[Job exposure matrices].

During the last years, the use of Job Exposure Matrices (JEMs) for the evaluation of past occupational exposures has grown. This approach is described in this paper. The definition of a JEM is provided. The development process of a JEM is discussed. It includes the choice of the job classification system which influences the degree to which the matrix can be used in other studies, the precise definition of exposures and the choice of entries which should provide more information than the oversimplified exposed/non exposed dichotomy. When comparing JEMs and experts' evaluation of exposure, all aspects of both methods have to be looked at and a balance struck between the advantages and shortcomings of each of them. The problem should not be reduced to a single choice between the precision provided by experts, and the cost savings and standardization offered by JEMs. Statistical analysis with a JEM must be adapted to the job classified information it provided. When an appropriate method is used, the loss of power with a JEM compared to experts' evaluation is rather small or even null. Finally, it is important to avoid a systematic opposition between JEMs and expert assessments, as one positive result of their comparison may be the improvement of both methods of exposure evaluation.