Methodology for collecting, storing, and analyzing human milk for volatile
Benjamin C. Blount,*aDavid O. McElprang,aDavid M. Chambers,aMichael G. Waterhouse,a
Katherine S. Squibbband Judy S. LaKindbc
Received 23rd December 2009, Accepted 4th March 2010
First published as an Advance Article on the web 31st March 2010
Biomonitoring, or the measurement of environmental chemicals in human tissues and fluids, is used to
supplement—and in some cases replace—more traditional exposure assessments which measure
chemicals in environmental media. Volatile organic compounds (VOCs) in physiological fluids are
biomarkers of exposure that present numerous challenges for sample collection and analysis. To date,
a thorough evaluation of methods for collection and analysis of breast milk samples for volatiles has
not been conducted. In this paper, we describe the development and validation of methods for
collecting, storing, and analyzing 36 volatile organic compounds (VOCs) in breast milk to assess VOC
exposure of lactating women and nursing infants. Volatile analyte loss was minimized by collecting and
storing samples in containers with small headspace volume resulting in recovery $70% for all 10 VOCs
detected in most breast milk samples. Potential contamination by chloroform, benzene, toluene,
ethylbenzene, xylenes, and methyl-tert-butyl ether was minimized by using specially treated sample
collection materials. Method detection limits in the low parts per trillion range were achieved by using
solid-phase microextraction headspace sampling, gas chromatography, and selective ion monitoring
mass spectrometry. We used this method to analyze 3 mL aliquots of breast milk collected from 12
women and found that 10 of the 36 VOCs were detectable in most samples (median values follow): m/p-
xylene, 0.539 ng mL?1; toluene, 0.464 ng mL?1; 1,4-dichlorobenzene, 0.170 ng mL?1;
tetrachloroethylene, 0.165 ng mL?1; o-xylene, 0.159 ng mL?1; ethylbenzene, 0.0149 ng mL?1; styrene,
0.129 ng mL?1; benzene, 0.080 ng mL?1; chloroform, 0.030 ng mL?1; and methyl-tert-butyl ether, 0.016
Volatile organic compounds (VOCs) are commonly found in the
environment, and originate from multiple natural and anthro-
pogenic sources. Common sources of VOCs include tobacco
petroleum products (including fuel and engine
exhaust),3,4building materials,5household and personal care
products,6and water disinfection byproducts.7The prevalence of
VOCs in the environment can result in human exposure through
ingestion, inhalation, and dermal absorption. Exposure can be
assessed by a traditional assessment in which VOCs in air are
measured and intake estimated using assumptions regarding
inhalation rates and other exposure parameters. Biomonitoring,
or the measurement of environmental chemicals in human tissues
and fluids, is used to supplement—and in some cases replace—
more traditional exposure assessments. While earlier bio-
monitoring efforts focused on persistent compounds, a growing
number of biomonitoring studies have included analyses of
short-lived volatile compounds. VOC measurements in blood8–11
indicate widespread non-occupational exposure to some VOCs
in US adults,12and it is likely that infants also experience wide-
spread exposures via the same routes as adults and also via breast
aUS Centers for Disease Control and Prevention, Division of Laboratory
Sciences, National Center for Environmental Health, Atlanta, GA,
30341, USA. E-mail: firstname.lastname@example.org; Fax: +1 404 638 5317; Tel: +1 770
bDepartment of Epidemiology and Preventive Medicine, University of
Maryland School of Medicine, Baltimore, MD, 21201, USA
cLaKind Associates, LLC, Catonsville, MD, 21228, USA
Biomonitoring, or the measurement of environmental chemicals in human tissues and fluids, is used to supplement—and in some
cases replace—more traditional exposure assessments which measure chemicals in environmental media. Biomonitoring of physi-
ologically short-lived volatile compounds presents numerous challenges. Sample collection and analysis for these compounds are
difficult and robust methods are needed for accurate human exposure assessments. To date, a thorough evaluation of methods for
collection and analysis of breast milk samples for volatiles has not been conducted. In this paper, we propose best practices for
collection and analysis of breast milk for volatile compounds to improve and advance the science of infant exposure assessment.
This journal is ª The Royal Society of Chemistry 2010 J. Environ. Monit., 2010, 12, 1265–1273 | 1265
PAPERwww.rsc.org/jem | Journal of Environmental Monitoring
VOCs are of particular interest because (i) there is known
widespread exposure to adults and children, (ii) there is known or
suspected toxicity for the relevant routes of exposure, and (iii)
there is a limited database on actual human exposures (especially
infant exposures) and body burdens.17Nationwide studies of
blood samples from individuals in the US have demonstrated
that exposure to VOCs can lead to measurable body burdens of
these compounds.12,18However, there is a dearth of information
on VOC exposures prenatally and during infancy, a time when
humans may be most sensitive to toxicants. The primary route of
exposure for the fetus is via maternal blood through the placenta,
whereas infants are exposed to VOCs via inhalation and dermal
exposures. VOCs have also been detected in human milk;13,15,19
thus, breast feeding is an additional route of perinatal exposure
to VOCs. While understanding the nature and quantifying the
magnitude of exposure are necessary to determine whether and
under what conditions chemicals in human milk may pose an
unreasonable risk to the health and development of breastfed
infants, there is a paucity of data regarding human milk as
a source of infant exposure to VOCs.
The limited extant information on VOC levels in human milk
derives from a small number of direct measurements19and from
estimates based on physiological models.20,21As a result, at
present, reliable exposure or risk estimates for infants from
lactational exposures to VOCs are not available. It is well-known
that exposure to VOCs can produce adverse effects on the central
nervous system (CNS).22These effects have been observed in
both adults23–25and young children,26although the preponder-
ance of evidence for neurotoxicity in humans relates to adult
exposures. Effects on the CNS range from minimal and revers-
ible to pronounced and irreversible.22The maturing fetus, infant
and child may be particularly vulnerable because their nervous
systems are thought to be more sensitive to the effects of neu-
rotoxicants.22In addition, chronic exposure to some VOCs has
been associated with increased risk for cancer.27,28
Thus, measuring milk VOC levels addresses an important gap
in our understanding of infant exposures and potential health
compounds are difficult and robust methods are needed for
accurate infant exposure assessments.
To date, a thorough evaluation of methods for collection,
handling, and analysis of breast milk samples for volatiles has
not been conducted. In this paper, we describe best practices for
collection and analysis of breast milk for volatile compounds to
improve and advance the science of infant exposure assessment.
Specifically, in early studies, analytical methods for measuring
Table 1 GC/MS parameters and reportable ranges achieved for VOCs in human blood
ISTD ions (m/z)
aISTD ¼ Isotopically Labeled Internal Standard.bSIM ¼ Selective Ion Monitoring.
1266 | J. Environ. Monit., 2010, 12, 1265–1273 This journal is ª The Royal Society of Chemistry 2010
VOCs in human milk involved extraction by purge and trap
(P&T) and dynamic headspace8methods. However, P&T
methods can be limited by breakthrough, foaming, and the
potential for contamination from the purging gas. For example,
a helium-purge gas can contain petroleum-based residue (e.g.,
monoaromatics and hydrocarbons), the levels of which can
become significant when working with lipophilic matrices at
trace concentrations (ng mL?1–pg mL?1).29Pellizari et al.13used
P&T methods to identify nearly 200 VOCs in human milk, but
their results were qualitative, and certain analytes were subject to
contamination during analysis. Liquid-extraction approaches
lacked adequate sensitivity for certain VOCs at non-occupa-
tional exposure levels.30Recently, Kim et al.19demonstrated the
use of solid-phase microextraction gas chromatography mass
spectrometry (SPME/GC/MS) to analyze four VOCs in 13
human milk samples. This static headspace sampling approach,
proven effective in analyzing other biological samples,31over-
comes many of the limitations associated with liquid extraction
and P&T. SPME eliminates the need for milk samples to come in
contact with potentially contaminating liquid and gas-extraction
matrices, while permitting sampling and preconcentration in an
air-tight vial. SPME analysis of VOCs also reduces sample
preparation time and method cost.
We used the SPME method in conjunction with GCMS to
develop a comprehensive method that quantifies 36 VOCs in
milk (Table 1). By quantifying a broad spectrum of VOCs, we
can assess exposure to cigarette smoke, fuel and engine exhaust,
water-disinfection byproducts, and solvents. In addition, we
explored methods for optimizing collecting and handling human
milk samples to minimize volatilization loss and contamination
gain. The result of this work is a method that permits accurate
measurement of 36 VOCs in the parts per trillion range (pg
mL?1). Following method validation, analysis was performed on
breast milk samples from 12 women, and results were compared
with those from other studies.
P&T-grade methanol (Burdick and Jackson, Muskegon, MI) was
used to dilute standard solutions and as a solvent for cleaning
glassware. HPLC-grade water (J.T. Baker, Phillipsburg, NJ) was
VOCs such as chloroform.32This VOC-free water was aliquoted
and flame sealed in glass ampoules (Wheaton, Millville, NJ).
from the Mother’s Milk Bank of Denver and donated by the
Cincinnati Children’s Research Human Milk Bank. Gray-top
glass Vacutainers? (16 ? 100 mm, Becton Dickinson, Franklin
Lakes, NJ) were disassembled, anticoagulant was removed, and
the butyl rubber stoppers were solvent extracted and vacuum
baked to reduce VOC contamination.33For the purposes of this
Glass bottles (30 mL, level 3 treated for EPA Protocol B) were
purchased from Qorpak (Bridgeville, PA). Gas-tight polytetra-
fluoroethylene (PTFE) Luer-Lok? glass syringes (5 mL) were
purchased from Hamilton (Reno, NV). Multipipettor and glass
Inc. (Leighton Buzzard, UK). Headspace vials (10 mL, beveled
top) were purchased from National Scientific Co. (Duluth, GA).
Septa (3.0 mm thick ? 20 mm diameter, PTFE-faced/silicone-
level 4 treated) and crimp seals (aluminium with magnetic insert,
open center) were purchased from Integrated Liner Technologies
GC inlet liners (0.75 mm ID) were purchased from Supelco
(Bellefonte, PA). Chromatographic separation wasperformed on
a J&W 40 m ? 0.18 mm I.D. 1 mm film DB-VRX
column (Agilent Technologies, Palo Alto, CA).
All analytes and internal standards (Table 1) were purchased as
neat compounds of the highest purity available and stored at
?20?C to minimize degradation. Stable isotope-labeled analogs
were13C-labeled primarily to bettermimic the native compounds.
used. Isotopically labeled internal standards are crucial for
maintaining precision when using non-equilibrium headspace
SPME, and thus justify the cost of these materials. Calibration
standards and internal standards were prepared from neat
chemicals and diluted with P&T-grade methanol to intermediate
concentrations. These stock solutions were sealed in glass
The internal standard stock solution was diluted with methanol
and added to blanks, standards, unknowns, or QC samples at
with glass capillary tips (VWR Scientific, West Chester, PA) were
used for all liquid transfers in the microlitre range.10,11
Participant selection and enrollment. Enrollment was open to
healthy women who were breast feeding, at least 30 days post-
partum, and capable of manually expressing milk. A convenience
sample of 12 participants was selected from the Baltimore,
Maryland, metropolitan area via announcements posted in one
pediatric office and by word of mouth. Participants completed
a questionnaire and signed an informed consent form. This study
was approved by the Institutional Review Board (IRB) of the
University of Maryland. The Centers for Disease Control and
Prevention (CDC) IRB relied on that review because CDC
involvement was limited to laboratory analysis of coded samples.
Human milk collection. Milk was collected by one of three
methods: manual expression directly into a Vacutainer?, manual
expression into a 30 mL glass jar then immediate transfer into
a Vacutainer?, or breast-pump (Hollister, Inc., Libertyville, IL)
expression then immediate transfer into a Vacutainer?. During
sample collection, the Vacutainer? was filled to a target volume
of ?7.5 mL. Once the milk was collected into the Vacutainer?,
the sample was sealed with a butyl rubber stopper and held in
place with paperbacked laboratory tape. The samples were
stored upright (to minimize sample contact with the rubber
stopper) at 4?C and shipped weekly via overnight courier in
insulated containers with cold packs to ensure that the samples
remained chilled. Upon arrival, the samples were placed in
a sample storage refrigerator at 4
samples were analyzed within 2 weeks of sample collection. Any
repeat analysis for data confirmation was conducted within
10 weeks of sample collection to ensure reliability of repeat data.
?C until analyzed. Most
This journal is ª The Royal Society of Chemistry 2010 J. Environ. Monit., 2010, 12, 1265–1273 | 1267
Human milk sample analysis. Milk samples were removed from
refrigerated storage (4?C) and allowed to equilibrate to room
temperature. The samples were then mixed with a vortex mixer
(Barnstead International, Dubuque, IA) at maximum agitation
speed for 30 s and placed on a rotary mixer (Robbins Scientific
Corp., Sunnyvale, CA) until time of preparation (0.5–1.0 h).
Subsequently, 3.0 mL of milk was transferred into a precleaned
10 mL headspace vial using a precleaned gastight glass syringe,
then spiked with 40 mL internal standard solution and crimp
sealed. The sample weight was gravimetrically confirmed. Blanks
and quality control samples were prepared in a similar manner.
The method used in this study was modified from an existing
high-throughput SPME-GC-MS method described by Blount
et al.11The system consisted of an Agilent 6890A GC coupled
with a 5973N quadrupole MS (Agilent Technologies). The GC
was equipped with a split/splitless injector operated in the pulsed
splitless mode using a 0.75 mm I.D. liner and a SIS cryotrap
(model 961, Scientific Instrument Services, Ringoes, NJ). The
SIS cryotrap was maintained at ?100?C with liquid nitrogen for
the first minute of analysis then ballistically heated to 225?C
(?25 s) while the GC oven was maintained at 0?C for an addi-
tional 30 s. During the initial 1.5 min of analysis inlet, pressure
wasincreased to 50 psi to maximize sample loading ontothe head
of the column. After this initial pulsed flow of 2.2 mL min?1,
carrier gas flow was maintained at 1.0 mL min?1. The helium
carrier gas (research grade, 99.9999%, Airgas South, Atlanta,
GA) was scrubbed using a two-stage oxygen, moisture, and
hydrocarbon trap manifold (VICI Mat/Sen Inc., Poulsbo, WA).
The GC oven was programmed for the following thermal
gradient: start at 0?C for 1.50 min, ramp at 7?C min?1to 140?C,
ramp at 40?C min?1to 220?C, and maintain this temperature for
8.5 min. SPME headspace sampling was conducted using
a CombiPAL autosampler (CTC Analytics AG, Zwingen,
Switzerland) equipped with two Peltier cooled trays (15 ? 1?C),
heated agitator station (40?C, 500 rpm), and 75 mm Carboxen/
polydimethylsiloxane fiber (15 min collection). This coating was
selected to optimize extraction efficiency and capacity for
a broad range of VOCs.4The fiber was desorbed in the GC inlet
at 250?C and remained there throughout the GC run to ensure
complete analyte desorption, reduce sample-to-sample carry
over, and minimize contamination from laboratory air. Mass
spectrometry was performed using electron ionization, and SIM
with mass resolution was set to a peak width of 0.50. Three ions
(Table 1) were monitored for each analyte: primary quantifica-
tion, confirmation, and internal standard. Dwell times varied
from 10 to 100 milliseconds, depending on the number of ions
monitored within a SIM window. Longer times were given to the
most important ions (e.g., primary ions) and to ions with lower
signals. Batches of unknown samples were bracketed with
blanks, calibrators, and QC materials.
Agilent Chemstation (G1701DA Ver. D.00.00.38, Agilent
Technologies) software was used for instrument control and data
collection. Data analysis was performed using Xcalibur
QuanBrowser (ThermoFinnigan, San Jose, CA), after the raw
data file was exported into ‘‘AIA/Net.CDF’’ format and con-
verted to Xcalibur ‘‘RAW’’ format. Compound identification
was confirmed by both retention time and the ratio of primary
quantification and confirmation ions. A full set of eight cali-
brators, including one blank, was analyzed with each set of
samples to generate calibration curves for that analytical run. A
linear 1/x fit was used to create calibration curves (r2typically
$0.99) that typically spanned three orders of magnitude. The
lowest calibrators ranged from 0.005 to 0.300 ng mL?1, as shown
in Table 1. The method detection limit (MDL) was defined as
three times the standard deviation at zero concentration (S0). S0
was determined by analyzing six sets of the lowest four calibra-
tion standards and plotting the standard concentration versus
standard deviation. When the calculated value for 3S0was lower
than the lowest standard concentration, the lowest reportable
level (LRL) was assigned the value of the lowest standard
concentration.34Repeat analysis was performed for samples with
analyte concentrations greater than the highest calibrator by
diluting with VOC-free water.
Each analysis batch was evaluated at two blind QC-concentra-
tion levels. Three QC pools were prepared, one made with bovine
milk (homogenized whole milk) and two made with pooled
human breast milk obtained from milk banks. Background VOC
levels in this milk were reduced by heating (?37
constantly stirring the milk while purging the headspace above
the milk with purified nitrogen gas for 24 h. The milk was then
divided into two pools and spiked to produce a low- and high-
concentration QC pool. Each QC pool was stirred for 30 min at
?4?C to facilitate equilibration of fortified VOCs. The spiked
milk was then dispensed (7.5 mL aliquots) into unstoppered,
precleaned Vacutainers?. These QC samples were numbered in
the order they were aliquotted, sealed and stored at 4?C.
VOC levels in QC pools were characterized based on 20
separate assays. An independent QC officer evaluated blind QC
results according to modified Westgard QC rules.34If QC sample
results did not meet acceptance criteria for an analyte, all results
for that analyte in that analytical batch were rejected.
VOC contamination was evaluated by analysis of blanks
prepared from VOC-free water. VOC-free blank water was
produced by inert gas sparging (filtered ultra high purity grade
nitrogen), distillation, and flame sealing (by water torch [Eagle
Research, Oroville, WA]) as described by Cardinali et al.32On
the day of analysis, water blanks were spiked with internal
standardand analyzedwith eachbatchof unknowns. Ifthe blank
contained analyte levels exceeding the LRL, the run was flagged
as contaminated for that analyte and those results were rejected.
Proficiency testing (PT) materials were prepared from third party
VOC reference solutions (Environmental Protection Agency Mix
524 rev A, Supelco) as previously described.11Additional ana-
lytes not found in this VOC mixture (e.g., 2,5-dimethylfuran)
1268 | J. Environ. Monit., 2010, 12, 1265–1273 This journal is ª The Royal Society of Chemistry 2010
were added using gravimetrically confirmed amounts of neat
material. Five ampoules at four different concentrations were
analyzed in blinded fashion at least twice per year. On the day of
analysis, PT materials were diluted with VOC-free water and
analyzed as unknowns. An analyte passed PT if blind-analyzed
concentrations fell within 25% of known values.
Results and discussion
Biomonitoring methods that measure VOCs in complex matrices
such as milk must address certain criteria essential for producing
acceptably accurate and precise data. In addition to an accurate,
precise, sensitive, and selective analytical method, the ultimate
goal of accuracyand precision canbe achievedonly bycollecting,
transporting, and storing biological samples reliably while
maintaining sample integrity. Additionally, contamination from
the surrounding environment and laboratory materials with
which the samples come in contact must be minimized. Once
these potential method biases have been minimized and charac-
terized, the method parameters can be reliably optimized.
Accuracy and precision
Accuracy and precision within and among quantitative runs were
characterized with third party verification standards (PT
samples), spiked matrix samples (spike recovery and QCs), and
water blanks. The PT samples were periodically analyzed on
multiple instruments throughout the study to test assay accuracy.
Results from a typical run for the lowest and highest PT levels are
given in Table 2. The error for these results fell below 15% for all
the analytes with the exception of methylene chloride, which is
biased by contamination from the laboratory air.
Method accuracy was also assessed by analyzing human milk
samples fortified with known amounts of VOCs. Five replicates
of the spiked and unspiked milk were prepared for instrumental
analysis. Because the precleaned milk contained significant
levels of some VOC analytes relative to the spiked concentra-
tion, including unspiked milk for background subtraction was
necessary. Results from this analysis are included in Table 2.
The data from each group of replicates were averaged before
calculating apparent recovery. As seen in Table 2, the apparent
recoveries for the majority of the analytes were within ?15% of
theoretical values. For those analytes with >15% error, there
appeared to be two major types of biases. Bias for the benzene
and methylene chloride data likely resulted from background
subtracting relatively high levels of these analyte that occurred
in the unspiked milk (0.046 and 0.109 ng mL?1, respectively).
The significant positive bias of 1,1-dichloroethylene and
trichloroethylene is unexplained at this time. No background
levels or coeluting interferents as determined from the unspiked
control milk sample existed that could explain these analyte
biases. Additionally, an interferent co-eluted with nitrobenzene
and produced a signal at m/z 129 that disproportionately
increased the ISTD signal and thus lowered the response ratio.
Additional spike experiments were not conducted to try to
correct this error.
Results from QC samples can provide a measure of both
precision and relative accuracy over time (e.g., drift). Table 3
shows QC means and relative standard deviations (RSDs)
calculated from 20 independent analyses collected on three
different instruments over a 1 week period. For the 36 VOC
analytes, the mean RSD was 8.6% and 9.1% for the high and low
QC pools, respectively. Regression analysis (p > 0.05) indicated
that RSDs are more a result of imprecision than drift, thus these
average RSD for each level will serve as future reference for
precision among runs.
Sample collection methodological issues
Sample collection container: wide-mouth jars vs. Vacutainers.
Wide-mouth glass jars are commonly used for breast milk sample
collection. While collection into a wide-mouth glass jar is
convenient, some of the volatile analytes listed in Table 1 can be
lost during collection and storage because of the large ratio of
headspace volume to sample volume. Additionally, the large
sealing boundary for wide-mouth jars can impede sealing the
samples hermetically. Due to these potential difficulties, we
investigated using Vacutainers? to allow samples to be stored
under hermetic conditions with a small headspace volume and
For this evaluation, we compared volatilization loss of VOCs
from spiked bovine milk (12 mL) stored in a pretreated
solutionsandbackground-subtractedapparent recoveryfor VOCsspiked
into human milk (n ¼ 5)
Accuracy of analysis of 36 VOCs in proficiency testing (PT)
This journal is ª The Royal Society of Chemistry 2010J. Environ. Monit., 2010, 12, 1265–1273 | 1269
Vacutainer? (?12 mL capacity) versus a wide-mouth glass jar
(?30 mL capacity). This milk was not precleaned but back-
ground corrected using control samples stored in 12 mL glass
screw-cap vials with a PTFE-lined cap. The milk in all containers
was spiked, capped, vortex mixed, and analyzed. Fig. 1 shows the
average percent loss for 36 analytes relative to the control
samples. Sample results were averaged from three replicates each
from duplicate containers (n ¼ 6). Best fit curves are shown in
Fig. 1 to emphasize the difference between the two sample
storage container types. For the first 15–20 eluting analytes, the
sample container loss from the jars exceeded that of the Vacu-
tainers?, with a maximum of 36% loss for furan. Losses for the
other analytes were within the method precision for both
collection container types. These results demonstrate the
importance of collecting milk samples in a low headspace
container, either Vacutainer? or screw-cap vial, for the more
volatile and nonpolar analytes.
Samples in wide-mouthed jars lost additional amounts of
VOCs during storage. Fig. 2 shows the percent lost after storage
for 3 weeks relative to the initial levels. These results exclude
methylene chloride, in which the concentration fell well below
the LRL; 1,1,2,2-tetrachloroethylene and 1,2-dichloropropane
could not be quantified because of coeluting interferences
apparently emanating from the jar materials. When stored in
Vacutainers?, most of the analytes maintained their initial
concentration levels, with the exception of nitrobenzene. In
Fig. 2, the results are plotted according to their calculated
octanol–water partition constant (KOW). The data in Fig. 2
emphasize the relationship between analyte lipophilicity and
loss during jar storage, despite vortex mixing prior to analysis.
One explanation for this lipophilic-dependent loss is that the
jar-headspace air is accelerating curdling, which then causes
partitioning of certain analytes into the coagulated portion of
the milk sample. This coagulated portion can coat the jar
surface and settle out of solution, making collecting a homoge-
neous sample difficult. Additional analyte loss is likely due to
inadequate sample sealing for wide-mouthed jars. These results
underscore the importance of storing milk samples in containers
with minimal headspace and sealing boundaries (e.g. Vacu-
tainer? or screw-cap vial).
Sample collection methodologic issues: manual expression vs.
breast pump. Two collection techniques were investigated:
manual expression into a glass jar and expression by breast pump
Reproducibility of analysis of VOCs in quality control milk (n
Milk: low QCa
Milk: high QC
ng mL?1RSD (%)
aQC ¼ Quality Control.bRSD ¼ Relative standard deviation.
samplesin30 mLjarsand10 mLVacutainers?encounteredafter3weeks
storage at 4?C.
Comparison of relative storage loss for identical human milk
collected samples (plotted by elution order as listed in Table 1).
Percentage sample container loss for glass-jar and Vacutainer?
1270 | J. Environ. Monit., 2010, 12, 1265–1273 This journal is ª The Royal Society of Chemistry 2010
with an attached jar. Manual expression directly into a collection
container allows for the least amount of sample handling and
minimizes contact with potentially contaminated materials.
Initially, we tested the feasibility of having a participant manu-
ally express milk directly into a Vacutainer? and immediately
replace the stopper after collection. This technique worked, but
expressing milk into the small glass vial (16 mm diameter) may be
challenging for some donors. Consequently, the remaining
participants manually expressed samples into wide-mouth glass.
Breast pumps are preferred by some women, but the samples
may be contaminated from contact with pump plastics.
Regardless of whether the samples were manually expressed or
pumped, they were collected in precleaned glass jars then trans-
ferred into precleaned Vacutainers? for storage and shipment.
Even though glass jars are preferred for easier sample collection,
the need for a low headspace container requires transferring the
sample, which likely results in increased handling volatilization
Because both of these sample-collection techniques presented
potential risks for analyte loss and/or contamination, we con-
ducted experiments to determine the extent of these biases. We
simulated breast milk secretion with a syringe pump to investi-
gate the magnitude of contamination gain and volatilization loss
duringthis process.To assess
from external sources (e.g., surrounding air and sample
collection supplies), we discharged precleaned and characterized
milk (N ¼ 3) at a rate of 0.8 mL min?1and measured the amount
of contamination gained relative to initial VOC levels. As shown
in Fig. 3, toluene levels increased during simulated manual
expression. Error bars represent variation of the triplicate results
and propagation of error from background subtraction. The
toluene contamination was near the MDL of 0.025 ng mL?1, and
may have resulted from partitioning of these VOCs into the
sample from the surrounding air.
A second experiment simulated secretion of VOC-spiked milk
(N ¼ 6) to characterize volatilization loss during manual
expression of milk into a Vacutainer?. As shown in Table 4,
analyte boiling point was related to loss during simulated milk
secretion. Analytes with lower boiling points likely volatilized
into the air as the milk dripped into the sample collection vial.
Breast-pump contamination issue. We poured precleaned
human milk through two breast pumps to evaluate the potential
for contamination from components of the breast pump with
VOCs. In these samples (Fig. 4), most analytes showed no
significant change from control levels. Relative to the control
sample (no contact with pump components), these samples had
lactational simulation by syringe pump spraying at 0.8 mL min?1(plotted
by elution order as listed in Table 1).
Contamination gain measured in precleaned milk following
before and after simulated milk expression using a syringe pump
Percent loss in VOC concentrations measured in spiked milk
Analyte Boiling point/?C % Loss
that were exposed to separate breast pumps normalized to a trip-control
sample (plotted by elution order as listed in Table 1).
Contamination gain measured in precleaned human milk samples
This journal is ª The Royal Society of Chemistry 2010 J. Environ. Monit., 2010, 12, 1265–1273 | 1271
slightly higher mean levels of toluene, m/p-xylene, styrene, and
o-xylene that were above the MDL. The relatively large error
bars for these data result from the propagation of error because
of background subtraction. The breast pump may be a preferred
method of collection for some study participants because of its
relative ease of use; thus, with proper characterization, this
sample collection approach appears acceptable.
Pilot study/method application
Milk samples were collected from 12 lactating women. To
identify any contamination bias that occurred during sample
collection, transportation, and storage, we included a character-
ized control human milk sample to assess contamination from
room air during sample collection. Each study participant left
this control sample open while collecting her milk sample to
allow exposure of the control sample to room air. When the
control samples were returned to the laboratory, they were
analyzed and final analyte concentrations were compared to
initial characterizations. These collection-control samples indi-
cated contamination from room air for furan, MTBE, toluene,
xylenes, ethylbenzene, and tetrachloroethylene. The magnitude
of collection-control contamination for these six analytes was
typically close to the LRL. These collection-control levels were
background subtracted from the matching unknown samples.
Table 5 shows VOC levels in milk from the 12 participants for
10 of the 36 compounds measured in the pilot study milk with
median concentrations higher than the LRL. Results determined
from this method are compared with other published data on
VOCs in breast milk in Table 5. Toluene, benzene, MTBE, and
chloroform have been previously quantified,15,19and those results
are generally similar to the results from this study. The only
exceptions were for chloroform and benzene reported by Kim
et al., in which milk concentrations were a factor of 18 and 1.5
higher, respectively, than those measured in this study. The
difference among benzene levels can be explained either in terms
of exposure differences or contamination bias. The chloroform
levels reported by Kim et al.19are high when considering non-
occupational exposure in the US population.12A likely reason
for the difference in the chloroform results is contamination bias,
which can be problematic because chloroform-containing
The lipid-rich character of milk presents many challenges to
ensure unbiased analysis of VOCs. With its relatively high lipid
content compared to other biological matrices (e.g., blood,
urine), milk is more susceptible to contamination through
contact with air and laboratory materials that are required for
sample collection, storage, and analysis. As a result, collecting
and storing samples into precleaned, gas-tight containers are
important. Additionally, suitable control samples are crucial for
identifying process biases for many analytes, including furan,
MTBE, tetrachloroethylene, and toluene. Conversely, the high
lipid content of milk also helps retain VOCs. Simulated lactation
experiments revealed minimal volatilization loss (recovery $
70%) for all 10 VOCs detected in most milk samples. Once
appropriate measures were taken to minimize volatilization loss
and contamination gain, we were able to expand analysis to
include quantification of 36 analytes representing exposure
associated with smoking, fuel and fuel emissions, water
disinfection, and solvent exposure. Analysis of 12 breast-milk
specimens revealed varying levels of 36 different analytes; median
levels for 10 of these analytes were above MDL. Using the
methods described in this paper, a broad range of VOCs can be
accurately quantified in human milk.
The authors thank the participating mothers and their families;
the Cincinnati Children’s Research Human Milk Bank and the
Mother’s Milk Bank of Denver for providing donated milk; and
John Morrow, Janice Menuel and Tim Hughes of the US Centers
for Disease Control and Prevention for technical assistance. The
findings and conclusions in this report are those of the authors
and do not necessarily represent the views of the Centers for
Disease Control and Prevention. Mention of trade names or
commercial products does not constitute endorsement or
recommendation for their use.
1 E. D. Pellizzari and L. A. Wallace, Toxicol. Lett., 1986, 35, 113–116.
2 G. M. Polzin, R. E. Kosa-Maines, D. L. Ashley and C. H. Watson,
Environ. Sci. Technol., 2007, 41, 1297–1302.
3 S. A. Batterman, C. Y. Peng and J. Braun, Atmos. Environ., 2002, 36,
comparison of median VOC levels measured in human milk with previously reported resultsa
Concentration ranges for 10 compounds in 12 manually expressed milk samples after background correction with control samples and
Compound Range/ng mL?1
Median,19n ¼ 8/ng mL?1
Median,15n ¼ 23b/ng mL?1
aPublished median milk VOC concentrations provided from Kim et al.19and Fabietti et al.15for comparison.bAssumes 1 mL milk ¼ 1 g.
1272 | J. Environ. Monit., 2010, 12, 1265–1273 This journal is ª The Royal Society of Chemistry 2010
4 D. M. Chambers, B. C. Blount, D. O. McElprang, M. G. Waterhouse Download full-text
and J. C. Morrow, Anal. Chem., 2008, 80, 4666–4674.
5 C. Yu and D. Crump, Build. Environ., 1998, 33, 357–374.
6 T. M. Sack, D. H. Steele, K. Hammerstrom and J. Remmers, Atmos.
Environ., Part A, 1992, 26, 1063–1070.
7 S. W. Krasner, H. S. Weinberg, S. D. Richardson, S. J. Pastor,
R. Chinn, M. J. Sclimenti, G. D. Onstad and A. D. Thruston, Jr,
Environ. Sci. Technol., 2006, 40, 7175–7185.
8 L. C. Michael, M. D. Erickson, S. P. Parks and E. D. Pellizzari, Anal.
Chem., 1980, 52, 1836–1841.
9 S. R. Antoine, I. R. Deleon and R. M. Odellsmith, Bull. Environ.
Contam. Toxicol., 1986, 36, 364–371.
10 D. L. Ashley, M. A. Bonin, F. L. Cardinali, J. M. Mccraw,
J. S. Holler, L. L. Needham and D. G. Patterson, Jr, Anal. Chem.,
1992, 64, 1021–1029.
11 B. C. Blount, R. J. Kobelski, D. O. McElprang, D. L. Ashley,
J. C. Morrow, D.M. Chambers
J. Chromatogr., B: Anal. Technol. Biomed. Life Sci., 2006, 832,
12 J. E. Churchill, D. L. Ashley and W. E. Kaye, Arch. Environ. Health,
2001, 56, 157–166.
13 E. D. Pellizzari, T. D. Hartwell, B. S. H. Harris, R. D. Waddell,
D. A. Whitaker and M. D. Erickson, Bull. Environ. Contam.
Toxicol., 1982, 28, 322–328.
14 J. Z. Byczkowski, J. M. Gearhart and J. W. Fisher, Nutrition, 1994,
15 F. Fabietti, A. Ambruzzi, M. Delise and M. R. Sprechini, Environ.
Int., 2004, 30, 397–401.
16 J. S. Schreiber, Risk Anal., 1993, 13, 515–524.
17 C. M. Berlin, S. Kacew, R. Lawrence, J. S. LaKind and R. Campbell,
J. Toxicol. Environ. Health, Part A, 2002, 65, 1839–1851.
18 D. L. Ashley, M. A. Bonin, F. L. Cardinali, J. M. Mccraw and
J. V. Wooten, Clin. Chem. (Washington, D. C.), 1994, 40, 1401–
andF. L. Cardinali,
19 S. R. Kim, R. U. Halden and T. J. Buckley, Environ. Sci. Technol.,
2007, 41, 1662–1667.
20 S. Batterman, L. Zhang, S. G. Wang and A. Franzblau, Sci. Total
Environ., 2002, 284, 237–247.
21 J. Fisher, D. Mahle, L. Bankston, R. Greene and J. Gearhart, Am.
Ind. Hyg. Assoc. J., 1997, 58, 425–431.
22 J. V. Bruckner and D. A. Warren, Toxic Effects of Solvents and
Vapors, in Casarett and Doull’s Toxicology: The Basic Science of
Poisons, ed. C. D. Klaasen, McGraw-Hill, New York, 6th edn,
2009, pp. 869–916.
23 E. L. Baker, Annu. Rev. Public Health, 1988, 9, 223–232.
24 F. Gobba and A. Cavalleri, NeuroToxicology, 2003, 24, 693–702.
25 G. D. Ritchie, K. R. Still, W. K. Alexander, A. F. Nordholm,
C. L. Wilson, J. Rossi and D. R. Mattie, J. Toxicol. Environ.
Health, Part B, 2001, 4, 223–312.
26 B.Saddik,I. Nuwayhid,A.
NeuroToxicology, 2003, 24, 733–739.
27 A. R. Schnatter, K. Rosamilia and N. C. Wojcik, Am. J. Epidemiol.,
2006, 163, S235.
28 K. P. Cantor, R. Hoover, P. Hartge, T. J. Mason, D. T. Silverman,
R. Altman, D. F. Austin, M. A. Child, C. R. Key and
L. D. Marrett, J. Natl. Cancer Inst., 1987, 79, 1269–1279.
29 D. M. Chambers, D. O. McElprang, M. G. Waterhouse and
B. C. Blount, Anal. Chem., 2006, 78, 5375–5383.
30 R. Kroneld and M. Reunanen, Bull. Environ. Contam. Toxicol., 1990,
31 F. L. Cardinali, D. L. Ashley, J. V. Wooten, J. M. Mccraw and
S. W. Lemire, J. Chromatogr. Sci., 2000, 38, 49–54.
32 F. L. Cardinali, J. M. Mccraw, D. L. Ashley and M. A. Bonin,
J. Chromatogr. Sci., 1994, 32, 41–45.
33 F. L. Cardinali, J. M. Mccraw, D. L. Ashley, M. Bonin and
J. Wooten, J. Chromatogr. Sci., 1995, 33, 557–560.
34 S. P. Caudill, R. L. Schleicher and J. L. Pirkle, Stat. Med., 2008, 27,
Williamson andD. Black,
This journal is ª The Royal Society of Chemistry 2010 J. Environ. Monit., 2010, 12, 1265–1273 | 1273