Analyttca ChunrcaActa, 258 (1992) 61-71 61

Elsevler Science Pubhshers B V , Amsterdam

Determination of arsenic in dry ashed seafood products

by hydride generation atomic absorption spectrometry
and a critical comparative study with platform furnace
Zeeman-effect atomic absorption spectrometry and
inductively coupled plasma atomic emission spectrometry
N Ybaiiez *, M L Cervera and R Montoro
Insfituto de Agroquinuca y Tecnologii de Alunentos (CSIC), Jaune Rorg 11, 46010 Vaktua (@am)
(Recemed 7th May 1991, revised manuscnpt received 26th September 1991)

Abstract

A hydnde generation atormc absorption spectrometnc (HG-AAS) method was developed for the determmatlon of
arsemc m mussel products as part of a general plan to control this element m seafood products Organic matter IS
destroyed by the dry ashmg techmque, the ash IS dissolved m HNO,, dduted m HCI, and hydrides of arsemc are
generated by the addltlon of sodmm tetrahydroborate pnor to atormzatlon m a flame-heated quartz cell and
measurement by atomic absorption spectrometry The detectlon hmit (0 017 gg g-l), precision (3%), recovery
(100 f 4%) and accuracy (detemuned by analysmg NIST Oyster Tissue) were evaluated and an Interference study
was camed out A comparative study of HG-AAS, platform furnace Zeemaneffect AAS and mductlvely coupled
atomic emlsslon spectrometry WP-AE?S) showed that the most sultable techmque for the determmatlon of arsenic m
seafood products 1s ICP-AES For samples wth arsemc levels below the detection lmut of this techmque (0 1 fig g-‘)
the altematwe 1s the use of HG-AAS
Keywords Atonuc absorption spectrometry, Arsemc, Seafood products

It has long been known that sea fish may
contain high levels of arsenic Results for arsenic
m a total-diet study carried out m vanous coun-
tries [l-3] have shown that fish and shellfish are
the most ngmficant dietary source of arsemc,
accountmg for nearly three quarters of the total
Intake [2] It 1s very dlffxult to quantify arseruc
mtake derwmg from seafood products owing to
the multlphclty of factors, such as species, age,
sex and place of ongm, which affect arsenic con-
tent Consequently, constant control of arsenic
levels m these products 1s necessary For the
various groups of marme ammals the arsemc lev-
els reported m the hterature vary considerably, as
follows On pg g-l) fish 006 [41-45 [2], lameh-
branchs 0 1 [4]-14 4 [5], cephalopods 0 1 [6]-49
[7], gasteropods 10 [2]-40 [2] and crustaceans
< 0 04 [8]-91 [9] This indicates the wide ranges
over which the arsenic concentrations m these
products vary, and consequently the need to adapt
the analytical method to these levels
Atonuc spectrometry 1s by far the most popu-
lar technique for determmmg arsenic m marme
orgamsms The hydnde generation (HG) tech-
mque using sodium tetrahydroborate as a reduc-
tant has been the most widely used [lo-223 be-
cause it 1s a rapid, sensitive and semi- or fully
automated system It 1s well known that arsme
generation is susceptible to mterference from var-
lous elements such as Cu, Fe, Mg, Ni, Sb, Se and

ooO3-2670/92/$05 00 8 1992 - Elsemer Science Pubhshers B V All rights reserved

62
Sn [lo-12,15,16,18] If they are present m fish m
greater concentrations than those established as
producmg interference To ehmmate or mmmuze
this mterferences m the analysis of seafood prod-
ucts, the use of an Ion-exchange resm and high
concentrations of acid have been recommended
ml
Graphite furnace atomic absorption spectrom-
etry (GF-AAS) has also been used [23-261 be-
cause It 1s a sensltlve techmque and can be auto-
mated at the instrumental stage Nevertheless,
several examples of spectral Interference by phos-
phates have been reported m the determmatlon
of arsenic m fish matrlces employmg deutermm
arc background correction [241 A dramatic re-
ductlon m interference has been demonstrated by
usmg Zeeman-effect background correctlon m
combmatlon with the stablllzed temperature plat-
form furnace (STPF) concept However, non-
spectral mterference cannot be ehmmated [23,26]
Inductively coupled plasma atomic emlsslon
spectrometry (ICP-AES) with conventlonal pneu-
matic nebuhzatlon has not been used for the
determmatlon of arsenic m seafood products with
the exception of work on the determmatlon of
arsenic m mussels [27] However, this techmque
offers substantial advantages, such as the absence
of chemical mterference and a wider dynamic
range 1281 Nevertheless, m ICP spectral mterfer-
ence IS a serious problem and control can only be
achieved by an accurate knowledge of the con-
centration levels of each of the potentially mter-
fermg elements 1281
As can be deduced from the above dlscussIon,
none of the three techniques consrdered is free
from possible Interferences and there 1s no unam-
mously accepted cnterlon for choosmg one spe-
cific techmque for determmmg arsemc m seafood
products With respect to sample preparation
procedure, dry ashmg Hrlth magnesium nitrate has
many advantages because of Its slmphcrty and
safety [15,18,22] Interference occurs when ml-
crowave dIgestIon, a very fast, efflclent and smt-
able method for decomposmg marme blologlcal
tissue [26,29,301, 1s combined with HG-AAS, ow-
mg to incomplete destructlon of orgamc matter
Perchlorlc acrd has to be Included, mtroducmg
special safety requirements and possible mterfer-
N YBAkZ El- AL
ence m HG-AAS [31] Moreover, the dIy ashmg
method allows a reduction m the acldlty of the
sample injected on to the platform, keepmg sam-
ple dllutlon at mmunum levels for ICP Thus
makes it possible to obtam acid solutions m which
arsenic can be determmed by the three tech-
mques, thus permlttmg the techmques to be com-
pared
In this work, experimental condmons for com-
plete destructlon by dry ashmg and arsenic deter-
mmatlon by HG-AAS m mussel products were
optlmlzed The analytlcal charactenstlcs of the
procedure developed were evaluated A crltlcal
comparative study was carned out of HG-AAS
and platform furnace Zeeman-effect (PFZ) AAS
[26] and ICP-AES [271 for the determmatlon of
arsenic m dry ashed seafood products, with a
view to proposing a methodology possessmg ap-
proprlate analytical characterlstlcs (detection
hmlt, preclslon and accuracy) and suitable for the
wide range of concentrations of arsemc found m
seafood products
EXPERIMENTAL
Apparatus
A Heraeus Model 1100/3 muffle furnace
equipped with a Jumo DPG-44/l dIgItal mlcro-
processor was used for sample ashmg A Perkm-
Elmer Model 5000 spectrophotometer equipped
with a Perhn-Elmer MHS-10 hydride genera-
tion attachment and a Perkm-Elmer Model 056
strip-chart recorder was used A Perkm-Elmer
Zeeman/3030 atonuc absorption spectrometer
eqmpped with an HGA-600 graphite furnace and
an AS-60 autosampler was employed This mstru-
ment includes graphics display, and highly tune-
resolved signals were plotted with a Perkm-Elmer
PR-100 prmter Pyrolytic graphite-coated graphite
tubes with mserted L’vov platform, Perkm-Elmer
Part No 112660, were utlhzed exclusively A
Perkm-Elmer ICP/6000 spectrometer controlled
by a Perkm-Elmer Model 7500 laboratory com-
puter and a Perkm-Elmer PR-210 printer was
also used The Instrumental characterlstlcs are
gwen m Tables l-3
All glass and polyethylene materials were
soaked in 10% (v/v) HNO, for 2 weeks and
DETERMINATION OF ARSENIC
rinsed three times wth analytical-reagent grade
water before bemg used for the first tune Be-
tween uses they were placed m mtrrc acid for 24
h sand-bath until totally dry, the residue undergoes
Reagents
Analytical-reagent grade water with a metered
reslstlvlty of 18 MR cm was used to prepare all
samples and standards All reagents used were of
the highest purity avaJable and at least of analyt-
I&-reagent grade A stock solution of A.&II)
was prepared by dlssolvmg arsemc tnoxlde
(Rledel-de Haen) m water. Ashmg ald suspension
was prepared by stu-rmg 20 g of Mg(NOJ, 6H,O
and 2 g of MgO m 100 ml of water until homoge-
neous Potassmm lodlde solution (2%, w/v) was
prepared and stored m a refruerator Soduun
tetrahydroborate was prepared by dlssolvmg 3 g
of NaBH, m 100 ml of 1% (w/v> NaOH solution
The mtiure was filtered through Whatman No
42 paper and stored m a refrigerator
Samples
Several cans of mussel products, m brme and
m pickled sauce, were purchased m local retall
outlets for analysis for arsemc The brme or sauce
was removed from the mussels by the method for
determmmg the dramed weight m canned foods
The total contents of the can were emptied mto a
sieve with a 5-mm stamless steel mesh made of
l-mm gauge wire, the sieve being tilted shghtly to
facilitate dramage To ensure complete dramage,
the mussels were allowed to dram for 5 mm The
drained mussels were then crushed and homoge-
mzed to a fine paste m a Mouhnex domestic
blender The resulting paste was stored m prevl-
ously decontammated glass flasks and kept m a
freezer until analysis
The methods presented m this study were vah-
dated by usmg Standard Reference Matenal
(SRM) 1566a Oyster Tissue from the National
Institute of Standards and Technology (NIST)
Procedure for determznatwn of As by HG-AAS
Dry ashzng mzneralzzatzon The recommended
procedure IS as follows A 0 250-g (dry weight)
portion of mussel product or certlfled sample or
100 g (wet weight) of mussel product sample IS
taken, 5 ml 50% (v/v> HNO, and 1 ml of ashmg
63
ard suspension contammg 20% (w/v> Mg(NO,),
and 2% (w/v) MgO are added and the compo-
nents are mixed well After evaporation on a
a first careful ashmg process m a muffle furnace
150°C for 1 h, 200°C for 30 mm, 250°C for 1 h,
300°C for 3 h, 350°C for 30 mm and 450°C for
12-14 h Generally it 1s necessary to wet the ash
with 50% (v/v) HNO,, evaporate on a sand-bath
and perform a second shorter ashmg process
(150°C for 1 h, 300°C for 30 mm and 450°C for
12-14 h), once or twice, until the ash IS com-
pletely white The 100°C temperature increases
are carned out over a penod of 15 mm and
mcreases of 150°C over 30 mm The ash IS dls-
solved m 2 ml of 50% (v/v) HNO,, washed wth
water and filtered through Whatman No 1
filter-paper mto a 50-ml volumetnc flask The
solution 1s stored m a polyethylene bottle Blanks
are prepared by charring 5 ml of 50% (v/v)
HNO, and 1 ml of ashmg aid suspension using
the dry ashmg procedure
Instrumental parameters The operatmg condo-
tlons for the atomic absorption spectrometer and
hydnde generatlon apparatus which gave the best
response were selected m previous studies of the
determmatlon of arsemc m beer [32] and tomato
products [33] and are gwen m Table 1
Analyszs of mmeralzzed samples Ahquots of 1
ml of mmerahzed sample solutron (contammg
< 60 ng of arsend and 4 ml of 62 5% (v/v) HCI,
m trlphcate, were plpetted mto the hydride reac-
tlon flask Then 1 ml of 2% KI solution was
added and allowed to stand for 5 mm (arsenate 1s
reduced to arsemte by KI) The NaBH, solution
was placed m the reactlon flask and the plunger
was released after the peak maximum had been
reached on the recorder (lo-15 s), the peak
signal height was measured The readmg was
compared with an arseruc cahbratlon graph ob-
tamed by plpettmg 0, 10,20,40 and 60 ~1 of 1 pg
ml-’ workmg standard solution mto 5 ml of 50%
(v/v) HCI and measurmg m the same way as for
the samples
Procedure for determznatzon of As by PFZ-AAS
Employ the same dry ashmg procedure as for
HG-AAS, take digest sample ahquots of 1 ml and
64 N YF&E~ETAL

TABLE 1
HG-AAS operabng condltlons for the detennmatlon of arsenic 111mussels
As ekctrodeless dmzharge lamp 85W
Wavelength 193 7 nm
Sht wgdth, low 07nm
Simple sht burner 10 cm
Acetylene flow-rate 21mm-’
An flow-rate 155lmm-’
Measurmg mode Peak height
Reading tie 20 s
Recorder input voltage 10 mV
Chart speed Smmmm-’

Condltlons m hydnde generation apparatus

Sample aliquot volume 1 ml
Solution volume m reaction flask 5ml
Sample acidity 50% (v/v) HCl
Reducing agent 1 ml, 2% (w/v) KI
NaBH, solution 3% (w/v>
Argon flow-rate 450 ml mm-’

add 1 ml of water and 20 ~1 of standard aqueous Procedure for deternunatwn of As by ICPAES

solutions contammg 0, 2, 4 and 6 pg ml-’ of Take 10 00 & 0 01 g (wet weight) of mussel
arsenic and analyse using the standard addltlon product sample, add 10 ml of ashmg ald suspen-
method with the expenmental conditions mdl- slon contammg 20% (w/v) Mg(NO,), and 2%
cated m Table 2 (w/v) MgO and 5 ml of 40% (v/v) HNO,, nux

TABLE 2
PFZ-AAS operating condltlons for the determmatlon of arsenic m mussels
Electrodeless discharge lamp 85W
Wavelength 193 7 nm
Spectral band mdth 0 7 nm (low)
Background correctIon Zeeman
Measunng mode Integrated absorbance calculated by HGA software
Graphite furnace HGA-600
Furnace tube Pyrolytlcally coated mth a L’vov platform
Injection mode Autosampler AS-60
Sample volume 2Opl
Matrur modlfier 10 /.~l 0 1% (w/v) NI
Furnace programme
Step Temperature Time (s) Internal
(“Cl Hold Ar flow-rate
Ramp
(ml mu-‘>
1 Dlylng 90 10 20 300
2 Drying 120 10 20 300
3 Ashmg 800 10 10 300
4 Ashmg 1100 10 10 300
5 Atonuzatlon 2300 0 5 0
6 Cleaning 2650 1 5 300
7 Cooling 20 10 10 300
DETERMINATION OF ARSENIC 65

TABLE3
ICP-AESoperatmg condlttons for the determmatlon of arsemc m mussels
Instrumentation
Spectrometer
Sequentlal scannmg, 0 025 nm band pass, holographic UV gratmg
ICP source
Plasmatherm 2 5 kW, 27 13 MHZ
Integral quartz torch
Sample introduction
Nebuhzer pneumatic cross-flow type
Spray chamber dual concentnc spray chamber type
Sample uptake pellstalhc pump
Operating condltlons
Vlewmg height 10 mm above the load cod
Coolant Ar flow-rate 181mm-’
Nebuhzer Ar pressure 20 ps1
Nebuhzer Ar flow-rate OSlmm-’
Auxdmry Ar flow-rate 041mm-1
R f forward power 1lkW
Sample solution flow-rate 11 ml mm-l
Integratron time 5S
Wavelength 193 696 nm

well and follow the procedure described for dry
ashmg Dissolve the completely whrte ash m 5 ml
of water and 5 ml of 50% (v/v) HCI, filter through
Whatman No 1 filter-paper mto a 25-ml volu-
metnc flask and dilute the filtrate to volume with
water Store the solution m a polyethylene bottle
Prepare reagent blanks by applymg the dry ashmg
procedure to 10 ml of ashmg ald suspension and
5 ml of 40% (v/v) HNO,
Determine arsemc by employmg the optlmlzed
analytical parameters mdlcated m Table 3, cah-
brate the spectrometer with a 10 pg ml-’ arsemc
standard solution m 10% (v/v> HCl, using a 10%
<v/v) HCl solution as a blank Check the calrbra-
tlon with standard solutions of 0 5, 2 and 5 pg
ml-’ of arsemc
RJLXJLTS AND DISCUSSION
Arsemc determmatwnby HG-AAS
To optmnze the methodology a mussel product
was selected that was also used to estabhsh the
methodolomes for arsemc determmatlon by PFZ-
AAS [26] and ICP-AAS [27]
Optlmlzatwnof dry ashmg mmeralizatwn To
optmuze the conditions for the destructlon of
orgamc matter and subsequent solublhzatlon of
the ash obtamed, the followmg condltlons were
established expenmentally mtnc acrd wet dlges-
tlon before ashmg, sample size required and ap
propnate type of contamer for mmerahzatlon,
amount of ashmg aid and mmerahzatlon stages m
the muffle furnace
According to other workers, mtnc acid wet
dIgestIon before ashmg helps to drsperse the sam-
ple m the MgO-Mg(NO,), suspension, elnm-
nates fluffing out and avoids sample combustion
losses during ashmg [22,25] On the other hand,
the high chlonde content 111mussel tissue and the
NaCl mgredlent m the brine or sauce m mussel
products could volatlhze arsemc as tnchlorlde
This 1s avolded by the use of mtrlc acid, which
volatlhzes as mtrosyl chlonde m the prehmmary
mmerahzatlon stage [341
Sample size was optimized on the basis of
obtammg a sufficient concentration of arsenic for
Instrumental readmgs and at the same time keep-
mg possible matrix mterference to a mmlmum
Samples of 0 250 g (dry weight) and 0 500, 100,
66
2 00,5 00 and 10 00 g (wet weight) were mmeral-
lzed A sample srze of 0 250 g (dry weight) or 100
g (wet werght), the mimum srze surtable for this
heterogeneous sample, was selected because it
pernuts the determmatton of arsenic by HG-AAS
and PFZ-AAS, using the same acid extract, wrth
good preclsron and recoverres
The type of contamer selected as appropriate
was a 250~ml glass vessel (tall form), which pro-
vides fast mmerahzatron without sample loss An-
other type of container (Erlenmeyer flasks)
proved unsuttable because the sample splashes
out and is more drffrcult to oxtdlze
The amount of ashmg atd [Mg(NO,),-MgO]
was carefully optimized because, m addrtlon to
being a volatrhzation mhhtor, it accelerates sam-
ple mmerahzation, producmg a spongier ash
which facihtates dlssolutron Nevertheless, we are
aware that Mg(I1) produces negatrve mterference
during arsme generation [33] It was confumed
that the arsemc recovery is quantitative up to a
level of 6 mg ml-’ of Mg(I1) m a hydride genera-
tion apparatus reaction flask, as reported by Tam
and Lacrorx [151 However, the concentratron cor-
respondmg to thus level of Mg(I1) m the mmeral-
rzed sample produces a film of magnesmm on the
graphite tube wall and also mcreases the back-
ground To muumlze the amount of ashmg aid
used, a general study of recovery tests usmg drf-
ferent amounts of magnesmm was carrred out
(Table 4) On the basis of the results obtained,
for a sample size of 100 g, 1 ml of suspension
[20% M&NO,),-2% MgO] was selected, corre-
spondmg to 0 62 mg ml-’ of Mg(II) 111the reac-
tion flask and yielding a recovery of 100%
TABLE 4
Effect of amount of ashmg ald on the recovery of arsemc m
mussels b
Ashmg ald Recovery of
Volume Mg(NO,), 6H,O MgO 10CLgg-’
(ml) (%, w/v) (%, w/v)
added As (%I
1 12 25
025 57
25 25 025 60
1 10 1 66
1 20 2 100
1 40 4 100
a Sample size = 1 g (wet weight)
N YBAIkZETAL
The stages m the muffle furnace prevent ar-
senic loss by sample splashmg or igmtion durmg
the evolutron of volatrle materials prior to the
final ashmg stage at 450°C The long charrmg
stage allows complete evolutron of the mtrogen
oxide gases
Interference study Among the mterference-
producmg elements mentioned, Fe and Sn can
occur m mussel at levels (1367 kg g-’ Fe, 14192
kg g-’ Sn) [35,36] that could potentrally interfere
wrth arsenic deternnnatlons m readmg reaction
flask solutions (2 /lg of Fe and 1 pg of Sn) [311
Therefore, it was decided to carry out an mterfer-
ence study of Fe and Sn m the mstrumental
reading of arsenic by comparing calibratron graphs
for arsemc without Fe and Sn with calibration
graphs obtamed usmg different amounts of Fe (0,
5, 10 and 15 pg) correspondmg to 0, 2000, 4000,
and 6000 pg g-l of sample, or Sn (0, 1,2 5 and 5
pg) correspondmg to 0, 200, 1000 and 2000 pg
g-’ of sample (dry weight) The statrstrcal method
used to estimate interference from Fe and Sn was
multiple linear regression analysrs performed by a
BMDPlR program [37] The resulting multtple
hnear regression models showed, m both m-
stances, a highly slgmficant relationship only be-
tween the concentratron of arsemc and ab-
sorbance (Tables 5 and 6) However, the other
regressors considered, the concentrattons of Fe
and Sn and then squares and mteractlons with
the concentration of As (Fe As and Sn As), did
not show a slgmfrcant relationship with ab-
sorbance These factors seem, therefore, to have
had no effect on absorbance, and it can be con-
firmed that levels 5-fold higher for Fe and 1Cfold
hrgher for Sn than the maxmum levels detected
m mussel have no effect on the arsemc signal
Stmdarly, no matrtx effect was detected, as the
slopes of the cahbratron graphs obtamed by the
standard addition method and the standards did
not show significant differences
Moreover, it was verified that background cor-
rection wrth a deutenum arc IS not necessary, as
samples, standard, reagent blanks and standard
blanks showed the same background signal
Analyttcal charactenstm of the HG-AAS
method To check the vahdity of the HG-AAS
method and to compare rt wrth the PFZ-AAS and
DETERMINATION OF ARSENIC 67

TABLE 5
Interference of Fe m the As Instrumental readmg
Analytical curves
AS y=OO44+95x,r=O9991
As+5pgFe y-0045+95x,r-09994
As+ 1OpgFe y=OO43+10Ox,r=O9995
As+ 15pgFe y=OO63+93x,r=O9993

Multiple hnear regressIon analysis

Absorbance = K, + k&is] + k,[Fe] + k.&4# + k,[Fe]’ + k,[AsXFe]
Multiple correlation coefficient 0 9986
Standard error of estimate 0 0132
F-value 994 476, P < 0 Oool, (F, 14,0 01) = 4 69
Summary table for the regresslon a
Variable Coefficient Standard error T P (2-W)
[ASI 98647 0 5462 18 06 000
[Fe1 0 0021 0 0019 107 030
[Fel* -06800~10-~ 0 1176 x 1O-3 -058 057
L4.d* - 3 8178 8 1750 -047 0 65
~FelA.4 -00015 0 0244 -006 0 95
Intercept 0 03755
a T-t test for the coefkent b,/s@,), and the associate two-taded probability value Tolerance (1 - Rf) where the multiple R: IS
the correlation of Independent variables

TABLE 6
Interference of Sn m the As mstrumental reading
AnalytIcal curves
As y=O027+8Ox,r=O9993
As+lpgSn y=O028+8Ox,r=O9988
As-k25figSn y=0026+77x,r=09985
As+SpgSn y=O032+79r,r=O9988
Multiple lmear regresslon analysis
Absorbance = K, + k,[Asl + k,[Sn] + k&& + k,[Sn]* + k,[AsXSn]
Multiple correlation coefflclent 0 9992
Standard error of estimate 0 0079
F-value 1863 297, P < 0 0001, (F, 14, 0 01) = 4 69

Summary table for the regresslon a

Vanable Coefficient Standard error T P (2-tad)
IAs1 8 8914 0 3223 27 58 000
ISnl -00043 0 0037 -117 0%
EW 0 1011 x 10-Z 0 6614 x 1O-3 153 0 15
L4s12 - 15 8633 4 9058 -323 001
LWAsl -00129 0 0435 -030 077
Intercept 0 02247
a T-r test for the coefficient b,/s(b,), and the associate two-taded probabdzty value Tolerance (1 - Rf) where the multlple R,” 1s
the correlation of Independent vmables
68 N YEdfEZETAL

ICP-AES methods developed for the determma-
tlon of arsemc m mussel products, analytical
charactenstlcs such as sensmvlty, detection Imut,
prec1slon, recovery and accuracy were evaluated
m accordance with IUPAC recommendations [38]
The results obtained are summarized 111Table 7
Sensltlvlty was established from the mean value
of the slopes of the cahbratlon graphs and ex-
pressed m absorbance umts per pg g-’ The
sensitlvlty obtained was greater than that re-
ported m the literature for the determmatlon of
arsemc m fish by HG-AAS [lo,391
The detection Inmt, established as the arsenic
concentration m pg g-l (wet weight) of mussel
product which provides an absorbance reading
statlstlcally different from that of the blank, was
calculated by dlvldmg three times the standard
deviation of the absorbance readmgs of eight
reagent blanks by the sensltlvity, and takmg mto
account the sample mass and ddutlon employed
The detectron lnnlt obtamed was lower than the
mmnnum content m fish and shellfish reported m
the hterature (0 04 pg g-l> [8]
Method precision 1s expressed as the relative
standard devlatlon of eight independent analyses
of the same sample of mussel product The
method precision obtained was suitable for the
level of arsenic concentration m seafood prod-
ucts
The percentage recovery was evaluated at two
levels of arsenic concentration (2 8 and 10 pg
g-l) and was determined as R = (X- Y/Z> x
100, where X= concentration of arsenic m the
spiked sample, Y = concentration m the sample
and 2 = concentration added All concentrations
were expressed m 1.18g-’ of mussel The percent-
age recovery was very satisfactory and the confi-
dence semi-mterval at 95% probabdlty IS wlthm
the precision for the method
The values found m the analysis of a certified
sample, SRM 1566a Oyster Tissue, were m close
agreement with the certified values
Arsemc content m canned mussel products of
natwnal manufacture
Seventeen different national brands purchased
m the retall market were analysed 111triplicate by
HG-AAS The arsenic contents found are sum-
marlzed m Table 8 The mean concentration of
all samples was 2 4 + 0 2 pg g-l (wet weight) or
6 8 f 0 6 pg g-’ (dry weight) Only s1x samples
analysed were outside the confidence interval for
the mean value at a 95% confidence level (four
higher and two lower) No differences were de-
tected which were related to the use of enamelled
or plam tmplate, type of oil employed (the only
different nigredlent specified m the od sauces m
the samples analysed) or mussel size

TABLE 7
Analykal charactensk of methods
Method Sensitivity a DetectIon hmit Preclslon b, Recovery ’ Accuracy d,
kg g-l) RSD (%) (%o) found value
(rg g-‘1 (dry weight)
HG-AAS 1760 f 120 0 017 3 (18) 102 f 3 (3 0,2 8) 137f12
(n = 8) = 97 f 4 (1 8, 10)
PEZ-AAS 1840 k 360 08 2 (2 9) 100 f 6 (2 9,2 8) 131+12
(n=8je
ICP-AES 182& 32 01 3 (2 7) 99 f 3 (2 7,5) 141*11
(n = 6) e 98 f 1 (2 7, 10)
a For HG-AAS and PFZ-AAS expressed m absorbance units per pg g- 1 For ICP expressed as 1,/I,, where 1, 1s the net As
enusslon and I, the background enusSlon obtamed for a standard solution of 100 pg ml- ’ b Values m parentheses correspond to
the average contents, m pg g-‘, for the sample analysed c Recovery percentages expressed unth the confidence mterval at the
95% confidence level Values m parentheses are the average concentration m pg g- ’ of the unsplked samples and of the As
added d Confidence Interval at the 95% confidence level obtamed m four analyses of NIST SRM 1566a Oyster ‘ksue, certified
value = 14 0 f 12 pg g-’ e II = Number of independent analyses for each parameter except sensltwty and accuracy
DETERMINATION OF ARSENIC 69

Comparison of the three me&oak Wlfh Zeeman-effect correction, provides control

The chorce of method 1s dependent on a num-
ber of factors, mcludmg sample matm mterfer-
ence, sensltnnty, detectlon hmlt, preaslon, accu-
racy, workmg analysis range and speed of analy-
SlS
The three methods are susceptible to mterfer-
ence, which must be controlled. For mussels, the
HG-AAS method 1s free from interference be-
tween elements, as Its high sensltlvrty pemuts
dlution of samples before mstrumental analysis,
reducmg or even ehmmatmg the maJor@ of m-
terference effects [ll] by lowermg the concentra-
tlon of mterfermg elements m the reaction flask
[40] There IS also no mterference from organic
matter as this 1s completely elnnmated by dry
ashmg However, for the apphcatlon of this tech-
ruque to seafood products m general, It IS neces-
sary to control the levels of potentmlly mterfermg
elements (Cu, Fe, Mg, Nl, Sb, Se and Sn) and
confum expernnentally that these do not exceed
the concentrations estabhshed as producmg mter-
ference m the reactlon flask
The PFZ-AAS method [26], usmg a combma-
tlon of a stablhzed temperature platform furnace
of spectral mterference, while non-spectral mter-
ference IS mmlrmzed by means of a suitable
tnne-temperature optumzatlon of the program-
me and the use of the standard addltlon method
for the determmatlon of arsenic m mussels Thrs
technique 1s apphcable to seafood products pro-
vided that agnal suppresslon caused by matrix
Interference 1s controlled by use of the standard
addltlon method and suitable sample dllutlon
The ICP-AES method [27] for the determma-
tlon of arsenic m mussels 1s free from chermcal,
physical and spectral interferences at the wave-
length used (193 696 nm) However, If this method
were to be used for the determmatlon of arsenic
m seafood products, it would be necessary to
check for potential mterference from Cr, Fe and
Sn occurrmg m the instrumental readmg ahquot
at concentrations greater than 50 pg ml-l for Cr
and 200 pg ml-’ for Fe and Sn
The sensltlvltles of the HG-AAS and PFZ-AAS
methods are comparable ICP-AES 1s a less sensl-
tlve technique In this instance, though, the dry
ashmg procedure makes It possible to obtam an
acid solution 160 times more concentrated than

TABLE 8
Arsemc content m canned mussel products of natlonal manufacture
Charactenstw na Range Mean
of values fcSI b
hg g-l) (1Lgg-l)
Contamer
Plam tmplate 13 18-3 3 24kO2
enamelled tmplate 4 23-26 24+02
011
Olive 011 3 24-25 25*02
Vegetable d 14 18-33 24kO2
Sue
Very large
(6-8 pieces per 75 g) 1 26 26
Large
(8-12 pieces per 70 g) 6 20-25 24rtO2
Medmm
(12-16 pteces per 70 g) 6 18-26 22f03
Small
(18-u) pieces per 70 g) 2 24-27 26f18
Ungraded 2 26-33 30*45
p n = Number of samples b CSI = confidence semi-Interval at the 95% confidence level
70 N YBAhZFiTAL

m the other two methods, which compensates for
the techmque’s lack of sensltlvtty In this m-
stance, for 10 g of sample 25 ml of 10% <v/v>
HCl 1s used
The detection hrmt of the three methods pro-
vldes louts below the arsenic levels detected m
the analysis of canned mussel products of na-
tlonal manufacture They are also below the mean
levels found m a review of the literature for the
various groups of marme ammals, v1z, m pg 8-l
fish 5, lamehbranchs 4, cephalopods 10, gas-
teropods 6 and crustaceans 14 However, the
nummum concentration detected according to the
review of the literature (< 0 04 pg g-‘1 1s below
the detection lumts for the PFZ-AAS and ICP-
AES techniques, so that for the analysis of sam-
ples with these low levels the HG-AAS techmque
1s required
The prec1slons, recovenes and accuracies of
the three methods are comparable, and this was
confumed expenmentally by comparmg the re-
sults obtained by usmg the three methods for the
analysis m tnphcate of five real mussel product
samples and the certified sample A one-way
analysis of variance (ANOVA) of the data by a
BMDP7D program [37], to test the mean concen-
tratlon obtained by the three methods, showed no
significant differences (based on a slgmfxance
level of 99%) The statIstica information ob-
tamed 1s given 111Table 9
The analysis ranges of the three methods, un-
der the operatmg condltlons employed (sample
size and dllutlon), permit the determmatlon of
arsemc over a concentration range which extends
from the detection llrmt to the end of the linear
workmg range and are, expressed m pg g-’ of
sample for HG-AAS 0 017-12 0, for PFZ-AAS
0 8-12 8 and for ICP-AES 0 l- > 91 (the maxi-
mum content quoted m the literature) For HG-
AAS the analysis can be extended to 24 0 ppm by
increasing the dilution without noticeably affect-
mg the analytical characterlstlcs However, m
PFZ-AAS greater sample dllutlons than those
used m the method quoted make the analytical
characterlstlcs of the method unacceptable From
this one can deduce that the only technique which
covers almost the whole concentration range de-
tected m fish and shellfish 1s ICP-AES
The tune required for the PFZ-AAS technique
IS lengthy, despite automation, owmg to the need
to use the standard additions method, with an
analysis rate of 1 sample h-’ Semi-automated
HG-AAS requires considerable sample mampula-
tlon and constant attention on the part of the
operator, but is much faster than PFZ-AAS ICP-
AES, featuring measurement against a single

TABLE 9
Companson of HG-AAS, PFZ-AAS and ICP-AES by a one-way analysis of variance
Analysis of vanance table for means
Source Sum of squares DF Mean square F value Tall
probablhty
Method 6 5450 2 3 2725 0 15 0 8590
Error 1159 6184 54 214744
(Fw, 0 01) = 5 03
Group means
Parameter HG-AAS PFZ-AAS ICP-AES All groups combmed
Mean 4 847 4602 5 412 4 954
Standard devlatlon 4 709 4 539 4 653 4 563
Maximum 14 419 13 926 15 542 15 542
Mmlmum 1 952 1765 2 152 1765
Cases Included 19 19 19 57
DETERMINATION OF ARSENIC
standard and not requu-mg sample mampulatlon
or the method of addltlons, 1s the fastest of the
three
Conclwwm
The most smtable technique for the deten-
nation of arsenic 111seafood products 1s ICP-AEB,
for the followmg reasons the absence of physical,
chemical and spectral interference, as long as the
levels of Cr, Fe and Sn are below the interference
levels (which are difficult to reach as they are
high), its excellent preclslon, accuracy and wde
analysis range, allowmg arsenic determmatlon
over almost the whole concentration range de-
tected m seafood products, and the absence of
the need for sample mampulatlon by the opera-
tor, and consequent speed of analysis For sam-
ples with arsenic levels below the detection limit
of this technique (0 1 pg g-l), the alternative 1s
the use of HG-AAS, allowing the determination
of arsenic levels over 0 017 pg g-’ with excellent
precision and accuracy
Funds to carry out this work were provided by
the Corn&n Intermnusterlal de Clencla y Tec-
nologia (CICyTj, Project AL189-0521, for which
we are deeply indebted
REFERENCES
1 C F Jehnek and P E Cornelmssen, En~ron Health Per-
spect, 19 (1977) 83
2 Mmlstry of Agnculture, Fishenes and Food, Food Surved-
lance Paper No 8, H M Stationary Office, London, 1982
3 W van Dokkun, R H de Vos, T Muys and J A Wesstra,
Br J Nutr , 610989) 7
4 H Yamauchl and Y Yamamura, Jpn J Pub1 Health, 27
(1980) 647
5 D Palermo, Arch Vet Ital, 30 (1979) 103
6 I Petrovlc and M Katalenuc, Hrana Ishrana, 22 (1981)
115
7 K Shlorm, A Slunagawa, H Yamanaka and T K&u&i,
Bull Jpn Sot Sa Fish ,49 (1983) 79
8 A F Mxho and M V H Tomotake, Rev Farm Bmqmm
Umv Sao Paula, 24 (1988) 128
9 P de Bat&&, G &ala and G Salzano, Zooprofdassi, 27
(1972) 405
10 J A Flonno, J W Jones and S G Capar, Anal Chem ,48
(1976) 120
11 CT Hu, S H Yu and Y T WeI, m H Cluba, M FujimaJu,
K Iwa~, H Mltsuda and Y Morlta (Eds ), Proceedmgs of
the F&h InternatIonal Congress of Food Scxence and
71
Technology, Kyoto, Japan, 17-22 September, 1978, Else-
ner, Amsterdam, 1979
12 W Evans, F J Jackson and D Dellar, Analyst, 104 (1979)
16
13 H Agerman and R Thomson, Analyst, 105 (1980) 902
14 J J Sanchez Saez, M D Santos Dlaz and M E Cirugeda-
Delgado, Boletin de1 Centro Naaonal de Ahmentaa6n y
Nutn&n, 6-8 (1981) 70
15 G KH Tam and G Lacrolx, J Assoc Off Anal Chem,
65 (1982) 647
16 K Jm, H Ogawa and H Taga, Bunseltl Kagaku, 32 (1983)
El71
17 WA Maher, Talanta, 30 (1983) 534
18 R E Sturgeon, S N W&e and S S Berman, J Anal At
Spectrom , 1 (1986) 115
19 I Santa Mana, M Gonzalez, W Lara and A Ober, Bull
Environ Contam Toxlcol ,37 (1986) 593
20 J W Hershey, T S Oostdyk and P N Kehher, J Assoc
Off Anal Chem, 71 (1988) 1090
21 M Lopez-Artlguez, ML Soria and M Repetto, Bull
EnvEon Contam Toxxol , 42 (1989) 634
22 W Brumbaugh and M J Walther, J Assoc Anal Chem,
72 (1989) 484
23 B Welz and G Schlemmer, J Anal At Spectrom, 1
(1986) 19
24 M Hoemng and P Van Hoeyweghen, Int J Environ
Anal Chem , 24 (1986) 193
25 W Dabeka and MA Lacrour, J Assoc Off Anal Chem ,
70 (1987) 866
26 N Yhaiiez, ML Cervera, R Montoro and M de la
Guarcha, J Anal At Spectrom, 6 (1991) 379
27 M L Cervera, Doctoral Dissertation, Umversity of Valen-
aa, Valenaa, 1991
28 P W J M Boumans, Inductively Coupled Plasma Emission
Spectroscopy, Wdey, New York, 1987
29 S Nakasluma, R E Sturgeon, S N W&e and S S Berman,
Analyst, 113 (1988) 159
30 H Matuslewxz, RE Sturgeon and S S Berman, J Anal
At Spectrom , 4 (1989) 323
31 KS Subramaman and J C Meranger, Analyst, 1Cn (1982)
157
32 ML Cervera, A Navarro, R Montoro and R Catala, At
Spectrosc ,5 (1989) 154
33 ML Cervera, A Navarro, R Montoro, C Catala and N
Ybafiez, J Assoc Off Anal Chem, 2 (1989) 282
34 T T Gorsuch, The Destruction of Organic Matter, Perga-
mon, Oxford, 1st edn , 1970
35 D J H Plulhps, Mar Biol , 46 (1978) 147
36 J Ezquerra and A Martm, AlimentarIa, XV/90 (1978) 55
37 W J Duron, M B Brown, L Engelman, MA Hdl and R I
Jennnch, BMDP Statistical Software Manual, Umvewty
of Cahforma Press, Berkeley, CA, 1988
38 H M NH Irvmg, H Freiser and T S West, Intematlonal
Umon of Pure and Applied Chemistry Compendmm of
Analytical Nomenclature, Pergamon, oxford, 1978
39 M Inhat and H J Miller, J Assoc Off Anal Chem, 60
(1977) 813
40 B Welz and M Melcher, Analyst, 109 (1984) 573