Under current European and national legislation in Europe, biocidal products must be registered based on a risk assessment for humans, animals and the environment. Especially human exposure assessment shows many knowledge gaps, which make it hard to properly assess health risks for many biocidal application scenarios. An EU-funded project is currently tackling this issue and guidance is developed to deal with the problem in an appropriate fashion.

Introduction
The Biocidal Product Directive 98/8/EC requires registration of biocidal products on the basis of a risk assessment for their uses. The Directive is currently being implemented in member state law, and initial studies have been carried out to investigate the approaches as laid down in the various regulations. There are 23 different biocidal product types, in four major groups:
- disinfectants and general biocidal products;
- preservatives;
- pest control products;
- other biocidal products.

A major shortcoming in our present knowledge is the assessment of human exposure for biocides. This was also shown in the pilot programme, indicated above. To fill this knowledge gap, a project proposal by Institutes/organisations from 6 European member states, together with representatives from CEFIC, the producing and formulating industry in Europe, was accepted by DG Environment (B4-3040/2000/291079/MAR/E2). The project is to deliver its final report mid 2002.

Results
The project focuses on exposures to workers and consumers. In this presentation only results with respect to worker exposure will be discussed. To assess exposure to biocidal products for risk assessment purposes, one needs to know the use pattern of the product:
- the product (physical state, package form, etc.) and its purpose;
- where, how and by whom it is used;
- expected control measures;
- tasks, frequencies and durations
- for mixing/loading
- for application
- for post-application activities (secondary exposure);
- who else may be exposed (bystanders).

Occupational exposure data have been requested from industry (sectors), governmental agencies, and academia from North America and Europe. All these publications have been quality-assessed based on pre-developed criteria, regarding aims, documentation and analytical-chemical quality aspects. From the study reports that were considered adequate, data were extracted and compiled in a series of databases.

For use of the databases two different approaches are taken. These are both based on the assumption that for most biocidal uses (applications), the exposure is task-based and not dependent on specific chemical properties of the biocide. This means that extrapolation from one study to another is possible when the exposure data have a suitable format and the exposure scenario (series of tasks, preferably one task) is similar.

All available study results were entered in the matrix at the right cell(s). Currently these cells are being filled with additional (new) exposure data.

The matrix has four typical central tendency values (GM), varying from 4 mg/min (low), 20 mg/min (medium), to 100 mg/min (high) and 500 mg/min (top). The other axis involves the width of the distribution and its GSD varies from 2.45 (narrow), via 3.36 (intermediate) to 6.04 (wide), as shown in the table. From this description it is clear that log-normal distributions have been assumed for these exposures, and the typical cell values have been chosen to accommodate the current experimental study results.

The databases are described with respect to the involved tasks and have been evaluated for relevant levels of exposure using 75-percentile values for typical exposure levels and 95-percentiles for ‘reasonable worst case exposures’. These values are used for chronic health effects, whereas the 95-percentiles are proposed for use in case of acute health effects. 95-Percentiles may be used to ascertain possible levels of exposure in case of foreseeable misuse.

If the task under consideration is not available in the matrix then the exposure assessor should present arguments for specific choices for an exposure cell in the matrix. If there is some doubt, a higher width and higher typical value should be chosen. If no arguments at all can be presented for specific choices, the assessor is to choose for the cell with the largest width and highest typical value. This forces industry (registrants) to produce data for that specific task or set of tasks (use scenario).

Second approach
Bayesian statistics are used to develop an exposure assessment for tasks that have no specific exposure model, but do have assessable (dis)similarities with all the other sets of data in the matrix. In this approach, called BEAT (Bayesian Exposure Assessment Toolkit), all databases (for the time being only for body exposure) have been computerised and are transformed into distributions with discrete GMs and GSDs.

For a new task (or scenario) that can be described in terms of similarity or dissimilarity a set of questions must be answered and entered into the model. Using Baye’s theorem, the (dis)similarities are calculated for all the databases in the model, leading to a new distribution which is then fitted to the cells of the above-mentioned matrix, since this is used to model the output in terms of maximum likelihood that the distribution fits into the various cells.

The so-called rule base for assessing the degree of similarity between tasks (scenarios) determines of course the output from the model. The rule base is determined on the basis of expert knowledge from field experience. Its validity is checked internally, by taking a study out and entering it again using the rule base. External validity can be determined using a new task (scenario) and assessing it on the basis of the rule base and concomitant comparison with actual field data to be collected.

In principle there is no problem to add a rule base for inhalation exposure data and hand exposure data. The databases for these exposures are already in the computerised approach.

There are currently hardly any data for post-application exposure. Some ‘reference scenarios’ have been developed that might help in estimating the most relevant exposures. These scenarios can only be handled for the exposure viewpoint with conservative assumptions and will thus require further exposure studies which will then help to obtain more realistic assumptions and possibly suitable databases.

Discussion and conclusion
The Bayesian approach for modelling, which combines objective and subjective data, is being considered and will be subject of validation exercises.

Exposure scenarios (“patterns of use”) are described for all 23 biocidal product types.

Guidance is written -with many worked out examples- to enable authorities and registrants to do exposure assessments in a harmonised way, throughout European member states, with quite a bit of specific expertise still required for use.

A draft report is prepared and discussed at EU level (Biocide Technical Meeting in May 2002). Based on these discussions, the document will be finished mid 2002. Development of the BEAT model will continue for quite some time afterwards. The current idea is to also include the exposure data from other European exposure databases under development, such as RISKOFDERM and EUROPOEM.

The project report is considered a ‘living document’ by the project team, underlining the need to update it and adapt it in view of further developments in experimental studies and scientific approaches for exposure modelling.

Partners in the project
1. TNO Chemistry, Zeist, the Netherlands (project co-ordination)
2. Finnish Institute of Occupational Health, KRIOH, Kuopio, Finland
3. Health and Safety Executive / Laboratories, Bootle/Sheffield, UK
4. Institute of Occupational Medicine, Edinburgh, UK
5. Bundesanstalt für Arbeitsschutz und Arbeitsmedizin, Dortmund, Germany
6. Rijksinstituut voor Volksgezondheid en Milieu, Bilthoven, the Netherlands
7. CEFIC representatives from BAYER, Germany and Rohm & Haas, France