1 This paper was presented at Mercury as a Global Pollutant: 5th International Conference, Rio de Janeiro, Brazil, May 23}28, 1999. Environmental Research Section A 84, 127 } 132 (2000) doi:10.1006/enrs.2000.4104, available online at http://www.idealibrary.com on Human Variability in Mercury Toxicokinetics and Steady State Biomarker Ratios 1 Scott M. Bartell, Rafael A. Ponce, Ravi N. Sanga, and Elaine M. Faustman Department of Environmental Health and Institute for Risk Analysis and Risk Communication, University of Washington, Seattle, Washington 98105-6099 Received July 30, 1999 Regulatory guidelines regarding methylmercury exposure depend on dose+response models relating observed mercury concentrations in maternal blood, cord blood, and maternal hair to develop- mental neurobehavioral endpoints. Generalized estimates of the maternal blood-to-hair, blood-to-in- take, or hair-to-intake ratios are necessary for link- ing exposure to biomarker-based dose+response models. Most assessments have used point esti- mates for these ratios; however, signiAcant inter- individual and interstudy variability has been re- ported. For example, a maternal ratio of 250 ppm in hair per mg/ L in blood is commonly used in models, but a 1990 WHO review reports mean ratios ranging from 140 to 370 ppm per mg/ L. To account for in- terindividual and interstudy variation in applying these ratios to risk and safety assessment, some researchers have proposed representing the ratios with probability distributions and conducting probabilistic assessments. Such assessments would allow regulators to consider the range and like- lihood of mercury exposures in a population, rather than limiting the evaluation to an estimate of the average exposure or a single conservative exposure estimate. However, no consensus exists on the most appropriate distributions for representing these parameters. We discuss published reviews of blood-to-hair and blood-to-intake steady state ra- tios for mercury and suggest statistical approaches for combining existing datasets to form generalized probability distributions for mercury distribution ratios. Although generalized distributions may not be applicable to all populations, they allow a more informative assessment than point estimates where individual biokinetic information is un- available. Whereas development and use of these dis- tributions will improve existing exposure and risk models, additional efforts in data generation and model development are required.  2000 Academic Press Key Words: biokinetic; methylmercury; exposure; risk; uncertainty. INTRODUCTION Most epidemiologic studies of methylmercury toxicity are based on biological markers of exposure, such as mercury in blood, hair, and urine (WHO, 1990). Fetal cord blood has also been used as a bio- logical exposure marker for gestational methylmer- cury exposure (Grandjean et al., 1997, 1998). These markers are presumed to re8ect relative methylmer- cury exposures in the individuals studied and have been used in categorical and continuous dose}re- sponse analyses (Clarkson et al., 1981; Cox et al., 1989; Crump et al., 1995). These analyses are then used as a basis for health risk assessment and regu- latory policy, which are commonly focused on envir- onmental and dietary sources of methylmercury exposure. When methylmercury is ingested, it rapidly binds to erythrocytes and is carried throughout the body. Some demethylation occurs, so that methylmercury and inorganic mercury are eventually excreted through the hair, urine, and feces (Smith et al., 1994; Gray, 1995). Although some recent models are more complex, mercury biokinetics are often described us- ing a single biological half-life of between 44 and 76 days (Lipfert, 1997). Because mercury is slowly ex- creted from the body, mercury biomarkers such as blood and urine are most strongly in8uenced by recent exposures. Nonetheless, mercury exposures are often assumed to have occurred with suf7cient 127 0013-9351/00 $35.00 Copyright  2000 by Academic Press All rights of reproduction in any form reserved.