Document MMYroxQnyp6NZY8XVny4ZgnbV
MICHIGAN DIOXIN EXPOSURE STUDY
PREDICTORS OF SERUM 2378-TCDD CONCENTRATION IN A BACKGROUND POPULATION IN MICHIGAN, USA
Gillespie BW1, Chang C-W2, Hedgeman E2, Hong B2, Chen Q1, Jolliet O2, Knutson K,2 Lee S-Y1, Lepkowski J3, Olsen K3, Adriaens P4, Demond A4, Towey T4, Zwica L2, LaDronka K3, Ward B3, Luksemburg W5, Maier M5, Franzblau A2, and Garabrant D2 1Department of Biostatistics, University of Michigan School of Public Health, 109 S Observatory, Ann Arbor, MI 48109 USA; 2Environmental Health Sciences, University of Michigan School of Public Health, 109 S Observatory, Ann Arbor, MI 48109 USA; 3Institute for Social Research, 426 Thompson St., University of Michigan, Ann Arbor, MI 48104 USA; 4Department of Civil and Environmental Engineering, University of Michigan College of Engineering, 1351 Beal, Ann Arbor, MI 48109; 5Vista Analytical Laboratory, Inc., 1100 Windfield Way, El Dorado Hills, CA 95762 USA
Keywords: Human samples, Serum, Food consumption, Environmental samples, Background
Abstract: Background U.S. levels of 2,3,7,8-TCDD were estimated for a 2-county area of Michigan (the University of Michigan Dioxin Exposure Study), and for the entire U.S. using a population-based sample (National Health and Nutrition Examination Survey). Mean levels by age and gender are presented. The additional effects of weight loss, smoking, breastfeeding, and other variables were also investigated using regression models. In general, the vast majority of background values are less than 10 ppt. Larger values tend to occur with older age and female gender. Weight loss tends to increase serum concentrations, and both smoking and breastfeeding tend to reduce serum concentrations.
Introduction: Serum dioxin concentrations in heavily exposed populations are often compared to populations exposed only to background levels. Modeling dioxin as a function of demographic and lifestyle variables in a background population can allow prediction of background values in different populations, and can aid in understanding the mechanism of dioxin elimination.
This investigation was carried out using two data sources: (1) the University of Michigan Dioxin Exposure Study (UMDES), and (2) the National Health and Nutrition Examination Survey (NHANES). The UMDES collected samples of serum, house dust, soil, and vegetation from subjects (and their homes and property) living in Michigan, USA in areas potentially exposed to dioxin-like compounds as well as areas presumably exposed only to background levels of these compounds. The NHANES collected data from non-institutionalized residents of the U.S., and measured serum dioxin levels in 1228 of these subjects. This study presents prediction models for serum 2,3,7,8-TCDD levels in both the UMDES background region and the NHANES study.
Materials and Methods: The UMDES was carried out in Michigan, USA, in Midland, Saginaw and parts of Bay County (all potentially exposed areas) and in Jackson and Calhoun Counties (control areas, and the focus of this analysis). A two-stage probability household sampling design was used.1 Eligible subjects were at least 18 years of age, lived in their current residence for at least 5 years, and provided written informed consent. A detailed exposure questionnaire asked several hundred questions including demographics, smoking history, pregnancy history, occupational exposure, food consumption, and other questions possibly related to human body burden. In addition, serum samples were collected from subjects who consented and were medically eligible to give blood as defined by the American Red Cross. Chemical analyses were performed by Vista Analytical Laboratory, Inc. (El Dorado Hills, California, USA) for the World Health Organization designated 29 PCDD, PCDF, and dioxin-like PCB congeners using US Environmental Protection Agency (EPA) methods 8290 and 1668.3,4 Values below the limit of detection (LOD) were identified, and the LOD was given in each case.
The National Health and Nutrition Examination Survey (NHANES) is a publicly available database of national health and vitality information for a sample of the United States population. The NHANES data were taken from the 2001-2002 sample release. All individuals answered a general questionnaire covering health, diet and social-demographics; medically eligible persons were asked to donate a blood specimen, and a sub-sample of 1228 persons aged 20 years and older was selected for additional analysis of serum dioxins and furans.
Organohalogen Compounds Vol 69 (2007)
P-303
2264
MICHIGAN DIOXIN EXPOSURE STUDY
Chemical analyses and lipid measurements were performed by the Centers for Disease Control and Prevention laboratories. Results below the LOD were identified, and reported as LOD/2.
Statistical Methods: For the UMDES data, the log10 serum dioxin level was the outcome variable in a linear regression model with demographic, lifestyle, food, and other variables as potential predictors. Values below the LOD were replaced with the LOD/2. Covariates included age, sex, BMI, BMI loss in the past year,
pack-years of smoking, months of breastfeeding the first child, and number of incomplete pregnancies.
Selected interactions were tested. Residual plots were examined to confirm that model assumptions of
normality and homoscedasticity were met. Backward selection was used to remove non-significant variables.
The regression models used survey weights to adjust for the probability of sample selection and allow inference
to the entire two-county population (SAS proc surveyreg).
The NHANES regression analyses required a different strategy because the majority of values (87%) were below
the LOD. We used likelihood-based regression methods for left-censored data, assuming a lognormal distribution (SAS proc lifereg). We used the generalized R2 to estimate explained variation, although it may
yield an underestimate with censored data. Unfortunately, this method did not permit the inclusion of survey
weights, which would have allowed inference to the entire U.S. population. We used the same covariates as in
the UMDES model, with one exception. The NHANES data did not collect the months of breastfeeding the first child, but instead collected the number of children breastfed. SAS version 9.1 was used for all analyses.
Results and Discussion: The UMDES interviewed over 1300 subjects, of whom 946 provided serum samples,
251 from Jackson/Calhoun. Of the Jackson/Calhoun samples, 21% fell below the LOD. The median LOD
was 0.5 ppt. In the NHANES study, 1228 subjects had serum 2,3,7,8-TCDD levels measured, although 87%
were below the LOD. The median LOD was 2.8 ppt, and the mean (s.d.) LOD was 2.9 (1.0). Figure 1 shows
similar overall effects of age and gender, although the UMDES estimates are slightly larger. The estimates
from UMDES are more precise because the data are based on larger sample volumes, which led to smaller LOD
values.
(a) UMDES
(b) NHANES
15 15
Observed values: Females Observed values: Males Predicted median: Females Predicted median: Males
LOD: Females LOD: Males Observed values: Females Observed values: Males Predicted median: Females
Predicted median: Males
10 10
55
0 20 30 40 50 60 70 80
0 20 30 40 50 60 70 80 Age
Figure 1. Plots of serum 2,3,7,8-TCDD data and model predicted means by age and gender for both (a) UMDES and (b) NHANES. In (b), for values below the LOD, the LOD itself is plotted (grey circles). One value of 42.7 ppt for a 65-year-old male (BMI=39.9, BMI loss in past year=1.1) was omitted from the NHANES plot to show more detail for the other values.
Characteristics of subjects from the two groups are given in Table 1 below. The two groups are reasonably similar with respect to demographic characteristics.
Serum TCDD (ppt) Serum TCDD (ppt)
Organohalogen Compounds Vol 69 (2007)
P-303
2265
MICHIGAN DIOXIN EXPOSURE STUDY
Table 1. Characteristics of subjects from two background population groups.
UMDES (n=251):
NHANES (n=1228):
Two Michigan Counties
U.S. Population
Age (years) Sex: Female Body Mass Index (BMI) BMI loss in past year Smoking (packyears) Breastfeeding (% of women) # Incomplete Pregnancies
Mean (s.d.) or %
52.3 (15.2) 63%
28.8 (6.3) 1.0 (1.7) 13.0 (18.6)
53% 0.5 (1.5)
Mean (s.d.) or %
49.1 (18.7) 54%
28.2 (6.1) 0.7 (1.8) 23.4 (28.3)
43% 0.8 (1.2)
Table 2 below shows the final models for the two groups. Both models show effects of age, sex, BMI, BMI loss, and breastfeeding, but the UMDES model had more power to detect interactions and other effects. The R2
for the UMDES model (54%) increases somewhat after adding diet and occupational variables (63%).
Table 2. Regression results predicting serum 2,3,7,8-TCDD levels from two background populations.
UMDES: Two Michigan Counties
Coefficient
p-value
NHANES: U.S. Population
Coefficient*
p-value
Age
Sex (female)
BMI
BMI loss (past yr) Breastfed Incomplete Pregnancy Pack-years Age*sex
BMI*sex
0.0117
-0.2424
0.0068
-0.0238 0.0153 -0.1824 -0.0031 0.0065
--
0.0000
0.0701
0.0381
0.3440 0.0361 0.0000 0.0108 0.0075
--
0.013
0.270
0.020
0.033 -0.053
---0.011
-0.021
<.0001
0.4755
0.0148
0.0176 0.0123
---0.0016
0.0377
Age*BMI loss
0.0011
0.0209
--
--
Age*Breastfed
-0.0006
0.0020
--
--
Age*Incomplete Pregnancy
0.0032
0.0000
--
--
R2 (*100%) R2 after addition of diet and
occupational variables (*100%)
Adjusted R2 = 54% Adjusted R2 = 63%
Generalized R2 = 22% ---
*Coefficients divided by 2.303 (=ln(10)) to make comparable to coefficients from UMDES model.
Figure 2 illustrates two interactions found in the UMDES model above: (a) the interaction between gender and BMI loss, with females having a more exaggerated elevation of serum dioxin with BMI loss than men; (b) the increasing effect of breastfeeding by age.
Organohalogen Compounds Vol 69 (2007)
P-303
2266
MICHIGAN DIOXIN EXPOSURE STUDY
Serum TCDD (ppt) Serum TCDD (ppt)
15 Predicted-BMI_loss=0 -Female Predicted--BMI_loss=0 -Male Predicted-BMI_loss=6 -Female Predicted--BMI_loss=6 -Male
10
5
8 Predicted-Months BreastFeeding for First Child=0 Predicted-Months BreastFeeding for First Child=18
6
4
2
00 20 30 40 50 60 70 80 20 30 40 50 60 70 80
Figure 2. Illustrations of (a) BMI loss by gender interaction, and (b) breastfeeding by age interaction effects.
In conclusion, the strongest predictors of higher background 2,3,7,8-TCDD serum levels were older age and female gender. Adjusting for these factors, the Jackson/Calhoun population appeared to have slightly higher serum levels (by a few ppt) than the U.S. population; however, precise estimation was impossible with the majority of NHANES values below the LOD. In addition, serum levels significantly increased with recent weight loss, and decreased with breastfeeding and smoking. These effects are consistent with previously reported effects. The effects of occupation and food consumption were small but non-trivial.
A strength of both studies was the population-based sampling, allowing valid inference back to the source population. The low LOD in UMDES allowed more detailed modeling. Limitations include the problem of high LOD values in NHANES.
Acknowledgements: Financial support for this study comes from the Dow Chemical Company through an unrestricted grant to the University of Michigan. The authors acknowledge Ms. Sharyn Vantine for her continued assistance and Drs. Linda Birnbaum, Ron Hites, Paolo Boffetta and Marie Haring Sweeney for their guidance as members of our Scientific Advisory Board.
References 1. American Society for Testing and Materials (ASTM), Standard Practice for Collection of Floor Dust for
Chemical Analysis, Designation D 5438-00, Reprinted from the Annual Book of ASTM Standards, Philadelphia, PA. 2. Van den Berg M, Birnbaum L, Bosveld AT, Brunstrom B, Cook P, Feeley M, Giesy JP, Hanberg A, Hasegawa R, Kennedy SW, Kubiak T, Larsen JC, van Leeuwen FX, Liem AK, Nolt C, Peterson RE, Poellinger L, Safe S, Schrenk D, Tillitt D, Tysklind M, Younes M, Waern F, and Zacharewski T. Environmental Health Perspectives 1998; 106:775-792. 3. United States Environmental Protection Agency. Method 1668, Revision A: Chlorinated biphenyl congeners in water, soil, sediment, and tissue by HRGC/HRMS. Washington, DC: Office of Water, 19997. 4. United States Environmental Protection Agency. Method 8290: Polychlorinated dibenzodioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs) by high resolution gas chromatography/high resolution mass spectrometry (HRGC/HRMS). Washington, DC: Office of Solid Waste and Emergency Response, 1994. 5. SAS Institute. SAS/STAT User's Guide Version 9. Cary, NC: SAS Institute Inc., 2004.
Organohalogen Compounds Vol 69 (2007)
P-303
2267