisers (by chemical structure
alone) were investigated. These comprised hazardous fragment identification by calculating odds ratios for hazard (HOR's), cluster analysis and the logistic regression analysis. Of these methods the most
effective approach was the logistic regression analysis. Using these
methods several known or suspected hazardous substructures were
confirmed to present statistically significant occupational asthma (OA)
hazard. These included isocyanates, acid anhydrides, acrylates and
(oligo)-amines. Furthermore, certain sub-structural fragments such as
chlorine atoms appeared to provide a protective effect from OA hazard. For differentiating between skin sensitisers and asthmagens it
was noted that fragments with carbon double bonded to nitrogen or
oxygen atoms were significantly more prevalent in the respiratory sensitisers set.
A predictive model of chemical asthma hazard was created using logistic regression and the model tested on a validation set of chemicals yielded a predictive kappa value greater than 0.7. This model is
available for predictive testing of compounds for asthma hazard via the
World Wide Web.
This work demonstrates that simple structural information may, in
conjunction with a well designed methodology, be used to identify occupational sensitisers with reasonable reliability.