Refinament of 2005.06

190 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
FOOD CHEMICAL CONTAMINANTS
Refinement and Extension of AOAC Method 2005.06 to Include
Additional Toxins in Mussels: Single-Laboratory Validation
ANDREW D. TURNER, DEIRDRE M. NORTON, ROBERT G. HATFIELD, STEVEN MORRIS, ALLAN R. REESE, MYRIAM ALGOET, and
DAVID N. LEES
Center for Environment, Fisheries and Aquaculture Science, Barrack Rd, The Nothe, Weymouth, Dorset, DT4 8UB,
United Kingdom
A single-laboratory validation study was
undertaken for the analysis of paralytic shellfish
poisoning (PSP) toxins in common mussels,
extending AOAC Official Method 2005.06 to include
the additional toxins dcNEO and dcGTX2,3. The
method was refined to improve toxin oxidation
product stability, analytical sensitivity of
N-hydroxylated toxins, and throughput. Validation
was performed to characterize the method for
selectivity, sensitivity, linearity, precision,
repeatability, recovery, ruggedness, and
uncertainty. Parallel testing of naturally
contaminated mussels enabled comparison of
sample toxicities obtained using mouse bioassay
(MBA) and high-performance liquid
chromatographic (HPLC) methodologies.
Performance characteristics of the method are
reported for all commercially available certified
reference toxins. Results from the MBA and HPLC
methods were well correlated, and the analytical
method has been instigated as the sole monitoring
tool for UK official control surveillance of PSP
toxins in common mussels.
P
aralytic shellfish poisoning (PSP) toxins are a major
group of shellfish toxins known to induce human
illness. Within the European Union (EU) and
elsewhere, monitoring of PSP toxins in shellfish is a statutory
requirement to protect shellfish consumers (1). The EU’s
reference method for detecting PSP toxins is the mouse
bioassay (MBA; 2, 3). An alternative method (AOAC Method
2005.06 or “Lawrence” method), that was developed during
the 1990s, applies high-performance liquid chromatography
with fluorescence detection (HPLC-FLD), and has gone
through single and interlaboratory validation (4–9). In 2005,
this method was adopted by AOAC as an Official First Action
method (10) and has recently been approved by the EU for
purposes of PSP monitoring (2). However, as an alternative to
the MBA, the HPLC-FLD method is still under investigation
Received July 30, 2008. Accepted by AP September 30, 2008.
Corresponding author’s e-mail: andrew.turner@cefas.co.uk
by a number of monitoring laboratories across Europe, and to
our knowledge no laboratories currently use this approach as
the sole statutory technique for monitoring production areas.
Previous studies have highlighted issues with the method,
such as throughput limitations, solid-phase extraction (SPE)
ion-exchange cleanup, co-eluting oxidation products, and the
effect of toxin profiles in quantitation (11).
The objectives of this study were to establish the full
performance characteristics of AOAC Method 2005.06 when
applied to common mussel (Mytilus edulis) and to refine and
validate it, via a single-laboratory validation (SLV) scheme,
for application in the high throughput Official Control
monitoring of PSP toxins in mussels sampled from England,
Wales, and Scotland. The validation was also extended to
include toxins only recently commercially available (dcNEO
and dcGTX2,3). The European Commission Regulation
882/2004 (12) requires that Official Control methods should
be validated and quality assured prior to adoption into EU
monitoring
programs.
The
method
performance
characteristics investigated included accuracy, applicability,
limit of detection (LOD), limit of determination, precision,
repeatability, reproducibility, recovery, selectivity, linearity,
measurement of uncertainty, and ruggedness. To achieve these
objectives, guidelines proposed by the International Union of
Pure and Applied Chemistry (13) were followed and the
HPLC-FLD method was assessed by comparing analytical
data with that derived from the MBA approach.
Experimental
Reagents and Chemicals
Certified reference toxins (GTX1,4, NEO, dcSTX,
GTX2,3, GTX5, C1,2, and STX, together with the additional
toxins dcNEO and dcGTX2,3) were obtained from the
Institute of Marine Biosciences, National Research Council
Canada (IMB, NRCC, Halifax, Nova Scotia, Canada).
Solvents were of HPLC grade and all chemicals were
analytical reagent grade. Primary toxin standards were diluted
in ~4.5 mL water to form concentrated stock standard
solutions
and
were
stored
following
NRCC
recommendations. They were subsequently diluted in
appropriate volumes of 0.1 mM acetic acid to produce
working analytical standards for instrument calibration
purposes. For purposes of extract oxidation and subsequent
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 191
Table 1. Toxin mixtures used in this study
Toxin mixture
Toxins
I
GTX1,4, NEO, dcNEO
II
STX, dcSTX, GTX2,3, GTX5
III
C1,2, dcGTX2,3
IV
dcGTX2,3 and dcSTX
V
All toxins
quantitation, toxin standard mixes (I–V) were prepared as
recommended (10) and are illustrated in Table 1. Periodate
oxidation of Mix IV was applied when naturally contaminated
samples were quantified to enable estimation of both the
dcSTX contribution to NEO/dcNEO and the dcGTX2,3
contribution to GTX1,4 as described previously (10).
Preparation and Extraction of Shellfish Samples
Approximately 0.5 kg of both common mussels (M. edulis)
and Pacific oysters (Crassostrea gigas, for matrix modifier)
were homogenized and stored at –20°C until required. To
ensure that these materials were free of PSP toxins prior to
their use in the validation schemes, 5.0 g (±0.1 g) samples
were transferred to 50 mL polypropylene centrifuge tubes,
extracted, and analyzed according to AOAC Method 2005.06
and chromatograms were compared against those derived
from the analysis of oxidized PSP standard solutions.
Archived mussel samples were acquired from shellfish
production areas sampled under the English, Welsh, and
Scottish Official Control monitoring programs and had been
stored at –20°C since arrival and homogenization. Matrix
modifier was prepared following a protocol modified from the
AOAC method.
Liquid Chromatography and PSP Toxin Quantitation
LC separation was performed using a Gemini C18
reversed-phase column (150 ´ 4.6 mm, 5 mm; Phenomenex,
Manchester, UK) with a Gemini C18 guard column (set
at 35°C). An Agilent (Stockport, UK) fluorescence detector
(1200 model FLD) was used to detect the oxidation products
of all PSP toxins. Fluorescence excitation was set to 340 nm
and emission to 395 nm. Mobile phase (A): 0.1 M ammonium
formate, adjusted to pH 6 ± 0.1 with 0.1 M acetic acid; (B):
0.1 M ammonium formate with 5% acetonitrile, also adjusted
to pH 6 ± 0.1 with 0.1 M acetic acid. The mobile phase was
delivered by an Agilent 1200 series LC at a flow rate of
2 mL/min. The LC gradient was as follows: 0–5% mobile
phase B in the first 5 min, 5–70% B for the next 4 min, hold at
70% B for 1 min, and back to 100% A over the next 2 min. The
100% A was held for an additional 2 min to allow for column
equilibration before subsequent sample injections.
Individual toxins and multiple toxin mixes were oxidized
using both periodate and peroxide oxidation (4–10) and
30–50 mL volumes were injected. Chromatographic data were
reviewed to ascertain retention times and relative peak area
responses of the toxins oxidized under both oxidation
conditions. The toxicity equivalence factors (TEFs) described
by Oshima (14) were used. These factors were incorporated
into the calculations for preparation of calibration solutions
for each toxin mix, so that the calibration range for each toxin
equated to 0–0.96 mg STX eq/g. The exception was GTX5,
where calibration solutions were prepared at 10% of the
concentration of other toxins (0–0.096 mg STX eq/g). In the
case of isomeric pairs (GTX1,4, GTX2,3, C1,2, and
dcGTX2,3), the highest TEF was used for each pair.
Individual toxin concentrations were reported as mg STX
dihydrochloride eq/g, and the total PSP toxicity was
calculated by summing the individual concentration
contributions from all quantified toxins and is quoted in terms
of mg STX di-HCl eq/100 g.
Extraction, Cleanup, and Oxidation of Samples
The extraction and extract oxidation procedures of
Lawrence (9) and AOAC Method 2005.06 (10) were followed
as closely as possible. Matrix modifier was prepared from
Pacific oyster homogenate shown not to contain PSP toxins,
using the same extraction method, without the 2–3-day
precipitation step. SPE ion-exchange cleanup was used for all
samples potentially containing any of the N-hydroxylated PSP
toxins GTX1,4, NEO, and dcNEO. Fraction F1 contains the
N-sulfocarbamoyl C-toxins (C1,2), F2 contains the
gonyautoxins (GTX) group of toxins (GTX1-5, and
dcGTX2,3) and the carbamates (STX, dcSTX, and NEO)
elute in F3. The AOAC Method 2005.06 can be divided into
2 parts. Initially, an HPLC-FLD qualitative screen, involving
the periodate oxidation of the pH-adjusted, cleaned-up
extracts, determines both the presence of PSP toxins in the
samples and the potential presence of N-hydroxylated toxins.
A sample was positive by the screening method when a peak
with a signal-to-noise (S/N) ratio of ³3 was present at the
same retention time (±2.5%) as in the analytical toxin
standard. For toxins with more than one oxidation product
(GTX1,4, GTX2,3, dcSTX, dcGTX2,3, and NEO), a positive
sample also required the presence of at least the most
significant secondary peak. For all screen-positive samples,
full quantitation was subsequently performed, with all
non-N-hydroxylated toxins quantified following peroxide
oxidation of the C18 cleaned-up extracts. GTX1,4 was
quantified following periodate oxidation of F2, and both NEO
and dcNEO were quantified following periodate oxidation
of F3.
Optimization of SPE Ion-Exchange Fractionation
AOAC Method 2005.06 describes the use of silica-bound,
ion-exchange SPE cartridges for the fractionation of sample
extracts. Because of concerns with method sensitivity due to
the extra dilution associated with this step, and the noted
impracticality of using additional evaporation steps suggested
in the AOAC method in a high-throughput monitoring
program, ion-exchange options were investigated to
determine whether other cartridge sorbents would enable the
192 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
Table 2. Experimental parameters chosen for 2 ruggedness experiments
Test 1 (GTX2,3 and STX)
Test 2 (GTX1,4 and NEO)
Extraction temperature (A = 100°C, a = 95°C)
C18 extract pH (A = pH 6, a = pH 7)
Vortex time for extraction (B = 30 s, b = 90 s)
Fractionation flow rate (B = 2 mL/min, b = 3 mL/min)
C18 cleanup flow rate (C = 2 mL/min, c = 3 mL/min)
pH of cleaned-up extract (D = pH 6, d = pH 7)
Ambient temperature during oxidation (E = 22°C, e = 25°C)
Oxidation time (F = 1 min 55 s, f = 2 min 5 s)
Chromatographic flow rate (G = 1.9 mL/min, g = 2.1 mL/min)
use of lower volumes of fraction elution solvents. A suitable
candidate was identified, and the elution conditions used for
fractionation were optimized.
Stability of Toxin Oxidation Products
AOAC Method 2005.06 describes issues with toxin
oxidation product stability, resulting in significant limitations
for the application of the method to a high-throughput routine
monitoring environment, where products must remian stable
over a suitable period of time without the need for additional
calibrants within the sequence. As such, the use of a cooled
autosampler was investigated and the subsequent stability of
toxin oxidation products was determined over a 24 h period.
Single-Laboratory Method Validation Scheme
C18 cleaned-up, extracted mussel tissue homogenate was
oxidized by both periodate and peroxide and analyzed by
HPLC-FLD. Periodate oxidation was performed in the
presence of matrix modifier (10). The selectivity of the
method involved analyzing oxidized sample aliquots
alongside unoxidized extracts and PSP calibration standards
in order to determine, qualitatively, whether mussel extracts
contained any naturally fluorescing compounds that had the
potential to interfere with the presence of PSP toxins.
Linear range was determined by spiking PSP toxins into
mussel extracts, fractions, and 0.1 mM acetic acid to produce
concentrations over a range of 0–2.0 mg STX eq/g; GTX5 =
0–0.2 mg STX eq/g. Periodate and peroxide oxidations were
performed in triplicate before HPLC-FLD analysis and linear
regression equations were generated. The linearity of the
analytical method was evaluated graphically, and
comparisons were made between calibrations from acetic acid
and mussel-extract-based standard solutions. Regression
coefficients and statistical F-test, goodness-of-fit were
examined for each calibration.
Method LODs were established from a PSP toxin S/N ratio
of 3:1 and were determined experimentally for both the
periodate screening step and the full quantitation method
using triplicate homogenate spikes at predicted LOD
concentrations. Triplicate oxidations were used to assess
variability of the amount. Limits of quantitation (LOQ) were
experimentally determined using an S/N ratio of 10:1.
Periodate oxidant pH (C = pH 8.15, c = pH 8.25)
Vortex mixing time (D = 3 s, d = 6 s)
Ambient temperature during oxidation (E = 22°C, e = 25°C)
Oxidation time (F = 55 s, f = 65 s)
Matrix modifier (G = Modifier 1, g = Modifier 2)
Without the availability of a mussel certified reference
material (CRM), method accuracies were investigated using 2
candidate and noncertified mussel homogenates (supplied by
IMB, NRCC).
Assessment of the recovery of PSP toxins from mussel
tissue involved the triplicate spiking of homogenates of each
toxin to provide expected concentrations relating to 0.16 mg
STX eq/g and 0.40 mg STX eq/g for each toxin (GTX5 at
0.016 and 0.04 mg STX eq/g). Toxins were divided into
groups of spiking mixtures (Table 1). One mixture (V)
contained all of the toxins for a limited number of tests,
including recovery assessment for dcNEO.
Instrumental peak area precision was assessed with the
repeat analysis of matrix-matched toxin standards in one
analytical batch. Repeatability of chromatographic retention
time was assessed with the repeat analysis of toxin standards
over a 2-month period. Short-term method repeatability
involved the analysis of triplicate homogenates spiked at both
0.16 and 0.40 mg STX eq/g concentration per toxin, analyzed
in one analytical batch. Medium-term repeatability involved
analysis of 2 batches of triplicate homogenates spiked at 0.16
and 0.40 mg STX eq/g, analyzed in 2 batches by more than one
analyst more than 2 weeks apart. Long-term repeatability was
assessed using the repeated extraction, cleanup, and analysis
of a naturally contaminated laboratory reference material
(LRM) over more than 2 months. Method precision was also
assessed with 10 replicate analyses of 6 naturally
contaminated mussel extracts. Intrabatch repeatability was
assessed with the analysis of each extract in one analytical
batch. Interbatch repeatability was investigated by analyzing
each of the extracts in different analytical batches.
Ruggedness was investigated by using the repeat analysis
of spiked mussel homogenates and followed a
Plackett-Burman experimental design (15) using key method
parameters (Table 2). Ruggedness tests were conducted for
model toxins (0.40 mg STX eq/g) of the non-N-hydroxylated
(STX and GTX2,3) and N-hydroxylated (GTX1,4 and NEO)
groups. Main effects were calculated as the difference of
means for each paired set of parameter levels (parameter
differences) and were tabulated in order of magnitude. The
parameter difference was also calculated as a percentage of
the spiked concentration (0.4 mg STX eq/g).
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 193
Figure 1. Typical chromatographic patterns obtained with 3 mixtures of analytical standards of PSP toxins,
including dcNEO and dcGTX2,3.
194 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
Results were used from the validation studies to calculate
an overall value of method uncertainty for the measurement of
PSP toxins in mussel tissue. Expanded uncertainties were also
calculated (16, 17). The contribution and effects of matrix
modifier and sampling uncertainty were not investigated in
this study.
Parallel Mussel Sample Analysis
Following SLV, the HPLC-FLD method was applied in
parallel with the MBA as used in the national Official Control
monitoring programs for PSP toxins in shellfish. All samples
were screened by HPLC-FLD and fully quantified using
AOAC Method 2005.06. PSP toxin concentration data
derived from HPLC-FLD analysis were compared with MBA
results, and the correlation between the 2 data sets was
determined statistically.
Results and Discussion
Method Modifications
Figure 2. Trellis plot for stability of oxidation products
of PSP toxins as shown by peak area responses minus
mean responses of diagnostic peaks measured
over 24 h.
(a) HPLC-FLD
profiles
of
PSP
toxin
standards.—Chromatographic elutions of each toxin were
similar to those described in the AOAC study (10). dcGTX2,3
and dcNEO were not included in previous AOAC
validation (10), although their chromatographic behavior has
been previously described (6, 18). Oxidations of dcGTX2,3
by periodate and peroxide each revealed 2 fluorescing
products, with retention times at 2.4 and 2.8 min. The 2.4 min
peak did not co-elute with fluorescing products of other toxins
(e.g., GTX1,4) and was thus selected for purposes of
quantitation. As an N-hydroxylated toxin, dcNEO was
expected to exhibit fluorescent products after periodate
oxidation, and a prominent peak was observed at ~4.5 min.
This peak co-eluted with the smaller of the 2 major dcSTX
peaks after periodate oxidation, and an additional, minor peak
was seen at ~5.2 min due to the dcSTX impurity (NRCC data
sheet). Because of the relative size of the dcSTX impurity to
the dcNEO, it is assumed that the effect of this impurity on
dcNEO quantitation was negligible. Chromatograms of the
3 mixes of toxins used as analytical standards are shown in
Figure 1.
(b) Toxin oxidation product stability.—AOAC Method
2005.06 highlights issues with the stability of toxin oxidation
products, advising that the stability should not be assumed
beyond 8 h (10). This limitation would result in either a
restricted sequence length, or the use of additional standards
throughout the sequence. However, our analysis of standard
stability using a temperature-controlled HPLC autosampler
(set at 4°C) showed that the oxidation products of all toxins
were stable for up to 24 h and potentially longer (Figure 2).
(c) Refinement to ion-exchange fractionation and
automated SPE.—Silica-bound ion-exchange SPE cartridges
were initially tested to assess the fractionation of the toxin
suite with the addition of both dcNEO and dcGTX2,3. The
new toxins eluted in the expected fractions, with dcGTX2,3
eluting in fraction F2 and dcNEO eluting in fraction F3
(Table 3). Several cartridges were subsequently considered
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 195
Table 3. Ion-exchange fractionation optimization
Method
AOAC 2005.06
Optimized
Fraction
Toxin
Elution solvent
Fraction volume, mL
Elution solvent
Fraction volume, mL
F1
C1,2
Water
6.0
Water
5.0
F2
GTX1,4, GTX2,3, dcGTX2,3, GTX5
0.05 M NaCl
4.0
0.3 M NaCl
3.0
F3
STX, NEO, dcNEO, dcSTX
0.3 M NaCl
5.0
2.0 M NaCl
3.0
Figure 3. Chromatograms showing fractions (F1, F2, and F3) collected after ion-exchange SPE using Strata-X-CW
of a solution containing all commercially available certified PSP reference standards.
196 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
Table 4. Linear regression parameters plus RSDsa of response factors and F-test of residual variance
(F-critical = 1.98)b
Toxin
GTX1,4
dcNEO
NEO
dcSTX
GTX2,3
Matrix
Calibration gradient
Intercept
r2
RSD% of
response factors
F-test
Solvent
3.06
0.13
0.968
13
0.27
Fraction F2
2.03
0.21
0.970
14
0.22
Solvent
12.61
0.40
0.995
6
0.22
Fraction F3
7.93
–0.15
0.989
8
0.52
Solvent
9.41
–0.09
0.996
8
0.13
Fraction F3
11.83
–0.50
0.963
17
0.42
Extract
68.6
0.92
0.994
5
0.44
Solvent
76.6
0.60
0.996
4
0.36
0.15
0.983
8
0.36
Extract
Solvent
GTX5
STX
C1,2
dcGTX2,3
a
b
6.87
0.01
0.968
13
0.33
Extract
108
6.64
–0.37
0.993
6
0.002
Solvent
119
–0.12
0.927
17
0.004
Extract
18.48
0.26
0.993
5
0.41
Solvent
15.91
0.25
0.956
13
0.45
Extract
21.06
1.87
0.970
13
0.17
Solvent
25.21
–0.14
0.986
7
0.53
Extract
6.82
0.72
0.986
16
0.04
Solvent
9.69
–0.80
0.902
20
0.60
RSD = Relative standard deviation.
Values calculated for each PSP toxin in mussel extract, solvent, and fractions (when applicable) over working calibration range (0–0.96 mg
STX eq/g; GTX5 0–0.01 mg STX eq/g).
with a view to increase method sensitivity and improve the
practicality of the method. The Strata-X-CW (Phenomenex)
SPE cartridge contains a polymer sorbent backbone, rather
than the silica stipulated by Lawrence, and an additional
reverse-phase element, resulting in a modified sorbent
selectivity and the need to optimize the concentrations of salt
solutions used for fraction elution. Optimization results
shown in Table 3 indicated that while different concentrations
of sodium chloride solution were used to elute F2 and F3, the
toxin compositions observed in each fraction were the same as
described previously (10) and observed in this study.
Furthermore, low levels of PSP toxin fraction
carryover (6, 11) were not observed when this cartridge was
used. Chromatograms of the 3 fractions collected following
ion-exchange SPE of a solution containing all toxin standards
are shown in Figure 3.
Although silica ion-exchange cartridges may still be used
for the method, including analyte extensions, this approach
represented an overall improvement to the fractionation
described in the AOAC method, both in terms of fractionation
selectivity and subsequent analytical sensitivity and toxin
recovery. A third advantage was the lack of cartridge
de-conditioning following the introduction of air into the
cartridge. Polymer SPE cartridges are ideally suited to
automated SPE systems, where instrument optimization will
not require checking that columns do not become dry each
time they are used. The Strata-X-CW is therefore the preferred
and recommended approach, although both silica-bound and
polymer-bound cartridges are usable.
The SPE (C18 cleanup and ion-exchange fractionation)
processes are time-consuming and labor-intensive to perform
manually; therefore, these processes were automated. Gilson
Aspec (Anachem, Luton, UK) liquid handlers were
programmed to perform both C18 cleanup and fractionation
procedures, with parameters defined to produce fast, reliable,
and repeatable cleanups. Volumes and flow rates were
programmed according to the parameters shown in Table 3.
Critical steps included proper positional adjustment of
needles, racks, and sample trays, the instigation of a full and
thorough system flush after each batch of cleanup, use of
regular system purges before use to ensure removal of all air
bubbles from the solvent flow system, and regular monitoring
of sample eluant collection volumes for quality
control purposes.
This instrumentation was utilized throughout the
subsequent validation program, enabling up to 32 C18
cleanups and 16 fractionations to be carried out within 1 h.
This represented a significant reduction in manual labor and
an essential requirement for application of this method to
high-throughput routine monitoring in a regulatory
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 197
Figure 4. Calibration plots of (a) dcSTX and (b) dcNEO concentration against detector response for standard
prepared in cleaned-up tissue extract or fraction and solvent over the working calibration range of 0–1.2 mg
STX eq/g.
environment. The instrumentation was robust over a
significant period of time (>2 years), and additional quality
control checks were instigated to highlight any potential
instrument failures.
Method Validation
(a) Selectivity.—The HPLC-FLD method was selective
for the analysis of the majority of PSP-contaminated mussel
extracts prepared in acetic acid. Chromatographic profiles of
naturally fluorescing matrix co-extractives were similar to
those highlighted in previous studies (6). Peaks were seen at
the same retention times as the early-eluting toxins
(dcGTX2,3 and GTX1,4) relating to co-extractives. Naturally
contaminated mussels contain variable amounts of these
interfering compounds, implying a need, as proposed in
AOAC Method 2005.06, to run unoxidized samples alongside
oxidized extracts. The nontoxin contributions are then
subtracted from overall toxicity. Despite a degree of
uncertainty, such calculations were shown to reduce the
likelihood of false positives.
(b) Linearity.—In all cases, results showed a linear fit to
be the preferred model, with separate slopes for each matrix
(solvent and extract/fraction). The results are summarized in
Table 4. Statistical analysis of calibration plots using both
correlation coefficients and F-test goodness-of-fit of the
residuals, indicated no significant, systematic deviations from
linearity within mussel extracts and fractions examined over
the concentration range of 0–0.96 mg STX eq/g (0 to 0.096 for
GTX5) for any of the PSP toxins studied. Differences in slope
of the calibrations between solvent and mussel extracts were
198 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
Table 5. Comparison of calibration slope gradients, intercepts, and correlation coefficients for C1,2 in C18-cleaned
mussel extract and solvent
(0 to 1.0 mg STX eq/g)
Toxin
C1,2
Matrix
Gradient
(0 to 2.0 mg STX eq/g)
r2
Intercept
Gradient
Intercept
r2
Extract
21.06
1.87
0.970
12.62
6.02
0.796
Solvent
25.21
–0.14
0.986
25.01
–0.16
0.991
noticeable for the N-hydroxylated toxins GTX1,4, NEO, and
dcNEO (Figure 4), whereas most non-N-hydroxylated toxins
did not exhibit any such differences (e.g., dcSTX). Deviations
for linearity above 0.8 mg STX eq/g were found only with
C1,2 in C18 cleaned mussel extract (Table 5).
(c) Limits of detection and quantitation.—LOD values
were calculated for the periodate oxidation of all toxins in a
cleaned-up mussel extract in order to describe the LODs for
the screening element of the method (Table 6). For toxins with
more than one oxidation product, the LOD values quoted
represent a cautious assessment of detection limits for the
primary toxin peaks, as they include confirmation of the
presence of secondary peaks in all cases. LOD values ranged
from ~0.03 to 0.14 mg STX eq/g with NEO and dcSTX
exhibiting the lowest LODs and C1,2 showing the highest
LOD at 0.14 mg STX eq/g.
Table 7 presents method LODs for the quantitation of PSP
toxins in mussels. The method was less sensitive for GTX1,4
(0.16 mg STX eq/g), whereas sensitivities for other toxins
(0.003–0.087 mg STX eq/g) were significantly better. Due to
toxin availability, the LOD value for dcNEO (0.16 mg
STX eq/g) was quoted conservatively as the concentration
spiked during recovery testing, where results indicated an S/N
ratio >3.0. The LODs were also compared (Table 7) against
the values quoted in AOAC Method 2005.06, and similar
results were seen, with the exception of GTX1,4. The LOD
obtained for dcGTX2,3 indicated a good level of sensitivity
for this toxin in comparison with the others.
LOQ values are also summarized in Table 7, ranging from
0.006 to 0.16 mg STX eq/g for the non-N-hydroxylated toxins.
Results obtained for the N-hydroxylated toxins show
acceptable LOQs for NEO and dcNEO (£0.16 mg STX eq/g),
whereas LOQs recorded for GTX1,4 were in the region of
0.38 mg STX eq/g. The degree of confidence in the method at
0.16 mg STX eq/g may therefore be questionable for GTX1,4.
However, results from the medium-term precision analysis of
GTX1,4 at 0.16 mg STX eq/g show a 16% relative standard
deviation (RSD) and a HorRat of 0.76; the long-term
precision GTX1,4 data taken from the analysis of the LRM
also showed an acceptable RSD and HorRat. These results
show that the level of precision is acceptable at 0.16 mg
STX eq/g and that of GTX1,4 can therefore be quantified with
an acceptable degree of precision below the validated level
of quantitation.
(d) Accuracy.—Table 8 shows the PSP toxicities obtained
by our laboratory (precolumn oxidation) and the preliminary,
noncertified values for a candidate mussel reference material
(CRM-PSP-Mus), as obtained by the NRCC (postcolumn
oxidation). Total toxicity values as determined by MBA for
the same pre-reference material [Canadian Food Inspection
Agency (CFIA)] are also quoted. On a qualitative basis, all
toxins identified using the AOAC method were present using
the postcolumn oxidation approach, with the exception of the
low-level dcNEO, which was not reported by NRCC. Relative
proportions of toxins showed near-identical profiles with
those calculated from the preliminary CRM. Concentrations
of NEO, dcSTX, and dcGTX2,3 detected using the AOAC
method were supported by the NRCC results. Concentrations
of these toxins fell between 0.01 and 0.04 mg STX eq/g, at
which concentrations a lower level of accuracy was expected,
although there was higher accuracy obtained for the very low
levels of dcGTX2,3.
Table 6. LOD (mg STX eq/g ± 1 SDa) of the HPLC-FLD
screening method for PSP toxins following periodate
oxidation of C18-cleaned mussel extracts
Peak
assignment
Rt, minb
LOD, mg STX
eq/g
GTX1,4
Primary
2.8
0.08 ± 0.014
GTX1,4
Secondary
7.0
Detected
dcNeo
Primary
4.6
0.05 ± 0.006
NEO
Primary
5.3
0.03 ± 0.002
NEO
Toxin
Secondary
9.6
Detected
dcSTX
Primary
4.5
0.03 ± 0.008
dcSTX
Secondary
5.3
Detected
GTX2,3
Unique
6.8
0.04 ± 0.009
GTX5
Unique
8.8
0.06 ± 0.005
STX
Unique
9.5
0.07 ± 0.020
dcGTX2,3
Secondary
2.4
Detected
dcGTX2,3
Primary
2.8
0.04 ± 0.010
C1,2
Unique
4.2
0.14 ± 0.025
a
b
SD = Standard deviation.
Rt = Response time.
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 199
Table 7. LOD and LOQ of the HPLC-FLD quantitation
method for PSP toxins following periodate oxidation of
fractions and peroxide oxidation of C18-cleaned
mussel extracts
Toxin
GTX1,4
dcNEO
LOD (mg STX
eq/g) ± 1 SDa
AOAC 2005.06
LOD (mg STX
eq/g)
LOQ (mg STX
eq/g) ± 1 SD
0.16 ± 0.04
0.05
0.38 ± 0.09
b
<0.16
NR
0.16 ± 0.04
NEO
0.068 ± 0.01
0.04
0.14 ± 0.04
dcSTX
0.007 ± 0.003
0.004
0.01 ± 0.002
GTX2,3
0.087 ± 0.03
0.08
0.17 ± 0.03
GTX5
0.003 ± 0.001
0.002
0.006 ± 0.001
STX
0.018 ± 0.007
0.022
0.04 ± 0.01
dcGTX2,3
0.053 ± 0.01
NR
0.11 ± 0.01
C1,2
0.018 ± 0.006
0.01
0.04 ± 0.01
a
b
SD = Standard deviation.
NR = No result available.
The overestimation of dcSTX using the AOAC method
also resulted in the significant underestimation of NEO due to
the subtraction of the dcSTX contribution. Without the
subtraction of the dcSTX contribution to the NEO peak, the
NEO value quantified was 0.06 mg STX eq/g, similar to the
value quoted by the NRCC. This may further highlight
potential issues with quantifying N-hydroxylated toxins in the
presence of dcSTX and dcGTX2,3, but more likely represents
the higher level of method uncertainty associated with low
concentrations using both methods. Values obtained for
GTX2,3 and STX using the AOAC method were both higher
than those quoted by the NRCC. However, without
certification of the NRCC values, it is impossible to determine
whether the inferred inaccuracy is due to either or both of the
2 techniques.
The MBA result was 2.60 mg STX eq/g, which compares
well with the total toxicity calculated using the AOAC
method. Table 9 shows the comparability between the
2 methods for a second candidate CRM (PSP-Mus-Pilot); all
toxins identified qualitatively by the AOAC method were also
detected by the postcolumn oxidation method. Proportions of
toxins quantified by the AOAC method were similar to those
reported by the NRCC values. Values for dcGTX2,3 derived
from the 2 methods were close; GTX2,3 and STX values
determined after precolumn oxidation were 10–30% higher
than those determined by the NRCC. Values for
N-hydroxylated toxins were less in agreement, but with the
uncertainties associated with each method and the
complexities associated with quantifying N-hydroxylated
toxins, the level of comparability was encouraging.
(e) Recovery from spiked mussel homogenate.—Table 10
presents the mean recovery percentages of PSP toxins from
mussel tissue homogenates spiked in triplicate at
0.40 mg STX eq/g and 0.16 mg STX eq/g, respectively.
Recoveries of non-N-hydroxylated toxins (STX, GTX2,3,
GTX5, dcSTX, C1,2, and dcGTX2,3) spiked at
0.40 mg STX eq/g following peroxide oxidation of
C18-cleaned mussel extracts range from 63 to 76%.
Recoveries were not thought to be matrix-influenced as no
signal suppression was identified. Similarly, the
N-hydroxylated toxins (GTX1,4 and NEO) did not show any
reduced recoveries at the 0.40 mg STX eq/g concentration
level, with recoveries of 81 and 107%, respectively, although
Table 8. Concentrations and relative proportions of PSP toxins in NRCC mussel reference material (mg STX eq/g;
± 1 SD) using Cefas AOAC method, postcolumn oxidation HPLC-FLD method (NRCC; preliminary noncertified results)a
Toxin
AOAC
(Cefas)
Postcolumn
(NRCC)
Accuracy using
postcolumn
FLD as reference, %
Relative toxin
proportions, %
(AOAC)
Relative toxin
proportions, %
(NRCC)
GTX1,4
NDb
ND
—
ND
ND
dcNEO
0.03 ± 0.012
NAc
—
1
NA
NEO
0.01 ± 0.004
0.06 ± 0.006
19
1
3
dcGTX2,3
0.04 ± 0.005
0.03 ± 0.003
139
2
2
C1,2
ND
ND
—
ND
ND
dcSTX
0.03 ± 0.001
0.01 ± 0.001
284
1
1
GTX2,3
0.30 ± 0.02
0.18 ± 0.01
165
12
10
ND
ND
—
ND
ND
1.96 ± 0.07
1.47 ± 0.03
134
83
84
2.38
1.75
GTX5
STX
Total
a
b
c
MBA results (CFIA) = 2.60 mg STX eq/g; NRCC = National Research Council Canada; Cefas = Center for Environmental Fisheries and
Aquaculture Science; MBA = mouse bioassay; FLD = fluorescence detection; CFIA = Canadian Food Inspection Agency.
ND = Not detected.
NA = Not analyzed for.
200 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
the poorer average recovery of dcNEO (53%) may relate
primarily to the fluorescence signal suppression seen in
mussel fraction extract as compared with solvent (Figure 4).
Further evidence for this is given in Table 4, which details
the calibration gradients measured from both solvent and
matrix-spiked dcNEO, with the matrix-spiked calibration
showing a gradient 63% of the solvent-spiked gradient.
Recoveries obtained for toxins GTX1,4 and NEO, both
commonly found in UK mussels, are higher than those
described in the AOAC method. This may relate to the
improved performance of the ion-exchange cartridges used in
our laboratory, as low recoveries were experienced with the
cartridges used in the original study (10). With the exception
of dcNEO, results obtained on mussel matrix spiked at
0.16 mg STX eq/g for each toxin showed mean recoveries in
the range of 79–136%, with STX and NEO recoveries of
>100%. Such values may reflect method uncertainties
associated with toxins at this level or a signal
enhancement phenomenon.
(f) Precision (short-, medium-, and long-term).—
Instrumental precision.—Table 11 shows that the level of
precision of chromatographic retention times was high over a
2-month period; thus, a high degree of confidence can be
placed upon the toxin peaks consistently eluting at repeatable
retention times. Results are also tabulated from the repeat
analysis (n = 6) in one sequence of one sample of a 0.40 mg
STX eq/g Mix V-spiked mussel homogenate, oxidized using
both periodate and peroxide oxidants. Relevant quantitation
peak areas were recorded and the RSD values of the replicate
analyses ranged between 0.2 and 3.5% (mean = 1.4% RSD).
Short-term repeatability: spiked mussel tissues.—Intrabatch,
RSDs determined from the analysis of mussel homogenates (n
= 3) spiked at 0.16 mg STX eq/g and 0.40 mg STX eq/g for
each PSP are shown in Table 11. With the exception of
dcGTX2,3, all RSD values were £10%. HorRat values (19)
were <2.0 at both concentration levels, although low HorRat
values <0.25 for some toxins are reported with caution.
Lawrence and Niedzwiadek (6) reported RSD values <10%
for the quadruplicate analyses of PSP toxins, although these
results related to spiked extracts rather than spiked tissue
samples. Medium-term repeatability: spiked mussel tissues.—
Table 11 shows the precision relating to the extraction,
cleanup, oxidation, and analysis of 6 replicate spiked samples
(both 0.16 and 0.40 mg STX eq/g) performed over a period of
>2 weeks. For the assessment of medium term precision,
Mix V spiking solution was used for all non-N-hydroxylated
toxins. Mussel homogenate was spiked separately with an
N-hydroxylated toxin mix (Mix I) to assess the method
precision for NEO and GTX1,4. RSD values for the
0.16 mg STX eq/g spiked tissues ranged from 9 (dcGTX2,3) to
41% (C1,2). At 0.40 mg STX eq/g, RSD values ranged from 4
to 32% for all toxins, with GTX1,4 and NEO showing a high
degree of medium-term precision. Overall, the precision at
0.40 mg STX eq/g was better than at 0.16 mg STX eq/g, with
mean RSDs of 15 and 24%, respectively. Acceptable
precision is further evidenced from the HorRat values, which
were <2.0 for most toxins at both concentration levels
(Table 11). Toxins which demonstrated low HorRat values for
the short-term repeatability exercise showed more realistic
HorRat values over the medium-term. Notable results with
RSDs >30% were C1,2 and NEO at 0.16 mg STX eq/g and
GTX5 at 0.04 mg STX eq/g, with NEO at 0.16 mg STX eq/g
giving a HorRat value of >2.0. Although this illustrates the
higher degree of medium-term variability associated with the
analysis of this toxin, HorRat values quoted in the AOAC
method were also >2.0 for NEO, and the range of RSD values
Table 9. Concentrations and relative proportions of PSP toxins identified in NRCC mussel pilot reference material
(mg STX eq/g; ± SD) using AOAC method (Cefas), postcolumn oxidation HPLC-FLD method (NRCC)a
Toxin
GTX1,4
dcNEO
AOAC
(Cefas)
Postcolumn
(NRCC)
Accuracy using
postcolumn
FLD as reference, %
1.75 ± 0.24
1.20
147
b
c
Relative proportions, Relative proportions,
% (AOAC)
% (NRCC)
32
26
ND
NA
—
0
NA
NEO
0.30 ± 0.02
0.49
61
5
11
dcGTX2,3
0.76 ± 0.01
0.78
98
14
17
ND
ND
—
ND
ND
C1,2
dcSTX
0.02 ± 0.001
NA
—
1
NA
GTX2,3
1.46 ± 0.02
1.14
129
27
25
GTX5
STX
Total
a
b
c
ND
ND
—
ND
ND
1.13 ± 0.01
0.97
116
21
21
5.42
4.57
MBA results = 6.40 mg STX eq/g; NRCC = National Research Council Canada; Cefas = Center for Environmental Fisheries and Aquaculture
Science; MBA = mouse bioassay; FLD = fluorescence detection.
ND = Not detected.
NA = Not analyzed for.
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 201
Table 10. Mean percentage recoveries and RSDs of PSP toxins from mussel (n = 3) homogenates spiked at expected
concentrations of 0.4 and 0.16 mg STX eq/g (GTX5 1/10 concentration), plus comparison with range of mean
interlaboratory recoveries reported by AOAC 2005.06 (at variable concentration levels)
0.40 mg STX eq/g, %
0.16 mg STX eq/g, %
AOAC (ref. 10), %
GTX1,4
81 (4)
112 (4)
67–79
dcNEO
53 (4)
29 (4)
NAa
107 (6)
136 (2)
53–62
dcSTX
68 (8)
85 (1)
64–84
GTX2,3
66 (9)
94 (7)
76–88
Toxin
NEO
GTX5 (1/10 concn.)
76 (7)
82 (1)
76–86
STX
68 (5)
122 (0.2)
74–93
dcGTX 2,3
70 (18)
82 (1)
NA
C1,2
63 (10)
79 (6)
74–78
a
NA = Not analyzed.
quoted in the method typically ranged between 10 and
>50% (10). Considering the high variability inherent in such a
multistep method, the majority of these values appeared to be
acceptable and indicated that the method was repeatable
within a single laboratory over the medium term.
Medium-term repeatability: analysis of naturally
contaminated
mussels.—Intrabatch
and
interbatch
repeatability was acceptable for most toxins at concentration
levels >0.16 mg STX eq/g, and most precision values were
<20% RSD over the 10 analyses in the same batch (Table 12).
As expected, longer-term, interbatch repeatability is of higher
variability, with RSD values ranging between 10 and 31% for
all toxins at concentrations >0.16 mg STX eq/g. Such values
appear reasonable given the variability associated with the
AOAC method, as exemplified by the medium-term
repeatability data generated from spiked mussel tissues
(Table 11). Even toxins present at levels <0.16 mg STX eq/g
exhibited a satisfactory degree of repeatability, for example
C1,2 (34%), NEO (30%), and GTX5 (34%). HorRat values
for intrabatch and interbatch repeatability were <2.0 for
all toxins. Long-term repeatability.—Table 13 shows the
realistic measure of long-term repeatability of the method.
Long-term analyses of the LRM showed GTX1,4, GTX2,3,
and STX were present in the testing material at concentration
levels >0.16 mg STX eq/g and all 3 analytes exhibited
long-term repeatability values between 11 and 27% (HorRat
Table 11. Instrumental precision, showing variability (RSD%) of toxin retention times and peak area responses,
short-term repeatability (n = 3, <2 weeks) and medium-term repeatability (n = 6, >2 weeks) of analysis of spiked mussel
homogenate at 0.16 and 0.40 mg STX eq/g per toxin (GTX5 1/10 concentration)
Short-term repeatability
Instrumental precision
Toxin
Toxin peak
area
Retention time
(RSD %, n = 7) (RSD %, n = 6)
0.16 mg STX eq/g
tissue spikes
Medium-term repeatability
0.40 mg STX eq/g
tissue spikes
0.16 mg STX eq/g
tissue spikes
0.40 mg STX eq/g
tissue spikes
RSD, %
HorRat
RSD, %
HorRat
RSD, %
HorRat
RSD, %
HorRat
GTX1,4
1
0.6
4
0.29
4
0.33
17
0.82
4
0.21
dcNEO
2
2.1
4
0.29
5
0.41
20
0.94
26
1.41
NEO
2
3.5
2
0.14
6
0.49
51
2.42
26
1.40
dcSTX
1
1.0
1
0.07
8
0.66
32
1.49
23
1.23
GTX2,3
2
1.0
7
0.50
9
0.74
22
1.06
10
0.53
GTX5 (1/10)
4
0.2
1
0.05
7
0.41
26
0.86
32
1.25
STX
1
3.0
0.2
0.01
5
0.41
16
0.77
5
0.29
dcGTX2,3
1
0.3
1
0.07
18
1.48
9
0.45
15
0.81
C1,2
2
0.6
6
0.43
10
0.82
41
1.94
10
0.54
202 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
Table 12. Summary of naturally contaminated mussel sample repeatability data showing mean RSDs, range of RSD
values obtained, and HorRat values for both intra- and interbatch repeatability data
Intrabatch precision
Toxin
Mean concn
(mg STX eq/g)
Mean RSD, % Range of RSD, %
Interbatch precision
HorRat
Mean RSD, % Range of RSD, %
HorRat
GTX1,4
0.77
16
NEO
0.11
C1,2
0.04
GTX2,3
0.51
9
4–17
0.77
26
22–31
1.47
GTX5
0.002
17
14–20
0.63
34
30–41
0.83
STX
0.49
10
4–16
0.85
18
14–25
1.01
1.75
8
2–12
0.82
14
10–22
0.95
Total
7–22
1.46
19
16–22
1.14
12
6–21
0.82
30
15–43
1.34
21
7 –39
1.23
34
24–48
1.31
<2.0). These were similar to those generated from the
medium-term repeatability studies (Table 11) and, as such,
give a further degree of confidence in the method for the
analysis of these 3 toxins. Conversely, NEO, C1,2, and GTX5
were determined at much lower concentrations and
consequently demonstrated higher variabilities. Nevertheless,
HorRat values were <2.0 for all toxins except C1,2 (mean
concentration 0.01 mg STX eq/g), indicating that the level of
repeatability is acceptable.
(g) Ruggedness.—Non-N-hydroxylated toxins.—Results
presented in Table 14 show that Vortex mixing time, a
parameter not defined in AOAC Method 2005.06, had the
most significant effect on method stability, with the negative
parameter difference confirming that a longer extraction
period resulted in improved extraction efficiency. To
determine whether such parameter differences are significant
and result in method instability, the results were compared
against method precision using a significance test (t-test).
Using 9 analyses of STX and GTX2,3 during the short-term
precision scheme, t-test values were calculated (Table 14),
with results showing all t-test values < t-critical (t-critical at
95% confidence = 4.3 for 3 single batch replicates). None of
the ruggedness parameters investigated had a statistically
significant effect on the method, suggesting that the
parameters investigated do not interact. However, because the
longer sample Vortex mixing time improved recovery and
precision, a 90 s Vortex mixing time was standardized.
N-hydroxylated toxins.—Table 15 lists the calculated
parameters in order of importance (magnitude) for both NEO
and GTX1,4, giving the result in terms of the parameter
difference, and the parameter difference as a percentage of the
spiked concentration (0.4 mg STX eq/g). These results
indicate that matrix modifier had the largest effect on the
method stability for GTX1,4, whereas the same parameter had
a relatively low effect on the stability of NEO. Oxidation
Vortex time also appeared to have systematic effect on
quantitation, with the parameter differences illustrating that
the longer mixing time results in improved oxidation
efficiency. Similarly, the closer the pH of the periodate reagent
is to 8.25, the more efficient the oxidation for both GTX1,4
and NEO.
The t-test values were tabulated for each toxin and were all
< t-critical, indicating that the method is robust for the analysis
of these toxins. However, a change in the matrix modifier did
result in noticeable differences in quantitation data, which,
although not statistically significant, may introduce further
uncertainty into the quantitation method. It is recommended
that further studies be conducted on this parameter alone to
establish appropriate method control limits.
(h) Method uncertainty.—An assessment of uncertainty
was based on the summation of standardized uncertainties
relating to the assessment of precision, recovery, and
reproducibility, with standardized uncertainty values
calculated for each individual toxin. The measurement
uncertainty inherent in the precision component was
evaluated from the statistical distribution of the results of a
series of measurements, and was characterized by standard
deviations (16, 17). Uncertainties were calculated for
medium-term precision, both at 0.16 and 0.40 mg STX eq/g
Table 13. Mean concentration ± SD [RSD (%)] and
%RSD data generated from long-term (>2 months)
extraction, cleanup, fractionation, oxidation, and
analysis of PSP LRMa
Mean concentration
(mg STX eq/g; ± SD)
RSD, %
HorRat
GTX1,4
0.40 ± 0.11
27
1.49
NEO
0.09 ± 0.042
45
1.98
C1,2
0.01 ± 0.008
69
2.19
Toxin
GTX2,3
GTX5
STX
Total
a
0.34 ± 0.061
18
0.96
0.002 ± 0.001
36
0.89
0.55 ± 0.061
11
0.63
1.41 ± 0.24
17
1.10
LRM = Laboratory reference material.
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 203
Table 14. Parameter differences, parameter difference percentages, and t-test values from ruggedness testing of
GTX2,3 and STX (t-critical = 4.3 for 3 replicates)
GTX2,3
Parameter
difference
Parameter
Vortex time (extraction)
Extract pH
LC flow rate
Parameter
difference, %
STX
t-Test value
Parameter
Parameter
difference
Parameter
difference, %
t-Test value
–0.055
–14
–0.87
Vortex time (extraction)
–0.056
–14
–1.57
0.039
10
0.62
Ambient temp.
–0.032
–8
–0.91
0.032
8
0.50
LC flow rate
0.031
8
0.88
Ambient temp.
–0.030
–8
–0.47
Oxidation time
0.023
6
0.66
C18 flow rate
–0.017
–4
–0.27
Extract pH
0.022
5
0.62
Extraction temp.
0.004
1
0.07
Extraction temp.
0.006
2
0.18
Oxidation time
0.003
1
0.05
C18 flow rate
–0.006
–2
–0.18
concentration levels. Sample repeatability data are not
included in preference to spiked sample data as the latter
include precision of extraction efficiency. RSDs were pooled
to give total standardized precision uncertainties (Table 16):
u c ( y) =
(n a -1) ´ a 2 + (n b -1) ´ b 2 +K
(n a -1) + (n b -1) +K
where uc(y) = pooled uncertainty of precision uncertainty
components; a,b = RSDs of components; and n = number of
replicates used in precision studies for each component.
The uncertainties associated with long-term precision were
estimated from the precision data generated by the repeated
extraction, cleanup, fractionation, and analysis of LRMs. For
toxins not present in the current LRM, reproducibility values
were taken directly from the mean of RSDR data from spiked
mussel matrix analyzed during the interlaboratory study of the
AOAC method (Table 16). A mean value of 0.27 was used for
toxins dcGTX2,3 and dcNEO, which were not included in the
AOAC 2005.06 interlaboratory study. The uncertainties
present in the determination of recovery were estimated by
calculating the RSD for each toxin at each spiking level.
Values were tabulated for each toxin at 0.16 and 0.40 mg STX
eq/g in Table 16. Pooled uncertainties were calculated for each
toxin as previously for precision uncertainty and were of
relatively small magnitude. Combined standardized
uncertainties for each PSP toxin were calculated (Table 16)
from the square root of the sum of squares:
u c = u 12 + u 22 + u 23 KK
where uc = combined standardized uncertainty; and u1 – un =
individual standardized uncertainties.
These uncertainties are preliminary and will change over
time as more method performance data are obtained through
routine implementation of the procedure and analytical
quality control. The results showed a combined standardized
uncertainty for individual toxins, ranging from 0.17 (for STX)
to 0.52 (for NEO). Expanded uncertainties calculated using a
coverage factor (k) of 2 subsequently result in a range of
Table 15. Parameter differences, parameter difference percentages, and t-test values from ruggedness testing of
GTX1,4 and NEO (t-critical = 4.3 for 3 replicates)
GTX1,4
NEO
Parameter
difference
Parameter
difference, %
t-Test
value
Parameter
Parameter
difference
Parameter
difference, %
t-Test
value
0.193
48
2.73
Vortex time (oxidation)
–0.036
–9
–0.86
Vortex time (oxidation)
–0.063
–16
–0.90
Periodate pH
–0.033
–8
–0.78
Periodate pH
–0.033
–8
–0.46
C18 extract pH
0.030
8
0.72
Parameter
Matrix modifier
Ambient temp.
0.016
4
0.23
Matrix modifier
0.018
5
0.43
Oxidation time
–0.007
–2
–0.10
Ambient temp.
–0.012
–3
–0.28
C18 extract pH
0.006
1
0.08
Oxidation time
–0.006
–1
–0.14
Fractionation flow rate
0.005
1
0.07
Fractionation flow rate
–0.006
–1
–0.13
0.68
0.34
0.083
0.099
0.063
—
0.15
0.30
LRM = Laboratory reference material.
a
0.41
C1,2
0.10
0.34
0.65
0.32
0.17
0.037
0.129
0.181
0.052
0.002
0.013
—
0.11
0.35
0.27
0.12
0.12
0.09
dcGTX2,3
0.15
0.16
STX
0.05
0.52
0.79
0.39
0.26
0.080
0.048
0.066
0.091
0.067
0.013
—
0.18
0.27
0.26
0.29
0.17
0.10
GTX5
0.32
0.22
0.26
GTX2,3
1.03
0.78
0.39
0.52
0.043
0.060
0.085
0.057
0.020
0.009
—
—
0.32
0.27
0.27
0.40
0.26
dcSTX
0.23
0.51
0.32
NEO
0.60
0.72
0.36
0.30
0.038
0.048
0.054
0.037
0.040
0.042
—
0.27
0.28
0.27
0.23
0.13
0.04
0.26
0.17
0.20
Toxin
dcNEO
LRMa
AOAC
Pooled
uncertainty
0.40 mg
STX eq/g
0.16 mg
STX eq/g
GTX1,4
Combined
standardized
uncertainty
Pooled
uncertainty
0.40 mg
STX eq/g
0.16 mg
STX eq/g
Uncertainty of recovery measurement
Reproducibility
Medium-term precision
Table 16. Standardized uncertainty components and combined standardized and expanded uncertainties for analysis of PSP in mussels
Combined
expanded
uncertainty (k = 2)
204 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
values from 0.34 (STX) to 1.03 (NEO). The coverage factor
(k) was taken to be 2 in order to provide a 95% confidence in
the distribution of values, assuming a normal
distribution (17).
(i) Parallel testing.—Twenty-one mussel samples, found
to be negative for PSP toxins by MBA, were analyzed by the
AOAC HPLC method, and all periodate screening analysis
showed either no PSP or low concentrations of toxins (0.06
and 0.11 total mg STX eq/g), below the limit of sensitivity of
the MBA. This confirmed the agreement between the
2 methods for samples containing toxins well below the
regulatory level. Results obtained from the quantitative
HPLC-FLD analysis of 40 mussel tissues (single analysis with
no repeats) found to contain PSP toxins by MBA (defined here
as MBA-positive) are summarized in Table 17. Inspection of
the data indicated some degree of correlation between data
derived from the 2 approaches (full data set not shown). All
MBA-positive samples were also positive for the presence of
PSP toxins by HPLC-FLD. MBA analysis indicated that 12 of
the samples contained total PSP toxin concentrations higher
than the regulatory action limit of 80 mg STX eq/100 g. When
analyzed by HPLC-FLD, 11 of these samples also produced
final toxin concentrations above action limit. One sample
resulted in an MBA of 89 mg STX eq/100 g, whereas the
analytical concentration was 77 mg STX eq/100 g. A certain
degree of variability is expected between the 2 methods, due
to the characteristics of each and the different basis for
determining toxicity.
Figure 5 shows the correlation between the HPLC and
MBA results along with a coefficient of 0.93. The mean of all
HPLC-FLD/MBA results (n = 40) is 101%, indicating that the
AOAC method agreed closely, on average, with the results
obtained by the MBA. The RSD of the HPLC/MBA results
ratios was 30%, which illustrates the degree of correlation
scatter as shown visually in Figure 5. A 2-tailed t-test
calculated results with a t-value of 0.76, which, when
compared with the t-critical value (n = 40) using a 2-tailed
t-test at 95% confidence of 2.02, indicates that there was no
significant statistical difference between the 2 analyses. The
results also show little difference between the correlations of
results obtained from fresh versus archived samples. Mean
values were similar for the data set, and the RSD of the
analysis ratios was also acceptable.
Throughout the parallel testing, unoxidized samples were
run alongside oxidized samples in order to determine the
presence of any co-extracted, naturally fluorescing
compounds. For a number of samples, levels of dcGTX2,3
and to a lesser extent, GTX1,4, interferences were detected,
thus requiring subtraction of the unoxidized peaks from the
oxidized peaks. The presence and levels of these components
varied throughout the exercise. Performing such subtractions
assumes that the signal response of naturally fluorescent peaks
will be consistent in both the unoxidized and oxidized
samples. However, the general approach is thought to be
valid, because without the use of interference subtraction,
concentrations of dcGTX2,3 would have been positively
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 205
Table 17. Summary of results from quantitative HPLC and MBA analysis of naturally contaminated mussel samples
Total number of mussel samples
61
Number of MBA positive samples
40
Number of MBA positive assigned HPLC
40 (100%)
Mean HPLC/MBA (n = 40)
101%
RSD of HPLC/MBA
30%
Correlation coefficient (r; all positives)
0.93
r; Fresh positive samples only; n = 11
0.88
r; 2007 archive positive samples only; n = 22
0.86
r; 2006 archive positive samples only; n = 7
t-Test of HPLC/MBA results
MBA >80 mg STX eq/100 g (action limit), HPLC <80 mg STX eq/100 g
MBA <80 mg STX eq/100 g, HPLC >80 mg STX eq/100 g
a
0.94
0.76 (t-criticala = 2.02)
1 (2.5%)
0 (0%)
t-Critical using 2-tailed t-test at 95% confidence.
biased in a number of positive samples, giving rise to
overestimation of concentrations.
Effects of Toxicity Factor Variability on Correlation of
Results
Table 18 summarizes the comparative results between
LC-FLD and MBA when different toxicity factors were used.
Use of the Genenah and Shimizu toxicities (20) resulted in a
higher mean analytical result, with the mean HPLC/MBA of
121% describing a positive HPLC bias. The t-test analyses of
the HPLC and MBA data failed for these results, indicating
that use of these values results in significantly different data
sets (Table 18). Additionally, 3 of the 40 samples exhibited
false positives (MBA <80 mg STX eq/100 g; HPLC >80 mg
STX eq/100 g; 7.5% of data set). Use of the lowest TEF for
each isomeric pair (TEF values taken from NRCC) resulted in
a negative HPLC bias as compared with the proposed use of
the higher TEFs. As a result, 10% of the data set showed
false negatives (MBA >80 mg STX eq/100 g; HPLC <80 mg
STX eq/100 g). Results therefore indicated a degree of
variability inherent in the use of TEFs and the importance of
using the most accurate values when modeling toxicity using
nonanimal methods. However, the results also indicated that
the best agreement between HPLC and MBA results was
obtained when TEFs quoted by Oshima (14) and the highest
TEF for each isomeric pair were used. These comparisons
depend on the overall toxin profile of the samples, with the
relatively low presence of C1,2 and dcGTX2,3 toxins
Figure 5. Comparison of total PSP toxin concentration results obtained by both MBA and quantitative HPLC-FLD
(natural log scale). Action limits of 1.0 and 0.5 (80 and 40 mg STX eq/100 g) are highlighted. MBA negatives/not
tested shown as MBA result at half MBA detection limit (18 mg STX eq/100 g).
206 TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009
Table 18. Effects of toxicity factor variability on HPLC/MBA results comparison (a) using standard Oshima TEFa
values and highest toxicity value for each isomeric pair, (b) using Genenah and Shimizu (20) and TEFs, and (c) using
lowest toxicity values on TEF for each isomeric pair, values quoted by NRCCb
(a) Oshima TEF +
highest isomeric toxicity
(b) Genenah and
Shimizu TEF
Number positve samples
(c) Lowest isomeric
toxicity on TEF
40
Mean MBA concentration (mg STX eq/100 g)
75
Mean HPLC concentration (mg STX eq/100 g)
83
99
Mean HPLC/MBA (n = 40), %
101
121
72
RSD% of HPLC/MBA results
30
31
32
Correlation coefficient (r ; all positives)
0.87
0.86
0.60
t-Test of HPLC/MBA results (t-critical = 2.0)
0.76
2.33
–1.64
2
58
MBA >80 mg STX eq/100 g, HPLC <80 mg STX eq/100 g
1 (2.5%)
0
4 (10%)
MBA <80 mg STX eq/100 g, HPLC >80 mg STX eq/100 g
0 (0%)
3 (7.5%)
0
a
b
TEF = Toxicity equivalence factors.
NRCC = National Research Council Canada.
reducing the overall influence of high TEF variability
(between isomers) on overall sample toxicity.
Implementation of AOAC Method 2005.06 in the UK
The UK Food Standards Agency (the UK Competent
Authority) has now approved the use of the modified AOAC
Method 2005.06 for Official Control monitoring of mussels
for PSP toxins under EU Regulation 854/2004 (1). The Center
for Environment Fisheries and Aquaculture Science
(Weymouth, Dorset, UK) implemented the method on May 5,
2008, for routine testing of all Official Control mussel
samples submitted to the laboratory from England, Wales, and
Scotland, about 3500 samples per year. Following the
guidance within the AOAC method, all samples are extracted,
cleaned up by C18 SPE, oxidized by periodate reagent, and
qualitatively screened by HPLC-FLD. Only samples showing
the potential presence of PSP toxins are progressed to full
quantitation. Extracts are then fractionated and the relevant
fractions oxidized by periodate before analysis, only if the
potential presence of N-hydroxylated toxins is shown.
Otherwise, all quantitation is performed with peroxide
oxidation of C18-cleaned extracts alongside unoxidized
extracts. Associated internal quality controls include the use
of toxin calibration standards, blanks, reference materials, and
calibration checks throughout every analytical batch. In the
absence of TEFs in current EU legislation, the Oshima
TEFs (14) are used to calculate sample toxicity, and PSP toxin
concentrations are calculated assuming the sole presence of
the highest relative toxicity for each isomeric pair. The
method is time-consuming but with the use of automated SPE
technology is capable of screening up to 36 samples and
quantifying up to 24 samples per day and as such is applicable
to the high-throughput UK biotoxin monitoring programs. To
our knowledge the United Kingdom is the first country to
implement AOAC Method 2005.06 for the sole determination
of PSP toxins in mussels for Official Control testing of
production areas and has thus moved away from reliance on
the MBA. Within the UK monitoring programs, mussels are
the most commonly produced species and in some areas are
used as an indicator species. Use of the HPLC-FLD method
will therefore substantially reduce MBA use in the PSP
Official Control testing.
Conclusions
The AOAC 2005.06 HPLC method was found to perform
adequately as an analytical procedure for the quantitative
analysis of PSP toxins in UK common mussels. The method
was refined to stabilize oxidation products and improve
ion-exchange fractionation, and the use of automated SPE has
improved sample throughput. Additionally the method was
extended to include the additional toxins dcGTX2,3 and
dcNEO. SLV of the method included linearity, detection
limits, recovery, precision, and uncertainty results for all
commercially available toxins. Ruggedness experiments
indicated the method to be robust for the parameters
investigated, although the duration of Vortex mixing time
during the sample extraction step was standardized to improve
method recovery and precision.
The method performed adequately as an analytical
procedure for the quantitative identification of PSP toxins in
UK common mussels and successfully identified
PSP-contaminated and noncontaminated samples. There was
no significant difference between the results produced by the
MBA and by quantitative HPLC analysis. These results
suggest, therefore, that implementation of the quantitative
HPLC method in the United Kingdom is safe in terms of the
ability of the method to replicate the values achieved by the
MBA for quantifying PSP toxins. The method was
implemented in the United Kingdom for Official Control
testing of mussels on May 5, 2008. Additional improvements
could be made to the method with further increases in
TURNER ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 1, 2009 207
sensitivity to the N-hydroxylated toxins and with a full
assessment of the requirements for use of matrix modifier. For
instance, it may be possible to substitute the cleaned-up
extract of Pacific oysters currently used as a modifier with a
nonbiological buffer. The throughput and turnaround of the
method could also be further improved with the shortening of
LC analysis time. Application of ultraperformance LC to this
method could potentially result in the shortening of LC times
by a factor of 2–3, significantly increasing the throughput in a
routine monitoring environment.
Acknowledgments
We gratefully thank and acknowledge the help of Kevin
Hargin and Claudia Martins (Food Standards Agency, UK)
for scientific advice and funding this study (Project code
ZB1807); Lorna Murray and Jacqui McElhiney (Food
Standards Agency, Scotland) for supply of samples;
James Lawrence for his ongoing help and encouragement;
Stephanie Rowland for her valuable help in programming the
automated fractionation; Michael Quilliam and team (NRCC,
IMB, Halifax, Canada) for supplying candidate CRMs and
sharing preliminary results; CFIA for MBA analysis of NRCC
CRM; and Begona Ben-Gigirey (Community Reference
Laboratory of Marine Biotoxins, Vigo, Spain) for supplying
Pacific oyster material used in the ruggedness study.
References
(1) EC Regulation No. 854/2004 (2004) Off. J. Eur. Commun.
L226, 83–127
(2) EC Regulation No. 1664/2006 (2006) Off. J. Eur. Commun.
L320, 13–45
(3) Official Methods of Analysis (2005) 18th Ed., AOAC
INTERNATIONAL, Gaithersburg, MD, Chapter 49, Method
959.08, pp 79–80
(4) Lawrence, J.F., & Menard, C. (1991) J. AOAC Int. 74,
1006–1012
(5) Lawrence, J.F., Menard, C., & Cleroux, C. (1995) J. AOAC
Int. 78, 514–520
(6) Lawrence, J.F., & Niedzwiadek, B. (2001) J. AOAC Int. 84,
1099–1108
(7) Vale, P., & Sampayo, M.A. (2001) Toxicon 39, 561–571
(8) Lawrence, J.F., Niedzwiadek, B., & Menard, C. (2004) J.
AOAC Int. 87, 83–100
(9) Lawrence, J.F., Niedzwiadek, B., & Menard, C. (2005) J.
AOAC Int. 88, 1714–1732
(10) Official Methods of Analysis (2005) AOAC
INTERNATIONAL, Gaithersburg, MD, Method 2005.06
(11) Ben-Gigirey, B., Rodriguez-Velasco, A., Villar-Gonzalez, A.,
& Botana, L.M. (2007) J. Chromatogr. A 1140, 78–87
(12) EC Regulation No. 882/2004 (2004) Off. J. Eur. Commun.
L191, 1–52
(13) Thompson, M., Ellison, S.L.R., & Wood, R. (2002) Pure
Appl. Chem. 74, 835–855
(14) Oshima, Y. (1995) J. AOAC Int. 78, 528–532
(15) Plackett, R.L., & Burman, J.P. (1946) Biometrika 33,
305–325
(16) ISO (1993) Guide to the Expression of Uncertainty in
Measurement, International Organization for Standardization,
Geneva, Switzerland
(17) EURACHEM/CITAC Guide (2000) 2nd Ed., Quantifying
Uncertainty in Analytical Measurement, S.L.R. Ellison, M.
Rosslein, & A. Williams (Eds), Eurachem/CITAC, St. Gallen,
Switzerland
(18) Lawrence, J.F., Wong, B., & Menard, C. (1996) J. AOAC Int.
79, 1111–1115
(19) Horwitz, W., Kamps, L.R., & Boyer, K.W. (1980) J. Assoc.
Off. Anal. Chem. 63, 1344–1354
(19) Genenah, A.A., & Shimizu, Y. (1981) J. Agric. Food Chem.
29, 1289–1291