Content uploaded by James Quackenboss
Author content
All content in this area was uploaded by James Quackenboss
Content may be subject to copyright.
Environmental Health Perspectives
•
VOLUME 109 | NUMBER 6 | June 2001
583
Measurement of Children’s Exposure to Pesticides: Analysis of Urinary
Metabolite Levels in a Probability-Based Sample
John L. Adgate,
1
Dana B. Barr,
2
C. Andrew Clayton,
3
Lynn E. Eberly,
1
Natalie C.G. Freeman,
4
Paul J. Lioy,
4
Larry L.
Needham,
2
Edo D. Pellizzari,
3
James J. Quackenboss,
5
Amit Roy,
4
and Ken Sexton
1
1
School of Public Health, University of Minnesota, Minneapolis, Minnesota, USA;
2
National Center for Environmental Health, Centers for
Disease Control and Prevention, Atlanta, Georgia, USA;
3
Research Triangle Institute, Research Triangle Park, North Carolina, USA;
4
Environmental and Occupational Health Sciences Institute, Piscataway, New Jersey, USA
5
U.S. Environmental Protection Agency, Las
Vegas, Nevada, USA
The potential health effects associated with
children’s exposure to pesticides are the sub-
ject of increasing concern. Children may be
exposed to unsafe levels of pesticides in the
air they breathe, the food they eat, the water
they drink, and the surfaces they touch
(1–4). Concerns about dietary exposures in
the general population resulted in the pas-
sage of the Food Quality Protection Act of
1996 [FQPA (5)]. Under FQPA, realistic
evaluation of the potential health risks to
children from pesticides requires informa-
tion about the full range of children’s expo-
sures. Unfortunately there is a scarcity of
data available to assess children’s actual
exposure to pesticides, which makes it diffi-
cult, if not impossible, to assess children’s
health risks realistically (6).
Urinary biomarkers of pesticides and their
metabolites have been used to characterize
body burden levels for adult populations as
well as trends over time in the United
States (7,8) and Europe (9). Under FQPA,
organophosphorus (OP) pesticides and car-
bamates are subject to increased scrutiny
because they are used widely in agricultural
and residential settings and they inhibit
cholinesterase, an enzyme essential for
proper functioning of the nervous system
(10). Most metabolites of OPs and carba-
mates have relatively short biological half-lives,
generally on the order of days, and are
excreted primarily in the urine (11,12). Past
studies of children’s OP pesticide exposure
either have been conducted outside the United
States (13) or have focused on special popula-
tions, such as children of agricultural worker
families (14), which are presumed to be more
highly exposed than average. There is a lack of
probability-based studies that provide baseline
data on the distribution of children’s expo-
sures in urban and nonurban settings.
As part of the Minnesota Children’s
Pesticide Exposure Study (MNCPES),
which was a probability-based sample of
children 3–13 years of age (15), urine sam-
ples were collected and analyzed for metabo-
lites of commonly used pesticides. The
MNCPES was a Phase III special study that
was part of the National Human Exposure
Assessment Survey (NHEXAS) (16). The
multiphase study was designed to identify
and select children with a known probability
by applying a predetermined set of eligibility
criteria and adjusting for subject nonre-
sponse within each phase. The sampling
process preferentially (but not exclusively)
selected a higher proportion of households
reporting more frequent pesticide use as well
as children more likely to experience expo-
sures to four target pesticides: the herbicide
atrazine and the organophosphates chlor-
pyrifos, diazinon, and malathion. It was pos-
tulated that children living in these homes
and participating in the MNCPES were
more likely to have measurable concentra-
tions of pesticide metabolites in their urine
(15). Individual-level statistical weights were
developed to adjust for oversampling these
nominally higher-pesticide-use households.
These weights were incorporated into the
analysis presented here so inferences could be
drawn about the prevalence and magnitude
of pesticide exposure for children within the
census tracts sampled.
Address correspondence to J.L. Adgate, University
of Minnesota School of Public Health, MMC 807,
Room 1260, 420 Delaware St. SE, Minneapolis,
MN 55455 USA. Telephone: (612) 624-2601.
Fax: (612) 626-0650. E-mail: jadgate@umn.edu
We thank the families who participated in the sur-
vey for their cooperation and M. Bollenbeck for his
assistance with the data analysis. MNCPES was
funded in part through the NHEXAS Pesticides and
PAH Module Cooperative Agreement R821902
between the U.S. Environmental Protection Agency
and the consortium of Research Triangle Institute/
Environmental and Occupational Health Sciences
Institute (RTI/EOHSI), by U.S. EPA STAR (Science
to Achieve Results) Grant R825283 to the University
of Minnesota, and by a grant from the Legislative
Commission on Minnesota Resources.
Received 20 October 2000; accepted 12 March
2001.
Children’s Health Articles
The Minnesota Children’s Pesticide Exposure Study is a probability-based sample of 102 children
3–13 years old who were monitored for commonly used pesticides. During the summer of 1997,
first-morning-void urine samples (1–3 per child) were obtained for 88% of study children and
analyzed for metabolites of insecticides and herbicides: carbamates and related compounds (1-
NAP), atrazine (AM), malathion (MDA), and chlorpyrifos and related compounds (TCPy).
TCPy was present in 93% of the samples, whereas 1-NAP, MDA, and AM were detected in 45%,
37%, and 2% of samples, respectively. Measured intrachild means ranged from 1.4 µg/L for
MDA to 9.2 µg/L for TCPy, and there was considerable intrachild variability. For children pro-
viding three urine samples, geometric mean TCPy levels were greater than the detection limit in
98% of the samples, and nearly half the children had geometric mean 1-NAP and MDA levels
greater than the detection limit. Interchild variability was significantly greater than intrachild
variability for 1-NAP (p = 0.0037) and TCPy (p < 0.0001). The four metabolites measured were
not correlated within urine samples, and children’s metabolite levels did not vary systematically
by sex, age, race, household income, or putative household pesticide use. On a log scale, mean
TCPy levels were significantly higher in urban than in nonurban children (7.2 vs. 4.7 µg/L; p =
0.036). Weighted population mean concentrations were 3.9 [standard error (SE) = 0.7; 95% con-
fidence interval (CI), 2.5, 5.3] µg/L for 1-NAP, 1.7 (SE = 0.3; 95% CI, 1.1, 2.3) µg/L for MDA,
and 9.6 (SE = 0.9; 95% CI, 7.8, 11) µg/L for TCPy. The weighted population results estimate
the overall mean and variability of metabolite levels for more than 84,000 children in the census
tracts sampled. Levels of 1-NAP were lower than reported adult reference range concentrations,
whereas TCPy concentrations were substantially higher. Concentrations of MDA were detected
more frequently and found at higher levels in children than in a recent nonprobability-based sam-
ple of adults. Overall, Minnesota children’s TCPy and MDA levels were higher than in recent
population-based studies of adults in the United States, but the relative magnitude of intraindi-
vidual variability was similar for adults and children. Key words: children’s health, exposure
assessment, Minnesota Children’s Pesticide Exposure Study (MNCPES), National Health and
Nutrition Examination Survey (NHANES), National Human Exposure Assessment Survey
(NHEXAS), organophosphate pesticides, urinary biomarkers. Environ Health Perspect
109:583–590 (2001). [Online 22 May 2001]
http://ehpnet1.niehs.nih.gov/docs/2001/109p583-590adgate/abstract.html
This study measured four specific urinary
metabolites: 1-naphthol (1-NAP), atrazine
metcapturate (AM), malathion dicarboxylic
acid (MDA), and 3,5,6-trichloro-2-pyridinol
(TCPy). These compounds are the primary
urinary metabolites of the parent compounds
naphthalene or carbaryl (1-NAP), atrazine
(AM), malathion (MDA), and chlorpyrifos,
chlorpyrifos-methyl, or triclopyr (TCPy)
(12). Naphthalene is used in mothballs and
found in cigarette smoke and petroleum
products, and carbaryl is a carbamate insecti-
cide used on turf and gardens (7). Atrazine is
a herbicide used on a wide variety of crops in
the United States, and malathion is an insec-
ticide used on fruits and vegetables as well as
in household products (12,17). Chlorpyrifos
and chlorpyrifos-methyl have had wide use
on fruits and vegetables and as treatments
inside and outside homes, although many
residential uses were recently restricted (18).
TCPy is the primary environmental degrada-
tion product of chlorpyrifos, chlorpyrifos-
methyl, and/or triclopyr and may persist in
soil for up to a year (19). Because of the
numerous uses of chlorpyrifos and chlorpyri-
fos-methyl as well as the ubiquity and stabil-
ity of TCPy in the environment, urinary
TCPy represents total intake of both “envi-
ronmental” TCPy and parent compounds.
Until more is known about the adsorption,
distribution, and metabolism of TCPy, the
relative contribution of environmental
TCPy to total urinary TCPy remains
unknown.
In this article we report measured uri-
nary metabolite levels and the variability of
1-NAP, AM, MDA, and TCPy within and
between individual children, and within and
between sociodemographic subgroups,
including urban and nonurban children.
Weighted distributions of urinary metabolite
levels for the general population of children
in the census tracts sampled are also pre-
sented. Additionally, we compare these pop-
ulation distributions to adult levels and
examine the implications of our findings for
exposure analysis and risk assessment.
Methods
Study design. MNCPES used a cross-sec-
tional design, and exposure-related measure-
ments were obtained for each subject during
a one-week sampling period between May
and September 1997. The sampled popula-
tion consisted of children 3–13 years of age
living in households located in either the
cities of Minneapolis and St. Paul (desig-
nated urban households), or Rice and
Goodhue Counties (designated nonurban
households), located just south of the Twin
Cities metropolitan area. A detailed descrip-
tion of the study design strategy, eligibility
criteria, household and subject sample selec-
tion, field sampling methods, monitoring
outcomes, and the development and applica-
tion of statistical weights has been published
(15,20). To maintain a probability-based
structure within a complex study design,
probabilities of subject selection by investi-
gators (based on application of eligibility cri-
teria described below) and subject response
(participation rate) were assessed at each
step. This information was used to develop
statistical weights to adjust the population
distributions for the oversampling of some
subpopulations. The probability-based sam-
ple was obtained through a systematic
process that occurred over three phases:
identification, household screening, and
intensive exposure monitoring. Table 1 sum-
marizes how participation rates were calcu-
lated over the two steps within each of the
three phases: the number of children listed
in the “number eligible” column reflects
both the process of subject selection within
each phase and the number of participants
electing to partake in the previous step(s).
Identification. We selected 2,303 tele-
phone numbers from a commercially avail-
able list of residences (Genesys Systems, Inc.,
Fort Washington, PA) predicted to have age-
eligible children based on birth records and
other publicly available data. Because of con-
cerns that the list might underrepresent fami-
lies from lower socioeconomic strata (SES),
we sampled telephone numbers from lower-
SES census tracts proportional to their rate of
occurrence in the 1990 Census. Of the initial
2,303 telephone numbers, 2,211 were deter-
mined to be residential, 2,057 were judged
eligible for screening, and telephone screening
was completed for 1,388 of these households.
Household Screening. Using a combina-
tion of selection criteria (i.e., residence
located in target areas, age-eligible child pre-
sent, reported use of pesticides, and use of a
well as a water source in nonurban house-
holds) and probability sampling, we deemed
477 families eligible to participate; 294 com-
pleted the screening-phase survey of in-home
pesticide storage and use (21). In the
MNCPES survey design, a larger proportion
of households with “more frequent pesticide
use” and with more than one eligible child
were selected for the household-screening
phase, and families with private wells in
nonurban areas were preferentially selected.
On the basis of screening results, those classi-
fied as having a greater potential for exposure
to target pesticides were selected at a higher
rate for the intensive-monitoring phase of
MNCPES. Identification of children consid-
ered likely to have higher exposures was
based on scoring, which integrated informa-
tion from a household roster, screening ques-
tionnaire, and the pesticide inventory (20).
The scoring factors used to assess exposure,
from highest to lowest weight, were use of a
primary pesticide (the OPs chlorpyrifos,
malathion, and diazinon) in the household in
the previous year; any pesticide use inside or
outside the home in the previous 6 months;
adult occupational pesticide contact; primary
pesticides present but not reported used in
the previous year; and use of only nonpri-
mary pesticides in the previous year.
Intensive monitoring. Of the 181 eligi-
ble families, 174 completed a baseline ques-
tionnaire (7 refused), and after selection
appointments were made with 109 to begin
intensive monitoring. Of these, 102 families
Children’s Health • Adgate et al.
584
VOLUME 109 | NUMBER 6 | June 2001
•
Environmental Health Perspectives
Table 1. Summary of participation rates for MNCPES.
Number Number Participation Cumulative
Type of participation eligible
a
participating rate (%) rate (%)
Determined if household or business phone 2,303 2,211 96.0 96.0
Completed telephone screening 2,057 1,388 67.5 64.8
Agreed to in-home screening 477 348 73.0 47.3
Completed in-home screening 335 294
b
87.8 41.5
Completed baseline questionnaire 181 173 95.6 39.7
Kept monitoring appointment 109 102
c
93.6 37.1
a
Number of subjects eligible for each phase based upon application of selection criteria described in the text as well as
participant response in each previous phase. See Adgate et al. (
20
) for a full description of the selection process.
b
14
additional cases were completed after the deadline for selection of monitoring subjects.
c
Seventy-two urban and 30
nonurban households.
Table 2. Percent relative SD and bias of field quality control samples of 1-NAP, AM, MDA, and TCPy.
Concentration in
Characterized MNCPES
Percent concentration control samples Mean
Analyte
n
a
relative SD (µg/L)
b
(µg/L) Bias percent bias
1-NAP 7 29 4.2 ± 1.3 3.4 –0.8 –19
AM 8 7.0 9.0 ± 1.1 5.8 –3.2 –36
MDA 8 8.0 10.3 ± 1.5 15 4.7 46
TCPy 6 23 10.3 ± 2.1 11 0.7 6.8
a
Number of field quality control samples with known reliable results.
b
No standard reference materials exist for these
metabolites in urine. Concentration characterized over time from a pool of urine with all four metabolites at approxi-
mately 8 µg/L.
with children were enrolled and subse-
quently completed the intensive-monitoring
phase of the study [the 7 remaining families
were not monitored because of constraints
on the number of monitoring appointments
available per week (15)]. During this phase a
combination of personal exposure measure-
ments (i.e., air, duplicate diet, hand rinse),
environmental measurements (i.e., residen-
tial indoor/outdoor air, drinking water, dust
on residential surfaces, soil), and data on
children’s activity patterns were collected, in
addition to blood and urine measurements
(22). During the week-long monitoring
period, individual children (sometimes with
help from their parents) completed time-
activity diaries, and a follow-up question-
naire was administered at the completion of
monitoring.
Response rates and sampling weights. As
shown in Table 1, subject response rates
ranged from 68% to 96% within each of the
three phases, with a compound response rate
of 37% for the children who completed all
three phases. On the basis of the sampling
scheme, we developed weights for house-
holds and children because not all members
of the population had an equal probability of
selection. Sampling weights were computed
based on probabilities of selection, were
adjusted for nonresponse, and incorporated
potential pesticide usage scoring factors and
household pesticide inventory information
(20,21). Therefore, weighted summary sta-
tistics can be used to provide unbiased esti-
mates of the general population in the
census tracts sampled.
Urine sampling and analysis. First-
morning-void samples were collected on
days 3, 5, and 7 of the week-long monitor-
ing period to bracket duplicate diet sample
collection. Samples were split into two sub-
fractions, frozen, and shipped by overnight
express mail to the National Center for
Environmental Health at the Centers for
Disease Control and Prevention (CDC) in
Atlanta, Georgia, where they were stored at
–70°C pending chemical analysis.
We determined concentrations of 1-
NAP and TCPy by capillary gas chromatog-
raphy and tandem mass spectrometry using
an isotope dilution technique with
13
C
and/or
15
N-labeled internal standards (23).
Similarly, we determined AM and MDA by
liquid chromatography and tandem mass
spectrometry using an isotope dilution tech-
nique with
13
C or deuterium-labeled inter-
nal standards (24). For most of the analytical
runs, the analytical detection limits (DLs)
were 1.3 µg/L for 1-NAP, 1.4 µg/L for
TCPy, and 1.0 µg/L for AM and MDA.
Lower DLs were obtained in some of the
analytical runs because of improved instru-
mentation and increased operator experi-
ence: 1.0 µg/L for 1-NAP (n = 62), 0.53
µg/L for AM (n = 135), and 0.49 µg/L for
MDA (n = 67).
We determined urinary creatinine con-
centrations using the established colorimetric
enzymatic method available on Vitros CREA
slides (Ortho Clinical Diagnostics, Raritan,
NJ). We made reflectance measurements
using a Kodak Ektochrome 250 analyzer
(Eastman Kodak Co., Rochester, NY) at
3.85 and 5 min. The difference in the two
reflectance measurements was directly pro-
portional to the creatinine concentration of
the urine.
We used field blanks and spiked samples
to assess the quality of sample collection and
analysis procedures. Field blanks consisted of
deionized 18 mega-ohm water poured from
a bulk container to a urine sample cup in the
residences of randomly selected children.
Eleven of the 13 field blanks contained con-
centrations of pesticide metabolites below
the DLs. Two field blank samples had
detectable levels of 1-NAP and TCPy, but
levels were just above the DL.
Using methods developed by CDC
investigators (25), samples from a character-
ized pool of urine containing 1-NAP, AM,
MDA, and TCPy at approximately 8 µg/L
were transported frozen to the field for use as
field quality control samples. These samples
were thawed and transferred to a sample cup
in the residence of randomly selected chil-
dren and then returned to the laboratory
with participant samples. The CDC techni-
cians conducting the analyses were blind to
the sample types. Precision, characterized by
calculating percent relative standard devia-
tion of repeat measurements, and bias (devia-
tion from the true value) of pre-characterized
metabolite concentrations in the field quality
control samples are summarized in Table 2.
Overall precision and bias were within the
range of MNCPES acceptance criteria, and
results were adjusted for recovery by using
isotope dilution mass spectrometry.
Statistical analysis. For summary statisti-
cal calculations, highly dilute urine samples
(< 0.3 g/L creatinine; n = 4) were excluded,
and samples with values that were less than
the DL were assigned half the limit of detec-
tion. All urine metabolite concentrations were
reported as mass per unit volume (µg/L). SAS
was used for tabulations and for weighted
analysis of variance between groups (PROC
GLM) (26). SUDAAN (27) was used for all
weighted summary statistical and sociodemo-
graphic comparisons, so that weighted means,
for example, were obtained by multiplying a
child’s mean metabolite concentration by the
weighting factor for that child and then divid-
ing by the sum of the weights for all children.
Reported p-values were not adjusted for mul-
tiple comparisons.
Results
At least one urine sample was collected for
90 of the 102 children who participated in
the week-long intensive monitoring phase of
MNCPES: 87 children provided three sam-
ples each, two children provided two, and
one child provided one sample. As shown in
Table 3, males and females were randomly
distributed among the children providing
urine samples. The weighted mean age of
the children providing samples was 7.4 years
[range 3–13, standard error (SE) = 0.40],
and mean age did not vary significantly
between urban (n = 62) and nonurban (n =
28) subjects nor between children who pro-
vided urine samples and those who did not.
We obtained 266 urine samples, and 24
(9%) had missing data for creatinine or at
least one metabolite, which was caused by
analytical problems, such as matrix interfer-
ence, or insufficient sample volume.
Table 4 summarizes the unweighted dis-
tribution of analytical results for 1-NAP,
AM, MDA, TCPy, and creatinine in these
urine samples. TCPy was present at detect-
able levels in nearly all samples, 1-NAP was
present at detectable levels in more than half
of the samples, MDA was present at detect-
able levels in more than a third of the sam-
ples, and AM was detected infrequently. The
distribution of creatinine results is also pre-
sented because few baseline distributional
data exist in the scientific literature on chil-
dren for this commonly used adjustment for
urinary dilution.
Intra- and interchild variability. The
distribution of metabolite levels was right
skewed for all metabolites, as shown by the
intrachild arithmetic means, standard devia-
tions, and CVs presented in Table 5 (AM
excluded because of infrequent detection).
For the 87 children with three samples,
intrachild mean levels ranged from 1.4 µg/L
Children’s Health • Children’s pesticide exposure
Environmental Health Perspectives
•
VOLUME 109 | NUMBER 6 | June 2001
585
Table 3. Age distribution of children providing urine samples by sex and household location.
Number of children in each age group
Age (years) 3 4 5 6 7 8 9 10 11 12 13 Total
Sex
Male 3 4 5 3 4 2 3 6 7 7 1 45
Female 5 3 5 11 5 4 6 1 3 2 0 45
Location
Urban 4 4 8 10 6 5 6 6 7 6 0 62
Nonurban 4 3 2 4 3 1 3 1 3 3 1 28
Children’s Health • Adgate et al.
586
VOLUME 109 | NUMBER 6 | June 2001
•
Environmental Health Perspectives
for MDA to 9.2 µg/L for TCPy. Mean intra-
child CVs were 74% (range 0–160) for 1-
NAP, 59% (range 0–158) for MDA, and
55% (range 5.9–134) for TCPy. The aver-
age intrachild range for the 89 children with
2 or 3 samples was 5.0 µg/L (range 0–53) for
1-NAP, 2.6 µg/L (range 0-22) for MDA,
and 9.0 µg/L (range 0.9-36) for TCPy.
Intrachild ranges were at least as great as the
overall pooled population means for 1-NAP,
MDA, and TCPy. Figure 1 displays TCPy
levels in all children providing samples (n =
90) sorted from lowest to highest mean lev-
els, and demonstrates the wide variability in
metabolite levels in this population of chil-
dren. Because of the skewed distribution of
all detectable metabolites, subsequent statis-
tical comparisons were conducted using log-
transformed values: a Box-Cox procedure
confirmed that the log transformation was
appropriate for both the intrachild and popu-
lation distributions of these metabolites (28).
A tabulation of the number of samples
with detectable concentrations is presented in
Table 6 for the 87 children who provided 3
urine samples: The number of children with
valid analytical results in all 3 samples ranges
from 80 to 83. Concentrations of AM were
lower than the DL in almost all children, but
all children had at least one TCPy measure-
ment greater than the DL. The percentage of
children with detectable metabolite concen-
trations in all three samples ranged from 0%
for AM to 83% for TCPy. Ninety-eight per-
cent of the intrachild geometric mean (GM)
TCPy levels were greater than the DL, and
slightly less than half of the children had GM
1-NAP and MDA concentrations greater
than the DL. Only three children had GM
AM levels above the DL.
To examine the correlation between
individual metabolites within urine samples,
we calculated bivariate correlation coeffi-
cients for all possible pairs of metabolites
with concentrations greater than the detec-
tion limit (AM was excluded because of the
high proportion of nondetectable samples).
Pearson’s r values for log-transformed
metabolite concentrations were low and not
statistically significantly different from 0:
–0.19 (p = 0.24) for 1-NAP versus TCPy;
0.14 (p = 0.39) for 1-NAP versus MDA; and
0.22 (p = 0.16) for MDA versus TCPy.
A weighted analysis of variance of log-
transformed metabolite levels indicated that
interchild variability was significantly greater
than intrachild variability for 1-NAP and
TCPy (1-NAP, F = 1.66, p = 0.0037; TCPy,
F = 2.79, p < 0.0001). For MDA, interchild
variability was greater than intrachild variabil-
ity, but only marginally so, perhaps as a result
of the relatively larger number of nondetects
(F = 1.23, p = 0.13). Intrachild TCPy vari-
ability increased with increasing concentra-
tion, which was determined by examining the
relations between the log intrachild variance
and the log intrachild mean using weighted
regression (Slope = 2.2, SE = 0.21, t = 10.3, p
< 0.01). A similar pattern was observed for 1-
NAP (Slope = 3.0, SE = 0.4, t = 7.5, p< 0.01)
and MDA levels (Slope = 3.0, SE = 0.20, t =
15.3, p < 0.01) for metabolite concentrations
greater than the DL.
Sociodemographic covariates and ques-
tionnaire responses. Mean log metabolite lev-
els did not vary significantly when stratified
by sex or by age (< 6 years versus > 6 years).
Weighted mean TCPy concentrations, how-
ever, were significantly higher on a log scale
in urban than in nonurban children (t = 2.1;
p = 0.036), and mean log 1-NAP and MDA
levels were both marginally higher in urban
areas (Table 7). Statistical comparisons for
AM were precluded because only 5 children
had detectable levels in their urine, although
four of the five were from urban areas.
The mean log metabolite levels did vary
between racial and income subgroups, but no
clear race- or income-related trends are evi-
dent for these metabolites. For example, log
1-NAP levels for children from households
with incomes between $30,000–$50,000
were significantly higher than levels in chil-
dren from households with incomes
> $75,000 (t = 2.28, p = 0.025), and white
children had significantly higher levels than
nonwhites (t = 2.7, p = 0.009). Log MDA
levels, however, were significantly higher in
nonwhite children when compared to white
children (t = 2.15, p = 0.035), and children
from households with incomes in the
$30,000–$50,000 range had lower log MDA
Figure 1. Mean and range of intrachild TCPy levels for urban and nonurban subjects sorted from lowest to
highest concentration.
50
40
30
20
10
0
TCPy (µg/L)
Subject
Urban
n
= 62
Nonurban
n
= 28
Table 5. Summary of intrachild metabolite levels (µg/L): mean, SD, and percent coefficient of variation
(CV) for 1-NAP, MDA, and TCPy in children with valid analytical results in three urine samples (
n
= 80 for
1-NAP;
n
= 83 for MDA and TCPy).
Intrachild mean
a
(µg/L) Intrachild SD (µg/L) Intrachild CV
b
Analyte Mean Min Max Mean Min Max Mean (%) Min (%) Max (%)
1-NAP 2.9 <1.3 20 2.8 0 31 74 0 160
MDA 1.4 <1.0 8.4 1.4 0 12 59 0 158
TCPy 9.2 <1.4 37 4.8 0.5 19 56 5.9 134
Abbreviations: Min, minimum; Max, maximum.
a
Values < the DL assigned one-half the DL associated with each sample to calculate mean and SD.
b
CV calculated by
dividing intrachild SD by intrachild mean.
Table 4. Unweighted summary statistics of the distribution of pesticide metabolites (µg/L) and creatinine
(g/L) for all urine samples collected from 90 children.
Frequency of GM
Analyte
n
a
detection (%)
b
Range Mean
c
SE GM
c
SE
1-NAP 258 45.3 < DL-55 3.0 0.34 1.4 1.1
AM 262 2.3 < DL-16 0.55 0.10 —
d
—
d
MDA 262 36.6 < DL-23 1.4 0.18 0.7 1.1
TCPy 261 93 < DL-45 9.2 0.48 6.4 1.1
Creatinine 263 100 0.32-3.4 1.1 0.031 1.0 1.0
GM, geometric mean.
a
Number of successful chemical analyses, with a maximum potential
n
= 266. Variation in numbers reflects lack of a valid
analytical result for a specific metabolite within samples as well as exclusion of 4 very dilute samples (creatinine values
less than 0.3 g/L).
b
For metabolites with two DLs (see “Methods”), frequencies were calculated based on the DL associ-
ated with each sample.
c
Pesticide metabolite values < the DL assigned half the DL.
d
Insufficient percent detectable for
calculation.
levels when compared to children from
households with incomes < $30,000 (t =
2.02, p = 0.047), $50,000–$75,000 (t =
1.84, p = 0.07), and > $75,000 (t = 2.68, p =
0.009). Children from households with
incomes between $50,000–$75,000 had sig-
nificantly lower log TCPy levels compared to
households with incomes < $30,000 (t =
2.56, p = 0.012) and $30,000–$50,000 (t =
2.55, p = 0.013).
Levels in “high use” households. It is
notable that the mean log intrachild metabo-
lite concentrations did not vary significantly
between children from “high pesticide use”
households [i.e., one of the four target
MNCEPS pesticides present and used inside
the home in the previous 12 months (20)]
and children from households where pesti-
cides use was less frequently reported.
However, the scoring process used to place
homes in the “high use” category was non-
specific: reported use of any single pesticide
resulted in inclusion in this group, so a
reported use does not necessarily correlate
with a specific metabolite.
To further explore the variability in expo-
sure we examined questionnaire responses for
the urban and non-urban groups, including
the 10 nonurban children residing on work-
ing farms. These questionnaires included
queries about recent pesticide exposures: a)
the number of days a child was present while
pesticides were being mixed or prepared
[from the time–activity diary (TAD)]; b) the
number of days a child was present while
pesticides were being applied (TAD); c)
whether chemicals for control of fleas,
roaches, ants, or other insects were used
inside the home in the previous month
[from the follow-up questionnaire (FQ)];
and d) whether chemicals for control of fleas,
roaches, ants, or other insects were used on
the exterior or foundation of the home in
the previous month (FQ).
Five children who provided three urine
samples reported being present while pesti-
cides were mixed for 1 day during the moni-
toring week. Three of the five children were
from nonurban households, but none were
from working farms. Weighted log TCPy
levels were noticeably higher (11.5 µg/L, SE
= 2.9) in these five children compared to lev-
els in children who were not around pesti-
cides being mixed (6.9 µg/L, SE = 0.58).
The levels of 1-NAP, AM, and MDA did
not vary systematically between children pre-
sent during mixing and those who were not
present during mixing.
Sixteen children reported being present
during pesticide applications inside or out-
side the home during the monitoring week.
Ten of the 16 children were from nonurban
households, and three were from working
farms. The number of the days they were
present during pesticide applications varied
from 1 to 3 days: There were 9 children (2
from working farms) present for 1 day, 4
children (0 from working farms) present for
2 days, and 3 children (1 from a working
farm) present for 3 days. Log metabolite lev-
els did not vary systematically between chil-
dren present during applications and those
who were not present during applications.
Forty-five of the 90 households reported
pesticide use inside the home in the previous
month, while 18 reported pesticide use out-
side the home in the previous month. The
mean log metabolite levels did not vary sys-
tematically between children from house-
holds reporting indoor or outdoor use in the
past month and those reporting no use.
Weighted population distribution. Table
8 summarizes the weighted distributions of
1-NAP, AM, MDA, and TCPy for all urine
samples, as well as comparable data from
studies performed in adults. As a conse-
quence of the weighting, the MNCPES
metabolite results estimate mean and
variability in a population of more than
84,000 children in the census tracts sampled.
As can be seen by comparing these results
with those in Table 4, the weighted and
unweighted distributions are similar in terms
of percent greater than the detection limit,
mean, and variability. The weighted 95%
confidence interval (CI) of the means are
skewed right and relatively small compared
to the overall range of 1-NAP, MDA, and
TCPy. Although the weighted estimates of
central tendency vary within relatively tight
bounds, the upper bound of the distributions
is known with less certainty. Nonetheless,
this is a reasonable estimate of the mean and
shape of the distribution for these metabo-
lites in the more than 84,000 children repre-
sented by this sample.
Discussion
This is one of the first probability-based sam-
ples of children’s urinary pesticide levels con-
ducted in the United States. One to three
first-morning-void urine samples were
obtained over 5 days from 90 children
between the ages of 3 and 13 and analyzed
for metabolites of commonly used pesticides.
To increase the likelihood of obtaining
detectable concentrations of target pesticides
in environmental and biological samples,
MNCPES oversampled households reporting
frequent pesticide use. Since this oversam-
pling was intentional, and probabilities of
selection and response were assessed system-
atically in each phase of the study, it is possi-
ble to extrapolate our results to obtain
estimates of the population distribution of
metabolites for similar-age children in the
sampled census tracts.
Comparison of MNCPES with recent
studies. The two largest studies that have
reported measurements of urinary pesticide
metabolites in children are not directly com-
parable to this study because they examined
total alkyl- or dialkylphosphate metabolites,
which are specific to OPs as a class but can-
not be traced to specific parent compounds
(13,14).
Most of the same urinary metabolites
measured in this study were also measured as
part of a) a probability-based cross-sectional
sample of 1,000 adults conducted between
1988 and 1994 as part of the National
Health and Nutrition Examination Survey
III (NHANES III) (7) and b) a convenience
sample of 80 adults from Maryland sampled
up to six times over 1 year beginning in
September 1995 as part of the National
Human Exposure Assessment Survey
(NHEXAS–MD) (12).
Children’s Health • Children’s pesticide exposure
Environmental Health Perspectives
•
VOLUME 109 | NUMBER 6 | June 2001
587
Table 6. Summary of urinary metabolite concentrations (µg/L) > the detection limit (DL) for children with
valid analysis in three urine samples.
Number of children
With valid analysis With With With With With GM metabolite
in 3 urine 0 samples 1 sample 2 samples 3 samples concentration > DL
Metabolite samples
a
> DL > DL > DL > DL (unweighted %)
b
1-NAP 80 14 37 15 14 38 (48%)
AM 83 78 4 1 0 3 (3.6%)
MDA 83 30 21 26 6 38 (46%)
TCPy 83 0 3 11 69 81 (98%)
a
Eighty-seven children provided three urine samples.
b
GMs calculated with values < the DL assigned one-half the DL
associated with each sample.
Table 7. Weighted log mean differences between pesticide metabolite levels (µg/L) for urban and nonur-
ban children.
Urban GM
a
Nonurban GM Mean SE of mean
Analyte (
n
)(
n
) difference
b
difference
b
t
-Statistic
p
-Value
1-NAP 1.7 (58) 1.2 (22) 0.33 0.21 1.6 0.13
MDA 0.77 (58) 0.61 (25) 0.24 0.14 1.7 0.099
TCPy 7.2 (60) 4.7 (23) 0.43 0.20 2.1 0.036
a
Group GM calculated from intrachild GM values presented in Table 6.
n
= the unweighted number of children in each
location category.
b
Log scale.
Table 8 summarizes and compares the dis-
tributions of metabolite levels in adults sam-
pled once during NHANES III, adults
sampled repeatedly as part of NHEXAS–MD,
and the children sampled 1–3 times in
MNCPES. In the NHANES III and
NHEXAS–MD studies, 1-NAP was detected
at a rate of 86%, with median concentration
of 4.4 and 4.2 µg/L, respectively. The fre-
quency of detection was 1.7 times greater
than that observed in MNCPES, with
median and 95th percentile values approxi-
mately four times the levels observed in
MNCPES children. 1-NAP was the only
metabolite for which the frequency of detec-
tion, median, and upper-bound urinary con-
centrations were higher for NHANES III and
NHEXAS–MD adults than for MNCPES
children. Levels of AM were not measured in
the NHANES III population, and it was
detectable in only 0.3% of adult samples
obtained over 1 year in NHEXAS–MD, com-
pared with 2.6% of children’s samples mea-
sured during the summer months. Levels of
MDA were detected in 36% of MNCPES
children, but only in 6.6% of NHEXAS–MD
adults. Median and 95th percentile MDA lev-
els were approximately 4 times higher in
MNCPES children than in NHEXAS–MD
adults. Levels of TCPy were detectable
in 97% of MNCPES children, 96%
of NHEXAS–MD adults, and 82% of
NHANES III adults, with median and 95th
percentile values 2.4 and 1.4 times higher in
MNCPES children than in NHANES III
and NHEXAS–MD adults, respectively. For
1-NAP, MDA, and TCPy, the highest
observed values in these studies were obtained
in samples from adult subjects.
Study design limitations. MNCPES was
a Phase III NHEXAS study, designed as a
pilot to develop and evaluate methods and
approaches for future large-scale national
exposure studies (16). Several limitations of
the design and issues of implementation pro-
vide important insights for investigators who
will conduct future probability-based studies
of children’s exposures. In this study the use
of a commercially available phone list pro-
duced a population with incomes approxi-
mately 40% greater than the median for the
census tracts sampled, as well as a relatively
low proportion of renters (21). Although
inner-city census tracts were oversampled,
and the population obtained reflected the
racial and ethnic makeup of the census tracts
as of 1990, this study obtained urine samples
from relatively few nonwhites (n = 13) and
Hispanics (n = 5). Additional studies need to
be performed to obtain statistically robust
and accurate characterization of these sub-
populations to address issues of income- and
race/ethnicity-related disparities in exposure
and health effects. Lastly, this study mea-
sured metabolite levels in a relatively wide
age range compared to other recent studies
of children’s pesticide exposures (13,14),
which were confined to children 3–6 years of
age; no studies published to date have
reported metabolite levels in children < 3
years of age. Although metabolite levels did
not vary systematically between children < 6
and older children in this study, obtaining
data on exposure levels for children < 3 will
require improved methods for both recruit-
ing children and obtaining urine samples
with sufficient volume for analysis of multi-
ple metabolites.
Implications for exposure and risk assess-
ment. Biomarkers of exposure can be used to
characterize the relative magnitude of expo-
sure within populations or population sub-
groups, as inputs to epidemiological
investigations of health effects associated
with chemical exposures, and as components
of risk assessments and risk management
decisions (29). In addition, they can be used
to check the validity of models and pathways
analyses. The results of this study supply a
cross-sectional snapshot of exposure levels in
a probability-based sample of children from
urban and nonurban areas of Minnesota.
These results can potentially provide a base-
line for evaluating trends over time—e.g.,
TCPy levels to compare with levels observed
in future studies examining the effect of the
U.S. Environmental Protection Agency’s
recent regulatory decision limiting indoor
and outdoor chlorpyrifos use (18).
Levels of 1-NAP were less frequently
detectable and were lower in children com-
pared to the NHANES III adult reference
range, but MDA was detected more fre-
quently and found at higher concentrations
than in NHEXAS–MD adults. TCPy was
detected more frequently in children than in
the NHANES III adult reference range popu-
lation, and levels were more than 2 times
higher at both the median and 95th percentile
compared to the adult reference range. Results
from the adults in the NHEXAS–MD study
indicated that TCPy levels vary seasonally,
with the highest levels occurring in the sum-
mer, while 1-NAP levels did not vary season-
ally. The MNCPES was conducted over a
single summer, which is likely to represent the
period of highest pesticide use in Minnesota.
In this study we measured urinary
metabolite levels up to 3 times over a 5-day
period. Individual metabolites did not vary
in concert, and the data are consistent with a
number of potentially overlapping explana-
tions: relatively infrequent domestic applica-
tions, varying residue levels in the diet, and
the rather short half-lives of these com-
pounds in the body. Interchild variability for
1-NAP, TCPy, and, to a lesser extent, MDA
was greater than intrachild variability, sug-
gesting that relatively large sample sizes and
Children’s Health • Adgate et al.
588
VOLUME 109 | NUMBER 6 | June 2001
•
Environmental Health Perspectives
Table 8. Comparison of weighted 1-NAP, AM, MDA, and TCPy distributions from MNCPES children with distributions observed in adults sampled once in NHANES
III (
7
) and repeatedly in NHEXAS–MD (
12
). All values in units of µg/L except as indicated.
Weighted 95% CI
Percent > of the mean
Analyte/Study
n
a
DL Mean
b
SE Lower Upper GM
b
25th% 50th% 75th% 95th% Max
1-NAP
MNCPES 258 52
c
3.9 0.7 2.5 5.3 1.6 < 1.3 1.0
c
4.0 14 55
NHANES III 983 86 17 NR NR NR NR 1.7 4.4 12 43 2,500
NHEXAS–MD 338 85 34 NR NR NR 4.2 1.6 4.2 9.3 52 2,500
AM
MNCPES 262 2.6
c
0.6 0.13 0.35 0.86 —
d
< 1.0 < 1.0 <1.0 <1.0 16
NHEXAS–MD 348 0.3 —
d
NR NR NR —
d
< 1.0 < 1.0 <1.0 <1.0 1.5
MDA
MNCPES 262 36
c
1.7 0.3 1.1 2.3 0.74 < 1.0 < 1.0 1.1 8.7 23
NHEXAS–MD 347 6.6 —
d
NR NR NR —
d
< 1.0 < 1.0 <1.0 2.0 51
TCPy
MNCPES 261 97
c
9.6 0.9 7.8 11 7.0 4.4 7.2 11 26 45
NHANES III 993 82 4.5 NR NR NR NR 1.3 3.0 5.9 13 77
NHEXAS–MD 346 96 6.8 NR NR NR 5.1 3.1 5.3 9.4 17 51
NR, not reported.
a
Unweighted number of urine samples.
b
Pesticide metabolite values < the DL assigned one-half the DL.
c
Weighted frequencies calculated and weighted percentiles interpolated based
on the DL associated with each sample.
d
Insufficient percent detectable to calculate.
multistage, probability-based designs are
needed to accurately characterize the
extremes of metabolite distributions for the
general population of children.
Considerable intrachild variability was
observed as well, with mean CVs ranging
from 56% to 74% and maxima ranging from
134% to 160% for 1-NAP, MDA, and
TCPy. The average intrachild range of values
was 1.3 times the weighted population mean
for 1-NAP, 1.5 times the weighted popula-
tion mean for MDA, and approximately
equal to the weighted population mean for
TCPy. As with the more robust repeated
measure design (up to 6 measurements over a
year) used in the NHEXAS–MD study, this
suggests that a single measurement of these
metabolites is insufficient to characterize the
relative magnitude of long-term exposure to
parent compounds (12).
Both time–activity diaries (filled out
daily by children, sometimes with parental
assistance) and follow-up questionnaires
(administered by study staff at the end of the
week of monitoring) were employed in this
study to obtain information on potential
exposures. It is notable that the only descrip-
tive data that appear to explain variability in
individual exposure levels were recorded by
children on their TADs, and that the two
follow-up questionnaires did not explain any
of the variability in urinary metabolite levels
in study children.
Existing pharmacokinetic models for
these compounds need to be developed
and/or updated for the parent compounds
measured in this study, and none of the
existing models have been validated for chil-
dren. Although we present the distribution
of creatinine concentrations for these chil-
dren, reported metabolite concentrations
were not adjusted for creatinine excretion.
At present it is not clear that creatinine
adjustment will necessarily improve the cor-
relation between exposure and dose (30),
because this assumption has not been sys-
tematically validated in children (31).
Although adjustment with creatinine appears
to introduce additional variability in
metabolite levels, it does not appear to affect
the trends we have observed. For example,
when metabolite levels in urban and nonur-
ban children are compared using creatinine-
adjusted data, urban children who provided
3 urine samples still had significantly higher
TCPy concentrations than nonurban chil-
dren (t = 2.39, p = 0.019), and there were
still no significant differences between 1-
NAP (t = 1.68, p = 0.097) and MDA (t =
1.42, p = 0.16) levels in urban and nonurban
children. Derivation of exposure and dose of
parent compounds must therefore account
for intrachild variability in elimination, the
relative contribution of pesticide metabolites
in the environment to concentrations mea-
sured in the urine after adsorption, distribu-
tion, and metabolism, and variability
potentially introduced by creatinine adjust-
ment if that is used to compensate for urinary
dilution.
Conclusions
We have presented individual and popula-
tion-level urinary metabolite data indicating
widespread exposure of children to the par-
ent compounds of carbaryl or naphthalene;
chlorpyrifos, chlorpyrifos-methyl, or tri-
clopyr; and malathion. In a population rep-
resenting more than 84,000 children,
intrachild GM urinary metabolite levels
measured over 5 days were greater than the
detection limit 98%, 48%, and 46% of the
time for TCPy, 1-NAP, and MDA, respec-
tively, whereas metabolites of atrazine were
detected 4% of the time. It is notable that
metabolite levels were not significantly
higher for subjects from households report-
ing higher than average pesticide use, likely
due to a lack of recent applications.
Metabolite levels varied widely between and
within children, and metabolite concentra-
tions were not correlated within a child’s
own urine samples. Although there was no
systematic relationship between metabolite
levels and most sociodemographic factors,
TCPy levels were higher for urban compared
to nonurban children for unknown reasons.
Ninety-fifth percentile MDA and median
and 95th percentile TCPy urinary concen-
trations were up to two times higher in chil-
dren than levels observed in two comparable
studies of adults. Overall, children’s metabo-
lite levels to these OP pesticides were greater
than in recent population-based studies of
adults in the United States.
REFERENCES AND NOTES
1. National Research Council. Pesticides in the Diets of
Infants and Children. Washington, DC:National Academy
Press, 1993.
2. Landrigan PJ, Claudio L, Markowitz SB, Berkowitz GS,
Brenner BL, Romero H, Wetmur JG, Matte TD, Gore AC,
Godbold JH, et al. Pesticides and inner-city children:
exposures, risks, and prevention. Environ Health
Perspect 107(suppl 3):431–437 (1999).
3. Gurunathan S, Robson M, Freeman N, Buckley B, Roy A,
Meyer R, Bukowski J, Lioy PJ. Accumulation of chlor-
pyrifos on residential surfaces and toys accessible to
children. Environ Health Perspect 106:9–16 (1998).
4. Zartarian VG, Ozkaynak H, Burke JM, Zufall MJ, Rigas
ML, Furtaw EJ Jr. A modeling framework for estimating
children’s residential exposure and dose to chlorpyrifos
via dermal residue contact and nondietary ingestion.
Environ Health Perspect 108:505–514 (2000).
5. Food Quality Protection Act of 1996. Public Law 104–170,
1996.
6. Cohen Hubal EA, Sheldon LS, Burke JM, McCurdy TR,
Berry MR, Rigas ML, Zartarian VG, Freeman NCG.
Children’s exposure assessment: a review of factors
influencing children’s exposure, and the data available
to characterize and assess that exposure. Environ Health
Perspect 108:475–486 (2000).
7. Hill R Jr, Head SL, Baker S, Gregg M, Shealy DB, Bailey
SL, Williams CC, Sampson EJ, Needham LL. Pesticide
residues in urine of adults living in the United States: ref-
erence range concentrations. Environ Res 71:99–106
(1995).
8. Hill RH Jr, Ashley DL, Head SL, Needham LL, Pirkle JL.
p
-
Dichlorobenzene exposure among 1,000 adults in the
United States. Arch Environ Health 50:277–280 (1995).
9. Aprea C, Betta A, Catenacci G, Lotti A, Magnaghi S,
Barisano A, Passini V, Pavan I, Sciarra G, Vitalone V, et
al. Reference values of urinary 3,5,6-trichloro-2-pyridinol
in the Italian population—validation of analytical method
and preliminary results (multicentric study). J AOAC Int
82:305–312 (1999).
10. Fenske RA, Kissel JC, Lu C, Kalman DA, Simcox NJ, Allen
EH, Keifer MC. Biologically based pesticide dose esti-
mates for children in an agricultural community. Environ
Health Perspect 108:515–520 (2000).
11. Gallo M, Lawryk N. Organic phosphorus pesticides. In:
Handbook of Pesticide Toxicology, Vol 2 (Hayes W Jr,
Laws E Jr, eds). New York:Academic Press, 1991;917–1123.
12. MacIntosh DL, Needham LL, Hammerstrom KA, Ryan PB.
A longitudinal investigation of selected pesticide
metabolites in urine. J Expo Anal Environ Epidemiol
9:494–501 (1999).
13. Aprea C, Strambi M, Novelli MT, Lunghini L, Bozzi N.
Biologic monitoring of exposure to organophosphorus
pesticides in 195 Italian children. Environ Health
Perspect 108:521–525 (2000).
14. Loewenherz C, Fenske RA, Simcox NJ, Bellamy G,
Kalman D. Biological monitoring of organophosphorus
pesticide exposure among children of agricultural work-
ers in central Washington State. Environ Health
Perspect 105:1344–1353 (1997).
15. Quackenboss JJ, Pellizzari ED, Shubat P, Whitmore RW,
Adgate JL, Thomas KW, Freeman NC, Stroebel C, Lioy
PJ, Clayton AC, et al. Design strategy for assessing
multi-pathway exposure for children: the Minnesota
Children’s Pesticide Exposure Study (MNCPES). J Expo
Anal Environ Epidemiol 10:145–158 (2000).
16. Sexton K, Callahan MA, Bryan EF, Saint CG, Wood WP.
Informed decisions about protecting and promoting pub-
lic health: rationale for a National Human Exposure
Assessment Survey. J Expo Anal Environ Epidemiol
5:233–256 (1995).
17. Kutz FW, Cook BT, Carter-Pokras OD, Brody D, Murphy
RS. Selected pesticide residues and metabolites in urine
from a survey of the U.S. general population. J Toxicol
Environ Health 37:277–291 (1992).
18. U.S EPA. Chlorpyrifos Revised Risk Assessment and
Agreement with Registrants. Washington, DC:U.S.
Environmental Protection Agency, Office of Prevention,
Pesticides, and Toxic Substances, 2000.
19. Racke KD. Environmental fate of chlorpyrifos. Rev
Environ Contam Toxicol 131:1–150 (1993).
20. Adgate J, Clayton C, Quackenboss J, Thomas K,
Whitmore R, Pellizzari E, Lioy P, Shubat P, Stroebel C,
Freeman N, et al. Measurement of multi-pollutant and
multi-pathway exposures in a probability-based sample
of children: practical strategies for effective field stud-
ies. J Expo Anal Environ Epidemiol 10:650–661 (2000).
21. Adgate JL, Kukowski A, Stroebel C, Shubat PJ, Morrell S,
Quackenboss JJ, Whitmore RW, Sexton K. Pesticide stor-
age and use patterns in Minnesota households with chil-
dren. J Expo Anal Environ Epidemiol 10:159–167 (2000).
22. Lioy PJ, Edwards RD, Freeman N, Gurunathan S,
Pellizzari E, Adgate JL, Quackenboss J, Sexton K. House
dust levels of selected insecticides and a herbicide mea-
sured by the EL and LWW samplers and comparisons to
hand rinses and urine metabolites. J Expo Anal Environ
Epidemiol 10:327–340 (2000).
23. Hill RH Jr, Shealy DB, Head SL, Williams CC, Bailey SL,
Gregg M, Baker SE, Needham LL. Determination of pesti-
cide metabolites in human urine using an isotope dilution
technique and tandem mass spectrometry. J Anal
Toxicol 19:323–329 (1995).
24. Beeson MD, Driskell WJ, Barr DB. Isotope dilution high-
performance liquid chromatography/tandem mass spec-
trometry method for quantifying urinary metabolites of
atrazine, malathion, and 2,4-dichlorophenoxyacetic acid.
Anal Chem 71:3526–3530 (1999).
25. Shealy D, Bonin M, Wooten J, Ashley D, Needham L,
Bond A. Application of an improved method for the
analysis of pesticides and their metabolites in the urine
Children’s Health • Children’s pesticide exposure
Environmental Health Perspectives
•
VOLUME 109 | NUMBER 6 | June 2001
589
of farmer applicators and their families. Environ Int
22:661–675 (1996).
26. SAS Institute Inc. Statistical Analysis Systems (SAS),
Version 8.01. Cary, NC:SAS Institute Inc., 2000.
27. Shah B, Barnwell B, Bieler G. SUDAAN. Research
Triangle Park, NC:Research Triangle Institute, 1997.
28. Neter J, Kutner M, Nachtsheim C, Wasserman W. Applied
Linear Statistical Models. Chicago:McGraw-Hill, 1996.
29. Sexton K, Callahan MA, Bryan EF. Estimating exposure
and dose to characterize health risks: the role of human
tissue monitoring in exposure assessment. Environ
Health Perspect 103(suppl 3):13–29 (1995).
30. Boeniger MF, Lowry LK, Rosenberg J. Interpretation of
urine results used to assess chemical exposure with
emphasis on creatinine adjustments: a review. Am Ind
Hyg Assoc J 54:615–627 (1993).
31. O’Rourke MK, Lizardi PS, Rogan SP, Freeman NC,
Aguirre A, Saint CG. Pesticide exposure and creatinine
variation among young children. J Expo Anal Environ
Epidemiol 10:672–681 (2000).
Children’s Health • Adgate et al.
590
VOLUME 109 | NUMBER 6 | June 2001
•
Environmental Health Perspectives