Background: Case-control studies rely on questionnaire data to assess participant exposures. Exposure information collected by job for specific agents can be used in conjunction with a database of measurements and associated exposure determinants to develop quantitative individual exposure estimates. Our objective was to develop an exposure determinants database and explore issues related its use for individual exposure estimation. Methods: The National Institute for Occupational Safety and Health and the National Cancer Institute have each conducted brain cancer studies and jointly developed methods for estimating individual exposures to six chlorinated solvents (perchloroethylene, chloroform, methylene chloride, tetrachloroethane, 1,1,1-trichloroethane, and carbon tetrachloride) employing a database of 5100 measurements and ten associated exposure determinants from 350 references. The determinants (number of levels assigned) include: closed/open process (3), temperature (3), release mechanism (6), quantity (3), ventilation type (6), indoor/outdoor location (3), confined space (3), source proximity (3), exposure route (3), and direct/indirect contact (3). For example, release mechanism has six levels: evaporation, spreading, manual agitation, rolling, mechanical agitation, and aerosolization. From the determinants and measurement data statistical models will be developed to predict exposure intensity. Results: Decision rules for defining determinants for specific operations, e.g. degreasing, foam manufacturing, and printing operations, are discussed. The distributions of determinants represented in the measurement database, are given by solvent. For example, for TCE, four of six possible ventilation determinant levels appear in the determinants database: 14% LEV and mechanical dilution, 55% LEV only, 9% LEV and no LEV combined, 12% dilution ventilation and no ventilation combined, and 10% unknown ventilation type. Since statistical models will link determinants to measurement data, determinants with little variation will have limited value in model development. For various exposure determinant combinations in the database, intensity data are ample (e.g. perchloroethylene exposures of dry cleaners) or sparse (e.g. outdoor exposures, or lower solvent quantities combined with local exhaust ventilation). Some combinations of determinants with abundant exposure measurements in the database comprise a small fraction of the 17,000 jobs (3726 unique job-industry combinations) in the case-control studies. The determinant distribution in the epidemiological studies will be compared to those in the measurement database. Conclusions: The strength of any associations found in the case-control studies depends on the validity and reliability of the exposure estimation process. We describe issues related to a systematic method for assessing such exposures. Use of these methods in our respective brain cancer studies will permit future combined epidemiological analyses.
Tijdschrift voor Toegepaste Arbowetenschap. Exposure Assessment in a Changing Environment, June 16-18, Utrecht, The Netherlands