Health effects after long-term exposure to subway particulate matter (PM) remain unknown due to the lack of individual PM exposure data. This study aimed to apply the job exposure matrix (JEM) approach to retrospectively assess occupational exposure to PM in the Parisian subway. Job, the line and sector of the transport network, as well as calendar period were four JEM dimensions. For each combination of these dimensions, we generated statistical models to estimate the annual average PM10 concentration using data from an exhaustive inventory of the PM measurement campaigns conducted between 2004 and 2020 in the Parisian subway and historical data from the Parisian air pollution monitoring network. The resulting JEM and its exposure estimates were critically examined by experts using the uncertainty analysis framework. The resulting JEM allows for the assignment of the estimated annual PM10 concentration to three types of professionals working in the subway: locomotive operators, station agents, and security guards. The estimates' precision and validity depend on the amount and quality of PM10 measurement data used in the job-, line-, and sector-specific models. Models using large amounts of personal exposure measurement data produced rather robust exposure estimates compared to models with lacunary data (i.e., in security guards). The analysis of uncertainty around the exposure estimates allows for the identification of the sources of uncertainty and parameters to be addressed in the future in order to refine and/or improve the JEM. The JEM approach seems relevant for the retrospective exposure assessment of subway workers. When applied to available data on PM10, it allows for the estimation of this exposure in locomotive operators and station agents with an acceptable validity. Conversely, for security guards, the current estimates have insufficient validity to recommend their use in an epidemiological study. Therefore, the current JEM should be considered as a valid prototype, which shall be further improved using more robust measurements for some jobs. This JEM can also be further refined by considering additional exposure determinants.
Read full abstract