Summary
This report describes a validation study of the dermal part of ECETOC TRA based on independent data. Literature studies and data from industry was gathered on exposure measurements for hands (and partly for underarms). The study included some types of situations that are not unconditionally inside the scope of the model, such as the use of pesticides and the use of powders in liquid products.
Data sets were derived from the sources and described as ‘exposure cases’, i.e. the data from one data source that describes measurements with the same input values for all relevant ECETOC TRA parameters. Contextual information was transformed to values for input parameters using expert elicitation. The resulting input values were used per exposure case to calculate the model estimate. The 75th percentile of the measured values for each exposure case was calculated from the geometric mean and geometric standard deviation and was compared with the model estimate.
Data sources with insufficient contextual information were used, as far as possible for indicative assessments. Also, comparisons between exposure cases for different, but similar activities, expressed as Process Categories (PROCs) and between different substance types were used as indicative information.
A total of 110 exposure cases from more than 30 data sources was used for direct comparison with model estimates, including at least 30 exposure cases each for liquids, solids and solids used in liquid products, while more than 60 studies were analysed for indicative information. Direct comparisons were possible for exposure situations in various industry sectors and for various product types. There was a relatively large number of exposure cases on product transfer, spray application and rolling and brushing.
The direct comparisons showed that in 80% of the cases the ECETOC TRA model estimate was higher than the corresponding 75th percentile of measured data. In more than half of the exposure cases, the model estimate was more than ten times the 75th percentile. There were also exposure cases where the model estimate was lower than the 75th percentile of measure data. These cases specifically occurred when the exposure level is high. In general it was seen that the model, compared to the 75th percentile of measured data, overestimates exposures substantially for very low exposure levels, while there is an increasing occurrence of underestimation of the model at higher measured exposure levels.
The very high overestimations at low exposure levels particularly occur when the concentration of substance in the product used is very low. This related to the conservative, e.g. less than linear, effect that concentration in the product has in the model. The underestimations occur to a large extent in situations with very high skin contamination due to e.g. dipping (part of) hands in a liquid, working in a cloud of dust or having contact with heavily contaminated surfaces.
By means of regression analyses, it was found that the ECETOC TRA model explains 37% of variance in the total dataset (between 28% and 46% for different substance types). Multivariate analyses of separate parameters indicated that the PROC, concentration of substance in product, use of gloves and the sampling method were significant determinants of the variance in exposure levels. Together they explained 62% of the total variance. The average protective effect of gloves was found to be a factor 34, while in our model estimates, we used factors of 5 and 10. Exposure cases sampled via an interception method (e.g. gloves) on average had a factor of 6 higher exposure levels than exposure cases sampled via a removal method (e.g. hand washing). Removal methods were used more often when sampling beneath protective gloves and in general, when sampling in cases with a low estimated exposure level.
Multivariate analyses were also used to study the factors that influence the ratio of model estimate/75th percentile of measured data. PROC, percentage of substance in the product, LEV use, glove use and sampling method appeared to be significant determinants of this ratio, together explaining 38% of the variance in the ratio. This analysis suggests that the effect of LEV on dermal exposure may, on average, be overestimated in the model, while the effect of glove use and concentration of the substance in the product may be underestimated.
It was concluded that the ECETOC TRA estimates correlate with the measured exposure levels and that they are in most cases conservative. However, in case of high measured exposure levels, the model estimates may not always be sufficiently conservative, specifically in case of relatively extensive contact with product.
Because the model is intended to provide conservative estimates, the assumed exposure reducing effect of gloves, though lower than found on average in practice appears to be quite reasonable, because it takes account of reasonable worst case situations.
This study also indicates that the ECETOC TRA model can be used to make estimates of dermal exposure to solids used in liquid products by considering them to be liquids with negligible volatility. These estimates are of similar quality as those for liquids.