THESE LIBRARY Michigan State University 100.5 This is to certify that the thesis entitled A LONGITUDINAL STUDY OF THE VARIABILITY AND PREDICTORS OF URINARY MERCURY LEVELS IN 6 - 11 YEARS OLD CHILDREN presented by XIAOBEI ZHU has been accepted towards fulfillment of the requirements for the MS. degree in Department of Epidemiology (2 (/1 (/j Major Professor’s Signature 5/9/2Wé Date MSU is an Affirmative Action/Equal Opportunity Institution -A _.---.-.-.-.-v PLACE IN RETURN BOX to remove this checkout from your record. To AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE JUN 20 2007 9007' 0': 2/05 p:/ClRC/DateDue.indd-p.1 A LONGITUDINAL STUDY OF THE VARIABILITY AND PREDICTORS OF URINARY MERCURY LEVELS IN 6 — 11 YEARS OLD CHILDREN By Xiaobei Zhu A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Epidemiology 2006 ABSTRACT A LONGITUDINAL STUDY OF THE VARIABILITY AND PREDICTORS OF URINARY MERCURY LEVELS IN 6 - 11 YEARS OLD CHILDREN By Xiaobei Zhu To exam the variability of urine mercury levels among children in three repeated measurements collected from three different areas in Germany during 1994-1997. Factors that affect urinary mercury levels among those children were also identified. Linear mixed mode] and Structure equation models were used to analyze the data. The Spearman correlation coefficients of the urinary mercury levels for the three mercury measurements were low, ranging from 0.212 to 0.307. In the present study, the intra-individual variation attributed more (approximately 72% - 74%) to the total variation compared with the inter-individual variation (approximately 26% - 28%) (ICC ranged from 0.26 to 0.28). The number of amalgam fillings had a significant effect on urinary mercury levels (p<0.001) controlling for other covariates. Residing in the toxic waste incinerator area posed a risk for increasing the urinary mercury level only in 1996, but seems to have a diminishing effect in 1997 from the results of both linear mixed mode] and SEMs. Age, the 24-hour total amount of urine, and urinary creatinine levels also have significant effect on urinary mercury levels. DEDICATION This work is dedicated to my beloved parents Shuyu Shi and Baofeng Zhu, my husband Yang Liang and my brother Yingnan Zhu. iii ACKNOWLEDGEMENTS I express appreciation towards my advisor, Professor Karmaus and my committee members, Dr. Todem and Dr. Kaneene, whose excellent guidance and instructions have made this work possible. I would also like to express my sincere gratitude to Dr. Velie, Dr. Gardiner, Dr. Luo and Dr. Liu, Dr. Kerver and Susan Davis for sharing their knowledge and helping me improve the manuscript. I thank my husband Yang Liang and my parents Mr. Zhu and Mrs. Shi and my brother Yingnan for their constant encouragement, motivation and love. iv TABLE OF CONTENTS LIST OF TABLES ............................................................................. vi LIST OF FIGURES ............................................................................... vii LIST OF ABBREVIATIONS ................................................................ viii CHAPTER 1 Introduction ...................................................................................... 1 CHAPTER 2 Population and methods .......................................................................... 7 2.1 Study population .................................................................... 7 2.2 Questionnaires ....................................................................... 7 2.3 Collection of urine samples and the determination of urinary mercury and creatinine .............................................................................. 8 2.4 Statistical methods .................................................................. 8 Data analysis ...................................................................... 8 Linear mixed model .............................................................. 9 Structural Equation Models (SEMs) to detect the stability of urinary mercury levels .................................................................... 11 CHAPTER 3 Results .............................................................................................. 12 3.1 Study population ................................................................... 12 3.2 Unadjusted Urinary Mercury (Hg) levels ........................................ 13 3.3 Urinary creatinine and 24 hour total urine in three different years ......... 13 3.4 The stability of the urinary mercury levels ..................................... 14 3.5 Association between adjusted urinary mercury levels and other factors 15 CHAPTER 4 Discussion ......................................................................................... 18 4.1 Study population .................................................................... 18 4.2 The variability of mercury levels in urine ........................................ 18 4.3 Factors that affect urinary mercury levels ........................................ 21 4.4 Structural Equation Models (SEMs) .............................................. 23 4.5 Summary ............................................................................. 24 REFERENCES ................................................................................... 49 LIST OF TABLES Table 1. Descriptive characteristics of the study population (n=663) .................... 27 Table 2. Crude urinary mercury level .......................................................... 28 Table 3. Crude urinary creatinine levels in three years ....................................... 31 Table 4. Crude 24 — hour total amount of urine samples in three consecutive years... . .32 Table 5. Spearman Correlation Coefficients of urinary mercury levels .................... 33 Table 6. Adjusted inter- and intra-individual variations (6' 2b and 6' 28) for log- transformed urinary mercury. ............................................................ ,, ...... 34 Table 7. Adjusted effects estimates and P-values for log urinary mercury (Hg) level. . .36 vi LIST OF FIGURES Figurel. Estimated adjusted means of log 24-hour urine mercury versus time in three different regions. ........................................................................... 38 Figure 2. Estimated adjusted means of log urine mercury divided by urine creatinine versus time in three different regions ..................................................... 42 Figure 3. Estimated adjusted means of log urine mercury concentration versus time in three different regions ........................................................................ 44 Path Diagram 1: The possible causal model for normally transformed creatinine corrected urinary mercury (Hg) levels ................................................................. 46 Path Diagram 2: The possible causal model for normally transformed 24-hour urinary mercury (Hg) levels .......................................................................... 47 Path Diagram 3: The possible causal model for normally transformed urinary mercury (Hg) concentrations ................................................................................. 48 vii LIST OF ABBREVIATIONS TWI Toxic Waste Incinerator group RVC Rhine Valley Control group OWC Odenwald Control group ICC Intra-class correlation coefficient viii CHAPTER 1 Introduction Mercury has been widely used in dental amalgam, pesticides, batteries, paint, and thermometers and barometers by humans for a long time. The ubiquitous and persistent nature of mercury has made it an environmental and human health concern. Over the past decades there has been an increasing awareness throughout the world regarding the health and developmental risks associated with environmental exposure to mercury (WHO, 2003). Exposure to mercury could be occupational or residential, acute or chronic. Kidney and brain are the main targets of mercury exposure (Clarkson, 2002; Counter and Buchanan, 2004). Mercury is part of the composition of the earth’s crust and may be found in air, water, soil, aquatic sediments, living plants and animals. The general population is exposed to mercury in everyday life. According to Dr. Clarkson (1983), three forms of mercury are found in the environment, including elemental mercury (quicksilver, HgO), organic mercury compounds (e.g., methyl mercury, phenyl mercury), and inorganic mercurial compounds (e.g., ngClz, HgClz) (Clarkson, 1983). Mercury combined with carbon is called organic mercury; methyl mercury is a common example of organic mercury. Mercury compounds that contain non-carbon substances such as chlorine, oxygen, or sulfur are called inorganic mercurials. Elemental mercury is usually referred to as metallic mercury or mercury vapor, which is nonflammable and has low solubility in both water and organic solvents. Elemental mercury vaporizes at room temperature and forms a heavy, shiny, silver-white, odorless liquid. The toxicities of mercury forms are different depending on their specific property, such as their solubility, and chemical reactivity (Clarkson, 2002). Mercury occurs naturally in the environment, and the levels are increased by certain human activities such as mining, the burning of fossil fuel by power plants, waste incineration, and other industrial activities (Trepka, Heinrich et al., 1997). These activities increase the amount of airborne mercury, which is eventually deposited in fresh and ocean waters (WHO, 1991). Inorganic mercury, present in water sediments, is subject to bacterial and microorganism conversion to methyl mercury compounds that bio- accumulate in the aquatic food chain, reaching the highest concentration in predatory fish (Clarkson, 1997). Methyl mercury compounds are found in seafood and freshwater fish. Mercury intoxication can cause mental retardation, cerebral palsy, seizures, nephrotoxicity and death (WHO, 1990; Tominack, Weber etal., 2002). Studies on the background level of mercury exposure, resulting mainly from amalgam fillings or contaminated fish consumption, indicated an elevated mercury level among exposed people, but no significant association could be found between exposure and human health effects (Tulinius, 1995; Langworth, Bjorkman et al., 2002; Becker, Schulz etal., 2003; Schober, Sinks et al., 2003; Kingman, Albers et al., 2005). Animal studies indicate that long-term oral exposure to mercury may have adverse effects on the kidney, stomach, blood pressure, and heart rate (WHO, 1990; WHO, 2003). Hultman et a1. (1994) did a study on mice and they reported that “accumulation of heavy metals, from dental amalgam and other sources may lower the threshold of an individual metal to elicit immunological aberrations” (Hultman, Johansson et al., 1994). In a review article, Tchounwou et a1. (2003) indicated that a lower level of mercury exposure might be unsafe among genetically susceptible populations and a noncytotoxic level of Hg2+ may pose a risk on a subset of populations with disease susceptibility (Tchounwou, Ayensu et al., 2003). Amalgam fillings are one source of elementary mercury exposure in the general population (WHO, 1991; ATSDR, 1999; ATSDR, 2002; Schober, Sinks et al., 2003). Mercury is the principal metal in most dental fillings (approximately 50% by weight) (WHO, 1991; Nadarajah, Neiders et al., 1996). Mercury release increases with chewing, followed by absorption and uptake by body tissues, in particular, the brain and kidneys (WHO, 1991). Humans with amalgam fillings have significantly elevated mercury levels, with 3-5 times more mercury in their urine and 2-12 times more mercury in their tissues than those without amalgam fillings (Drasch, Schupp et al., 1994; Becker, Kaus etal., 2002; Zimmer, Ludwig et al., 2002; Becker, Schulz et al., 2003). Consumption of methyl mercury-contaminated food, especially fish, is a common route of acquiring methyl mercury. In addition, living in an industrial area with a hazardous waste site or an incinerator may also add to mercury exposures (Kurttio, Pekkanen et al., 1998; ATSDR, 1999). Industry waste and the combustion of fossil fuels released tons of mercury into the environment from 1970 to 1990, although Mercury losses from incineration processes decreased by 47 percent between 1990 and 1996 in the United States (Sznopek and Goonan, 2000). The resulting mercury vapor exerts a significant vapor pressure. When inhaled, elemental mercury vapor is nearly completely absorbed across the alveolar membrane, with a resulting retention of 75—80%. Children also have higher minute ventilation, which increases most inhalation exposures (Forman, Moline et al., 2000). School-aged children (6-12 years of age) may be exposed to more mercury vapor since they spend greater amounts of time outdoors. Exposure assessment is an important part in epidemiological studies (Symanski, Sallsten et al., 2000). Usually it is not possible to measure the exposure directly, so biomarkers (such as mercury in urine) are used as surrogates and are measured during a limited time period (Brunekreef, Noy et al., 1987). Such measurements can only approximate the individual exposure level. In environmental epidemiological studies, the variability of exposure measurements is of critical importance and needs to be considered and addressed during study design and data analysis, since the variation of exposure could introduce biases in estimating the effect (Brunekreef, Noy et al., 1987; Symanski, Sallsten et al., 2000). Studies based on the occupational exposure to mercury suggest that the intra—individual variation of the exposure measurement seems to be an important issue and may induce errors if not appropriately corrected when assessing the exposure effect (Barregard, 1993; Symanski, Sallsten et al., 2000). Similar findings have been reported by Bore] and his co-workers based on an iron-status assessment and the authors indicated that “day-to-day biological variation is a major component of the variability in the iron- status indicators and must be considered when assessing iron status” (Borel, Smith et al., 1991). Unlike other environmental pollutants, such as polychlorinated biphenyls (PCB) and dichlorodiphenyl dichlorethylene (DDE), which have long half-lives, up to several years (Longnecker, Rogan et al., 1997), mercury has a very short half-life, up to approximately 2 months (Counter and Buchanan, 2004). The shorter half-life of mercury in the human body combined with other factors (such as diet and number of amalgam fillings) may produce large intra-individual variations of mercury level over time. It is important to acquire knowledge about the magnitude of the variation in order to more accurately estimate the true exposure level. To date, most of the mercury exposure assessment studies or studies on the health effects due to mercury exposure were based on only one biological monitoring sample during the study period (Olstad, Holland et al., 1987; Schulte, Stoll et al., 1994; Tulinius, 1995; Batista, Schuhmacher et al., 1996; Khordi-Mood, Sarraf-Shirazi et al., 2001; Pesch, Wilhelm et al., 2002; Levy, Schwartz et al., 2004), which can not account for the temporal variation (usually called the ‘time’ effect). In addition, measurement error in cross sectional study can cause attenuation bias. Hence, a single measurement of the biomarkers is very likely to be a poor approximation of exposure. The primary objective of the present study is to exam the variability of urine mercury levels in three consecutive measurements among children collected from three different areas in Germany during 1994-1997. The second objective is to identify factors that affect urinary mercury levels among those children. Our first hypothesis is that the intra-individual variation of mercury level among the children is higher than the inter- individual variation after controlling for confounders. Analysis of variance for repeated measurements will be used to investigate the intra-individual or “error” variance and the inter-individual or true variance (Brunekreef, Noy et al., 1987; Symanski, Sallsten etal., 2000). Our second hypothesis is that living in the region with a toxic waste incinerator will result in increased mercury levels among the study children. In the south of the federal state of Hessen, Germany, an industrial toxic waste incinerator (TWI) was situated in the Rhine Valley with low mountains on each side. Other industries such as a chemical plant were nearby. Besides the incinerator, the region was also used for agriculture, including the production of vegetables. Children from the TWI region comprise the exposure group. The first comparison group was 20 km north of the incinerator, in the Rhine Valley, and was also industrial and agricultural in nature (Rhine Valley control, RVC). Southeast of the incinerator region was the second comparison group, with low mountains separating it from the industrial areas (Odenwald control, OWC). For the study population, we assume that mercury exposure via the breathing of the contaminated air from the TWI, the release of mercury vapor from dental amalgam, and fish consumption were the sources of mercury exposure. Hair, blood, and urine can be used to monitor exposure to mercury. Urinary mercury is considered to reflect the chronic exposure to inorganic mercury, while mercury in blood primarily reflects recent exposure and mercury in hair is used in monitoring long term methyl mercury exposure (Barregard, 1993; Pesch, Wilhelm et al., 2002; Gabrio, Benedikt et al., 2003). We used urine samples to monitor the mercury level in the study population. Mercury concentration in urine samples shows a circadian rhythm with highest concentrations in morning samples (Araki, Murata et al., 1983). For this reason, 24-hour urine samples were collected. In our analysis, three measures of urine mercury were used to test our hypotheses: 24-hour urinary mercury level, urinary mercury concentration, and the creatinine corrected urinary mercury. CHAPTER 2 Population and methods 2.1 Study population After approval by the data-protection agency in Hamburg, the Ministry for Cultural Affairs, and the local school committees, 1,091 second—grade school children in 18 townships were approached. The townships/primary schools were selected with respect to the willingness of the local school committees to cooperate. Informed consent according to requirements of the Ethical Committee of the Board of Physicians of the State of Hamburg and the data-protection agency in Hamburg was obtained from all participating parents. Children and their parents from nine townships/schools in the TWI, five in the RVC, and four in the OWC region were recruited for our study. 2.2 Questionnaires Three questionnaires were used in 1994-5, and were repeated in 1996 and 1997. One questionnaire asked for the family’s living conditions, including cigarette consumption in the home during the preceding year, place of residence, and fish consumption in the last year. From the questionnaire that was attached to the container for the 24-hour urine samples, the child’s fish consumption in the two days preceding urine collection were obtained. Therefore, fish consumption was assessed first by questions of in general how often the child ate different type of fish, and second, by fish consumption in two days before the collection of the urine sample. In the third questionnaire the parents were asked for the child’s gender, age, how many amalgam fillings the child had and when the most recent one was put in. In 1997, to test the reliability of the questionnaire, the questionnaires were administered a second time. Fifty-two families, including 53 children, returned the questionnaires. For the question “when did the child receive the new amalgam”, the Kappa is 0.85, for the question “whether the child has amalgam”, the Kappa is 0.89. For smoking information the weighted Kappa is 0.78. A complete agreement for Rhine River fish consumption were also found. 2.3 Collection of urine samples and the determination of urinary mercury and creatinine At the time of the physical examination one accompanying parent was instructed on how to collect a 24-hour urine sample. Flyers, suggested collecting the urine on a weekend, were also handed to their parents. Parents brought the urine sample to the clinic, where it was weighed, aliquoted, and frozen at -30°C. The samples were sent to the Institute for Toxicology, University of Kiel, Germany. Aliquots of 5 ml of urine were processed with 0.5 ml of 32% HCL and 0.2 ml of potassium-bromide/Potassium bromate (2 g KBr and 0.56 g KBr03 in 25 ml H20), and analyzed by flow injection atomic absorption spectroscopy (Perkin Elmer). The detection limit of Hg was 0.15 Mg Hg/L. Urine creatinine (mg/L) was determined using the Jaffe-Method. The 24-hour mercury level was calculated by multiplying the urinary mercury concentration (pg/L) times the amount of urine collected in the 24-hour period (U24-hour). The creatinine corrected urinary mercury (rig /g) was calculated using urinary mercury level (pig/L) divided by the urinary creatinine level (g/L). 2.4 Statistical methods 2.4.]. Data analysis To approximate normality, geometric means and corresponding 95% confidence intervals were calculated. The log-transformed urinary mercury levels were used as the dependent variables. Exposures of interest were the place of residence (Toxic Waste Incinerator region, Rhine Valley comparison group, Odenwald comparison group) and amalgam fillings. Potential confounders, such as gender, age, fish consumption, passive smoking status, urinary creatinine, and 24-hour amount of urine were included in the models. The interactions between the time when the sample was collected (1994-5, 1996, 1997) and number of amalgam fillings, and between the time of sample collecting time ion and place of residence were also assessed. Gender, place of residence, and the time of sample collection were categorical variables in the model. Age, number of amalgam fillings, fish consumption (meals/month), smoking (number of cigarettes per day), urinary creatinine (mg/L) and 24-hour amount of urine (mL) were used as continuous variables. To describe the data, fish consumption was grouped as S 0.05 (meals/month), 0.05 to S 0.1 (meals/month), 0.1 to S 0.15 (meals/month), and > 0.15 (meals/month). The number of cigarettes smoked per day by the children’ parents were divided in to 4 groups (0 cigarette /day, 5 cigarettes /day, 15 cigarettes /day and 25 cigarettes Iday). Urinary mercury levels in the three consecutive years could not be considered to be independent for the same child, which violates the “independent” assumption in the general linear model. Therefore, linear mixed modeling was used and restricted maximum likelihood estimates of the variance components were obtained using PROC MIXED, SAS software. 2.4.2. Linear mixed models: The linear mixed model used is: In (32,.) = X,- B + b,- + 8,, bi ~ N (0, 02b), eij ~ N (0, 02.) in here: i = l, 2, 3, ...... n children; j = 1”", 2nd, 3rd measurement index; In (yij) = the natural logarithm of the urinary mercury level; X ij = the fixed design vector for the ith child at the jth time point, and B is the corresponding slope. (X03 gender, age, place of residence, number of amalgam fillings, fish consumption and passive smoking status); I),- = the random effect, which measure subject heterogeneity; 8,]- = error term. Y 2 2 . . It follows that Var( ij) = 0' b + 0' 8 for all 1,}. It is assumed that b ,- and Sij are mutually independent and normally distributed . . 2 0,2 . 2 With zero means and variances 0 b» and 8, respectively. The termO' b represents the vanance between rndrvrduals; 0' 8 measures trme pornt variability and represents the . . . . . . 2 2 . . . . . variance wrthrn rndrvrduals. Thus, 0' b and 0' 8 represent the rnter-rndrvrdual and intra- individual variance components, respectively. We used the linear mixed modeling procedure (PROC MIXED) in SAS to estimate the variance (02b) for inter-individual vanatron and (0' 8) for the rntra-rndrvrdual variation. The intra-class correlation (ICC), which is the ratio of the inter-individual . . 2 2 2 . variance to the total variance [ICC= 0' b / (0’ b + 0' 8)], was used to quantify the 10 proportion of total outcome variation that was due to inter-individual variation. The ratio . 2 2 . . . A . . of the variances (0' 8 / 0' b) rs called variance ratro or/I . Computation of the variance ratio is a useful technique to evaluate the potential bias in regression coefficients obtained with measures of exposure (Brunekreef, Noy et al., 1987). Therefore, we reported the variance ratio (A ) in our results. 2.3 Structural Equation Models (SEMs) to detect the stability of urinary mercury SEM is a generalization of regression, path analysis, and factor analysis, and provides a way to formulate a model that integrates mediating variables. In SEM multiple regression equations can be written simultaneously, and dependent variables in one regression can be independent variables in others (Lehert and Dennerstein, 2002). For the analysis, we transformed urinary mercury values to normally distribute by computing the normal scores from the ranks (Blom, 1958). SEM was performed to test our theoretical model. All the causal factors were selected according to the literature and our data. The number of changes in amalgam fillings in one year compared to the previous year was represented by Al and A2 respectively. For example, Al was calculated using the number of amalgam fillings in 1996 minus the number of amalgam fillings in 1995; A2 was calculated using: the number of amalgam fillings in 1997 minus the number of amalgam fillings in 1996. Since both the number of amalgam fillings at baseline and the changes in number of fillings (Al and A2) will affect the mercury levels, both were included in the SEM. All analyses were conducted using the SAS System’s CALIS procedure. These analyses used the maximum likelihood method of parameter estimation. ll Chapter 3 Results 3.1 Study population Of the 1,091 children in the second grade, 671 (61.5%) participated in the study. The characteristics of the study population are displayed in Table 1. Among them, 663 children had at least one urine sample during the three years. In 1994-5, 623 children provided urine samples. Sixty-six out of 623 children who had provided urine samples in 1994-5 did not give a urine sample in 1996. However, 37 additional urine samples were collected among the rest of the children resulting in a total of 594 urine samples available in 1996. In 1997, 68 out of 594 children who had provided urine samples in 1996 did not provide urine samples. Twenty-four children who gave urine sample, in 1994-5 but not in 1996, provided urine samples in 1997. Additionally, 3 children without urine samples in both 1994-5 and 1996 had provided urine samples in 1997. A total of 553 urine samples were available in 1997. Out of 663 children, 496 (74.8%) participated in all urine collection during the 3-year study period. The majority were around 8 or 9 years of age in 1994-5, 9 or 10 years of age in 1996 and > 10 years of age in 1997. Of 663 children, 362 (54.5%) resided in the toxic waste incinerator (TWI) region and 360 of these children were there during the entire study period. In 1994-5, 173 (26.1%) children had amalgam fillings and most of them (n=l43) had between one to four amalgam fillings. Eighty-one (12.4%) children in 1996 and 45 (6.7%) children in 1997 had at least one amalgam filling. The percentage of amalgam fillings went down during the 3-year period. Fish consumption is low, with approximately 50% of children having less than 0.1 meal per month. Approximately half of the children had not been exposed to smoke during the 12 months before the urine sample was collected (Table l). 12 3.2 Unadjusted urinary mercury (Hg) levels Table 2 displays the median and geometric means of the urinary Hg concentration (jig/L), the creatinine corrected urinary Hg (pg/g creatinine), and 24-hour total urinary Hg (ugl24-hour) for the three years in the three different regions. For all children, both urinary Hg concentration and 24-hour urinary Hg increased during the three years. The creatinine corrected urinary Hg (pg/g creatinine) did not show this increasing trend (Table 2). There was no apparent difference in the three urine mercury levels in 1994-5 among the three research regions. In 1996, children in the TWI group showed higher mercury level in all three outcomes compared with the other two regions. However, in 1997 children in the OWC group had the highest mercury levels for all three outcomes (Table 2). Data on urinary mercury levels in children with and without amalgam fillings are shown in Table 2 with higher mercury levels in individuals with amalgam fillings. The urinary Hg levels for all three outcomes increased with each calendar year for both children with and without amalgam fillings. 3.3 Urinary creatinine and 24 hour total urine in three different years In 1996, boys had a higher urinary creatinine level than girls (p-value = 0.02) (Table 3). Children with amalgam fillings had a higher urinary creatinine level than children without amalgam fillings in 1994-5 (p-value = 0.05), but not in 1996 and 1997. In 1997 the mean urinary creatinine was higher in the RVC group (1.09 mg/L) compared with the OWC group (0.98 mg/L). l3 The mean24-hour amount of urine increased in the three years from 639 mL in 1994-5 to 704 mL in 1996, and 779 mL in 1997. The increases were expected since the amount of urine increases with age. Similar trends were found for each region and for children with and without amalgam groups, respectively. Boys had a higher urine amount than girls in 1994-5 (p-value = 0.04) (Table 4). 3.4 The stability of the urinary mercury levels The Spearman correlation coefficients of the urinary mercury levels for the three mercury measurements were low, ranging from 0.212 to 0.307 (Table 5). The low correlation coefficients indicate that there was substantial variation in mercury levels for each child over the three years. Adjusted estimates of the variance components on the log-transformed urinary mercury are shown in Table 6. There seems to be as much or greater variation within individuals as there is between individuals. A large variance ratio (/I ) indicates a larger intra-individual variation comparing to inter-individual variation. The intra-class correlation coefficient (ICC) is the proportion of the total mercury variance due to the variance between children. Hence a higher ICC indicates a higher inter-individual variation and a lower ICC indicates a higher intra-individual variation. A smaller ICC results in a larger/I. For example, in table 6, if one uses the log-transformed A urine concentration as the outcome, based on all children, A = 2.75 and ICC = 0.28. A variance ratio (A ) of 2.75 means that the intra-individual variance is 2.75 times the interdndividual variance. The ICC of 0.28 indicates that only 28% of the total variability was due to inter-individual variation and 72% can be attributed to other variances, including intra-individual variance. Similar patterns were shown when stratifying by 14 region and amalgam filling status. Children without amalgam fillings during the study period showed greater intra-individual variation (Table 6). 3.5 Associations between adjusted urinary mercury levels and other factors Adjusted for the other covariates listed in Table 7, the number of amalgam fillings had a significant effect on the three log-transformed urine Hg measurements. There were also significant interactions between time of measurement and number of amalgam fillings, and time of measurement and living region. The effect of amalgam fillings varied depending on the time of measurement. For instance, in 1994-5, with an increase of one amalgam filling, the log- transformed urine mercury concentration will increase 19%: (0.36 - 0.17 * time) * 100%: 19%, where time=l. Since there are significant interactions between the time when the sample was collected and number of amalgam fillings, the time when the sample was collected and living region, the changes of urinary mercury level between two years depend on both the region and the number of amalgam fillings (Table 7). For example: in TWI region, the log-transformed urine Hg concentration in 1994-5 versus 1997 is equal to (- 0.09) — 0.17*number of amalgam fillings + 0.33 *TWI, where TWI=1. According to the formula above, if the number of amalgam fillings is l, the log urine Hg concentration in 1994-5 minus the log urine Hg concentration in 1997 = 0.07, indicating the log urine mercury concentration in 1994-5 is 7% higher than it was in 1997. When the number of amalgam fillings is 2 or more, the log urine mercury concentration in 1994-5 minus the log urine Hg concentration in 1997 will be (- 0.09) - 0.l7*2 (or larger number) + 0.33 *TWI = - 0.1 or lower, which shows that the log urine Hg concentration in 1994-5 is 10% lower than that in 1997. 15 Figure 1 shows the changes of the adjusted means of log 24-hour total urine mercury in different years with different number of amalgam fillings at the mean values of all the other covariates. When the number of amalgam fillings increases, the changes in the log 24-hour total urinary mercury became larger from year to year. In 1996, the adjusted mean was higher among children living in the TWI area compared to the OWC group (p-value = 0.0023). However the TWI adjusted mean decreased in 1997 becoming the lowest of the three regions (Figure 1). Age had a significant effect on the three log urinary mercury outcomes (Table 7). Girls had higher urinary mercury levels when using log urine mercury/creatinine (jig/g) as the outcome (p = 0.0033). The 24-hour total amount of urine had a significant effect on the three outcomes. The results did not change when restricted to the children with all three measurements during the study period (data not shown). Urine creatinine also had a significant effect on log urine Hg concentration (p < 0.0001) and on log 24-hour urine Hg levels (p = 0.02). Path coefficients (pcoe) are shown in Path Diagrams 1, 2, and 3. The number of amalgam fillings (pcoe = 0.48, p-value < 0.05) and gender (pcoe = 0.12, p-value < 0.05) had a significant effect on the urine mercury/creatinine (jig/g) during the study period. The change in number of amalgam fillings in 1996 had an effect on the urine mercury/creatinine (jig/g), with pcoe = 0.25 (p-value < 0.05). The number of fish meals had a significant effect in 1995 (pcoe = 0.81, p-value < 0.05), but not in 1996 or 1997. We did not find a fish meal effect in our mixed model. Urinary mercury (Hg) level divided by urine creatinine in 1995 had a significant effect on the level in 1996 (pcoe = 0.14, p < 0.05) and 1997 (pcoe = 0.13, p < 0.05), and the urine mercury/creatinine (jig/g) l6 in 1996 also affected the level in 1997 (pcoe = 0.22, p < 0.05). Living in the TWI area had no effect on the urine mercury/creatinine (jig/g) in 1995 (pcoe = 0.10, p-value > 0.05), but had a significant risk effect in 1996 (pcoe = 0.33, p-value < 0.05), and a significant but diminishing effect in 1997 (pcoe = - 0.21, p < 0.05) (Path Diagram 1), which confirms the findings from the linear mixed model. The amount of total urine also significantly affected the urinary mercury level divided by urine creatinine (pcoe = 0.22, p < 0.05). When using normal-transformed 24-hour total urine mercury as the dependent variable, our findings were similar to those when we used normal-transformed urine mercury/creatinine (jig/g) as the outcome, with several differences: there was no gender effect, but age and urine creatinine showed significant effects on the outcome. The age effect was weak with pcoe=0.04, p <0.05, based on the one-sided t-test, which is consistent with our mixed model using the two-sided test. The path coefficients for urine creatinine were pcoe=0.09, (p <0.05) in 1994-5, pcoe=0. 13, (p <0.05) in 1996 and pcoe=0.l3, (p <0.05) in 1997 (Path Diagram 2). Path Diagram 3 displayed the factors affecting the normal-transformed urinary mercury concentration. 17 CHAPTER 4 Discussion 4.1 Study population Of the 1,091 eligible second grade children, 61.5% participated in the study (n = 671), and, therefore, it is unlikely that a selection bias can explain our findings. The distributions of factors, which may affect the mercury level in children, were similar among the three regions during the study period (data not shown). The reproducibility of the information was high. Kappas on questions of amalgam fillings, fish consumption, and smoking status were higher than 75%. Hence, we do not suspect an information bias could explain our findings. 4.2 The variability of mercury levels in urine The importance of assessing the intra- and inter-individual variation in biomarkers of exposure has been recognized in occupational and environmental epidemiology for a long time (Brunekreef, Noy et al., 1987; Rappaport, Symanski et al., 1995; Symanski, Sallsten et al., 2000; Symanski, Bergamaschi et al., 2001; Symanski and Greeson, 2002). We are not aware of any studies to date that have quantified the variability in urinary mercury levels in children. In the present study, we investigated the variability of the mercury levels among school-aged children residing in three different areas, who participated in a longitudinal study in Hesse, Germany. The low Spearman correlation coefficients of the urinary mercury levels (0.212 to 0.307) indicate that the mercury level in the human body was not stable. After adjusting for the other covariates in Table 7, the ‘time’ variable showed a significant effect on the three mercury measurements, which denotes that at least in two years the mercury levels were statistically different. The 18 adjusted point estimates of the variance components in the log-transformed urinary mercury showed a greater variation within children than between children. Children without amalgam fillings showed larger variance ratios than children with amalgam fillings. Since children with amalgam fillings had higher urinary Hg levels, the increase of number amalgam fillings posed a relatively smaller impact on urinary Hg levels comparing to children without amalgam fillings whose urinary Hg levels were 2 -3 times lower. In addition, in each year more boys (56.7%) gave urine samples among children without amalgam fillings, meanwhile approximately 50% of children who provided urine sample with amalgam fillings are boys. Since gender has significant effect on urine mercury levels, the greater within children variation may be caused by gender. Due to the large intra-individual variation, a single measurement of urinary mercury may not be adequate for estimating the “true” individual exposure level. Symanski et a1. quantified the intra- and inter-individual variation in urinary levels of mandelic acid and phenylglyoxylic acid among workers exposed to styrene (Symanski, Sallsten et al., 2001 ). They found that due to the intra-individual variation, estimates of workers' exposures that rely on single measurements would perform poorly in a regression analysis designed to examine effects resulting from chronic exposure. However, the bias in an observed slope coefficient would be diminished if a second or third urine sample were collected. Brunekreef et a1. (Brunekreef, Noy et al., 1987) indicated that when the intra-individual variation is large, repeated measurements for each individual could reduce the bias in a regression coefficient. The authors also suggested that the average of a number of measurements for the same subject was more reliable than one measurement. However, we think the average of several measurements 19 may also produce bias in estimating the “true” level because one outlier could drag the average level away from the “true” level. Instead of taking the average of several repeated measurements of exposure, one can use statistical methods for repeated exposure measurements to analyze the data, such as the mixed effect model, which can check if there is outlier problem by exploring the residual distribution. In case there are several exposure measurements in a long period and only one outcome measured at the end of the study, one could use other methods to handle the repeated exposure measurements, such as using “area under the curve” to estimate the cumulative exposure status. Given the advantages of repeated measurements, the following questions should be considered: 1) How large should the inter-measurement interval be? 2) How many measurements are needed? Since the half-life of urinary mercury is approximately two months and some factors among children, such as the number of amalgam fillings and diet, may change within one half-life, we suggest that at least three 24-hour urine samples should be collected in the first two-month period in order to monitor the variation of urinary mercury within one half-life. Then, two or more urinary mercury samples should be collected in subsequent two-month periods. By using this strategy we expect to monitor the variation of urinary mercury within one half-life and between two or three additional half-lives. Our suggestion is consistent with the recommended biological mercury monitoring strategy in the medical guideline produced in 1996 by the Health & Safety Executive (HSE) in the UK, which suggests using a usual urinary sampling frequency of 20 between 1 to 3-month, and more frequent sampling for subject close to the health guideline value (Mason, Hindell et al., 2001). A strength of our study is the large sample size (N=663), which increases the power to detect a large intra-individual variation. A limitation in our study is the long interval (approximately 1 year apart) between measurements. Symanski et al., in one study investigating the variance of aggregated data from five different data sets based on 53 workers with 123 measurements, indicated an increasing variability with the increased interval between measurements after controlling for factors that were likely to contribute to variability (S ymanski and Rappaport, 1994). In our study the observed large intra-individual variation could be caused by the long interval between measurements. We hope that future studies will incorporate a greater number of repeated measurements with shorter measurement intervals in order to provide detailed information on the choice of the optimal number of measurements and time intervals. 4.3 Factors that afi‘ect urinary mercury levels Among the study population, the number of amalgam fillings was strongly associated with the urinary mercury levels. After adjusting for place of residence, gender, age, fish consumption, smoking status, urine creatinine and 24—hour urine amount, the number of amalgam fillings had the most significant effect on the urinary mercury levels in both the linear mixed model and the structural equation model. Our results were consistent with previous findings that people with amalgam fillings have higher urinary mercury level compared to people without amalgam fillings (Olstad, Holland et al., 1987; Olstad, Holland et al., 1990; Schulte, Stoll et al., 1994; Trepka, Heinrich et al., i997; 21 Counter and Buchanan, 2004). Kingman et a1 estimated that 10 amalgam surfaces would raise urinary concentration by 1 ug of mercury per liter, which is twice the normal environmental background concentration (Kingman, Albertini et al., 1998). In our study, the increase in log mercury levels due to one amalgam filling is approximately 36%. Alternatively, an increase in 1 unit of the log urinary mercury concentration is associated with an increase of 2.7 amalgam fillings (without considering the interaction between the time of measurement and the number of amalgam fillings). Given the operation of the incinerator and other industrial activities in the TWI region, it was expected that the mercury would be higher in the TWI area than in the OWC area. Compared to the mercury levels in 1994/5, residing in the toxic waste incinerator area posed a risk of increased urinary mercury levels in 1996. However in 1997, the mercury levels were lower than those in 1994/5. The reason for this could be due to the percentage of children having at least one amalgam fillings were different in three regions in different years. In 1994/5, the percentage of children having at least one amalgam fillings is 25.9% in TWI, 30.2% in RVC, and 30.1% in OWC. In 1997, the percentage decreased to 5% in TWI, 3.6% in RVC, and 16.8% in OWC. The higher percentage of children having amalgam fillings in OWC in 1997 may produce the higher urinary Hg levels compare to children in TWI region. Other possible explanation may be that parents were aware of the possible risk of increasing the internal body burden of mercury in the incinerator area. So they tried to protect their child from mercury exposure, for instance, they may have suggested that their child not play outdoors, etc. The finding of decreased mercury levels in 1997 may also have occurred by chance. 22 Consistent with this result, Trepka et al. also failed to find higher mercury level in the polluted compared to the control region (Trepka, Heinrich et al., 1997). In the present study, gender showed a significant effect on mercury level when using the log creatinine corrected urine mercury as the outcomes. However, using the log creatinine corrected urine mercury as the dependent variable may produce bias and thus an artificial difference in urinary mercury levels between boys and girls can be generated because the urine creatinine level differed by gender (Mage, Allen et al., 2004; Barr, Wilder et al., 2005). We found that boys had a significantly higher creatinine concentration compared to girls in 1996. Recently, Schisterman et a1 evaluated four statistical methods for the analysis of PCB exposure, serum lipid and health risk (Schisterman, Whitcomb etal., 2005). They suggested that lipid standardization (the division of serum PCB concentration by serum lipids) is highly prone to bias. Hence, simple division by creatinine may also produce a bias. It is likely that the observed gender effect on urinary mercury levels results from a different creatinine concentration in boys and girls. A solution for this problem is not to divide urinary mercury level by urinary creatinine level, but to use urinary creatinine levels included in the model as a confounder. Hence the statistical significance of the independent variables would not be due to an association with the urinary creatinine level. 4.4 Structural Equation Models (SEMs) Using SEMs, we found that residing in the toxic waste incinerator area posed a risk of increasing the urinary mercury level in 1996 compared to 1995; the risk seems to be diminished in 1997. These results are consistent with those from the linear mixed model. Path Diagrams 1-3 showed that the urinary mercury level in 1995 can predict 10 - 23 16% of the increase in mercury levels in 1996 and l3 -l7% of the increase in 1997; and the mercury levels in 1996 predicted an 18 — 22% increase of the mercury levels in 1997. Those results indicated a low stability with a 10 — 22% fluctuation in the urinary mercury levels among the study population, which is in agreement with the results from mixed linear model. 4.5 Summary Analyses using both linear mixed models and SEMs showed that the number of amalgam fillings had a significant effect on urinary mercury levels. Residing in the toxic waste incinerator area posed a risk for increasing the urinary mercury level only in 1996, but seems to have a diminishing effect in 1997 from the results of both linear mixed model and SEMs. Based on our data we found that the intra-individual variation attributed more (approximately 72% - 74%) to the total variation compared with the inter-individual variation (approximately 26% - 28%). Since the large intra-individual variance may induce a biased estimation of the effect, repeated determinations of the biomarkers of the exposure with a shorter interval between two measurements may be one method that could be used to reduce the likelihood of the attenuation biases. Because the half-life of urine mercury is approximately two months, we recommend taking at least three measurements within one half-life (approximately two months), and then taking two additional measurements in at least two-month periods. We suggest that a repeated measurement approach with shorter time intervals between two measurements should be used to increase the accuracy of exposure assessments. Considering the characteristics and living conditions of different populations, some factors might only be generalized to children aged 8-11 years in Germany. However, the information regarding the intra- 24 individual and inter-individual variation provided in our study and our suggestions for future study designs should be useful to other investigators planning prospective studies for assessing mercury exposure among school aged children. 25 Tablel. Descriptive characteristics of the study population (n=663) Variable Categories 94/95 96 97 n (%) n (%) n (%) Gender of the Male 336 (50.7) 317 (47.8) 298 (45.0) Child Female 292 (44.0) 275 (41.5) 254 (38.3) Missing 35 (5.3) 71 (10.7) 111 (16.7) Age S 8 250 (37.7) 0 0 > 8 - 9 349 (52.6) 225 (34) 0 > 9 - 10 27 (4.1) 346 (52.2) 153 (23.1) > 10 3 (0.5) 24 (3.6) 399 (60.2) Missing 34 (5.1) 68 (10.2) 111 (16.7) Place of WI“ 360 (54.3) 360 (54.3) 362 (54.6) resident at each [web 117 (17.7) 116 (17.5) 108 (16.3) year owc° 168 (25.3) 155 (23.4) 139 (21.0) Missing 18 (2.7) 32 (4.8) 54 (8.1) Number of 0 448 (67.6) 510 (77.0) 497 (75.0) amalgam 1 35 (5.3) 41 (6.2) 31 (4.7) “km at ”C“ 2 53 (8.0) 25 (3.8) 9 (1.4) year 3 31 (4.7) 7 (1.1) 1 (0.2) 4 24 (3.6) 5 (0.8) 1 (0.2) > 4 30 (4.5) 3 (0.5) 3 (0.5) Missing 42 (6.3) 72 (10.9) 121 (18.3) Fish 3 0.05 141 (21.3) 156 (23.5) 174 (26.2) consumption > 0.05 - 0.1 260 (39.2) 247 (37.3) 247 (37.3) mi?) P“ > 0.1 - 0.15 97 (14.6) 102 (18.4) 50 (7.5) > 0.15 129(195) 86(13.3) 80(12.1) Missing 36 (5.4) 72 (10.9) 112 (16.9) Number of 0 338 (51) 347 (52.3) 318 (48.0) cigarettes per 5 142 (21.4) 130 (19.6) 136 (20.5) day (smomg 15 87 (13.1) 74 (11.2) 58 (8.8) by the parents dun-“g the last 25 62 (9.4) 45 (6.8) 41 (6.2) 12 months) Missing 34 (5.1) 67 (10.1) 110 (16.6) a: Toxic Waste Incinerator group (TWI) 26 b: Rhine Valley Control group (RVC) C: Odenwald Control group (OWC) 27 Amend 2mm; Acme.— w_ _.o -mmodv 036 .959 am 5 .o .898 -586 me .o mmod mmm -mw _ .o 2 Nd 02.0 mmm -03 .o 09 .o o3 .o mam 331 002 8006 8mm; 80— .— 3 _ .o .399 vmmd 68.9 VS .o .398 -NSd owed mood m2 lo: .o oomd om _ .o m2 .03 .o on 5 .o mbod 02 030 £33 ammo 83o mood 48.8 End .309 vm _ .o $8.8 Hood wad ~56 S _ Am— .o >26 mEd C 5 -m ~ mo mm 5 .o de C _ O>m Ammcd Among 83”; wood $8.9 a 5 Nd 58.9 52 .o $8.9 -056 $56 mood Sum $5 .o 3 _ .o 02 .o mvm -wm _ .o em _ .o flied New 3:. 9&3 aowd Swv. _ 85v. — vQ .o - 5 8.8 w: .o $8.8 3N6 63.8 4436 o:.o acme mmm .ofld mono mid Nmm -oEd 926 02.9 mmm 30— Ammcd Chm; Good 09 .o - _ No.8 >2 .o -Nvodv 02 .o 63.8 .mde Sod whod van Am _ A m3 .0 mm _ .o can -02 .o mm _ .O on _ .o 00m 092 Ammcd 3%; 8m _ A mood .33: OS .o -mmodv oomd -mhodv m -086 £56 need 80 -wm _ .o 3: .o hm _ .o #8 -0: .o m3 .o pod 80 L32 Com Com Com 53-53 53-50 53-50 620-5 9 ED .8602 : 620-5 9 20 «8:002 : 620-5 R20 5602 a 35530.8 33c em: Evaav Soném 05 EA 0:53 Set: 0080500 05530.8 05. SEE 5535:0080 mm 05.5 25,—. n_0>0_ b50505 505.5 09:0 .N 030,—. 28 88; 68.8 68.8 as Ed 888 End 888 Ed 888 mean 83d Ed Ed 8 -88 84d 88 8 88d and dead 8 58> 88 Ed 88d 88d 88d ween 48d 88.8 Ed 88.8 Ed 88.8 :5 88d 88d Sod 83. .33 Ed Ed 83 -Ed Ed 88d :4 as» 88d 88.8 SEN :8 Ed 88.8 88 48.8 dmmd 88.8 mans -Ed Ed Ed E 88d :8d Ed E .88 88d d8~d E 5:5 8888 E? 383 888 Ed 888 8de 28.8 88 888 -Ed Ed Ed E .83 add 88d E 82d Ed 8nd E 030 :e8d :edd 888d Ed 888 Ed 88.8 Ed 888 88d Ed Ed E -Ed Ed Ed E -Ed Ed Ed E u>m 88d 88? 8:: E _d 88.8 E d 888 Ed 888 28d Ed Ed 88.. -Ed Ed Ed 88 .83 Ed Ed 88 5:. 88 8:: 88c 68; Ed 888 8de -888 Ed 888 88d 88d 88d E .83 Ed Ed E -Ed Ed Ed 885 030 8:8 88d 888d 38d -88 Ed 888 Ed 888 88d Ed wood 4: .83 Ed Ed 8: 88.8 Ed Ed 8: u>m 3.308 N 030% 29 .CZOAUV 20 3% BEBE 8:02.300 oxoma no .229 :38 050880 ”a .388 _u>2 055820 East: 05 an 323v 3R3 coup—E850 3:82: East: as 6823—8 “w: bust: @88on 055880 of. ”a Em 88d dad E d wags 2 _.o .809 cg .o 63.8 mg .o .089 So -cood me .o we .o now -ofld m2 .c on ~ .0 c3 82d 9: .o v2 .o 8v 53> Awomd Good Good Ea 933 .969 wood A 3.8 m _ 5.0 68.8 wand -22 .o owmd Numd ow -gmd mcvd wcvd 0v -wmmd Need mwmd 9» 5;? 32 Ed 888d 88; 888d 83:3 $56 48.9 m: .o 63.8 Q: .o .83: So -Eod :wcd $5.0 mow 63 .o 02.0 mg .0 2m .: fio 546 OS .o E m 5;» 8.288 N 288 30 828338 Emu—Em ”a 5.8 3.8 28.8 883:.“ cad cow cad o_ m Nwd 2% 5053? nod @md 31de wmd and Amvdv *mod Sum 2‘de Saw—«Ea we." 3 mwd 5 $6 N: 559 A838 €28 608 wad om. mwd m2 $6 of 030 88.8 388 608 386 36 34 we mmd co; ad 3 _ Euo Omd 93 2 ~ U>m 88.8 €38 38.8 mad mam cod mmm mwd 91m 3:. €38 808 $8.8 wad me $6 an mwd 3m 2:0 2 .o MON 3:: *Nod cw.m 288 vmd amd Ammdv 3 Ed 8d :8 8d 88 88m Amvdv 93.8 $2: saw—Eu - - had mmm - - and can - - vwd owe =< 58 58 p.88 eigd 88 -m :32 c 028?; 88 -m :32 = 033.3 88 i :82 : A138 :85 509— Em Oamcmuwoho OGCD 88 E 28:86 8.5 38.8 8-38 5 82:85 2:5 88% 8:: E 20>»: 0558.20 bust: 35.5 .m 035. 31 853850 nawcflm ”a 888 88 GE 8 883:3 £5 SV 85 00m 3% 3 50:55 8d 8d :88 2d 8d 888 Ed 2 d :88 8 .5838“ SH mv 0 _ N. 0m 30 2 £5, Awhmv C 30 309 N. NR 02 mg mg 0m0 E 030 888 A88 A38 8 :0 8.0 005 mg 3.0 3.0 m: v: 3.0 a“ .0 mmc I U>m $me Ammmv 300 m awn 00m mac mmm wmo Vm SE. 888 A80 :88 _ 0E. vmm 50 EN NE 0m 8:5 mod mmd 20$ 3.0 $4 30$ 30.0 mad Am 8 O m 8: 8m 88 8 8 so 8 88m 8me A30 :00 0 56:5 - - 0P mmm - - wow wow - - one No =< a8 58 A. 88 853-0 88 -n. :32 : 028>d 88 -m 2882 : 0285-0 88 i :32 a 30 30 30 Bamfi :~ Oct: %0 u=508< 003 5 8E: 0o E=OE< 8-88 5 2E: .6 asoe< 8888 0358886 025 E 889:8 8E: 00 8585 ~88 .50: .. vm 88:0 .v 808,—. 32 $338888“ 33 US“ 83 .méoa E cots—828:8 mm bust: of. no 5038838 39 can 82 .méoa E 8.3838 .8». @8333 £38 w: Duct: 2F b 5382.88 32 Ba 82 .Eoe 5 222 m: $5.5 52$ a wmm _ooo.v * :VNd NNm * ~80.V :Nd cam coo; 3m “LOOOV o—Nd hum * SOOV camd ONW * ~OOO.V thd com coo; com * ~OOOV VmNd omm * ~OOO.V ENC own * ~OOO.V owmd vow o2: 5mm .LOOOV 2N6 : 8.138 .800 cocoUIOm c 033$ #80 053:8.me : 8527a .boU couUDlOm : 033d .boU anvm .53!va = 883?; .boU omDvmlOm : oflgd .boU «332$..me 0 380-0: o 8:86: p bauUDIOI a cQUDIOm N 323103 a cabana: “838 >588: Sat: .8 856580 sous—8.80 58.8on .m 8an 33 vmd mm.m md 2d 2: U>m wmd mmd mud omd com 5:. mcofiom cmd ad add vmd new :8220 =< a Evfiwé >588: bust: Sop—LVN cogemcmb-w3 cod o2; wmd vod on». 3223 wvd 2 .8 dwd mud mdm 5;? mwcac Saw—«:2 3d afim and Ed m2 030 vmd afim omd 2d 2: U>m wad ohm mud wad com 3:. 8.23m cmd $.N nod vmd new :8230 =< 105.588 whé 055380 cat: .3 BEZU m: 05.5 85.89.388-34 wdd Sufi end mod 03 50:55 on d 2 8 5d dud mom 5;? 8.55 883:2 Ed 34» :d 2d m3 030 mmd Sim E d m 8 d 2: U>m wmd mmd mud wad den :59 mcommom wmd Qua :d omd 80 58:5 =< a SE3 :osmhcoocoo 8E: n~§ofi§§w3 w 3855; 3:232: 3889; 126365 A «0+ 388 $8353 -885 c8333 .88.: n8mz€ meo 050:3... 3 0.50 .0> 0o owd mod 5d 00d mod wod who mod mod 0.50... 355.0 505050 00 n no 0.50 .m> meo *0000 00.0 2 .0- 300.0 00.0 £0- .. .00 00.0 t .0- 2:0... 00500 8000050 00 0 .Looo. v wooood ooood .Looo. v nooood woood *mood oooood moood- 0.5050 05.5 *mod nod w_.o I I I * dooo. v nod omd 0550080 E05: *wood Ed omd- *mod 2d wad- *wood Q d mmd- no .m> 00 85.0. _w.o 2d 2d- mud 2d cod hmd Ed ood- 8 .0> meo 0.5... NNd wood wood 5d mood wood woo mood wood w5onm Nd 2d mmd m _.o omd omd 2 .o 3 .o omd 00005 50E ._. _o.o cod 2d *mod cod id *mood cod 2d 0m< 2d cod mod *mood cod 2d ood cod o_.o 0.8m .0> 0:5 Ed 2d 8d- cod 2d wmd- mod N. d “No. 030 .0> U>m *mood _.o omd- ._. 8d 3 .o mmd- *wood 2 .o omd- 030 .m> SE. .2000. v 00.0 00.0 1000. v 00.0 00.0 300. v 00.0 00.0 00500 800080 00 0 .Looo. v Nod ow- .Looo. v moo. Nam- .Looo. v Ed m.m- 50885 0000-: c.0003 9000-: 050>¢ 0 um 80.500m 050>¢ 0 mm 8055M 050>d 0 Mm 85500m 08050200 Ewflwav 8550080 @qu 0550080 3E3 50:-wm 0020 mm 05.5 mod 05.5 3 5055—0 m: 05.5 m3 5808:0050 mm 05.5 mo..— u_0>0_ Ammo 55008: 0055 m2 .08 0050>i 5.0 08005000 808.00 0855< H 030,—. 36 000:0 00000000 ”0 $306 wofimd Sod V 2d 2d Nfid Nod 3d omd cmwmd woomd Soc. V woo woo Nod 3d 26 and oowmd woomd Boo. V 2d 26 Qd 2d a .o cod 8 0E: .0> co 08:...U>m 8 0E: .m> meo 0E_0*U>m 3 0:5 .0> do 08:0;39 8.308 h 030% 37 Estimated adjusted means of log 24-hour urine mercury Figurel. Estimated adjusted means of log 24-hour urine mercury versus time in three different regions: When number of amalgam fillings = O 0‘ -O.5000‘ —1.0000‘ -1300. -2.0000‘ _2.50004"--'--.0,,,m__::::: - - - .. A .. .- —- "‘ ‘flfl'.’ ’ :: ‘7' ..................... —3.0000< I 95 96 97 "‘TWI RVC ~~ owe Time (in year) 38 Estimated adjusted means of log 24—hour urine mercury Figurel. Estimated adjusted means of log 24-hour urine mercury versus time in three different regions (cont’d): When number of amalgam fillings = l —O.5000‘ -1.0000‘ -1.5000* -2.CX)00‘ —2.5000‘ — (3.00001 . Y % w 97 ""TW| ""RVC "°'°OWC Time (in year) 39 Figurel. Estimated adjusted means of log 24-hour urine mercury versus time in Estimated adjusted means of log 24-hour urine mercury When number of amalgam fillings = 4 (n=29) three different regions (cont’d): 0. —0.5000‘ -1.0000* -1.5(D0‘ — 2.0000‘ —2.5000‘ — 3.0000 I 96 "'" TWI "" RVC "°'° OWC Time (in year) 40 97 Estimated adjusted means of log 24-hour urine mercury Figurel. Estimated adjusted means of log 24-hour urine mercury versus time in three different regions (cont’d): When number of amalgam fillings = 5 (n=15) —0.5000‘ -1.0000* —1.5000‘ — 2.0CO0‘ -2.5000‘ - 3.0000 * v v 95 % 97 "'" TWI "’" RVC "'” OWC Time (in year) 41 Figure2. Estimated adjusted means of log urine mercury divided by urine creatinine versus time in three different regions: When number of amalgam=0 -O.5000‘ -1.0000‘ Estimated adjusted means of log urine mercury d1v1ded by unne creatmine -8.CD00* . 95 % 97 "" TWI "'" RVC ”'" OWC Time (in year) 42 Figure2. Estimated adjusted means of log 24-hour urine mercury versus time in three different regions (cont’d): When number of amalgam=4 (n=29) —1.oooo-~"" -1.5000* -2.0000* —2.5000* Estimated adjusted means of log urine mercury d1v1ded by unne creatlnine - 3.CX)00‘ r . 95 % 97 ""' TWI “"'" RVC "°"° OWC Time (in year) 43 Figure3. Estimated adjusted means of log urine mercury concentration versus time in three different regions: When number of amalgam=0 -0.5000* -1.0000+ -1.5000‘ — 2.0000‘ -2.5000‘ Estimated adjusted means of log urine mercury concentratlon —3.0000‘ . . % m 97 "" TWI ”“'" RVC ’°'° OWC Time (in year) Figure3. Estimated adjusted means of log urine mercury concentration versus time in three different regions (cont’d): When number of amalgam=4 (n=29) —0.5000‘ —1.0000‘ -1.5000‘ -2.0000* -2.5000‘ Estimated adjusted means of log urine mercury concentratlon — 3.0000 ‘ 1 . 95 % 97 "" TWI "‘ RVC "°'° OWC Time (in year) 45 Path Diagram 1: The possible causal model for normally transformed creatinine corrected urinary mercury (Hg) levels: #ofarnalgamsI I GenderI I Age 7 I Alj I f2 I I# ofamalgarrEI I Gender—I L Age I I A1 I 045* 012* 0.03 0.14 0.03 # of amalgamsI I Genderj I Age 0 54* 0 12* 0.03 0 25* 048* 012* 0.’/| Normally transformed urinary Normally transformed urinary Normally [ramf‘med 0 14* Hg level divided by urine 0 22* . . . urinary Hg level divided by > creatinine (1996) ' Hg 16%] d1v1ded by unne urine creatinine (1994-5) creatinine (1997) A 0.810.10 -00x00:\:&64 IFish mealsI ITW II two I ISMKII IAmountI 0' 033* 0.08 0.18* 0.13* . 021* —00 0.03 0015 IFish mealsI II TWI I I RVC I I—ihf-K—‘I IAmountI Fish meals TWI RVC I SMK I IAmountI *: p < 0.05 two-sided test. A1 = the number of amalgam fillings in 1996 minus the number of amalgam fillings in 1995; A2 = the number of amalgam fillings in 1997 minus the number of amalgam fillings in 1996; TWI: Toxic Waste Incinerator Group; RVC: Rhine Valley Control Group; Cre: creatinine in urine; Amount: 24—hour total urine amount. SMK: Number of cigarette smoked by their parents. 46 tinine corrected urinary mercury (Hg) levels: # of amalgams Gender Age Al A2 ] A1 045* 012* 0.03 0-14 0.03 0.25”“ o I ’ rlnary * Normally transformed urinary *6 0-22 Hg level divided by urine creatinine (1997) 0. 18* O. 13* 0.1 .0 - (1 Amount .21* 0. 0.03 0015 Fish meals TWI R VC SMK Amount linus the number of amalgam fillings in 1995; A2 = the . in 1996; TWI: Toxic Waste Incinerator Group; RVC: :al urine amount. SMK: Number of cigarette smoked by in “ma Path Diagram 2: The possible causal model for normally transformed 24—hour urinary mercury (Hg) levels: # of amalgams I# of amalgamgI I GendeTI I—Ag:I # of amalgamsI 0.49* Normally transformed 24— 010* Gender I Age I 058* 0.08 0.04** 0.08 00/ , , J 030* Al L hour urinary Hg level (1994-5) Normally transformed 24—hour 048* 0.08 0.04** 0-16 0.04 / I Fish meals I 0.59 031* 0 El l1] Hg level (1996) 018* .04 0.06 0.13* 0.39”< 015* C ISMKI ‘ mount 0.06 Normally transformed 24—hour Hg level (1997) -0.27* -0.06 —0.002 016* FiJmeals TWI RVC SMK L 7| H H L 1 **: p < 0.05 one—sided test; *: p < 0.05 two-sided test. A1 = the number of amalgam fillings in 1996 — the number of amalgam fillings in 1995; A2 : the number of amalgam fillings in 1997 — the number of amalgam fillings in 1996; TWI: Toxic Waste Incinerator Group; RVC: Rhine Valley Control Group; Cre: creatinine in urine; Amount: 24-hour total urine amount; SMK: Number of cigarette smoked by their parents. 47 0.4l* ur urinary mercury (Hg) levels: # of amalgams Gender Age A 1 A2 A1 048* 0.08 0.04** 0-16 0.04 0* 7 hour * Normally transformed 24—hour 0-13 Hg level (1997) 0.41* 0.15* 0.06 -0.06 -0.002 016* -0.27 Amount Amount \ . Fish meals TWI RVC SMK Cre nalgam fillings in 1996 - the number of amalgam fillings :am fillings in 1996; TWI: Toxic Waste Incinerator ,: 24-h'ottr total urine amount; SMK: Number of cigarette Path Diagram 3: The possible causal model for normally transformed urinary mercury (Hg) concentrations: # of amalgams Gender Age Al A2 Imamalgamsj IiGender I I Ag€4I A1 55* 0- _ 0.07 004* 0-16 0.04 #of amalg;i:sI I Gender I I Age 0 07 0 04* 0 30* 049* 0.07 004* ' V V J Normally transformed * Normally transformed urinary * Normally transformed urinary Urinary mercury (Hg) 0-16 r mercury (Hg) concentration 0-18 r mercury (Hg) concentration concentration (1994-5) (1996) (1997) —0.06 0.002 013* . o (%4 0.07 0.13*}—0.07 017* Tim] ITish mealsI VTWIJ I RVC I I SMK I *: p < 0.05 two-sided test. A1 = the number of amalgam fillings in 1996 - the number of amalgam fillings in 1995; A2 = the number of amalgam fillings in 1997 - the number of amalgam fillings in 1996; TWI: Toxic Waste Incinerator Group; RVC: Rhine Valley Control Group; Cre: creatinine in urine; Amount: 24—hour total urine amount; SMK: Number of cigarette smoked by their parents. 48 References Araki, S., K. Murata, et al. (1983). "Circadian rhythms in the urinary excretion of metals and organic substances in "healthy" men." Arch Environ Health 38(6): 360-6. ATSDR (1999). Toxicological profile for mercury (update). Atlanta, US. Dept. Public Health, Atlanta, GA, US Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry. ATSDR (2002). Pediatric environmental health. Atlanta,U.S. Dept. Public Health, Atlanta, GA, US Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry. Barr, D. B., L. C. Wilder, et al. (2005). "Urinary creatinine concentrations in the US. p0pulation: implications for urinary biologic monitoring measurements." Environ Health Perspect 113(2): 192-200. Barregard, L. (1993). "Biological monitoring of exposure to mercury vapor." Scand J ng Environ Health 19 Suppl 1: 45-9. Batista, J ., M. Schuhmacher, et al. (1996). "Mercury in hair for a child population from Tarragona Province, Spain." Sci Totaflinviron 193(2): 143-8. Becker, K., S. Kaus, et al. (2002). "German Environmental Survey 1998 (GerES III): environmental pollutants in the blood of the German population." Int J Hyg Environ Health 205(4): 297-308. Becker, K., C. Schulz, et al. (2003). "German Environmental Survey 1998 (GerES III): environmental pollutants in the urine of the German population." Int J Hyg Environ Health 206(1): 15-24. Blom, G. (1958). Statistical estimates and transformed beg-variables. New York, John Wiley & Sons, Inc. Borel, M. J ., S. M. Smith, et al. (1991). "Day-to-day variation in iron-status indices in healthy men and women." Am J Clin Nutr 54(4): 729-35. Brunekreef, B., D. Noy, et al. (1987). "Variability of exposure measurements in environmental epidemiology." Am J Epidemiol 125(5): 892-8. Clarkson, T. W. (1983). "Mercury." Annual Review of Public Hefl 4: 375-380. Clarkson, T. W. (1997). "The toxicology of mercury." Crit Rev Clin Lab Sci 34(4): 369- 403. 49 Clarkson, T. W. (2002). "The three modern faces of mercury." Environ Health Perspect 110 Suppl 1: 11-23. Counter, S. A. and L. H. Buchanan (2004). "Mercury exposure in children: a review." Toxicol Appl Pharmacol 198(2): 209-30. Drasch, G., I. Schupp, et al. (1994). "Mercury burden of human fetal and infant tissues." Eur J Pediatr 153(8): 607-10. Forman, J ., J. Moline, et al. (2000). "A cluster of pediatric metallic mercury exposure cases treated with meso-2,3-dimercaptosuccinic acid (DMSA)." Environ Health Perspect 108(6): 575-7. Gabrio, T., G. Benedikt, et al. (2003). "10 years of observation by public health offices in Baden-Wurttemberg--assessment of human biomonitoring for mercury due to dental amalgam fillings and other sources." Gesundheitswesen 65(5): 327-35. Hultman, P., U. Johansson, et al. (1994). "Adverse immunological effects and autoimmunity induced by dental amalgam and alloy in mice." Faseb J 8(14): 1183-90. Khordi-Mood, M., A. R. Sarraf-Shirazi, et al. (2001). "Urinary mercury excretion following amalgam filling in children." J Toxicol Clin Toxicol 39(7): 701-5. Kingman, A., J. W. Albers, et al. (2005). "Amalgam exposure and neurological function.‘ Neurotoxicologv 26(2): 241-55. Kingman, A., T. Albertini, et al. (1998). "Mercury concentrations in urine and whole blood associated with amalgam exposure in a US military population." J Dent Res 77(3): 461-71. Kurttio, P., J. Pekkanen, et al. (1998). "Increased mercury exposure in inhabitants living in the vicinity of a hazardous waste incinerator: a 10-year follow-up." Arch Environ Health 53(2): 129-37. Langworth, S., L. Bjorkman, et al. (2002). "Multidisciplinary examination of patients with illness attributed to dental fillings." J Oral Rehabil 29(8): 705-13. Lehert, P. and L. Dennerstein (2002). "Statistical techniques for the analysis of change in longitudinal studies of the menopause." Acta Obstet Gynecol Scan_d 81(7): 581-7. Levy, M., S. Schwartz, et al. (2004). "Childhood urine mercury excretion: dental amalgam and fish consumption as exposure factors." Environ Res 94(3): 283-90. Longnecker, M. P., W. J. Rogan, et al. (1997). "The human health effects of DDT (dichlorodiphenyltrichloroethane) and PCBs (polychlorinated biphenyls) and an overview of organochlorines in public health." Annu Rev Public Health 18: 211- 44. 50 Mage, D. T., R. H. Allen, et a1. (2004). "Estimating pesticide dose from urinary pesticide concentration data by creatinine correction in the Third National Health and Nutrition Examination Survey (NHANES-III)." J Expo Anal Environ Epidemiol 14(6): 457-65. Mason, H. J., P. Hindell, et al. (2001). "Biological monitoring and exposure to mercury." Occup Med (Lond) 51(1): 2-11. Nadarajah, V., M. E. Neiders, et al. (1996). "Localized cellular inflammatory responses to subcutaneously implanted dental mercury." J Toxicol Environ Heal—th 49(2): 113- 25. Olstad, M. L., R. 1. Holland, et a1. (1990). "Effect of placement of amalgam restorations on urinary mercury concentration." J Dent Res 69(9): 1607-9. Olstad, M. L., R. I. Holland, et al. (1987). "Correlation between amalgam restorations and mercury concentrations in urine." J Dent Res 66(6): 1179-82. Pesch, A., M. Wilhelm, et a1. (2002). "Mercury concentrations in urine, scalp hair, and saliva in children from Germany." J Expo Anal Environ Epidemiol 12(4): 252-8. Rappaport, S. M., E. Symanski, et al. (1995). "The relationship between environmental monitoring and biological markers in exposure assessment." Environ Health Perspect 103 Suppl 3: 49-53. Schisterman, E. F., B. W. Whitcomb, et al. (2005). "Lipid Adjustment in the Analysis of Environmental Contaminants and Human Health Risks." Environmental Health Perspectives 113: 853-857. Schober, S. B., T. H. Sinks, et al. (2003). "Blood mercury levels in US children and women of childbearing age, 1999-2000." J AMA 289(13): 1667-74. Schulte, A., R. Stoll, et a1. (1994). "Mercury concentrations in the urine of children with and without amalgam fillings." Schweiz Monatsschr Zahnmed 104(11): 1336-40. Symanski, E., E. Bergamaschi, et al. (2001). "Inter- and intra-individual sources of variation in levels of urinary styrene metabolites." Int Arch Occup Environ Health 74(5): 336-44. Symanski, E. and N. M. Greeson (2002). "Assessment of variability in biomonitoring data using a large database of biological measures of exposure." AIHA J (Fairfax, Ya) 63(4): 390-401. Symanski, E. and S. M. Rappaport (1994). "An investigation of the dependence of exposure variability on the interval between measurements." Ann Occup Hyg 38(4): 361-72. 51 Symanski, B., G. Sallsten, et al. (2000). "Variability in airborne and biological measures of exposure to mercury in the chloralkali industry: implications for epidemiologic studies." Environ Health Perspect 108(6): 569-73. Symanski, B., G. Sallsten, et al. (2001). "Heterogeneity in sources of exposure variability among groups of workers exposed to inorganic mercury." Ann Occur; Hyg 45(8): 677-87. Sznopek, J. L. and T. G. Goonan. (2000, June 14). "The Materials Flow of Mercury in the Economies of the United States and the World." Retrieved Sep. 18, 2005, from http://pubs.usqs.qov/circ/2000/c1 197/c1 197. pdf. Tchounwou, P. B., W. K. Ayensu, et al. (2003). "Environmental exposure to mercury and its toxicopathologic implications for public health." Environ Toxicol 18(3): 149- 75. Tominack, R., J. Weber, et al. (2002). "Elemental mercury as an attractive nuisance: multiple exposures from a pilfered school supply with severe consequences." Pediatr Emerg Care 18(2): 97-100. Trepka, M. J ., J. Heinrich, et al. (1997). "Factors affecting internal mercury burdens among eastern German children." Arch Environ Health 52(2): 134-8. Tulinius, A. V. (1995). "Mercury, dental amalgam fillings and intellectual abilities in Inuit school children in Greenland." Arctic Med Res 54(2): 78-81. WHO (1991). Environmental Health Criteria 118: Inorganic mercury. Geneva, World Health Organization: 168. WHO (2003). Elemental mercury and inorganic mercury compounds: human health aspects. Geneva, World Health Organization. WHO, W. H. O. ( 1990). Methylmercury, International programme on chemical safety. 101: 1-144. Zimmer, H., H. Ludwig, et al. (2002). "Determination of mercury in blood, urine and saliva for the biological monitoring of an exposure from amalgam fillings in a group with self-reported adverse health effects." Int J Hyg Environ Health 205(3): 205-11. 52 IIIIIIIIIIIIIIIIIIIIIIIIIIIII 1|iIIIHIIIJIIIIIIIIIIIIIIIIIIIII'll