[Retrospective estimate of occupations exposure to solvents. Methodology used in the Epilymph study].

To assess occupational exposure to solvents in the EPILYMPH multicentre case-control study on lymphoma, we combined experts assessment, based on detailed description of work tasks from questionnaires, with the use of job-exposure matrices (jem). Whenever occupations and/or industries conveying exposure to occupational risk factors for lymphoma occur in their work history, the interviewer selects the respective questionnaire in a set of 15 special questionnaires and applies it to the study subject. All jobs lasting one year or more in the work histories are coded using the 1968 ILO ISCO, and the 1996 EEC Classification of Economic activities (NACE). A priori jems for 3 categories of solvents (solvents in general, chlorinated aliphatic hydrocarbons and benzene-toluene-xylene combination) and 10 individual agents have been constructed based on the same coding systems, by which each occupationa/industry combination receives a unique score for confidence and intensity of exposure to each agent and group of solvents. Trained industrial hygienists (IH's) in each participating country examine the general and special questionnaires and classify three indices of exposure for each agent under scrutiny: confidence, which represents the IH degree of certainty that the worker has been exposed to the agent; frequency, which is expressed in a 4-step scale according to the proportion of working time involving contact with the agent; and intensity, which is also expressed in a scale, defined by quantitative cut-points, when available, or benchmark occupations, when no quantitative information is available from the questionnaires. The IH finally notes his degree of reliance on his own exposure estimate. The IH supports his exposure assessment with confidence and intensity scores provided by the jem when questionnaire information is so poor that it does not provide any clue about exposure, or he relies poorly on his own exposure estimate. Such a procedure is conceptually alike to statistical procedures to fill missing data in a given series.