You are on page 1of 10

Analytical

Methods
View Article Online
PAPER View Journal
Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

Performance improvement and single laboratory


validation of classical qualitative methods for the
Cite this: DOI: 10.1039/c5ay01822f
detection of adulterants in milk: starch, chlorides
and sucrose
Carina de Souza Gondim, Roberto Cesar Santos de Souza, Marina de Paula Penna e
Palhares, Roberto Gonçalves Junqueira and Scheilla Vitorino Carvalho de Souza*

Milk stands out among the foods that are subject to adulteration. Several studies have shown that milk
adulteration is a reality in many countries, despite the existing inspection routines. Furthermore, there is
an aggravating factor in that there are no reports in the literature concerning the validation of the
classical qualitative methods that are commonly used to detect fraud in milk and that are described in
the legislation. In the present study, a novel validation approach was applied on classical qualitative
methods related to the detection of the density restoratives starch, chlorides and sucrose in raw milk,
considering the official and modified versions. Rates, unreliability regions, detection limits, accordance,
Received 13th July 2015
Accepted 5th October 2015
concordance, selectivity and robustness were estimated. Although simple, the proposed modifications
significantly increased the efficiency of the methods, generating a lower frequency of errors or false
DOI: 10.1039/c5ay01822f
results and a reduction in the detection limits and uncertainty ranges, consequently enhancing the
www.rsc.org/methods capability of detecting adulterations.

Although they are published in the legislation of several


1. Introduction countries,8,9,14,15 validation studies of the qualitative methods
The foods or ingredients that are most likely to be targets for for detecting adulterants in milk are not reported in the litera-
adulteration include those that are of a high value and undergo ture. Some works involve a limited evaluation of classical
a number of processing steps before reaching the consumer methods for the detection of density restoratives in milk.16
market.1 This context highlights why milk and milk products In several sectors of the food chain, the importance of
are so vulnerable to several types of adulterations.2 Milk can be qualitative methods has increased. These methods have oen
adulterated with water and then mixed with starch, salt or been selected for screening and can be implemented in labo-
sugar.3 These density restoratives are one class of substances ratories of different sizes and personnel qualications because
that are more frequently used in milk fraud.3,4 they provide objective and rapid results, with low costs and
Recent studies reported adulterations in milk despite exist- simplicity of use.17,18 Despite recent publications focusing on
ing legislation and inspections by federal and state agencies, the proposal for harmonized and detailed protocols, validation
and they still represent a challenge to the productive sector and of qualitative methods is still a critical issue for food analysis
regulatory bodies.2,5–7 laboratories.18–21
Most of the quantitative methods related to the quality All the work required to maintain a framework for moni-
control of milk are normalized. However, the qualitative toring food quality should be based on the degree of reliability
methods for the detection of adulterations in milk are not of the analytical methods.22 Thus, in the dairy industry,
normalized.8–10 Chemometric strategies and infrared spectros- adequate control methods are required to evaluate the
copy have been jointly used to detect different adulterants in authenticity of milk and milk products.11 The complexity, vari-
milk.4,11–13 However, the classical qualitative tests are still widely ability and instability of the milk matrix in association with
used to detect milk adulteration by official laboratories and a lack of expertise in analyzing milk products are important
dairy industries. aspects that can highly affect the reliability of the analytical
control, even if consolidated methods are adopted.23
This paper presents the optimization and application of
Federal University of Minas Gerais (UFMG), Faculty of Pharmacy (FAFAR), Department a novel validation approach in the tness for purpose assess-
of Food Science, Av. Antônio Carlos, 6627, Campus da UFMG, Pampulha, 31270-010, ment of classical qualitative methods related to the detection of
Belo Horizonte, MG, Brazil. E-mail: scheilla@bromatologiaufmg.com.br; Fax: +55-31- starch, chlorides and sucrose in raw milk by considering both
3409-6908; Tel: +55-31-3409-6918

This journal is © The Royal Society of Chemistry 2015 Anal. Methods


View Article Online

Analytical Methods Paper

versions as described in the legislation and with modications solution). The heating causes the opening of the helical struc-
for performance improvement. ture of starch, allowing for the adsorption of iodine and the
development of a characteristic blue color aer cooling.9 The
2. Experimental identication of chlorides was based on a reaction with silver
nitrate in the presence of a potassium chromate indicator. The
2.1 Samples development of the yellow color indicated a positive result or
Raw cow milk samples (3 L) were obtained from the experimental the presence of chloride in amounts greater than the normal
farm of the Ministry of Agriculture Livestock and Supply/Veterinary range in milk.9 Sucrose was detected in milk through the
Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

School at the Federal University of Minas Gerais State by milking development of a deep-red color that resulted from the
20 different cows that were managed under strictly controlled condensation of ketoses with resorcinol in an acid medium.25
livestock conditions. These samples were homogenized, packaged
in polypropylene gallons and kept under refrigeration (2 to 7  C) 2.4 Experimental design
until the beginning of the experiments, which were performed
The optimization and validation approach, performed as
within a maximum period of 48 hours. Each sample was analyzed
described by Gondim et al.,18 included eight analytical batches
for its density, freezing point and titratable acidity, with the aim of
for each method and their respective versions (official and
checking its consistency with regulated ranges.24
modied) (Fig. 1):
Sample collection was performed for the eight analytical
(i) Optimization;
batches involved in the optimization and validation of each
(ii) Validation – preliminary tests;
method (starch, chlorides and sucrose) by considering both
(iii) Validation – study of rates, uncertainty or unreliability
versions (official and modied) (Fig. 1).
regions (UR), detection limits (DL), accordance (ACO) and
concordance (CON) – run 1;
2.2 Chemicals and equipment
(iv) Validation – study of rates, UR, DL, ACO and CON – run 2;
All reagents used were of analytical grade. Carboxymethylcel- (v) Validation – study of rates, UR, DL, ACO and CON – run 3;
lulose, sodium chloride and starch were purchased from Vetec (vi) Validation – study of selectivity in the presence of known
Quı́mica Fina Ltd. (Rio de Janeiro, RJ, Brazil). Sucrose and interferences;
pectin were supplied by Synth (Diadema, SP, Brazil) and xan- (vii) Validation – study of robustness; and
than gum and maltodextrin by Sigma-Aldrich Co. Ltd. (Gil- (viii) Validation – tness for purpose assessment.
lingham, Dorset, UK). Iodine, potassium iodide, silver nitrate,
potassium chromate, resorcinol and hydrochloric acid were
2.5 Optimization
purchased from FMaia (São Paulo, SP, Brazil), Vetec Quı́mica
Fina Ltd. (Rio de Janeiro, RJ, Brazil) and Synth (Diadema, SP, The optimization steps were conducted to test changes in the
Brazil). Puried water was obtained using a Milli-Q Direct analytical procedures of the methods described in legislation
system (Billerica, MA, USA). aiming to reduce false results and increase the sensitivity and
The equipment was calibrated, including water bath (314- selectivity of the methods.
8DN, Nova Ética), micropipets (LM 10000, HTL Lab Solutions; Raw milk samples were spiked with starch levels between
Finnpipette F3, Thermo Scientic; and Transferpette S, Brand), 0.01 and 1.2 g L1, between 0.92 and 2.27 g L1 for chlorides
weighing machines (AUX 220, Shimadzu; and BK 300, Gehaka), (considering the native concentrations determined9); and
freezing point equipment (PZL 900, PZL), milk densimeter (5784, between 0.3 and 3.0 g L1 for sucrose, plus the blanks (unspiked
Incotherm) and thermometers (7665.02.0.00, Incotherm). samples).

2.3 Methods 2.6 Validation


The presence of starch in milk was determined aer heating the 2.6.1 Preliminary tests. Spiked and unspiked samples were
milk and adding an iodine/potassium iodide solution (Lugol's prepared and analyzed randomly in a single analytical batch,

Fig. 1 Representation of the raw milk sampling procedure used for the optimization and validation of the methods for detecting starch, chlorides
and sucrose. Method 1.1: official method for starch detection; method 1.2: modified method for starch detection; method 2.1: official method for
chloride detection; method 2.2: modified method for chloride detection; method 3.1: official method for sucrose detection; and method 3.2:
modified method for sucrose detection.

Anal. Methods This journal is © The Royal Society of Chemistry 2015


View Article Online

Paper Analytical Methods

with 10 replicates per level, blind to the analysts. These exper- potassium iodide (C and D) brands used in the preparation of
iments were started with the same concentration ranges Lugol's solution, and the heating time of the sample (8 and 12
described in Section 2.5. When necessary, new ranges were min), for a total of eight treatments.
tested to obtain positive results between 0 and 100%.18 Two factors and two levels were considered for the chloride
2.6.2 Rates, unreliability regions, detection limits, accor- and sucrose methods, for a total of four treatments. For the
dance and concordance. Raw milk samples were spiked with chloride method, potassium chromate (A and B) and silver
the analytes at different concentrations, with 30 independent nitrate (A and B) brands were used, and for the sucrose method,
replicates per level, plus the unspiked samples. The samples resorcinol brands (A and B) and the heating times for the
Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

were prepared and analyzed randomly in three different samples (3 and 7 min) were used.
analytical batches, which involved three different times and The factors and levels were considered to be critical for
analysts, with 10 replicates per level plus the unspiked samples, method performance when the RLR was less than 90%, corre-
which was blind to the analysts. This experiment covered the sponding to the possibility of obtaining a single false negative
conditions for repeatability and intermediate precision.18 result.18
Starch was added at 21 levels between 0.01 and 1.20 g L1, 2.6.5 Assessment of tness for purpose. Milk samples were
sodium chloride was added to produce 15 levels between 0.92 adulterated with 150 g L1 water.7 These samples were spiked
and 2.24 g L1 chlorides, and sucrose was added at 12 with each analyte at 3 concentrations, in 3 replicates, including
concentrations between 0.3 and 3.6 g L1. the following: unspiked samples, samples spiked at the DL of
The rates were calculated by using contingency tables. The the modied methods, and samples spiked at the DL of the
UR limits were estimated by non-linear regression, based on the official methods. The density was measured for each replicate
probabilities of 0.05 and 0.95 of having positive results. The DL and compared with the permitted range as described in the
was stated as the upper limit of the UR. For the ACC and CON legislation, or 1028 to 1034 g L1.24
values, calculated by equations based in combinatorial The tness for purpose of the method was considered
concepts, the acceptability criterion was stated as a minimum of satisfactory if the starch, sodium chloride and sucrose
0.8, which was outside the UR, given that this value corre- concentrations necessary to restore the density of the samples
sponded to the possibility of obtaining a single false negative that had been adulterated with water were equal to or greater
result in each analytical batch.18,26 than the DL of the methods.
2.6.3 Selectivity – interferents. Unspiked and spiked
samples at levels corresponding to the lower concentrations at 3. Results and discussion
which 100% reliability rates (RLRs) were obtained in the
previous step of the validation process were used for the selec- 3.1 Starch
tivity evaluation, with 10 independent replicates per level. These 3.1.1 Optimization. Considering the occurrence of false
concentrations were set as 0.3 and 1.0 g L1 of starch; 1.51 and negative results, a new criterion for positive results was
1.87 g L1 of chlorides; and 2.4 and 3.6 g L1 sucrose, respec- proposed (Fig. 2a.1). A greyish-green color was observed for
tively, for the modied and official methods. Samples were positive samples spiked at lower levels. This color was different
analyzed randomly in a single analytical batch, which was blind from the blue color observed for samples spiked at the upper
to the analyst.18 levels and from the yellow color formed for unspiked samples
Carboxymethylcellulose, xanthan gum, maltodextrin and (Fig. 3a). Then, the criterion suggested was the observation of
pectin were investigated as potential interferents. These a color different from that obtained for a non-blind unspiked
carbohydrates were spiked at a 0.12 g L1 level, which was the sample, as shown in Fig. 2a.2.
highest concentration at which the raw milk sample showed no 3.1.2 Validation
visually noticeable change. Starch, chlorides and sucrose were 3.1.2.1 Rates, unreliability regions, detection limits, accor-
also tested as potential interferents in other qualitative dance and concordance. The preliminary tests indicated the
methods, and set at 1.0, 1.87 and 3.6 g L1, respectively, to range of starch concentrations between 0.01 and 1.2 g L1 for
correspond to the lowest level with a 100% RLR for the official the subsequent steps of the validation study, i.e., provided 0%
methods. and 100% of the positive results for both methods.
The studied compound was designated as an interferent if For both studied criteria, the analysis of unspiked samples
the RLR was less than 90%, corresponding to the possibility of resulted in a 100% selectivity rate (SLR) and a 0% false positive
obtaining a single false negative result.18 rate (FPR), indicating that the matrix components did not
2.6.4 Robustness. Full factorial experiments were per- interfere in the SLR. Considering the official method as well as
formed. For each experiment, the analyte in question was added the modied method, a trend of larger percentages of false
to raw milk samples at concentrations corresponding to the results was evidenced by the lower concentrations. For the
lowest levels in which RLRs of 100% were obtained for each official method, 100% false negative rates (FNRs) were ob-
method, with 10 independent replicates per treatment. The tained at levels from 0.01 to 0.2 g L1, and no false negative
samples were analyzed randomly in a single analytical batch, results were observed for levels corresponding to starch at 1.0,
and they were blind to the analyst.18 1.1 and 1.2 g L1. For the modied method, 100% FNRs were
For the starch detection method, the experiment involved achieved only at the 0.01 g L1 starch level, and false negative
three factors and two levels as follows: the iodine (A and B) and results were not observed from the 0.3 g L1 level (Table 1).

This journal is © The Royal Society of Chemistry 2015 Anal. Methods


View Article Online

Analytical Methods Paper


Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

Fig. 2 Flowchart of the qualitative analytical methods for starch (a), chlorides (b) and sucrose (c) detection with 1 being the official method and 2
being the modified method (modified step marked in gray).

These ndings demonstrated that the proposed modication The performance curves are shown in Fig. 4a.1 and a.2. In
reduced the false negative errors, enhancing the sensitivity of considering the blue color observation criterion, the UR was
the method. estimated to be between 0.2 and 0.8 g L1 starch. A lower UR

Anal. Methods This journal is © The Royal Society of Chemistry 2015


View Article Online

Paper Analytical Methods


Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

Fig. 3 Colors observed in the optimization of the method for detecting starch, chlorides and sucrose in milk. (a) 1 – milk sample without addition
of Lugol's solution: observation of a white color (color index: 255r 255g 255b); 2 – the unspiked sample: observation of a yellow color (color
index: 255r 246g 133b) with the Lugol's solution; 3 – a spiked sample at 0.3 g L1: observations of a greyish-green color (color index: 208r 197g
151b or 230r 217g 165b) different from the blue color and from that obtained for the unspiked samples; and 4 – a spiked sample at 1.2 g L1:
observation of a blue color (color index: 107r 120g 164b) as established in the official method. (b) 1 – unspiked sample with a native chloride
concentration of 0.9 g L1: observation of brick-red color (color index: 200r 80g 60b); 2 and 3 – spiked samples at 1.0 and 1.1 g L1 chlorides:
observation of intermediate colors (color indexes: 250r 168g 95b and 254r 207g 140b); 4 and 5 – spiked samples at 1.2 and 1.5 g L1 of chlorides:
observation of yellow color (color index: 255r 246g 133b). (c.1) 1 – unspiked sample: observation of white color (color index: 255r 255g 255b); 2, 3
and 4 – spiked samples at 0.3 1.2 and 2.1 g L1 sucrose: observation of intermediate colors (color indexes: 225r 163g 128b, 201r 116g 105b and
245r 132g 102b); and 5 – spiked samples at 3.0 g L1 of sucrose: observation of pinkish color (color index: and 220r 86g 66b). (c.2) 1 – unspiked
sample: observation of a colorless liquid; 2 and 3 – spiked samples at 0.3 and 1.2 g L1 sucrose: observation of intermediate colors (color indexes:
142r 107g 75b and 117r 73g 48b); and 4 and 5 – spiked samples at 2.1 and 3.0 g L1 sucrose: observation of deep-red color (color index: 95r 46g
32b and 100r 35g 17b).

was achieved for the comparison criterion with the unspiked 3.1.2.2 Selectivity – interferents. The methods were regarded
sample, ranging from 0.02 to 0.2 g L1. In addition to having as selective for the potential interferents carboxymethylcellu-
presented a signicant reduction in the DL value (approxi- lose, xanthan gum, pectin, sodium chloride and sucrose, having
mately four times), the narrower UR demonstrated that the obtained 100% of negative and positive results for samples
proposed modication, although simple, also contributed to an without and those spiked with starch, respectively. A lack of
improvement in the uncertainty. selectivity was observed for maltodextrin. This could be because
Generally, adequate precision was observed for both the starch had not been fully hydrolyzed during maltodextrin
methods because the ACC and CON values outside the UR were production30–32 or because of the capability of these methods to
higher than 0.8, except for one value of ACC at the 0.9 g L1 detect the products of incomplete starch hydrolysis. This
level, when using the official method. For the unspiked second alternative could be investigated because it would allow
samples, these parameters were estimated as 1 (Table 1). an expansion of the scope of these methods for the detection of
Silva16 studied the starch detection method in pasteurized starch and its derivatives.
milk with 27 assays. This author observed 100% negative results 3.1.2.3 Robustness. Robustness was demonstrated for the
for unspiked samples analyzed in three replicates. Positive reagent brands and heating time for both methods because
results of 100% were achieved when this sample was split, and 100% positive results were obtained for each treatment, i.e.,
the samples were spiked at eight levels between 0.1 and 25 g L1 there was no change in the RLRs.
and analyzed in three replicates. In the present study, no posi-
tive results were obtained for the 0.1 g L1 level when the official
method was applied. When the modied method was consid- 3.2 Chloride
ered, a 53.3% rate of positive results was achieved. However, it is 3.2.1 Optimization. When the chloride content in a milk
important to highlight the importance of the representative- sample is considered normal, part of the silver nitrate reacts
ness, number of samples and replicates in studies involving with chloride to form silver chloride, and the excess silver
qualitative methods; and considered in the present study.18,27,28 nitrate reacts with potassium chromate, forming silver chro-
The qualitative method for starch detection in milk mate with a brick-red color. When the chloride content is high,
proposed by India-Ministry of Health and Family Welfare of the consumption of silver nitrate will be greater. The quantity of
India,14 similar to the Brazilian official method, is described at silver chromate and the intensity of the brick-red color will
a DL of 1 g L1 that is consistent with the 0.8 g L1 estimated in consequently be low, and the yellow color of potassium chro-
the present study, considering the official procedure. It is mate will predominate.10
important to note that the 2005 valid version of the Indian In the rst analyses employing the official method, a brick-
manual has no data about the DL.29 red color was observed for the blank samples with the addition
Infrared spectroscopy techniques associated with chemo- of silver nitrate. However, this color faded depending on the
metric strategies were used to discriminate between milk agitation to which the sample was subjected, which led to false-
samples that were adulterated with starch when the concen- positive results.
tration of the adulterating solutions equalled or exceeded 1.5 Silver chromate has a solubility product constant (Ksp) lower
and 5 g L1.12,13 These new methodologies could detect adul- than that of silver chloride (1.2  1012 and 1.8  1010 mol
terated samples at levels signicantly higher than the DL esti- L1, respectively), but its solubility is approximately ve times
mated in this study for both classical methods. greater than that of silver chloride (6.5  105 and 1.3  105

This journal is © The Royal Society of Chemistry 2015 Anal. Methods


View Article Online

Analytical Methods Paper

Table 1 False negative rate (FNR), sensitivity rate (SNR), false positive rate (FPR), selectivity rate (SLR), accordance (ACC) and concordance (CON)
obtained for unspiked and spiked samples with starch, sodium chloride and sucrose, for both official and modified methods

Official method Modied method

Density ACC (analytical batch) ACC (analytical batch)


restorative FNR or SNR or FNR or SNR or
(g L1) FPRa SLRa 1 2 3 CON FPRa SLRa 1 2 3 CON
Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

Starch
0.0a 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
0.01 100.0 0.0 1.0 1.0 1.0 1.0 100.0 0.0 1.0 1.0 1.0 1.0
0.02 100.0 0.0 1.0 1.0 1.0 1.0 93.3 6.7 0.6 1.0 1.0 0.9
0.03 100.0 0.0 1.0 1.0 1.0 1.0 93.3 6.7 0.6 1.0 1.0 0.9
0.04 100.0 0.0 1.0 1.0 1.0 1.0 96.7 3.3 0.8 1.0 1.0 0.9
0.05 100.0 0.0 1.0 1.0 1.0 1.0 83.3 16.7 0.5 1.0 0.8 0.7
0.06 100.0 0.0 1.0 1.0 1.0 1.0 90.0 10.0 0.6 0.8 1.0 0.8
0.07 100.0 0.0 1.0 1.0 1.0 1.0 73.3 26.7 0.4 0.6 0.8 0.6
0.08 100.0 0.0 1.0 1.0 1.0 1.0 60.0 40.0 0.5 0.5 1.0 0.4
0.09 100.0 0.0 1.0 1.0 1.0 1.0 53.3 46.7 0.6 0.5 1.0 0.4
0.1 100.0 0.0 1.0 1.0 1.0 1.0 33.3 66.7 1.0 1.0 1.0 0.3
0.2 100.0 0.0 1.0 1.0 1.0 1.0 20.0 80.0 1.0 1.0 0.5 0.6
0.3 86.7 13.3 1.0 0.6 0.6 0.8 0.0 100.0 1.0 1.0 1.0 1.0
0.4 60.0 40.0 1.0 0.5 0.8 0.4 0.0 100.0 1.0 1.0 1.0 1.0
0.5 36.7 63.3 1.0 0.8 1.0 0.3 0.0 100.0 1.0 1.0 1.0 1.0
0.6 33.3 66.7 1.0 1.0 1.0 0.3 0.0 100.0 1.0 1.0 1.0 1.0
0.7 23.3 76.7 0.5 1.0 1.0 0.5 0.0 100.0 1.0 1.0 1.0 1.0
0.8 6.7 93.3 0.6 1.0 1.0 0.9 0.0 100.0 1.0 1.0 1.0 1.0
0.9 6.7 93.3 0.6 1.0 1.0 0.9 0.0 100.0 1.0 1.0 1.0 1.0
1.0 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
1.1 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
1.2 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0

Chloride
0.90 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
0.92 100.0 0.0 1.0 1.0 1.0 1.0 100.0 0.0 1.0 1.0 1.0 1.0
0.93 100.0 0.0 1.0 1.0 1.0 1.0 80.0 20.0 0.8 1.0 0.8 0.9
0.95 100.0 0.0 1.0 1.0 1.0 1.0 100.0 0.0 1.0 1.0 1.0 1.0
0.96 100.0 0.0 1.0 1.0 1.0 1.0 60.0 40.0 0.6 0.8 0.8 0.8
1.02 100.0 0.0 1.0 1.0 1.0 1.0 50.0 50.0 0.4 0.5 0.4 0.5
1.14 90.0 10.0 0.5 1.0 1.0 0.8 50.0 50.0 0.5 0.4 0.5 0.5
1.26 70.0 30.0 0.4 0.6 0.6 0.6 20.0 80.0 0.8 0.6 0.5 0.7
1.39 60.0 40.0 0.5 0.5 0.6 0.5 10.0 90.0 0.8 1.0 0.5 0.5
1.51 10.0 90.0 1.0 0.6 0.8 0.8 0.0 100.0 1.0 1.0 1.0 1.0
1.63 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
1.75 10.0 90.0 0.8 1.0 0.6 0.8 0.0 100.0 1.0 1.0 1.0 1.0
1.87 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
1.99 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
2.11 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
2.24 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0

Sucrose
0.0a 6.7 93.3 1.0 1.0 0.6 0.9 0.0 100.0 1.0 1.0 1.0 1.0
0.3 93.3 6.7 1.0 0.8 0.8 0.9 100.0 0.0 1.0 1.0 1.0 1.0
0.6 80.0 20 1.0 0.5 0.6 0.6 90.0 10.0 1.0 1.0 1.0 1.0
0.9 53.3 46.7 0.5 0.5 0.5 0.5 66.7 33.3 1.0 0.6 0.8 0.8
1.2 50.0 50.0 0.5 0.5 0.4 0.5 30.0 70.0 0.5 0.5 0.5 0.6
1.5 23.3 76.7 0.6 1.0 0.4 0.6 6.7 93.3 0.4 0.8 0.5 0.6
1.8 36.7 63.3 0.5 0.5 0.4 0.5 10.0 90.0 0.8 1.0 0.8 0.9
2.1 30.0 70.0 0.5 0.8 0.5 0.6 6.7 96.7 0.8 0.8 0.8 0.8
2.4 23.3 76.7 0.8 0.6 0.5 0.6 0.0 100.0 1.0 0.8 1.0 0.9
2.7 16.7 83.3 1.0 0.8 0.5 0.7 0.0 100.0 1.0 1.0 1.0 1.0
3.0 10.0 90.0 1.0 1.0 0.5 0.8 3.3 96.7 1.0 1.0 1.0 1.0
3.3 3.3 96.7 1.0 1.0 0.8 0.9 0.0 100.0 0.8 1.0 1.0 0.9
3.6 0.0 100.0 1.0 1.0 1.0 1.0 0.0 100.0 1.0 1.0 1.0 1.0
a
FPR and SLR for unspiked samples, i.e., 0 g L1 density restorative; number of replicates for each level: 30; criterion for ACC and CON evaluation:
minimum of 0.8 outside the unreliability region (marked in bold).

Anal. Methods This journal is © The Royal Society of Chemistry 2015


View Article Online

Paper Analytical Methods


Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

Fig. 4 Performance curves with their respective experimental data (A), unreliability regions and detection limits for starch (a), chlorides (b) and
sucrose (c) for 1 – the official method and 2 – the modified method.

mol L1, respectively33). Thus, by adding silver nitrate to an which were above the normal range. However, intermediate
aqueous medium with a low chloride content, the silver chro- colors were observed even for 1.0 g L1 to 1.1 g L1, which
mate can be solubilized aer manually shaking the test tube, were concentrations within and above the range stated as
making it difficult to visualize the characteristic color of this normal for milk, respectively. Considering these ndings, it
salt, and then giving false positive results. was not possible to suggest a new criterion for positive
A full factorial experiment was designed with two factors and samples (Fig. 3b.1 and b.2).
three levels for each one, including the following: the type of
movement (pendular, circular and inversion) and the number
of movements (10, 20 and 30), with triplicates for each treat- Table 2 Evaluation of different types and numbers of movements for
ment. The experimental design was set up with an analysis of the manual homogenization of test tubes, when considering unspiked
and spiked (2.27 g L1 of chlorides) samplesa
the blank samples and samples with 2.24 g L1 of chlorides and
the higher concentration of this analyte was chosen for the Unspiked Sample spiked at 2.27
optimization study. Homogenization with 30 pendular move- sample g L1 of chlorides
ments was considered the best condition, i.e., it did not lead to Type of Number of
movement movements R1 R2 R3 R1 R2 R3
false or inconclusive results (Table 2).
False negative results were also obtained for some spiked Circular 10 I I I P P P
samples, when an intermediate color was observed, different 20 I I I P P P
from both characteristic colors yellow and brick-red. Subse- 30 I P I P P P
quent studies were performed to dene the expected color for Pendular 10 N N N I I I
20 N N N P P I
samples containing chloride within and above the range 30 N N N P P P
considered normal for milk (0.8 to 1.0 g L1), in triplicate. The Inversion 10 P P P P P P
unspiked sample, with a native chloride concentration of 0.9 g 20 P P P P P P
L1, was spiked with sodium chloride solutions to obtain six 30 P P P P P P
nal concentrations between 1.0 and 1.5 g L1. a
R: replicate; N: negative result (brick red color characteristic of silver
The characteristic brick-red color appeared in the chromate – color index: 200r 80g 60b); I: inconclusive result (color
unspiked sample. The yellow color was observed for all index: 250r 168g 95b or 254r 207g 140b); P: positive result (yellow
color – color index: 255r 246g 133b).
samples containing chlorides between 1.2 and 1.5 g L1,

This journal is © The Royal Society of Chemistry 2015 Anal. Methods


View Article Online

Analytical Methods Paper

In the official method, potassium chromate solution is the official Brazilian method. A DL of 0.2 g L1 sodium chloride
added, followed by silver nitrate solution and homogenization. (0.12 g L1 chlorides), which was lower than the amount esti-
Thus, silver chromate precipitation can occur even for milk mated in the present study, was reported. However, the exper-
samples containing high levels of chlorides, generating a brick- iments or criteria adopted for the estimation of the limit were
red color, which can lead to false-negative results. not described.
With the aim of improving the sensitivity of the method and 3.2.2.2 Selectivity – interferents. The selectivity was observed
ensuring that all the chlorides present in the sample will for carboxymethylcellulose, xanthan gum, maltodextrin, pectin,
precipitate as silver chloride before the reaction between the starch and sucrose for both chloride detection methods with
Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

chromate and the silver nitrate, an inversion in the solution RLRs of 100%, considering the unspiked and spiked samples
addition sequence was proposed, as described by Tronco.10 with sodium chloride in the presence of these compounds.
Furthermore, a manual homogenization step aer the addition 3.2.2.3 Robustness. In a robustness assessment of the offi-
of the rst solution was suggested, considering the standard- cial method, brand B of silver nitrate was indicated as a factor
ized conditions (Fig. 2b.1 and b.2). that affected the performance of the method because RLRs of
3.2.2 Validation 80% were obtained with the experiments involving this brand.
3.2.2.1 Rates, unreliability regions, detection limits, accor- When considering the experiments performed with brand A,
dance and concordance. The range of chlorides from 0.92 to 2.24 RLRs $ 90% were observed. For the modied method, the
g L1 tested in the preliminary tests was used in the validation robustness was demonstrated for all factors and levels studied
process. when RLRs $ 90% were obtained in all treatments.
In the analysis of unspiked samples, SLR and RLR were null
for both methods. With respect to spiked samples, the FNR
ranged from 100 to 0%, resulting in sensitivity rates (SNRs) 3.3 Sucrose
between 0 and 100%, which showed higher FNRs for the lowest 3.3.1 Optimization. The results obtained for the unspiked
concentrations, and also for both methods. However, FNRs of samples indicated a lack of selectivity for this method, with an
0% were observed for the official method at levels between 1.87 FPR of 10%. FNRs between 100 and 50% were also observed in
and 2.24 g L1 chlorides, and for the modied method, the all studied levels.
same FNR value was found over a wider range (from 1.51 g L1). When heated with acid, the poly- and oligosaccharides are
These results indicated that the proposed modications hydrolyzed to simpler sugars and dehydrated to furfural and
contributed to an increase in the sensitivity of the method by derivatives. Ketoses are more rapidly dehydrated. Dehydrated
the reduction of false negative errors (Table 1). ketose reacts with resorcinol producing a deep red color.
The prole of the performance curves showed that the Aldoses may react and produce a faint pink color. The method
changes suggested have also led to an improvement in the described in the Brazilian legislation (Fig. 2c.1) therefore
parameter uncertainty and DL (Fig. 4b.1 and b.2). Considering indicated the wrong principle and criterion for positive
the official method, the UR was estimated to be between 1.08 results, i.e., the reaction of the resorcinol with aldoses and the
and 1.65 g L1, and for the modied method, the UR was formation of a pinkish color, respectively. These misconcep-
between 0.8 and 1.42 g L1 chlorides. The DLs were estimated to tions could explain the occurrence of false positive results
be 1.65 and 1.42 g L1 for the official and modied methods, when the described criteria were applied. The observed
respectively. These limits were near the 1.6 g L1 chlorides that pinkish color was likely formed from the reaction of aldoses
represented a theoretical limit based on the stoichiometry of with resorcinol, which was derived from the hydrolysis of
the reaction, considering the volume (4.5 mL) and the concen- lactose.
tration (0.1 mol L1) of the silver nitrate solution. The procedure described by India-Ministry of Health and
The standardization of both procedures for chloride detec- Family Welfare of India29 was examined and modications
tion was evidenced by the satisfactory values (greater than or described in Fig. 2c.2 were proposed based on this method. It is
equal to 0.8) for ACC and CON at levels outside the UR, except at important to consider that the nal resorcinol concentration
1.75 g L1 for the official method. was increased in the modied method, with the aim of
Silva16 evaluated the sensitivity of the chloride detection improving the sensitivity.
method in pasteurized milk samples spiked with commercial In Fig. 3c.1 and c.2, the colors observed for milk samples
common salt at nine concentrations between 0.1 and 1 g L1. For containing sucrose concentrations from 0.0 to 3.6 g L1 for both
each level, an analysis was performed in three replicates. The methods are presented. In the official method, there is a slight
results demonstrated that this test was capable of detecting increase in the intensity of the pink color with the increasing
concentrations as low as 0.27 g L1 salt (0.16 g L1 chlorides). concentration of sucrose. This nding occurs because the color
Positive results were observed for 1.14 g L1 chlorides, when visualization is hampered by milk coagulation as a result of the
considering the official method. In the Silva16 study, the native addition of concentrated hydrochloric acid to the samples. In
chloride concentration was not considered. As mentioned previ- the modied method, the results can vary for unspiked samples
ously, this work did not account for the importance of the repre- between a colorless liquid and an intermediate slightly
sentativeness and number of replicates in qualitative methods. yellowish color, which does not characterize a positive result.
The chloride detection method described by India-Ministry The orange to deep-red color results could be observed in
of Health and Family Welfare of India14 was similar to that of samples containing approximately 1.2 to 3.6 g L1 sucrose.

Anal. Methods This journal is © The Royal Society of Chemistry 2015


View Article Online

Paper Analytical Methods

Then, the observation of a color from orange to deep-red was restore the density. These results indicated that greater levels of
suggested for reporting positive results, as shown in Fig. 2c.2. these adulterants would be necessary and detectable.
3.3.2 Validation For the official method, the density was restored for sodium
3.3.2.1 Rates, unreliability regions, detection limits, accor- chloride and sucrose when the samples contained density
dance and concordance. In the preliminary tests, the sucrose restoratives at the DL, indicating no need for greater amounts of
concentrations between 0.3 and 3.6 g L1 provided 0% and these restoratives. However, the analytes were detected at these
100% positive results for both methodologies and were selected levels.
for the next step of the validation process. These results indicated that the modied methods for
Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

In the analysis of blank samples, SLRs of 93.3 and 100% and detecting starch, chlorides and sucrose and the official method
FPRs of 6.7 and 0% were obtained for the official and modied for starch detection were able to detect milk fraud at concen-
methods, respectively. For the official method, these results trations lower than the levels required for masking fraud by
could be attributed to the reaction of aldoses, resulting in adding 150 g L1 water.
lactose hydrolysis, with resorcinol discussed above. In the For all adulterated samples no false results were obtained,
spiked samples, the percentage of positive results was enhanced considering the official and modied versions, conrming their
with the increasing concentration, for both methods. The FNR applicability.
and SNR ranged from 93.3 to 0% and 6.7 to 100% for the official
method, and from 100 to 0% and 0 to 100% for the modied 4. Conclusions
method (Table 1). The frequency of positive results was higher
for the modied method, indicating an improvement in the For the starch detection method, a new criterion for positive
sensitivity. results was suggested. The standardization of the manual
The URs were estimated between 0 and 3.2 g L1 and from homogenization and an inversion in the sequence of reagent
0.3 to 1.9 g L1 sucrose for the official and modied methods, addition were the proposed modications for the chloride
respectively (Fig. 4c.1 and c.2). The range of uncertainty was method. Considering the method used for sucrose detection,
reduced with the proposed modications because this param- a ltration step and the use of a resorcinol solution were
eter was estimated at a lower and narrower concentration range incorporated; moreover, a change was made in the criterion for
for the modied method. positive results.
The DLs were established as 3.2 g L1 for the official A performance improvement was observed for the modied
method and 1.9 g L1 for the modied method. The dra of methods proposed in the present study, specically with respect
the Indian manual14 describes a method with a DL of 1 g L1. to the reduction of false responses, uncertainty and detection
Liu, Ren, Liu & Guo4 suggested a method that could accu- limit.
rately detect sucrose in milk with a concentration as low as The main weak point of these methods is that they are based
0.5% (w/w), using a Fourier transform infrared spectroscopic on sensorial detection, depending on human senses to record
instrument. and interpret the response; so that the experience and training
With the exception of an ACC value obtained for unspiked of analysts may affect the results.
samples by the official method and despite the difficulty of Despite the current trend of replacing the classical qualita-
interpreting the results inherent in this method because of the tive methods with instrumental techniques, high performance
milk coagulation, the values presented in Table 1 for ACC and in detection of adulterants in milk was demonstrated, which
CON indicated a satisfactory standardization of both methods, can signicantly impact on protecting the health and rights of
in terms of repeatability and intermediate precision conditions, consumers.
respectively.
3.3.2.2 Selectivity – interferents. In both the official and Acknowledgements
modied methods, the selectivity in the presence of the density
restoratives (carboxymethylcellulose, xanthan gum, maltodex- The authors would like to acknowledge Otávio Augusto Mazzoni
trin, pectin, starch and sodium chloride) studied was observed, Coelho, Ronália Leite Alvarenga and Linda Lys Calderaro
with RLRs $ 90%. Lafayette for their assistance in the rst stages of the execution
3.3.2.3 Robustness. Robustness was demonstrated for the of this project; the experimental farm of the Ministry of Agri-
official method for all the studied factors and levels (100% of culture Livestock and Supply (MAPA)/Veterinary School at the
RLR). For the modied method, RLRs of 60% were obtained in Federal University of Minas Gerais State (EV/UFMG) for allow-
all treatments with 3 minutes of heating and the robustness of ing the use of their facilities and providing the raw cow milk
this method was then assured when the heating time varied samples; the Ezequiel Dias Foundation (Funed) for their
from 5 to 8 minutes. support of the density and freezing point analyses; and the
Brazilian agency CAPES for nancial support.

3.4 Complementary evaluation of tness for purpose


References
In considering the modied method, it was observed that aer
the fraud with 150 g L1 water, the addition of starch, sodium 1 R. Karouri and J. Baedemaeker, Food Chem., 2007, 102, 621–
chloride and sucrose at the respective DLs was not sufficient to 640.

This journal is © The Royal Society of Chemistry 2015 Anal. Methods


View Article Online

Analytical Methods Paper

2 S. S. Souza, A. G. Cruz, E. H. M. Walter, J. J. F. Faria, 18 C. S. Gondim, O. A. M. Coelho, R. L. Alvarenga,


R. M. S. Celeghini, M. M. C. Ferreira, D. Granato and R. G. Junqueira and S. V. C. Souza, Anal. Chim. Acta, 2014,
A. S. Santana, Food Chem., 2011, 124, 692–695. 830, 11–22.
3 M. Kartheek, A. A. Smith, A. K. Muthu and R. Manavalan, J. 19 S. L. R. Ellison and T. Fearn, Trends Anal. Chem., 2005, 24,
Chem. Pharm. Res., 2011, 3, 629–636. 468–476.
4 J. Liu, J. Ren, Z. M. Liu and B. H. Guo, Food Chem., 2015, 172, 20 R. Macarthur and C. Holst, Anal. Methods, 2012, 4, 2744–
251–256. 2754.
5 A. Afzal, M. S. Mahmood, I. Hussain and M. Akhtar, Pak. 21 Association of Analytical Communities, International
Published on 28 October 2015. Downloaded by UNIVERSITY OF NEBRASKA on 04/11/2015 23:09:42.

J. Nutr., 2011, 10, 1195–1202. Stakeholder Panel on Alternative Methods (ISPAM),


6 R. M. Balabin and S. V. Smirnov, Talanta, 2011, 85, 562– Guidelines for Validation of Qualitative Binary Chemistry
568. Methods, AOAC, Gaithersburg, 2013, version 12.3.
7 P. M. Santos, P. D. Wentzell and E. R. Pereira-Filho, Food 22 P. Gowik, J. Chromatogr. A, 2009, 1216, 8051–8058.
Anal. Method, 2012, 5, 89–95. 23 I. de Noni, Accredit. Qual. Assur., 2004, 9, 226–231.
8 Brasil-Portaria no 1 de 7 de outubro de 1981, http:// 24 Brasil-Instrução Normativa no 62 de 29 de dezembro de
www.agricultura.gov.br, (accessed April 2015). 2011, http://www.agricultura.gov.br, (accessed April 2015).
9 Brasil-Instrução Normativa no 68 de 12 de dezembro de 25 A. Nigam and A. Ayyagari, in Laboratory Manual in
2006, http://www.agricultura.gov.br, (accessed April 2015). Biochemistry: Immunology and Biotechnology, Tata McGraw-
10 V. M. Tronco, in Manual para inspeção da qualidade do leite, Hill, Education, New Delhi, part 2, 2007.
Editora da UFSM, Santa Maria, 4th edn, 2010. 26 S. D. Langtonn, R. Chevennement, N. Nagelkerke and
11 P. M. Santos, E. R. Pereira-Filho and L. E. Rodriguez-Saona, B. Lombard, Int. J. Food Microbiol., 2002, 79, 175–181.
Food Chem., 2013, 138, 19–24. 27 B. M. Simonet, A. Rı́os and M. Valcárcel, Anal. Chim. Acta,
12 L. G. Zang, X. Zang, L. J. Ni, Z. B. Xue, X. Gu and S. X. Huang, 2004, 516, 67–74.
Food Chem., 2014, 145, 342–348. 28 S. Cárdenas and M. Valcárcel, Trends Anal. Chem., 2005, 24,
13 B. G. Botelho, N. Reis, L. R. Oliveira and M. M. Sena, Food 477–487.
Chem., 2015, 181, 31–37. 29 India-Ministry of Health and Family Welfare of India,
14 India-Ministry of Health and Family Welfare of India, Manual of Methods of Analysis of Foods-milk & Milk
Manual of Methods of Analysis of Foods–Milk and milk Products, New Delhi, 2005.
products–Dra, http://fssai.gov.in/Portals/0/Pdf/15Manuals/ 30 N. Harnkarnsujarit, S. Charoenrein and Y. H. Roos,
MILK%20AND%20MILK%20PRODUCTS.pdf (accessed Carbohydr. Polym., 2012, 88, 734–742.
March 2015). 31 E. Franceschinis, A. C. Santomaso, A. Trotter and
15 IPQ (Instituto Português da Qualidade)-Leite: ensaios N. Realdon, Food Res. Int., 2014, 64, 711–717.
preliminares de análise-exame prévio (NP 467), 1979. 32 W. Kearsley and S. Z. Dziedzic, in Handbook of Starch
16 L. C. C. D. Silva, PhD thesis, Universidade Estadual de Hydrolysis Products and their Derivatives, Chapman & Hall,
Londrina, 2013, http://www.bibliotecadigital.uel.br/document/ London & New York, 1995.
?code¼vtls000185906, (accessed April 2015). 33 W. M. Haynes, in CRC Handbook of chemistry and physics:
17 A. Pulido, I. Ruisánchez, R. Boqué and F. X. Rius, Trends a ready reference book of chemical and physical data, CRC
Anal. Chem., 2003, 22, 647–654. Press, Boca Raton, 93rd edn, 2012.

Anal. Methods This journal is © The Royal Society of Chemistry 2015

You might also like