Ion chromatographic determination of perchlorate in foods by on-line enrichment and suppressed conductivity detection.

Abstract

Systemic uptake of perchlorate anion, a rocket fuel component and potential thyroid function disruptor, by leafy vegetables and other crops grown in contaminated waters is a public health concern. A column-switching anion-exchange chromatographic method with suppressed conductivity detection, described in this paper, achieved a 3-6 microg/kg method limit of quantitation in analysis of the wet weight edible portion of cantaloupe, carrots, lettuce, and spinach samples with field-incurred perchlorate. A test portion was blended with dilute nitric acid, and the extract was filtered under vacuum. A portion of the measured filtrate was acidified to pH approximately 2 by addition of cation-exchange resin, 4 mL was passed through a graphitized carbon cleanup column, and an aliquot of a collected fraction was pushed through a short precolumn for anion extraction, enrichment, and injection onto the analytical column. Statistical comparison with determination by tandem mass spectrometry-ion chromatography analysis of untreated filtrate revealed that the difference between means was not significant at the 95% confidence level (P value > or = 0.12) for crops tested. In addition, the method was applied to cooked vegetables processed as baby food.

Full text

Page 1

Ion Chromatographic Determination of Perchlorate in Foods by
On-Line Enrichment and Suppressed Conductivity Detection
RICHARD A. NIEMANN,* ALEXANDER J. KRYNITSKY, AND DAVID A. NORTRUP
Center for Food Safety and Applied Nutrition, Food and Drug Administration,
5100 Paint Branch Parkway, College Park, Maryland 20740
Systemic uptake of perchlorate anion, a rocket fuel component and potential thyroid function disruptor,
by leafy vegetables and other crops grown in contaminated waters is a public health concern. A
column-switching anion-exchange chromatographic method with suppressed conductivity detection,
described in this paper, achieved a 3-6 µg/kg method limit of quantitation in analysis of the wet
weight edible portion of cantaloupe, carrots, lettuce, and spinach samples with field-incurred
perchlorate. A test portion was blended with dilute nitric acid, and the extract was filtered under
vacuum. A portion of the measured filtrate was acidified to pH ∼2 by addition of cation-exchange
resin, 4 mL was passed through a graphitized carbon cleanup column, and an aliquot of a collected
fraction was pushed through a short precolumn for anion extraction, enrichment, and injection onto
the analytical column. Statistical comparison with determination by tandem mass spectrometry-ion
chromatography analysis of untreated filtrate revealed that the difference between means was not
significant at the 95% confidence level (P value g 0.12) for crops tested. In addition, the method was
applied to cooked vegetables processed as baby food.
KEYWORDS: Perchlorate; ion chromatography; suppressed conductivity; anion exchange; determination;
food
INTRODUCTION
Perchlorate anion (ClO4-) competitively inhibits iodide uptake
by the thyroid gland’s sodium-iodide symporter (NIS) protein
and, at high levels, may disrupt production of two iodine
hormones [triiodothyronine (T3) and thyroxine (T4)] essential
for metabolic activity. Thiocyanate (SCN-) and nitrate (NO3-)
anions are less potent competitors. When Chinese hamster ovary
cells stably expressing human NIS were exposed to varying
concentrations of each anion, the relative potency of ClO4-to
inhibit125I-uptake at the NIS was found to be 15, 30, and 240
times that of SCN-, I-, and NO3-, respectively, on a molar
concentration basis (1). Recast on a mass basis, 1 µg of
perchlorate was as potent as 9 µg of thiocyanate, 38 µg of iodide,
or 150 µg of nitrate in vitro uptake by human NIS transfected
into animal cells.
At particular risk to perchlorate exposure are fetuses and
infants because neural, brain, and skeletal development depend
on sufficient levels of T3. Pregnant and nursing women transfer
perchlorate by placental and mammary NIS systems, respec-
tively. In a small study of 36 lactating women from 18 states
in the United States, perchlorate levels in breast milk up to 92
µg/L and averaging 10.5 µg/L have been reported (2). The U.S.
National Academies of Science, in reviewing studies on human
health effects from perchlorate ingestion, found that a reference
dose (RfD) of 0.7 µg/kg per day was protective of public health
for even the most sensitive groups of people over a lifetime of
exposure (3). In February 2005, the U.S. Environmental
Protection Agency (EPA) established this finding as the official
RfD for perchlorate (4).
Ammonium perchlorate is widely used as an oxidant/
propellant in solid rocket fuels and in explosives such as
fireworks and munitions. Production and military facilities have
contaminated ground and surface waters through leaching,
seepage, and improper disposal of material past its shelf life.
Known perchlorate releases have been identified in 35 states
(5, 6). As an environmentally stable, strong electrolyte, per-
chlorate is highly mobile in a water system from a point source
of entry. Perchlorate in Lake Mead and the lower Colorado
River, sources of drinking water for ∼15 million people in the
southwestern United States, came from a production plant in
Henderson, Nevada, that had been making ammonium perchlo-
rate for four decades (7). Since 1999, when controls began at
the plant, >3 million pounds of perchlorate have been captured,
and annual average concentrations at sampling points have
declined substantially (8). In 2004, nine monthly samples taken
at a Colorado River aqueduct intake for the Metropolitan Water
District (serving Los Angeles and Southern California) were
below the 4 µg/kg limit of detection, a 50% decline since 2000.
A portion of this water irrigates lettuce, cantaloupe, alfalfa, and
other crops grown in the Imperial Valley of California and
Arizona near Yuma. Studies (9, 10) have shown that growing
* To whom correspondence should be addressed. Tel: 301-436-1999.

Page 2

plants can systemically accumulate perchlorate from water.
Bioconcentration factors [(µg/kg fresh weight)/(µg/L)] may
reach many 100-fold, with perchlorate preferentially translocated
to leaves. For example, higher concentrations were found in
the outer leaves of lettuce, which are more extensively removed
from iceberg than from romaine and leaf varieties before human
consumption (11). The U.S. winter lettuce crop comes pre-
dominantly from this growing region; sampling and analysis
of the edible portion of lettuce by the U.S. Food and Drug
Administration (FDA) in 2004 (12) found perchlorate up to 129
µg/kg, although the mean (N ) 138) was more than an order
of magnitude lower at 10.5 µg/kg. Forage plants such as alfalfa
may be a source of perchlorate found in dairy milk (12-14).
Perchlorate occurs naturally in salt deposits in the Atacama
Desert of Chile where nitrate is mined for fertilizer. Although
U.S. fertilizers are made from synthetic materials and presently
do not contain perchlorate (15), some fertilizers acceptable for
organic farming may contain potash and saltpeter from natural
deposits with significant quantities of perchlorate (16). Oxygen
isotopic abundance, which showed a large and positive17O
anomaly in natural as compared to synthetic perchlorate (17),
may help identify environmental contamination from suspected
atmospheric sources, such as perchlorate in the groundwater of
the southern high plains (includes the Texas Panhandle) where
clear evidence of anthropogenic sources was missing (18).
Early analytical methods development (19) focused on detect-
ing perchlorate in water to provide data for the EPA’s un-
regulated contaminants monitoring rule program (ClO4-was
placed on the candidate contaminant list). Sample cleanup was
usually minimal despite the wide use of ion chromatography
(IC) with conductivity detection (CD), which cannot confirm
analyte identity. In contrast, chemically diverse matrices such
as food have taken days to clean up an extract for chromato-
graphic measurement (10, 11, 13, 20). Generally, freeze drying
the sample, a time-consuming step, was employed to concentrate
the perchlorate residue before extraction and cleanup. Recently,
speed, specificity, and sensitivity were achieved with IC inter-
faced by electrospray ionization (ESI) to a tandem mass spec-
trometer (MS/MS) and use of18O4-labeled perchlorate internal
standard (IS) for analysis of food (14) or urine (21) without
extract cleanup. Without the labeled IS, standard additions were
necessary to mitigate groundwater matrix effects on the per-
chlorate signal. Flow injection (FI)-ESI-MS/MS (22) enabled
detection to 0.5 µg/kg. Similarly, high-field asymmetric wave-
form ion mobility spectrometry (FAIMS), a technique that separ-
ates gas-phase ions by ion mobility in an applied electric field,
was performed without liquid chromatography and with a single
stage quadrupole. FI-ESI-FAIMS-MS analysis of waters (23)
achieved µg/kg detection limits from 0.5 (river water) to 0.050
(tap water); application to food extracts has not been reported.
Methodology employing costly MS instrumentation may
restrict the number of government laboratories gathering level
and incidence data on perchlorate in the food supply. Concur-
rently with our MS methodology, we developed an alternative
IC method for the more widely available and simpler conductiv-
ity detector. Concentration and enrichment of perchlorate were
accomplished quickly by passage of cleaned-up food extract
through a short anion-exchange high-performance liquid chro-
matography (HPLC) precolumn, from which column switching
injected the retained contents onto an analytical anion-exchange
column by coupling the two columns. Use of a graphitized
carbon solid-phase extraction (SPE) column efficiently removed
potentially interfering conductive coextractives and saved
considerable time over published alumina cleanup methods.
Perchlorate passed through the column and was collected in a
fraction held for instrumental analysis.
Over many years, the FDA has applied column-switching ion-
exchange for on-line extraction, enrichment, and partial cleanup
in the precolumn to determine cation-forming analytes such as
pesticide residues in fruit (24, 25) and the naturally occurring
ephedrine alkaloids and synephrine stimulants in dietary supple-
ments (26). Unlike UV and fluorescence detectors used in
previous work, the suppressed conductivity detector, when
optimized to detect a nanogram of perchlorate, severely limited
the mobile phase concentration of ionic (hydroxide) eluent that
could be neutralized by the self-regenerating suppressor operated
at sufficiently low applied current in order to reduce suppressor
noise. A reasonable retention time under these conditions could
be achieved with columns of considerably lower ion-exchange
capacity than in previous work, a disadvantage that would reduce
the volume of extract passed through the precolumn before anion
breakthrough from heterogeneous anion-exchange equilibrium.
Achieving a method determination limit for perchlorate in food
that rivaled 1 µg/kg obtained with our IC-MS/MS method was
important for data consistency in FDA exposure assessment and
risk management programs.
A preconcentration/preelution IC-suppressed CD method for
perchlorate in high salinity water (27)slater adapted to milk
(13) and complex samples (28)swas similar in instrumental
approach. However, we used a less concentrated hydroxide (40
vs 100 mM) mobile phase modified with 5% ACN to shorten
chromatographic retention, a self-regenerating suppressor oper-
ated at lower current (100 vs 300 mA) for reduced amplitude
sinusoidal noise, a lower capacity ion-exchange chromatography
column set, and a more efficient extract cleanup provided by
graphitized carbon SPE column. Briefly, the edible portion of
lettuce, cantaloupe, spinach, or carrot samples was analyzed on
a wet weight basis. After sample compositing and chopping, a
test portion was blended with dilute acid and filtered, a small
volume of measured filtrate was acidified to pH ∼2, several
milliliters was passed through a carbon SPE column, and
perchlorate in an aliquot from a collected fraction was measured.
Analysis was complete in hours as compared to days with
existing methods based on alumina cleanup. For statistical
comparison, an additional analysis was done by IC-MS/MS (14)
on untreated filtrate diluted with an isotopically labeled per-
chlorate IS. Also, cooked vegetables processed as baby foods
were analyzed for perchlorate.
MATERIALS AND METHODS
Chemicals, Reagents, and Solutions. Nitric acid (70%) and 50%
sodium hydroxide solution were analyzed reagents (J. T. Baker,
Phillipsburg, NJ). Water resistivity was 18 MΩ cm (Milli-Q cartridge
system, Millipore, Billerica, MA). A 1.0 M nitric acid stock solution
(90.3 g of reagent per 1 L aqueous solution) was used to prepare a 10
mM nitric acid solution for use as the extraction solvent and diluent
for standard solutions. A 100 mM NaOH solution (8.00 g of reagent
diluted to make 1 L of aqueous solution), a component of the LC mobile
phase, was prepared fresh weekly. Diluted acid and base solutions were
stored in wide-mouth polypropylene bottles (Nalgene). A potassium
perchlorate stock standard aqueous solution obtained from Alltech
(Deerfield, IL) was certified 997 ( 6 µg/mL in ClO4-; for clarity,
reported standard concentrations were based on the nominal 1000 µg/
mL. A 1.00-mL aliquot was diluted to make a 10.0 µg/mL working
standard solution, from which perchlorate calibration standards were
prepared by serial dilution to make 1.00, 0.100, and 0.0100 µg/mL.
Acetonitrile (ACN) and methanol were UV and HPLC grade (Honey-
well Burdick & Jackson, Muskegon, MI). Reagent grade Celite 545-
AW filtering aid, DOWEX 50WX8-200 cation-exchange resin (H+
form) and Supelclean ENVI-Carb graphitized carbon SPE columns (500

Page 3

mg per 6 mL) were purchased from Supelco (Bellefonte, PA). The
resin was cleaned and conditioned batchwise (50 g) by magnetic stirring
with 1.0 M nitric acid for 30 min, followed by ∼seven batch rinses
with water while stirring for ∼15 min each. The final rinse was pH
∼5 (indicator strips pH 0-14, EM Science, Gibbstown, NJ). The resin
was transferred to 50-mL glass-stoppered flasks, and the water layer
was withdrawn by disposable pipet and evaporation until the moist
resin would clump onto a small spatula for removal. Carbon SPE
columns were conditioned as needed with one column volume each of
methanol and extraction solvent.
Instrumentation. The column-switching ion chromatograph shown
in Figure 1 was assembled from a gradient pump and vacuum
membrane degasser (respectively, models P4000 and SCM 1000,
ThermoFinnigan, San Jose, CA), an electrically actuated injection valve
(model E4C6W, VICI, Houston, TX) mounted inside a convective
column oven (old DuPont Instruments or equivalent), an anion self-
regenerating suppressor (ASRS, ULTRA II, Dionex, Sunnyvale, CA)
operated in the external water mode, and a Dionex thermostated
conductivity cell (model DS3 Detection Stabilizer) with electrochemical
detector (model ED40). A coil of PEEK tubing (0.010 in. i.d. × 2-2.5
ft) provided backpressure (30-40 psig) to suppress air bubble formation
in the cell. Filters (A-316, Upchurch Scientific, Oak Harbor, WA) with
replaceable 0.5-µm frits (A-103x) were installed before the valve’s inlet
filling port and before the analytical column.
A Dionex IonPac AG16 anion-exchange guard column (4 mm ×
50 mm) functioned as a precolumn to extract and enrich perchlorate
and other anions from test solutions for injection (reverse flush) onto
an AS16 anion-exchange analytical column (4 mm × 250 mm). The
columns, chosen for their greater ion-exchange capacity than the
comparable 2 mm i.d. set, were held at 40 °C. During injection, the
coupled columns were eluted at 1.0 mL/min with a mobile phase
dynamically generated from (A) 100 mM NaOH, (B) water, and (C)
ACN at 40:55:05 (A-B-C). During loading of the precolumn, 0.5
mL of water was pushed through to flush out the mobile phase and
then 50-750 µL of test solution was delivered, followed by another
0.5 mL of water. The ASRS was operated at 100 mA, and a 4-5 mL/
min water regenerant flow rate was measured with the current on.
Conductance of the suppressed background signal inside a 40 °C cell
was 2.2-3.2 µS with the suppressor noise averaging 0.5-0.8 nS peak-
to-valley. The detector signal was acquired and processed by a
chromatography data station (model SS420X analog-to-digital converter,
Scientific Software, Inc., Pleasanton, CA; Chromquest ver. 4.0 software,
ThermoFinnigan).
Analysis. Most samples of lettuce (iceberg, green leaf, red leaf, and
romaine varieties), cantaloupe, spinach, and carrots were collected by
FDA inspectors from growers or packing sheds as part of the 2004
“Collection and Analysis of Food for Perchlorate”, a 500 sample field
assignment. Samples (5-20 lb each) were shipped overnight to FDA’s
Southeast Regional Laboratory, for compositing in a food chopper and
analysis by IC-MS/MS methodology (14, 29). At our request, frozen
(-20 °C) portions of certain sample composites were sent to us for
IC-CD method development research and quality assurance check
analysis. Additional samples (1-2 lb each), especially winter lettuce
from the 2004 crop, were purchased locally at suburban Washington,
DC, supermarkets, chopped in a food processor, and stored at -20 °C.
Samples were thawed before a test portion was removed for analysis.
Glassware was cleaned in perchlorate-free detergent (MICRO-90,
International Products Corp., Burlington, NJ) and rinsed with extraction
solvent before use. A 100-g test portion (weighed to 0.01 g) of thawed,
thoroughly mixed sample and ∼10 g of Celite (omitted for iceberg
lettuce samples) were blended with 150 mL of 10 mM HNO3 in a
Waring glass jar (rubber gasket replaced with PTFE) for 6 min at high
speed. The contents plus two 20-mL rinses of the jar were filtered under
vacuum through a rinsed glass fiber filter (#30, 11 cm, Schleicher &
Schuell MicroScience, Inc., Riviera Beach, FL). Volume recovery was
usually complete in 30 min, but filtration of some extracts exceeded
90 min, despite use of the filtering aid. The filtrate volume was
measured. A portion (25 mL) was poured into a beaker and stirred
magnetically while cation-exchange resin was added until pH ∼2 was
reached (combination glass electrode, model 91-55, Thermo Orion,
Beverly, MA), after which stirring continued for 15 min. After settling,
supernatant was poured onto a conditioned ENVI-Carb SPE column
placed on a vacuum manifold (Visiprep, Supleco), two 2-mL fractions
were collected, and an aliquot of the second fraction was analyzed by
IC-CD. The maximum aliquot size depended upon the crop: 400 µL
for spinach; 500 µL for red leaf lettuce; and 750 µL for cantaloupe,
carrots, and lettuce of iceberg, green leaf, and romaine varieties. Raw
and treated extracts kept well for days stored in a refrigerator.
The IC columns, stored in water, were cleaned of strongly retained
crop material before daily use by pumping 100 mM NaOH (base) at
1.5 mL/min for 30 min, followed by a 5-min gradient to 35:65 base/
water (hold 15 min) and 1.0 mL/min flow rate. The ASRS and
conductivity cell were disconnected from the system. Note that when
cooked foods had been analyzed, columns were cleaned with 10:90
base/ACN mobile phase for 30-60 min at 1.0 mL/min. Daily calibration
began after equilibrating the columns to the 1.0 mL/min 40:55:05
mobile phase proportioned from base, water, and ACN, respectively.
The perchlorate retention time was stable during the day. Note that
mobile phases were never premixed and 100 mM NaOH was prepared
fresh at least weekly.
A six-point linear regression calibration of peak area was performed
at 50.0, 30.0, 10.0, 5.00, 2.50, and 1.00 ng of perchlorate in this order
using 50-500-µL aliquots of calibration standard solutions. The low-
frequency sinusoidal suppressor noise pattern modulating the detector
signal often caused automatic integration to begin and end the
perchlorate peak at a valley point of the noise cycle, a positive
integration bias that included the full peak-to-valley noise in the
baseline. If needed, peak start and end points were reset manually at
the noise cycle midpoint. The mass of perchlorate Q (ng) in the analyzed
test solution TS (mL) was quantitated from peak area A and the linear
regression coefficients slope m and intercept b according to Q ) (A -
b)/m. Determination (µg/kg) of perchlorate in the sample was calculated
from (Q/TS)(V/W), where V was the final extract volume (mL) and W
was the test portion weight (g).
Comparative Analysis by IC-MS/MS. Slight changes were made
to the method (14) to compensate for perchlorate losses experienced
with some disposable syringe filters. Filtrate with no cleanup was diluted
9:10 with 1.00 mL of 0.100 µg/mL of aqueous Cl18O4-IS in a 15-mL
graduated (0.1 mL) glass-stoppered centrifuge tube. Diluted solution
was drawn up into a syringe and pushed through a 0.2-µm nylon filter
(30 mm, Titan2, Sun-SRi, Duluth, GA), diverting ∼six drops to waste
before collecting 0.5-1.0 mL in an autosampler vial.
RESULTS AND DISCUSSION
Chromatography. Suppressed conductivity chromatograms
in Figure 2 compare extracts and the 1.0 ng lowest calibration
standard, which at S/N ∼ 10 was the measurement limit of
quantitation (LOQ). Perchlorate eluted ∼11 min, separated from
similar-sized peaks and far away from intense peaks, in a
chromatographic region supportive of accurate integration.
Figure 1. Column-switching ion chromatograph assembled from HPLC
syringe and adapter (A), in-line filters (B), injection valve (C) shown in
the inject position, precolumn (D), analytical column (E), convective oven
(F), solvent degasser (G), multisolvent high-pressure pump (H), anion
self-regenerating suppressor (I) in external water mode, conductivity cell
(J), and cell backpressure tubing (K). During loading, the test solution
followed the dashed-line flow path through the valve and entered the
precolumn at the outlet end for the mobile phase to avoid chromatography
through the precolumn at injection.

Page 4

Quantitation of the suspected perchlorate peak in these food
extracts exceeded the measurement LOQ and was equivalent
to a sample level in the 7-14 µg/kg range. Findings were
confirmed by IC-MS/MS with the following comparative results
(µg/kg CD vs MS/MS): (a) cantaloupe with seeds, 9.64 and
9.87; (b) seedless cantaloupe, 14.0 and 13.9; (c) carrots, 7.39
and 9.22; (d) iceberg lettuce, 10.5 and 8.69; and (e) spinach,
9.59 and 12.6. Determinative results were within (25% relative
to mass spectrometry, with no evidence of systematic high bias
due to conductive coeluting crop material. Chromatography of
cantaloupe extract was very long, mainly due to coextractives
from seeds included in the composited sample; analysis of just
the fleshy fruit produced a cleaner, simpler chromatogram (cf.
traces A and B).
Graphitized carbon removed organic material while allowing
perchlorate and ions to pass through the column for collection
and analysis. Organoanions, conceivably conjugate bases of
organic acids, potentially interfered with conductivity measure-
ment. Therefore, to suppress weak acid ionization and thereby
enhance likelihood of removal by carbon adsorption of the
neutral form, the pH was lowered to ∼2 before cleanup. Extract
acidification by release of resin-bound hydrogen ions exchanged
for metal ions introduced no additional anion to compete with
perchlorate for the fixed number of anion-exchange sites on
the precolumn stationary phase. Alternatively, the addition of
an acid solution, assuming minimal dilution, would increase the
anion concentration and thus reduce the aliquot size for
precolumn extraction and enrichment under nonequilibrium
conditions; moreover, a smaller quantity of perchlorate injected
onto the analytical column would raise the LOQ. Once the
precolumn reached heterogeneous ion-exchange equilibrium,
proportionality between analyte signal and volume passed was
lost, quantitation became independent of aliquot size, and
determination failed.
The average method LOQ for perchlorate determination in a
crop was calculated from the 0.0010 µg measurement LOQ
assumed to be in the recommended extract volume passed
through the precolumn, multiplied by the average final volume
for that crop (values ranged from 247 to 270 mL), and divided
by the test portion weight (0.100 kg). For example, the method
LOQ for cantaloupe averaged 3.5 µg/kg [i.e., (0.0010 µg/0.750
mL) × (266 mL/0.100 kg)]. Method LOQs (µg/kg) for the other
crops were 3.3 for carrots; 3.6 for iceberg, green leaf, and
romaine lettuce but 5.0 for red leaf lettuce; and 6.4 for spinach.
Calibration and Recovery. Calibration of peak area response
(y) from suppressed CD of 1-50 ng of perchlorate (x) was
described by the following linear regression equation (mean (
SD) for N ) 51: y ) (1.55 ( 0.088) × 105x - (0.32 ( 0.20)
× 105. Over this long period, the RSD of the slope was (6%,
and the 95% confidence limits (mean ( tdf)50‚SD) about the
intercept included zero. Further evidence supporting consistently
linear calibration came from goodness-of-fit R2averaging 0.9997
( 0.0003. Results of duplicate recovery trials at the 10 µg/kg
fortification level are presented in Table 1 after correction for
incurred level (only control lettuce was blank). The corrected
recovery averaged 93-96% for cantaloupe, carrots, and lettuce
and 81% for spinach. The spinach recovery values may have
been influenced by a high control level correction obtained by
single analysis, which insufficient sample prevented checking.
The grand mean (corrected) recovery of 91.7% and RSD of
10.6% (N ) 8) were acceptable at 10 µg/kg.
Comparative Analysis. Determinative accuracy was tested
statistically against the result by IC-MS/MS analysis of raw
filtrate (no cleanup) with18O4-labeled perchlorate added as an
IS. Linear regression plots of determination by IC-CD against
the gold standard IC-MS/MS are shown in Figure 3 for 17
cantaloupe samples, 10 carrot samples, 27 lettuce samples (four
Figure 2. Suppressed conductivity IC of extracts prepared from the edible
portion of (A) cantaloupe with seeds, (B) cantaloupe without seeds, (C)
carrots, (D) iceberg lettuce, and (E) spinach. Samples had 7−14 µg/kg
incurred perchlorate. Chromatographic analysis was done on 750 µL
except spinach, which was 400 µL. For comparison, a chromatogram of
(F) 1.0 ng of perchlorate standard is included.
Table 1. Recovery (%) of Perchlorate Added at 10 µg/kg to Crop
Controlsa
crop trial 1 trial 2average
cantaloupe
carrots
lettuce
spinach
106
95.6
97.6
85.0
80.5
95.9
95.1
77.5
93.3
95.8
96.4
81.3
a% recovery ) [(F − U)/A] × 100, where F and U are determinations (µg/kg)
of perchlorate in the fortified and unfortified test portions, respectively, and A is
the concentration added. U ) 3.61 for cantaloupe; 3.17 and 3.41 for carrot trials
1 and 2, respectively; 0 (not detected) for lettuce; and 7.64 for spinach.
Figure 3. Linear regression statistics for perchlorate determination by
conductometric vs mass spectrometric measurements for (a) cantaloupe,
(b) carrots, (c) lettuce (four varieties), and (d) spinach sample sets.

Page 5

varieties included), and 22 spinach samples. Goodness-of-fit R2
was >0.980 in all crops, indicating that results by IC-CD were
a strong predictor of the results by IC-MS/MS. Bias, however,
was indicated by regression coefficients that differed from unity
slope and zero intercept. Proportional ((4%) and fixed ((1.00
µg/kg) bias were small for cantaloupe, lettuce, and spinach.
However, a large negative proportional bias (17%) was observed
for carrots, although the fixed bias was similar to the other crops.
One data point (88.4, 65.2) was responsible for 7% of this bias
(i.e., m ) 0.90 if these data were omitted), but no justification
for dropping the data was evident. Doubling the number of carrot
samples to bring the set size in line with the other crops may
provide better regression statistics later. Our data indicate that
about one-third of the spinach samples exceeded 100 µg/kg as
compared to only one sample of cantaloupe, carrot, or lettuce.
The highest perchlorate finding in this study was ∼800 µg/kg
in spinach.
A two-tailed paired Student’s t-test was performed to test
the null hypothesis that there was no difference between the
means of the IC-CD and IC-MS/MS determinations at the 95%
confidence level (P value > 0.05). Carrots showed no statisti-
cally significant difference between the means of CD and MS/
MS determinative values (P ) 0.12), despite large deviation
from the ideal regression slope. Likewise, the null hypothesis
was true for cantaloupe (P ) 0.51), spinach (P ) 0.55), and
lettuce (P ) 0.93). Therefore, when a sample set of these crops
is analyzed by the IC-CD method, the sample set average
determination 95% of the time will be no different from that
average obtained had the IC-MS/MS method been used,
whenever perchlorate exceeded the IC-CD method LOQ.
Presented in Table 2 are comparative findings of low-level
(<11 µg/kg) perchlorate in crops culled from the results
displayed in Figure 3. The mean difference (0.06 µg/kg)
between determinative results was not significant at the 95%
confidence level (P value ) 0.89). Statistically, determination
by IC-CD was equivalent on average to determination by IC-
MS/MS. Nevertheless, on an individual analysis basis, compara-
tive results differed up to (5 µg/kg, and triple quadrupole mass
spectrometry with labeled perchlorate IS provided the best
specificity and accuracy. If the average of perchlorate findings
in a sample dataset is sought, then the IC-CD method may be
fit-for-purpose. Although the conductivity detector was not
specific for perchlorate, the carbon column cleanup and column-
switching IC appeared to impart high selectivity to perchlorate
detection by removing endogeneous conductive interferences.
Ruggedness. Extending the method to other crops must be
done cautiously. The conductivity detector provided no molec-
ular fingerprint for identification and always will be susceptible
to false positives or interference from conductive crop material
eluting with perchlorate. This method was also susceptible to
negative bias from samples that provided large quantities of
coextractive anions that competed with perchlorate for the 3.5
µequiv of AG16 exchange sites. For these crops, the precolumn
and extract ion-exchange (IX) equilibrated sooner. This condi-
tion is explained by a simple case of univalent-univalent anion
exchange between P-(predominant level) and T-(trace level)
in solution and a solid phase ion exchanger (IXer) initially in
the P-ion form. At heterogeneous IX equilibrium, the IXer
ceased uptake of T from solution, its concentration in the solid
phase fixed according to [T h] ) KIX[P h][T]/[P], where the
overhead bar refers to the IXer phase and KIX is the IX
equilibrium constant. Because T was a trace, then [T h] + [P h] ≈
[P h] ) C h, the IX capacity, and [T h] ) K[T]/[P], where constant
K ) KIXC h. Therefore, for a NO3--ClO4-(P-T) system, the
perchlorate enrichment factor [ClO4
j]/[ClO4-] at IX equilibri-
um was inversely proportional to [NO3-], a constant. Before
equilibrium, perchlorate uptake by the exchanger proceeded
linearly with depletion in solution passed through the precolumn
because of fast ion-exchange kinetics and lack of evidence of
a competing separation mechanism (e.g., perchlorate does not
form complexes).
Ion-exchange dynamics in the precolumn are plotted in
Figure 4 at various loadings of a perchlorate standard solution
(∼100 ng/mL) prepared in 10, 15, 25, or 50 mM nitric acid.
As predicted, perchlorate enrichment increased linearly regard-
less of nitrate level until IX equilibrium, which was reached at
-
Table 2. Perchlorate Findings between 3 and 11 µg/kg by IC-CD As
Compared to Independent Analysis by IC-MS/MS
cropIC-CD (y)IC-MS/MS (x) difference (y − x)
−1.41
−1.81
1.52
0.67
0.42
1.37
−2.20
−0.88
−2.66
−0.27
−0.79
−4.54
−0.93
−1.83
−0.68
2.52
1.25
4.19
0.93
1.74
−3.01
−0.23
5.13
−0.30
1.81
1.58
26
0.060
2.185
0.143
0.888
cantaloupe
carrots
carrots
lettuce
carrots
cantaloupe
spinach
carrots
spinach
lettuce
lettuce
cantaloupe
lettuce
carrots
spinach
cantaloupe
lettuce
cantaloupe
lettuce
lettuce
spinach
cantaloupe
lettuce
lettuce
lettuce
lettuce
2.78
3.17
3.41
4.00
4.05
4.58
5.13
5.71
5.77
6.04
6.27
7.26
7.38
7.39
7.64
7.80
8.18
8.49
8.54
9.25
9.59
9.64
10.1
10.2
10.5
10.6
4.19
4.98
1.89
3.33
3.63
3.21
7.33
6.59
8.43
6.31
7.06
11.8
8.31
9.22
8.32
5.28
6.93
4.30
7.61
7.51
12.6
9.87
4.97
10.5
8.69
9.02
N
mean
SD
t
P value
Figure 4. Perchlorate uptake by the precolumn from test solutions of
various anion concentrations. Main graph: plots for 100 ng/mL perchlorate
standard solutions prepared in nitric acid concentrations of (O) 10, (4)
15, (0) 25, or (]) 50 mM. (Inset) Plots for extracts prepared from (b)
green leaf lettuces, (9) iceberg lettuce (offset 2 µg/kg), and (2) spinach
(offset 15 µg/kg).