Assessing variability and comparing short-term biomarkers of styrene exposure using a repeated measurements approach

The aim of this work is to compare several short-term biomarkers of styrene exposure, namely urinary styrene (StyU), mercapturic acids (M1+M2), mandelic acid (MA), phenylglyoxylic acid (PGA), phenylglycine (PHG), and 4-vinylphenol conjugates (VP), for use as biomarkers of exposure in epidemiologic studies. A repeated measurements protocol (typically 4 measurements per worker over 6 weeks) was applied to measure airborne styrene (StyA) and urinary biomarkers in 10 varnish and 8 fiberglass reinforced plastic workers. Estimated geometric mean personal exposures to StyA were 2.96mg/m(3) in varnish workers and 15.7mg/m(3) in plastic workers. The corresponding levels of StyU, M1+M2, MA, PGA, MA+PGA, PHG and VP were 5.13microg/L, 0.111, 38.2, 22.7, 62.6, 0.978, and 3.97mg/g creatinine in varnish workers and 8.38microg/L, 0.303, 146, 83.4, 232, 2.85 and 3.97mg/g creatinine in plastic workers. Within-worker (sigma(wY)(2)) and between-worker (sigma(bY)(2)) variance components were estimated from the log-transformed data as were the corresponding fold ranges containing 95% of the respective lognormal distributions of daily levels ((w)R(0.95)) and subject-specific mean levels ((b)R(0.95)). Estimates of (w)R(0.95) (range: 4-26) were generally smaller than those of (b)R(0.95) (range: 5-790) for both environmental and biological markers; this indicates that exposures varied much more between workers than within workers in these groups. Since attenuation bias in an estimated exposure-response relationship increases with the variance ratio lambda=sigma(wY)(2)/sigma(bY)(2), we estimated values of lambda for all exposure measures in our study. Values of lambda were typically much less than one (median=0.220) and ranged from 0.089 for M1+M2 in plastic workers to 1.38 for PHG in varnish workers. Since values of lambda were 0.147 and 0.271 for StyA in varnish workers and plastic workers, respectively, compared to 0.178 and 0.210 for MA in the same groups, our results suggest that either air measurements or conventional biomarker measurements (urinary MA) would be comparable surrogates for individual exposures in epidemiologic studies.