Assessment of Variability in Biomonitoring Data Using a Large Database of Biological Measures of Exposure

Abstract

Although intra- and interindividual sources of variation in airborne exposures have been extensively studied, similar investigations examining variability in biological measures of exposure have been limited. Following a review of the world's published literature, biological monitoring data were abstracted from 53 studies that examined workers' exposures to metals, solvents, polycyclic aromatic hydrocarbons, and pesticides. Approximately 40% of the studies also reported personal sampling results, which were compiled as well. In this study, the authors evaluated the intra- and interindividual sources of variation in biological measures of exposure collected on workers employed at the same plant. In 60% of the data sets, there was more variation among workers than variation from day to day. Approximately one-fourth of the data were homogeneous with small differences among workers' mean exposure levels. However, an almost equal number of data sets exhibited moderate to extreme levels of heterogeneity in exposures among workers at the same facility. In addition, the relative magnitude of the intra- to interindividual source of variation was larger for biomarkers with short compared to long half-lives, which suggests that biomarkers with half-lives of 7 days or longer exhibit physiologic dampening of fluctuations in external levels of the workplace contaminant and thereby may offer advantages when compared to short-lived biomarkers or exposures assessed by air monitoring. The use of biological indices of exposure, however, places an additional burden on the strategy used to evaluate exposures, because data may be serially correlated as evidenced in this study, which could result in biased estimates of the variance components if autocorrelation is undetected or ignored in the statistical analyses.

Variation of metabolites in normal human urine

“…Thus, differences in metabolite excretion should be noted during metabolic studies using urine as the biofluid. Urine metabolite variability has been seen in earlier metabolomic studies; such as public health and exposure limits to certain agents (Symanski and Greeson 2002). The metabolite variability observed in our human study is similar to that found in studies by Tate (Tate et al 2001) and Holmes (Holmes et al 2000).…”

Evaluating measurement error in estimates of worker exposure assessed in parallel by personal and biological monitoring

“…This becomes critical when evaluating the magnitude of exposure measurement error since improperly specifying the mean structure can result in biased estimates of the variance components [Symanski et al, 1996]. In addition, if measurements closer together in time are more highly correlated than those spaced farther apart, then variance components and functions thereof estimated from models that ignore serial correlation may be biased as shown previously [Symanski and Greeson, 2002] and in this investigation. While we found little evidence of serially dependent measurements (<15% of the data), the residence time of a biomarker should be factored into decisions regarding the longitudinal design of a study and the analytic approaches used for modeling the correlation structure among multiple measurements collected on the same individual.…”

“…The database contributing measurements analyzed in this study has been described elsewhere [Symanski and Greeson, 2002]. Briefly, biological monitoring data were abstracted from 53 studies reported in the peer-reviewed literature that examined workers' exposures in myriad industries worldwide.…”

Urine stability for metabolomic studies: effects of preparation and storage

“…Many studies have found correlations of key metabolites with particular diseases, metabolites that may be used for diagnosis, or changes that may arise due to laboratory conditions of the study (Brindle et al, 2002;Coen et al, 2005;Sabatine et al, 2005). Human urine studies have found changes in metabolite concentrations as a result of age and diet, variation in urine metabolite concentrations within large populations, and variation of particular urine metabolites in individuals over time (Guneral and Bachmann, 1994;Dyer et al, 1997;Zuppi et al, 1997;Symanski and Greeson, 2002;Lenz et al, 2003Lenz et al, , 2004Bollard et al, 2005). Studies of animal urine have found metabolic differences within genetically similar animal models (Holmes et al, 2000;Tate et al, 2001).…”