You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264120481

Detailed Validation of a Method for the Determination of Nitrate in Water by


UV/Vis Spectroscopy

Article  in  Journal of AOAC International · July 2014


DOI: 10.5740/jaoacint.12-007

CITATIONS READS
0 1,913

3 authors:

Vanessa Guimarães Manuel Azenha


Universidade de Trás-os-Montes e Alto Douro University of Porto
18 PUBLICATIONS   43 CITATIONS    59 PUBLICATIONS   943 CITATIONS   

SEE PROFILE SEE PROFILE

Helder Durão

2 PUBLICATIONS   2 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Beneficial effects of nano-SiO2 on the tolerance of barley exposed to high levels of nano-NiO and paracetamol View project

BPD/UTAD/INNOVINE&WINE/WINEMAKING/754/2016 (NORTE 2020) View project

All content following this page was uploaded by Vanessa Guimarães on 22 July 2014.

The user has requested enhancement of the downloaded file.


Guimarães et al.: Journal of AOAC International 2014  1

RESIDUES AND TRACE ELEMENTS

Detailed Validation of a Method for the Determination of


Nitrate in Water by UV/Vis Spectroscopy
Vanessa Guimarães
University of Porto, Faculty of Sciences, CIQ-UP, Department of Chemistry and Biochemistry, Rua do Campo Alegre 687, 4169-
007 Porto, Portugal; IDIT—Institute for Development and Technological Innovation, Rua do IDIT–Espargo, 4520-102 Santa Maria
da Feira, Portugal
Helder Durão
IDIT—Institute for Development and Technological Innovation, Rua do IDIT–Espargo, 4520-102 Santa Maria da Feira, Portugal
Manuel Azenha
University of Porto, Faculty of Sciences, CIQ-UP, Department of Chemistry and Biochemistry, Rua do Campo Alegre 687, 4169-
007 Porto, Portugal

N
A spectroscopic method is reported for the itrate levels in water may pose concerns to human and
determination of nitrate in water samples based on environmental health mainly due to its conversion to
the reaction of nitrate and concentrated hydrochloric nitrite, a hematoxic anion that compromises the oxygen-
acid. Nitrous acid is obtained with the subsequent transport function and that may also be implied in the generation
diazotization of sulfanilamide. In brief, the method of N-nitroso compounds, some of which are considered as human
introduces a simpler, faster, and cheaper nitrate carcinogens (1–2). Nitrate admissible concentration values
reduction step, as compared to metal reduction. (ACV) are thus established in the legislation for water pollution
Such described advantages prompted us to control. In Portugal, the ACV for drinking water and wastewater
undertake a validation and quality control study is 50 mg/L, the same as 11.3 mg/L nitrogen (N), which is in
aimed at the assessment of its fitness-for-use agreement with the European Union legislation (3–4).
The determination of nitrate in water is a task performed
as a routine laboratory procedure. The assessed
thousands of times daily in the world, mostly at routine
method exhibited a satisfactory performance under
accredited laboratories employing validated analytical
indirect evaluation for drinking and wastewater
methods. Comprehensive reviews (not limited to water matrixes
samples within the linear range between 0.5 to
or specific analytical techniques) of methods for nitrate
5 mg/L nitrogen (N), sensitivity of 0.095 a.u. mg/L,
determination can be found in the literature (5–6). A large
discrimination threshold at 0.06 mg/L N, LOD of diversity of methods is compiled there, relying on spectroscopic
0.05 mg/L N (estimated in repeatability conditions), (UV/Vis absorption, fluorescence, chemiluminescence, IR,
LOQ of 0.5 mg/L N (estimated in intermediate Raman, and molecular cavity emission), electrochemical
precision conditions), repeatability and intermediate (voltammetry and potentiometry) and chromatographic
precision coefficients were estimated to be below (HPLC and ion chromatography), or electrophoretic analysis.
10%. In the case of wastewater samples, due to Chemiluminescent and chromatographic techniques are
the observed significant matrix effects, the original envisaged as offering superior performance. However, the
procedure was modified. The dilution and simple instrumental complexity and cost required for their successful
spike recovery test were included into the method in implementation limit their widespread operation. The use of
order to decide the method’s applicability for each spectroscopic analysis appears to be the method of choice due
sample. A recovery within 90–110% is considered to the simplicity of the protocols and the wide availability of the
acceptable; while a recovery outside of that range instrumentation involved.
is not. Multiple standard addition quantification For routine laboratory work, the ability to provide reliable
recommended but not validated. Direct evaluation output with high-throughput and cost effectiveness is of utmost
of the method’s validation was accomplished by importance to be competitive in today’s market. Reference
successfully participating in a proficiency test and in methods for the determination of nitrate, established by several
a comparison with a standard method. The combined institutions [International Organization for Standardization,
uncertainty was estimated at a 7% maximum. (ISO); U.S. Environmental Protection Agency], assure the
The internal quality control provided positive potential reliability of such methods, which must be validated
indications pointing to long-term stable behavior within every individual laboratory. These methods are a
of the procedure. Therefore, the proposed method common choice in most labs, but any other method can be
constitutes a reliable, simple, rapid, and inexpensive used, provided that an acceptable validated performance can be
demonstrated. It is thus of great relevance for labs providing
procedure which may advantageously replace other
analytical services to be aware of potentially advantageous novel
methods of nitrate determination in routine work.
methods, but also to know in advance if these new proposals
Received January 5, 2012. Accepted by AK March 8, 2012. comply with strict quality assurance checks. In this context, we
Corresponding author’s e-mail: mazenha@fc.up.pt decided to carry out a validation study of a recently proposed
DOI: 10.5740/jaoacint.12-007 method (7) claiming to provide a simple, rapid, and inexpensive
2  Guimarães et al.: Journal of AOAC International 2014

procedure, and selective in the presence of nitrite, for the routine distilled water. Both stock solutions were preserved with
determination of nitrate in the range between 0.5–10 mg/mL N, 0.2% (v/v) chloroform.
in different aqueous samples. This is a spectroscopic method The sulfanilamide solution concentration was 0.5% (w/v)
based on the reaction of nitrate and concentrated HCl to obtain sulfanilamide in 0.5% (v/v) hydrochloric acid. The NEDA
nitrous acid (Equation 1), and the subsequent diazotization of solution concentration was 1% (w/v) NEDA in 1% (v/v)
sulfanilamide (Equation  2). The diazo-compound reacts then hydrochloric acid. The urea solution was 2% urea (w/v) in
with N-1-(naphthyl)ethylenediamine dihydrochloride (NEDA) distilled water.
(Equation  3) to create the diazo-dye compound which can The coloring reagent for nitrite determination was prepared
be analysed by UV/Vis spectroscopy, using a wavelength of by diluting 100 mL 85% phosphoric acid and 10 g sulfanilamide
540 nm. in 800 mL of distilled water. After complete dissolution, 1 g of
NEDA was added and then the solution was further diluted
HNO3 (aq) + 2HCl (concn) → Cl2 (g) + to 1 L with distilled water. The coloring reagent is somewhat
HNO2 (aq) + H2O (l) (1) unstable, so this solution was prepared every 15 days and was
always kept away from light.
Water samples from several different sources (cork industry,
metal finishing industry, car wash facilities, wood industry,
paint industry, textile industry, etc.) were collected, preserved,
(2) and stored according to standard procedures. Before the nitrate
determinations, wastewater samples were diluted in the region
of the lower end of the working range so as to minimize
interferences while working above the LOQ. If required, these
samples were subjected to nitrite removal. Unless otherwise
stated, the experiments were carried out at ambient temperature.

Nitrite Determination

Nitrite was determined in each sample, prior to the


application of the nitrate procedure. In this process, 0.2 mL
of coloring reagent was added to 5 mL sample, or to a portion
(3) diluted to 5 mL. Blanks and standards were treated using the
same procedure. After a period of 10 min (but before 2 h),
absorbance was measured at 543 nm using quartz cuvettes with
a 1 cm pathlength (9). Beer’s law was obeyed up to 180 µg /L N.

This method falls within the class of methods that require Nitrite Removal
reduction of nitrate to nitrite prior to analysis. Nitrite thus
becomes an obvious interferent that can be previously removed When nitrite concentration is greater than 200 µg /L N, it
by urea treatment (8). Basically, this proposed method introduces must be removed with urea. This treatment protocol was
a simpler, faster, and cheaper nitrate reduction step, as compared described previously for selective determination of nitrate
with the most popular technique of metal reduction. The next in the presence of nitrite (7). Each milliliter of nitrate sample
sections demonstrate, for the first time, that the proposed method was treated with 0.1 mL urea solution and 0.2 mL HCl, and
is in fact validated for the determination of nitrate in drinking incubated for 10 min in a boiling water bath. The nitrite- free
water samples. However, for wastewater samples, a sample- samples were, thereafter, processed for nitrate determination.
dependent behavior was found, some possible interferents were These samples were quantified against standards subjected to
identified, and modifications were incorporated in the procedure the nitrite removal treatment just described.
in order to reach an acceptable performance.
Nitrate Determination
Experimental
The method under study involves a sulfanilamide diazotization
Reagents and Solutions process, which must be carried out at a high temperature. To
accomplish this, all samples and standards were immersed in a
All reagents used were of analytical grade purity, and included boiling water bath, using heat-resistant 20 mL tubes. The tubes
sodium nitrite (Panreac, Cascais, Portugal), potassium nitrate were maintained in place using an improvised aluminum rack.
(Merck, Algés, Portugal), N-1-(naphthyl)ethylenediamine The water level in the water bath remained 2–3 cm above the
dihydrochloride (Riedel-de Haën, through Sigma-Aldrich, level of solution in the test tubes.
Sintra, Portugal), sulfanilamide (Himedia, Mumbai, India), The experimental procedure was as follows: 0.2 mL
hydrochloric acid 37% (Panreac), and urea (Panreac). Distilled sulfanilamide solution and 2 mL HCl were added to each
water was used for making solutions, dilutions, and for material milliliter of nitrite-free samples. The mixture was incubated
washing. in a boiling water bath for 3 min, cooled with cold water to
Working solutions of nitrate and nitrite were made by diluting room temperature, and 0.2 mL of coupling agent (NEDA)
stock solutions of 100 mg/L and 250 mg/L, respectively, in was added. The absorbance of the solution was measured
Guimarães et al.: Journal of aoaC international 2014 3

coefficients supported the acceptance of linearity for the


obtained calibration curve, the ISO 8466-1/2 standards (10, 11)
recommend testing linearity by taking the variances of the
2
residuals obtained when assuming the linear (S y/x) or the
2
non-linear (second order polynomial, S y2) models and then
2 2 2
Parameter Value calculating the difference ∆S = (N-2) * S y/x – (N-3) * S y/x.
2 2
This difference is then used for calculating F as ∆S /S y2 and
Slope (b) 0,0952
Sb 0,0004
Origin (a)
Sa
0,0005
0,001
finally this “experimental” F value is compared with the critical
Sy/x 1,48x10-3 Snedecor/Fisher value, Fcrit. In this case, F is 0.08, which is
R 0.9999
much smaller than Fcrit (7.7), thus the linearity is verified.

Sensitivity, LOD, and LOQ

Sensitivity, as given by the calibration slope, was


Figure 1. Averaged (n = 9) calibration curve obtained with the estimated as 0.095 a.u. mg/L (taken from the average line of
proposed method. 9 calibration curves as shown in Figure 1). The discrimination
threshold (DT) is a useful parameter that provides the lowest
Figure 1.

at 540 nm after 30 min standing, at room temperature (7). difference in analyte concentration that the method is able to
Quantification in drinking water samples was simply made by discriminate (12). It is closely related to sensitivity and residual
calibration against external standards prepared in distilled water. standard deviation, Sy/x. Admitting that the errors associated
However, for wastewater samples the procedure was found not with the concentrations within the working range are Gaussian,
to perform acceptably for every sample. Consequently, prior to then, to have statistical significance, two concentrations must
quantification, a recovery test was done with such wastewater differ more than 3 * √2 * Sy/x/b (for a 99.7% confidence interval).
samples. The method was considered acceptable if sample A DT of 0.06 mg/L N was thus obtained, meaning that with such
recoveries were between 90–110%. If recoveries were outside calibration curves it is possible to significantly discriminate
of this range, the sample would then be analyzed for nitrate by between 75 different concentration levels within the working
an alternative procedure. range.
Safety notes: Hydrochloric acid causes burns and is irritating The LOD was determined as 3.3 * Sy/x/b, corresponding to
to respiratory system. Avoid contact with skin and eyes. 1 0.05 mg/L N, while the LOQ was determined according to the
Avoid inhalation of vapor or mist. NEDA is irritating to eyes, International Union of Pure and Applied Chemistry criterion (13)
respiratory system and skin. Avoid breathing dust and wear of taking LOQ as the lowest concentration at which acceptable
suitable personal safety equipment. levels of precision and accuracy (<10%) are observable.
Ten replicated measurements performed under intermediate
precision conditions showed that a variation coefficient of 6.6%
Results and Discussion
and bias of 1% were obtained at a concentration of 0.5 mg/L N,
and so the lower limit of the working range was also considered
Indirect Evaluation of the Method
the method’s LOQ.
The indirect evaluation involved the different assessments
of the calibration curve, such as operating range, linearity, Precision
sensitivity, detection and quantification limits, precision
(repeatability, intermediate precision), accuracy, and selectivity. Precision was evaluated under repeatability and intermediate
precision conditions. Reproducibility, on the other hand, would
Operating Range, Calibration Curve and Linearity have required a between-lab scheme which unfortunately was
not possible to setup.
The intended working range, 0.5 to 5 mg/L N, was submitted The repeatability study was carried out with eight real
to a variance homogeneity test according to the ISO 8466-1 samples [drinking water (DW) and wastewater (WW)] and
standard (10). The variances obtained for the signal from two standard solutions. Each sample was analyzed 10 times in
independently prepared standards (n = 10) at the lower and the same working session and, before proceeding any further,
–5 2
upper range limits were 1.04 × 10 absorbance units (a.u.) the Grubbs’ test was applied so as to rule out possible outliers,
–5 2
and 1.69 × 10 (a.u.) , respectively. The corresponding F in compliance with standard ISO 5725-2 (14). After checking
value was 1.63, which is below the critical F (3.18) for 95% for outliers, the repeatability standard deviation (Sr) and the
degree of confidence; thus, warranting the applicability of the coefficient of variation (CVr) were computed according to the
non-weighed least-squares method for obtaining the calibration standard ISO 5725-6 (15), and the corresponding values are
equation. compiled in Table 1. Method repeatability, evaluated from
The calibration curves were obtained from the signals the CV values, is generally considered acceptable under 10%.
measured for six different concentration levels (0.5, 1.0, 2.0, The CV values obtained lie at ≤5% (with one exception with
3.0, 4.0, 5.0 mg/L). Using linear regression, in accordance with the nitrate at the LOQ concentration level), and therefore the
the ISO 8466-1 standard (10), the average (of 9 calibration method under study has demonstrated acceptable repeatability
curves) regression parameters obtained are presented in Figure 1 for real samples and standard solutions.
(averaged calibration curve). The intermediate precision may be ascertained by several
Although both residual analysis (not shown) and correlation different ways (ISO 5725-3, 16). The approach selected here
4  Guimarães et al.: Journal of AOAC International 2014

Table  1.  Results from repeatability studies, using real (DW size), and subjected to nitrite removal when necessary, as a
and WW) samples and standard solutions) first step of minimization of potential matrix effects. For some
samples, their intense color/turbidity could interfere with the
Sample |Nitrate|/mg/L N Sr/mg/L N CVr, %
standard addition procedure, even after being greatly reduced
DW1 0.51 0.023 4.6 by dilution and filtration. Therefore, the sample background
DW2 1.56 0.049 3.1 signal was systematically checked at 540 nm before NEDA
DW3 1.68 0.049 2.9 addition (Equation 3), to monitor any significant contribution
to the signal attributed to nitrate. We found that none of the
WW1 2.50 0.042 1.7
analyzed sample backgrounds contributed significantly (all
WW2 3.68 0.035 0.96
<0.003 a.u.), but we could not obviously encompass all possible
WW3 0.99 0.037 3.8 types of wastewater samples, so this check-up should not
WW4 0.90 0.026 2.9 be neglected. The results are presented in Table 2. All of the
WW5 0.87 0.043 4.9 DW samples appear not to induce matrix effects, but from the
Standard solution 0.50 mg/L N 0.51 0.034 6.6
12  WW samples tested, five produced significantly different
results (Student’s-t-test, a  =  0.05), by the two quantification
Standard solution 5.00 mg/L N 5.01 0.043 0.86
methods. These five samples (WWh to WWk) were sampled
from miscellaneous types of industries. Both colorless and
consisted on taking two standard solutions at concentration markedly colored, low and high organic contents, low and
levels located at the upper (5 mg/L N) and lower (0.5 mg/L high nitrogen contents, and low and high turbidity, were
N) limits of the working range and subjecting them to characteristics found among the water samples exhibiting
31 determinations performed over a three month period. Each matrix effects. None of the commonly known characteristics
of the 31 determinations of nitrate relied on an independent of the samples seemed to be, by themselves, dictating the
calibration prepared at the same session. In this manner, variance appearance of a matrix effect. Nevertheless, an expedited way to
contributions from sources over relatively long periods of time deal with the potential problem of significant matrix-effects in
(e.g., calibration curve parameters, preparation of new reagent wastewaters was necessary. It was decided, from this point on,
solutions, and others), may be accounted for (the operator was to include the previous dilution into the final method (referred
above plus a single standard addition check-up for recovery).
kept constant). After outlier identification and rejection (using
Recoveries with 90–110% limits were considered acceptable
the Grubbs test previously mentioned), intermediate precision
and calibration curve quantification was performed. Otherwise,
values of 0.033 mg/L N (CV 6.6%) and 0.037 mg/L N (CV
samples were ruled-out from the method’s applicability and
0.74%) were obtained for 0.5 and 5 mg/L N, respectively.
a modified method, employing multiple standard addition
The corresponding variances were found statistically identical
quantification, is suggested as a feasible alternative for handling
(a = 0.05) which implies that the intermediate variance
these samples. It is critical to notice that all figures of merit
appears to be sufficiently homogeneous over the working
presented above are not applicable to this alternative method,
range. Additionally, no statistically significant differences
which must be submitted to an independent validation process
were found between intermediate precision and repeatability,
which is not covered here. Therefore it must be emphasized that
which suggests that the time-dependent sources of variance
all further data presented here relate to samples not ruled out by
had no significant additional influence on the total variance.
the recovery check-up.
We speculate that since a calibration curve was prepared each
Aiming at the identification of some possible interferents when
day for each sample set, that this must have cancelled most
using external calibration, it was decided to evaluate the effect of
of such possible contributions. For example, when a reagent
typical concentrations of species commonly found in wastewater
solution was refreshed, the new solution was applied both to the samples. For inorganic species, we selected iron(II), sodium,
calibration solutions and the sample, and this source of variance sulfate, ammonium, chloride, and zinc(II). For organic species,
would be mostly cancelled. in order to simulate chemical and biochemical oxygen demand
(COD and BOD) and also organic nitrogen, we chose potassium
Matrix Effects, Selectivity and Recovery hydrogen phthalate (KHP), glutamic acid (GTA)/glucose (GLU)
mixture, and glycine. Table 3 shows the recovery results for
Although a good precision under repeatability conditions was 0.50 mg/L N nitrate solutions containing different concentrations
found for a set of different DW and WW samples (see Table 1), of potential interferents. With the exception of iron(II), the
the matrix components may deleteriously affect the accuracy inorganic salts, tested at a relatively high concentration level
of the determinations. This issue is especially relevant for (1 g/L), did not interfere; iron (II) exhibited an interfering effect
WW samples, which often exhibit quite complex and diverse above 100 mg/L, but a lower interfering threshold (>20 mg/L) is
matrixes. To check for a significant influence of the matrix on probable. Concerning the effect of GTA/GLU, an abnormally high
the accuracy of the results, 5 DW and 12 WW samples were recovery was observed at 3.5 + 3.5 g/L (corresponding to BOD
subjected to both the proposed method of determination (whose ≈ 5000 mg O2/L), suggesting that high levels of organic acids
quantification relies on an external calibration curve), and to and/or sugars may present an interfering effect. KHP at 17 g/L
a modified method where the quantification was performed (COD = 20000 mg O2/L) was also an issue as far as interferences
by successive standard addition; an approach consensually are concerned, though such high values of COD are rare in real
regarded as much less susceptible to matrix effects. The samples wastewater samples. Nevertheless, GTA, GLU and KHP are
were previously diluted with distilled water to the neighborhood only surrogates for BOD and COD originating compounds, so
of the lower end of the working range, filtered (0.45 mm pore we cannot exclude the possibility that real matrices with lower
Table  2.  Results from matrix effect studies, using real (DW and WW) samples analysed by direct reading and by standard additions

Significant
|NO3–|/mg/L Nmg/L N |NO3–|/mg/L Nmg/L N Dilution (α = 0.05)
a
Sample Calibration curve (*) Standard addition (*) Recovery, % factor difference (Y/N)** Observations*

DWa 0.47 0.46 102 4 N Source: cork industry (well water).


DWb 0.56 0.53 106 30 N Source: cork industry.

DWc 0.47 0.50 94 16.7 N Source: metal finishing industry.

DWd 0.68 0.65 105 20 N Source: car wash.

DWe 0.85 0.85 100 10 N Source: cork industry.

WWa 0.79 0.79 100 50 N Source: wood industry; after microbiological treatment; yellow color; COD < 4 mg O2/L, BOD < 0.4 mg O2/L.

WWb 0.38 0.39 97 50 N Source: paint industry; light yellow color; COD < 4 mg O2/L, BOD < 0.4 mg O2/L.

WWc 0.50 0.51 98 6 N Source: textile industry; brown color.

WWd 0.49 0.51 96 20 N Source: cork industry; light brown color; COD < 2 mg O2/L.

WWe 0.55 0.56 98 40 N Source: aluminum processing industry; light yellow color; COD < 4 mg O2/L, BOD < 0.4 mg O2/L.

WWf 0.70 0.72 97 50 N Source: paint industry; light yellow color; COD < 4 mg O2/L.

WWg 0.50 0.58 86 20 Y Source: cork industry; colorless, COD < 2 mg O2/L, BOD < 0.4 mg O2/L.

WWh 0.66 0.74 89 10 Y Source: municipal wastewater treatment plant; colorless, COD < 2 mg O2/L, BOD < 0.4 mg O2/L.

WWi 0.23 (<LOD) 0.56 41 40 Y Source: waste management (oils and fats); yellow color (colorless after dilution), foaming, intense smell, COD
625 mg O2/L, containing oils and fats
WWj 0.58 0.46 126 10 Y Source: metal ware industry; colorless.

WWk 0.56 0.44 127 10 Y Source: wood industry; dark-brown color, COD 1650 mg O2/L, ammonia nitrogen 23 mg/L.
a –
 
(*) = |NO3 | in the diluted sample. COD (Chemical Oxygen Demand) and BOD (Biochemical Oxygen Demand) values after sample dilution. ** = Student’s t-test comparison
Guimarães et al.: Journal of AOAC International 2014  5
6 Guimarães et al.: Journal of aoaC international 2014

Table 3. Recoveries obtained for 0.5 mg/L N nitrate 0.81. In the case of the WW sample, z-score/En values could
solutions in the presence of potential interferents not be calculated since the reference value is indeterminate
(<1.51 mg/L). However, it should be noted that the analyzed
Potential interferent Concentration, g/L Recovery, %
concentration was also indeterminate (<0.50 mg/L) and that the
FeCl2 0.010 104 two results are therefore compatible. The method under study
0.020 102 satisfactorily passed this proficiency test, which was based on a
0.10 140 single z-score/En set. Further tests, such as the comparison with
a standard method, are thus required to ultimately determine the
1.0 29
adequacy of this method.
Na2SO4 1 107
NH4Cl 1 98 Standard Method Comparative Study
ZnCl2 1 98
Glycine 0.35 113 Five samples, two DW and three WW, analyzed by the
3.5 94 current method, were sent to an accredited lab (GBA, Hamburg,
Glutamic acid + Glucose
a
0.07+0.07(BOD ≈100) 109
Germany) for nitrate determination using an accredited standard
method based on ion chromatography (18). Table 5 shows the
0.35+0.35(BOD≈500) 109
obtained results. A very close agreement (relative deviation
3.5+3.5(BOD≈5000) 193 ≤7%) between the results was observed for this set of samples.
a
Potassium hydrogen phthalate 0.85 (COD ≈1000) 105 The most statistically meaningful comparison would be
8.5 (COD≈10000) 105 achieved through a metrological compatibility test (19) but for
17 (COD≈20000) 128 that the uncertainties for each individual measurement needed
a
to be known, and unfortunately the contracted laboratory
COD (Chemical Oxygen Demand) and BOD (Biochemical Oxygen
Demand); Values in mg/L O2.
only provided a general estimate for this (uncertainty = 4%).
Alternatively, the results of the two methods were found to be
statistically (a = 0.05) indistinguishable, both by the paired
BOD or COD values may originate an interfering effect. Glycine, t-test and regression analysis (slope not statistically different
a compound containing nitrogen, appears not to interfere up to from 1). It is reasonable to consider then that the method under
653 mg/L N but this does not necessarily imply that other amino- study passed this comparative test, which supports its ability to
compounds do not interfere at lower concentrations. provide true results.

Direct Evaluation of the Method Estimation of Combined Uncertainty

This kind of evaluation aims primarily at the direct The estimation of combined uncertainty followed Eurachem’s
determination of the accuracy of the method by resorting to guidelines (20). Standard uncertainties are denoted, u, while
certified reference materials, proficiency tests or standard relative standard uncertainties are denoted u′.
methods. Herein we present results of a proficiency test and The uncertainty associated with precision, u′p, was estimated
comparative results against a standard method. at the two ends of the working concentration range (0.500 and
5.000 mg/L N). The data previously acquired for the intermediate
precision (Sip) was used for the purpose. The values of u′p were
Proficiency Test
computed simply as u′= Sip/|NO3–| and the obtained values were
0.07 (0.50 mg/L N) and 0.007 (5.00 mg/L N).
Two water samples (DW and WW) from the Association
The uncertainty associated with accuracy , u′t, was estimated
of Portuguese Accredited Labs (RELACRE) wastewater
from the information collected when comparing against the
proficiency test were subjected to the method under scrutiny reference method (Table 5). Estimated uncertainty attributed
in 2011. The results are shown in Table 4. A z-score between to recovery (defined as R=Ccurrent method/Creference method) was first
–2 and 2 or |En|<1 (normalized error) are generally considered calculated as
indexes of satisfactory performance of a laboratory (17). For
the DW sample, the reference and sample concentrations were  S ip2   u(C ref.method ) 
2

u(R m ) = R m ⋅  + 
almost identical, corresponding to a z-score of 1.3 and to En =  n ⋅ C curr .method 2   
   C ref.method 

Table 4. Results obtained in the RELACRE Wastewater Proficiency Test 2011


Sample Reference concn/mg/L NO3– Sref/mg/L NO3– Analyzed concn/mg/L NO3– z-scorea Enb

DW 10.1 0.4 10.6 1.3 0.81


WW < 1.51 0.51 < 0.50 — —
a
z = (Vlab-Vref)/S; where S stands for the standard deviation associated to the reference value (Vref) and Vlab: the value obtained by the participating
laboratory.

b
where Ulab and Uref are the expanded uncertainties associated to Vlab and Vref, respectively. Ulab was considered 0.14 (see Table 6)
Guimarães et al.: Journal of AOAC International 2014  7

Table  5.  Comparison of nitrate concentration obtained by 0,1040

a standard method and by the proposed method 0,1020

Standard method Current method Relative 0,1000


Sample ID concn/mg/L N concn/mg/L N deviation, %
0,0980 +3s

DW I 1.90 1.84 3

Slope / L/mg N
+2s
0,0960
DW II 2.70 2.89 7
0,0940
WW I 2.00 1.92 4 +2s
+3s
0,0920
WW II 2.70 2.82 4
0,0900
WW III 0.75 0.77 2
0,0880

0,0860
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75

where u(Rm) is the uncertainty of the mean recovery Rm and n Time / days

the number of recovery tests (5 in this case). The value used as Figure  2.  Individual control chart for calibration slope.
u(Cref.method)/Cref.method) = 0.04 was provided by the contracted Figure 2.
lab. A Student’s-t test showed that Rm is not significantly
(α  =  0.05) different from 1. In such case, the uncertainty chart was inspected at each new point according to the criteria
u′t is simply equivalent to the uncertainty of recovery, i.e.: recommended in the ISO 8258 standard (21). Abnormally high
u′t = u′(Rm) = u(Rm)/∙Rm. The average uncertainty of recovery values were observed at several points for the first twenty days.
value of 0.03 (see Table 6), was then taken as an estimate of u′t The cause for these anomalies was found to be related to the
for the whole working range of concentrations. short-term stability of the colorant solution, which prompted a
The combined (u′c) and expanded [u′c, (95%) confidence change to the method of a maximum stability period of 15 days
interval], uncertainties were finally estimated for the two for this solution (protected from light ). Another individual
ends of the working range, according to the relationship; value of the ICC (47th day) fell off-limits but then the slope
u′c = ((u′p)2+(u′t)2)1/2 and U′c(95%) = k.u′c, where k stands for the returned to the acceptable region within the prescribed control
coverage factor (20). The results are presented in Table 6. The limits until the end of the study. It can finally be stated that,
estimated values locate the expanded uncertainty of the method as long as the previously mentioned corrective measure was
between 6% (upper end of the working range; uncertainty maintained, the calibration remained under statistical control
dominated by accuracy ) and 14% (lower end of the working during the remaining enclosed period.
2

range; uncertainty dominated by precision). These values The control line of the RCC (Figure 3) is located at range 3.27
may be considered quite satisfactory in the light of the main Rm (Rm stands for the mean range observed from a previous
purpose of such a method which is to verify the compliance set of 9 triplicate determinations). The triplicate determinations
with a regulatory upper limit of 11.3 mg/L N in water. In this were performed under repeatability conditions on incoming
case, uncertainty in the upper working range, due to a question wastewater samples. Throughout the whole period the range
of eventual proximity to the decision threshold (the ACV), is remained below the control line and therefore the short-term
especially relevant and the value estimated is adequately low. precision was under statistical control, despite the much
diversified provenance of the wastewater samples analyzed.
Internal Quality Control
Conclusions
Both indirect and direct evaluation of the method as well
as the respective expanded uncertainty estimation provided The assessed method exhibited a satisfactory performance
satisfactory outputs that allowed for assuming the overall under indirect evaluation (operating range, calibration curve,
reliability of the method. It was further decided to advance linearity, sensitivity, LOD and LOQ, precision, selectivity, and
to the internal quality control stage, which can assure that the recovery), for drinking and wastewater samples. In the case of
method performance remains valid throughout a relatively wastewater samples, due to the possibility of significant matrix
long period of operation. An individual control-chart (ICC) of
calibration slope and a range control-chart (RCC) of triplicate
determinations were constructed in the course of 2.5 months. 0,3

Control and warning upper and lower limits of the ICC 0,25
3.27 Rm*

(Figure 2) were obtained from the mean and standard


deviation of an initial set of 9 calibrations, after passing the 0,2
Kolmogorov-Smirnov test for a normal distribution. The
Amplitude

0,15

0,1
Table  6.  Combined and expanded uncertainties estimated
for the proposed method
0,05
U′c (95%) = k∙u′c
Concn level/mg/L N u′p u′t =u′ (Rm) u′c (k = 2) 0
0 2 4 6 8 10 12 14 16
0.500 0.07 0.03 0.07 0.14
Sample number
5.000 0.007 0.03 0.03 0.06
Figure  3. * Range control chart for triplicate determinations.
Rm stands for the mean range observed from a previous set of nine triplicate
determinations.

Figure 3
8  Guimarães et al.: Journal of AOAC International 2014

effects, dilution and a simple spike recovery test were included Standardization, method ISO 8466-2. https://www.iso.org/obp/
in order to decide the method’s applicability to each sample; ui/#iso:std:iso:8466:-2:ed-2:v1:en
a recovery within 90–110% is acceptable; while a recovery (12) EURACHEM Guides (1998), 1st Edition, EURACH EM, The
outside of that range is not and a multiple standard additions Fitness for Purpose of Analytical Methods. A Laboratory Guide
quantification would be recommended but unable to be to Method Validation and Related Topics. http://www.eurachem.
validated). Results from a proficiency test and comparison with a org/images/stories/Guides/pdf/valid.pdf
standard method provided a direct evaluation of the method, and (13) Thompson, M., Ellison, S.L.R., & Wood, R. International Union
further emphasized the fitness-for-use of the proposed method. of Pure and Applied Chemistry (2002) Pure Appl. Chem. 74,
Acceptable estimated values of the combined uncertainty were 835–855. http://www.iupac.org/publications/ci/2002/2405/
pac_7405x0835.html
obtained, and a 2.5 month period study of the internal quality
(14) Accuracy (trueness and precision) of Measurement Methods
control of the procedure provided positive indications pointing
and Results - Part 2: Basic Method for the Determination of
to long-term stable behavior of the procedure. Therefore, the
Repeatability and Reproducibility of a Standard Measurement
proposed method constitutes a reliable, simple, rapid, and
Method (1994), International Organization for Standardization,
inexpensive procedure which may advantageously replace other
Method ISO 5725-2. http://www.iso.org/iso/home/store/
methods of nitrate determination in routine work.
catalogue_tc/catalogue_detail.htm?csnumber=11834
(15) Accuracy (trueness and precision) of Measurement Methods
References and Results - Part 6: Use in Practice of Accuracy Values
(1994) International Organization for Standardization, method
  (1) Ting, D.S., Song, Z.Y., & Yong, L. (2007) Agric. Sci. China 10, ISO 5725-6. http://www.iso.org/iso/home/store/catalogue_tc/
1246–1255 catalogue_detail.htm?csnumber=11837
  (2) Krause, S., & Clarck, H.M. (2002) Chemistry of the (16) Accuracy (trueness and precision) of Measurement Methods
Environment, Elsevier Science & Technology Books, and Results - Part 3: Intermediate Measures of Precision
Amsterdam, The Netherlands
of a Standard Measurement Method (1994) International
 (3) Official Journal of the European Communities (1991) European
Organisation for Standardization, method ISO 5725-3. http://
Union, Council Directive 91/676/EC. http://eur-lex.europa.eu/
www.iso.org/iso/home/store/catalogue_tc/catalogue_detail.
legal-content/EN/TXT/?uri=CELEX:31991L0676
 (4) Official Journal of the European Communities (1998) European htm?csnumber=11835
Union, Council Directive 98/83/EC. http://faolex.fao.org/cgi- (17) Conformity assessment — General requirements for proficiency
bin/faolex.exe?rec_id=013657&database=FAOLEX&search_ testing (2010) International Organization for Standardization
type=link&table=result&lang=eng&format_name=@ERALL and International Electrotechnical Commission, method
  (5) Moorcroft, M.J., Davis, J., & Compton, R.G. (2001) Talanta 54, ISO/IEC 17043. https://www.iso.org/obp/ui/#iso:std:iso-
785–803 iec:17043:ed-1:v1:en
 (6) Standard Methods for the Examination of Water and Wastewater (18) Water Quality – Determination of dissolved anions by liquid
(2005), 21st Ed, American Public Health Association APHA, chromatography of ions – Part 1: Determination of bromide,
Washington DC chloride, fluoride, nitrate, nitrite, phosphate and sulfate (2007),
  (7) Mir, S.A. (2008) Anal. Chim. Acta 620, 183–189 International Organization for Standardization, method ISO
  (8) Mir S.A. (2007) Asian J. Chem. 19, 5703–5710 10304-1. http://www.iso.org/iso/home/store/catalogue_tc/
 (9) Standard Methods for the Examination of Water and Wastewater catalogue_detail.htm?csnumber=46004
(2005) American Public Health Association, Washington DC, (19) International Vocabulary of Metrology — Basic and general
method SMEWW 4500-NO2-B
concepts and associated terms; (VIM) (2008) 3rd Edition, Joint
(10)  Water quality - Calibration and evaluation of analytical
Committee for Guides in Metrology. http://www.bipm.org/utils/
methods and estimation of performance characteristics - Part 1:
common/documents/jcgm/JCGM_200_2012.pdf
Statistical evaluation of the linear calibration function (1990),
International Organization for Standardization, method ISO (20) EURACHEM Guides (2000) 2nd Edition, EURACHEM,
8466-1. https://www.iso.org/obp/ui/#iso:std:iso:8466:-1:ed- Quantifying Uncertainty in Analytical Measurement. http://
1:v1:en www.iswa.uni-stuttgart.de/ch/aqs/pdf/quam2000-p1.pdf
(11)  Water quality - Calibration and evaluation of analytical (21) Shewhart control charts (1991) International Organization
methods and estimation of performance characteristics - for Standardization, method ISO 8258. http://www.iso.
Part 2: Calibration strategy for non-linear second-order org/iso/home/store/catalogue_ics/catalogue_detail_ics.
calibration functions (2001) International Organization for htm?csnumber=40174

View publication stats

You might also like