INSTITUTE
Draft Report
Statistical Analysis of Mirex Special Study Data
June 3, 1980
by
Carol Leininger
Donna Lucas
Charles Sparacino
Stephen Williams
Prepared for
Task Manager:
Project Officer:
Cindy Stroup
Dr. Joseph Carra
Design and Development Branch
Survey and Analysis Division
Office of Pesticides and Toxic Substances
U. S. Environmental Protection Agency
401 M Street, SW
Washington, B.C. 20460
Under EPA Contract Number 68-01-5848
DRAFT CON
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27709
-------
-ii-
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
1. INTRODUCTION 1
1.1 Foreword 2
1.2 Summary of Study 2
1.3 RTI Objectives and Findings 3
1.3.1 Objectives 3
1.3.2 Findings 3
2. THE NATIONAL HUMAN MONITORING PROGRAM'S SPECIAL STUDY OF MIREX. . 5
2.1 Detailed Description of Study 5
2.1.1 Introduction 5
2.1.2 Sample Design 5
2.2 National Human Monitoring Program 7
2.2.1 Introduction 7
2.2.2 Strata 7
2.2.3 Eligible Places 7
2.2.4 Pathologists 9
2.2.5 Quotas 9
2.3 Selection Weight Construction 10
3. EVALUATION OF CHEMICAL ANALYSIS 15
3.1 Introduction 15
3.2 Discussion 15
4. UNWEIGHTED ANALYSIS 28
4.1 Data Description 28
4.2 Data Analysis 31
4.2.1 Data Distribution 31
4.2.2 Proportions of Positive Values 35
4.2.3 Quantifiable Positive Values 44
4.2.4 Unweighted Analysis Conclusions 47
5. WEIGHTED ANALYSIS 48
5.1 Introduction 48
5.2 Weight Formation 48
5.3 Estimates of Positive Mirex Detection 49
5.4 Estimates of Quantifiable Positive Values 51
5.5 Weighted Analysis Conclusions 56
-------
-iii-
TABLE OF CONTENTS (Coat.)
Page
REFERENCES 60
APPENDIX A: State Maps Showing Counties Where Mirex was Applied at
Some Time During 1965-1974 A-l
APPENDIX B: Counties Where Mirex was Applied at Some Time During
1965-1974 B-l
APPENDIX C: Definition of Mirex Treated Areas C-l
APPENDIX D: MSS Sample Selection Method D-l
APPENDIX E: List of MSS Collection Sites E-l
APPENDIX F: Instructions to Pathologists F-l
APPENDIX G: Quotas for Mirex Special Study Specimens G-l
APPENDIX H: Sample Mirex Chromatograms Made Available to RTI. . . . H-l
APPENDIX I: Modification of Mirex Analysis Protocol - GC Column
Temperature 1-1
APPENDIX J: Confirmation Analyses of Mirex in Human Adipose Tissue
Extracts J-l
-------
-iii-
LIST OF TABLES
Table Page
3.1 Comparison of Mirex Peak Heights (nun) and Calculated
Amounts (pg) as Determined by Welch and RTI 24
3.2 Interperson Comparison of Mirex Peak Height Measurements . . 26
4.1 Specimens Collected for Each Census Division by Age, Race,
and Sex 30
4.2 Detection of Mirex Residue by State: Number of Specimens
with Positive and Zero Detectable Levels 38
4.3 Detection of Mirex Residue by Race and Age: Number of
Specimens with Positive and Zero Detectable Levels 39
4.4 Chi-Square Analysis Results for Zero/Positive Values:
Geographic and Demographic Variables 40
4.5 Regression Model for Race, Age, Sex, and State: Zero/
Positive Residue Values 42
4.6 Regression Model for Race, Age, Sex, and Census Divisions:
Zero/Positive Residue Values 43
4.7 Summary for Quantifiable Positive Residue Values: Geometric
Means and Standard Errors 45
4.8 Summary for Quantifiable Positive Residue Values: Geometric
Means and Standard Errors 46
5.1 Weighted Chi-Square Analysis Results for Zero/Positive
Values: Geographic and Demographic Variables 50
5.2 Regression Model for Race, Age, Sex, and Census Division:
Zero/Positive Residue Values 52
5.3 Weighted Regression Model for Race, Age, Sex, and State:
Zero/Positive Residue Values 53
5.4 Estimated Summary Statistics for Quantifiable Positive
Residue Values: Weighted Geometric Means and Standard
Errors 54
5.5 Estimated Summary Statistics for Quantifiable Residue
Values: Weighted Geometric Means and Standard Errors ... 55
5.6 Weighted Regression Model for Location, Age, Race, and Sex:
Quantifiable Positive Residue Amounts 57
5.7 Weighted Regression of Quantifiable Positive Residue
Amounts: Location, Age, Race, and Sex 58
-------
-V-
LIST OF FIGURES
Figure Page
2.1 Sampling frame and selection procedures for Mirex Special
Study 8
3.1 Gas chromatogram - Aroclor 1260 standard 17
3.2 Gas chromatogram - mirex standard 18
3.3 Gas chromatogram - mixture of Aroclor 1260 and mirex
standards 19
3.4 Gas chromatogram - mirex containing sample (moderate level)
with interference peak (moderate level) 20
3.5 Gas chromatogram - mirex containing sample (low level), with
interference peak (low level) 21
3.6 Gas chromatogram - mirex containing sample (moderate level)
with interference (high level) 22
4.1 Frequency Distribution for Raw Values of Residue Amount. . 33
4.2 Frequency distribution for transformed values of mirex
residue 34
4.3 Frequency distributions of lipid adjusted mirex residue
amounts and transformed values (n=123 quantifiable
positives) 37
-------
ABSTRACT
Mirex is an insecticide which has a specific geographic region of
application in the United States. The Mirex Special Study was undertaken
in 1975 as part of the National Human Monitoring Program in an attempt to
obtain information about the levels of mirex in human adipose tissues in
areas of mirex application. A sample of human adipose tissue specimens was
selected from the mirex treated areas in eight southern States. Detectable
levels of mirex were found in 141 chemical analyses of 624 human adipose
tissue specimens in the sample. The data were analyzed with respect to
three demographic variables (race, age, and sex) and two geographic vari-
ables (Census Division and State). Geographic divisions appear to be the
most salient factors in levels of concentration of mirex in the sample
studied.
-------
-2-
1. INTRODUCTION
1.1 Foreword
This report presents a synopsis of the research activities involved in
Task 17 of EPA contract number 68-01-5848, Statistical Analysis of Mirex
Special Study Data. Chapter 2 describes the survey design employed in the
Mirex Special Study. A rationale of the statistical analysis approach is
also included in Chapter 2. Chapter 3 is a discussion of the chemical
analysis. Chapters 4 and 5 are presentations of unweighted and weighted
statistical analyses, respectively.
1.2 Summary of Study
Mirex is an organochlorine insecticide widely used for the control of
the imported fire ant in large areas of the southern United States. It is
chemically resistant to degradation, which, in combination with the sub-
stance's pharmacodynamics, leads to environmental mobility and potential
incorporation into various food chains.
Mirex residues have been found in specimens of human adipose tissue, a
discovery which led to the Mirex Special Study within the organization of
the National Human Monitoring Program (NHMP). From October 1975 through
September 1976, 630 human adipose tissue specimens were collected from
selected surgical patients and cadavers. The sample was selected within a
study area defined as all counties having some application of the insecti-
cide mirex between 1965 and 1974. Mirex residues were found in 141 of the
624 tissue specimens analyzed.
The definitions of the study area, sample design, sample selection,
specification of the procedures for collecting specimens in the field, and
specification of chemical analysis were undertaken by the NHMP within the
Environmental Protection Agency (EPA).
-------
-3-
The statistical analyses and assessment of chemical analysis were
performed by Research Triangle Institute, Research Triangle Park, North
Carolina, under EPA contract number 68-01-5848. EPA Task Manager was Cindy
Stroup and EPA technical assistant was Sandra Strassman-Sundy.
RTI personnel with major responsibilities were: S. R. Williams,
Project Director; Dr. D. Lucas, Task Leader; Dr. C. Sparacino, chemist; and
C. Leininger, statistician.
1.3 RTI Objectives and Findings
1.3.1 Objectives
RTI had several primary objectives for Task 17, Statistical Analysis
of MSS Data. These were: investigate and document the MSS survey design,
calculate quasi selection weights, assess the assumptions and limitations
of the design, verify the chemical analysis results, and analyze the weight-
ed and unweighted data for all 624 specimens and for the 123 quantifiable
positive values.
1.3.2 Findings
This report and the accompanying appendices draw together the informa-
tion about the MSS provided to RTI by EPA staff and the results of the RTI
staff work on design assessment. The method of selection weight calculation
is outlined briefly.
The chemical analysis results are quite solid and are not a problem in
this study.
RTI considers the weighted regression and weighted chi-square to be
the most sensitive and most appropriate tests for describing the study
population. Usually, results of an unweighted analysis should not even be
presented, especially in the face of extreme disproportionate sampling,
because it is easy to misinterpret such results. The unweighted results
-------
-4-
are presented here because the selection weights could not be founded
entirely on probability of selection.
The results of the unweighted analysis indicate that the location of
the collection site is the most important factor in the differences in
number of positive detections of mirex in human adipose tissue. In un-
weighted regression analysis of the 123 quantifiable observations, demo-
graphic factors and geographic factors did not show any effect in the
degree of mirex concentration in the sample specimens. The results of the
unweighted chi-square tests are significant (p<0.0001) for State and Census
Division, and marginally significant for the factors age and race (p<0.10).
In the weighted analyses location of the selection site is again the
major variable of interest. State is a significant (p<0.0001) factor in
both weighted chi-square and weighted regression analyses of the proportion
of positive mirex detection. The weighted regression analysis for the
quantifiable residue amounts indicates an interaction between geographic
location and race. This suggests that there is a race effect for some
geographic divisions that is not particularly strong throughout the mirex
treated area. This interpretation is corroborated by the results of the
weighted chi-square tests.
-------
-5-
2. THE NATIONAL HUMAN MONITORING PROGRAM'S
SPECIAL STUDY OF MIREX
2.1 Detailed Description of the Study
2.1.1 Introduction
Human adipose tissue specimens for the Mirex Special Study (MSS) were
collected from October 1, 1975 through September 30, 1976 (FY76). The
target population was the human population of all counties where mirex was
sprayed at any time during the 10-year period 1965 to 1974 [1]. A list of
these counties was provided to the NHMP manager by the U.S. Department of
Agriculture. Maps of the eight southern States containing these counties
are displayed in Appendix A. A list of the mirex counties, by State, is
given in Appendix B. The pattern of mirex use can be charted directly to
the areas of infestation of imported fire ants in this country. This
infestation pattern does not conform to large-scale geographic or political
boundaries, such as Census divisions or States, as can be seen by examining
the maps in Appendix A. No information was provided concerning the amount
or frequency of mirex application over the 10-year period. Such informa-
tion might be extremely important in determining levels of expected ex-
posure to this insecticide; for further discussion of this point see Appen-
dix C.
2.1.2 Sample Design
Mirex residues were found in human adipose tissue specimens submitted
to the NHMP, which was designed to provide estimates at the national level.
There were eight NHMP cities in mirex areas where established contacts
(hospital pathologists or medical examiners) were already submitting adi-
pose tissue specimens for the NHMP. A special study was designed to in-
crease representation from mirex regions by increasing the number of sample
-------
-6-
sites in mirex areas. 32 supplementary sites were selected, bringing the
total number of sites contacted in the MSS to 40.
To select the 32 supplementary areas for the MSS, a listing was made
of all 1974 Ranally Metropolitan Areas (RMA's) in the mirex area: RMA's
were defined on a township and locality basis. All areas with a total
population of 50,000 or more were considered to be RMA's. Other areas that
presently had populations close to 50,000, had populations greater than
50,000 in the past, or had special significance in their States were also
defined as RMA's [2]. The complete listing of the population from which a
sample will be selected is called a frame. The eight NHMP sites were on
the frame of 68 RMA's located in the mirex area. These 8 cities were not
included as possible selections for the mirex supplementary sample because
they were already included as part of the MSS sample by virtue of their
initial selection into the NHMP.
Of the 60 RMA's in the mirex supplementary list, 32 were apparently
selected by simple random sampling (SRS). The 60 RMA's were listed in
alphabetical order (not pertinent to the selection process), and then a SRS
was selected from the list. See Appendix D for a discussion of the site
selection methodology employed. Nine additional sites were selected, in
the same manner, to be used as alternates in case there were nonresponding
sites. These alternates were to be substituted for nonresponding sites in
the order in which they were selected. One site was replaced by an alter-
nate selection. A list of the selected sites and alternates is given in
Appendix £.
The selection of sites within mirex areas produced 40 sample RMA's.
Of these, 5 of the 8 NHMP sites and 23 of the 32 mirex supplementary sample
RMA's responded with valid tissue specimens. It is not known if any of the
-------
-7-
nonresponding RMA's seat tissue samples that did not fit EPA specifica-
tions. The treatment of nonresponding sites is described in section 2.3.
Figure 2.1 outlines the path of the sample selection procedures for obtain-
ing tissue specimens for the MSS.
Data collection for the MSS after the selection of first-stage RMA's
follows the methods used in the Adipose Tissue Survey (ATS) of the NHMP.
For the convenience of the readers, a brief discussion of the NHMP is now
included; a more detailed description is contained in a 1979 EPA document
[3].
2.2 National Human Monitoring Program
2.2.1 Introduction
The statistical design used to collect data for the ATS of the NHMP
has two stages. The contiguous 48 states were stratified into several
regions. Sampling sites were selected from a list of eligible places with
probability proportional to population. Within the sampling sites, the
subsampling was performed by cooperating pathologists and medical ex-
aminers .
2.2.2 Strata
In FY70-FY72 the contiguous 48 states were stratified by Census Re-
gion. Beginning in FY73 the strata were changed to Census divisions. In
the earlier period, the number of sites selected within each stratum was
determined by its population as given in the 1960 Census. In FY73, the
allocations were revised based on 1970 Census data.
2.2.3 Eligible Places
In FY70-72, the eligible places were cities with populations greater
than 25,000 persons based on the 1960 Census. In FY73-76, the eligible
places were cities greater than 25,000 based on the 1970 Census. Sample
sites were selected in FY73 and followed through FY76. The sample sites
-------
-8-
23,557,425
17,085,700
13,486,900
unknown
624
Target population*—total population of counties
where mirex has been used.
Purposive exclusions of cities and towns not
defined as RMA's.
1st stage sampling frame and inference population-
the human population of 68 RMA's.
Selection of a subset of 40 of the 68 RMA's.
• *
1st stage sample of RMA's.
Purposive exclusion of some hospitals, pathologist*
and medical examiners, selection of only
cadavers and surgical patients.
Sampled population—includes a subset of only
surgical patients and persons who died in FY76.
Subsampling of surgical patients and cadavers
by cooperating professionals.
Sample.
Figure 2.1. Sampling frame and selection procedures for Mirex Special Study.
Source Based on information contained in references 1,4.
'Population values based on 1970 census data.
-------
-9-
were independently selected from each stratum with probability proportional
to size of the city. This was done by listing the sites in random order
along with their cumulative population totals. An interval for each stra-
tum was calculated by dividing the total population for all sites listed by
the number of sites to be selected in the stratum. A random number was
obtained between 0 and the length of the interval to give a starting point.
The sites were selected by matching their cumulative totals with the start-
ing point or integer multiples of the interval plus the starting point.
2.2.4 Pathologists
The subsampling within each site was done by a cooperating pathologist
or medical examiner. When no cooperative pathologist or medical examiner
could be found in a selected site, an alternate site was selected. Alter-
nate sites were chosen by position in the listing of eligible sites with
respect to the nonrespondent site. The first alternate was the site im-
mediately below the originally selected site, the second alternate the one
immediately above, the third alternate was the second site following and so
on.
2.2.5 Quotas
Each site was assigned a quota based on the demographic characteristics
of age, sex, and race. The ages were grouped into three ranges: 0-14,
15-44, and greater than 44. The races were classified as white and nonwhite.
The quotas for FY70-FY72 were based on the national demographic character-
istics according to the 1960 Census. The quotas for FY73-FY76 were deter-
mined by the appropriate demographic characteristics for each Census division.
Below the sample city level there was no probability structure for
obtaining the participants in the survey. The hospitals, pathologists and
medical examiners were selected subjectively. Also, the tissue samples
-------
-10-
from cadavers and surgical patients were selected subjectively by the
cooperating professionals. Guidelines for selecting patients were distri-
buted by EPA. Appendix F contains a copy of the letter sent to each co-
operating professional containing these guidelines, which outline the
procedures for screening specimens and the types of individuals from whom
tissue specimens should not be obtained (for example, suspected victims of
pesticide poisoning and institutionalized patients). The MSS quotas are
given in Appendix G.
2.3 Selection Weight Construction
In order to use the information obtained in the MSS to estimate the
level of mirex concentration in the population of RMA's in the target area
and to account for disproportionate sampling rates, it is necessary to
weight the original values before analysis. This presents some problems,
for the nature of the selection of sites, which includes at some points
purposive selection (subjective, not based on any known probability of
selection), is incompatible with the formation of statistically valid
sampling weights. A method of "quasi" weighting has been devised and used
for the weighted analyses. A simple method of weighting was chosen to
facilitate the weighted analyses, but it is by no means necessarily the
best or only way to weight the data. The method of weighting chosen and
some of that particular method's limitations are described here.
The first stage selection frame, consisting of 68 1974 RMA's from the
designated mirex counties in eight southern States, was actually split
between two lists, which were considered separately in constructing
weights. One list consisted of eight mirex area cities that were original-
ly part of the NHMP, and the second list consisted of 60 sites (RMA's),
-------
-11-
from which the supplementary MSS sample was drawn. The eight cities that
had been part of the NHMP were included with certainty in the sample.
These eight cities are in RMA's, but were not considered among the possible
selections for the supplementary sample.
Even if one assumes the selection process within sample sites is
approximately random, the sample design of the MSS does not assign equal
"probabilities" of selection to all elements (cadavers and surgical pa-
tients) in the sample. This introduces the possibility of sampling bias in
calculating population estimates. Bias refers to the difference between an
estimated population value and the true population value. It can occur as
a result of sampling error or measurement error, or a combination of both.
To facilitate bias reduction in this situation, a sample weight for each
observation was calculated, which attempts to reflect its approximate
probability of selection. Including weights in the analysis should reduce
bias in estimating means or proportions. For probability sampling this
results in strictly unbiased estimators for such statistics as means,
proportions, and totals. In the following paragraphs, procedures are given
for computing weights for the MSS. Again, note that true sampling weights
cannot be calculated because some stages of selection involve non-probabil-
ity sampling.
The population from which tissue specimens might be obtained is not
actually all people residing in a selected RMA; only surgical patients or
individuals who die from certain causes at the selected hospital in any
sample RMA have a chance of being selected. The surgical patients and
cadavers that are available to the selected pathologist may not necessarily
be residents of the RMA that they are purported to represent, because
individuals may travel to another city for medical treatment. The possi-
bility of medical care migration is particularly problematic for RMA's that
-------
-12-
are located on or near the borders of mirex areas (such as Dallas).
To estimate the probability of a given individual being selected
within a sample site, one might use the number of pathologists and hos-
pitals, so that the probability of selecting any one pathologist at any
hospital in an RMA could be estimated. Then the population of suitable
surgical patients and cadavers to which the pathologist had access would
need to be estimated, so that a probability of selection of a tissue spec-
imen by that pathologist could be estimated. Even with this type of in-
formation, the estimates of selection probability would be very rough
because the pathologists and tissue specimens in the MSS were not selected
in any random or systematic method but by purposive selection. At this
time, there is no reasonable adjustment that can be made to take the above
facts into consideration. Hence, equal probability of selection is assumed
for the within-site stage of sampling. This method involves dividing the
number of specimens received by the population of the RMA.
The probability of an RMA being selected in the MSS is dependent upon
the study for which the site was originally selected because the selection
criteria for the sites that were already part of the NHMP were different
from those for the sites that were sampled specifically for the MSS.
The probability of selection of a site at the first stage is the
number of selections made divided by the number of possible selections;
this is the number of selected RMA's divided by the number of RMA's in the
frame, or
p 'fist stage) number (n) of sample RMA's
r (selection/ number (N) of RMA's on the li
list
-------
-13-
For each of the eight RMA's that are part of the NHMP, the probability
of selection is equal to 8/8, or a probability of selection of one. These
eight sites were included with certainty in the sample because data collec-
tion systems already existed at these locations. For RMA's that were
selected as part of the supplementary or additional MSS sample, the proba-
bility of selection is 32/60, or 0.53. The supplementary sites were se-
lected by simple random sampling, which does not consider the size of the
population of the RMA's in the selection process. When equal selection
probability is used at the first stage and approximately equal sample sizes
are selected from each of the primary sampling units, this can result in
oversampling small RMA's and cause underrepresentation of RMA's that have
large populations.
Weighting the observations within the sample site is used to adjust
for this. The weight of a given sampling unit is the inverse of the
"quasi" sampling rate. Hence, the weight for a given observation is
c , .. . ,k number of RMA s on list population of RMA
Selection weight = r ^ ;—-.,., x —r f, ^ :
6 number of sample RMA s number of specimens
selected in RMA
In this study, as noted previously, some of the sample sites selected
did not submit valid tissue samples for analysis. To compensate for this
nonresponse, the inverse of the proportion of selected sample sites which
responded was used to adjust the weights. For example, data were submitted
from five of the eight NHMP sites ( | or 62.5 percent responded), which
results in a nonresponse adjustment factor of 1.6 for NHMP sites. The
nonresponse adjustment for supplementary RMA's is 1.23, the inverse of
26/32. This factor, multiplied by the selection weight for an observation,
produces a weight adjusted for nonresponse at the first stage level. There
-------
-14-
was no information upon which to base any nonresponse adjustments for
selection of tissue specimens within selected sites.
-------
-15-
3. EVALUATION OF CHEMICAL ANALYSIS
3.1 Introduction
The objective of this portion of the study is to assess the reli-
ability of the results obtained from the chemical analytical procedures
employed in the Mirex Special Study. The analysis of adipose tissue speci-
mens for mirex was carried out at the Michigan Department of Public Health
under the direction of Mr. Robert L. Welch.
The analytical procedure used for the determination of mirex is de-
scribed in the EPA Pesticides Methods manual [5]. Specific details re-
garding the collection, storage, and shipping of specimens were described
in the NHMP study report for mirex [1]. In conformance with the Work Plan
requests for needed data, Mr. Welch shipped a number of raw mirex chroma-
tograms to RTI. These chromatograms were examined in sufficient detail to
allow for a meaningful assessment of the quality of reported mirex levels.
The identification numbers (not patient identifers) and dates of analysis
for these chromatograms are provided in Appendix H.
3.2 Discussion
The method used for the determination of mirex levels was based on the
use of an external standard. Reference standards were injected before,
during, and at the end of each mirex specimen group. The standard con-
sisted, usually, of a mixture of known amounts of mirex and Aroclor 1260.
This mixture served not only for purposes of quantitation, but also pro-
vided a constant check on the degree of separation between mirex and a
late-eluting Aroclor peak. This peak represents an interference found in
most tissue specimens due to the ubiquitous nature of the Aroclors. In an
effort to maximize the separation of the two peaks, the gas chromatographic
-------
-16-
pcocedure was modified, and the analysis was conducted at a temperature
slightly below that recommended in the Pesticides Manual. This procedure
was approved by memo (Appendix I) prior to chemical analysis. Chromatograms
of Aroclor 1260, mirex, and a mixture of 1260/mirex are depicted in Figures
3.1 through 3.3. Baseline separation was achieved for mirex and the Aro-
clor 1260 peak, and this separation was maintained throughout the analysis
period covered by the chromatograms examined (May 1976-November 1977).
The same quantity of Aroclor 1260 and mirex was injected for each
check. Thus, a measure of the reproducibility of injection could be deter-
mined by measuring detector response (peak height) for these standards. A
check of 17 such injections showed acceptably close agreement (differences
< 10 percent) between injections for a given specimen group. This agree-
ment and the reproducible retention times observed for the compounds over a
period of several months are indicative of a stable analytical system.
Greater reliability can be attached to results obtained from such a system.
For a significant number of chromatograms, an interference peak other
than Aroclor 1260 was noted. Of some 50 chromatograms from which a quant-
itative assessment of mirex was made, approximately half (26) showed mirex
peak interference. The degree of interference varied (Figures 3.4 through
3.6) but was not so severe as to prevent estimation of levels of target
compound.
Of the chromatograms examined, only one blank and two control speci-
mens were included. These specimens were not described so no significance
could be attached to them, except that the blank chromatogram was "clean"
in the area of interest.
To assess the reliability of the chroraatographic and, to a limited
extent, the quantitative procedures used, all mirex peaks were measured
-------
Figure 3.1. Gas chromatogram - Aroclor 1260 standard.
-------
00
Figure 3.2. Gas chromatogram - mirex standard. Conditions as in Figure 1.
-------
-19-
Figure 3.3. Gas chromatogram - mixture of Aroclor 1260 and mirex
standards. Conditions as in Figure 1 except chart
speed is 0.5X.
-------
I
r-o
o
Figure 3.A. Gas chromatogram - mirex containing sample (moderate level)
with interference peak (moderate level). Conditions
as in Figure 3.1.
-------
Figure 3.5. Gas chromatogram - mirex containing sample, (low level) with Interference peak,
(low level). Conditions as in Figure 3.1.
-------
I
ro
K>
I
Figure 3.6. Gas chroma togram - mirex containing sample (moderate level) with
interference (high level). Conditions as in Figure 3.1.
-------
-23-
(peak heights) and the results compared with those obtained from the chro-
matograms. Comparisons were also made of the amount of mirex as calculated
from the chromatograms and as determined by the RTI staff. These data are
shown in Table 3.1. As can be seen, the results are consistent with aver-
age absolute errors and standard errors (A%, see Table 3.1) of 9.2 + 2.1%
for peak height measurements, and 8.1 + 1.2% for calculated amounts. A
check on the interpersonal variability was determined by measuring and
calculating the same parameters, i.e., peak height and amount of mirex,
using a different RTI staff member. Some 10 percent of the RTI-measured
chromatograms were replicated in this manner. These results are shown in
Table 3.2. Again the data are very consistent, with the average absolute
error and standard error calculated at 3.0 + 1.0 %. Thus, the measurement
of mirex peaks and the calculation of raw mirex amounts were straight-
forward, and, at least for the chromatograms examined, no gross errors
(inaccurate measurement, faulty calculation, peak assignment error) were
evident.
As no completely adequate method exists for the quantitative statis-
tical treatment of "trace" levels of compound, an attempt was made to
ascertain the minimum detectable quantity (MDQ) of mirex as analyzed by
Welch. Since the MDQ was not calculated or made available for RTI use, an
approximation was made adopting the usual assumption of MDQ = 3 (signal/
noise). Since two detection systems were used in this study, MDQs were
determined for each at the beginning of the analytical runs (May 1976) and
again at the end (November 1977). For system 1, the MDQs were 16 pg at the
beginning and 20 pg at the end of the study; for system 2, the MDQs were 9
pg and 12 pg at the beginning and end, respectively. The similarity of
these values bespeaks an extremely stable analytical system, with more than
-------
-24-
TABLE 3.1. COMPARISON OF MIREX PEAK HEIGHTS (mm) AND
CALCULATED AMOUNTS (pg) AS DETERMINED BY
WELCH AND RTI
Specimen
MA76-98 (1:2)
MA76-100 (1:5)
MA76-101 (1:2)
MA76-123 (1:2)
MA76-124 (1:2)
MA76-125 (1:2)
MA76-126 (1:2)
MA76-127 (1:5)
MA76-128 (1:2)
MA76-129 (1:2)
MA76-130 (1:2)
MA76-131 (1:2)
MA76-160 (1:2)
MA76-161 (1:2)
MA76-162 (1:2)
MA76-164 (1:2)
MA76-165 (1:2)
MA76-166 (1:2)
MA76-167 (1:2)
MA76-80 (1:5)
MA76-81 (?)
MA76-82 (1:5)
MA76-83 (1:5)
MA76-157 (1:2)
MA76-158 (1:2)
MA76-159 (1:2)
MA77-9 (1:2)
MA77-11 (1:1)
MA77-12 (1:2)
Ht (Welch)
17
108
19
54
58
43
13
75
75
20
20
78
15
98
28
0
0
0
0
15.5
32.5
118
6.5
22
0
32
12
60
21
Ht (RTI)
16.5
113.0
20.5
57.0
77.0
43.5
13.0
76.5
82.0
20.0
20.0
81.5
15.0
99.5
33.0
0
0
0
0
15.0
32.5
119.0
6.5
23.5
0
32.5
16.5
67.0
23.0
A%*
-2.9
4.6
7.9
5.6
32.8
1.2
0
2.0
9.3
0
0
4.5
0
1.5
17.9
-
-
-
-
-3.2
0
0.8
0
6.8
-
1.6
37.5
11.7
9.5
Amt (Welch)
187
2967
284
9
X
X
236
X
986
213
274
829
221
2020
412
0
0
0
0
470
198
2450
198
224
0
327
102
254
244
Amt (RTI)
159
2729
271
1326
488
423
236
1855
982
185
240
755
190
2030
418
0
0
0
0
458
198
2466
198
209
0
289
115
205
217
A%*
-15.0
-8.0
-4.6
-
-
-
0
-
-0.4
-13.1
-12.4
-8.9
-14.0
0.5
1.5
-
-
-
—
-2.6
0
0.7
0
-6.7
—
-11.6
12.7
-19.5
-11.1
-------
-25-
TABLE 3.1. (cont'd)
Specimen
MA77-13 (1:2)
MA77-14 (1:2)
MA77-60 (1:2)
MA77-61 (1:2)
MA77-129 (1:2)
MA77-130 (1:2)
MA77-133 (1:2)
MA77-134 (1:2)
Ht (Welch)
71
124
56
107
15
18
12
54
Ht (RTI)
80.0
126.5
60.0
120.0
11.0
17.5
17.0
70.5
A%*
12.7
2.0
7.1
12.4
-26.7
-2.8
41.7
30.6
Amt (Welch)
602
1442
577
973
?
7
120
378
Amt (RTI)
557
1193
517
906
282
280
142
400
A%*
-7.5
-17.3
-10.4
-6.9
-
-
18.3
5.8
A % =
X (RTI) - X (Welch)
X (Welch)
(100)
-------
-26-
TABLE 3.2. INTERPERSON COMPARISON OF MIREX PEAK HEIGHT MEASUREMENTS
Specimen
MA 76-128 (1:2)
MA 76-129 (1:2)
MA 76-130 (1:2)
MA 76-131 (1:2)
MA 76-12 (1:2)
MA 76-11 (1:5)
MA 76-60 (1:2)
MA 76-60 (1:5)
Peak height
KB
85.0
21.0
20.0
79.5
22.5
25.5
60.0
29.0
(mm)
SF
82.0
20.0
20.0
81.5
23.0
23.5
60.0
29.5
A %
3.7
5.0
0
-2.5
-2.2
8.5
0
-1.7
-------
-27-
adequate sensitivity for these types of analyses. Unfortunately, these
values could not be translated into the concentration values (ppm) reported
on the data file. This was due to the inability to relate the specimen
identifiers (specimen numbers) on the same chromatograms with the identi-
fiers (patient numbers) reported on the raw data file. In addition, in-
formation was not available regarding the exact analytical procedures
followed, and, thus, specimen dilution factors could not be ascertained.
The absolute amounts of mirex, as shown on most of the chromatograms, could
not, therefore, be converted to ppm fat.
Although an assurance check could not be made on the final, reported
mirex concentration levels, the raw data indicate that sound analytical
quantitation procedures were followed, and, in general, the results are of
acceptable quality. Further assurance of the quality of the data is pro-
vided by independent confirmation of the mirex levels using, primarily, gas
chromatography/mass spectrometry (GC/MS). Results of analysis of some 23
specimens (Appendix J) showed reasonably good agreement between the values
reported by Welch and those reported by the confirming laboratories.
The quality of the raw data, the results of the confirmational anal-
yses, and the stability of the analytical system are taken as good in-
dicators of reliable and accurate results. Although every aspect of the
analysis could not be checked (for example, calculations leading to con-
centration values), no errors were noted from the available chemical analy-
tical data. There is no indication that the information as reported on the
raw data file..is suspect.
-------
-28-
4. UNWEIGHTED ANALYSIS
4.1 Data Description
The MSS data received by RTI consisted of 664 observations from 31
responding sites. 40 observations were excluded from analysis: 34 records
were of specimens collected outside the reference time period, and the
lipid extractable amount for 6 observations was less than 10 percent.
There were no data exclusions on the basis of inadequate chemical analysis
confidence codes. After exclusions, there were 624 observations available
for statistical analysis.
In Chapter 2, there is brief mention of desired specimen quotas by
age, race, and sex (Appendix G). These quotas guided the field contact in
submitting specimens from a range of individuals. It should be kept in
mind that representativeness for factors of occupation and length of resi-
dence in an area is not accomplished through balancing on age, race, and
sex information. The quotas were established in an attempt to obtain
information that would reflect the age, race, and sex distribution of the
living population.
Tissue specimens that fit the quota specifications were referred to as
design specimens, and those tissue specimens received in excess of the
quota guidelines were called surplus specimens. None of the sites filled
their quota of 27 design specimens even though 6 of the 31 responding sites
submitted 27 or more tissue specimens. There were 479 design specimens and
145 surplus specimens. There is no difference in methodology of selection
or collection between the design and surplus specimens; the designation was
made in order of date of receipt of a specimen as the study year progressed.
-------
-29-
There seems to be no obvious reason, to exclude the surplus tissue specimens
(23.2 percent of the data) from weighted analysis and probably not from
unweighted analysis because design quotas were never satisfied.
Table 4.1 contains classifications for all tissue specimens within
Census Divisions by age, race, and sex. The desired quotas for racial
representation were met reasonably well within the South Atlantic and West
South Central Census Division and almost exactly within the East South
Central Division. The Census Division level desired distribution by age
group was not met; the 0-14 age group was consistently underrepresented in
all three divisions. The actual distribution of specimens classified by
sex was roughly in keeping with the desired quotas on the Census Division
level.
For the various sample sites the number of valid tissue specimens
submitted ranged from 1 to 37. Dallas, Texas, one of the NHMP sites and
the largest RMA in the MSS (2,455,000 people), was the location that con-
tributed only one tissue specimen. Dallas presents further analysis prob-
lems because of its location, for it falls on the edge of a mirex area and
part of the defined RMA including Dallas falls outside of the mirex bound-
ary. Because Dallas is the largest city in that part of Texas (see maps,
Appendix A), the population that relies on the medical facilities in Dallas
extends far beyond the boundaries of the mirex counties. Houston and San
Antonio are similarly situated in both geographical and medical terms. The
locations of other RMA's are subject to the same cautions in interpreting
the residence of the individuals from whom adipose tissue specimens were
obtained. Some sites, such as those in Louisiana, are clearly located in a
heavily fire-ant-infested region, and for these, little ambiguity exists
about the potential mirex exposure of the population.
-------
-30-
Table 4.1 Specimens Collected for Each Census Division
By Age, Race, and Sex*
East South Central; Alabama and Mississippi
Number of Specimens
Caucasian
Non-Caucasian
Age Group
0-14
15-44
>45
Male
5
22
36
63
Female
7
29
43
79
142
Male
1
6
7
14
Female
2
5
12
19
Total
15
62
98
33
175
desired Non-Caucasian quota = 5/27 = 18.5%
actual Non-Caucasian representation = 33/175 = 18.9%
South Atlantic: Florida, Georgia, North Carolina, and South Carolina
Number of Specimens
Caucasian
Non-Caucasian
Age Group
0-14
15-44
>45
Male
13
35
55
103
Female
8
29
52
89
Male
6
25
18
49
Female
10
11
20
41
192
90
Total
37
100
145
282
desired Non-Caucasian quota = 6/27 = 22.2%
actual Non-Caucasian representation = 90/282 = 31.9%
West South Central: Louisiana and Texas
Number of Specimens
Caucasian
Non-Caucasian
Age Group
0-14
15-44
>45
Male
7
18
28
53
Female
9
34
41
84
137
Male
2
7
7
16
Female
0
8
6
14
Total
18
67
82
30
167
desired Non-Caucasian quota = 4/27 = 14.8%
actual Non-Caucasian representation = 30/167 = 18.
Based on the results of the MSS and information in Appendix G.
-------
-31-
In future investigations it would be useful to obtain the zip code of
the home address of the patients from whom the tissue specimens are obtain-
ed so that residence in a mirex area can be verified. However, this still
would not give any indication of the length of time an individual resided
in an area exposed to mirex. Level of exposure and intensity of appli-
cation of mirex to an area are also not known but are of interest.
4.2 Data Analysis
4.2.1 Data Distribution
In the interpretation of estimates and statistical tests in this
section, which are based on unweighted analyses of MSS data, it must be
recognized that at least two categories of individuals are not as prevalent
(underrepresented) in the sample as in the population. These are people
under age 15 and residents of the larger RMA's. In light of the correla-
tions between geographic location and mirex positives, the latter imbalance
may cause unweighted analysis to be seriously biased for purposes of ex-
tending the results to all persons in the study population. The weighted
analyses in the following chapter should reduce this risk of bias.
In examining the data, it is important to consider the distribution of
mirex levels before choosing summary statistics or conducting descriptive
analyses. Of the 624 specimens, 141 (22.6 percent) contained detectable
amounts of mirex. Eighteen (12.8 percent) of the 141 detectable values
were trace quantities of mirex, present but not in sufficient quantity to
be accurately measured (see discussion in section 3.2). These trace quanti-
ties were included in analyses where presence or absence of mirex was
measured but not included in those statistical analyses where a specific
numerical value was required.
-------
-32-
The two histograms in Figure 4.1 illustrate the distribution of the
frequency of mirex residue values (in parts per million) for all 624 speci-
mens and for the 123 specimens for which there were definite measurable
quantities. It is obvious that the levels of mirex in the sample do not
follow a normal distribution (example at the bottom of Figure 4.1), even
when the 501 zero (nondetectable) and trace values are excluded. The
skewing of the distribution to the right indicates that the residue amounts
follow a lognormal or exponential distribution, and that a transformation
of the data might yield a distribution of 'values that would be more appro-
priate for meeting the assumption of normality underlying many statistical
tests. Figure 4.2 contains histograms of the results of two logarithmic
transformations of the measurable positive residue amounts: natural log
(log ) and log base 10 (logxo)- The logio transformation is slightly
closer to normal than the log transformation and was chosen as the trans-
formation to be used in computing summary statistics and in analyses in-
volving the quantifiable residue amounts.
It should be noted that the actual distribution of mirex levels contain-
ed in human adipose tissues probably consists of fewer true zeros than
current analytical techniques indicate. The skewing to the right would
still exist, but the distribution of values would more closely resemble a
true continuous distribution if the presence of mirex in any quantity,
however small, could be measured.
No minimum detectable quantity could be determined for mirex in human
adipose tissues; consequently, no specific numerical value could reasonably
be substituted or inserted for the 18 trace observations. The zero values
and trace values could be transformed to a logarithmic scale if a constant
term was added to all 624 values before transformation. This was not done
-------
-33-
500 -
number of
specimens 400 -
300 -
200 -
100 -
n = 624 sample values:
zeros, traces, and
quantifiable values
zeros and traces' 0.5 1.0
Residue Amount, ppm
1.5
2.0
2.5
90 •
80 •
70 •
number of
specimens 60 •
50 -
40 -
30 -
20 -
10 -
n = 123 quantifia
positive valu
.A.
<0.4
0.8 1.6 2.4 3.2
Residue Amount, ppm
4.0
4.8
68% of
values
normal curve
Figure 4.1 Frequency distribution for raw values of residue amount
*Based on results of Mirex Special Study, FY76.
-------
-34-
60
50
number of
specimens 40
30
20
10
•
,
•
^ , 1 , _
-4
-3
-2
-1
log mirex residue amounts
50
number of 40
specimens
30
20
10
•
•
»
l
-1.6 -1.2 -0.8 -0.4 0.0
l°8iO mirex residue amounts
0.4
0.8
Figure 4.2 Frequency distributions for transformed values of mirex residue'
Based on results of Mirex Special Study, FY76.
-------
-35-
because there are so many nondetectable values that summary statistics for
the entire sample would provide little information. Instead, it was de-
cided to look at detectable versus nondetectable values and at the quant-
ifiable residue amounts.
Both the proportion of positive values and the geometric mean of the
transformed data were used as summary statistics. The geometric mean is an
appropriate summary statistics for logarithmic data [6].
It has been specified by EPA that part of the analysis in this task is
to be carried out on the residue amount divided by the percent lipid ex-
tractible material (LEM) in the specimen. This quantity is indicated by
the term Y in the following pages. The natural log, log , was obtained for
all 123 quantifiable positive values of Y. The two histograms in Figure
4.3 are the distributions of Y and log Y.
4.2.2 Proportions of Positive Values
In Tables 4.2 and 4.3 are some simple classifications of the number of
detectable and nondetectable values by demographic and geographic cate-
gories. In Table 4.2, which is a summary by State, the two States with the
largest percent of detectable mirex levels are Mississippi and Louisiana,
46.3 and 35.7 percent positive, respectively. Georgia has the third larg-
est percentage of detectable mirex levels, at 20.9 percent. The percentage
of detectable mirex values drops to less than 8 percent for the 5 remaining
States. There were no positive values for the specimens submitted from
North Carolina and Texas.
This information by State suggests a possible connection between
residence in a widespread mirex-treated area with detectable mirex levels
in human adipose tissue. This is borne out by the highly significant
results of a chi-square test on the zero/positive values by State in Table
4.4. The empty cells for positive values for Texas and North Carolina make
-------
-36-
the chi-square test results less reliable than if there were some positive
values for these states. However, it can still be used to infer that there
are significant differences in the percent of positive mirex residues
detected by State.
Census Division was the only other variable investigated which shows
significant results in the chi-square test. Race, age, and sex did not
show significant differences in the percent positive mirex residue values
for the chi-square test results in Table 4.4.
The information in Table 4.3, which is a three-way classification by
race, age, and detectable mirex residues, is less easily summarized than
the data in the two-way classification in Table 4.2. When considering
additional categories, it becomes necessary to consider possible inter-
actions between variables as well as simple main effects. This can be done
by computing an analysis of variance (ANOVA) or a regression equation. A
model such as the following can be used.
P. ., a = |J + A. + S. + R, + L. + £. .
ijklra r i j k S,
where
P... „ = mirex level in tissue for the m individual, the
ijk£m Otn , _, /--IN
£ location, in the (ijk) age-race-sex group;
[i = mean mirex level;
A. = age effect for the i age group;
S. = sex effect for the j sex group;
R, = race effect for the k race group;
L, = location effect of the £ location;
£..,„ = random error.
ijk£m
In the above model, testing for age, sex, race, and locations effects
corresponds to testing the equality of the A., S., R, , and L.. Standard
1 J K *•
-------
-37-
Number of
specimens
80 -
70 -
60 -
50 -
40 -
30 -
20 -
10 -
AAA . . . I i l . l i
<0.5 "'123456
Y
50
40
30
Number of
Specimens 20
10
•
~~i
-4
-3
-2
-1 0
LogY
Y = Mirex Residue Amount (ppm)/% Lipid Extractible Material
Figure 4.3 Frequency distributions of lipid adjusted mirex residue amounts
and transformed values (n=123 quantifiable positives)*
*
Based on results of Mirex Special Study, FY 76.
-------
-38-
Table 4.2 Detection of Mirex Residue by State: Number of Specimens
with Positive and Zero Detectable Levels*
State/number of responding RMA's
Alabama (3)
Florida (6)
Georgia (5)
Louisiana (5)
Mississippi (6)
North Carolina (1)
Texas (3)
South Carolina (2)
Zero
36
129
72
83
73
18
38
34
483
Positive
3
8
19
46
63
0
0
2
141
Total
39
137
91
129
136
18
38
36
624
Percent
Positive
7.7
5.8
20.9
35.7
46.3
0.0
0.0
5.6
*
Based on results of the Mirex Special Study, FY 76.
-------
-39-
Table 4.3 Detection of Mirex Residue by Race and Age:
Number of Specimens with Positive and Zero Detectable Levels*
Race
Caucasian
Non-Caucasian
Total over race and age
Age
Group
0-14
15-44
>45
Total
0-14
15-44
>45
Total
Number
Zero
42
128
202
372
19
44
48
111
483
of Specimens
Positive
7
39
53
99
2
18
22
42
141
Total
49
167
255
471
21
62
70
153
624
Based on the results of the Mirex Special Study, FY76.
-------
-40-
Table 4.4 Chi-Square Analysis Results for Zero/Positive Values:
Geographic and Demographic Variables'"
2
Variable Number of % Positive within X df
specimens each level
AGE
0-14
15-44
>45
CENSUS DIVISION
East South Central
South Atlantic
West South Central
RACE
Caucasian
Non-Caucasian
SEX
Male
Female
STATE
Alabama
Florida
Georgia
Louisiana
Mississippi
North Carolina
South Carolina
Texas
70
229
325
175
282
167
471
153
298
326
39
137
91
129
136
18
36
38
12.7 4.53 2
24.5
23.1
37.7 49.65 2
10.3
27.5
21.0 2.73 1
27.5
24.2 0.80 1
21.2
7.7 105.8 7
5.8
20.9
35.7
46.3
0.0
5.6
0.0
Significance
level (p)**
0.103
0.0001
0.098
0.371
0.0001***
Calculated at RTI, total number of specimens (N) = 624.
** 2
Probability that a X this large or larger might occur by chance, under
the assumption of no difference among categories.
***
Over 5% of the cells have expected counts less than 5. Sparse table
indicates chi-square may not be a valid test.
-------
-41-
statistical computer software was used to carry out these tests for un-
weighted data, and specialized software is available at RTI for carrying
out the analyses for weighted data (see Chapter 5). A model similar to the
above is used for testing the proportion of positive values and the geo-
metric means of the positive values for mirex residues found in human
tissue.
A regression model was tested for the main effects of race, age, sex,
and location, with location being alternately defined as State or Census
Division. Then the model was expanded to include all possible interaction
terms. The dependent variable for this model was presence or absence of
mirex, indicated as 0 or 1 in the model. The results of this analysis are
summarized in Tables 4.5 and 4.6. When State was the location factor, the
model gave a strong main effect (p<0.0001) for location, even considering
interaction effects. When interaction terms were not considered, race and
sex were both significant. After interaction terms were included in the
model, these significant main effects for race and sex disappeared. When
Census Division was used as the location factor, Census Division was highly
significant (p<0.0001) for the main effects of the all possible interaction
terms model. Race was the only other significant effect in either model
involving Census Division, and there were no significant interaction terms.
In addition to examining the significance levels of the individual F
ratios, the fit of the model to the data overall should be considered. The
2
value of R is one indication of the appropriateness of the model to de-
2
scribe the data. In general, the larger the value for R , the better the
model is as a description of the relationship of the dependent variable (in
this case, presence or absence of mirex residue in human tissue) to the
2
independent factors (age, race, sex, and location). R can range from zero
-------
-42-
Table 4.5 Regression Model for Race, Age, Sex, and State:*
Zero/Positive Residue Values
Source
df
F Value
**
Pr>F
Significance
Model 11
Error 612
Corrected Total 623
State
Race
Age
Sex
7
1
2
1
Model 73
Error 550
Corrected Total 623
12.92
18.89
.48
,45
,64
2.46
0.0001
0.0001
0.006
NS
0.057
0.0001
0.188
0.246
State
Race
Age
Sex
St x R
St x A
St x Sx
R x A
R x Sx
A x Sx
St x R x A
St x R x Sx
St x A x Sx
R x A x Sx
7
1
2
1
7
13
7
2
1
2
10
7
11
2
6.85
1.12
1.18
0.21
0.80
0.37
0.21
0.60
0.00
0.58
1.02
0.89
0.37
0.53
0.0001
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
Calculated at RTI - unweighted and ignoring design effect.
Probability that this might occur by chance under the assumption of
no difference among categories.
-------
-43-
Table 4.6 Regression Model for Race, Age, Sex, and
Census Division: Zero/Positive Residue Values'"
Source
Model
Error
Corrected Total
Census Division
Race
Age
Sex
Model
Error
Corrected Total
Census Divsion
Race
Age
Sex
CD x R
CD x A
CD x S
R x A
R x S
A x S
CD x R x A
CD x R x S
CD x A x S
R x A x S
CD x R x A x S
df
6
617
623
2
1
2
1
34
589
623
2
1
2
1
2
4
2
2
1
2
4
2
4
2
3
F Value
11.45
29.86
8.05
1.74
2.54
2.73
20.07
6.82
0.46
0.78
1.10
0.89
0.59
0.19
0.50
1.11
0.98
0.23
0.35
0.41
1.64
Pr>F** 2
Significance R
0.0001 0.100
0.0001
0.005
NS
NS
0.0001 0.136
0.0001
0.0009
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
Calculated at RT1 - unweighted and ignoring design effect.
Probability that this might occur by chance under the assumption of
no difference among categories.
-------
-44-
2
to one; a good fit is indicated by values for R that are greater than 0.4.
2
The value for R within a series of models of the same variable will always
increase as more interaction or quadratic terms are added. This change in
2
the magnitude of R can be used to determine the relative contribution of
2
an addition or deletion of a term to the model. In this case, the R
values for the 4 models fitted are small; the two models including State
2
have higher R values than do the models including Census Division as the
geographic boundary of interest. This is logical, as the Census Division
boundaries group States on a proximity basis that does not necessarily
reflect concentrated mirex treatment area divisions. If percentage of a
State's area treated by mirex were considered, it would be more reasonable
to group Louisiana and Mississippi together for analysis than the Census
Division grouping of Louisiana and Texas.
4.2.3 Quantifiable Positive Values
The summary statistics for the 123 quantifiable positive values are
displayed in Tables 4.7 and 4.8. The data in these two tables are the
summary statistics for both the Iog10 of the raw residue amount and the
log of the lipid extractable material adjusted residue amount (Log Y).
Both of these values were analyzed in an unweighted regression analysis.
The model for this analysis is similar to the model used to detect differ-
ences in the presence or absence of mirex in the entire sample (624 observa-
tions) .
•i
The unweighted regression analysis did not show any significant differ-
ences either in a simple main effects model or in examination of interaction
terms in the model. This indicates that if there are any statistically
significant differences for the level of concentration of mirex residues
among the categories of age, race, sex, and location of selection sites,
these difference do not fit a linear model.
-------
-45-
Table 4.7 Summary for Quantifiable Positive Residue Values:
Geometric Means and Standard Errors*
Factor
All
AGE
CENSUS
RACE
SEX
STATE
Factor
Levels
0-14
15-44
>45
East South Central
South Atlantic
West South Central
Caucasian
Non-Caucasian
Male
Female
Alabama
Florida
Georgia
Louisiana
Mississippi
North Carolina
South Carolina
Texas
N
123
8
47
68
61
23
34
86
37
65
58
3
30
14
39
58
0
1
0
LOGY
Geometric Mean (ppm)
0.319
0.226
0.290
0.356
0.278
0.368
0.365
0.299
0.372
0.362
0.278
0.340
0.262
0.419
0.365
0.275
-
0.897
~
Standard Error
0.029
0.084
0.041
0.044
0.035
0.098
0.050
0.032
0.065
0.049
0.033
0.280
0.116
0.148
0.050
0.035
-
-
—
Calculated at RTI-unweighted and ignoring design effect.
-------
-46-
Table 4.8 Summary for Quantifiable Positive Residue Values;
and Standard Errors'"
Geometric Means
Factor
All
AGE
CENSUS
RACE
SEX
STATE
Factor
Levels
0-14
15-44
>45
East South Central
South Atlantic
West South Central
Caucasian
Non-Caucasian
Male
Female
Alabama
Florida
Georgia
Louisiana
Mississippi
North Carolina
South Carolina
Texas
N
123
8
47
68
61
23
34
86
37
65
58
3
30
14
39
58
0
1
0
LOG 0 Residue Amount
Geometric Mean (ppm)
0.220
0.117
0.205
0.250
0.190
0.237
0.267
0.211
0.245
0.239
0.201
0.290
0.177
0.269
0.267
0.186
-
0.420
™
Standard Error
0.011
0.019
0.012
0.013
0.010
0.026
0.016
0.010
0.017
0.013
0.010
0.106
0.033
0.039
0.016
0.010
-
-
•
Calculated at RTI-unweighted and ignoring design effect.
-------
-47-
4.2.4 Unweighted Analysis Conclusions
For this section, the results do not generalize to the target popula-
tion but reflect only the actual sample selected. Within this group, it is
observed that the most important factor in an individual showing rairex
residues in adipose tissue is the location of the selection site. In this
sample, there were no significant differences discovered differentiating
any of the groups analyzed among the 123 observations with measurable
levels of mirex.
-------
-48-
5. WEIGHTED ANALYSIS
5.1 Introduction
The analyses presented in this chapter are weighted to produce popula-
tion estimates for the target population. Quasi selection weights were
calculated for each observation and used in estimating population totals.
RTI has developed several software programs specifically for weighted
analysis of sample survey data. These programs are used to produce the
summary statistics and regression analyses [7,8].
5.2 Weight Formation
Weights were calculated separately for each site, as described in
section 2.4. The observations within each site have the same selection
weight, with no adjustments for age, race, and sex distribution within the
site. The quotas established for each site were intended to create a
self-weighting sample with respect to age, race, and sex distribution.
These quotas were based on the demographic distribution of the living
population of the Census Divisions in which the sample sites were located,
rather than on the distribution of the population within the individual
sites. These quotas were not achieved within site, although on a division
level the desired proportions of race and sex were approximately met (see
Table 4.1). The age distribution was not met, and this undoubtably reflects
the source of the selection population. The population that is young is
not proportionately represented in the segment of the population that
undergo surgery or die in a large hospital.
After weights were calculated and added to the data file for each
observation, the weighted analyses for the percent positive values and the
quantifiable positives were carried out.
-------
-49-
5.3 Estimates of Positive Mirex Detection
The parameters and statistical tests of interest are the same for the
weighted and unweighted analyses. The use of weights to produce population
based estimates affects the calculation of statistics and thus allows
generalization of the statistics to the target population.
A weighted chi-square test was used to examine the estimated proportions
of the target population with detectable mirex residues. The weighted
chi-squares were produced by a weighted chi-square SAS procedure [9]. This
produced an inflated statistic. To adjust for this inflation before evaluat-
ing the resulting statistic, the computed chi-squares were multiplied by
the sampling fraction and then divided by the design effect. The sampling
fraction is the number of specimens in the sample divided by the target
population estimate. In this case, the sampling fraction = 624/23,371,372
= 2.67 x 10* . The design effect (DEFF) is the inverse of the ratio of the
variance that would be obtained for a simple random sample to the variance
for the design employed.
The results of this analysis are shown in Table 5.1. State was the
only significant factor for differences in the proportion of the target
population that has some detectable mirex residue in adipose tissue.
Mississippi was the state that had a statistically significantly larger
proportion of positive mirex detection.
A Taylorized weighted least squares analysis was used, which incorpor-
ates .the sampling structure into the regression model calculations [8].
This analysis will not compute valid F statistics if the number of PSU's is
smaller than the combined levels of the factors in the model. In the case
of a model which included State as the location factor, there are 8 levels
for the factor State. When interaction terms are included in this model,
-------
-50-
Table 5.1 Weighted Chi-Square Analysis Results for Zero/Positive Values:
Geographic and Demographic Variables*
% Estimated Positive 2
Variable Estimates
AGE
0-14
15-44
>45
CENSUS DIVISION
East South Central
South Atlantic
West South Central
RACE
Caucasian
Non-Caucasian
SEX
Male
Female
STATE
Alabama
Florida
Georgia
Louisiana
Mississippi
North Carolina
South Carolina
Texas
of n
6,606,385
7,529,303
9,235,645
1,727,307
9,726,438
11,917,587
18,517,425
4,853,897
8,653,061
14,718,264
503,079
2,477,302
5,642,299
3,229,595
1,224,229
210,692
1,396,152
8,687,994
within each X df
level
1.6 2.72 2
11.3
15.4
37.6 2.6 2
13.3
3.6
9.8 0.04 1
11.5
16.1 1.8 1
6.7
4.6 32.8 7
5.4
19.6
13.4
51.1
0.0
3.9
0.0
Significance
level (p)**
0.257
0.273
0.844
0.180
0.0001***
Calculated at RTI, estimate of N = 23,371,332.
** 2
Probability that a X this large or larger might occur by chance, under
the assumption of no difference among categories.
Over 5% of the original cells have expected counts less than 5. Sparse table
indicates chi-square may not be a valid test.
-------
-51-
the sum of levels in the analysis rapidly increases beyond the limits
imposed by the sampling structure. Consequently, only main effects models
involving State as the location variable were computed at this time.
The interpretation of the weighted regression model requires some
caution. The presence of significant main effects can be misleading in
some cases. It is important to consider possible interaction effects in
the model as well. Significant main effects in the presence of significant
interaction terms involving the same factors that are in the main effects
are not interpreted independent of the interaction terms.
The results for a weighted regression analysis for the estimated
presence or absence of mirex when Census Division is the location of interest
are in Table 5.2. A model containing all possible two-way interactions
produced a significant sex and Census Division interaction (p<0.017).
The model containing State was limited to a simple main effects model,
due to the restrictions noted above. Results for this model are shown in
Table 5.3. The significant main effect for State (p<0.0001) in this analysis
confirms the results of the weighted chi-square, which also shows State to
be a significant factor in differences in proportion of the population with
positive mirex residues. The main effect for sex in this model is marginally
significant (p<0.109). There may be some interaction between sex and
State; however, this was not tested.
5.4 Estimates of Quantifiable Positive Values
The estimates of the geometric means and standard errors for the log
of the lipid extractable material (LEM) adjusted residue amounts and the
the log,0 residue amount are in Tables 5.4 and 5.5. These variables were
also analyzed by the weighted regression technique described in section
5.2.
-------
-52-
Table 5.2 Regression Model for Race, Age, Sex, and
Census Division: Zero/Positive Residue Values*
Degrees of Pr>F**
Source freedom F Value Significance
Overall Model 6 9.30 0.0001
Census Division 2 3.53 0.042
Race 1 0.14 NS
Age 2 1.93 0.162
Sex 1 3.03 0.092
Overall Model 19 46.41 0.0001
Census Divsion
Race
Age
Sex
CD x R
CD x A
CD x S
R x A
R x S
A x S
2
1
2
1
2
4
2
2
1
2
***
«nt"&
•ff^n
'&'M\
0.26
1.13
3.45
0.28
0.11
1.22
NS
NS
0.044
NS
NS
NS
*
Calculated at RTI.
**
Probability that this might occur by chance under the assumption of
no difference among categories.
ff
Not testable in the presence of interactions.
-------
-53-
Table 5.3 Weighted Regression Model for Race, Age,
Sex, and State: Zero/Positive Residue Values*
Source
Overall Model
Total
State
Race
Age
Sex
Degrees of
freedom
11
30
7
1
2
1
F Value
11.38
16.11
0.23
0.91
2.72
Pr>F**
Significance
0.0001
0.0001
NS
NS
0.109
Calculated at RTI.
**
Probability that this might occur by chance under the assumption of
no difference among categories.
-------
-54-
Table 5.4 Estimated Summary Statistics for Quantifiable Positive Residue Values:
Weighted Geometric Means and Standard Errors*
Factor
All
AGE
CENSUS
RACE
SEX
STATE
Factor Estimate of LOGY
Levels N** Geometric Mean (ppm)
0-14
15-44
>45
East South Central
South Atlantic
West South Central
Caucasian
Non-Caucasian
Male
Female
Alabama
Florida
Georgia
Louisiana
Mississippi
North Carolina
South Carolina
Texas
1,919,220
88,474
504,045
1,326,701
621,846
924,120
373,254
1,427,135
492,085
1,033,219
886,001
23,193
132,962
764,235
373,254
598,654
***
26,923
TrTTnT/t
0.286
0.172
0.283
0.297
0.292
0.266
0.331
0.252
0.406
0.322
0.249
0.340
0.298
0.250
0.331
0.290
***
0.897
***
Standard Error
0.048
0.037
0.039
0.077
0.018
0.083
0.049
0.030
0.102
0.085
0.023
^nWf
0.105
0.085
0.049
0.019
^r*Wr
•WrtSr
TrrtWf
Calculated at RTI.
Estimate of total population for the region of interest = 23,371,372.
No information in sample on which to base estimate.
-------
-55-
Table 5.5 Estimated Summary Statistics for Quantifiable Positive Residue Values:
Weighted Geometric Means and Standard Errors*
Factor
All
AGE
CENSUS
RACE
SEX
STATE
Factor Estimate of Log.. Residue Amount
Levels N** Geometric Mean (ppm)
0-14
15-44
>45
East South Central
South Atlantic
West South Central
Caucasian
Non-Caucasian
Male
Female
Alabama
Florida
Georgia
Louisiana
Mississippi
North Carolina
South Carolina
Texas
1,919,220
88,474
504,045
1,326,701
621,846
924,120
373,254
1,427,135
492,085
1,033,219
886,001
23,193
132,962
764,235
373,254
598,654
TnrtWif
26,923
irfcn
0.197
0.095
0.192
0.209
0.189
0.187
0.244
0.184
0.240
0.208
0.186
0.290
0.190
0.181
0.244
0.185
^WSrtir
0.420
^TrtTnT
Standard Error
0.010
0.009
0.012
0.017
0.003
0.017
0.014
0.011
0.020
0.018
0.006
^SrjWr
0.030
0.018
0.014
0.003
T-WrsV
^WWr
«"«Wif
Calculated at RT1.
Estimate of total population for the region of interest = 23,371,372.
No information in sample on which to base estimate.
-------
-56-
The variable of most interest is the LEM adjusted residue amount;
three models involving this variable are listed in Table 5.6. The first
two models in the table are main effects models, one with State as the
location factor and one with Census Division. Neither model has any signi-
ficant main effects. The third model in Table 5.6 contains all possible
second order interactions. There are marginally significant interactions
for Census Division by age group (p<0.069) and for Census Division by race
(p<0.0106). This indicates that there may be differences in levels of
mirex concentraction among subgroups of the target population. However,
these differences are not straight-forward in the sense that it can be
asssumed that the same subgroup had the highest mirex levels in every
Census Division. For example, Caucasians in the East South Central and
West South Central Census Divisions have higher levels of mirex concentra-
tions than do Non-Caucasians. In the South Atlantic Census Division Non-
Caucasians have higher mirex residue levels than do Caucasians.
The same three models were analyzed for the unadjusted residue amount
values. These results are in Table 5.7. There is a significant effect for
State (p<0.0027) and for age group (p<0.0138) when State is the location
tested. In the second order interaction model, with Census Division as the
location factor, the interactions are similar to those obtained when the
LEM adjusted residue amount is the dependent variable in the same model.
5.5 Weighted Analysis Conclusions
The two weighted analyses are conducted so that inferences can be made
about the population of the mirex treated areas. The use of two different
analyses provides a check on the test results.
The chi-square statistic is relatively assumption-free, but it is less
sensitive to differences than regression analysis. Regression analysis, in
-------
-57-
Table 5.6 Weighted Regression Model for Location, Age, Race, and
Sex: Quantitative Positive Residue Amounts*
Source
Degrees of
freedom
F Value
Pr>F**
Significance
Overall Model*** 9
State 5
Age 1
Race 1
Sex 2
Overall Model 6
Census Division 2
Age 2
Race 1
Sex 1
Overall Model*** 18
7.42
1.34
0.66
1.65
2.22
6.25
0.85
2.02
2.25
0.77
129409
0.0001
NS
NS
NS
0.1322
0.006
NS
0.1561
0.1481
NS
0.0001
Census Division
Age
Race
Sex
CD X A
CD X R
CD X S
A X R
A X S
R X S
2
2
1
1
3
2
2
2
2
1
Not Testable
Not Testable
Not Testable
Not Testable
2.72
5.63
0.43
0.55
0.50
2.17
0.0690
0.0106
NS
NS
NS
0.155
Calculated at RTI: dependent variable in model is the LEM adjusted
residue amount.
Probability that this might occur by chance under the assumption of
no difference among categories.
Full higher order interaction models not estimable.
-------
-58-
Table 5.7 Weighted Regression of Quantifiable Positive
Residue Amounts: Location, Age, Race, and Sex*
Source
Degrees of
freedom
F Value
Pr>F**
Significance
Overall Model***
34.18
0.0001
State
Age
Race
Sex
5
1
1
2
.20
.24
.34
0.01
0.0027
0.0138
NS
NS
Overall Model
13.70
0.001
Census Division 2
Age 2
Race 1
Sex 1
Overall Model***
1.31
4.83
1.70
0.02
NS
0.0182
NS
NS
Census Division
Age
Race
Sex
CD X A
CD X R
CD X S
A X R
A X S
R X S
3
2
2
2
2
1
Not Testable
Not Testable
Not Testable
Not Testable
4.03
6.59
0.84
1.34
0.43
1.70
0.020
0.0057
NS
NS
NS
NS
^Calculated at RT1. Variable Iog10 residue amount.
Probability that this might occur by chance under the assumption of
no difference among categories.
***Models containing higher level interactions not computed.
-------
-59-
turn, requires more assumptions to be met about the underlying structure of
the data before it can be considered an appropriate test.
It appears that location is an important variable in the presence or
absence of mirex in human adipose tissue. Mississippi is the State which
shows the largest proportion of presence of mirex in human adipose tissue.
There is less clear cut evidence that there are any large differences in
level of residue amount in any particular category studied, although margin-
ally statistically significant results are found for several categories.
-------
-60-
REFERENCES
[1] Kutz, F. W., and Strassman, S.C. National Human Monitoring Program
Study Plan for Mirex Special Project. EPA internal document, not
dated.
[2] Commercial Atlas and Marketing Guide, Rand McNally and Company,
New York, New York. 1974.
[3] Hartwell, T. et. al (1979). Analysis of EPA Pesticides Monitoring
Networks. EPA, Office of Toxic Substances, Washington, D.C.,
560/13-79-014.
[4] U. S. Department of Commerce (1972). 1970 Census of Population.
Characteristics of the Population (Volume I, Part A, Section 1
and 2).
[5] Analysis of Pesticide Residues in Human and Environmental Samples,
J. F., Thompson, ed., U. S. Environmental Protection Agency,
Research Triangle Park, NC, 1974, Section 5 A [1].
[6] U. S. Environmental Protection Agency (1979). Analysis of EPA
Pesticides Monitoring Networks. ADDENDA. EPA. Office of Toxic
Substances, D.C. 5601/13-79-014. December, 1979.
[7] Shah, B. V. (June, 1979). SESUDAAN: Standard Errors Program for
Computing of Standardized Rates from Sample Survey Data. Research
Triangle Institute, Research Triangle Park, North Carolina.
RTI/1789/00-01F.
[8] Holt, M. M. (May 1977). SURREGR: Standard Errors of Regression
Coefficients From Sample Survey Data. Research Triangle Institute,
Research Triangle Park, North Carolina.
[9] Barr, Antony J., et al. (1979). A User's Guide to SAS-79, SAS
Institute, Inc., North Carolina.
-------
A-l
APPENDIX A
State Maps Showing Counties Where Mirex Was Applied
At Some Time During 1965-1974
-------
A-2
Shaded areas had some application of mirex during the
period 1965-1974
-------
A. iBAMA ^3
Counties, Standard Metropolitan Statistical Areas, and Selected Places
TT___
L-
N«l
B
LEGEND
(•) Places of 100,000 or more inhabitants
• Places of 50.000 to 100,000 inhabitants
O Places of 25,000 to 50,000 inhabitants outside SMSA's
f] Standard Melropolitan
Statistical Areas (SMSA's)
M
-------
LEGEND
(•i Ptjces o( 100,000 or more inhjbrunll
• Putt-. 0( 50.000 10 100.000 ifiluruunl!
O Pljcei ol 25.OOO lo W.OOO lnhJ^.unli oullidt SMSA »
| !>t jrulif J Metropolitan
St«t'Shc«l Are» ISMSA'%)
VV™'
/.°"r"" _,
WEST PALM BEACH
FORT LAUOERDALE HOLLYWC
0>
to
CO
ff
Q.
CD
•-*-
o
"O
Q
oT
CO
ST
«-*-
55"
o"
5L
>
fD
Q.
CO
5.
S
i-t-
(D
Q.
CO
O
o>
10
11
12
13
14
15
-------
EORGIA
A-5
Counties, Standard Metropolitan Statistical Areas, and Selected Places
LEGEND
(•) Places ol 100,000 or more inhabitant
• Places of SO.000 to 100.000 mhabitai
O Places ol 25,000 to 50.000 inhabitant
ints
50,000 to 100.000 inhabitants
25,000 to 50,000 inhabitants outside SMSA's
Standard Metropolitan
Statistical Areas (SMSA's)
-------
10
11
12
13
LEGEND
® Pl«cn of 100.000 or more inhabitants
• PUCM of M.OOO to 100,000 inhabitant!
O Placti of 25,000 10 iO 000 inhabitants oulside SMSA's
SUndird Metropuhlan
-. C,M':,A ..)
>« CH««fS • „
OIOSIIU • 1).
LAKE CHARLES
co
Q.
CO
*-*-
&
33
Q>
O
-------
I^'SSISSIPPI
A-7
Counties, Standard Metropolitan Statistical Areas, and Selected Places
7 8
10
. DC SOTO
\
J^-x_y —
BENTON 1 ALCORN
MARSHALL
1
,-T
PRENTISS
V
o
j
l/
LEGEND
® Places of 100,000 or more inhabitants
Q Central cities of SMSA's with fewer than 50,000 inhabitants
O Places of 25,000 to 50,000 inhabitants outside SMSA's
Standard Metropolitan
Statistical Areas (SMSA's)
M
10
-------
10 11
12
13 14
15
16 17 18
C/)
r-^
0)
Q.
Q)
B
H
GREENSBORO-WINSTON-SAL
LEGEND
® Places of 100.000 or more inhabitants
• Places ot 50.000 to 100.000 inhabitants
O Central cities of SMSA's with fewer than 50,000 inhabitants
O Places ol 25.000 to 50.000 inhabitants outside SMSA's
jg$ %m
D
H
O>
r-+>
o
o
cu
Cr>
Q)
rj
Q.
(D
r>
*— *•
a
Q)
O
CD
to
Standard Metropolitan
Statistical Areas (SMSA'sl
O 10 20 10 40 SO Mil IS
10 11 12 13 14 15 16 17
18
-------
10
11
12
13
CHARLESTON
LEGEND
® Places ol 100.000 or more inhabitants
• Places ol 50.000 lo 100,000 inhabitants
O Places ol 25.000 lo 50,000 inhabitants outside SMSA's
Standard Metropolitan
s-;-.-l Statistical Areas (SMSA's)
00 10 «0 30 40 MILES
I
#-4-
CD"
CO
ST
Q.
O)
CD
i-t-
O
f
I
CO
«-*
U)
^*
Crt'
i-t-
o"
CL)
0)
CD
cn
cu
3
Q.
CO
-------
10 11 12 13
14
16 17
-»<" "S.i. •—> '"" "" [-"^V
-T -"T" ~ T ..L.^SP/1'
->J""'"; "'-»"> K>" w°"f ••)«rj'»ii.
0" 0'iM
iBEAUMONT
PORT ARTHUR
Z ORANGE
„\ V
LEGEND
(•) PUcts ot 100 000 or mo>e .n
• PUces of MJ.OOO 10 IOO.OOO
O Centrjl Citm o< SMS* i with l«wer \tttn 5Q 000
O Places ol 25.000 lo 50.00O mnjb.ums ounme SMS* s
i Slandird Metropolitan
latiitKal Areas (SHSA s|
CO
<-»•
CD
3
Q.
Q>
a
3
ta
CD
0)
CO
CD
SL
o>
n
r^-
CD
Q.
TJ
Q>
O
•Ntj(N BROWNSVILLE HARLINGEN SAN BENITO
10 12 13 M IS 16 17
-------
B-l
APPENDIX B
Counties Where Mirex was Applied at Some Time During 1965-1974
Note: Population values based on 1970 Census Data
-------
B-2
Alabama
Total Population 3,444,165
All counties except
Colbert
Lauderdale
Lawrence
Limestone
Madison
Morgan
Marshall
Jackson
DeKalb
Cherokee
Mirex Area Popultation
Population
49,632
68,111
27,281
41,699
186,540
77,306
54,211
39,202
41,981
15.606
601,569
2,842,596
Florida
Total Population 6,789,443
All counties except
Monroe
Dade
Collier
Breward
Hendry
Palm Beach
Glades
Martin
St. Lucie
Indian River
Population
52,586
1,267,792
38,040
620,100
11,859
348,753
3,669
28,035
50,836
35.992
2,457,662
Mirex Area Population
4,331,781
-------
B-3
Georgia
Total Population 4,589,575
All counties except
Dade
Walker
Catoosa
Whitfield
Murray
Fannin
Gilmer
Union
Towns
Rabun
Habersham
White
Lmnpkin
Dawson
Pickens
Gordon
Chattooga
Floyd
Polk
Haralson
Carroll
Douglas
Paulding
Bartow
Cherokee
Forsyth
Hall
Banks
Stephens
Franklin
Hart
Elbert
Madison
Jackson
Barrow
Walton
Morgan
Oconee
Clarke
Oglethorpe
Wilkes
Lincoln
Columbia
McDuffie
Warren
Taliaferro
Greene
Hancock
Glascock
Population
9,910
50,691
28,271
55,108
12,986
13,257
8,956
6,811
4,565
8,327
20,691
7,742
8,728
3,639
9,620
23,570
20,541
73,742
29,656
15,927
45,404
28,659
17,520
32,663
31,059
16,928
59,405
6,833
20,331
12,784
15,814
17,262
13,517
21,093
16,859
23,404
9,904
7,915
61,177
7,598
10,184
5,895
22,327
15,276
6,669
2,423
10,212
9,019
2,280
Jefferson
Burke
Jenkins
Emanuel
Candler
Mirex Area
Population
Population
17,174
18,255
8,332
18,189
6.412
1,031,614
3.557.961
-------
B-4
Louisiana
All counties mirex treated
Total Population 3,641,306
Population
3,641,306
Mississippi
Total Population 2,216,912
All counties except
Sunflower
Leflore
Humphreys
Sharkey
Issaquena
Bolivar
Coahoma
Quitman
Tallahatchie
Tunica
Panola
Tate
DeSoto
Marshall
Benton
Tippah
Alcorn
Prentiss
Tishomingo
Carroll
Holmes
Attala
Clarke
George
Stone
Pearl River
Hancock
Population
37,047
42,111
14,601
8,937
2,737
49,409
40,447
15,888
19,338
11,854
26,829
18,544
35,885
24,027
7,505
15,852
27,179
20,133
14,940
9,397
23,120
19,570
15,049
12,459
8,101
27,802
17.387
566,148
Mirex Area Population
1.650,764
-------
B-5
North Carolina
Total Population 5,082,059
New Hanover
Brunswick
Columbus
Robeson
Bladen
Duplin
Onslow
Lenoir
Jones
Craven
Pamlico
Cateret
Population
82,996
24,223
46,937
84,842
26,477
38,015
103,126
55,204
9,779
65,554
9,467
31.603
Mirex Area Population 578,223
South Carolina
Total Population 2,590,516
Marlboro
Dillon
Marion
Horry
Georgetown
Williamsburg
Florence
Clarendon
Suoiter
Kershaw
Fairfield
Richland
Calhoun
Orangeburg
Berkeley
Charleston
Dorchester
Colleton
Bamberg
Barnwell
Aiken
Allendale
Edgefield
Lexington
Hampton
Jasper
Beaufort Point
Population
27,151
28,838
30,270
69,992
33,500
34,243
89,636
25,604
79,425
34,727
19,999
233,868
10,780
69,789
56,199
247,650
32,276
27,622
15,950
17,176
91,023
9,692
15,692
89,012
15,878
11,885
51.136
Mirex Area Population 1,469,023
-------
B-6
Texas
Total Population 11,196,730
Denton
Collin
Dallas
Bowie
Harrison
Smith
Orange
Jefferson
Polk
San Jacinto
Walker
Madison
Grimes
Harris
Galveston
Brazoria
Fort Bend
Waller
Austin
Washington
Wharton
Lavaca
DeWitt
Victoria
Bexar
Nueces
Montgomery
Brazos
Population
75,633
66,920
1,327,321
67,813
44,841
97,096
71,170
244,733
14,457
6,702
27,680
7,693
11,855
I,74lj912
169,812
108,312
52,314
14,285
13,831
18,842
36,729
17,903
18,660
53,766
830,460
237,544
49,479
57.978
Mirex Area Population 5,485,781
-------
C-l
APPENDIX C
Definition of Mirex Treated Areas
-------
C-2
Definition of Mirex-Treated Areas
The definition of mirex-treated areas for the MSS is global; a mirex
treated county is a county including an area having any application of
mirex during a 10-year period. This definition does not take into consider-
ation factors of repeated applications, amount of active substance displaced,
or land area covered.
Before being banned from sale and use in 1977, mirex was sold over the
counter to private individuals and applied by both State action agencies
and the States and Federal Government working together. It should be
possible to obtain mirex use information from individual State action
agencies and from the United States Department of Agriculture's Animal and
Plant Health Inspection Service (USDA, APHIS) which would help answer some
of the above questions about the frequency and intensity of mirex app-
lication. For example, this kind of historical information about mirex
application in the State of North Carolina could be gathered from existing
records in the North Carolina Department of Agriculture's Pest Control
Division.- RTI has information from the APHIS of USDA concerning the
number of mirex-treated acres by State; this only includes those acres
treated from 1964 to 1973 by the Federal Government in connection with the
2/
individual States.- Additional mirex treatment must have varied from
State to State, dependent upon the magnitude of the fire ant problem and
the resources of the State.
- Howard M. Singletary, North Carolina State Department of Agriculture,
Pesticide Division. Personal communication, May 8, 1980. C. Leininger.
- Table of Aggregate Acres Treated, provided to EPA by APHIS.
Revised date 8/1/73.
-------
C-3
In an epidemiological study such as this, it is still more important
to obtain specific residence information about the individuals from whom
specimens are collected than a better definition of the area in which the
individuals are assumed to reside. At this time it would be inconclusive
to attempt to correlate more detailed information about mirex use with this
human adipose tissue data.
-------
D-l
APPENDIX D
MSS Sample Selection Method
-------
D-2
MSS Sample Selection Method
D.1 Introduction
In order to calculate weights that can be used in making population
estimates from sample data, the method used to select the sample must be
known. This appendix is a discussion of the method used to select the
sample RMA's for the MSS. For the NHMF, the method used for selection of
sites was probability proportional to the size of the site's population
(PFS). There was some question about the method used to select sites for
the MSS. The selection procedure was checked by calculating the expected
outcome of selecting 32 out of 60 RMA's by both PPS and simple random
sampling (SRS). It was concluded that SRS was the method used.
D.2 The NHMP and Probability Proportional to Size
The method of sample selection used in the NHMP was selection with
probability proportional to size (PPS). This means that the population of
a city was a factor in the possibility of the city being selected in a
sample. PPS sampling with equal-size final stage samples results in a
self-weighting sample. When n areas are to be selected with PPS, the
selection probability for a given area is calculated by dividing n times
the population of that area by the total population of all the areas in the
sampling frame. This is expressed as
i ^- v. i-.-i.^ />nr.r.\ f -i n population of area i
selection probability (PPS) of area i' = —K—^
N
I population of area i
i=l
where N is the number of areas in the sampling frame.
-------
D-3
NHMP cities were selected by PFS, within strata defined by Census
Divisions. As discussed in Chapter 2, the 8 NHMP cities located in the
mirex area were included in the MSS. This was an effort to use the exist-
ing tissue collection network. The MSS supplementary sample was selected
from 60 RMA's not in the NHMP. The majority of the existing documentation
of the MSS indicates that SRS was used to select the 32 MSS sites.
D.3 Simple Random Sampling
In SRS, any RMA has the same chance of being selected as any other RMA
in the list of 60 RMA's in the mirex area. A listing of the eligible sites
might be made and a consecutive number assigned to each site. A sample of
sites can then be selected by drawing numbers (sites) from the list using a
table of random numbers. This is the method documented on page 3 of the
NHMP study plan for the MSS- . There was one instance of conflicting MSS
documentation, suggesting PPS selection for the MSS supplementary sites.
This was sufficient reason to investigate the sample selection method.
If SRS was used to choose the 32 RMA's, then any one RMA was as likely
as any other to be chosen. To test this, the 60 RMA's were first ordered
by population size, and then divided into 4 groups, each with 15 RMA's.
There were 32 RMA's actually selected. Under SRS it would be expected that
approximately the same number of RMA's would be selected from each of the 4
divisions that have been arbitrarily made in the ordered listing. This can
be seen more easily in the line chart below, where the 4 segments or groups
are indicated by Roman numerals, with a subscript "s1 or 'f to indicate
membership in the sample or in the frame.
II 8 II 16 III 24 IV 32 = sample RMA's
s s s s
15 IIf 30 IIIf 45 IVf 60 = frame RMA's
-------
D-4
In repeated sampling, the average proportion of selected RMA's from a
segment of the frame will be one-fourth. Since it is a random process,
there will be some variation in the number of selected RMA's from a segment
of the frame. A 95 percent confidence interval (C.I.) will give the range
of values that will be encountered 95 out of 100 times a given sample se-
lection procedure is used. The expected number of sites selected by SRS in
any of the 4 segments of the frame is one-fourth of 32, or 8. To find a 95
percent confidence interval for the number of sites chosen from each
segment, computations were based on the variance of the number of sites
selected per segment and the properties of the normal curve. These cal-
culations are given in detail in section D.6. For this case, where 8 is
the expected number of sites selected from a segment, the 95 percent C.I.
is 8 ± 1.96/T" = 8 ± 4.8. The number of sites selected under SRS can
range from 3.2 to 12.8 sites within a 95 percent C.I. This agrees with the
actual distribution of sites among the arbitrary segments, as indicated in
the chart below.
I II III IV
s s s s
Expected number
of sites (SRS) 8 ± 4.8 8 ± 4.8 8 ± 4.8 8 ± 4.8
and the 95% C.I.
Number of sites 8 5 8 11
actually in the MSS
D.4 Probability Proportional to Size
The same arbitrary segments were created, and 95 percent C.I.'s for
the number of sites selected from each segment were calculated, assuming
selection PPS. The selection rates varied from segment to segment, because
-------
D-5
the combined population of the areas within a segment determined the sam-
pling rate for that segment. The combined population for each of the four
divisions varied considerably: IF = 730,600, HF = 1,159,800, IIIj. =
2,124,100, and IV., = 6,779,200, for a frame total of 10,793,700. The
T
selection probabilities also varied accordingly, from Pr(I ) = 0.07 to
s
Pr(IV ) = 0.63. C.I.'s for the number of selected sites from each segment
S
under PPS were computed and are displayed in the chart below.
Js Hs HIs IVs
Expected number
of sites (PPS) 2.2 ± 2.8 3.4 ± 3.5 6.3 ± 4.4 20.1 ± 5.4
and the 95% C.I.
Actually selected 858 11
for MSS
The number of sites actually selected in 2 of the 4 segments falls
outside the 95% C.I. for number of selected sites from the segment when
selection is PPS.
D.5 Conclusions
Based on the computations described in Section D.3 and D.4, it appears
that the MSS sample sites were selected by simple random sampling. There-
fore, the selection rates used in computing the quasi selection weights
were calculated under the assumption of SRS. This agrees with the majority
of the documentation for the MSS.
D.6 Calculations
This section contains a detailed description of the calculation of
confidence intervals for the number of sites selected per segment when
selecting 32 of 60 RMA's by SRS. The method used to determine C.I.'s when
-------
D-6
using PPS sampling is similar to that used in SRS, and will be described
briefly.
Confidence intervals are calculated using the properties of the normal
curve and the variance of a random variable. In this case, the random
variable, which can be called y, is the number of sites that will be se-
lected from a given segment. The distribution of y is binomial. Variances
are calculated for binomial random variables (in this case, the Bernoulli
event is that a site is selected in a given segment or it is not selected
in that segment) by multiplying the total number of sites selected by the
probability (p) that the selected site is in the segment by the probability
(q) that it is not. Thus, the variance of y can be expressed as V(y) =
npq.
The probability that a selected site falls in a given segment is 8/32
= 1/4 = p. The value of q is then 24/32 or 3/4. The variance will be the
same for the number of sites selected in each segment: 32(1/4)(3/4)=6. To
find the 95 percent C.I. for y (where y is the number of sites selected for
a segment) requires multiplying the square root of the variance (this is
the standard deviation) by 1.96. The constant 1.96 is a value (called a z
score) which corresponds to 95 percent of the area under the normal curve.
Other constants, which correspond to different percentages of the area
under the normal curve, can be found in a table of z scores in any standard
statistical text. In this case, where the probability that a selected site
falls in the given segment is the same for all segments, the 95 percent
C.I. is 8 ± 1.96/1T = 8 ± 4.8 for each segment.
-------
D-7
To find the 95% confidence interval for number of selected sites
falling in a given segment, assuming PPS, the same procedure is followed as
for SRS, except that the probability of a selected site falling in a given
segment is different for each segment. This reflects the difference in the
population size of the sites in the frame. To calculate the expected
number of sites in a segment, the probability of selection for a segment is
multiplied by the number of sites desired in the final sample. As an
example, for segment If, with Pr(selection) = 730,600/ 10,793,700 = 0.07,
the expected value of y = (0.07)(32) =2.2. The variance of y was
calculated separately for each segment, and the C.I.'s computed for each
segment as was done when assuming SRS.
- Kutz, F. W. and Strassman, S. C. (not dated). National Human
Monitoring Program. Study Plan for Mirex Special Project. EPA
internal document.
-------
E-l
APPENDIX E
List of MSS Collection Sites
-------
E-2
COLLECTION SITES
Alabama
Anniston
0 Birmingham
Dothan
°*Mobile
0 Montgomery
* Tuscaloosa
Florida
Fort Myers
Fort Walton Beach
Lakeland
Orlando
°*Panama City
Sarasota
0 St. Petersburg
Tallahassee
°*Tampa
Georgia
Atlanta
0 Brunswick
Columbus
Macon
Savannah
x Valdosta
Louisiana
Alexandria
0 Lafayette
Lake Charles
Monroe
* New Orleans
Shreveport
Albany, Georgia
Opelika, Alabama
Melbourne, Florida
Bryan, Texas
RMA Population-/
93,100
679,000
51,200
312,500
201,000
106,300
83,500
76,500
97,000
473,000
72,000
222,000
625,000
121,500
495,000
1,720,000
54,300
226,700
219,500
184,000
50,100
97,500
128,800
128,500
110,800
1,126,000
282,000
Mississippi RMA Popu
Columbus 43
0 Greenville 51
Hattiesburg 57
Jackson 258
Laurel 41
Meridian 56
Pascagoula 74
North Carolina
Wilmington 91
South Carolina
Charleston 290
Columbia 315
Texas
* Dallas 2,455
* Houston 2,085
* San Antonio 890
lati(
,200
,100
,200
,000
,000
,200
,900
,300
,000
,000
,000
,000
,000
ALTERNATE COLLECTION SITES -/
94,800
46,700
100,000
60,700
Gadsden, Alabama
Jacksonville, Florida
Baton Rouge, Louisiana
Pensacola, Florida
Gulfport, Mississippi
7?
585
361
21/
172
79,000
*
x
o
1
2
National Human Monitoring Site
Replaced by Albany, Georgia, in survey
Nonresponding Site
1974 Rand McNally Population Estimates
Listed in Order of Selection
-------
E-3
RMA Remainder Frame After Selection of
Collection Sites and Alternates
Florida
Cocoa
Daytona Beach
Gainesville
Titusville
Winter Haven
Georgia
Augusta
Louisiana
Houma
New Iberia
North Carolina
Jacksonville
South Carolina
Florence
Sumter
RMA Population -•
87,000
117,000
97,000
39,000
64,800
215,500
53,000
40,500
71,000
49,600
69,600
I/
Texas
RMA Population -'
I/
Beaumont-Port Arthur 226,000
Corpus Christi 257,000
Freeport-Lake Jackson 58,700
Galveston-Texas City 135,000
Orange 51,600
Texarkana 75,000
Tyler 79,200
Victoria 45,400
1 1974 Rand McNally Population Estimates
* Remainder Frame: The sites remaining after sample selection has been
made from the frame of target sites.
-------
F-l
APPENDIX F
Instructions to Pathologists
-------
F-2
Dear :
has been selected as one of 40
representative cities in which the U. S. Environmental Protection
Agency would like to conduct a special research study to determine
the frequency of finding residues of the insecticide, mirex in
humans. This letter is an invitation to you to participate
in this study. This project is an activity of the National
Human Monitoring Program for Pesticides and will be ongoing for
the next 6 to 9 months.
Mirex is an organochlorine insecticide which has been used
for the control of the imported fire ant in large areas of the
southern United States. Our information indicates that over
14 million acres have been treated to date. Mirex is extremely
resistant to chemical degradation and thus its persistence enables
it to accumulate in certain organisms in the food chain. As you
will note from the enclosed reprint, mirex has been found in
humans. We now need to make a more precise estimate of the
epidemiology of this chemical.
As a mechanism for estimating exposure to mirex, samples of
human adipose tissue will be chemically analyzed in our laboratories.
Reports of findings will be furnished to each participating physician.
Data from these analyses are useful in evaluating various factors
and conditions pertaining to human health and pesticide regulation.
-------
F-3
We will need samples of human adipose tissue from previously
excised surgical specimens and/or postmortem examinations. The
samples should be collected in accordance with the enclosed
guidelines. Each participant is being asked to collect specimens
according to the enclosed age/sex/race quota form. We furnish
all supplies (mailers, forms, bottles, postage-paid address cards,
etc.) and are able to remunerate you or your designee for professional
services at the rate of $10.00 per acceptable sample.
We hope that you will help in this study by providing selected
adipose tissues obtained during routine pathology. We will telephone
you about this in a few days.
Sincerely yours,
Frederick W. Kutz, Ph.D.
Project Officer
National Human Monitoring Program (WH-569)
Enclosures
-------
F-4
INSTRUCTIONS FOR THE COLLECTION OF
HUMAN ADIPOSE TISSUE FOR MI REX
ANALYSIS
Mi rex is an organochlorine insecticide used for the control
of the imported fire ant in large areas of southern United States.
The chemical is extremely resistant to degradation and this
persistence and its chemical characteristics enable it to
accumulate in certain organisms in the food chain.
The objective of this special environmental study is to
determine the incidence and levels of mirex in human adipose
tissue collected from areas which have been treated with this
chemical. The results from the program are used in evaluating
various factors and conditions pertaining to human health and effective
pesticide regulation.
The adipose tissue for this project is secured through the
cooperation of participating pathologists located throughout
southern United States. The tissue is obtained from surgical
specimens previously excised for pathological examination and from
postmortem examinations. The specimens are sent to the program
office in Washington, D. C., from which they are then forwarded to
contract laboratories for chemical analysis. Periodic reports of
the laboratory results are sent to each participating pathologist
for the tissues which were submitted under his auspices. Summaries
comparing results with other regions are also provided as they become
available.
-------
F-5
In order to develop precise epidemiologic data, collections
must be made according to an experimental design which dictates
the number of samples required according to the demographic
distribution of the population of your area. You should have a
copy of the quota of samples expected to be collected from your
location on a fiscal year basis. All collections should be made
according to this age/sex/race distribution. You should be able
to collect the number of samples required in each category in the
next 6 months. If you feel that you will be unable to collect the
number of samples required, please let us know.
Criteria for Selection of Patients to be Sampled
Since the program objective is to reflect pesticide incidences
and levels in the general (man-on-the-street) population, a few
suggestions are listed here for your guidance:
. The highest priority should be given to satis-
fying the number and demographic distribution
of your quota. This quota should be completed
as soon as possible after the start of the project.
. Patients having known or suspected pesticide
poisoning should not be sampled. If you are
involved with a potential pesticide poisoning,
we would like to assist you in any way possible.
However, samples should not be taken for the
National Human Monitoring Program.
-------
F-6
. Patients with cachexla or who have been Institu-
tionalized for long periods should not be sampled
for the national program.
Collection of Surgical Adipose Tissue
Collect samples of adipose tissue from living patients from
unfixed specimens which have been surgically excised for therapeutic
reasons. Take special care to keep samples from different patients
separate and correctly and securely labeled and to avoid their con-
tact with other chemicals such as paraffin, disinfectants or plastics.
At least five grams of good quality (subcutaneous, perl renal,
mesenteric) adipose tissue should be collected from any part of the
specimen; avoid fibrous or connective tissue, I.e., omenturn. Place
the fat, without any fixatives or preservatives, into the chemically-
cleaned container that you have been given and legibly complete, in
ball-point pen or pencil, and attach the self-adhesive label. The
bottle labels should be affixed before freezing. Store the specimens
up-right in a freezer at -4°F (-20°C) until shipment.
Collection of Postmortem Adipose Tissue
Adipose tissue samples must be obtained only from unembalmed
cadavers. The interval between death and the collection of tissue
should be as short as possible and must not exceed 24 hours,
assuming refrigeration during the interval. Samples of adipose
tissue must weigh at least five grams and should be placed in the
-------
F-7
supplied chemically-clean container with a completed label affixed.
Specimens should be stored at -4°F (-20°C) without any fixative or
preservative until shipment. Submit only good quality fat; do not
submit omentum as it contains too much connective tissue for satis-
factory analysis.
Adipose should be taken dry, and should not be rinsed before
placing in the containers provided. Many water supplies contain
materials which would interfere with chemical analysis.
Instruments should be well-rinsed with distilled water and
dried before taking the adipose sample.
Completion of the Patient Summary Report
A Patient Summary Report should be completed for each patient
from whom a sample was taken. Special attention should be given to
the completeness of the information. First and last Initials, in
that order, should be used Instead of the complete name to Insure
that confidentiality is maintained. The Initials, along with the
date of birth, sex, and race, are used 1n this office to compose the
AMA identification number. The patient's Identification number and/or
the pathology department's accession number are for your Information
in referring back to the individual patient when you receive the
results of the pesticide analysis.
Confirmed diagnosis should be detailed 1n the spaces provided.
Only the major ones should be supplied.
-------
F-8
Other information required should be completed as accurately
as possible. The completed forms should be held and sent under the
lid of the insulated container when shipment 1s made.
Packing and Shipping
Tighten all lids on the specimen bottles carefully. This is
important since we are required to use special aluminum foil cap
liners which make tightening a little difficult. Be certain that
a completed bottle label is firmly attached to each specimen bottle.
Wrap each bottle in gauze or paper to prevent breakage during ship-
ment and to keep the label on the container. Place the specimen
bottles in the Insulated mailer and fill it with dry ice. If you
have difficulty obtaining dry ice, please call us and we can arrange
alternative methods of refrigeration for you.
A franked addressed label is on the reverse side of the address
card. This card is marked AIR MAIL - SPECIAL DELIVERY. (Do not
send Air Express, please). There is no cost to the sender because
of the franked label. All insulated mailers should have a PERISHABLE
PACKED IN DRY ICE label visible from all sides on the outside.
Specimens should be mailed on a Monday or Tuesday of a week with
no federal holidays. This assures that they will arrive before the
end of the work week on Friday.
-------
F-9
Patient Summary Reports should be sent In the carton with
the specimens when possible. They can be folded and placed on
the top of the styrofoam lid.
Only samples which meet our cnfna and are handled according
to the guidelines can be accepted. No substitute containers will
be accepted.
For Further Information
If you have any questions or comments, please contact us.
Telephone (collect): 202-755-8060.
Frederick W. Kutz, Ph.D.
Sandra C. Strassman
National Human Monitoring Program
-------
G-l
APPENDIX G
Quotas for Mirex Special Study Specimens
-------
G-2
NATIONAL HUMAN MONITORING PROGRAM
U. S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D. C. 20460
QUOTA FOR MIREX SPECIAL PROJECT
Each collection site in the states of Alabama and Mississippi is
requested to collect human adipose tissue from the following
age/sex/race groups:
QUOTA AGE GROUP SEX
4 0-14 Male
4 0-14 Female
5 15-44 Male
15-44 Female
4 45 and above Male
4 45 and above Female
Quota of 27 samples should include 5 non-Caucasian specimens.
Proportional distribution of these non-Caucasian specimens
among the age/sex groups is preferable but not required.
-------
G-3
NATIONAL HUMAN MONITORING PROGRAM
U. S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D. C. 20460
QUOTA FOR MIREX SPECIAL PROJECT
Each collection site in the states of Louisiana and Texas
is requested to collect human adipose tissue from the following
age/sex/race groups:
QUOTA AGE GROUP SEX
4 0-14 Male
4 0-14 Female
6 15-44 Male
6 15-44 Female
3 45 and above Male
4 45 and above Female
Quota of 27 samples should include 4 Non-Caucasian specimens.
Proportional distribution of these non-Caucasian soecimens amooo
the aqe/sex qrouos is preferable but not required.
-------
G-4
NATIONAL HUMAN MONITORING PROGRAM
U. S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D. C. 20460
QUOTA FOR MIREX SPECIAL PROJECT
Each collection site in the states of North Carolina, South Carolina,
Georgia and Florida is requested to collect human adipose tissue from
the following age/sex/race groups:
QUOTA AGE GROUP SEX
4 0-14 Male
4 0-14 Female
5 15-44 Male
6 15-44 Female
4 45 and above Male
4 45 and above Female
Quota of 27 samples should include 6 non-Caucasian specimens. Prooortional
distribution of these non-Caucasian specimens among the age/sex groups
is preferable but not required.
-------
H-l
APPENDIX H
Sample Mirex Chromatograms Made Available To RTI
-------
H-2
SAMPLES AND DATES OF ANALYSIS FOR CHROMATOGRAMS MADE AVAILABLE TO RTI
Samples Date of Analysis
MA 76-77 May 27, 1976
MA 76-78 May 27, 1976
MA 76-79 May 27, 1976
MA 76-80 May 27, 1976
MA 76-82 May 27, 1976
MA 76-83 May 27, 1976
MA 76-84 May 27, 1976
MA 76-85 May 27, 1976
MA 76-86 May 27, 1976
MA 76-98 June 4, 1976
MA 76-99 June 4, 1976
MA 76-100 June 4, 1976
MA 76-101 June 4, 1976
MA 76-71 Repeat June 4, 1976
MA 76-102 June 4, 1976
MA 76-103 June 4, 1976
MA 76-104 June 4, 1976
MA 76-123 June 15, 1976
MA 76-124 June 15, 1976
MA 76-125 June 15, 1976
MA 76-126 June 15, 1976
MA 76-127 June 15, 1976
MA 76-124 June 16, 1976
MA 76-125 June 16, 1976
MA 76-126 June 16, 1976
MA 76-127 June 16, 1976
MA 76-128 June 16, 1976
MA 76-129 June 16, 1976
MA 76-130 June 16, 1976
MA 76-131 June 16, 1976
MA 76-153 June 24, 1976
MA 76-155 June 24, 1976
MA 76-157 June 24, 1976
MA 76-158 June 24, 1976
MA 76-159 June 24, 1976
-------
H-3
SAMPLES AND DATES OF ANALYSIS FOR CHROMATOGRAMS MADE AVAILABLE
TO RTI
(Cont'd)
Samples Date of Analysis
MA 76-160 June 25, 1976
MA 76-161 June 25, 1976
MA 76-162 June 25, 1976
MA 76-163 June 25, 1976
MA 76-164 June 25, 1976
MA 76-165 June 25, 1976
MA 76-166 June 25, 1976
MA 76-167 June 25, 1976
MA 77-9 September 20, 1977
MA 77-10 September 20, 1977
MA 77-11 September 20, 1977
MA 77-12 September 20, 1977
MA 77-13 September 20, 1977
MA 77-14 September 20, 1977
MA 77-57 October 12, 1977
MA 77-58 October 12, 1977
MA 77-59 October 12, 1977
MA 77-60 October 12, 1977
MA 77-61 October 12, 1977
MA 77-62 October 12, 1977
MA 77-129 November 8, 1977
MA 77-130 November 8, 1977
MA 77-131 November 8, 1977
MA 77-132 November 8, 1977
MA 77-133 November 8, 1977
MA 77-134 November 8, 1977
-------
1-1
APPENDIX I
Modification of Mirex Analysis Protocol - GC Column Temperature
-------
ENVIRONMENTAL TOXICOLOGY DIVISION
HEALTH EFFECTS RESEARCH LABgAJjgJ
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
Research Triangle Park, North Carolina 27711
' 1976
FROM- Chief, Quality Assurance Section, ETD, EPA, HERL,
Research Triangle Park, N.C. /J (J
TO: Michigan Project, Mr. R. L. Welch
After receiving a copy of Bob Welch's letter to DrKutz (2/11/76)
ranges incorporating the derivatives of heptachlor/chlordane, mirex, and
Aroclor 1260.
Frankly, we were not overjoyed by the elution patterns produced
by Aroclor 1260 and Mirex. As Bob commented in his letter, at ZOO c
Mirex overlaps nearly completely with the next to !«' Ar°9lorP"*
(a major one). However, at 195- . although the separation is si ghtly
better we could not agree that it was anything remarkable on our
o"mn! Vr separation^ was not nearly as good as the separation
in Bob's chromatogram. We dropped the column temp. to. ^Sl'^lor it
the mirex peak was clearly visible on the early side of the Aroclor, it
was not a quantifiable separation.
Undoubtedly, the relative efficiency of his column vs ours was a
factor We estimated his column to be yielding about 3,200 T.P. witn
his carrie? flow obviously cranked down. To compensate for the exces-
s ve"ime consumption at 190', we raised the flow to 80 m with a
reciting EFF. of about 2,700 T.P. Undoubtedly, our resolution
characteristics suffered accordingly.
Just so long as Bob's column holds up and he can continue to
obtain the separations shown in his sample chromatograms we see no
objection at all to making temperature and/or carrier flow adjustments
in any way that it will best get the job done. There is nothing
sanctified in the recoimended parameters in the manual. They were
"nten for a shotgun approach to general multiresidue work, but if
«me other combination is more suitable for special purposes, then by
all means, appropriate modifications should be made.
Even with our poor separation, I believe the chromatographer would
be alerted io the pSssibl /presence of mirex then by further confirmation
via ECD and GC/MS the final answers can be obtained.
I'm sure we will want to get together for further discussions
of this topic the week of the chemists' meeting.
cc: Dr. F. W. Kutz, Washington
Dr. H. F. Enos, Athens
EPA Form 1320.4 (R»v. 6-72)
-------
J-l
APPENDIX J
Confirmation Analyses of Mirex in Human Adipose Tissue Extracts
-------
J-2
Analytical Results
Confirmations! Analyses for
Mirex in Human Adipose Tissue Extracts
Analytical Chemistry Branch
Environmental Toxicology Division
Health Effects Research Laboratory
Research Triangle Park, North Carolina 27711
Date: 30 November 1976
-------
DATE
SUBJECT.
J-3
Health Effects Research Laboratory
Research Triangle Park, North Carolina 27711
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
November 30, 1976
Confirmational Analyses for Mirex in Human Adipose Tissue Extracts
FROM. Dr. Edward 0. Oswald, Chief ___
Analytical Chemistry Branch, ETD (MD-69)
TO: Dr. Frederick W. Kutz
Ecological Monitoring Branch (WH-569)
OPP, Headquarters i
? ^ r
THRU: Director, HERL (MD-51) 1 'J
Attached you will find the results of confirmational analyses for
Mirex as generated by members of the Analytical Chemistry Branch, ETD,
HERL, RTF on extracts of human adipose tissue which were received from
the Michigan OPP lab on the 2 September 1976. Three (3) separate mass
spectrometry groups analyzed selected extracts by gas liquid chroma-
tography interfaced either with low resolution or with high resolution
mass spectrometry in the electron impact mode.
The results of the present report confirm both the qualitative and
the comparable quantitative presence of Mirex in human adipose tissue as
indicated by the electron capture results from the Michigan lab.
As indicated to you in a similar report of the 23 June 1976, the
GC/EC results agree quite favorably with the confirmational analyses by
GC/MS.
If additional GC/MS confirmctional analyses are needed by OPP for
Mirex in human adipose tissue, then I would suggest that only about 10%
of the total extracts be analyzed.
If I or other members of the Branch can be of assistance on this or
other matters, please feel free to contact me.
Enclosure
EPA FOIL. 1370 6 (Re. 3 761
-------
J-4
Mirex in Human Adipose Tissue Extracts
(^Concentrations expressed in parts per million—ppm)
Sample
Identification
*GC/EC Analyses
(Michigan)
Confirmational GC/MS Analyses
Research Triangle Park, N. C.
MA76-51
MA76-62
MA76-72
MA76-75
MA76-80
MA76-81
MA76-84
MA76-93
MA76-101
MA76-102
MA76-116
MA76-123
MA76-124
MA76-129
MA76-131
MA76-146
MA76-147
MA76-159
MA76-161 ,'ii
Control 41 #1
Control 41 #11 •'*
0.10
0.20
0.39
0.16
0.20
0.07
Negative
0.09
0.09
Negative
0.12
0.22
1.10
0.11
0.15
0.12
1.13
0.16
0.69 , /
i.oo ->'/./
.-•' ' i/i**''
i.oo -
**GC/LRMS
Group A
0.14
—
0.42
0.17
0.25
0.08
Negative
—
—
0.02
0.01
—
0.56
—
0.60
—
1.97
0.28
0.27
0.83
0.21
**GC/LRMS
Group B
0.08
0.26
—
—
0.15
0.06
Negative
0.07
0.09
0.08
—
0.21
--
0.36
—
0.14
—
—
0.74
0.81
***GC/HRMS
0.08
0.18
0.36
0.15
0.15
0.03
0.01
0.04
0.08
0.02
0.01
0.18
0.43
0.16
0.45
0.14
1.12
0.16
0.13
0.61
0.62
Note: A number of the analyses were hindered by the high fat content
of the sample extract.
*GC/EC = Analyses by gas liquid chromatography using the electron
capture detector.
**GC/LRMS = Analyses by gas liquid chromatography interfaced with low
resolution (500-1000) electron impact mass spectrometry.
***GC/HRMS = Analyses by gas liquid chromatography interfaced with high
resolution (7200-8500) electron impact mass spectrometry.
-------
J-5
Analytical Methodology
Confirmational Analyses for
Mir« in Human Adipose T.ssue Extracts
Analytical Chemistry Branch
Environmental Toxicology Division
Health .Effects Research Laboratory
Research Triangle Park, North
Date: 30 November 1976
-------
J-6
Methods^
Extracts of human adipose tissue were received in sealed glass
ampules. At the time that the sample vial was opened, aliquots of each
selected sample extract were prepared for analysis by the various mass
spectrometry groups. All samples were analyzed by at least two (2)
of the analytical teams which generated multiple analyses. All
quantitative results are expressed as mean concentrations in parts per
million (ppm) for each analytical team. The three (3) mass spectrometry
teams were coordinated by: (a) Mr. Robert Harless; (b) Dr. Wayne Sovocool;
and (c) Dr. Lynn Wright.
Below are listed the detailed analytical conditions utilized for
the GC/MS confirmational analyses of Mirex in extracts of human adipose
tissue.
-------
J-7
Group A.
Low Resolution GC/MS
Instrument: Hewlett Packard 5930A Mass Spectrometer equipped with a
Hewlett Packard 5700 gas chromatograph
GC Column: 183 cm x 2 nun (i.d.) glass column packed with 3% OV-17 on
100/120 mesh Gas Chrom Q.
Column temperature: 235°C
Injector temperature: 235°C
Transfer line temperature: 250°C
Membrane separator temperature: 235°C
Carrier gas: Helium
Flow rate: 42 ml/min
lonization voltage: 70 eV
Criteria for Analyses:
These low resolution mass spectrometric analyses (GC-LRMS) are
based upon: (a) correct absolute retention time as compared to Mirex;
(b) presence of intense fragments at m/e 270 with a 6-chlorine isotope
cluster and at m/e 235 with a 5-chlorine cluster (corresponding to the
main electron impact ion fragments of Mirex); and (c) the proper ratios
of the m/e 270 and m/e 235 ion clusters.
Quantitative GC-LRMS measurements were accomplished by single ion
monitoring of the m/e 272 ion in comparison to an external Mirex standard
and to internal spiked samples. All samples were analyzed in at least
duplicates.
-------
J-8
Group B.
Low Resolution GC/MS
Instrument: Finnigan 3200 Mass Spectrometer equipped with a Finnigan
6100 data system and Finnigan 9500 gas chromatograph
GC Column: 122 cm x 2 mm (i.d.) glass column packed with 3% OV-17
on 100/120 mesh Gas Chrom Q
Column temperature: 240°C
Injector temperature: 240°C
Transfer line temperature: 240°C
Glass jet separator temperature: 240°C
Carrier gas: Helium
Flow rate: 20 ml/min
lonization voltage: 70eV
Criteria for Analyses:
These low resolution mass spectrometric analyses (GC-LHMS) are
based upon: (a) correct absolute retention time as compared to Mirex;
and (b) presence of intense fragment*, in the m/e 270 region representing
the most prominent ions of the c5c]-6 cluster.
The mass spectrometer was operated in the multiple ion dejection
mode monitoring the three (3) most prominent ions of the C5C1& cluster—
m/e 270, m/e 272, and m/e 274. A fourth ion m/e 405~not present in the
spectrum of Mirex—-was monitored to provide a background. Any sample
which gave a negative or trace response upon the initial dilution
conditions was then further reduced in volume to approximately 1/10 of
its original volume. The sample was then reanalyzed. Negative responses
to the latter conditions were reported as negative.
Samples were quantitated by summating the areas the mass chromato-
grams for m/e 270, 272, and 274 and comparisons of these values with the
average response obtained for an external Mirex standard which was
injected either before or after each residue sample.
High Resolution GC/MS
Instrument: Varian MAT 311A Mass Spectrometer- equipped with a Varian
2700 gas chromatograph and a Varian C-1024 time averaging
computer (CAT).
GC Column: 91 cm x 2 mm (i.d.) glass column packed with 2% Dexil 410
on 90/100 mesh Anakrom Q.
-------
J-9
Column temperature: 243°C
Injector port temperature: 255°C
Transfer line temperature: 255°C
Slit separator temperature: 255°C
Carrier gas: Helium
Flow rate: 10-12 ml/min
lonization voltage: 70eV
Criteria for Analyses:
The high resolution mass spectrometnc analyses (GC-HRMS) are based
upon: (a) correct absolute retention time as compared to Mirex; (b)
isotope ratios of two fragments of the molecular ion cluster—m/e 539.6'62
{C10 C112) and m/e 543.6203 (C^Cl, J7C1 ); and (c) analysis with
internal and external standards of Mirex.
The mass spectrometer was operated at a high sensitivity in order
to observe the molecular ion m/e 543.6203 (Cin CLln Cl,) with a mass
resolution capability of 7200-8500. Using tne timeaveraging computer
(CAT) system, the mass spectrometer was operated in a double ion moni-
toring mode for the molecular ion m/e 539.6262 and m/e 543.6203. The
quantitative analyses were accomplished by the use of internal and
external standards of Mirex combined with the double ion monitoring of
two specific ion fragments in the molecular cluster of Mirex in the high
resolution GC/MS mode.
-------
J-lO
Health Effects Research Laboratory
Environmental Research Center
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
Research Triangle Park, North Carolina
SUBJECT: Confirmation of Mir ex in Human Adipose Tissue DATE: June 23, 1976
Extracts _ .....
FROM: Dr. Robert G. Lewis, Chief 11 VIP"'
Chemical Characterization Section, ACB/ETD (MD-69)
TO: Dr. Frederick W. Kutz
Ecological Monitoring Branch (WH-569)
OPP, Headquarters
This is to confirm my telephone transmittals to you of the results
of our confirmatory analysis of mirex in human adipose tissue extracts
received from the Michigan project on June 15, 1976. Mirex was confirmed
qualitatively and determined quantitatively in all five samples by each
of three independent methods; i.e., photochemical, GC-low resolution
mass spectrometry (GC-LRMS) and GC-high resolution mass spectrometry
(GC-HRMS). Data are presented below:
GC-ECD Mirex Concentration, ppm
Sample I.D. (Michigan) Photochem. GC-ECD GC-LRMS GC-HRMS
A75-95 0.5 0.4 0.5 0.7
75-14673 1.3 0.8 0.7 0.9
A104-75 1.3 1.3 1-2 1.6
575-6796 0.4 0.3 0.2 0.4
575-6999 1.2 0.4 0.8 1.2
Samples were received in glass vials sealed with Teflon-lined caps.
Volumes ranged from 3.6 to 14.6 ml. Aliquots of 1 to 4.5 ml were with-
drawn for photochemical confirmation. In this procedure, the aliquot
was added to 10% diethylamine in hexane sufficient to make a total
volume of 5 ml. These solutions were analyzed for GC-ECD before and
after 100 min. exposure in quartz tubes to a 275W sunlamp. Analytical
conditions were as follows:
Instrument ~ Tracer 222 - Ni BCD
Column — 183-cm x 4-mm (i.d.) glass, packed with 1.5% OV-17/1.95%
OV-210 on 80/100 mesh Gas Chrom Q
Attenuation — 10 x 8
Carrier Gas — Ar/CH4/ 95:5
Carrier Flow — 75 ml/min
Column temp — 200°C
Detector temp — 270°C
Inlet temp — 225CC
EPA Form 1320-6 (R... 6-72)
-------
J-ll
2
B.G. Current — 46% f.s.d. @ 10 x 128 D.C.
Voltage — 50v
Pulse Width — 5 psec
Pulse Rate — 150 ysec
Photochemical confirmation work was performed by R. C. Hanisch.
Portions of each sample were taken for GC-LRMS analysis. These
were concentrated to 100 pi by evaporation under a gentle stream of
nitrogen at room temperature. Portions corresponding to ca. 50 ng of
mirex were injected into the GC-LRMS for determination of mass spectra
over the 195 to 560 amu range. GC-LRMS confirmation was based on (1)
correct absolute retention time; (2) presence of intense fragments at
270 m/e with 6-chlorine isotope cluster and at 235 m/e with 5-chlorine
isotope cluster (corresponded to the main electron impact fragments of
mirex); (3) presence of other fragments derivable from mirex; (4) absence
of possible interferring compounds; and (5) proper ratios of the 270 and
235 m/e clusters. Quantitative GC-LRMS measurements were accomplished
through single ion monitoring of the 272 m/e peak and the use of an
external mirex standard. All samples were analyzed in duplicate or
better. Precision with standards was £10 to 15%. GC-LRMS parameters
were as follows:
Instrument: Hewlett Packard 5930A quadrapole focusing mass spec-
trometer equipped with an HP 5700A gas chromatograph and an HP
5932A data system.
Column: 183-cm x 2-mm (i.d.) glass, packed with 3% CV-17 on 100/120
mesh Gas Chrom Q.
Carrier Gas: Helium
Flow rate: 41 ml/min
Injection port temperature: 200°C
Column temperature: 230°C
Transfer line temperature: 250°C
Membrane separator temperature: 230°C
Ion source temperature: 190°C
Mass filter temperature: 120°C
lonization voltage: 70eV
Filament emission: 160 yamp
-6
Ion source pressure: 2 x 10 mm Hg
-------
J-12
Scan rate: 360 amu/sec
GC-LRMS measurements were performed by G. W. Sovocool and M.
Simpson.
The same concentrated samples used for GC-LRMS were also used for
GC-HRMS. The mass spectrometer was operated at extreme sensitivity in
order to observe the molecular ion at 543.6203 m/e (corresponding to
C1035Clln37Cl2). Qualitative and quantitative analyses were performed
by double ion monitoring of the 539.6262 and molecular ions. GC retention
times and internal standards were used for identification and quantifi-
cation. GC-HRMS parameters are given below:
Instrument: Varian MAT 311A double focusing mass spectrometer
equipped with a Varian gas chromatograph and a Varian data system.
Column: 91-cm x 2-mm (i.d.) glass, packed with 5% OV-101 on Gas
Chrom Q.
Carrier gas: Helium
Flow rate: 15 ml/min
Column temperature: 235°C
Injection port, transfer line and separator temperatures: 250°C
Ion source temperature: 220°C
Ion source pressure: 5 x 10 mm Hg
GC-HRMS measurements were performed by R. L. Harless.
I trust that you will find these data useful. If we can be of
further assistance, please call on us.
------- |