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

net/publication/233982995

EC50 estimation of antioxidant activity in DPPH assay using several statistical

programs

Article  in  Food Chemistry · May 2013

DOI: 10.1016/j.foodchem.2012.11.001 · Source: PubMed

CITATIONS READS

191 12,350

3 authors:

Zheng Chen Riccardo Bertin

University of Padova University of Padova
14 PUBLICATIONS   303 CITATIONS    30 PUBLICATIONS   287 CITATIONS   

SEE PROFILE SEE PROFILE

Guglielmina Froldi
University of Padova
101 PUBLICATIONS   1,165 CITATIONS   

SEE PROFILE

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

Cardiac activity and inotropic effect View project

Medicinal plants View project

All content following this page was uploaded by Riccardo Bertin on 20 October 2017.

The user has requested enhancement of the downloaded file.

 Food Chemistry 138 (2013) 414–420

Contents lists available at SciVerse ScienceDirect

Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem

EC50 estimation of antioxidant activity in DPPH assay using several

statistical programs
Zheng Chen, Riccardo Bertin, Guglielmina Froldi ⇑
Department of Pharmacology and Anaesthesiology, University of Padova, Largo E. Meneghetti 2, 35131 Padova, Italy

a r t i c l e i n f o a b s t r a c t

Article history: DPPH assay is a reliable method to determine the antioxidant capacity of biological substrates. The DPPH
Received 13 January 2012 radical scavenging activity is generally quantified in terms of inhibition percentage of the pre-formed free
Received in revised form 26 July 2012 radical by antioxidants, and the EC50 (concentration required to obtain a 50% antioxidant effect) is
Accepted 3 November 2012
a typically employed parameter to express the antioxidant capacity and to compare the activity of
Available online 12 November 2012
different compounds. In this study, the EC50 estimation was performed using a comparative approach
based on various regression models implemented in six specialized computer programs: GraphPad
Keywords:
PrismÒ version 5.01, BLeSq, OriginProÒ 8.5.1, SigmaPlotÒ 12, BioDataFitÒ 1.02, and IBM SPSS StatisticsÒ
DPPH assay
Antioxidant activity
Desktop 19.0. For this project, quercetin, catechin, ascorbic acid, caffeic acid, chlorogenic acid and
EC50 estimation acetylcysteine were screened as antioxidant standards with DPPH assay to define the EC50 parameters.
Statistical programs All the statistical programs gave similar EC50 values, but GraphPad PrismÒ five-parameter analysis
pointed out a best performance, also showing a minor variance in relation to the actual EC50.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction (ABTS+) assay (Miller, Rice-Evans, Davies, Gopinathan, & Milner,

1993), the oxygen radical absorbance capacity (ORAC) assay (Cao,
Oxidative stress is involved in the production of potentially Alessio & Cutler, 1993; Glazer, 1990), the cupric reducing antioxi-
harmful free radicals which play a pivotal role in the pathogenesis dant capacity (CUPRAC) assay (Apak, Güçlü, Ozyürek, & Karademir,
of various diseases including atherosclerosis, neurodegenerative 2004), the electrochemical estimation of total reducing capacity
diseases, rheumatoid arthritis, age-related degeneration and can- (Chevion, Roberts, & Chevion, 2000) and the 1,1-diphenyl-2-pic-
cer initiation. These effects are achieved by action of radical oxygen rylhydrazyl radical (DPPH) assay (Bondet, Brand-Williams, &
(ROS) and reactive nitrogen (RNS) species, generally named as Berset, 1997; Brand-Williams, Cuvelier, & Berset, 1995). Among
reactive species (RSs), responsible for cellular disorders through these methods, the last is one of the most extensively used to
their action on proteins, lipids and DNA. RSs modify the oxidative evaluate antioxidant activity of plant extracts, foods and single
balance within the cells and thus they are also important media- compounds, thanks to its stability and its easiness (Krishnaiah,
tors of inflammatory-like injuries (Valko et al., 2007). The natural Sarbatly, & Nithyanandam, 2011). This assay is based on the
antioxidants are believed to be very important for human diet, measurement of the reducing ability of antioxidants toward DPPH
especially in relation to oxidative stress prevention; it has been radical, through electron spin resonance (EPR) detection or by
recognised an inverse association between the consumption of measuring the decrease of its absorbance (Prior, Wu, & Schaich,
fruits and vegetables and mortality from senescence-related disor- 2005). DPPH is a stable and commercially available organic nitro-
ders, which could be partly attributed to the presence of antioxi- gen radical, which reacts with hydrogen/electron donor com-
dant compounds, mainly with phenolic structure (Blokhina, pounds and has a maximum UV–Vis absorption within the range
Virolainen, & Fagerstedt, 2003; Scalbert, Johnson, & Saltmarsh, of 515–520 nm. Upon reduction, the radical solution becomes
2005). For this reason, extensive investigation of natural sources discoloured according to the number of electrons paired, and the
of efficient radical-scavenging compounds is receiving high atten- reaction progress is conveniently monitored by a spectrophotome-
tion in health research. A wide variety of methods have been devel- ter (Blois, 1958; Brand-Williams et al., 1995; Gil, Tomás-Barber,
oped for the antioxidant assessment, including the ferric reducing Hess-Pierce, Holcroft, & Kader, 2000). The analysis is simple, sensi-
antioxidant power (FRAP) assay (Benzie & Strain, 1996), the 2,20 - tive and fairly rapid and needs only a UV–Vis spectrophotometer to
azino-bis(3-ethylbenzothiazoline-6-sulphonate) radical cation perform, which probably explains its widespread use in antioxi-
dant screening. The result is normally expressed using the EC50
value, defined as the concentration of antioxidant that causes a
⇑ Corresponding author. Tel.: +39 0498275092; fax: +39 0498275093.
50% decrease in the DPPH absorbance. Several experimental
E-mail address: [email protected] (G. Froldi).

0308-8146/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.foodchem.2012.11.001
 Z. Chen et al. / Food Chemistry 138 (2013) 414–420 415

evidences have indicated a non-linear relationship between the 2.3.1. GraphPad Prism 5.01
antioxidant concentration and the DPPH radical scavenging activ- The program is implemented with built-in equations for non-
ity (Eklund et al., 2005; Villano, Fernández-Pachon, Troncoso, & linear regression; log (inhibitor) vs. normalised response (variable
García-Parrilla, 2005); as a consequence, the determination of slope) and asymmetric (five-parameter, 5P) were the statistical
EC50 becomes quite problematic, revealing a variable goodness of models we used for data elaboration of DPPH assay.
fit for the plotted regression models. The discrepancy between The first option, usually known as four-parameter regression
the observed values and the estimated values under the statistical (4P), is given by the following equation:
models can be overcome through the construction of an appropri- 1
ate mathematical equation to optimise the antioxidant profile of Y¼
1 þ 10f½log EC 50 log ðxÞHillslopeg
the samples. The aim of this study was to identify the more suit-
able program for the EC50 estimation from experimental data ob- where Y is the response and x is the concentration of the agonist,
tained by DPPH assay by comparing various statistical programs. while the asymmetric model is represented by:
For this, six computational programs and four different regression Top  Bottom
models were employed to estimate the EC50 value, using various Y ¼ Bottom þ
ð1 þ 10½ðlog xb log xÞHillslope Þs
standard natural antioxidants.
These two equations share three common parameters: the bot-
tom asymptote, which refers to the baseline and the top asymptote
2. Materials and methods that represents the maximum response, while the Hill slope is re-
lated to the steepness of the antiradical curve. In the first equation,
2.1. Materials and instruments EC50 stands for the actual EC50 value, while in the second equation
xb is the concentration of the sample at the inflection point. The
1,1-Diphenyl-2-picrylhydrazyl (DPPH) and the reference stan- five-parameter equation adds another variable, ‘‘s’’, which quanti-
dards quercetin, (+)-catechin, L-ascorbic acid, caffeic acid, chloro- fies the asymmetry of the curve.
genic acid and N-acetyl-cysteine were purchased from Sigma
Chemical Co. (St. Louis, MO, USA). HPLC grade methanol was pur- 2.3.2. BLeSq 0.9.1
chased from Carlo Erba (Milano, Italy). Dimethyl Sulfoxide (DMSO) BLeSq is a software specifically developed for the calculation of
was purchased from J.T. Baker (Deventer, Holland). Fresh stock EC50 in DPPH scavenging assay; it was designed to perform the
solutions were prepared before each analysis. The absorbance determination of outliers, the calculation of EC50 values and the
measurements were performed using a Beckman Coulter DU 800 comparison of the curves – and their corresponding straight lines
spectrophotometer (Fullerton, CA, USA). – obtained from the regression analysis of the data (Locatelli
et al., 2009). After data loading, even in replicates, the program
marks the antiradical curve with the EC50 value expressed as con-
2.2. DPPH method centration range within a 95% confidence interval. There are three
different models implemented for the curve fitting: ‘‘probit’’, ‘‘lo-
The DPPH radical scavenging assay was performed according to git’’ and ‘‘angular regression’’. The probit model [Pr(Y = 1|X) = uX0 b]
the method reported by Brand-Williams et al. (1995) with some was introduced for the calculation of the sigmoid dosage-mortality
modifications (Mishra, Ojha, & Chaudhury, 2012; Sharma & Bhat, curve (Bliss, 1935), and later on was applied also in other statistical
2009). Briefly, standard solutions or DMSO (control) were prepared fields. In this equation Y is a (0,1) binary, where Pr denotes proba-
and added to a 70 lmol/l DPPH methanolic solution; the mixtures bility and u is the Cumulative Distribution Function (CDF) of the
were shaken vigorously and left to stand in the dark for 30 min at standard normal distribution. The b parameter typically refers to
room temperature, then absorbance was read at 517 nm. Radical a maximum-likelihood estimation, which provides a method for
scavenging capacity was expressed as percentage effect (E%) and estimating different variables of a statistical model.
calculated using the following equation: BLeSq uses the form below, where N is the standard normal dis-
  tribution and ln is the natural logarithm:
Abscontrol  Abssample Z
Percentage effect ðE%Þ ¼  100 bþaln ½x
Abscontrol YðxÞ ¼ probitðb þ a  ln ½xÞ ¼ N½tdt
1
Different sample concentrations were used in order to obtain
antiradical curves for calculating the EC50 values. Antiradical The logit model [y = ln (p/1  p)] was introduced by Berkson
curves were plotted referring to concentration on the x axis and (1944). This transformation is more arbitrary but has important
their relative scavenging capacity on the y axis. The EC50 values advantages, such as the easiness of calculation and the wide appli-
were processed using six different statistical programs. cations in epidemiological studies: logit is defined as the logarithm
of the odds, and logit differences are logarithms of odds ratios
(Armitage, Berry, & Matthews, 2005).
2.3. Data analysis The angular model [y = sin2(x)] was developed to simplify all the
biological assays having quantal responses. Probit and angular
Results were expressed as Mean ± Standard Deviation (SD) of model give practically the same results, even if probit calculation
three independent experiments for each antioxidant; following algorithm appears about twelve times longer than the angular
the statistical interpretation of data, EC50 values were expressed transformation. BLeSq used the modified form [Y = sin2(b + aln [x])]
as 95% confidence interval. (Knudsen & Curtis, 1947).
Data analysis was performed using the algorithms implemented The authors of the software suggested the use of the probit fit-
in six frequently used statistical programs: GraphPad PrismÒ ver- ting mode, but in our evaluation logit option worked with a better
sion 5.01 (San Diego, CA), BLeSq (implemented from Microsoft Vi- accuracy. For this reason we chose logit for our comparative study,
sual BasicÒ 6.0, Redmond WA), OriginProÒ version 8.5.1 following the transformation presented below:
(Northampton, MA), SigmaPlotÒ version 12 (San Jose, CA), BioData-
FitÒ version 1.02 (Castro Valley, CA) and IBM SPSS StatisticsÒ Desk- e½bþaln ðxÞ
Y¼
top version 19.0 (Armonk, NY). e½bþaln ðxÞ þ 1
416 Z. Chen et al. / Food Chemistry 138 (2013) 414–420

Table 1
The EC50 for the standard antioxidants expressed as pD2 (log EC50) obtained by statistical elaborations and experimental procedures (actual pD2).

Quercetin (+)-Catechin L-Ascorbic acid Caffeic acid Chlorogenic acid N-Acetylcysteine

a
GraphPad 5.392 ± 0.046 5.130 ± 0.047 4.869 ± 0.032 5.007 ± 0.038 5.309 ± 0.035 4.466 ± 0.039
GraphPad 5Pb 5.316 ± 0.046 5.109 ± 0.021 4.830 ± 0.058 4.993 ± 0.011 5.293 ± 0.023 4.578 ± 0.035
Blesqc 5.272 ± 0.041 5.097 ± 0.054 4.851 ± 0.049 5.018 ± 0.051 5.246 ± 0.042 4.388 ± 0.050
Blesqd 5.304 ± 0.076 5.000 ± 0.076 4.893 ± 0.039 5.018 ± 0.051 5.305 ± 0.035 4.496 ± 0.010
BioDataFit 5.411 ± 0.030 5.166 ± 0.035 4.881 ± 0.023 4.992 ± 0.024 5.314 ± 0.016 4.576 ± 0.023
OriginPro 5.411 ± 0.030 5.166 ± 0.035 4.881 ± 0.023 4.992 ± 0.024 5.314 ± 0.016 4.576 ± 0.023
SigmaPlot 5.316 ± 0.046 5.109 ± 0.021 4.830 ± 0.058 4.993 ± 0.011 5.293 ± 0.023 4.578 ± 0.035
SPSS 5.354 ± 0.212 5.159 ± 0.131 4.870 ± 0.136 5.023 ± 0.138 5.287 ± 0.125 4.440 ± 0.130
Actual pD2 5.261 ± 0.021 5.095 ± 0.004 4.793 ± 0.006 4.930 ± 0.016 5.203 ± 0.008 4.521 ± 0.028
a
GraphPad log (inhibitor) vs. normalised response model (variable slope).
b
GraphPad five-parameter regression model.
c
Blesq logit regression model.
d
Blesq logit regression model with outlier elimination.

that is supposed to be the equation of a symmetrical sigmoid curve, 2.3.6. IBM SPSS Statistics Desktop19.0
where x represents the concentration of the sample, Y is the inhibi- The Statistical Package for Social Scientists (SPSS) is a compre-
tion ratio on a (0,1) scale, and two variables, a and b, together decide hensive statistical and data management package for analysts
the EC50 value and the slope of the curve. This program can be freely and researchers. Its statistical capabilities range from simple per-
downloaded at http://www.pharm.unipmn.it/rinaldi/software/ centages to complex analyses of variance, multiple regressions
blesq/BleSq.html (Locatelli et al., 2009). and general linear models, as well as to generate tabulated reports,
charts and plots of distributions and trends.
2.3.3. OriginPro 8.5 SPSS was originally designed for social science survey, but has
OriginPro is a versatile and resourceful software endowed with found wide applications in general statistics for a variety of scien-
tools for data analysis, publication-quality graphing and program- tific disciplines and businesses. Over the years, the developers
ming that is used in different scientific branches. Among the added many tests and diagnostics that greatly extended the range
suggested models, we used the pharmacological dose–response and capabilities of the base program (Norusis, 2002).
template for curve fitting, with a classic four-parameter formula
expressed as: 2.4. Actual EC50 determination
A2  A1
Y ¼ A1 þ Once the EC50 falls in a very narrow experimental range, it may
1 þ 10½ðlog x0 log xÞp be calculated from the data by using a simple mathematical meth-
where A1 is the baseline and A2 represents the maximum response; od based on the principle of right-angled triangle. To apply this
p is the slope of the curve and x0 is the concentration at the inflec- method, two assumptions have to be accepted: (1) that the maxi-
tion point, also recognised as EC50 value. mum response is reached, and (2) that the responses to the exper-
imental concentrations of the two recorded points on either side of
the 50% maximal response should be as close as possible to the
2.3.4. SigmaPlot 12
point of the 50% maximal response, in order to consider the
SigmaPlot 12 is a scientific graphing and data analysis software,
sigmoid curve as a straight line (Alexander, Browse, Reading, &
but presents more sigmoid fitting choices than OriginPro. Sigma-
Benjamin, 1999). As reported in results, Da and Db are two similar
Plot offers seamless Microsoft OfficeÒ integration, for an easy data
triangles, where the corresponding sides have lengths in the same
access from Microsoft ExcelÒ spreadsheets (Microsoft Corporation,
ratio. A, B, C, and D are known values from the experimental data;
Redmond, WA). It requires more expertise than the other programs
since we have already normalised the data in percentage, the EC50
and it does not express the EC50 values directly but through an
response is the 50% of the maximal response. Therefore, we applied
individual analysis of each single data point; thus, a further
the follow equation (Alexander et al., 1999):
calculation is necessary, which may be problematic and time con-
suming. There are different available equations for statistical ðA  50% max responseÞ  ðD  CÞ
assessment; in this article we have employed the one reported EC50 ¼ D 
ðA  BÞ
below:
The experiments were performed in triplicate, and the actual
a
Y ¼ Y0 þ EC50 was carried out by calculating the average value from three
ð1 þ e½ðxx0 Þ=b Þc different EC50 extrapolations. The results are reported in the last
where x0 is the concentration at the inflection point, c is the asym- row of Table 1.
metry factor, b determines the slope while Y0 and a determine the
min and max of Y values, respectively. 3. Results and discussion

2.3.5. BioDataFit 1.02 3.1. EC50 determination

BioDataFit is the only software we tested which is based on
java-script; it is specifically designed for biological research, freely Drug potency or efficacy may be measured in several ways;
downloadable and independent from the operating system. This many techniques, widely used in biology and pharmacology, rely
program uses a four-parameter formula for generating the curve upon the analysis of concentration-effect curves through the
and provides for loading the data as BLeSq software. Since it uses measurement of the drug concentration necessary to give the max-
the same equation of OriginPro, the statistical analysis given by imal effect (Gmax) or 50% of the maximum response (EC50) (Alexan-
this algorithm remains unchanged. der et al., 1999). Regarding DPPH assay, the EC50 is the antioxidant
 Z. Chen et al. / Food Chemistry 138 (2013) 414–420 417

concentration required to obtain a 50% radical inhibition. All the caffeic acid (a propenoic acid), chlorogenic acid (a quinic acid),
programs we have considered in this study provide the EC50 values and finally the aminoacid N-acetyl-cysteine. They are all standard
through a mathematical calculation whereby they generate a curve compounds frequently used in antioxidant assays, endowed with
fitting as close as possible to the experimental data points, using an a well-known antioxidant profile.
appropriate equation model. The algorithm adjusts the parameters Each compound was assayed at eight different concentrations,
by minimizing the sum-of-squares, to reach a more accurate good- within the range of 0.1–300 lmol/l. The concentrations were cho-
ness of fit to the data points (Motulsky & Ransnas, 1987). Since the sen according to their antiradical capacity towards the DPPH rad-
non-linear regression has a greater complexity than linear regres- ical on the basis of two pilot determinations and following the
sion, several factors such as the regression model, the initial esti- literature data (Nenadis, Lazaridou, & Tsimidou, 2007; Tabart, Ke-
mated value chosen and the weighting options may influence the vers, Pincemail, Defraigne, & Dommes, 2009). The results obtained
final results. from the DPPH assay are showed in Fig. 1: the graph above shows
In the present research, a wide range of antioxidant compounds the DPPH scavenging effect vs. the antioxidant concentration
were analysed to obtain various examples of antiradical activity vs. (log [mol/l]), where each point is linked to the next ones by a sim-
antioxidant concentrations curves. We tested natural compounds ple connecting line, while the graph below shows a five-parameter
belonging to different chemical families, such as quercetin (a flavo- transformation of the inhibition curve obtained with GraphPad
nol), (+)-catechin (a flavonoid), L-ascorbic acid (a hydroxyl acid), Prism, as an example of the statistical elaboration of the programs
we have considered. In fact, the experimental data were processed
by six different statistical programs to obtain estimated EC50 val-
ues (Table 1). However, these ones may be considered as theoret-
ical values, because they are derived from a range of antioxidant
concentrations, where experimental points are rather far from
the actual EC50 value. For this reason, in order to determine the
most reliable EC50 value, DPPH scavenging assay was still per-
formed for each antioxidant using several concentrations closer
to the estimated EC50. As an example, we report the experimental
procedure performed for catechin, a compound endowed with an
intermediate antioxidant capacity among the ones we have stud-
ied. Catechin showed an estimated EC50 value within the range
of 6.82–10.00 lmol/l (pD2 = 5.000–5.166, Table 1). In order to
determine the actual EC50 value, we chose five different concentra-
tions (6, 7, 8, 9 and 10 lmol/l) and we repeated in triplicate the
analysis (Fig. 2). The results demonstrated that the actual EC50 va-
lue of catechin really fell within the estimated range, but in a more
accurate approximation was closer to 8 lmol/l (pD2 = 5.096). The
evaluation of the antiradical curve stabilized in a smaller range
of antioxidant concentrations, as near as possible to the estimated
EC50 value, enables a more accurate specification of the EC50 for the
mathematical interpretation. For this last point, to perform a more
accurate analysis, we enclosed the EC50 value within a narrow
range. This approximation might represent a limitation for the
analysis, because the setting of an EC50 interval instead of a single
point seemed incompatible for a comparative evaluation among
different statistical programs. For this reason a simple, accurate,
and speedy non-computational technique for the EC50 calculation
is required. Once the EC50 fell in a narrow range, it may be calcu-
lated by using a simple mathematical method based on the princi-
ple of right-angled triangle (Alexander et al., 1999). Fig. 3 shows
the actual EC50 obtained for catechin and Table 2 reports the values
used in the calculation. In the same way, we applied this method to
all the antioxidant compounds to obtain the actual EC50 values (Ta-
ble 1, last row). Following this experimental strategy, the order of
potency for the reference compounds was quercetin > chlorogenic
acid > catechin > caffeic acid > ascorbic acid > acetylcysteine.

3.2. Comparative data analysis

Table 1 shows the EC50 values of the antioxidants calculated

with several programs and different statistical options; among
these, GraphPad Prism five-parameter model gave the same results
of SigmaPlot processing, as for BioDataFit and OriginPro (because
of the common equation they use). For this reason, there are no
real differences among these programs for the EC50 calculation:
values depend on the choice of the equation used in the regression
Fig. 1. Antiradical activity (%) of antioxidant compounds: above, XY scatters with
connecting lines; below, XY curves with a five-parameter regression transforma-
model. Furthermore, for a detailed comparison, we estimated the
tion, using GraphPad Prism. Data are expressed as mean ± SD of three independent goodness of the regression for the programs, adopting the follow-
experiments for each antioxidant; undisclosed SDs fall within respective symbols. ing equation:
418 Z. Chen et al. / Food Chemistry 138 (2013) 414–420

Fig. 2. In the antiradical curve of catechin above, the parallel dotted lines define the
concentration range for the determination of the actual EC50, graph below (see
results). Data are expressed as mean ± SD of three independent experiments;
undisclosed SDs fall within respective symbols.

1X N
r2 ¼ ðxi  lÞ2 Fig. 3. Above: EC50derivation from the concentration–response curve of catechin,
N i¼1 obtained by GraphPad Prism five-parameter regression, with highlighted the
nearest actually recorded responses (A and B) of each experimental concentration
(C and D) on either side of the EC50 forming a right-angled triangle, method of
where l is the actual EC50, N is the number of reference compounds Alexander et al. (1999). Below: graphic magnification of experiment No. 1 (Table 2)
(i.e. 6) and xi is the estimated EC50 value for each antioxidant taken showing the EC50 obtained for catechin using the right-angled triangle method.
from Table 1. The relative variance of the estimated EC50 of each
antioxidant was calculated for the different statistical programs
(Table 3); SigmaPlot and GraphPad Prism 5P implemented with a
Table 2
five-parameter equation showed a minor variance and seemed to
The EC50 values obtained for (+)-catechin using the right-angled triangle method
work with a better approximation in relation to actual EC50 values. (Alexander et al., 1999), expressed as log of the antioxidant concentration.
In order to determine correctly the EC50 values, data have to be
log [antioxidant] Experiment No. Mean ± SD
interpolated by an appropriate curve. However, the goal of non-lin-
ear regression is not only the fitting of a model to data, but also 1 2 3
interpreting them, thus the choice should not be based only on 5.22 39.36% 41.58% 39.04% 40.00 ± 1.38%
the shape of the graph. Various antioxidant mechanisms are 5.15 45.09% 47.88% 43.19% 45.39 ± 2.36%
5.10 49.30% 50.01% 49.99% 49.77 ± 0.40%
involved in DPPH radical scavenging assay, and they are not the
5.05 54.63% 56.35% 54.35% 55.11 ± 1.08%
same for every compound; for this reason we studied several anti- 5.00 56.71% 60.51% 56.17% 57.80 ± 2.37%
oxidants to find the appropriate statistical program of general EC50 5.090 5.097 5.097 5.095 ± 0.004
application in DPPH assay. During the regression we also noticed
that not all the programs were performing in the same way; some
of them, like GraphPad Prism, show an exhaustive list of the values
for all the parameters together with EC50, even if it is not present in as a sigmoidal curve, after the logarithmic transformation of the
the equation as an index. Whereas other programs, such as Origin- horizontal axis. Hyperbolic dose–response curves on linear scale
Pro and SigmaPlot, show only parameters listed within the are considered less convenient to work out than the logarithmic-
function. transformed functions, because they may be graphically difficult
The DPPH radical scavenging assay of an antioxidant may be to represent if the curve spans many orders of drug dose magni-
plotted as a typical rectangular hyperbole (defined by the tude (Christopoulos, 1998). The standard sigmoidal dose–response
inhibition ratio in relation with the antioxidant concentration), or model (based on the Hill equation) refers on the assumption that
 Z. Chen et al. / Food Chemistry 138 (2013) 414–420 419

Table 3 as an efficient statistical strategy for curve-fitting, EC50 determina-

Variance obtained from a comparative statistical analysis of antioxidant standard tion and data processing.
compounds with different statistical programs.
Differences in results among statistical processing programs re-
Software Variance vealed the need of an appropriate regression model as a preferen-
GraphPada 0.007 tial choice, even for other chemical or biological antioxidant
GraphPad 5Pb 0.003 assessments.
Blesqc 0.005
Blesqd 0.007
OriginPro 0.009
SigmaPlot 0.003 Acknowledgements
BioDataFit 0.009
SPSS 0.007 This research was supported by a Grant MIUR 2009-2010
a
GraphPad log (inhibitor) vs. normalised response model (variable slope). (Italy).
b
GraphPad five-parameter regression model. The authors are grateful to Dr. R. Sato and Mrs. F. Chinaglia for
c
Blesq logit regression model. reference assistance.
d
Blesq logit regression model with outliers elimination.
The interpretation of the results obtained in this research only
reflects the authors’ view; the authors report no conflict of interest,
financial or otherwise.
the log dose vs. response curve is symmetrical around its midpoint,
even if in some cases dose–response curves are not symmetrical.
Regarding this point, Van der Graaf and Schoemaker (1999) References
showed that the application of the Hill equation to asymmetric
dose–response data can lead to a quite erroneous assessment of Alexander, B., Browse, D. J., Reading, S. J., & Benjamin, I. S. (1999). A simple and
drug potency. These authors suggested an alternative model, accurate mathematical method for calculation of the EC50. Journal of
Pharmacological and Toxicological Methods, 41(2–3), 55–58.
known as Richards equation, which could provide a more suitable Apak, R., Güçlü, K., Ozyürek, M., & Karademir, S. E. (2004). Novel total antioxidant
fit to asymmetric dose–response data by adding a further parame- capacity index for dietary polyphenols and vitamins C and E, using their cupric
ter, S, which quantifies the asymmetry. Consequently, this equa- ion reducing capability in the presence of neocuproine: CUPRAC method.
Journal of Agricultural and Food Chemistry, 52(26), 7970.
tion is sometimes referred to a five-parameter logistic equation Armitage, P., Berry, G., & Matthews, J. N. S. (2005). Statistical methods in medical
and is the equation implemented in GraphPad and SigmaPlot. Dur- research. Oxford: Blackwell Science.
ing our comparative study of statistical programs, we also noticed Benzie, I. F., & Strain, J. J. (1996). The ferric reducing ability of plasma (FRAP) as a
measure of ‘‘antioxidant power’’: the FRAP assay. Analytical Biochemistry, 239(1),
that Xb is the inflection point and EC50 is the point where the ago- 70–76.
nist effect reaches 50%, but they are not necessarily concurrent, Berkson, J. (1944). Application to the logistic function to bio-assay. Journal of the
though they are very near in horizontal axis in most of the cases, American Statistical Association, 39(227), 357–365.
Bliss, C. I. (1935). The calculation of the dosage–mortality curve. Annals of Applied
using similar equation for regression. GraphPad calculates the ac- Biology, 22(1), 134–167.
tual EC50 values while SigmaPlot just calculates the x values of Blois, M. S. (1958). Antioxidant determinations by the use of a stable free radical.
the inflection point as EC50. Nature, 181(4617), 1199–1200.
Blokhina, O., Virolainen, E., & Fagerstedt, K. V. (2003). Antioxidants, oxidative
Five-parameter model assumes that the SD of the scatter is the
damage and oxygen deprivation stress: A review. Annals of Botany, 91(2),
same for all x values (homoscedastic data). In the example of a 179–194.
standard curve, the SD for all the replicates of a low sample con- Bondet, V., Brand-Williams, W., & Berset, C. (1997). Kinetics and mechanisms of
centration is equal to the SD of the replicates for the high concen- antioxidant activity using the DPPH free radical method. LWT – Food Science
and Technology, 30(6), 609–615.
tration, but this is rarely true in a dose–response curve fitting; it is Brand-Williams, W., Cuvelier, M. E., & Berset, C. (1995). Use of a free radical method
recommended applying weighting algorithms for five-parameter to evaluate antioxidant activity. LWT – Food Science and Technology, 28, 25–30.
curve fitting. In the present study, we tried it, but without any Cao, G., Alessio, H. M., & Cutler (1993). Oxygen-radical absorbance capacity assay
for antioxidants. Free Radical Biology & Medicine, 14(3), 303–311.
improvement of EC50 elaboration. Chevion, S., Roberts, M. A., & Chevion, M. (2000). The use of cyclic voltammetry for
the evaluation of antioxidant capacity. Free Radical Biology & Medicine, 28(6),
860–870.
Christopoulos, A. (1998). Assessing the distribution of parameters in models of
4. Conclusions ligand–receptor interaction: To log or not to log. Trends in Pharmacological
Sciences, 19(9), 351–357.
These findings underscore the importance of an appropriate Eklund, P. C., Långvik, O. K., Wärnå, J. P., Salmi, T. O., Willför, S. M., & Sjöholm, R. E.
(2005). Chemical studies on antioxidant mechanisms and free radical
analysis for the interpretation of data obtained with DPPH scav- scavenging properties of lignans. Organic & Biomolecular Chemistry, 3(18),
enging assay; since a wide scientific production on the reliability 3336–3347.
of the most commonly used statistical packages is available, the Gil, M. I., Tomás-Barber, F. A., Hess-Pierce, B., Holcroft, D. M., & Kader, A. A. (2000).
Antioxidant activity of pomegranate juice and its relationship with phenolic
present study is focused on the definition of the most accurate sta- composition and processing. Journal of Agricultural and Food Chemistry, 48(10),
tistical program for EC50 determination, which is generally used as 4581–4589.
an indicator of antioxidant capacity for plant extracts and pure Glazer, A. N. (1990). Phycoerythrin fluorescence-based assay for reactive oxygen
species. Methods in Enzymology, 186, 161–168.
compounds. In the present research, we performed experimental Knudsen, L. F., & Curtis, J. M. (1947). The use of the angular transformation in
and statistical studies using single antioxidant compounds since biological assays. Journal of the American Statistical Association, 42(238),
the implications are conveyable to mixtures of compounds such 282–296.
Krishnaiah, D., Sarbatly, R., & Nithyanandam, R. (2011). A review of the antioxidant
as plant extracts and foods. potential of medicinal plant species. Food and Bioproducts Processing, 89(3),
GraphPad Prism is a flexible and worldwide used software for ba- 217–233.
sic biostatistics, curve fitting and scientific graphics, and is almost Locatelli, M., Gindro, R., Travaglia, F., Coïsson, J.-D., Rinaldi, M., & Arlorio, M. (2009).
Study of the DPPH center dot-scavenging activity: Development of a free
exclusively developed for biological and pharmacological use.
software for the correct interpretation of data. Food Chemistry, 114(3), 889–897.
Among the programs considered in the present study, GraphPad Miller, N. J., Rice-Evans, C., Davies, M. J., Gopinathan, V., & Milner, A. (1993). A novel
Prism 5P was the only that could easily load all datasets together method for measuring antioxidant capacity and its application to monitoring
and could plot them on the same graphic, with an automatic update the antioxidant status in premature neonates. Clinical Science, 84(4), 407–412.
Mishra, K., Ojha, H., & Chaudhury, N. K. (2012). Estimation of antiradical properties
after every data changing in the spreadsheet cells. From the results of antioxidants using DPPH assay: A critical review and results. Food Chemistry,
presented above, we suggest the five-parameter regression model 130, 1036–1043.
420 Z. Chen et al. / Food Chemistry 138 (2013) 414–420

Motulsky, H. J., & Ransnas, L. A. (1987). Fitting curves to data using nonlinear Sharma, O. P., & Bhat, T. K. (2009). DPPH antioxidant assay revisited. Food Chemistry,
regression: A practical and nonmathematical review. FASEB Journal, 1(5), 113(4), 1202–1205.
365–374. Tabart, J., Kevers, C., Pincemail, J., Defraigne, J.-O., & Dommes, J. (2009). Comparative
Nenadis, N., Lazaridou, O., & Tsimidou, M. Z. (2007). Use of reference compounds in antioxidant capacities of phenolic compounds measured by various tests. Food
antioxidant activity assessment. Journal of Agricultural and Food Chemistry, Chemistry, 113(4), 1226–1233.
55(14), 5452–5460. Valko, M., Leibfritz, D., Moncol, J., Cronin, M. T., Mazur, M., & Telser, J. (2007). Free
Norusis, M. J. (2002). SPSS 11.0 guide to data analysis. Upper Saddle River, NJ: radicals and antioxidants in normal physiological functions and human disease.
Prentice Hall. The International Journal of Biochemistry & Cell Biology, 39(1), 44–84.
Prior, R. L., Wu, X., & Schaich, K. (2005). Standardized methods for the Van der Graaf, P. H., & Schoemaker, R. C. (1999). Analysis of asymmetry of agonist
determination of antioxidant capacity and phenolics in foods and dietary concentration–effect curves. Journal of Pharmacological and Toxicological
supplements. Journal of Agricultural and Food Chemistry, 53(10), 4290–4302. Methods, 41(2–3), 107–115.
Scalbert, A., Johnson, I. T., & Saltmarsh, M. (2005). Polyphenols: Antioxidants and Villano, D., Fernández-Pachon, M. S., Troncoso, A. M., & García-Parrilla, M. C. (2005).
beyond. The American Journal of Clinical Nutrition, 81(1), 215S–217S. Comparison of antioxidant activity of wine phenolic compounds and
metabolites in vitro. Analytica Chimica Acta, 538(1–2), 391–398.

View publication stats