United States
Environmental Protection
Agency
Toxic Substances
Office of
Toxic Substances
Washington, D C. 20460
EPA 560'5-90-001
October 1989
Washington, D.C. 20460
&EPA NHATS Broad Scan Analysis
Population Estimates from
Fiscal Year 1982 Specimens
•lie
-------
EPA 560/5-90-001
October 1989
FINAL REPORT
NHATS BROAD SCAN ANALYSIS:
POPULATION ESTIMATES FROM FISCAL TEAR 1982 SPECIMENS
Prepared by:
Battelle
Arlington Office
2101 Wilson Boulevard
Arlington, VA 22201
Contract No. 68-02-4294
for the:
Design and Development Branch
Exposure Evaluation Division
Office of Toxic Substances
Office of Pesticides and Toxic Substances
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20460
-------
This document has been reviewed and approved for publication by
the office of Toxie substances* offie* of Pesticides and Toxic
substances, u.3. Environmental Protection Agency. Th« ua« of
trkd« nam«i or oommareial products do«a not constitute Agency
endorsement or recommendation for uae.
-------
AUTHORS AND CONTRIBUTORS
The Broad Scan Study described in this report was a
cooperative undertaking that benefitted from the contribution* of
many SPA and contract support staff. EPA participation cam* from
th« Design and Development Branch (ODB) and the Field Studies
Branch (FSB) of the Exposure Evaluation Division (BED), Office of
Toxic Substancet (OTS). Contract support to OTS was provided by
Battalia and the Midwest Research Institute (MRI).
Battalia
Developed the statistical methodology for data analysis;
designed the specimen compositing plan; created and maintained
the computer files of Patient Summary Reports (PSRs); analysed
the chemical measurement and demographic data; prepared the final
Broad Scan Report.
Midwest Research Institute (MRI)
Prepared the composite samples of adipose tissue; developed
the methodology and carried out the chemical analysis of the
samples.
EPA Exposure Evaluation Division (BSD)
Participated in development of the study plan; managed and
coordinated the overall study; reviewed, edited, and finalized
the report. Kay staff included:
Joseph Breen John Schwemberger
Mary Frankenberry Cindy Stroup
Janet Reminera
Philip Robinson
iii
-------
TABLE OF CONTENTS
AUTHORS AND CONTRIBUTORS iii
EXECUTIVE SUMMARY X
1.0 INTRODUCTION 1
1.1 National Human Monitoring Program 1
1.2 National Human Adipose Tissue Survey 2
1.2.1 NHATS Objectives 2
1.2.2 NHATS Data Uses 3
1.3 Broad Scan Analysis Study 4
1.3.1 Study Objectives 6
1.3.2 Study Schedule 7
1.3.3 Report Overview 7
2 .0 RESULTS 9
2.1 Population Estimates of Average Concentration
Level 13
2.1.1 National Estimates 14
2.1.2 Geographical Estimates 14
2.1.3 Age Group Estimates 19
2.1.4 Comparison of Estimated Average
Concentration Levels Across Sex and
Race Groups 55
2.1.5 Relative Standard Errors 57
2.2 Incidence of Detection for Compounds Identified
in the Composite Samples 58
2.2.1 Volatile Organic Compounds 58
2.2.2 Semi-Volatile Organic Compounds 65
2.2.3 Dioxins and Furans 66
3. 0 QUALITY ASSURANCE 77
3.1 Volatile Organic Compounds 79
3.2 Semi-Volatile Organic Compounds 82
3.3 Dioxins and Furans 83
4.0 SAMPLING AND COMPOSITING DESIGNS 91
4.1 Sampling Design 91
4.2 Compositing Design 94
5.0 SPECIMEN COLLECTION AND STORAGE 99
6.0 CHEMICAL ANALYSIS PROCEDURES 101
7.0 DATA PREPARATION AND MANAGEMENT 107
8 . 0 STATISTICAL ANALYSIS APPROACH 109
8.1 Selection and Development of the Statistical
Model 109
-------
TABLE OF CONTENTS
(Continued)
8.2 Application of the Statistical Model 113
8.3 Statistical Estimation of Average Concentration
Levels for the Entire Nation and Various
Subpopulations 115
8.4 Significance Testing of Differences Between
Subpopulations 116
8.5 Detection and Exclusion of Outliers Among PECDD
Measurements 116
8.6 Concentration Estimates and Hypothesis Tests
for Total Equivalent DDT 117
8.7 Considerations in the Use of the Broad Scan
Analysis Study Statistical Analysis Approach... 122
9 . 0 REFERENCES 127
LIST OF APPENDICES
Appendix A. Statistical Estimates 129
Table A-l. Weighted Estimates and Their Associated
Standard Errors of the Average Concentration
Levels for the Entire Nation and for Each
Census Region/ Age Group, Race Group,
and Sex. 131
Appendix B. Percentage Detected Data 135
Table B-l. Volatile Organic Chemicals Identified in
the Broad Scan Analysis Study 137
Table B-2. Semi-Volatile Organic Chemicals Identified
in the Broad Scan Analysis Study 138
Table B-3. Dioxins and Furans Identified in the
Broad Scan Analysis Study 139
Appendix C. F782 NHATS Sampling Design SMSAs 141
Table C-l. SMSAs Selected for the F782 NHATS Sample.. 143
Appendix D. Broad Scan Analysis Study Compositing Design. 145
Table D-l. Demographic Characteristics for Each
Broad Scan Analysis Study Sample -
Volatile Analysis 147
Table D-2. Demographic Characteristics for Each
Broad Scan Analysis Study Sample -
Semi-Volatile Analysis 148
Appendix E. Glossary of Terms 149
Appendix F. Statistical Analysis Methodology 153
vi
-------
TABLE OF CONTENTS
(Continued)
LIST OF FIGURES
Figure 2-1. Weighted estimates of the average con-
centration levels of volatile (wet weight,
/*g/g) for the U.S population. (Standard
errors of the of the estimates are in
parentheses.) 21
Figure 2-2. Weighted estimates of the average concentration
levels of semi-volatiles (lipid adjusted, pg/g)
for the U.S. population. (Standard errors of
the estimates are in parentheses.) 23
Figure 2-3. Weighted estimates of the average concentration
levels of dioxins and furans (lipid adjusted,
pg/g) for the U.S. population. (Standard
errors of the estimates are in parentheses.)... 25
Figure 2-4. United States Census regions 27
Figure 2-5. Weighted estimates of the average concentration
levels of volatiles ( wet weight, pg/g) for
each census region. (Standard errors of the
estimates are in parentheses.) 29
Figure 2-6. Weighted estimates of the average concentration
levels of semi-volatiles (lipid adjusted,
for each census region. (Standard errors of
the estimates are in parentheses.) 33
Figure 2-7. Weighted estimates of the average concentration
levels of dioxins and furans (lipid adjusted,
P9/9) f°r each census region. (Standard errors
of the estimates are in parentheses.) 37
Figure 2-8. NHATS age groups 41
Figure 2-9. Weighted estimates of the average concentration
levels of volatiles (wet weight, pg/g) for each
age group. (Standard errors of the estimates
are in parentheses.) 43
Figure 2-10. Weighted estimates of the average concentration
levels of semi-volatiles (wet weight, j*g/g) for
each age group. (Standard errors of the
estimates are in parentheses.) 47
vii
-------
TABLE OF CONTENTS
(Continued)
Figure 2-11. Weighted estimates of the average concentration
levels of dioxins and furans (lipid adjusted,
pg/g) for each age group. (Standard errors of
the estimates are in parentheses.) 51
Figure 2-12. Percentage of FY82 composite samples in which
benzenes were detected 61
Figure 2-13. Percentage of FY82 composite samples in which
trihalomethanes and halocarbons were detected.. 63
Figure 2-14. Percentage of FY82 composite samples in which
PCB homolog groups were detected 67
Figure 2-15. Percentage of FY82 composite samples in which
organochlorine pesticides were detected 69
Figure 2-16. Percentage of FY82 composite samples in which
aromatics and chlorinated benzenes were
detected 71
Figure 2-17. Percentage of FY82 composite samples in which
phthalates and phosphates were detected 73
Figure 2-18. Percentage of FY82 composite samples in which
dioxins and furans were detected 75
Figure 4-1. Overview of the FY82 NHATS sampling design 93
Figure 4-2. NHATS FY82 collection map 96
Figure 6-1. Chemical anaysis steps for semi-volatiles,
dioxins and furans 104
LIST OF TABLES
Table 2-1. Target Compounds Identified in the Broad Scan
Analysis Study 10
Table 2-2. Compounds for Which Statistical Analyses Were
Performed 12
Table 2-3. Weighted Estimates and Their Associated
Relative Standard Errors of the Average
Concentration Levels for the Entire Nation
and for Each Census Region, Age Group,
Race Group, and Sex 15
viii
-------
TABLE OF CONTENTS
(Continued)
Table 2-4. Summary of Significance Testing for Differences
Between Subpopulations 18
Table 3-1. Summary of QA Results for Selected Volatile
Organic Analytes in Spiked 20 Gram Aliquots of
Human Adipose Tissue 80
Table 3-2. Ranges of Estimated Levels of Detection of
Volatile Organic Compounds for Composite
Samples Whose Reported Concentration Levels
Were Declared Not Detected or Trace 81
Table 3-3. Summary of QA Results for Selected Volatile
Organic Analytes in Spiked 20 Gram Aliquots of
Human Adipose Tissue 85
Table 3-4. Ranges of Estimated Levels of Detection of Semi-
Volatile Organic Compounds for Composite
Samples Whose Reported Concentration Levels
Were Declared Not Detected or Trace 88
Table 3-5. Ranges of Estimated Levels of Detection of
Dioxins and Furans for Composite Samples Whose
Reported Concentration Levels Were Declared Not
Detected or Trace. 90
Table 4-1. Geographic and Demographic Counts for
Specimens 97
Table 6-1. Pairing of Target Analytes Versus Internal
Quantitation Standards for Volatile Organic
Compounds Analysis 102
Table 8-1. Comparison of Average Concentration Estimates
and Significance Test Results for 1,2,3,7,
8-PECD Including, and Excluding Outliers 118
Table 8-2. Comparison of Average Concentration Estimates
and Significance Test Results for Alternative
Ways of Computing Total Equivalent DDT
(TEDDT) 121
ix
-------
EXECUTIVE SUMMARY
BACKGROUND
The National Human Monitoring Program (NHMP), operated
by the United States Environmental Protection Agency's Office of
Toxic Substances (USEPA/OTS) under the 1976 Toxic Substances
Control Act (TSCA), is an ongoing national chemical monitoring
program. The main operative program of the NHMP is the National
Human Adipose Tissue Survey (NHATS). The NHATS is an annual
survey to collect and analyze a nation-wide sample of adipose
tissue specimens from autopsied cadavers and surgical patients.
The purpose of the NHATS is to identify and quantify the
prevalence and levels of selected compounds in human adipose
tissue. The analysis results are used to establish an exposure-
based chemicals list and to estimate baseline levels and trends
of the selected chemicals.
In the past/ NHATS data have been used to monitor levels
of organochlorine pesticides and polychlorinated biphenyls (PCBs)
in the U.S. NHATS data have shown that the estimated percentage
of individuals with levels of PCBs greater than three parts per
million decreased from 1977 to 1983. This decrease occurred
after the passage of legislation in 1976 which limited the
production of PCBs (USEPA 1985). NHATS studies on
hexachlorobenzene and mirex have helped to identify regions of
the country where relatively high levels of these pesticides were
found in human tissue.
METHODS
Although the NHATS data have proved useful in the past,
the chemicals that could be monitored were limited .to selected
semi-volatile organic compounds. To broaden the range of
chemicals, EPA proposed to analyze specimens through high
resolution gas chromatography/mass spectrometry (HRGC/MS).
-------
The HRGG/MS method, however, required considerably more
tissue mass than the previous method of analysis. In addition,
the HRGC/MS protocol was significantly more expensive than the
previous protocol, thereby limiting the number of analyses that
could be performed. For these reasons, the individual adipose
tissue specimens were physically mixed to form composite samples.
The composite sample, rather than the individual specimen, was
analyzed. The use of composite samples created a need to develop
a new statistical analysis approach.
NHATS specimens collected in Fiscal Tear 1982 (FT82)
were selected for the Broad Scan Analysis Study, the first
application of HRGC/MS to NHATS. For this analysis, 763
individual specimens were combined into two sets of composite
samples: 46 composite samples used for analysis of volatile
organic compounds, and 46 composite samples used for the analysis
of semi-volatiles, dioxins, and furans. In total, 57 compounds,
including some homolog groups and isomers, were target analytes
for the composite samples analyzed for the study. Of these, 17
were volatile organic compounds; 30 were semi-volatile organic
compounds; 5 were dioxins (polychlorinated dibenzo-para-dioxins,
or PCDDs); and 5 were furans (polychlorinated dibenzofurans, or
PCDFs).
RESULTS
Compounds Detected
Volatile Organic Compounds
Results of the analysis indicated that eight of the nine
benzene related volatile organic compounds were detected in more
than 90% of the composite samples. • For instance, benzene was
found in 96% of the composite samples and 1,4-dichlorobenzene was
found in all the composite samples.
XI
-------
Semi-Volatile Organic Compounds
The incidence of detection of the composite samples
varied considerably for the organochlorine pesticides; mirex
was detected in 14% of the composite samples while B-BHC and
£,p/-DDE were detected in 93% and 100%, respectively. PCBs were
detected in 86% of the composite samples.
Dioxins and Purans
Four out of five dioxins were detected in more than 90%
of the composite samples. The one exception, 2,3,7,8-TCDD, was
found in 74% of the composite samples. The incidence of
detection for the furans ranged from 26% for 2,3,7,8-TCDF to 93%
for 1,2,3,4,6,7-HPCDF.
Average Concentrations
To form the required tissue composites it was sometimes
necessary, because of the limited number of individual samples
available, to mix male and female, and Caucasian and non-
Caucasian specimens, in the same composite. The need to estimate
average concentration levels using measurement data on these
mixed composites required a model-based approach to the analysis.
A multiplicative statistical model, which relates average
concentration levels of the composite samples to demographic
characteristics of constituent specimens, was developed for this
purpose.
The FY82 NHATS survey initiates new data series for the
dioxins and furans, as well as members of the following semi-
volatile classes: PCB homologs, aromatics, chlorinated benzenes,
phthalates, and phosphates. Volatile organic compounds were also
measured in the FY82 survey, but there are no plans to measure
this class of chemicals in subsequent years. Comparisons to past
years' results for organochlorine pesticides are limited in this
report because of the change in chemical methods.
xii
-------
Estimates of average concentration levels in the nation
and various geographic and demographic subpopulations (i,e.,
Census region, age group, sex, and race group) were derived for
22 of the 57 target compounds, those for which more than half of
the composites had measured concentration values above the
analytical limit of quantification. This restriction was adopted
to avoid possible bias in estimating average concentrations for
compounds where most of the measurements were imputed (a compound
not detected in a particular composite was assumed to be present
at a level of one half the limit of detection). The average
concentration estimates serve as baseline levels against which
data from other sources can be compared.
The national average concentrations for selected
compounds of current interest to EPA were:
Benzene, 0.014 /*g/g (wet weight);
1,4-dichlorobenzene, 0.12 /ig/g (wet weight);
PCBs, 0.33 /ig/g (lipid adjusted weight); and
2,3,7,8-TCDD, 6.1 pg/g (lipid adjusted weight).
Regional Differences
There were statistically significant differences between
regional concentrations for five compounds: benzene,
chlorobenzene, 1-4 dichlorobenzene, 3-BHC, and tetrachloroethene.
The West and Northeast Census regions had the highest average
levels for benzene, while the South had the highest levels for
chlorobenzene and 1,4-dichlorobenzene, two volatile organics, and
for 8-BHC, an organochlorine pesticide. Average levels for
tetrachloroethene, a volatile organic, were higher in the
Northeast and North Central Census regions than in the South and
West.
xiii
-------
Age Differences
Eight compounds had statistically significant
differences between age groups. Average levels for ethylphenol,
a volatile organic compound, significantly decreased with age
group. Among the semi-volatiles, average concentration levels of
total PCBs, pp'-DDE, and B-BHC, an organochlorine pesticide,
significantly increased with age group. Total equivalent DDT
also significantly increased with age group. For the dioxins and
furans, levels were highest in the "15-44 years" age group for
2,3,7,8-TCDD, 1,2,3,7-PECDD and 2,3,4,7,8-PECDF. Levels for OCDD
were higher in the "15-44 years" and in the oldest age group than
in the youngest age group.
Sex Differences
Eight compounds were statistically significant with
respect to sex differences. Males had significantly higher
average levels than females for five volatile organics:
chloroform, styrene, tetrachloroethene, toluene, and xylene; and
for one semi-volatile compound, p_,p_'-DDE. The result for p,p'-
DDE appears anomalous and is primarily attributable to a very low
concentration in one pure female composite. Females had
significantly higher levels of the dioxin, HXCDD, and the furan,
HXCDF.
Racial Differences
Five compounds were statistically significant with
respect to race differences. Caucasians had significantly higher
average levels than non-Caucasians for toluene, <:hlorobenzene, 8-
BHC, butyl benzyl phthalate, and 2,3,4,7,8-PECDF. There were no
compounds for which non-Caucasians had significantly higher
levels than Caucasians. Because the non-Caucasian sample size
was too small to create composites that adequately represented
this race group, the estimated race group effects should be
interpreted cautiously.
xiv
-------
Qualifications
In interpreting the statistical results of the survey,
the reader should be aware of the following characteristics of
its design. First, it is assumed that the average concentration
levels of chemicals in the adipose tissue of surgical patients
and autopsied cadavers is approximately equal to the average
concentration levels in the U.S. population. Second, the survey
is voluntary, and depends on the active participation of hospital
pathologists and medical examiners (collectively known as
"cooperators") who collect the adipose tissue samples that are to
be analyzed. The cooperators are given quotas of specimens to
fill, defined in terms of the age, race, and sex of donors;
little additional information on donors is collected. In Fiscal
Tear 1982 approximately 50% of the planned number of samples were
actually submitted for analysis by the survey cooperators.
Third, the hospitals in the NHATS sample are all located in
Standard Metropolitan Statistical Areas (SMSAs), and it is
therefore plausible to expect that the distribution of survey
specimens collected at these hospitals will be skewed toward
individuals living in urban rather than rural areas. The impact
on the estimated average concentrations, if any, attributable to
these factors is not known.
xv
-------
1.0 INTRODUCTION
1.1 National Human Monitoring Program
The Toxic Substances Control Act (TSCA), enacted by
Congress in 1976 as Public Law 94-469, directs the United States
Environmental Protection Agency (USEPA) to prevent unreasonable
chemical risk to the human population and the environment. To
prevent or reduce such risk, it is necessary for the EPA to
identify and evaluate those chemicals which contribute to
unreasonable levels of risk to the human population or the
environment.
EPA evaluates risk using both toxicity and exposure
data. EPA determines whether a chemical is toxic enough to be
harmful to human health or the environment through toxicological
studies, quantitative assessments, and phanoacokinetic modeling.
In addition, EPA determines if there is sufficient opportunity
for humans or the environment to be exposed. Monitoring of both
the environment and the population is one approach used by the
EPA to estimate exposure. TSCA Section 10 (Research,
Development, Collection, Dissemination, and Utilization of Data)
allows the EPA to develop monitoring data, techniques, and
instruments to detect toxic chemicals and to assess the degree of
chemical risk they represent.
In response to TSCA, the EPA's Office of Toxic
Substances (OTS) operates the National Human Monitoring Program
(NHMP). The NHMP was first established by the U.S. Public Health
Service in 1967. It was transferred to the EPA in 1970 and
operated by the Office of Pesticide Programs (OPP) until 1979,
when the program was assigned to the Exposure Evaluation Division
(EED) of the newly created OTS.
The NHMP is an ongoing chemical monitoring program in
which human media are sampled and analyzed to determine the
extent of human exposure to toxic substances in the environment.
By measuring the concentrations of toxic chemicals in human
tissue and fluids, evidence of actual exposure is obtained.
-------
Monitoring these levels over time provides the EPA with a means
to assess and subsequently to address, through TSCA Section 4
(Testing of Chemical Substances and Mixtures) and TSCA Section 6
(Regulation of Hazardous Chemical Substances and Mixtures) those
chemicals that are most likely to be associated with significant
health concerns. Historically, the EPA has prioritized chemicals
on the basis of significant toxicological findings and surrogate
measures of exposure, such as production volume. The NHMP offers
the EPA a means to prioritize chemicals using direct measures of
exposure.
1.2 National Human Adipose Tissue Survey
The National Human Adipose Tissue Survey (NHATS) is the
main operative program of the NHMP. The NHATS is an annual
survey, conducted since 1970, which collects and chemically
analyzes adipose tissue specimens for the presence of selected
compounds. The tissue specimens are collected by pathologists
and medical examiners, whose participation in NHATS is voluntary,
from a national sample of autopsied cadavers and surgical
patients in Standard Metropolitan Statistical Areas (SMSAs) in
the continental United States. Past NHATS monitoring efforts
have focused on the monitoring of organochlorine pesticides and
polychlorinated biphenyls (PCBs). The analysis results have been
used to provide information on U.S. population exposure to the
pesticides and PCBs.
1.2.1 NHATS Objectives
The primary purpose of the NHATS program is to collect
data for the detection and quantification of selected toxic
residues in the adipose tissue of the general population of the
United States. The specific objectives are to:
Identify the presence of toxic chemicals in human
adipose tissue;
-------
Establish baseline levels of the selected chemicals
in the U.S. population and various demographic
subpopulations;
Measure time trends of these levels; and
Make statistical comparisons of these results across
the various geographic regions and demographic
groups.
1.2.2 MEATS Data Uses
The chemicals identified through the NHATS provide
information on human exposure. Population estimates establish
baseline levels and trends of these chemicals in adipose tissue.
Baseline levels serve as values against which other exposure
levels can be compared. NHATS data can be used to assist in
prioritizing the EPA's chemical screening and testing activities.
Trend estimates of changes in prevalence and levels are
used to help identify the need for regulatory action or, in the
case of existing regulations, to assess the efficacy of such
regulations. Observed decreases in human monitoring data provide
evidence that chemical risk has been reduced. For instance, in
1976, legislation limiting the production and usage of PCBs was
passed. Through NHATS monitoring of PCB levels, it was observed
that the estimated percentage of individuals having total PCB
levels greater than 3 parts per million (ppm) decreased during
the period from 1977 to 1983 (USEPA 1985). This result
demonstrated the efficacy of the 1976 legislation. On the other
hand, increasing trends may help to uncover emerging problems-.
Demographic and geographic data are used to estimate
baseline levels and trends for various subpopulations of interest
to EPA. This information identifies exposed segments of the
population for further investigations of chemical risk and to
supports resulting regulatory actions. Several past NHATS
studies have resulted in the identification of such high risk
populations. A geographical evaluation of NHATS data on
hexachlorobenzene (HCB) levels in the U.S. found a high incidence
-------
of levels greater than 0.09 parts per million, the ninetieth
percentile of the data observed throughout the nation, in the
western region of the country (Leczynski and Stockrahm 1985).
Although the direct use of HCB as a pesticide decreased sharply
through the 1970's, further investigation discovered that
pesticides containing HCB were still used in several Pacific
Northwest areas (USEPA 1986f). A follow-up study to investigate
evidence of mirex exposure observed in the NHATS verified the
increased prevalence of mirex in a section of the southern U.S.
(USEPA 1980). Thus, the estimation of levels and trends through
periodic monitoring provides an effective means to maintain
surveillance of both the general population and selected
subpopulations with respect to chemical exposure (Mack and
Stanley 1984).
1.3 Broad Scan Analysis Study
Upon assuming the responsibility of operating the NHMP
in 1979, the OTS decided to expand the usefulness of the program
by broadening the range of chemicals monitored by the NHATS. OTS
proposed a Broad Scan Analysis Study of additional semi-volatile
organic compounds, including the dioxins and furans, as well as
volatile organic compounds and trace elements (Mack and Stanley
1984).
Previous NHATS chemical analyses were carried out by
packed column gas chromatography/electron capture detector
(PGC/ECD) methods. These methods permitted analysis of
individual tissue specimens. However, since the PGC/ECD protocol
was limited to the analysis of selected organochlorine pesticides
and PCBs and was not readily expandable to additional chemicals
(USEPA 1986c), several changes in the approach to analyzing
adipose tissue specimens were required.
First, analytical methods based on high resolution gas
chromatography/mass spectrometry (HRGC/MS) techniques were needed
for the detection of semi-volatile and volatile organic compounds
-------
(USEPA 1986a). The HRGC/MS method provided a greater degree of
certainty in compound determination than PGC/ECD since
identification is based on matching both retention time and mass
spectra.
The HRGC/MS method, however, required more tissue mass
per analysis sample than was collected from each individual NHATS
donor. Furthermore, additional sample preparation work and
sophisticated analytical equipment were needed to perform the
chemical analyses. These factors greatly increased the cost of
analyzing each sample and thereby reduced the number of samples
that could be analyzed. For these reasons, individual tissue
specimens had to be composited prior to chemical analysis.
Compositing is a process in which a specific amount of
tissue is taken from each of several individual specimens and
physically mixed to form a single sample. The composite sample,
rather than the individual specimens, is then chemically
analyzed. A compositing design was needed to ensure that each
composite sample would have sufficient tissue mass available for
analysis and that estimates of average concentration levels for
aubpopulations and the general population could be obtained. The
design specified which types of specimens, in terms of their
geographic and demographic makeup, to include in each composite
sample. The composite design led to a major change in the
statistical analysis of the NHATS chemical analysis data. A
statistical model was developed to make inferences concerning
average concentration levels for subpopulations and the general
population.
To do this, a relationship was assumed between the
concentration of a composite and its geographic and demographic
make-up. That is, the concentration of a composite was assumed
to have a geographic component, age group, sex, and race
components and random error components. The statistical model
made it possible to estimate the components from the observed
concentrations of the composites. Once the components were
-------
estimated, estimates of the average concentration levels of
geographic and demographic subpopulations and the national
population could be made.
Reliable estimates of prevalence, the proportion of the
population with concentration levels above a specific threshold,
cannot be computed from the composited data using the
multiplicative model adopted for the NHATS F782 survey. Work is
currently underway on the development of a new modeling approach
which will allow such prevalence estimates to be made in future
surveys.
To further expand the range of chemicals monitored by
the NHATS, multi-elemental techniques were needed for the
detection of trace elements. The two procedures that were
identified, however, were only used for the analysis of nine
selected individual specimens (USEPA 1986e). Thus, compositing
was unnecessary. Average concentration levels of trace elements
for the U.S. population were not estimated.
1.3.1 Study Objectives
The specific objectives of the Broad Scan Analysis Study
were to:
• Identify the presence of a wider range of chemicals
in the adipose tissue of the U.S. population than
had been identified in the past;
• Estimate the FY82 average levels of the chemicals
for the entire U.S. and for selected geographic and
demographic subpopulations; and
• Make comparisons of the estimated average levels
across these various demographic and geographic
subpopulations.
To accomplish these objectives, several activities were
required. They were to:
• Develop, refine, and conduct a preliminary
evaluation of appropriate analytical protocols based
on HRGC/MS and the two proposed multi-elemental
techniques;
-------
• Derive a statistically based compositing design for
the F782 specimens that would provide a high degree
of sensitivity for detecting chemicals (Mack and
Stanley 1984), and permit appropriate estimates to
be made for populations of interest.
• Develop, implement, and initially assess an
appropriate statistical analysis methodology.
1.3.2 Study Schedule
Specimens collected during Fiscal Tear 1982 (F782) were
selected from the NHATS repository for use in the Broad Scan
Analysis Study. These specimens were collected from October,
1981 through September, 1982. Two sets of composite samples were
prepared for chemical analysis, one set for the semi-volatile
analyses, including the dioxins and the furans analyses, and one
set for the volatiles analyses. Both sets of composite samples
were prepared for chemical analysis during February and March of
1984. The semi-volatiles analyses were performed between April
and June, 1984; the volatiles analyses were performed during June
and July, 1984; the dioxins and furans analyses were performed
from October, 1984 through March, 1985. The chemical analysis
results for the semi-volatile and volatile compounds were
completed in November, 1985. The dioxin and furan results were
completed in March, 1986. The statistical analysis procedures
were performed between December, 1985 and May, 1989.
1.3.3 Report Overview
This report summarizes the analysis approach adopted for
the Broad Scan Analysis Study. It describes the statistical
methodology and provides population estimates of the average
concentration levels obtained for the volatile and semi-volatile
organic compounds and the dioxins (polychlorinated dibenzo-para-
dioxins, or PCDDs) and furans (polychlorinated dibenzofurans, or
PCDFs). Additional information on the chemical analysis
procedures used in the analysis of these compounds, the results
-------
for trace elements, and related quality assurance efforts is
found in the five volume series/ "Broad Scan Analysis of Human
Adipose Tissue* (USEPA 1986a-e).
8
-------
2.0 RESULTS
Fifty-seven compounds, including some homolog groups and
isomers, were the target compounds for the composite samples
analyzed for the Broad Scan Analysis Study. Seventeen of these
were volatiles; thirty were semi-volatiles; five were dioxins;
and five were furans. The volatile organics are members of six
chemical classes: benzene, substituted benzenes, alkyl benzenes,
chlorinated benzenes, trihalomethanes, and halocarbons. The
semi-volatile organics are also members of six chemical classes:
PCB homologs, organochlorine pesticides, aromatics, chlorinated
benzenes, phthalates, and phosphates. The results presented in
this report are grouped into these chemical classes. The 57
compounds and their CAS numbers are listed in Table 2-1. The
compound 1,2-dichlorobenzene was included in both the volatile
and semi-volatile compound lists because it could be measured by
both the volatile and semi-volatile protocols.
Estimates of population average concentration levels
derived from the statistical model for analysis of composite
sample data were obtained for those compounds for which at least
half of the reported concentrations were above the limit of
quantificaton. Twenty-two of the 57 compounds met this
condition. These compounds are listed in Table 2-2.
For the volatile organic compounds, the statistical
analyses were based on wet weight concentration levels. For the
semi-volatile compounds, including the dioxins and furans, the
analyses were based on lipid adjusted concentrations. The
concentrations for the volatiles were not lipid adjusted. To do
so would have required further handling of samples, which
increases the potential for volatile compounds to escape.
Concentrations were reported in parts per million (ftg/g) for
volatiles and semi-volatiles and in parts per trillion (pg/g) for
the dioxins and furans.
-------
Table 2-1.
Target: Compounds Identified in the Broad Scan
Analysis Study
Class
Compound:
CAS Number
VOLATILE ORGANIC COMPOUNDS
Benzene
Substituted Benzenes
Alkyl Benzenes
Chlorinated Benzenes
Trihalomethanes
Halocarbons
Benzene
Styrene
Ethylphenol
Toluene
Ethylbenzene
Xylene
Chlorobenzene
1,2-Dichlorobenzene
1,4-Dichlorobenzene
Chloroform
Bromodichloromethane
Dibromochloromethane
Bromoform
1,1,1-Trichloroethane
1,1,2-Trichloroethane
1,1,2,2-Tetrachloroethane
Tetrachloroethene
SEMI-VOLATILE ORGANIC COMPOUNDS
PCBs
Organochlorine
Pesticides
PCBs
Trichlorobiphenyl
Tetrachlorobiphenyl
Pentachlorobiphenyl
Hexachlorobiphenyl
Heptachlorobiphenyl
Octachlorobiphenyl
Nonachlorobiphenyl
Decachlorobiphenyl
8-BHC
p_,p/-DDE
E/E'-DDT
Mirex
trans-Nonachlor
Heptachlor Epoxide
Dieldrin
71-43-2
100-42-5
25429-37-2
188-88-3
100-41-4
1330-20-7
108-90-7
95-50-1
106-46-7
67-66-3
75-27-4
124-48-1
75-25-2
71-55-6
79-00-5
79-34-5
127-18-4
1336-36-3
25323-68-6
26914-33-0
25429-29-2
26601-64-9
28655-71-2
31472-83-0
53742-07-7
2051-24-3
319-85-7
72-55-9
50-29-3
2385-85-5
39765-80-5
1024-57-3
60-57-1
10
-------
Table 2-1.(Continued) Target Compounds Identified in the Broad
Scan Analysis Study
Class
Compound
CAS Number
Aromatics
Chlorinated Benzenes
Phthalates
Phosphates
Naphthalene
Phenanthrene
Pyrene
1,2-Dichlorobenzene
1,2,4-Trichlorobenzene
Pentachlorobenzene
Hexachlorobenzene
Diethyl Phthalate
Di-n-butyl Phthalate
Di-n-octyl Phthalate1
Butyl Benzyl Phthalate
Triphenyl Phosphate
Tributyl Phosphate
Tris (2-Chloroethyl)
Phosphate
91-20-3
85-01-8
129-00-0
95-50-1
120-82-1
608-93-5
118-74-1
84-66-2
84-74-2
117-84-0
85-68-7
115-86-6
126-73-8
115-96-8
DIOXINS AND FURANS
Dioxins
Furans
2,3,7,8-TCDD
1,2,3,7,8-PECDD
HXCDD
1,2,3,4,7,8,9-HPCDD
OCDD
2,3,7,8-TCDF
2,3,4,7,8-PECDF
HXCDF
1,2,3,4,6,7,8-HPCDF
OCDF
1746-01-6
40321-76-4
34465-46-8
35822-46-9
3268-87-9
51207-31-9
57117-31-4
55684-94-1
67562-39-4
39001-02-0
The chemical actually identified was diethyl hexyl phthalate,
an isomeric compound to di-n-octyl phthalate, that exhibits
many of the same chemical and physical properties.
11
-------
Table 2-2.
Compounds for Which Statistical Model Analyses
Were Performed
Class
Compound
VOLATILE ORGANIC COMPOUNDS
Benzene
Substituted Benzenes
Alkyl Benzenes
Chlorinated Benzenes
Trihalomethanes
Halocarbons
Benzene
Styrene
Ethylphenol
Toluene
Ethylbenzene
Xylene
Chlorobenzene
1,4-Dichlorobenzene
Chloroform
Tetrachloroethene
SEMI-VOLATILE ORGANIC COMPOUNDS
Organochlorine Pesticides
PCBs
Phthalates
B-BHC
2/2'-DDE
PCBs ^
Butyl Benzyl Phthalate
DIOXINS AND FURANS
Dioxins
Furans
2,3,7,8-TCDD
1,2,3,7,8-PECDD
HXCDD
1,2,3,4,7,8,9-HPCDD
OCDD
2,3,4,7,8-PECDF
HXCDF
1,2,3,4,6,7,8-HPCDF
12
-------
Average concentration levels were estimated for the
entire nation and for various geographic (Census region) and
demographic (age group, sex, and race group) subpopulations.
Data on the percentage of composite samples having detectable
levels of the 57 compounds were also obtained.
2.1 Population Estimates of Average Concentration Level
For the Broad Scan Analysis Study, 763 individual
specimens, collected from the nine U.S. Census divisions, three
age groups, two sexes and two race groups, were composited into
two sets of 46 composite samples each prior to chemical analysis.
One set was prepared for the analysis of the volatiles, and the
other for the analysis of the semi-volatiles and the dioxins and
furans. The compositing procedures were performed following a
design which ensured that estimates of the average concentration
levels for populations of interest could be obtained from an
eight parameter statistical model. The model has eight
parameters to be estimated from 46 samples. Estimates for 48
target subpopulations, corresponding to the 4 Census regions, 3
age groups, 2 race groups and 2 sexes (4x3x2x2-48), were
derived from the model. These estimates were then weighted to
obtain estimated average concentrations for the selected
subpopulations, as well as for the entire nation. The weights
corresponded to the 1980 U.S. Census population counts for the 48
target subpopulations.
The estimated average concentration levels as well as
their relative standard errors are provided in Table 2-3. This
table is reproduced in Appendix A, Table A-l with standard errors
rather than relative standard errors. A summary of the results
of significance testing for differences between populations is
presented in Table 2-4.
13
-------
2.1.1 National Estimates
Figures 2-1, 2-2, and 2-3 graphically show the national
estimates of the average concentration levels for 22 compounds
classified as volatiles, semi-volatiles, and dioxins
and furans, respectively. For selected compounds of current
interest to EPA, the estimated average concentration levels were
0.014 /*g/g for benzene, 0.12 pg/g for 1,4-dichlorobenzene, 0.33
ftg/g PCBs, and 6.1 pg/g for 2,3,7,8-TCDD.
2.1.2 Geographical Estimates
The graphs in Figures 2-5, 2-6 and 2-7 depict the
estimated average concentration levels of the volatiles, semi-
volatiles, and dioxins and furans, for each of the four Census
regions shown in the map of the continental United States in
Figure 2-4. There were statistically significant regional
differences at the p-values shown in Table 2-4 for five of the 22
compounds: benzene, chlorobenzene, 1,4-dichlorobenzene,
tetrachloroethene, and B-BHC:
• Benzene concentrations ranged from a low of
0.010 pg/g in the North Central and South
regions to a high of 0.019 pg/g in the West;
• Chlorobenzene concentrations ranged from 0.0025 ftg/g
in the North Central region to 0.0072 pg/g in the
South;
• 1,4-dichlorobenzene concentrations ranged from 0.052
pg/g in the West to 0.20 pg/g in the South;
• Tetrachloroethene levels were higher in the North
Central region (0.044 pg/g) and in the Northeast
(0.041 pg/g) than they were in the South (0.016
or the West (0.0086 pg/g)} and
8-BHC concentrations ranged from 0.097 pg/g in the
West region to 0.31 pg/g in the South.
14
-------
Table t-S. Weighted Estimates (and Associated Relative Standard Errors)1
of tho Average Concentration Levels for the Entire Nation and
for Each Census Region, Age Group, Race Group, and Set
Coapound
'Population Percentages
VOLATILE3
Benzene
Benzene
Substituted Benzenes
Styrene
Ethylphenol
Alkyl Benzenes
Toluene
Ethyl benzene
Xylene
Chlorinated Benzenes
Chlorobenzene
1,4-Dichlorobenzene
Relative standard error
2HE • North East S
NC • North Central 1
Entire
Nation
0.014
(12)
0.096
(20)
0.086
(26)
0.046
(37)
0.077
(41)
0.30
(42)
0.0044
0.12
(17)
expressed
• South
• lest
Census Region
ME
22
0.018
(21)
0.096
(38)
0.13
(47)
0.023
(62)
0.072
(72)
0.20
(71)
0.0033
(30)
0.075
(32)
as s percentage
NC
26
0.010
(19)
0.069
(37)
0.029
(46)
0.062
(48)
0.076
(70)
0.26
(71)
0.0026
(24)
0.11
(29)
S
33
0.010
(16)
0.10
(31)
0.090
(37)
0.046
(47)
0.10
(69)
0.49
(69)
0.0072
(24)
0.20
(23)
I
19
0.019
(28)
0.13
(40)
0.10
(61)
0.061
(69)
0.039
(75)
0.12
(72)
0.0030
(33)
0.062
(39)
1-14 yrs
23
0.016
(19)
0.12
(24)
0.17
(33)
0.036
(48)
0.063
(47)
0.27
(46)
0.0038
(24)
0.12
(28)
Age Qroups
16-44 yrs
46
0.014
(17)
0.10
(23)
0.065
(31)
0.066
(44)
0.090
(44)
0.33
(43)
0.0061
(20)
0.13
(24)
Race Oroups
46* yrs
31
0.012
0.076
(22)
0.060
(31)
0.038
(43)
0.066
(44)
0.26
(43)
0.0037
(22)
0.11
(24)
•hits
83
0.016
(12)
0.095
(21)
0.079
(27)
0.063
(38)
0.078
(42)
0.31
(42)
0.0048
(17)
0.11
(19)
Non-lhite
17
0.0096
(29)
0.10
(33)
0.12
(47)
0.013
(63)
0.070
(58)
0.23
(61)
0.0018
(28)
0.19
(39)
Sex
Uale
49
0.017
(20)
0.14
(24)
0.096
(34)
0.080
(46)
0.11
(47)
0.43
(44)
0.0067
(23)
0.13
(29)
Feeala
61
0.010
(21)
0.060
(26)
0.076
(37)
0.014
(47)
0.048
(48)
0.17
(46)
0.0032
(26)
0.11
(31)
of the estinte.
'Volatile average concentrations are expressed in vet Might in parts per ail I ion
-------
TibU 2-3. (continued)
Coapound
Population Percentages
VOLATILE OMMIICS*
TrilnloMthaiMs
Chlorofora
Malocarbona
Tetrachloroethene
SEW-VOLATILE ORMNICS4
PCBa
PCBa
Organociilorino Pesticides
Beta-BHC
E-E'-K*
Total DOT
Phthalatea
Butyl benzyl
phthalate
Entire
Nation
0.047
(42)
0.027
(29)
0.33
(23)
0.19
(16)
1.3
(23)
1.6
(26)
0.39
(60)
Census Region
HE
22
0.021
(67)
0.041
(46)
0.31
(42)
0.19
(26)
1.1
(36)
1.4
(37)
0.11
(73)
NC
26
0.041
(«4)
0.044
(42)
0.23
(38)
0.11
(21)
0.73
(32)
0.87
(33)
0.46
(68)
S
33
0.049
0.016
(37)
1.61
(32)
0.81
(22)
1.9
(31)
2.4
(33)
0.62
(66)
1
19
0.081
(72)
0.0086
(62)
0.20
(63)
0.097
(33)
1.3
(46)
1.7
(47)
0.21
(86)
0-14 yre
23
0.063
(61)
0.017
(37)
0.071
(34)
0.071
(23)
0.76
(32)
0.98
(36)
0.46
07)
Age Groups
16-44 yra
46
0.063
(49)
0.030
(36)
0.30
(31)
0.17
(20)
l.S
(30)
1.7
(32)
0.31
(68)
Race Groups
46* yra
31
0.033
(47)
0.031
(84)
0.67
(30)
0.31
(21)
i.a
(29)
2.1
(30)
0.46
(61)
•hi to
83
0.046
(44)
0.029
(30)
0.32
(26)
0.21
(17)
1.4
(26)
1.7
(27)
0.46
(62)
Non-lhite
17
0.062
(69)
0.019
(62)
0.41
(47)
0.088
(30)
0.73
(44)
1.1
(62)
0.096
(84)
Sex
Vale
49
0.081
(49)
0.044
(37)
0.36
(38)
0.19
(24)
2.0
(32)
2.4
(33)
0.64
(67)
Feaale
61
0.014
(63)
0.011
(40)
0.32
(40)
0.19
(29)
0.64
(40)
0.93
(46)
0.24
(69)
9 Volatile average concentrations are expressed in wet Might in parts per Million (pg/g).
4 Seat-volatile average concentrations are expressed in lipid adjusted might in parts per Billion (pg/g).
-------
Table 2-3. (continued)
Coapound
Population Percentages
•pyjffljjP
2,8,7,8-700
1,2.1,7,6-PBCDD
HXCOD
1,2,8,4,7,8,9-
OCDO
FuwiB1
2,3,4,7,6-PKDF
HXCDF
1,2,3,4,6,7,6-
Bit ire
Nation
6.1
(13)
76
(23)
120
(20)
140
(19)
820
(13)
40
(16)
24
21
(13)
Census Region
NE
22
6.6
(26)
120
(39)
160
(32)
160
(32)
760
(24)
49
(30)
20
(24)
18
(24)
HC
26
7.1
(22)
62
(34)
110
(29)
180
(28)
920
(21)
38
(26)
29
(22)
26
(21)
S
33
6.1
(19)
60
(30)
100
(26)
110
(24)
780
(18)
30
(24)
24
(19)
22
(17)
19
4.1
(28)
73
(42)
120
(39)
100
(39)
860
(30)
62
(34)
23
(30)
16
(29)
0-14 yra
23
4.1
(20)
64
(30)
92
(29)
89
(29)
410
(21)
36
(26)
18
(22)
19
(21)
Age Groups
16-44 yra
46
7.8
130
(27)
120
(28)
160
(26)
920
(18)
63
(22)
27
(19)
22
(IB)
46* yra
31
6.0
(19)
11
(31)
130
(26)
160
(26)
990
(18)
26
(24)
26
(20)
20
(18)
Race
White
83
6.4
(14)
83
(24)
120
(21)
140
(21)
810
44
(18)
26
(16)
20
(14)
Groups
Non-Mite
17
4.3
(30)
39
(46)
110
(42)
140
(42)
880
(31)
22
(39)
18
(31)
26
(30)
Sex
Hale
49
6.7
(21)
100
(34)
70
(27)
89
(27)
760
(22)
61
(27)
13
(20)
16
(20)
Feaala
61
6.6
(24)
49
(39)
160
(31)
180
(31)
880
(24)
30
(31)
36
(23)
26
(23)
*Dioxin and furan avaraga concantrationa ara expressed in lipid adjusted weight in parta par trillion (pg/g).
-------
Table 2-4.
Summary of Significance Testing for
Differences Between Subpopulations"
Chemical
Effect Due To...
Census Region
Age
Race
Sex
VOIATILES
Benzene
Benzene .059*
Substituted Benzenes
Styrene .502
Ethylphenol .138
Alkyl Benzenes
Toluene .505
Ethylbenzene .757
Xylene .450
Chlorinated Benzenes
Chlorobenzene .049**
1,4-Dichlorobenzene .038**
Trihalomethanes
Chloroform .524
Halocarbons
Tetrachloroethene .078*
SEMI-VOLATILES
PCB8
PCBs .385
Organochlorine Pesticides
Beta-BHC .012**
2/£'-DDE .145
Total DDT .166
Phthalates
Butyl benzyl phthalate .232
.411
.161
.194 .882
.002*** .595
DIOXINS
2,3,7,8-TCDD
1,2,3,7,8-PECDD
HXCDD
1,2,3,4,7,8,9-
OCDD
HPCDD
FURANS
2,3,4,7,8-PECDF
HXCDF
1,2,3,4,6,7,8-HPCDF
.569
.464
.712
.412
.875
.506
.663
.339
.621
.526
.438
.242
.879
.596
.219
.034**
.779
.244
.158
.001***
.674
.016**
.116
.006***
.001*** .108
.310 .734
.954
.695
.000*** .545
.000*** .022**
.064* .169
.010*** .254
.732
.078*
.029** .276
.000*** .102
.643 .936
.215 .734
.003*** .658
.043** .091*
.349 .333
.815 .440
.011**
.013**
.823
.958
.045**
.141
.384
.590
.230
.078*
.142
.751
.280
.008***
.244
* Indicates significance at the .10 level.
** Indicates significance at the .05 level.
*** Indicates significance at the .01 level.
The table entries are p- values, which indicate the exact level of significance at which a statistical difference
can be declared, given the observed data.
18
-------
Note in Table 2-4 that the estimated regional effect on
average concentration was not declared to be statistically
significant for the other compounds. In these cases the total
survey and laboratory variability may preclude detecting
differences that may be present. See section 2.1.4 below for a
discussion of race group influences on regional estimates.
2.1.3 Age Group Estimates
The NHATS classifies specimens into one of three age
groups: 0 to 14 years, 15 to 44 years, and 45 years and older, as
displayed in Figure 2-8. The age group estimates for the
volatiles, the semivolatiles, and the dioxins and furans are
displayed in Figures 2-9, 2-10, and 2-11, respectively. Eight of
the twenty-two chemicals had statistically significant
differences between age groups. Only one volatile organic
compound, ethylphenol, had statistically significant age group
differences. For ethylphenol, average concentration decreased as
age group increased.
In the semi-volatiles, FCBs, p_,p/-DDE, and Q-BHC had
statistically significant differences between age groups. For
these three chemicals, average concentrations increased for the
older age groups. Total Equivalent DDT, which was calculated
from measured concentrations, also was significant with respect
to age, and average concentrations increased as age group
increased.
Four of the eight chemicals with statistically
significant age group differences were dioxins and furans. These
four chemicals were:
2,3,7,8-TCDD,
1,2,3,7,8-PECDD,
OCDD, and
2,3,4,7,8-PECDF.
19
-------
0.35
0.30-
0.25-
0.20-
Concentration
(pg/g)
0.15H
0.10^
0.05-
0.0 -
0.047
(0.020)
Chloroform Benzene
Tetra-
chloro-
ethene
Toluene
Chloro-
benzene
Ethyl-
benzene
Styrene
1,4-
Dichloro-
benzene
Ethyl-
phenol
Xylene
Compound
Figure 2-1. Weighted estimates of the average concentration levels of volatiles
(wet weight, jig/g) for the U.S. population. (Standard errors of the
estimates are in parentheses.)
21
-------
2
2.0 -
1.5 -
Concentration
(pg/g)
1.0 -
0.5-
Beta-BHC
p,p'-DDE
Total DDT
Compound
PCBs
Butyl Benzyl
Phthalate
Figure 2-2. Weighted estimates of the average concentration levels of semi-volatiles
(lipid adjusted, ng/g) for the U.S. population. (Standard errors of the
estimates are in parentheses.)
23
-------
200
180-
160 -1
140-
120-
100-
Concentration
(pg/g) so -
so -
40 -
20 -
Dioxins
1 Furans
*OCDD = 820(100)
6.1
(0.76)
2,3,7,8- 1,2,3,7,8- HXCDD 1.2,3,4, 2,3,4,7,8-
TCDD PECDD 7-8-9- PECDF
HPCDD
Compound
HXCDF
1,2,3.4.6,7,8-
HPCDF
Figure 2-3. Weighted estimates of the average concentration levels of dioxins and f urans
(lipid adjusted, pg/g) for the U.S. population. (Standard errors of the
estimates are in parentheses.)
25
-------
Figure 2-4. United States Census regions
27
-------
05
0.4-
0.3-
Concentration
(pg/g)
0.2-
0.1 -
Census Region = Northeast
0.20
(0.14)
0.021 Q.018
(0-01*) (0.0038)
0.041
(0.019)
Chloroform Benzene Tetra- Toluene Chloro- Ethyl- Styrene 1,4-
chloro-
ethene
benzene benzene
Compound
Dichloro- phenol
benzene
Ethyl- Xylene
0.5
0.4-
0.3-
Concentration
(ng/g)
0.2 H
0.1-
0.0-
0.49
(0.29)
Census Region = South
0.20
(0.047)
0.049
(0.027)
0.010
0.016
0.046
(0.022)
(0.0017) (0-0058>
0.0072
(0.0017)
0.10
(0.061)
Chloroform Benzene
Tetra- Toluene Chloro- Ethyl- styrene 1,4- Ethyl- Xylene
chloro- benzene benzene Dichloro- phenol
ethene benzene
Compound
Figure 2-5. Weighted estimates of the average concentration levels of
volatiles (wet weight, pg/g) for each census region. (Standard
errors of the estimates are in parentheses.)
29
-------
0.5
0.4 -
0.3 -
Concentration
0.2 -i
0.1 -1
Census Region = North Central
0.041
(0.026
Chloroform Benzene Tetra- Toluene Chloro- Ethyl- styrene 1-4- Ethylphenol Xylene
chloro- benzene benzene Dkhloro-
ethene ,— __, benzene
Compound
0.5
0.4
0.3
Concentration
0.2-
0.1-
Census Region = West
0.081
(0.058)
0.13
(0.054)
0.052
(0.020)
°-10 (0084)
(0.053) MM
11
Chloroform Benzene Tetra- Toluene Chloro- Ethyl- styrene 1.4- Ethylphenol Xylene
chloro- benzene benzene Dichloro-
ethene benzene
Compound
Figure 2-5 (Continued)
31
-------
1.5 -
Concentration
(pg/g)
1.0 H
Census Region = Northeast
Beta-BHC p,p'-DDE Total DOT PCBs
Compound
Butyl Benzyl
Phthalate
2.5
2.0 -
1.5 -
Concentration
1.0 H
0.5-
Beta-BHC
Census Region = South
p.p'-DDE
Total DDT
Compound
PCBs
Butyl Benzyl
Phthalate
Figure 2-6. Weighted estimates of the average concentration levels of semi-volatiles
(lipid adjusted, ng/g) for each census region. (Standard errors of the
estimates are in parentheses.)
33
-------
2.5
2.0 ~
1.5-
Concentration
1.0 -\
Beta-BHC
p,p'-DDE
Census Region = North Central
Total DDT
Compound
PCBs
0.45
(0.31)
Butyl Benzyl
Phthalate
2.5
2.0 -
1.5 -
Concentration
Beta-BHC
p.p'-DDE
Total DDT
Compound
Census Region = West
0.20
(0.11)
PCBs
0.21
(0.18)
Butyl Benzyl
Phthalate
Figure 2-6. (Continued)
35
-------
200
150
Census Region = Northeast
*OCDD = 750(180)
Concentration
(pg/g) 100 -
50-
2.3,7,8-
TCDD
1,2,3.4,6,7,8-
HPCDF
Compound
200
150-
Concentration
(pg/g)
Census Region = South
*OCDD = 780(140)
2,3,7.8-
TCDD
HXCDF 1,2,3,4,6,7,8-
HPCDF
Compound
Figure 2-7. Weighted estimates of the average concentration levels of dioxins and
furans (lipid adjusted, pg/g) for each census region. (Standard errors of
the estimates are in parentheses.)
37
-------
200
150 ~
Concentration
(Pg/9) 100
50 -
180
(51)
Census Region = North Central
*OCDD = 920(190)
2,3,7,8-
TCDD
1.2.3,7,8-
PECDD
HXCDD
1,2,3,4,
7,8,9-
HPCDD
2,3,4,7,8-
PECDF
HXCDF 1,2.3,4.6.7,8-
HPCDF
Compound
200
150 ~
Concentration
(pg/g) 100
50 -
Census Region = West
*OCDD = 850(250)
1.2,3,4, 2,3,4,7,8-
7,8,9- PECDF
HPCDD
Compound
HXCDF 1,2,3,4.6,7,8-
HPCDF
Figure 2-7. (Continued)
39
-------
0-14 Years
Figure 2-8. NHATS age groups
41
-------
0.35
0.30-
0.25-
0.20-
Concentration
(jig/g) 0.15-
0.10-
0.05-
Age = 0-14 years
o.o
0.053
(0.027)
Chloroform Benzene
Tetra-
chloro-
ethene
Toluene
Chloro-
benzene
Ethyl-
benzene
Styrene 1.4- Ethyl-
Oichloro- phenol
benzene
Xylene
Compound
0.35-
0.30-
0.25-
0.20-
Concentration
(pg/g) 0.15-
0.10-
0.05-
Age = 15-44 years
o.o
0.053
(0.026)
Chloroform Benzene
Tetra-
chloro-
ethene
Toluene
Chloro-
benzene
Ethyl-
benzene
Styrene 1.4- Ethyl-
Dichloro- phenol
benzene
Xylene
Compound
Figure 2-9. Weighted estimates of the average concentration levels of volatiles
(wet weight, ng/g) for each age group. (Standard errors of the
estimates are in parentheses.)
43
-------
0.30
0.251
0.20-
Concentration
0.15-
o.io-
0.05
Age = 45 + years
0.0
0.033
Chloroform Benzene Tetra- Toluene Chloro- Ethyl- Styrene 1,4- Ethyl- Xylene
chloro- benzene benzene Dichloro- phenol
ethene benzene
Compound
Figure 2-9. (Continued)
45
-------
2.5
2.0 -
1.5 -
Concentration
(pg/g)
1.0 -1
0.5-
Beta-BHC
Age = 0-14 years
0.75
(0.24)
p,p'-DDE
Total DDT
Compound
PCBs
Butyl Benzyl
Phthalate
2.5
2.0 -I
1.5 -
Concentration
1.0 -I
0.5-
Beta-BHC
p,p'-DDE
Total DDT
Compound
Age = 15-44 years
PCBs
0.31
Butyl Benzyl
Phthalate
Figure 2-10. Weighted estimates of the average concentration levels of semi-volatiles
(wet weight, pg/g) for each age group. (Standard errors of the estimates
are in parentheses.)
47
-------
2.5
2.0 -
1.5 -
Concentration
(ng/g)
1.0 -i
0.5-
0.31
(0.065)
Beta-BHC
Age = 45+ years
p,p'-DDE
Total DDT
Compound
PCBs
Butyl Benzyl
Phthalate
Figure 2-10. (Continued)
49
-------
200-
150^
Concentration
100-
Age = 0-14 years
*OCDD = 410(85)
50-
HXCDF 1,2,3,4.6,7,8-
HPCDF
Compound
200
150-
Concentration
100-
Age = 15-44 years
*OCDD = 920(170)
2.3,7,8-
TCDD
Figure 2-11.
Compound
Weighted estimates of the average concentration levels of dioxins and
furans (lipid adjusted, pg/g) for each age group. (Standard errors of the
estimates are in parentheses.)
51
-------
200-
150-
Concentration
(pg/g) 100-
Age = 45 + years
'OCDD = 990(180)
2,3,7,8- 1,2,3.7,8- HXCDD
TCDD PECDD
1-2,3 A.
7-8,9-
HPCDD
2,3,4,7,8-
HXCDF 1,2,3,4,6,7,8-
HPCDF
Compound
Figure 2-11. (Continued)
53
-------
Concentrations of 2,3,7,8-TCDD, 1,2,3,7,8-PECDD, and 2,3,4,7,8-
PECDF were higher in the 15 to 44 age group than in other two age
groups. Concentration levels of OCDD were higher in both the
adult groups (15-44 and 45 years and older) than in the youngest
(0-14 years) age group. For the dioxin 1,2,3,7,8-PECDD, the
oldest age group had a much lower concentration than either of
the other two age groups. This did not appear to be an artifact
of the model, as the concentrations of the composite samples from
the oldest age group, with one exception, were clustered at the
lower end of the range. Other researchers have found that
2,3,7,8-TCDD concentration levels increase with age (Patterson et
al., 1986). Future years' surveys will be closely monitored to
see if the results of the FY82 NHATS survey are replicated.
2.1.4 Comparison of Estimated Average Concentration Levels
Across Sex and Race Groups
Table 2-4 shows the p-values for comparisons of
average concentration levels by race and sex for each of the 22
compounds. For five of the ten volatile organic compounds there
was a statistically significant sex effect, with the estimated
average concentration levels for males being greater in each
case. These compounds were styrene, toluene, xylene, chloroform,
and tetrachloroethene.
Among the semi-volatiles, the average concentration of
p_,p_'-DDE was significantly greater for males. The sex difference
in p_,p_'-DDE appears anomalous and is primarily attributable to a
very low concentration in one pure female composite. The model-
based average concentration estimate was less than twice as high
for males as for females when this composite was omitted, and the
difference was no longer statistically significant.
Historically, p_,p_'-DDE estimates from the Human Monitoring
Program for the two sexes have been about the same.
55
-------
Among the dioxins and furans, the average concentration
levels for females were significantly greater for the dioxin
HXCDD and the furan HXCDF.
The average concentration levels of five compounds
different significantly by race group. Toluene, chlorobenzene,
B-BHC, butyl benzyl phthalate, and 2,3,4,7,8-PECDF each showed an
average concentration level for Caucasians that was greater than
the average level for non-Caucasians.
However, because all six of the pure non-Caucasian
composites were located '- the South and no composite outside of
the South was more than .1^ non-Caucasian, the estimation of the
race group effect in the model was driven almost entirely by the
data in the South. Therefore, the validity of the national race
group effect estimated by the multiplicative model depends on
accepting the assumption that there is no interaction between
race and region, and that race-group effects observed in the
South apply elsewhere.
In fact, for the chemicals B-BHC and chlorobenzene,
where the model declared a statistically significant race group
effect on concentration, the arithmetic mean concentrations of
non-Caucasian composites in the South were not much different
than the mean concentrations of the Caucasian composites outside
the South. In comparison, the means of the Caucasian composites
in the South were higher than those of the non-Caucasians in the
South. The combination of these data result in a model-based
estimate of a significant regional effect—the South higher than
the other regions—and a significant race group effect—
Caucasians higher. This situation did not occur for the three
other chemicals showing significant differences by race; the
average concentrations of these chemicals were less for non-
Caucasian composites in the South than for Caucasian composites
in all regions. Also, all chemicals showing significant average
concentration differences by region but not by race group
(benzene, 1,4-dichlorobenzene, and tetrachloroethene) had mean
56
-------
concentrations in Caucasican composites in the South that were
similar to the levels observed in non-Caucasian composites in the
South.
Because of the lack of empirical support for estimating
race group differences in regions other than the South, the
model-based estimate of a national race group effect should be
interpreted cautiously. Race group effects estimated in future
surveys will be monitored closely to see if the results of the
NHATS FY82 survey are replicated.
2.1.5 Relative Standard Errors
The precision of each of the average concentration
estimates at the national level, and by region, age group, race
group, and sex is measured by an associated standard error.
Contributing to the standard error of the estimated average
concentration are errors due to sampling Standard Metropolitan
Statistical Areas (SMSA's) and individual specimens within SMSA,
errors in tissue preparation and chemical analysis, and possible
errors introduced because of model mis-specification. The
standard error of the estimated average concentration is also a
function of the underlying variability of the concentration in
the population. Estimates of the standard errors are shown in
Table 2-3, where they are expressed as relative standard errors,
which are percentages of the associated average concentration
estimates. Note that at the national level the relative standard
errors range from a low of 12 percent for benzene, to a high of
50 percent for butyl benzyl phthalate; 10 of the relative
standard errors are less than 20 percent, 8 are between 20 and 30
percent, and the other 5—toluene, xylene, ethylbenzene,
chloroform, and butyl benzyl phthalate—exceed 30 percent. The
high relative standard errors of 37 percent for toluene and 42
percent for chloroform are primarily attributable to a few
composites with extreme concentration values. The relative
standard errors of 50 percent for butyl benzyl phthalate and 42
57
-------
percent for xylene were high because the concentrations of these
chemicals had a greater degree of variability across the
composites than the other chemicals and their distribution did
not fit the multiplicative model as well. In the case of
1,2,3,7,8-PECDD the two highest measured concentrations were so
much greater than the rest of the distribution that they were
deemed to be outliers and removed from the analysis to avoid
distorting the estimate of average concentration level. A
detailed explanation of the rationale for this decision is
contained in Section 8.5.
2.2 Incidence of Detection for Compounds Identified
in the Composite Samples
Results on the percentage of composite samples having
detectable levels do not necessarily imply that the percentages
of detected levels for individual samples were similar. For
example, if a compound is detected in all of the composite
samples, it may or may not be present in all of the individual
specimens contained in the composite samples. The estimation of
prevalence in composite samples is addressed in a separate study
(Orban et al. 1987).
2.2.1 Volatile Organic Compounds
The incidence of detection for volatile organic
compounds varied across the chemical classes. Eight of the nine
compounds from the four benzene related classes were detected in
greater than 90% of the composite samples. For instance, benzene
was detected in 96% of composite samples and 1,4-dichlorobenzene
was found in all the composite samples. The compound 1,2-
dichlorobenzene, which was detected in 63% of the composites, was
the lone exception.
Several compounds, specifically styrene, ethylphenol,
xylene, and 1,4-dichlorobenzene, were detected in all of the
composite samples. Bromodichloromethane, dibromochloromethane,
58
-------
bromoforra and 1,1,2-trichloroethane from the trihalomethane and
halocarbon chemical classes were not detected in any of the
composites. Incidence data for the volatile compounds are
provided in graphical format in Figure 2-12 for benzenes/ and in
Figure 2-13 for trihalomethanes and halocarbons. These data are
listed in tabular format in Appendix B, Table B-l.
59
-------
Benzene Styrene Ethylphenol Toluene Ethylbenzene Xylene Chloro- 1,2-Dichloro- 1,4-Dichloro-
benzene benzene benzene
Compound
Figure 2-12. Percentage of FY82 composite samples in which benzenes were detected
61
-------
Trihalomethanes
Halocarbons
46
Chloroform
Bromo-
dichloro-
methane
Dibromo-
chloro-
methane
Bromo-
form
1,1,1-
Trichloro-
ethane
1,1,2-
Trichloro-
ethane
1,1,2,2-
Tetrachloro-
ethane
Tetrachloro-
ethene
Compound
Figure 2-13. Percentage of FY82 composite samples in which trihalomethanes and
halocarbons were detected
63
-------
2.2.2 Semi-Volatile Organic Compounds
The semi-volatile organic compounds were detected in
various degrees across and within the chemical classes. For the
PCB homolog groups, the incidence of detection ranged from 7% for
decachlorobiphenyl to 75% for hexachlorobiphenyl. PCBs were
detected in 38 out of 44 of the composites, or approximately 86%
of the sample. All six of the composites where PCBs were not
detected were from the youngest age group. This result is at
variance with what was observed in the survey of FY81—when all
422 individual samples including 94 from the youngest age were
found to have detectable levels of PCBs—and the survey of FY83—
when all 407 specimens including 63 from the youngest age group
had detectable PCB levels. The surveys from these other years
were not strictly comparable to FY82 since they employed a
different analytical procedure and measured concentration levels
on individual specimens, not composites. PCB levels will be
closely monitored in future surveys to clarify the trend.
Percentages for the organochlorine pesticides ranged from 14% for
mirex to 100% for p_,p_'-DDE. The incidence for p_,p_'-DDT was 68%.
Only these two of the six DDT isomers were identified in the
Broad Scan Analysis Study composite samples. Of the aromatics
and chlorinated benzenes, only hexachlorobenzene (79%) was
detected in more than 50% of the composite samples.
The phthalates were detected in more of the composite
samples than the phosphates. Tributyl phosphate and
tris(2-chlorethyl) phosphate were detected in only 2% of the
composite samples. Graphs depicting the incidence of detection
for semi-volatile compounds are provided in Figures 2-14, 2-15,
2-16, and 2-17 for PCBs, organochlorine pesticides, aromatics and
chlorinated benzenes, and phthalates and phosphates,
respectively. These data are listed in tabular format in
Appendix B, Table B-2. In general the detection percentages were
lower for the FY82 composite samples than they were for the FY81
and FY83 individual specimens. For example, p,p'-DDT was
65
-------
detected in only 68% and hexachlorobenzene in only 79% of the
FY82 composites, in contrast to the other two years when more
than 99% of the survey specimens were found to have detectable
amounts of both compounds. One exception to the general rule is
the compound mirex, which was detected in 14% of the FY82
composites but in less than 1% of the FY81 and FY83 specimens.
2.2.3 Dioxins and Furans
Four of the five dioxins were detected in more than 90%
of the composite samples. The exception was 2,3,7,8-TCDD which
was detected in 74% of the composites. The percentage detected
for the furans ranged from 26% for 2,3,7,8-TCDF to 93% for
1,2,3,4,6,7,8-HPCDF. Graphs for percentage detected data are
provided in Figure 2-18. These data are listed in tabular format
in Appendix B, Table B-3.
66
-------
PCBs Tr'- Tetra- Pcnta- Hcxa- Hepta- Oeta- Nona- Deea-
chloro- chloro- chloro- chloro- chloro- chloro- chloro- chloro-
biphenyl biphenyl biphenyl biphenyl biphenyl biphenyl biphenyl biphenyl
Compound
Figure 2-14. Percentage of FY82 composite samples in which PCB homolog groups were
detected
67
-------
Beta-BHC p,p'-DDE p,p'-DDT Mirex
trans- Heptachlor
Nonachlor Epoxide
Oieldrin
Compound
Figure 2-15. Percentage of FY82 composite samples in which organochlorine pesticides
were detected
69
-------
100
90 ~
80
70
60
Percent
so
40 -
Aromatics
Chlorinated Benzenes
Naphthalene phe"-
anthrene
Pyrene 1,2-Dichloro- 1,2,4-
benzene Trkhloro-
benzene
Compound
Pentachloro- Hexachloro-
benzene benzene
Figure 2-16. Percentage of FY82 composite samples in which aromatics and chlorinated
benzenes were detected
71
-------
100-
90-
80-
70-
60-
Phthalates
Phosphates
Diethyl
Phthalate
Di-n- Diethyl
butyl Hexyl
Phthalate Phthalate
Butyl Triphenyl
Benzyl Phosphate
Phthalate
Compound
Tributyl Tris-
Phosphate (-2-Chloroethyl)-
Phosphate
Figure 2-17. Percentage of FY82 composite samples in which phthalates and phosphates
were detected
73
-------
100-
90 -
80 -
70 -
60 -
50 -
Percent
40 -
30 -
20 -
10 -
0
98 98
100
93
74
Dioxins
Furans
2,3,7,8- 1,2,3,7,8- HXCDD 1,2,3,4, OCDD 2,3,7,8- 2,3,4,7,8- HXCDF 1,2,3,4, QCDF
TCDD PECDD 7,8,9- TCDF PECDF 6.7.8-
7,8,9-
HPCDD
HPCDF
Compound
Figure 2-18. Percentage of FY82 composite samples in which dioxins and furans were
detected
75
-------
3.0 QUALITY ASSURANCE
An extensive quality assurance/quality control (QA/QC)
effort accompanied the analysis of the Broad Scan Analysis Study
composite samples. The results of this effort were used to
assess the performance of the HRGC/MS method. Percent recovery,
precision and estimated limit of detection (LOD) data were
reported.
Information on method percent recovery was obtained by
spiking QA/QC samples with a known amount of analyte and
determining what percentage of that amount was estimated as
present in the sample. Percent recovery (PR) was calculated by
the formula:
PR = Amount of Analyte Found in QA/QC Sample x 100%
Known Spike Amount
(Equation 3-1)
Average percentage recoveries (PR) were also calculated.
Method precision information was obtained using the
percent recovery data. It was expressed in terms of standard
deviation and percent relative standard deviation. The standard
deviation (S) was calculated by the formula:
9 N —
S* - (1/(N-1)) _Z (PR± - PR)«
(Equation 3-2)
where
PR^ is the percent recovery for the ith QA/QC sample;
PR is the average percent recovery of the QA/QC samples;
and
N is the number of QA/QC samples.
77
-------
The relative standard deviation (RSD) was calculated as:
RSD » 100 S/PR
(Equation 3-3)
It is expressed as a percentage.
Estimated limits of detection (LODs) were reported only
for those composite samples whose observed concentration level
was below the level of quantification; that is, those composite
samples whose concentration levels were determined to be either
not detected or trace. The LODs were reported as total mass of
target analyte detectable and as the equivalent concentration
level. LOD concentrations were calculated based on wet weight
for the volatiles analyses and extractable lipid weight for the
semi-volatiles, dioxins and furans analyses. In this report, the
maximum and minimum of the reported LOO amounts and equivalent
concentrations are presented. For some chemicals, there were
composite samples whose reported concentrations were above the
limit of quantification but were also below the maximum LOD as
presented in this section. This occurred because of variation
between samples with respect to sample weight, minimum analyte
mass detectable, or both.
This report presents a summary of some of the QA results
obtained in the Broad Scan Analysis Study. Data for those QA/QC
samples whose results are most comparable to the results obtained
from the Broad Scan Analysis Study composite samples, that is,
spiked human adipose tissue QC samples, are provided here.
Percent recoveries calculated for the human adipose tissue QC
samples include any possible background contribution from the
adipose tissue itself. Additional information from results of
other QA/QC samples is provided in USEPA a-e (1986).
78
-------
3.1 Volatile Organic Compounds
Several types of QA/QC procedures were performed with
the volatiles analyses of the Broad Scan Analysis Study composite
samples. They included daily instrument performance checks
through analyses of internal QC and external QC (performance
audit) samples, analyses of spiked human adipose tissue samples,
and analyses of internal standard responses. Concentrations were
calculated based on wet weight.
Data from the analyses of five human adipose tissue QC
samples, run with the first sample batches, were obtained for 14
of the 17 target analytes. Each QC sample was made up of 20
grams of spiked human adipose tissue. Spike levels ranged from
0.20 ftg to 1.4 jig per 20 grams. These levels were equivalent to
concentrations ranging from 0.010 /tg/g to 0.070 /ig/g. Results of
these analyses found that the HRGC/MS method performed quite well
with respect to bias. Average percent recoveries ranged from 85%
for chloroform to 141% for styrene. The high recoveries for
chemicals such as styrene may be due to background contribution
from the adipose tissue itself. Precision data, however, were
quite variable for the different compounds. The relative
standard deviations ranged from 4% for benzene to 54% for
bromodichloromethane. Styrene, tetrachloroethene, and 1,1,1-
trichloroethane had both high recoveries and high relative
standard deviations. In general, precision was better for
compounds that were quantitated versus their associated
deuterated analog than for compounds that were quantitated versus
bromochloropropane, the internal standard for analyses in which a
deuterated analog was not available. These results are
summarized in Table 3-1.
Ranges of reported limits of detection for the 17
volatile compounds are provided in Table 3-2. In general, the
LOD target of .05 to .10 /ig/g was met.
79
-------
Table 3-1. Summary of QC Results for Selected Volatile Organic Analytes
in Spiked 20 Gram Aliquots of Human Adipose Tissue"
Precision
Number of
Chemical QC Samples"
Benzene
Benzene
Substituted Benzenes
Styrene
Alkyl Benzenes
Toluene
Ethylbenzened
Chlorinated Benzenes
Chlorobenzene
I f 2-Dichlorobenzened
Trihalomethanes
Chloroform
Bromodichloromethane
Dibromochloromethane
Bromoform
Halocarbons
1,1, 1-Trichloroethane
1,1, 2-Trichloroethane
1,1,2, 2-Tetrachloroethaned
*
Tetrachloroethene
4
5
5
5
5
3
2
5
5
5
5
5
5
5
Average
Percent
Recovery
100
141
97
105
104
103
85
111
111
99
131
94
103
122
Standard
Deviation
(Units -
Percent)
4
71
27
10
21
8
14
60
20
25
59
9
8
55
Relative
Standard
Deviation6
4
50
27
10
21
7
17
54
18
25
45
10
8
45
Thes* samples were analyzed with the first sample batchaa of the Broad Scan Analysis Study composite siaplas. Spika levels
•era equivalent to concentration levels ranging froe 0.01 jig/g to 1.07V /JQ/g.
For half of the cheaicals listed in Table 3-1, there are five QC samples. For the other chemicals, the nuaber of QC saaples
is less than five for one of tvo reasons: percent recovery for the QC saiple was not determined or percent recovery for the
QC sasple was determined by a sethod not comparable to other recoveries for that chemical.
c Relative standard deviation is the standard deviation expressed as a percentage of the Average Percent Recovery.
Quint Station for these analytes was performed versus the deuterated analog of the specific compound. All other calculations
were performed versus the internal standard, bromchloropropane.
80
-------
Table 3-2. Ranges of Reported Limits of Detection of Volatile Organic
Compounds for Composite Samples Whose Concentration Levels
Were Declared Not Detected or Trace
Reported Limit of
Detection
(UQ)
Chemical
Benzene
Benzene
Substituted Benzenes
Styrene3
Ethylphenol
Alkyl Benzenes
Toluene
Ethylbenzene
Xylene*
Chlorinated Benzenes
Chlorobenzene
1 , 2-Dichlorobenzene
I , 4-Dichlorobenzenea
Trihalomethanes
Chloroform
Bromodichlororaethane
Dibroraochlororaethane
Bromoforra
Halocarbons
1,1, 1-Trichloroethane
1,1 , 2-Trichloroethane
1,1,2, 2-Tetrachloro-
ethane
Tetrachloroethene
Minimum
0.080
c
0.005
0.003
0.010
—
0.004
0.001
—
0.020
0.53
0.030
0.008
0.04
0.021
0.001
0.020
Maximum
0.095
0.005
0.005
0.050
—
0.040
0.020
—
0.74
5.4
0.50
0.67
2.7
0.50
0.090
0.80
Equivalent Concentration
Wet Weight Basis
(na/cr)
Mininum
0.0044
0.0002
0.0002
0.0009
—
0.0003
0.0001
—
0.0008
0.021
0.0013
0.0004
0.0022
0.0010
0.0001
0.0009
M •y'miim
0.013
0.0002
0.0004
0.0027
—
0.0026
0.0015
—
0.10
0.50
0.033
0.050
0.24
0.050
0.0052
0.033
This compound ms detected in all composite s»plea.
81
-------
3.2 Semi-Volatile Organic Compounds
Quality assurance/quality control procedures for the
semi-volatiles analyses . included analyses of method blanks,
spiked blanks, porcine fat samples prepared by EPA/EMSL-LV,
spiked human adipose tissue samples, and replicate analyses of
homogenized human adipose tissue samples. Analyses for surrogate
compounds and anthracene-d10, the internal standard, were also
performed for each composite sample. Results for adipose tissue
samples were adjusted for extractable lipid weight. The results
of the analyses of the spiked human adipose tissue samples are
described in this section.
Fifty-two target analytes were analyzed in five human
adipose tissue QC samples at spiking levels equivalent to
concentrations of 0.10 /ig/g. Each QC sample was a 20 gram
aliquot of adipose tissue. Results for these samples were
considerably more variable than the results for the volatile QC
samples cited in Section 3.1. Average percent recoveries were
lower for semi-volatiles than for volatiles. Seventeen of the 52
semi-volatiles listed in Table 3-3 had average percent recoveries
of less than 50%. Thirty-three semi-volatiles had recoveries
between 51 and 100%, while only two had recoveries exceeding
100%. For the fourteen volatiles listed in Table 3-1, five had
recoveries between 85 and 100%, while nine had recoveries over
100%. The average recovery for 1,2-dichlorobenzene was 48% for
the semi-volatiles protocol compared to 103% for the volatiles
protocol.
The semi-volatile, p_* p_' -DDE, had an unusually high
recovery of 204%. This recovery is believed to be due to the
background contribution from the bulk adipose tissue itself.
Relative standard deviations (RSDs) for the semi-
volatiles were higher in general than those for the volatiles.
The relative standard deviations for semi-volatiles ranged from
12% to 74%. About three-fourths (38/52) of the semi-volatile
82
-------
RSDs were 20% or greater. For volatiles, the range was 4% to 50%
and half of the volatiles had RSDs of 20% or more.
There appeared to be differences in method performance
between the compound classes. Highest average percent recoveries
were observed for the organochlorine pesticides, ranging from 56%
for o-p/-DDE to 93% for p_,p/-DDD, as well as the 204% for
p_,p_'-DDE. The phosphates, with average recoveries from 22% to
29%, had the lowest recoveries. The results are presented in
Table 3-3.
Ranges of reported limits of detection are provided in
Table 3-4. The reported limits of detection (LODs) for diethyl
phthalate, di-n-butyl phthalate, and di-n-octyl phthalate are
relatively high. These compounds were detected in the associated
method blanks. This is not an unusual situation. (McLafferty
1980). In general, the target LOD concentrations of .05 to .10
pg/q were achieved.
3.3 Dioxinsand Furans
The QA/QC procedures for the dioxins and furans analyses
included the following:
• Analysis of method blanks;
• A check on the response factor each day;
• A check on the column resolution for 2,3,7,8-TCDD
each day;
• Estimation of recovery of internal standards; and
• Qualitative verification of 2,3,7,8-TCDD in certain
extracts.
Unlike the volatiles and semi-volatiles analyses, measurements on
spiked human adipose tissue samples were not reported for the
dioxins and furans analyses. Therefore, equivalent information
on the bias and precision of the dioxins and furans method is not
83
-------
available. The Limits of Detection for dioxins and furans were
reported as lipid adjusted concentrations in picograms per gram.
The minimum and maximum of these concentrations are presented in
Table 3-5. The LOD goals for the analyses were 0.1 to 1 nanogram
per gram. This is equivalent to 100 to 1000 picograms per gram.
An inspection of Table 3-5 indicates that the LOD goals of the
analysis appear to have been met.
84
-------
Table 3-3. Summary of QC Results for Selected Semi-Volatile Organic
Analytes in Spiked 20 Gram Aliquots of Human Adipose Tissue*
Precision
Chemical
organochlorine Pesticides
a-BHC
B-BHC
A-BHC
0-D/-DDE
E-P/-DDE
o-o '-DDD
£-p/-DDD
O-p/-DDT
B-E'-DDT
Mirex
trans -Nonachlor
Heptachlor
Heptachlor Epoxide
Dieldrin
Aldrin
uromatics
Naphthalene
dg -Naphthalene
Phenanthrene
Pyrene
Dimethyl phthalate
Number of,
QC Samples
5
4
5
5
5
5
5
5
5
5
4
5
5
4
5
3
5
5
5
5
Average
Percent
Recovery
83
66
62
56
204C
61
93
90
84
58
76
73
91
74
74
120
39
70
78
56
Standard
Deviation
19
21
8
10
20
13
31
35
39
28
12
13
18
45
9
22
17
11
16
19
Relative
Standard
Deviation
23
32
13
18
10
21
33
39
46
48
16
18
20
61
12
18
44
16
21
34
85
-------
Table 3-3. (continued)
Precision
Average Standard Relative
Number of. Percent Deviation Standard
Chemical QC Samples0 Recovery Deviation
Phthalates
Diet by 1 phthalate
Di-n-butyl phthalate
Butyl benzyl phthalate
Di-n-octyl phthalate
Phosphates
Tris ( 1 , dichloropropyl ) phos-
phate
Triphenyl phosphate
Tri-m-tolyl phosphate
Chlorinated Benzenes
1 , 2-Dichlorobenzene
1,2, 4-Trichlorobenzene
Cg-1 ,2,4, 5-Tetrachloro-
benzene
Pentachlorobenzene
Hexachlorobenzene
Cg-Hexachlorobenzene
Bromobiphenyls
4 -Br omobipheny 1
4,4' -Dibromobiphenyl
2,4, 6-Tribromobiphenyl
2 , 2 ' , 4 ' , 5-Tetrabroraobiphenyl
5
5
5
5
5
5
4
3
3
5
5
5
5
5
5
5
5
65
85
41
63
22
29
22
48
51
39
34
49
48
52
95
66
87
13
34
23
46
11
15
7
8
10
16
14
12
9
13
14
10
33
20
40
56
73
50
52
32
17
20
41
41
24
19
25
15
15
38
86
-------
Table 3-3. (continued)
Precision
Average Standard Relative
Number of. Percent Deviation Standard
Chemical QC Samples0 Recovery Deviation
Chlorodiphenyls
4-Chlorodiphenyl ether
2,2', 4 , 4 ' , 5-Pentachloro-
diphenyl ether
Chloroterphenyls
4-Chloro-p_-terphenyl
2 , 5-Dichloro-o-terphenyl
2,4' r 5-Trichloro-o-terphenyl
2,4,4', 6-Tetrachloro-o-
terphenyl
Chlorophenols
2 , 4-Dichlorophenol
2,4, 6-Trichlorophenol
Other Compounds
Acenaphthylene
Acenaphthene
Fluorene
Fluoranthene
r-Chlordane
Chrysene
d12-Chrysene
5
5
5
5
5
5
4
4
5
5
5
5
5
5
5
41
73
45
58
72
80
74
37
44
46
50
82
71
61
37
14
21
18
25
53
57
24
8
13
16
15
14
10
10
14
34
29
40
43
74
71
32
22
30
35
30
17
14
16
38
Spile* lava I a war* equivalent to concentrations of 0.10
In cases where tha nirabar of RC sasples analyzed is less thin 5, the percent recovery value was not determined for those
rniining saaplea.
°High recovery rate due to contribution fros the adipose tissue eatrix.
87
-------
Table 3-4. Ranges of Reported Limits of Detection of Semi-Volatile Organic
Were Declared
Chemical
Organochlorine Pesticides
8-BHC
£,2 '-DDE
p_,p_'-DDT
Mirex
trans -Nonachlor
Heptachlor Epoxide
Dieldrin
PCBs
PCBs
S
V "' '
Trichlorobiphenyl
Tetrachlorobiphenyl
Pentachlorobiphenyl
Hexachlorobiphenyl
Heptachlorobiphenyl
Octachlorobiphenyl
Nonachlorobiphenyl
Decachlorobiphenyl
Chlorinated Benzenes
1 , 2-Dichlorobenzene
1,2 , 4-Trichlorobenzene
Pentachlorobenzene
Hexachlorobenzene
Not Detected or Trace
Reported Linu
Detection
Ucrt
M\niwnm
0.20
0.20
0.20
0.20
0.40
0.20
1.0
0.20
0.20
0.20
0.40
0.40
0.40
0.40
* 0.40
1.0
0.20
0.20
0.20
0.20
Lt Of
Maximum
0.20
0.20
0.20
0.20
0.40
0.20
1.0
0.20
0.20
0.20
0.40
0.40
0.40
0.40
0.40
1.0
0.20
0.20
0.20
0.20
Equivalent Concentration
Lipid Weight Basis
(tfQ/Q)
' Mininum
0.012
0.011
0.0092
0.0088
0.018
0.0088
0.044
0.014
0.0088
0.0088
0.018
0.018
0.018
0.018
0.018
0.044
0.0088
0.0088
0.0088
0.0088
Maximum
0.036
0.016
0.036
0.036
0.071
0.036
0.18
0.036
0.036
0.036
0.071
0.071
0.071
0.071
0.071
0.18
0.036
0.036
0.036
0.036
88
-------
Table 3-4. (continued)
Reported Limit of
Detection
(ucrt
Chemical
Aromatics
Naphthalene
Phenanthrehe
Pyrene
Phthalates
Diethyl Phthalate
Di-n-butyl Phthalate
Diethyl Hexyl Phthalate
Butyl Benzyl Phthalate
Phosphates
Triphenyl Phosphate
Tributyl Phosphate
Minimum
0.20
0.20
0.20
0.20
0.20
0.20
0.20
0.40
1.0
Maximum
0.20
0.20
0.20
1.3
3.3
12.2
0.20
4.4
1.0
Equivalent Concentration
Lipid Weight Basis
(HQ/Q)
Mininum
0.0088
0.0088
0.0088
0.0089
0.0095
0.0089
0.0089
0.018
0.044
Maximum
0.036
0.036
0.036
0.077
0.18
0.69
0.019
0.59
0.13
Tris (2-Chloroethyl)
Phosphate
0.80
0.80
0.035
0.14
89
-------
Table 3-5.
Ranges of Reported Limits of Detection of
Dioxins and Furans for Composite Samples
Whose Concentration Levels Were Declared
Not Detected or Trace
Chemical
Number of
Samples with
Concentration
Concentration
Lipid Adjusted
Dioxins
2,3,7,8-TCDD
1,2,3,7,8-PECDD
HXCDD
1,2,3,4,7,8,9-HPCDD
OCDDa
Furans
2,3,7,8-TCDF
2,3,4,7,8-PECDF
HXCDF
1,2,3,4,6,7,8-HPCDF
OCDF
ND or TR
13
15
4
1
0
34
9
19
7
33
Minimum
1.3
1.3
13
26
—
1.3
1.3
3.0
3.5
1.2
Maximum
24
140
49
26
.... ••-
45
46
51
19
200
aThis compound was detected in all composite samples.
90
-------
4.0 SAMPLING AND COMPOSITING DESIGNS
4.1 Sampling Design
The human adipose tissue specimens analyzed in the Broad
Scan Analysis Study were collected from October, 1981 through
September, 1982, following the NHATS sampling design. The NHATS
program uses a statistically based survey design to obtain
adipose tissue specimens from autopsied cadavers and surgical
patients. Although the NHATS target population is the general,
non-institutionalized U.S. population, the sampling population is
limited to cadavers and surgical patients due to the invasive
nature of the process required to collect the adipose specimens
from living persons.
The FY82 NHATS sampling design involved a two-stage
selection process. In the first stage, Standard Metropolitan
Statistical Areas (SMSAs) were randomly selected from the nine
Census divisions of the continental United States, with
probabilities proportional to their 1970 census population size.
The number of SMSAs selected from each Census division was
proportional to the population of the Census division relative to
the total U.S. population. In the second stage, individual
tissue specimens were collected by cooperating medical examiners
and pathologists within the selected SMSAs using target quotas
for various age, race and sex categories. The categories were:
• Age ("0-14 years," "15-44 years," "45+ years")
• Race (white, non-white)
• Sex (male, female)
The SMSA target quotas were proportional to the 1970
U.S. Census population counts for the Census Division in which
the SMSA is located. The tissue specimens were selected in a
nonprobalistic manner based on the judgment of the medical
examiner or pathologist involved (Lucas et al. 1982). An
91
-------
overview of the FT82 sampling design is provided in Figure 4-1.
A map of the SMSAs selected for the F782 NHATS collection is
provided in Figure 4-2. These SMSAs are listed in Appendix C,
Table C-l.
Because the survey required some divergence from strict
probabilistic sampling, the validity of the statistical estimates
derived from the data depends on several assumptions. First, the
concentration of toxic substances in the adipose tissue of
cadavers and surgical patients is assumed to be the same as in
the general population. Second, it is assumed that the level of
toxic substances in urban residents is approximately the same as
in rural residents, and therefore the selection of only urban
hospitals (i.e., located in SMSAs) does not introduce any
significant bias into the estimates of average concentrations
levels. Finally, it is assumed that no systematic bias is
introduced by the fact that the participating pathologists and
medical examiners were self-selected, and the specimens were non-
probabilistically sampled according to pre-specified quotas.
The FY82 sampling design specified the collection of 40
specimens from each of 35 SMSAs, five of which were double
collection sites. Double collection sites are SMSAs whose
populations are so large that their proper representation in the
sample requires that they be sampled twice. In a double
collection center, either one cooperator provides twice the
number of specimens (80) or two cooperators each provide the
standard quota of specimens (40). Sixteen hundred specimens were
designated for collection. However, due to incomplete
fulfillment of target quotas and no response from several medical
examiners/ pathologists, only 827 specimens were collected, from
26 SMSAs.
92
-------
FY82 NHATS Sampling Design
Stratify by
Geographical
Region
Nine Census Divisions
Select Probability
Sample of SMSAs
Within Each
Census Division
Select One or
More
Cooperators
From Each SMSA
Select Specimens
According to
Age, Race, and
Sex Quotas
Two to Seven SMSAs Within
Each Census Division
45 Cooperators in
Total Survey
1600 Specimens Requested
827 Specimens Collected
Figure 4-1. Overview of the FY82 NHATS Sampling Design
93 /
-------
4.2 Compositing Design
The HRGC/MS analytical method and associated protocol
required approximately 20 grams of tissue for each analysis
sample. The amount of tissue available for each NHATS specimen
ranged from two-tenths of a gram to thirty grams. Hence, it was
necessary to composite individual specimens prior to chemical
analysis. By requiring more tissue per analysis sample, the
protocol also reduced the level of detection.
The variables Census division, age group, sex, and race
group were selected as the design variables for the composite
design. These four variables led to a structure of 108 cells (9
x 3 x 2 x 2 » 108). However, a design with 108 cells was not
practical because of budgetary constraints and because of the
target of twenty grams per sample. Census division and age group
were chosen as the nesting variables. These variables created a
structure of twenty-seven cells. The individual specimens were
classified into these cells, and composites of specimens were
created within this cell structure. The percentages of specimens
in a composite from the sex and race groups were deliberately
designed to vary over the set of composite samples. The
variation in the sex and race percentages was designed to
facilitate the estimation of the effect of sex and race on
concentration levels.
The target for the composition of the composite samples
was one gram apiece from twenty individual specimens. However,
it was not possible to achieve the design in the preceding
paragraph, to use all available FY82 specimens, and still meet
the target of twenty specimens of one gram per composite.
Moreover, as noted above, some specimens had weights as small as
two-tenths of a gram. Therefore, the actual number of specimens
in a composite, the weight of the composite, and the equality of
individual specimen weights within a composite were allowed to
vary over the set of composites.
94
-------
A total of 763 specimens were assigned to the 46
composites dedicated to the semi-volatiles and dioxins and furans
analyses. A total of 697 specimens were assigned to the 46
composites dedicated to the volatiles analysis. The geographic
and demographic composition of the two specimen sets are shown in
Table 4-1. The characteristics of the composite samples for the
volatiles, semi-volatiles, and dioxins and furans analyses are
listed in Appendix D.
95
-------
Figure 4-2. NHATS FY82 collection map.
96
-------
Table 4-1. Geographic and Demographic Counts for Specimens
Category Volatiles Semi-volatiles, Requested
Specimens Dioxins and Furans Specimens
Specimens
Census Region
Northeast 161 166 320
North Central 188 206 480
South 295 331 520
West 53 60 280
Total 697 763 1,600
Age Group
0-14 years 134 178 463
15-44 years 301 312 662
45 + years 262 273 475
Total 697 763 1,600
Sex
Male 364 412 788
Female 333 351 812
Total 697 763 1,600
Race Group
White 590 632 1,420
Non-White 107 131 180
Total 697 .763 1,600
Note: The number of individual specimens comprising the composite
samples used in analysis of semi-volatiles, dioxins, and
furans was larger than the number of specimens in
composites samples used to analyze volatile organic
compounds. Since the composite samples used for analysis
of semi-volatiles were formed first, several of the
volatile organic composite samples contained fewer
specimens.
97
-------
5.0 SPECIMEN COLLECTION AND STORAGE
The 827 individual specimens collected for the Broad
Scan Analysis Study were obtained by medical examiners and
pathologists either during regularly scheduled surgical
procedures, that is, from tissue excised for therapeutic or
elective purposes, or as part of routine postmortem examinations.
If the specimen was collected postmortem, the tissue was obtained
from an unembalmed cadaver which had been dead for less than
twenty-four (24) hours and had been kept under refrigeration
during that time. The death should have been caused by sudden
traumatic injury, ,such as cardiac arrest, car accident, or
gunshot wound.
The following groups were excluded from data collection:
institutionalized individuals;
persons known to be occupationally exposed to toxic
chemicals;
persons who died of pesticide poisoning; and
persons suffering from cachexia.
These guidelines were stipulated so that the levels of the
substances detected in the specimens were a result of
environmental exposure.
All NHATS cooperators in the selected SMSAs were
provided with target quotas for specimen collection from age, sex
and race groups. The cooperators were asked to obtain at least
five grams of adipose tissue from each donor. The cooperators
were asked to guard against contamination through contact with
disinfectants, paraffins, plastics, preservatives, and solvents.
After collection, the adipose tissue specimens were placed in
glass jars frozen to -20* centigrade. These jars were packed on
dry ice in insulated containers for transport and delivery to the
Toxicant Analysis Center at Bay St. Louis, Mississippi. In
September, 1983, the frozen specimens were transferred to Midwest
Research Institute (MRI) in Kansas City, Missouri. At MRI, the
specimens were placed in freezers maintained at a temperature of
99
-------
-20* centigrade. The adipose tissue specimens were kept frozen
during the transfer from the Toxicant Analysis Center to the
Midwest Research Institute (USEPA 1986b).
All tissue specimens remaining after the completion of
the FT82 analytical effort are stored at the Midwest Research
Institute in Kansas City, Missouri. All remaining NHATS tissue
specimens from 1970 to the present are stored at the same
location. The specimens are kept frozen at a temperature of
approximately -20 degrees centigrade.
At the time the FY82 Broad Scan Analysis was conducted
and since that time, there has not been a comprehensive study to
evaluate the stability of volatile organics in fatty tissues.
Following the completion of the FY82 analytical effort, there
have been studies by EPA to assess the stability of
organochlorine pesticides and dioxins and furans in human adipose
tissue. The issue of storage and stability for the NHATS is not
fully resolved at this time.
100
-------
6.0 CHEMICAL ANALYSIS PROCEDURES
Midwest Research Institute (MRI) conducted the chemical
analysis of the FY82 composites. For the volatiles analysis, MRI
developed a "dynamic headspace purge and trap system" to extract
the volatile target compounds from the composite samples. The
target compounds were directed into a Finnigan 9610 gas
chromatograph and a Finnigan 4000 quadruple mass spectrometer for
analysis. Target volatile compounds were identified by the
response time of the primary characteristic ion relative to
either a deuterated analog of the target compound or to the
internal standard, 'bromochloropropane. The complete mass spectra
at the appropriate points in time were reviewed to confirm the
identification. The quantitation of the volatile compounds was
carried out by comparison of peak areas for the compounds to the
peak area for the associated deuterated counterpart, if one were
available, or to the peak area for bromochloropropane, the
internal standard for this analysis. (USEPA 1986b)
The deuterated compounds and bromochloropropane were
added to the system by a ten railliliter syringe. This syringe
was first filled with three milliliters of water free of volatile
organics. The deuterated compounds and bromochloropropane were
inserted into the ten milliliter syringe from a five micro liter
syringe. An additional two milliliters of water and one
milliliter of air were drawn into the ten milliliter syringe, and
the syringe was inverted several times to allow mixing. Finally,
the contents of the ten milliliter syringe were transferred to
the sample vessel in the "purge and trap" system. The sample
vessel was tightly capped and allowed to remain at room
temperature for thirty minutes before initiating the analysis.
Refer to Table 6-1 for the quantitation standard associated with
each of the target volatile compounds.
101
-------
Table 6-1.
Pairing of Target Analytes Versus Internal
Quantitation Standards for Volatile Organic
Compounds Analysis
Target Analyte
internal Quantitation
Standard
Benzene
Benzene
Substituted Benzenes
Styrene
Ethylphenol
Alkyl Benzenes
Toluene
Ethylbenzene
Xylene
Chlorinated Benzenes
Chlorobenzene
1,2-Dichlorobenzene
1,4-Dichlorobenzene
Trihalomethanes
Chloroform
Bromodichloromethane
Dibromochloromethane
Bromoform
Halocarbons
1,1,1-Trichloroethane
1,1,2-Trichloroethane
1,1,2,2-Tetrachloroethane
Tetrachloroethene
d^-Benzene
Bromochloropropane
d^-Ethylbenzene
dg-Toluene
du -Ethy Ibenz ene
ds-Chlorobenzene
d4-l , 4-Dichlorobenzene
d4-l , 4-Dichlorobenzene
d-Chloroform
Bromochloropropane
Bromochloropropane
Bromochloropropane
Bromochloropropane
Bromochloropropane
dj-1 , 1,2, 2-Tetra-
chloroethane
Bromochloropropane
102
-------
For the semi-volatiles analysis, the composites went
through an initial extraction step (see Figure 6-1). One percent
(1%) of this extract was set aside to determine the percentage of
extractable lipid tissue in the composite sample. The ninety-
nine percent (99%) of the extract left underwent a gel permeation
chromatographic step to separate the lipid tissue from the target
compounds. After the gel permeation step, ten percent (10%) of
the resulting extract was reserved for the dioxins and furans
analysis. The ninety percent (90%) aliquot of the extract was
partitioned through Florisil fractionation to create "fractions",
each with different sets of target semi-volatile compounds. A
sample from the fractions was injected to a Finnigan MAT 31 LA
double focusing magnetic sector mass spectrometer for analysis.
Target semi-volatile compounds were identified by the response
time of the primary characteristic ion relative to the internal
standard, anthracene-d1Q. The ratios of the peak areas of two
secondary ions to the peak area for the primary ion were computed
to further verify the identification. Review of mass spectra at
appropriate points in time was carried out to confirm
identification. Quantitation of the semi-volatile target
compounds was carried out by comparison of peak areas for the
compounds to the peak area for anthracene-d, Q, the internal
standard for this analysis. (USEPA 1986c)
For the dioxins and furans analysis, the fractions
obtained in the semi-volatiles analysis were recombined. These
fractions represented the ninety percent aliquot mentioned in the
preceding paragraph. The recombined fractions went through a
further "clean up" step. The ten percent aliquot was subject to
separate and different "clean up" step. The ninety percent
aliquot was earmarked for analysis for the tetra- and penta-
chloro dioxins and furans. The ten percent aliquot was earmarked
for analysis for the hexa-, hepta-, and octa- chloro dioxins and
furans. Although a number of exceptions were necessary, in
general the analysis was carried out according to the plan for
103
-------
FLOW CHART OF CHEMICAL ANALYSIS STEPS
FOR SEMI-VOLATILES, DIOXINS AND FURANS
CREATE COMPOSITE SAMPLE
ADD C«-» — LABELED TCDD
13
AND C13 - LABELED OCDD
AND SEMI-VOLATILES
SURROGATES
99%
DETERMINE PERCENTAGE OF
UPID EXTRACTABLE
GEL PERMEATION
CHROMATOGRAPHY
90X
FLORISIL FRACTIONATION
i
RESERVE FOR DIOXINS
AND FURANS ANALYSIS
ADD ANTHRACENE -
i
SEMI-VOLATILES GC/MS
ANALYSIS (SCANNING)
RECOMBINE FRACTIONS
CLEAN-UP FOR
DIOXINS AND FURANS
(AMOCO PX-21/GLASS FIBER)
i
CLEAN-UP FOR
DIOXINS AND FURANS
(CARBOPAK C/CEUTE)
TETRA- AND PENTA- DIOXINS
AND FURANS GC/MS ANALYSIS
(SELECTED ION MONITORING)
i
HEXA-. HEPTA-. AND OCTA-DIOXINS
AND FURANS GC/MS ANALYSIS
(SELECTED ION MONITORING)
Figure 6—1. Chemical analysis steps for semi—volatiles, dioxins
and furans.
104
-------
the ninety and ten percent aliquots. Analysis was performed by a
Kratos MS-50 double focusing mass spectrometer functioning in the
Selected Ion Monitoring phase. Analysis of ash from an
incinerator was used to determine the time frame during which
dioxin and furan ions were likely to appear. If characteristic
ions for dioxins and furans appeared in this time frame and
theoretical ion ratios were achieved within certain limits, a
dioxin or furan was identified. More specific identification of
dioxin and furan chemicals was carried out through comparison of
response times to selected internal standards. The guantitation
of the tetra- and penta- chloro dioxins and furans was
accomplished by comparison of peak areas of the target compounds
to the peak area for carbon-13 labeled 2,3,7,8-TCDD. The
guantitation of the hexa-, hepta-, and octa- compounds was
accomplished through comparison of peak areas of the target
compounds to the peak area for carbon-13 labeled OCDD. (USEPA
1986d)
105
-------
7.0 DATA PREPARATION AND MANAGEMENT
Data were reported on 46 composite samples for the
analysis of volatile organic compounds and 46 composite samples
for the analysis of semi-volatile organic, dioxin, and furan
compounds. Altogether, analyses were conducted on 57 compounds:
17 volatile organics, 30 semi-volatile organics, 5 dioxins and 5
furans.
Each concentration was reported in one of three
categories:
• Not Detected (ND) if the analytical instrument
response was below the limit of detection;
• Trace (TR) if the result was between the limit of
detection and the limit of quantitation; and
• Positive Quantifiable (PQ) if results were above the
limit of quantitation.
The limit of detection (LOD) is a threshold value below
which the presence or absence of a compound cannot be determined.
For this study the LOD was calculated as 2.5 times the estimated
average background signal. The limit of quantitation (LOQ) is a
threshold value below a detected compound cannot be accurately
quantified. The LOQ was calculated as four (4) times the LOD.
For positive quantifiable measurements, the actual concentration
was reported. For readings that were either trace or not
detected, the LOD was reported.
The reported concentrations were converted to the
following concentrations:
actual concentration, if positive quantifiable;
LOD ^ L°Q , if trace; and
—=— , if not detected.
107
-------
The data were stored in three Statistical Analysis
System (SAS) data bases, one each for volatiles, semi-volatiles,
and dioxins and furans on the National Computer Center IBM
computer system in Research Triangle Park, North Carolina. Model
fitting was performed using the Biomedical Statistical Software
System, BMDP (Dixon 1981). The program BMDP3V, Mixed Model
Analysis of Variance, was used to fit the model parameters.
Output from BMDP was then entered into SAS to generate variance
estimates.
108
-------
8.0 STATISTICAL ANALYSIS APPROACH
8.1 Selection and Development of the Statistical Model
Since one of the objectives of the Broad Scan Analysis
Study was to estimate average levels of toxic chemicals in the
general, non-institutionalized U.S. population, it was necessary
to derive information about population averages from the chemical
analyses of the composite samples. This required an assumption
relating the chemical level of a composite sample to the chemical
levels of the individual specimens which comprised the composite
sample. Average concentration levels for the demographic
subpopulations, as, well as for the nation, could then be
estimated.
In developing the compositing design, it was first
necessary to assume that the amount of a chemical in a composite
sample was equal to the sum of the amounts contributed by each of
the individual specimens that comprise the composite sample.
This assumption is quite sound provided the compositing procedure
does not result in any synergistic effect that chemically alters
the specimens. Second, a review of the composites indicated
that, in general, specimens in the same composite contributed
approximately the same weight of tissue to that composite.
Hence, it was assumed that the concentration of the composite
sample equaled the average of the concentrations of the
individual specimens that comprised it. Accordingly, the
concentration level of the composite sample, calculated by simply
dividing the total amount present by the total tissue mass of the
composite sample, was assumed to equal the average of the
concentrations of the individual specimens.
A statistical model was developed to permit data from
the chemical analyses of composite samples to be used to estimate
average chemical levels in the U.S. population and its various
geographic and demographic subpopulations. The model postulates
how the chemical level of a composite sample varies as a function
of the geographic and demographic characteristics of the
109
-------
individual specimens that comprise it. It assumes that the
chemical level of any individual specimen is a function of the
characteristics (i.e., Census region, age group, sex, and race)
of the specimen donor. This implies that the chemical level of a
composite sample is also a function of the geographic and
demographic characteristics of the specimens that comprise it.
The model is a multiplicative one. It assumes that the
effect of each geographic or demographic factor (e.g., Census
region) is to proportionally increase or decrease the expected
average concentration level of the composite sample. For
example, if specimens from the North East Census region have an
average concentration level that is ten percent higher than the
overall average level, composite samples from the North East
Census region will tend to have average concentration levels ten
percent higher than composite samples from other regions of the
country. The model further assumes that the standard deviation
of the measured concentrations increases with the mean (Snedecor
and Cochran 1967). This type of model is common for models used
in the analysis of data on toxic pollutants, where the
distribution of concentration levels is typically asymmetric or
skewed (Gilbert 1987).
Since geography and age were the primary factors of
interest to EPA, the composite design stipulated that individual
specimens be composited within Census division and age group
combinations. However, the effects of race and sex on average
concentration levels were still a concern. The compositing
design needed to provide information on these factors by
purposefully mixing individual specimens of both race groups and
sexes within a composite sample and varying the race and sex
proportions across the composite samples. The race and sex
makeups of the composite samples were either homogenous or mixed
depending on the availability of individual specimens.
110
-------
These concepts lead to the statistical model:
E(CiJk) - M CR, A,
(Equation 8-1)
where
. E(C;jk) is the expected average concentration level
for the distribution of all composite samples formed
from specimens collected from the ith Census region,
jth age group, and having race and sex proportions
given by R-,jk and S;jk, respectively, with:
Rjjk « proportion of white specimens minus the
, proportion of non-whites; and
S-,jk = proportion of male specimens minus the
proportion of females;
. M is the overall average effect of all demographic
factors ;
is the effect of the ith Census region;
, A; is the effect of the jth age group; and
. 3} and B2 are parameters which describe the
relationship between the chemical level and the
race and sex makeup of the composite sample.
The Census region, age group, sex, and race parameters
of the model indicate how the average concentration levels differ
across the various demographic subpopulations . The parameter
estimates are interpreted as follows: if CRj is greater than
one, composite samples formed from specimens collected from the
ith Census region will tend to have higher than average
concentration levels. If CRj is less than one, the concentration
levels will tend to be lower than average. The age parameter Aj
is interpreted similarly. For the race parameter, if Ql is
positive, the expected concentration level of the composite
sample will increase as the proportion of white specimens
increases. If Ql is negative, the expected concentration level
decreases as the proportion of whites decreases. The parameter
111
-------
B2 is interpreted similarly for sex. A positive value for Q2
indicates that higher expected concentration levels are
associated with males. .A negative value for 32 indicates that
lower concentration levels are associated with males or,
conversely, that higher levels are associated with females.
It should be noted that variation in the number of
specimens comprising the composites was not taken into account in
estimating the model parameters; each composite was given equal
weight in the computation. This does not create any bias in the
estimated parameters or average concentrations, but may entail a
loss of efficiency in the estimators. As a practical matter the
increase in estimator variance due to not taking this into
account is minimal, since the measurement error of a composite is
independent of the number of its constituent specimens, and this
is the dominant component of variance across composites.
Census region, rather than Census division, was used in
the statistical estimation analysis procedures. Although Census
division was used to specify the collection and compositing
procedures, the nine Census divisions were collapsed into the
four U.S. Census regions for statistical analysis purposes. This
reduced the number of subpopulations and hence the number of
model parameters that needed to be estimated. Originally 108
subpopulations (corresponding to the 9 Census divisions, 3 age
groups, 2 sexes and 2 race groups) were defined. Collapsing
resulted in a total of 48 target subpopulations. Only eight
model parameters were then needed to estimate the 48
subpopulations, since the model estimated subpopulation averages
without interaction effects. This was a reasonable number of
model parameters given that 46 composite samples were available
for each analysis set.
112
-------
8.2 Application of the Statistical Model
The expected value of the concentration level expressed
as a function of the effects of Census region, age, race, and sex
is given by Equation 8-1. The assumed error structure of this
model is given in Equation F-l of Appendix F, incorporating into
the model terms that explicitly reflect the variance components
attributable to the complex NHATS sample design as well as
measurement error.
The statistical model given in Equations 8-1 and F-l
is called a mixed model since it includes several factors whose
effects on the composite concentration levels are considered to
be random in addition to factors whose effects are considered to
be systematic rather than random. The BMDP program P3V (Dixon,
1981) can perform mixed model regression analysis and was used to
fit the Broad Scan analysis data. P3V uses maximum likelihood
estimation (MLE) techniques to fit linear models under the
assumption that the random factors are normally distributed. MLE
techniques were used in the Broad Scan analysis because they are
more flexible for fitting mixed models with unbalanced data.
Because the model in Equation F-l assumed that the
composite concentration levels have a lognormal distribution, the
parameters of the model were estimated by taking logarithms (base
e) on both sides of Equation F-l and then fitting the logarithms
of the measured concentration levels to the transformed model.
The log-concentration model met the assumptions of the P3V
analysis because the model was linear in the unknown parameters
and the assumptions of Equation F-l implied that the log-
concentrations are normally distributed.
Goodness-of-fit tests on the residuals from the fitted
model confirmed that the assumption of normality on the log-
concentrations was reasonable in 19 of 22 compounds analyzed. A
Chi-square goodness-of-fit test rejected the hypothesis of
lognormality at the .10 level for three compounds, xylene,
2,3,4,7,8-PECDF and OCDD. In tests of 22 compounds, by chance
113
-------
alone, it is expected that two compounds would have been rejected
at a .10 level. The distribution of values for 2,3,4,7,8-PECDF
had a significantly lighter right tail than a lognormal
distribution. The distributions of values for xylene and OCDD
had significantly heavier right tails than a lognormal
distribution. Since the mean estimates generated by the model
are fairly robust in the face of variability in distributional
form, the model was used with these compounds as well.
Results from the P3V analysis included maximum
likelihood estimates of all the parameters of the log-
concentration model including estimates of the variances for the
random factors log SMSA and log E. The output also included
standard errors for all estimated parameters and statistical
tests of significance.
Because the composite concentration levels were assumed
to be lognormally distributed, a maximum likelihood estimate of
the expected concentration level of any composite sample can be
calculated by
log M+log CR.+log Aj+V^jk'V^jk** (
E(Cijk> * e (8-2)
where each " denotes an MLE for the corresponding parameter.
Equation 8-2 was used to calculate an MLE of the
average concentration level for each of the 48 demographic
subpopulations defined by the 4 Census regions, 3 age groups, 2
race groups, and 2 sexes. Each MLE was obtained by substituting
into Equation 8-2 estimates for the corresponding Census region
and age group effects and setting the race proportion, Rj^fc/
equal to either +1 (to indicate an all white subpopulation) or -1
(all non-white subpopulation), and setting the sex proportion,
sijk' e
-------
individual specimens. This is because the expected value of the
average of random variables with identical means (in this case,
the concentration level of a composite formed entirely from
specimens from a single demographic subpopulation) is the same as
the expected value of the individual random variables (in this
case, the concentration levels of specimens from that demographic
subpopulation) .
8.3 Statistical Estimation of Average Concentration Levels for
the Entire Nation and Various Subpopulatxons
The estimated average concentration levels for the 48
target subpopulations were used to construct estimates of average
levels for other subpopulations of interest. Of particular
interest were the estimates for each Census region, age group,
race group, and sex, as well as estimates for the entire nation.
These estimates were calculated as weighted averages of the
individual 48 subpopulation estimates, where the weights were
proportional to the population of each target subpopulation. For
example, the estimated average concentration level for the ith
Census region was calculated as:
where (Equation 8-3)
WJJ|B is the population proportion for the jth age
group, 1th race group, and mth sex group relative
to the total ith Census Region, and
A
ft JJ|B is the estimated average concentration level
from Section 8.2 for the respective subpopulation.
This procedure involves summing the weighted average
concentration level estimates for each of the twelve
subpopulations within the Census Region. The national estimate
was obtained by summing the weighted estimates over all 48 target
subpopulations .
115
-------
8.4 Significance Testing of Differences Between Subpopulations
Significance testing was conducted to determine whether
any significant differences existed among the levels of any of
the geographic or demographic factors (i.e., Census region, age
group, race group, or sex) . The results of the model fitting and
parameter estimation were used to test the following hypotheses
concerning the parameters estimated in Equation 8-1:
Ho: CRi=l for all i=l,2,3,4 versus Ha: CRi^l, for some i
Ho: Aj-1, for all j=l,2,3 versus Ha: Aj#l, for some j
Ho: 8^0 versus Ha: 6^0
Ho: 82=0 versus Ha: 82*0
Each of these tests was conducted on the model parameters to
determine whether the corresponding factor had a significant
effect on concentration levels. The results of the significance
tests were presented in Table 2-4.
8.5 Detection and Exclusion of Outliers Among PECDD Measurements
The two largest measured values of PECDD in the sample
of composites were determined to be outliers and excluded from
the data used to estimate average concentration and test
hypotheses. The decision to declare them outliers was based both
on evidence of internal inconsistency with the rest of the
distribution, and the implausibility of their magnitude when
compared to external sources of PECDD concentration data.
First, the two largest PECDD concentration measurements
of 5300 and 5200 pg/g were each more than six times the next
highest value. For no other chemical detected in the study was
there an interval as great between the maximum values and the
rest of the distribution. The existence of such an extremely
large gap in a measurement distribution is an indication of the
probable presence of outliers.
116
-------
Second, for the other chemicals the arithmetic averages
of the composite measurements by age group were close to the
model-based, maximum-likelihood estimates, a result which would
be expected for large samples if the measurements were in fact
identically, lognormally distributed within age group. However,
approximate equality between the arithmetic averages and the
model estimates of concentration was not observed for PECDD when
the two largest values were included, but was when they were
excluded. This is further indication that these large values
were generated from a contaminant distribution, rather than the
distribution associated with the other PECDD data.
Finally, other data sources have reported average
concentrations of PECDD in the 20 pg/g range, with maximum
concentrations that are almost two orders of magnitude less than
the two suspect outliers. Since the Broad Scan composite
measurements are themselves averages of the concentrations of
their constituent specimens they should tend to cluster around
population means, and it is therefore extremely implausible that
they would have values as high as 5000 pg/g.
The effect on concentration estimates and hypothesis
tests due to excluding the two outliers is shown in Table 8.1.
8.6 Concentration Estimates and Hypothesis Tests for Total
Equivalent DDT
Levels of Total Equivalent DDT (TEDDT) can be estimated
from the concentrations of the congeners in the DDE, DDT, and DDD
families, through the following formula:
TEDDT » p,p'-DDT + o,p'-DDT
+ 1.114 (0,p'-DDE + p,p'-DDE + o,p'-DDD + p,p'-
DDD)
Of the six chemicals appearing in the formula only p,p'-DDT and
p,p'-DDE were found to have detectable levels in the analyzed
117
-------
Table 8-1. Comparison of Average Concentration Estimates
and Significance Test Results for 1,2,3,7,8-
PECDD Including, and Excluding outliers
Estimate wita TWO outliers*
Nation
Census Region
NE
NC
S
W
Age Group
0-14 yrs
15-44
45+
Race Group
White
Non-White
Sex
Male
Female
Included
Averaoe Concentration and Relative
190
(43)
420
(61)
170
(54)
100
(50)
130
(70)
200
(53)
290
(48)
44
(56)
210
(44)
91
(78)
140
(55)
240
(62)
Excluded
Standard Error2
75
(23)
120
(39)
62
(34)
60
(30)
73
(42)
54
(30)
130
(27)
11
(31)
83
(24)
39
(46)
100
(34)
49
(39)
Siqnif icance Test p-Value
Census region
Age
Race
Sex
.214
.001
.261
.505
.464
.000
.102
.230
1 Outl iers are composite IDs 82168 ind 82IS1
2
Concentration is in units of pg/g. Relative standard errors are shorn in parenth
118
-------
composites. If zero is substituted for the non-detected
congeners the following approximate formula results:
TEDDT = p,p'-DDT + 1.114(p/p'-DDE).
There are two possible ways to apply this formula to
the Broad Scan data to estimate the average concentration of
Total Equivalent DDT for the nation and for demographic
subgroups: (1) TEDDT is first estimated for each composite by
substituting the values of p,p'-DDE and p,p'-DDT into the
formula, and the resulting composite-level data are then analyzed
by the multiplicative model to yield average concentration
estimates for subgroups; (2) the average concentrations of p,pf-
DDE and p,p'-DDT are first estimated for each subgroup by the
multiplicative model/ and these subgroup estimates are then
substituted into the formula to yield average concentration
estimates of TEDDT.
Although the defining equation implies that the
population concentration of TEDDT must be greater than or equal
to the concentrations of p,p'-DDT and p,p'-DDE, the estimated
values produced by approach (1) may not satisfy this constraint
because the multiplicative model is a nonlinear function of the
composite measurements. Approach (2) has the desirable property
of producing estimates that always satisfy the constraint, and
for this reason was used to generate the average concentration
estimates of TEDDT shown in this report. The associated standard
error (SE) for the estimated average concentration of TEDDT was
estimated as:
SE(TEDDT) - SE(p,p'-DDT) + 1.114 SE(p,p'-DDE).
To test hypotheses on the significance of effects due to
census region, age group, sex, or race group on TEDDT levels, an
analysis of composite-level data was necessary, and for this
purpose individual composite estimates of TEDDT were computed by
substituting the concentrations of p,p'-DDT and p,p'-DDE into the
119
-------
formula for TEDDT. Hypothesis testing was then carried out for
the derived values of TEDDT in the same manner as was done for
the chemicals that were measured directly.
When TEDDT is estimated for composites, the problem
arises of what values to substitute in the formula when one or
both of the components p,p'-DDE and p,p'-DDT are not detected. A
comparison was made of the results obtained from the following
two alternative approaches: (a) if for a given composite either
compound is not detected, a value of zero is substituted into the
formula, and if the derived concentration of TEDDT is zero after
this substitution, it is replaced with an LOD/2 value of .005;
(b) if one or both of the components are not detected in a given
composite, the LOD/2 values associated with the non-detected
compounds are substituted into the formula. Since very little
difference was observed between these two approaches, method (b)
was adopted for this study because it was consistent with the
statistical treatment of the directly measured compounds.
Table 8.2 compares estimation and hypothesis test
results for the alternative ways of computing TEDDT.
120
-------
Table 8-2. Comparison of Average Concentration Estimates and
Significance Teat Results for Alternative Way? of
Computing Total Equivalent DDT (TEDDT)
""•'^•••"•••••"••^^"•••••"••"••••••"•"••••••^ TEDDT FormuTa""5ppTTecT>l^""""""
Composite Level Subgroup
Using Method1: Level
(a) (b)
Nation
Census Region
NE
NC
S
w
Age Group
0-14 yrs
15-44
45+
Race Group
White
Non-White
Sex
Male
Female
Census
Age
Race
Sex
Averaae Concentration and
1.4
(19)
1.2
(29)
0.74
(29)
1.9
(27)
1.4
(38)
0.7
(27)
1.2
(27)
2.0
(25)
1.5
(21)
0.85
(39)
1Q 1 •
. y • " , I v -:-
(30)
0.87
(36)
Siqnificance
0.159
0.010
0.256
0.152
Relative Standard Error
1.4
(19)
1.2
(29)
0.75
(29)
1.9
(27)
1.4
(38)
0.7
(27)
1.2
(27)
2.0
(25)
1.5
(21)
0.85
(39)
1.9
(30)
0.86
(36)
Test
0.166
0.010
0.254
0.141
1.6
(26)
1.4
(37)
0.87
(33)
2.4
(33)
1.7
(47)
0.98
(36)
1.7
(32)
2.1
(30)
1.7
(27)
1.1
(52)
2.4
(33)
0.93
(46)
p-Value
„.
—
—
Method (a) rap lieu non-detected p,p'-DDE or p.p'-DDT with zaro in tin foraula for TEDDT; if the resulting value
of TEDDT is zero it is replaced with the value .MS. Uethod (b) replaces non-detected p,p'-DOE and p.p'-DDT
with their LOD/2 values.
2 Lipjd-idjusted weight in parts per si 11 ion (pg/g). Relative standard errors are shown in parenthesis.
121
-------
8.7 Considerations in "the Use of the Broad Scan Analysis
Study Statistical Analysis Approach
There are two general classes of assumptions that have
been made in the statistical estimation approach used in the
Broad Scan Study: assumptions about the form of the statistical
model used in the analysis, and assumptions about the
characteristics of the sample of specimens and composites.
The first set of assumptions is embodied in Equations
8-1 and F-l, which describe the relationship between the
concentration level of a composite sample and the demographic
characteristics of the individual specimens which comprise it.
The model assumes the effect of each demographic factor was
multiplicative, rather than additive, that the factors acted
independently without interactions, and that the composite
concentration levels have a lognormal distribution. Orders of
magnitude differences between some of the composite concentration
levels suggested the use of such a model and normal probability
plots and goodness-of-fit tests on the log-concentration data
confirmed the reasonableness of this assumption. A model which
assumes that the composite concentration levels have a normal
distribution might have seemed appropriate since each composite
concentration level was essentially an average of concentration
levels of individual specimens. However, the measurement error,
which was likely a substantial component of the distribution of
measured concentration levels, was not averaged since each
composite was analyzed only once. Measurement error
distributions often tend to be skewed and could have accounted
for the fact that a lognormal model fit the composite
concentration level data better than a normal model. Note that
the standard procedure of replacing non-detects by LOD/2 and
trace observations by (LOD+LOQ)/2 (LOD is the analytical
procedure's limit of detection and LOQ is the limit of
quantisation), may create artificial non-lognormal distributions
of concentrations, if for example, all the LCDs were
122
-------
approximately equal. However, this seems not to have occurred,
and the statistical tests that were performed on the model
residuals did not reject the hypothesis of lognormality for 19 of
the 22 chemicals.
The model used in this report is a main effects model
with eight parameters. The model did not include interaction
effects because of budget limitations, because of the need to
make inferences for a large number of chemicals, and because the
number of model parameters would have exceeded the number of
observations. The discussion in Section 2.1.4 suggests
consideration be >given to a model which includes two-way
interactions. Including all two-way interactions would increase
the number of model parameters to twenty-five. The problem of
interactions is one that would require funding for development
and testing.
Several important assumptions have been made about the
sample of tissue specimens constituting the NHATS data. First,
because practical considerations dictated that tissue sampling
had to be limited to surgical patients and autopsied cadavers,
it was assumed that the average concentration levels in this
sampling population would be approximately equal to the average
concentration levels in the U.S. population. Second, it was
assumed that nonresponse due to incomplete fulfillment of planned
quotas by participating medical examiners and pathologists would
not significantly bias the results. Finally, it was assumed that
the concentration of a compound in a composite would be equal to
the average of the concentrations of the constituent specimens;
that is, the compounds retain their identities, and synergistic
effects do not occur between chemicals as a result of the
compositing. This assumption was necessary to justify equating
the estimated average concentration level of a homogeneous
composite population to the average concentration level of the
individual specimen population.
123
-------
It should be noted that if it were feasible to carry
out a probability sample survey to collect adipose tissue from
the general population EPA would prefer that approach. The NHATS
approach is a practical solution to the problem of obtaining
tissue samples. Its potential biases are mitigated by: 1)
selection of areas of the country by a probability mechanism; 2)
use of quotas based on population statistics for donor selection;
3) program preference for specimens from donors whose death was
sudden and unexpected; 4) use of population statistics to
weight average concentrations; and 5) emphasis on comparisons
across years, which of course is not possible in this report.
Several projects are currently underway to improve the
chemical and statistical approaches employed in the NHATS
program. An alternative statistical model has been developed
that represents a composite concentration as a linear function of
the demographic and geographic descriptors of the constituent
specimens. This additive model has been evaluated in comparison
to the multiplicative model and found to perform well. The
ultimate objective is to replace the multiplicative with the
additive model so that prevalence as well as average
concentration levels can be estimated from the composited NHATS
data. The results and status of this project are described in
the draft EPA report, "Statistical Methods for Analyzing NHATS
Composite Sample Data—Evaluation of Multiplicative and Additive
Model Methodologies."
Significant changes continue to be made in the chemical
analysis approach, especially in the methods of calibration,
quantitative procedures, qualitative identification, internal
standards, and use of spiked and unspiked samples. These
modifications have been implemented in the FT87 and FT86 surveys,
and are expected to provide demonstrated improvements in data
quality.
124
-------
Work is also underway to determine what effect the
change in chemical analysis methods had on the NHATS estimates of
average concentration levels. Composite samples from the F784
NHATS survey were analyzed by both the old chemical analysis
method—packed column gas chromatography/electron capture
detector (PGC/ECD)—and the new method—high resolution gas
chromatography/mass spectrometry (HRGC/MS)—to determine their
comparability. A report on this project is expected by the end
of the year.
125
-------
9.0 REFERENCES
Dixon WJ. 1981. Mixed model analysis of variance programs.
Biomedical Statistical Software. California: University of
California Press.
Gilbert RO. 1987. Statistical methods for environmental
pollution monitoring. New York: Van Nostrand Reinhold Company,
Inc.
Leczynski BA, Stockrahm J. 1985. Battelle Columbus Division.
An evaluation of hexachlorobenzene body burden levels in the
general US population-. Draft report. Washington, DC: Office of
Toxic Substances, U.S. Environmental Protection Agency. Contract
68-01-6721.
*
Lucas RM, Melroy DK, Immerman FW. 1983. Research Triangle
Institute. National adipose tissue survey statistical analysis
file. Draft final report. Washington, DC: Office of Pesticides
and Toxic Substances, U.S. Environmental Protection Agency.
Contract No. 68-01-5848.
McLafferty FW. 1980. Interpretation of Mass Spectra, (Third
Edition), Mill Valley, California: University Science Books.
Mack GA, Leczynski B, Chu A, Mohadjer L. 1984. Battelle
Columbus Division. Survey design for the national human adipose
tissue survey. Draft final report. Washington, DC: Office of
Pesticides and Toxic Substances, U.S. Environmental Protection
Agency. Contract No. 68-01-6721.
Mack GA, Panebianco DL. 1986. Battelle Columbus Division.
Statistical analysis of the FY82 NHATS broad scan analysis data.
Draft final report. Washington, DC: Office of Pesticides and
Toxic Substances, U.S. Environmental Protection Agency. Contract
NO. 68-02-4243.
Mack GA, Stanley J. 1984. Battelle Columbus Division, Midwest
Research Institute. Program strategy for the national human
adipose tissue survey. Final report. Washington, DC: Office of
Pesticides and Toxic Substances, U.S. Environmental Protection
Agency. Contract Nos. 68-01-6721 (BCD) and 68-02-3938 (MRI).
Orban J, Leczynski BA, Lordo R. 1987. Battelle Columbus
Division. Estimation of prevalence using composited samples.
Draft final report. Washington, DC: Office of Pesticides and
Toxic Substances, U.S. Environmental Protection Agency. Contract
No. 68-02-4243.
Patterson, D.G. et al. 1986. Human Adipose Data for 2,3,7,8,-
tetrachlordodibenzo-p-dioxin in certain U.S. samples.
Chemosphere, 15: 2055-2060.
127
-------
Public Law 94-469, Toxic Substances Control Act, Enacted by the
Senate and House of Representatives, October 11, 1976.
SAS Institute, Inc. 1985. SAS user's guide: basics and
statistics, version 5. North Carolina: SAS Institute, Inc.
Snedecor GW, Cochran WG. 1967. Statistical methods. Ames,
Iowa: The Iowa State University Press.
USEPA. 1980. U.S. Environmental Protection Agency. Mirex
residue levels in human adipose tissue: a statistical
evaluation. Washington, DC: Office of Toxic Substances, U.S.
Environmental Protection Agency. EPA 560/13-80-024.
USEPA. 1985. U.S. Environmental Protection Agency. Baseline
estimates and time trends for beta-benzene hexachloride,
hexachlorobenzene, and polychlorinated biphenyls in human adipose
tissue 1970-1983. Washington, DC: Office of Toxic Substances,
U.S. Environmental Protection Agency. EPA 560/5-85-025.
USEPA. 1986a. U.S. Environmental Protection Agency. Broad scan
analysis of human adipose tissue: volume I: executive summary.
Washington, DC: Office of Toxic Substances, USEPA. EPA 560/5-
86-035.
USEPA. 1986b. U.S. Environmental Protection Agency. Broad scan
analysis of human adipose tissue: volume II: volatile organic
compounds. Washington, DC: Office of Toxic Substances, USEPA.
EPA 560/5-86-036.
USEPA. 1986c. U.S. Environmental Protection Agency. Broad scan
analysis of human adipose tissue: volume III: semivolatile
organic compounds. Washington, DC: Office of Toxic Substances,
USEPA. EPA 560/5-86-037.
USEPA. 1986d. U.S. Environmental Protection Agency. Broad scan
analysis of human adipose tissue: volume IV: Polychlorinated
dibenzo-ja-dioxins (PCDDs) and polychlorinated dibenzofurans
(PCDFs). Washington, DC: Office of Toxic Substances, USEPA.
EPA 560/5-86-038.
USEPA. 1986e. U.S. Environmental Protection Agency. Broad scan
analysis of human adipose tissue: volume V: trace elements.
Washington, DC: Office of Toxic Substances, USEPA. EPA 560/5-
86-039.
USEPA. 1986f. U.S. Environmental Protection Agency. Exposure
Assessment for Hexachlorobenzene, Washington, DC: Office of
Toxic Substances, USEFA. EPA 560/5-86-019.
128
-------
APPENDIX A
STATISTICAL ESTIMATES
-------
Table A-l.
•eighted Eatiaatea (and Their Associated Stwidard Errors)
of tha Average Concontratioa Level* for the Entire Nation
and for Each Canada Region, A0a Group, Raca Croup, and Sei
Coa pound
Population Percentages
VOLATILE ORGANICS2
Benzene
Benzene
Substituted Benzene*
Styrene
Ethylphenol
Alkyl Benzene*
Toluene
Ethyl benzene
Xylene
Chlorinated Benzene*
Chlorobenzene
1,4-Dichlorobenzene
1NE • North East
NC - North Central
Entire
Nation
0.014 ,
(0.00 IB)1
0.096
(0.019)
0.086
(0.022)
0.046
(0.017)
0.077
(0.032)
0.30
(0.12)
0.0044
(0.0007)
0.12
(0.021)
S <> South
1 = lest
Census Region1
HE
22
0.018
(0.0038)
0.096
(0.036)
0.13
(0.061)
0.023
(0.012)
0.072
(0.062)
0.20
(0.14)
0.0033
(0.0010)
0.076
(0.024)
NC
26
0.010
(0.0020)
0.069
(0.022)
0.029
(0.013)
0.062
(0.029)
0.076
(0.064)
0.26
(0.18)
0.0026
(0.0006)
0.11
(0.032)
S
33
0.010
(0.0017)
0.10
(0.032)
0.090
(0.034)
0.046
(0.022)
0.10
(0.061)
0.49
(0.29)
0.0072
(0.0017)
0.20
(0.047)
19
0.019
(0.0060)
0.18
(0.064)
0.10
(0.063)
0.061
(0.030)
0.039
(0.029)
0.12
(0.084)
0.0030
(0.0010)
0.062
(0.020)
0-14 yra
23
0.016
(0.0029)
0.12
(8.028)
0.17
(0.068)
0.036
(0.017)
0.063
(0.029)
0.27
(0.12)
0.0038
(0.0009)
0.12
(0.033)
Age Groups
16-44 yrs
48
8.014
(0.0024)
0.10
(0.028)
0.066
(0.020)
0.066
(0.024)
0.090
(0.040)
0.33
(0.14)
0.0061
(0.0010)
0.13
(0.031)
46* yrs
31
0.012
(0.0020)
0.076
(0.017)
0.060
(0.016)
0.036
(0.017)
0.068
(0.030)
0.26
(0.11)
0.0037
(0.0008)
0.11
(0.027)
Race
Ihite
83
V
0.015
(0.0018)
0.096
(0.020)
0.079
(0.021)
0.063
(0.020)
0.078
(0.033)
0.31
(0.13)
0.0048
(0.0008)
0.11
(0.021)
Groups
Non-Ihite
17
0.0096
(0.0028)
0.10
(0.033)
0.12
(0.066)
0.013
(0.083)
0.070
(0.040)
0.23
(0.12)
0.0018
(0.0005)
0.19
(0.072)
Sex
Male
49
0.017
(0.0035)
0.14
(0.036)
0.096
(0.033)
0.080
(0.036)
0.11
(0.061)
0.43
(0.19)
0.0057
(0.0013)
0.13
(0.039)
Feaale
61
0.010
(0.0022)
0.060
(0.013)
0.076
(0.028)
0.014
(0.0066)
0.048
(0.023)
0.17
(0.077)
0.0032
(0.0008)
0.11
(0.035)
^Volatile average concentrations are expressed in mt wight in parts per ail I ion (pg/g).
^Standard error expressed in the saee units as the average concentration.
-------
Table K-l. (continu^)
Compound
Population Percentages
VOLATILE OMANICS1
Tribe loMthanoa
Chlorofora
HalocufcoM
Tetrach loroethene
SBO-VOUTILE OMANICS4
PCBe
Total PCBe
Organochlorine Pesticides
Beta-BHC
MM.
To til DDT
Phthalatee
Butyl benzyl
phthalato
Entire
Nation
0.047
(0.020)
0.027
(0.0070)
0.33
(0.078)
0.10
(0.020)
1.3
(0.30)
1.8
(0.41)
0.30
(0.10)
Census Region
NE
22
0.021
(0.014)
0.041
(0.010)
0.31
(0.18)
0.18
(0.047)
1.1
(0.38)
1.4
(0.62)
0.11
(0.083)
NC
26
0.041
(0.028)
0.044
(0.010)
0.28
(0.086)
0.11
(0.023)
0.73
(0.23)
0.87
(0.29)
0.46
(0.31)
S
33
0.040
(0.027)
0.018
(0.0068)
0.61
(0.17)
0.31
(0.068)
1.0
(0.80)
2.4
(0.79)
0.82
(0.41)
I
10
0.081
(0.068)
0.0086
(0.0046)
0.20
(0.11)
0.007
(0.032)
1.3
(0.80)
1.7
(0.80)
0.21
(0.18)
0-14 yrs
23
0.063
(0.027)
0.017
(0.0084)
0.071
(0.024)
0.071
(0.016)
0.76 .
(0.24)
0.98
(0.36)
0.48
(0.81)
Age Groups
16-44 yre
48
0.063
(0.026)
0.030
(0.011)
0 30
(0.003)
0.17
(0.033)
1.3
(0.30)
1.7
(0.64)
6.81
(0-18)
Race Groups
4t» yre
81
8.038
(0.010)
8.031
(0.011)
•
0.67
(0.17)
0.31
(8.066)
1.8
(0.62)
2.1
(0.64)
0.46
(0.28)
Ihite
83
0.048
(0.020)
0.020
(0.0087)
0.82
(0.084)
0.21
(0.086)
1.4
(0.36)
1.7
(0.46)
0.46
(0.23)
Non-lhite
17
0.082
(0.036)
0.010
(0.0008)
0.41
(0.10)
0.088
(0.026)
0.73
(0.32)
1.1
(0.67)
0.086
(0.080)
Sex
Hale
40
0.081
(0.030)
0.044
(0.016)
0.36
(0.12)
0.10
(0.044)
2.0
(0.86)
2.4
(0.70)
0.64
(0.38)
Feaale
61
0.014
(0.0076)
0.011
(0.0044)
0.32
(0.13)
0.10
(0.066)
0.64
(0.26)
0.03
(0.43)
0.24
(0.17)
'Volatile average concentratione ire expressed in vet night inparte per ail I ion (pg/g).
Seai-volatilo average concent rations are lipid id jus ted Mights expressed in parte per ail I ion (pg/0)-
-------
TabUA-1. (continued)
Coepound
Population Percentages
NOONS*
2,3,7,8-TCOO
1,2,8,7,8-PQCDD
HXCDD
51,2,3,4,7,8,9-
HPCDO
OCOO
FURN6*
2,3,4,7,6-PBCDF '
HXCOF
1,2,3,4,6,7,8-
HPCOF
Entire
Nation
6.1
(0.78)
76
(17)
120
(24)
140
(27)
820
(100)
40
(6.7)
24
(».7)
21
(2.6)
Census Region
HE
22
6.6
(1.6)
120
(46)
160
(60)
160
(62)
760
(180)
49
(16)
20
(6.0)
18
(4.3)
NC
26
7.1
(1.6)
62
(21)
110
(30)
180
(61)
920
(190)
38
(9.8)
29
(6.3)
26
(6.4)
S
33
6.1
(1.2)
60
(18)
100
(26)
110
(27)
780
(140)
30
(7.1)
24
(4.7)
22
(3.8)
1
19
4.1
(1.2)
73
(30)
120
(47)
100
(41)
860
(260)
62
(18)
23
(6.6)
15
(4.3)
0-14 yrs
23
4.1
(0.80)
64
(16)
92
(27)
89
(26)
410
(86)
36
(8.9)
16
(4.1)
19
(4.0)
Age Qroups
16-44 yrs
46
7.8
(1.8)
130
(»4)
120
(32)
160
(38)
920
(170)
63
(12)
27
(6.1)
22
(4.0)
Race Qroups
46* yrs
31
6.0
(0.98)
11
(3.4)
130
(33)
160
(39)
990
(180)
26
(6.0)
26
(6.0)
20
(3.6)
In its
83
6.4
(0.90)
83
(20)
120
(26)
140
(28)
810
(110)
44
(7.8)
26
(4.1)
20
(2.8)
Hon-lhite
17
4.8
(1.3)
39
(18)
110
(47)
140
(60)
880
(280)
22
(8.6)
18
(6.6)
26
(7.6)
Hale
49
8.7
(1.4)
100
(36)
70
(19)
89
(24)
760
(160)
61
(14)
13
(2.7)
16
(3.3)
Sex
Feaale
61
6.6
(1.3)
49
(19)
160
(60)
180
(68)
880
(210)
30
(9.3)
36
(7.9)
26
(6.8)
*Dioiin and furan avenge concentration* are lipid adjusted weights expressed in parta per trillion (pg/g).
-------
APPENDIX B
PERCENTAGE DETECTED DATA
-------
Table B-l. Volatile Organic Chemicals Identified in the
Broad Scan Analysis Study
Class Cheaical
Nuabar of
Coaposite
Sup I*
CAS Nuaber Mesaureaents*
Percentage
Detected
Benzene
BMZMM
71-43-2
48
96
Substituted Benzenes Styrene
Ethylphenol
1M-42-5
26429-37-2
46
46
1M
101
A Iky I Benzenes
Toluene
Ethylbenzene
Xylene
188-88-3
10f-41-4
1330-20-7
46
46
46
93
96
1M
Chlorinated Benzenes
Chlorobenzene
1,2-Dichlorobenzene
1,4-DichIorobenzeno
ll8-9f-7
95-61-1
106-48-7
46
4ft
46
96
83
100
Trihaloaethines
Halocarbons
Chlorofora
Brosod i ch 1 oroaothane
Dibroaochloroaathana
Broaofora
1, 1, 1-Tr ichloroathana
1,1,2-Trichloroethane
1, 1.2, 2-Tatrsch loroethana
Tatrsch loroathene
67-86-3
76-27-4
124-48-1
75-26-2
71-66-6
79-00-6
79-34-6
127-18-4
46
46
46
46
48
46
46
48
78
0
0
0
46
0
9
61
• The nuabar of eoapoaita aaaples aassuraaants (out of a possible 46) asy vary par coapound due to
ehaaieal probleas in retrieving or suaaarizing the data.
137
-------
Table B-2.
Semi-Volatile Organic Chemicals Identified in
the Broad Scan Analysis Study
Class
PCBs
Organochlorine
Pesticides
Arosatics
Chlorinated Benzenes
Phthalatos
Phosphates
Chesicsl
PCBs
Trichlorobiphenyl
Tetrach lorobi pheny 1
Pentach lorobi pheny 1
Hexach 1 orob i pheny 1
Heptach lorobi pheny 1
Detach lorobi pheny 1
Nonsch lorobi pheny 1
Decachlorobiphenyl
Beta-BHC
E,E'-ME
g,g'-OOT
Him
trana-Honachlor
Heptach lor Epoxide
Dieldrin
Naphthalene
Phenanthrene
Pyrene
1,2-Oichlorobenzene
1,2,4-Trichlorobenzene
Pentach lorobeozens
Kexach 1 orobenzene
Diethyl Phthalate
Di-n-butyl Phthalate
Diethyl Hexyl Phthalate
Butyl Benzyl Phthalate
Tri pheny 1 Phosphate
Tributyl Phosphate
Tris (2-Chloroethyl) Phosphate
CAS Huaber
1336-36-3
25323-68-6
26914-33-f
25429-29-2
26M1-64-9
28655-71-2
31472-83-1
53742-17-7
2151-24-3
319-86-7
72-65-9
51-29-3
2386-86-5
39766-81-5
1124-57-3
61-57-1
91-21-3
as-fi-a
129-M-l
96-61-1
121-82-1
618-93-6
118-74-1
84-66-2
84-74-2
117-81-7
85-68-7
116-86-6
128-73-8
116-96-8
Umber of
Coaposite
Saiple
Utaaureaanta*
44
44
44
44
44
44
44
44
44
43
45
37
43
42
43
43
43
43
43
43
46
44
43
42
42
42
42
42
43
48
Percentage
Detected
86
23
65
73
75
52
41
14
7
93
IN
68
14
57
71
33
42
14
§
12
4
I
79
48
51
33
74
38
2
2
* The nuaber of coaposite seeplee eeasuresenta (out of a possible 46} say vary per coapound due to
cheaical problem in retrieving or sussarizing the data. Several composite saaples were not
analyzed due to the unavailability of sufficient tissue
138
-------
Table B-3.
Dioxins and Furans Identified in the
Broad Scan Analysis Study
Class
Chwieil
CAS Nuibar
Nuiber of
Composite
SupU
Measurements*
Percentage
Detected
Oiexirw
Furaiw
2,3,7,6-TCDD
1,2,3,7,8-PBCDO
HXCOD
1,2,3,4,7,8,9-HPCOD
ocoo
2,3,7,8-TCOF
2,3,4,7,8-PBCOF
HXCOF
1,2,3,4,8.7,8-HPCDF
OCOF
1746-11-6
41321-76-4
34465-46-8
36022-46-0
3268-67-9
51207-31-9
67117-31-4
55664-94-1
67662-39-4
39Ml-fl2-fl
43
43
46
45
45
43
43
45
46
46
74
93
98
96
1M
26
88
71
93
41
The nueber of cooposite seep lee eeasurosonts (out of a poeeible 46) say vary psr eospeund due to cheeical
probleu in retrieving or susnrizing the data. Several composite aasples were not analyzed due to the
unavailability of sufficient tissue sass. Additionally, a sessurosent for several coeposite ssaples could not
be calculated for soee dioxins and furans due to low response observed for the internal standard used in the
analysis procedures.
139
-------
APPENDIX C
FY82 NHATS SAMPLING DESIGN SMSAs
-------
Table C-l. SMSAs Selected for the FY82 NHATS Sample
Census Division
SMSA
New England
Middle Atlantic
South Atlantic
East South Central
East North Central
West North Central
West South Central
Mountain
Springfield, MA
Boston, MA
Albany, NY
New York, NY (2)
Binghamton, NY
Philadelphia, PA
Pittsburgh, PA
Washington, DC
Norfolk, VA
Orlando,PL
Ft. Lauderdale, FL
Greenville, SC
Miami, FL
Memphis, TN
Lexington, KY
Birmingham, AL
Detroit, MI (2)
Cleveland, OH
Dayton, OH
Akron, OH
Chicago, IL (2)
Madison, Wl
Moline, IL
Rochester, MN
Omaha, NE (2)
Lubbock, TX
El Paso, TX
San Antonio, TX
Dallas, TX
Salt Lake City, UT
Denver, CO
143
-------
Table C-l. (Continued)
Census Division SMSA
Pacific Sacramento, CA
Los Angeles, CA (2)
Portland, OR
Spokane, WA
(2) Indicates a double collection site. A double collection
site is an SMSA whose population relative to its Census
Division population is so large that its proper
representation in the sample required it to be selected
twice.
144
-------
APPENDIX O
BROAD SCAN ANALYSIS STUDY COMPOSITING DESIGN
-------
Table D-l.
Demographic Characteristics for Each Broad Scan
Analysis Study Composite Sample - Volatile Analysis
Census
Region*
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NE
NE
NE
NE
NE
NE
NE
NE
NE
S
s
S
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
* NC
NE
S
i
Census
Division**
BK
ENC
BK
BK
BK
BK
BK
INC
INC
•NC
•NC
UA
UA
HA
UA
UA
UA
NE
NE
NE
ESC
ESC
esc
esc
esc
SA
SA
SA
SA
SA
SA
SA
SA
SA
SA
vsc
ISC
ISC
ISC
in
in
in
PA
PA
PA
« North Central
» North East
• South
• Vest
Age Group***
1
1
2
2
3
3
3
1
2
.,3
3
1
1
2
2
3
3
1
2
3
1
2
2
3
3
1
1
2
2
2
2
3
3
3
3
1
2
2
3
1
2
3
1
2
3
*• ENC
INC
UA
NE
esc
SA
ffSC
in
PA
Coeposit* N
Ninber S
1
2
1
2
1
2
3
1
1
1
2
1
2
1
2
1
2
1
1
1
1
1
2
1
2
1
2
1
2
3
4
1
2
3
4
1
1
2
North Central
West North Central
Uiddle Atlsntic
New 019 land
Esst South Central
South Atlmtie
Vest South Central
Uountain
Pacific
iwber of
MCIMIUI
12
16
19
19
18
16
13
10
17
IS
15
7
12
24
22
20
20
16
21
19
25
16
17
17
9
12
11
24
22
13
12
21
21
7
3
6
19
IB
23
2
12
10
6
8
16
Percent
•hit*
83.3
75.1
94.7
89.6
94.4
93.8
100.0
90.0
100.0
93.3
93.3
100.0
83.3
83.3
86.4
96.0
86.0
87.5
95.2
100.0
88.0
87.5
94.1
100.0
0.0
100.0
0.0
100.0
100.0
0.0
0.0
100.0
100.0
0.0
0.0
100.0
78.9
83.3
87.0
100.0
100.0
100.0
83.3
100.0
80.0
*•• 1 * 0-14
2 a 15-44
3 = 46- y
Percent
Uale
100.0
0.0
47.4
42.1
55.6
82.5
46.2
50.0
64.7
60.0
53.3
71.4
58.3
50.0
54.6
50.0
40.0
56.3
57.1
47.4
62.0
100.0
0.0
52.9
56.6
60.0
63.6
100.0
0.0
100.0
8.3
100.0
0.0
100.0
0.0
40.0
52.6
50.0
43.5
60.0
58.3
70.0
66.7
60.0
46.7
years
years
ears
Tissue
Mass (g)
12.7
17.3
20.8
21.1
18.6
22.6
21.4
18.9
21.8
21.6
18.3
20.3
18.1
25.0
25.3
16.3
17.8
20.0
23.8
25.6
26.6
19.0
24.3
20.6
19.3
12.6
16.7
22.8
18.7
10.1
17.8
16.4
23.2
13.8
11.8
6.0
22.4
21.9
22.0
5.1
18.8
22.4
15.0
17.4
20.7
1
Census Division, Age Group and Composite Nuaber uniquely identify each coaposite.
147
-------
Table D-2.
Demographic Characteristics for Each Broad Scan
Analysis Study Composite Sample - Semi-Volatile and
Dioxin and Furan Analyses
Census
Region*
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NC
NE
NE
IE
NE
NE
NE
NE
NE
NE
S
s
S
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
* NC
NE
S
•
Census
Division**
BtC
BK
BtC
BK
ENC
BK
BK
BK
•NC
•NC
VNC
•NC
HA
HA
UA
UA
UA
HA
NE
NE
NE
esc
esc
ESC
ESC
esc
SA
SA
SA
SA
SA
SA
SA
SA
SA
SA
•sc
ISC
•sc
•sc
HO
in
HO
PA
PA
PA
• North Central
• North East
» South
« VMt
Coaposito Nusbor of
Ago Group***
1
1
2
2
2
3
3
3
1
2
3
3
1
1
2
2
3
3
1
2
3
1
2
2
3
3
1
1
2
2
2
2
3
3
3
3
1
2
2
3
1
2
3
1
2
3
«• BK
me
UA
NE
esc
SA
•sc
uo
PA
Nuabor Specimens
1
2
1
2
3
1
2
3
1
1
1
2
1
2
1
2
1
2
1
1
1
1
1
2
1
2
1
2
1
2
3
4
1
2
3
4
1
1
2
1
1
1
1
1
1
1
North Central
•eat North Central
Middle Atlantic
NOT DlQ 1 *lfM
East South Central
South Atlantic
Vert South Central
Uountain
Pacific
18
21
20
19
19
18
18
14
13
17
16
15
11
13
24
22
2f
20
16
21
19
28
16
17
17
9
21
14
28
22
19
12
25
21
9
6
13
19
18
23
7
12
11
7
9
16
Pcrcwifc
•hite
77.8
81.1
M.I
89.6
89.5
94.4
88.9
1H.I
92.3
101.0
93.3
93.3
90.9
76.9
83.3
86.4
96.0
86.0
87.5
96.2
100.1
88.5
87.5
94.1
100.0
0.1
100.0
0.0
100.0
100.1
0.0
0.0
100.0
1N.0
0.0
0.0
69.2
78.9
83.3
87.0
85.7
100.0
100.0
85.7
88.9
80.0
PsrcMv
Hale
100.0
0.0
50.0
42.61
62.6
65.6
66.6
50.0
53.8
64.7
60.0
53.3
81.8
61.5
50.0
54.6
50.0
40.0
58.3
67.1
47.4
53.8
100.9
0.0
62.9
66.6
46.0
64.3
100.0
0.0
100.0
0.0
1M.0
0.0
100.0
0.0
63.8
62.6
60.0
43.5
71.4
68.3
70.0
71.4
55.6
46.7
Tissue
Haas (g)
18.1
21.2
21.6
21.4
20.2
19.8
28.2
23.2
23.4
29.6
22.6
21.4
23.0
20.2
25.2
26.1
16.2
18.0
19.1
21.9
26.7
28.1
19.9
26.7
20.7
21.1
20.7
19.1
26.4
19.5
17.9
18.2
20.0
26.1
18.0
17.6
11.1
22.7
21.9
22.4
9.0
18.3
21.0
19.7
21.6
22.0
**• i * 0-14 years
2 • 16-44 years
3-46-
years
Census Division, Ags Group and Coeposite Nuaber uniquely identify each coeposite.
148
-------
APPENDIX B
GLOSSARY OF TERMS
-------
BMDP Biomedical Statistical Software System
ECO Electron Capture Detection
EED Exposure Evaluation Division
FT82 Fiscal Tear 1982
GPC Gel Permeation Chromatography
HPCDD Heptachlorodibenzo-para-dioxin
HPCDF Heptachlorodibenzofuran
HRCG High Resolution Gas Chromatography
HXCDD Hexachlorodibenzo-para-dioxin
HXCDF Hexachlorodibenzofuran
LOD > Limit of Detection
LOQ Limit of Quantification
MS Mass Spectrometry
NCC National Computer Center
NHATS National Human Adipose Tissue Survey
NHMP National Human Monitoring Program
OCDD ^ Octachlorodibenzo-para-dioxin
OCDF Octachlorodibenzofuran
OTS Office of Toxic Substances
PCS Polychlorinated Biphenyls
PCDD Polychlorinated dibenzo-para-dioxin
PCDF Polychlorinated dibenzofuran
PECDD Pentachlorodibenzo-para-dioxin
PECDF Pentachlorodibenzofuran
PGC Packed Column Gas Chromatography
SAS Statistical Analysis System
SIM Selected Ion Monitoring
SMSA Standard Metropolitan Statistical Area
TCDD Tectrachlorodibenzo-para-dioxin
TCDF Tetrachlorodibenzofuran
TSCA Toxic Substances Control Act
151
-------
APPENDIX F
STATISTICAL ANALYSIS METHODOLOGY
-------
Statistical Model
The statistical analysis assumed that the chemical
concentration level of each composite can be expressed as
C..k - M • CR, ' Aj * exp(8lR,jk) • exp
. C|jk is the concentration level for the kth composite
from the jth age group and ith census region;
, M is a constant;
. CR, is the effect of the ith Census Region
. AJ is the effect of the jth age group;
. R,,k equals the proportion of white specimens in the
composite minus the proportion of non-whites;
. S,jk equals the proportion of male specimens in the
composite minus the proportion of females;
. BI and 8t are parameters which describe the
relationship between the chemical level and the race
and sex makeup of the composite sample;
. SMSA|:k is a random variable representing the random
effect due to the cluster of SMSAs which contributed
specimens to the composite. SMSA was assumed to
have a lognormal distribution where
In SMSA..k M N (0,
-------
The model assumes that a composite's concentration level is
systematically affected by the demographic characteristics of the
donors which contributed specimens to the composite. The model
further assumes that the concentration level is also randomly
affected by unknown exposure factors unique to the SMSAs and
individuals which contributed to the composite.
Estimation Approach
Let /tjjii be the population average concentration level
for the ith census region, jth age group, 1th race and mth sex.
Our goal was to estimate the average concentration levels for the
entire nation and various subpopulations, respectively defined by
432
£ £ £
i-1 j»l 1-1 m-1
- £ £ £ £ W..|B . /t|j|B (entire nation)
322
£ £ £ Wm . ,..,
-------
where each W,j(l is a weight proportional to the population census
count for the (i,j,i,«) subpopulation .
Estimates of the national and means estimates were
obtained by substituting estimated values of /*JJIB'S into each of
the equations. The approach used to estimate each /i^, is as
follows .
The statistical model in (F-l) implies that
log C..r N(log M+log CR.+log Aj+V^VV^jk' ffZ^+ff\) (F-2)
*
and therefore,
' (F-3)
E(Cijk)
When a composite consists only of specimens from the same
subpopulation (i,j ,!,•)/ the expected concentration level of the
composite is equivalent to the expected concentration level of an
individual specimen. That is
This follows from the statistical result that when Xlf * ' * , Xn
are random variables with identical means,
Xx + ••• -»• X.
- E ( - ).
A maximum likelihood estimate for each p was therefore
obtained from (F-3) by the equation
log M+log CRj+log Aj+8li(±l)-«-82i(±l)+i (^2sysA+^2E> '
E(CiJk) = e (P-4)
where the denotes an MLE and the ± 1 indicates that R,jk and
157
-------
Sjjk were set equal to either +1 or -1 to correspond to a pure
race (all white or all non-white) and pure sex (all male or all
female) composite.
The BMDP program P3V was used to obtain the MLEs given
in (F-4) by fitting the linear model in Equation F-2 to log-
concentrations of the composites. P3V also yielded the variance-
covariance matrix of the estimated parameters in Equation F-4.
A
The standard error of E(Cjjk) was obtained by substituting the
elements of the variance-covariance matrix into a Taylor Series
linearization of (F-4), as described below.
Define the random variables X and 7 for a fixed set of i, j,
and k, as follows
X- log M + log CRj_ + log Aj + 81(±1) + 82(±1)
A o
Y* **a SMSA * ° E>
A
Then the estimator E(Cjjk) in (F-4) can be written as
E(Cijk)-eXeY. (F-5)
X Y
Expand e and e in Taylor Series about the means Xo»EX and
Yo»EY, respectively, giving the linear approximations
ex-ex° + ex°(X-Xo)
eY-eYo + eYo(Y-Yo)
Substitute these formulas into equation (F-5) to give the
following approximation to E(C|jk)
E(Cjjk)-eX°eYo[l+(X-Xo)+(Y-Yo) + (X-Xo) (Y-Yo) ]
which can be written as
158
-------
E (C, jh)-eXoeYo-eXoeYo[(X-Xo) + (Y-Yo) + (X-Xo)(Y-Yo)] (F-6)
Square both aides of (F-6) and take expectations, noting that
because the MLEs of X and Y are independent, the expectation of
the cross-product terms on the right-hand side is zero. This
yields the following approximation to the variance of E(Cijk)
Var(E(Cjjk))=e2Xoe2Y°(Var(X)+Var(Y)+Var(X)Var(Y)) (F-7)
Finally, substitute estimates generated from the BMDP program P3V
for the unknown means, Xo and To, and variances, Var(X) and
Var(Y), to give a numerical approximation for the variance and
A
standard error of the estimated average concentration E(C;jk).
159
-------
DOCUMENTATION
MAC
EPA 560/5-90-001
S. NM00M'* ^kCCOMion Mo.
o. rmoMM »uotnn
NHATS Broad Scan Analysis: Population Estimates
From Fiscal Year 1982 Specimens
S. Booort MM
October, 1989
7. AwttMrtti
Alan anger, Gregory A. Hack
Battelle
Columbus Division
SOS King Avenue
Columbus, Ohio 43201
10. PlU|OU/To>»/Wom Unit No.
1L ContraeHO or QrwitfO) No.
<0 68-02-4294
(0)
u.
U.S. Environmental Protection Agency
Office of Toxic Substances
Exposure Evaluation Division (TS-798)
401 M Street, SW, Washington, DC 20460
13. Typo of ftonott 4 ••nod Covered
peer-reviewed report
is.
(Urn* 200«Mf«tt
Human adipose specimens were collected for the fiscal year 1982 National Human
Adipose Tissue Survey (NHATS). The specimens were combined into composite
samples, which were chemically analyzed for the presence and level of a number
of potentially toxic chemicals. The chemical classes monitored were:
volatile organic compounds, semi-volatile organic compounds, and dioxins and
furans. Average concentrations of chemicals in the human adipose tissue of
the general U.S. population are estimated. The estimation technique is
maximum likelihood. Comparisons are made between Census regions/ age groups,
sex groups, and two race groups.
17.
human adipose tissue* composite samples, toxic chemical monitoring, volatile
organics, seal-volatile organics, dioxins and furans, PCBs, maximum likelihood
estimation, lognormal distribution
NHATS, National Human Adipose Tissue Survey
Unclassified
a.
160
I Z7Z («—»''
------- |