1131 Combining lead exposure measurements and experts’ opinion through a bayesian framework

Abstract

Objectives CAREX (CARcinogen EXposure) is a carcinogen surveillance system employed in many countries. To initiate Korean CAREX, we focused on estimating the exposure intensity of lead across industries, which is a suspected carcinogen. Methods We extracted airborne lead measurements from the work environment measurement database (WEMD) which is the Korean nationwide measurement database. In addition, we elicited the experts’ opinion about lead exposure intensity across industries by conducting a questionnaire. Experts provided estimates of lead exposure levels as the boundary of the 5th and 95th percentiles. We assumed that experts provided their estimates based on the assumption of log-normal distributions of exposure. First, for each industry, estimates of log-transformed geometric means (logGM) and log-transformed geometric standard deviations (logGSD) were extracted from the experts’ responses, followed by combining them to quantify the experts’ prior Normal-Inverse-Gamma prior distribution. Then, the corresponding logGM and logGSD from lead measurement data for each industry were updated with the experts’ prior distribution through a Bayesian framework, yielding posterior distributions of logGM and logGSD. Results WEMD contains 83 035 airborne lead measurements collected between 2002–2007. Total 17 occupational hygiene professionals with more than 20 year experience provided lead exposure estimates. In industries where measurement data is abundant, the measurement data dominate the posterior exposure estimates, while in industries with a limited number of measurements, experts’ opinion played an important role in determining posterior exposure estimates. For example, rubber manufacturing industry with 246 measurements (GM 1.72; GSD 1.94) and 6 experts’ responses (GM 0.79; GSD 6.73) showed posterior exposure estimates of GM 1.60 and GSD 2.34. Conclusions Our method of estimating the exposure intensity of CAREX may introduce an unbiased approach to the development process by utilising both prior knowledge of experts and measurement data simultaneously. In addition, it will supply a framework for future updates.