Retrospective benzene exposure assessment for a multi-center case-cohort study of benzene-exposed workers in China

Abstract

Quality of exposure assessment has been shown to be related to the ability to detect risk of lymphohematopoietic disorders in epidemiological investigations of benzene, especially at low levels of exposure. We set out to build a statistical model for reconstructing exposure levels for 2898 subjects from 501 factories that were part of a nested case-cohort study within the NCI-CAPM cohort of more than 110,000 workers. We used a hierarchical model to allow for clustering of measurements by factory, workshop, job, and date. To calibrate the model we used historical routine monitoring data. Measurements below the limit of detection were accommodated by constructing a censored data likelihood. Potential non-linear and industry-specific time-trends and predictor effects were incorporated using regression splines and random effects. A partial validation of predicted exposures in 2004/2005 was performed through comparison with full-shift measurements from an exposure survey in facilities that were still open. Median cumulative exposure to benzene at age 50 for subjects that ever held an exposed job (n=1175) was 509 mg/m(3) years. Direct comparison of model estimates with measured full-shift personal exposure in the 2004/2005 survey showed moderate correlation and a potential downward bias at low (<1 mg/m(3)) exposure estimates. The modeling framework enabled us to deal with the data complexities generally found in studies using historical exposure data in a comprehensive way and we therefore expect to be able to investigate effects at relatively low exposure levels.

Benzene Exposure Response and Risk of Myeloid Neoplasms in Chinese Workers: A Multicenter Case–Cohort Study

“…Factory records were the primary source for historical benzene, toluene, and xylene air measurements and associated data, production processes, and complete job histories. See details described elsewhere (15) and in Supplementary Methods (available online). Questionnaires administered to subjects or next of kin were used to identify jobs held outside the cohort factories.…”

“…To estimate individual monthly benzene exposures, a Bayesian hierarchical model was built from the historic monitoring data. This model allowed for clustering of measurements by factory, workshop, job, and date (15). The exposure assessors were blinded to exposure status.…”

Retrospective Exposure Assessment Methods Used in Occupational Human Health Risk Assessment: A Systematic Review

“…The use of statistical models to assign estimated exposure intensities is a reliable practice that allows the development of quantitative and more reproducible methods for REA [ 34 ]. For example, hierarchical models can be used in order to obtain clusters of data concerning a specific factory, task or job [ 43 , 44 , 45 ]. Finally, questionnaires are a useful tool to gather a-posteriori information about a certain exposure scenario [ 13 ], especially when these can provide information about every task held by the subject in his working schedule.…”

“…The amount and quality of data available for the development of JEM is usually a crucial element in REA studies: data may be too few for a categorization of workers by exposure levels [ 33 ], the time period over which the data were collected may be unsuitable for the purpose of the study [ 45 ] and the data may not be enough for developing a quantitative cumulative exposure matrix or not detailed enough for a good characterization of a specific scenario [ 28 ]. The REA conducted using data from a specific occupation or job title is a source of well-established and detailed information of an exposure scenario.…”

Prostate cancer characteristics in the World Trade Center cohort, 2002–2013

“…Personal protective equipment was not worn for the majority of workers and individuals in the area 19 and our study has found significant association of higher clinical stage with the highest exposure level that includes exposure to the 9/11 dust cloud. The dust cloud that resulted after the collapse of the buildings, exposed respondents to soot, 20, 21 benzene, 22 WTC dust and smoke, which contained asbestos, 23 silica, 24 cement dust, 25 glass fibers, 26 heavy metals, 27 polycyclic aromatic hydrocarbons (PAHs), 20, 28 phthalates, 29, 30 polychlorinated biphenyls, 31 and polychlorinated dibenzofurans, 32, 33 and dioxins 34 from the burning and collapse of the planes and the towers. 35 Other suspected carcinogens, such as fine particulate matter (PM2.5) 36 were elevated above that of normal levels of 2 to 6 times higher in some areas.…”