http://informahealthcare.com/dctISSN: 0148-0545 (print), 1525-6014 (electronic)

Drug Chem Toxicol, Early Online: 1–6! 2014 Informa Healthcare USA, Inc. DOI: 10.3109/01480545.2014.922096

RESEARCH ARTICLE

Chemometrics models for assessment of oxidative stress risk inchrome-electroplating workers

Rezvan Zendehdel1,2, Seyed Vahid Shetab-Boushehri3,4, Mansoor R. Azari1, Vajihe Hosseini1, andHamidreza Mohammadi5,6

1Department of Occupational Hygiene, School of Public Health, Shahid Beheshti University of Medical Sciences, Tehran, Iran, 2Occupational Hygiene

Laboratory, Deputy Chancellor of Health, Shahid Beheshti University of Medical Sciences, Tehran, Iran, 3Department of Medical Nanotechnology,

School of Medicine, Iran University of Medical Sciences, Tehran, Iran, 4Razi Drug Research Center, Iran University of Medical Sciences, Tehran, Iran,5Department of Toxicology and Pharmacology, Faculty of Pharmacy, Mazandaran University of Medical Sciences, Sari, Iran, and 6Center for Air

Pollution Research (CAPR), Institute for Environmental Research (IER), Tehran University of Medical Sciences, Tehran, Iran

Abstract

Oxidative stress is the main cause of hexavalant chromium-induced damage in chromeelectroplating workers. The main goal of this study is toxicity analysis and the possibility oftoxicity risk categorizing in the chrome electroplating workers based on oxidative stressparameters as prognostic variables. We assessed blood chromium levels and biomarkers ofoxidative stress such as lipid peroxidation, thiol (SH) groups and antioxidant capacity of plasma.Data were subjected to principle component analysis (PCA) and artificial neuronal network(ANN) to obtain oxidative stress pattern for chrome electroplating workers. Blood chromiumlevels increased from 4.42 ppb to 10.6 ppb. Induction of oxidative stress was observed byincreased in lipid peroxidation (22.38 ± 10.47 mM versus 14.74 ± 4.82mM, p50.0008), decreasedplasma antioxidant capacity (3.17 ± 1.35 mM versus 7.74 ± 4.45 mM, p50.0001) and plasma totalthiol (SH groups) (0.21 ± 0.07 mM versus 0.45 ± 0.41 mM, p50.0042) in comparison to controls.Based on the oxidative parameters, two groups were identified by PCA methods. One categoryis workers with the risk of oxidative stress and second group is subjects with probable risk ofoxidative stress induction. ANN methods can predict oxidative-risk category for assessmentof toxicity induction in chrome electroplaters. The result showed multivariate modeling canbe interpreted as the induced biochemical toxicity in the workers exposed to hexavalentchromium. Different occupation groups were assessed on the basis of risk level of oxidativestress which could further justify proceeding engineering control measures.

Keywords

Artificial neuronal network, chromeelectroplaters, oxidative stress, principlecomponent analysis

History

Received 18 February 2014Revised 29 April 2014Accepted 5 May 2014Published online 27 May 2014

Introduction

The toxicity of hexavalent chromium (Cr VI) is one of the

major health concerns in recent years. Occupational exposure

to Cr (VI) cause ulcerations, chronic bronchitis, decreased

pulmonary function and pneumonia (Bradshaw et al., 1998).

On the other hand, epidemiologic evidences show that Cr (VI)

compounds are human carcinogen (Caglieri et al., 2006;

Linos et al., 2011). It is considered as a group 1 human

carcinogen by the International Agency for the Research on

Cancer (De Flora, 2000). Occupational exposure to Cr (VI)

occurs mainly in the working groups such as chromium

compound manufacturing, chrome electroplating, and leather

tanning, and welding (Myers & Myers, 2009).

Oxidative stress is defined as a toxicity mechanism in

which reactive oxygen species (ROS) are actively generated.

Excess levels of ROS disrupt cell macromolecules and lead

to critical damage in biological function; metal-induced

lung damage and bronchitis associated with inflammation,

and oxidative stress (Bartoli et al., 2011; Fischer et al., 2011;

Rahman, 2008). Moreover, reports have shown that Cr (VI)-

induced oxidative stress was linked to DNA damage, altered

gene expression and carcinogenicity (Bagchi et al., 2002;

Grass et al., 2010; Zhang et al., 2011). Oxidative stress

induction with Cr (VI) was reported by assessing factors such

as Ferric reducing ability of plasma (FRAP), superoxide

dismutase (SOD), ascorbate peroxidase (APX), and reduced

glutathione content (GSH) in biological samples (Długosz

et al., 2012; Liao et al., 2012; Sawicka et al., 2008; Yao et al.,

2008; Zendehdel et al., 2012).

Multivariate methods such as principal components ana-

lysis (PCA) and artificial neural networks (ANN) have been

used as a low biased tool for the analysis and interpretation of

complex datasets. Some authors challenged multivariate

methods for analyzing such datasets (Sawicka et al., 2008;

Address for correspondence: Dr. Hamidreza Mohammadi, Department ofToxicology and Pharmacology, Faculty of Pharmacy, MazandaranUniversity of Medical Sciences, PO Box- 48175/861, Sari, Iran. Tel:+98 151 3543082. Fax: +98 151 3543085. E-mail: hmohammadi@farabi.tums.ac.ir

Dru

g an

d C

hem

ical

Tox

icol

ogy

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Mem

oria

l Uni

vers

ity o

f N

ewfo

undl

and

on 0

7/11

/14

For

pers

onal

use

onl

y.

Zendehdel et al., 2012). PCA and ANN have successfully

been applied to various environmental, chemical, biological,

and ecotoxicological studies (Antolini et al., 2005;

Hertsenberg et al., 2007; Sinha et al., 2009; Sirven et al.,

2006; Vermeulen et al., 2004).

The aim of this study was to explore oxidative stress in

chrome electroplating worker through application of ferric-

reducing ability of plasma (FRAP), thiol (SH) content and

lipid peroxidation of plasma. PCA analysis was applied to

classify workers in different groups in cases where their

oxidative stress correlated. Moreover, oxidative stress param-

eters and blood chromium concentration followed by ANN

analysis reveal oxidative stress pattern in chrome electroplat-

ing workers.

Materials and methods

Chemicals

In this research dithionitrobenzoic acid (DTNB), tris base, 1,

1, 3, 3-tetraethoxypropane, 2-thiobarbituric acid (TBA), 2, 4,

6-tripyridyl-s-triazine (TPTZ), ferric chloride, ethylenediami-

netetraacetic acid (all from Sigma Chemical Co. (USA),

n-butanol, and chromium standards, nitric acid, perchloric

acid, magnesium nitrate, (Merck, Germany) all in extra pure

grade were used.

Human subjects

In this study, male chrome electroplaters (n¼ 30) were

selected in 7 workshops in Tehran. Age of workers was

35 ± 9.6 years and they had work history between 1 to 10

years. Age and sex was matched in controls (n¼ 30) from

dairy production workshops who were not occupationally

exposed to Cr (VI) or any other physical or chemical

hazardous compounds. Control and exposed population were

matched in socioeconomically statue. All subjects were

required to fill a questionnaire considering their home

addresses, salaries, states of health’s, use of medications,

history of occupation, and alcohol consumption. Based on the

questionnaire, subjects with history of cigarette smoking in

the last year or use of drugs were excluded from the study.

Tehran University of Medical Sciences (TUMS) Institutional

Review Board (IRB) approved all study procedures. Five

milliliter of heparinized blood sample was obtained from

selected subjects.

Assessment methods

Blood chromium

Blood samples were ashed and digested for chromium

analysis (Eller & Cassinelli, 1994; Nduka & Orisakwe,

2009). Digestion was done by addition of 10 mL concentrated

acids (3:1 (v/v) Nitric acid: Perchloric acid) to blood samples.

They were then heated on a hot plate at 200 �C to dryness.

Then 0.5 ml deionized water was added, stirred

and filtered. Chromium concentration was assayed with a

flameless atomic absorption spectrophotometer (Canadian

AL2200 Aurora spectrometer). Limit of chromium detection

in blood (LOD) was 0.6 mg/L and the quantification limit

was 1 mg/L. The recovery of standard addition samples was

96–99.5%.

Ferric reducing ability of plasma (FRAP)

The FRAP test was carried out to determine antioxidant

capacity of plasma samples by measuring the reduction

of Fe+3 to Fe+2 In this test, plasma is exposed to Fe+3 and the

antioxidants present in plasma reduce it to Fe+2. The reaction

mixture contains of 2,4,6-tripyridyl-s-triazine (TPTZ), acetate

buffer (pH 3.6) and FeCl3. The complex between TPTZ and

Fe+2 creates a blue color and read by spectrophotometer

at 593 nm (Mohammadi et al., 2011; Ranjbar et al., 2002).

Total SH groups of plasma

Total SH content of plasma was determined by visible

spectrophotometer at 412 nm. The reagent contained DTNB

and Tris-EDTS buffer (Tris base, ethylenediamine tetraacetic

acid (pH¼ 8.2) (Mohammadi et al., 2011).

Lipid peroxidation of plasma

For this assay, trichloroacetic acid (20%) and TBA solutions

(0.67% in sodium sulfate 2 M) were added to plasma samples.

The mixture is heated in a boiling water bath for 30 min.

The resulting color complex was extracted by n-butyl alcohol

and the absorbance of the organic phase is measured at the

wavelength of 530 nm (Mohammadi et al., 2011).

Data analysis

Statistical analysis

Statistical analysis was performed using the JMP-7 (2010

version) software. The results are expressed as means ± stand-

ard deviation. The difference between subject and control

groups was achieved with two samples student t-test in

normal distribution of samples. If data haven’t normal-like

distribution Mann–Whitney U test was used. p Values50.05

were found to be significant.

Principle component analysis

Method of PCA applied for classification and dimension

reduction of data set, whilst retaining as much as possible the

variation present in the original predictor variables (Sawicka

et al., 2008; Zendehdel et al., 2012). In this research, the

results of assessment among the exposed group containing

blood chromium concentration and oxidative stress param-

eters have been used as data set for the PCA analysis using

MATLAB (2010 a) software. PCA curved four measured

parameters for all subjects in four-dimensional plot. The best

lines in this plot create principle components (PC) in this

analysis. Based on the selected PC, subjects were categorized

in several groups. Thus PCA classifies exposed subjects based

on the measured parameters.

Artificial neural networks analysis

Artificial neural networks (ANN) are computerized models

capable of identifying complex nonlinear relationships

between input and output data sets (Park et al., 2004).

Neurons are the main processing units in ANN modeling that

process the data using a variety of mathematical functions.

Neurons are organized in parallel layers: input, hidden (single

2 R. Zendehdel et al. Drug Chem Toxicol, Early Online: 1–6

Dru

g an

d C

hem

ical

Tox

icol

ogy

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Mem

oria

l Uni

vers

ity o

f N

ewfo

undl

and

on 0

7/11

/14

For

pers

onal

use

onl

y.

or multiple), and output (Zendehdel et al., 2012). In this

analyzing a weight was assigned to each data. ANN could be

learn and adapt the selected weight to predict the output layer.

In this study, multiple layer neuronal networks were

designed using MATLAB software. The same data sets used

for PCA analysis was used as input layer where 20% of data is

prepared for testing model. The number of output neurons

related to groups was categorized by PCA modeling.

Results

Oxidative stress parameters

Distribution of data for blood chromium concentration, FRAP

and Thiol group was normal. Three samples of control people

have lower chromium concentration than LOD. Chrome

electroplaters contained significantly higher levels of blood

chromium concentration (p50.0001) than control population.

The mean ± SD values in plating workers and control

were 5.97 ± 1.74 ng/ml and 4.22 ± 0.08 ng/ml, respectively

(Table 1).Ability of plasma for ferric reducing in exposed

group was significantly (p50.0001) lower than control

group (p50.0001). The mean ± SD values of electroplating

workers and control were 3.17 ± 1.35 mM and 7.74 ± 4.45 mM,

respectively (Table 1). The concentrations of two thiol

samples were not detectable. Total thiol groups of plating

workers (mean ± SD: 0.21 ± 0.07mM) were also significantly

(p50.0042) lesser than that of control group (mean ± SD:

0.45 ± 0.41mM). Plasma lipid peroxidation level was signifi-

cantly (p50.0008) higher in chrome electroplating workers

than control (Table 1). The mean ± SD values for workers and

control were 22.38 ± 10.47mM versus 14.74 ± 4.82 mM.

PCA analysis

The data matrix formed by four measured parameters (blood

chromium concentration, FRAP, thiol content, lipid peroxida-

tion) in the 30 chrome electroplating workers (Table 2). PCA

can be used to extract groups of people with the most

correlation between measured parameters. PC Score plot

provide visualization of the all the measured data for each

person. Figure 1 shows the data plot in a 2-dimensional

projection derived from the first and third PC scores. Each

spot in this plot summarized oxidative stress features for each

chrome electroplater with different chromium exposure. The

subjects in the left area of the score plot in the circled domain

were named as group 1 where other subjects of the projection

are second category. Different templates of PCs were plotted

where PC1 and PC3 has the best data classification (there

was not shown). Blood chromium concentration in the first

category was 7.92 ± 1.7 mg/L in the second group was

5.01 ± 1.74 mg/L. It seems, blood chromium values in group

2 is significantly (p50.0001) lower than group 1. Since

ability of plasma for ferric reducing in group1 (2.3 ± 1.1 mg/L)

was significantly (p50.01) lower than group 2 (3.5 ± 1.02 mg/

L) and plasma lipid peroxidation level was significantly

(p50.04) higher in workers of group 1 (24.95 ± 7.02mg/L)

than second category (19.5 ± 5.4 mg/L).

Based on this approach, PCA as unsupervised model

provide two different groups for representing the variety of

oxidative stress parameters in chromium exposed workers.

First category is the workers with significantly greater

oxidative parameters than group 2.

Loading plot of data in PCA analysis determine the

parameter which have the greatest effect for discrimination.

The Loading plot of PC1 from four measured parameters

was observed in Figure 1. This diagram shows plasma thiol

content have the greatest impact for group discrimination.

Artificial neural network

Artificial neural network (ANN) was applied to anticipate the

subjects of PCA categorized group. The same data sets used

for PCA analysis were used for input layer. The output layer

consisted of two output neurons, one to classify the first

category and the other for second group. We ran ANN on the

dataset applying Feed-forward back propagation using

Levenbery-Marqwardt as the training algorithm. To optimize

the structure of the networks, the selected error goal was

0.0005. The parameters of the optimized neural network are

listed in Table 3. When the model is accomplished for the

training dataset, the risk pattern of oxidative stress for testing

the dataset is predicted using the learned rules derived from

the ANN model. The result showed that ANN model is able

to anticipate 100% of training and testing data set correctly.

Discussion

This work was based on the fact that the main cause of

chromium-induced cellular damages is through ROS gener-

ation and oxidative stress (Bagchi et al., 2002; Grass et al.,

2010; Zhang et al., 2011). In the present work, FRAP, SH

content and lipid peroxidation of plasma as oxidative stress

biomarkers were studied. Classifying oxidative stress among

the exposure subjects predict categorized people as higher

priority group for monitoring. With regards to identifying

different occupation groups and their prioritization based

on the risk of oxidative damage, management control and

engineering control measures could be justified. Other studies

exhibited chromium induced oxidative stress through param-

eters such as gluthathione-thiyl radicals (Myers & Myers,

2009), FRAP (Grass et al., 2010) and lipid peroxidation

(Długosz et al., 2012). The results of this study demonstrated,

Table 1. Blood Chromium concentration and plasma Oxidative stressparameters in control and chrome-plating workers.

Control(n¼ 30)

Workers(n¼ 30) t-test

Mean ± SD Mean ± SD t df p

TBAR (mM) 14.74 ± 4.82 22.38 ± 10.47 3.6 40 50.0008FRAP (mM) 7.74 ± 4.45 3.17 ± 1.35 5.3 34 50.0001Thiol group (mM) 0.45 ± 0.41 0.21 ± 0.07 3.09 30 50.0042Cr Concentration (PPb) 4.22 ± 0.08 5.97 ± 1.74 5.5 29 50.0001

Table 2. Eigen value of covariance.

PC Eigen value

1 13.22 10.63 0.944 0.68

DOI: 10.3109/01480545.2014.922096 hemometrics models for assessment of oxidative stress risk 3

Dru

g an

d C

hem

ical

Tox

icol

ogy

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Mem

oria

l Uni

vers

ity o

f N

ewfo

undl

and

on 0

7/11

/14

For

pers

onal

use

onl

y.

blood chromium concentration and lipid peroxidation in

plasma were significantly higher, however, ferric reducing

ability of plasma and plasma thiol content was lower among

electroplating workers compared to control group (Table 1).

In this study, the concentration of blood chromium

monitoring in electroplating workers were obtained

to10.6 ppb which were higher than the amount reported in

previous studies. These differences may be due to electro-

plating process difference and types of hygiene control

suggested in these two studies. As stated by Zhang et al.

blood chromium concentration about 4.41 ppb could be

produce DNA oxidation (Zhang et al., 2011). Our results

confirm oxidative damage in lipids and proteins.

PCA is a well-known approach of chemometric methods.

One of the advantages of this method is reducing of the

dimensionality of a data set. Although most of the statistical

approach considers the mean of data, all of data was applied

by this analysis individually. The results of this study

highlighted the potential of PCA modeling for discrimination

of exposed subjects with higher oxidative parameters.

First category is the exposed subjects with significantly

higher oxidative stress that emphasized as people with the

risk of damage induction (in risk group: R). Otherwise,

second group are the people with probably in risk group (PR)

who are subjects with lower oxidative parameters. It was

suggested that proceeding engineering controls have to be

prepared for R subjects as soon as possible.

In other studies, PCA method used for determination

relationships between oxidative stress parameters in different

patients (Vermeulen et al., 2004). Also, several studies used

PCA method for relationships between different compounds

in different groups (Sinha et al., 2009; Zendehdel et al., 2012).

Some authors challenged with multivariate analysis to

discriminate chromium-induced oxidative stress where chro-

mium was supplied in plant samples (Sirven et al., 2006).

Although they only used of specific selected chromium

concentrations and oxidative stress were investigated by

researcher interferences, but in our study oxidative stress were

determined in human chromium exposed without any inter-

vention. PCA analysis can be used to identify significant

variables which are responsible for the discrimination of

oxidative-risk in workers (Figure 2). Moreover, ANN

modeling as supervised method predicts worker groups with

different oxidative-risk grade which could justify proceeding

engineering control program.

Evaluation of human health risk has become a worthy

subject of great concern throughout the world. In conventional

risk assessment, environmental concentrations of chemicals

such as acceptable daily intake (ADI) and reference doses

(RD) are used for evaluation of health risks (Satoh, 1978).

The using of RD is required air monitoring of exposure and

ADI considered oral intake which they were not included all

of route exposure. In this study biological monitoring was

used to discriminate groups by different oxidative stress risk

evaluation. Exposure assessment in biological samples such

as blood chromium concentration and oxidative parameters

consider oral, skin and respiratory intake. So risk assessment

based on the biological monitoring has lower bias than air

evaluation.

To estimate excess lifetime risk of lung cancer death, some

authors challenged cumulative exposure to hexavalent chro-

mium (Park et al., 2004). Throughout recent years; a variety

of biomarker concentration has been developed in a risk

assessment (Bailer & Hoel, 1989; Calafat & McKee, 2006;

Cox Jr, 1996). This study could be regarded as pioneer

investigation for development of advanced risk assessment

with internal dose measurement.

In conclusion multivariate modeling can be interpreted the

induction of biochemical toxicity in the workers exposed to

Figure 1. Score plot of PCA analysis forchromium induced oxidative stress in thechrome electroplating workers.

Table 3. Optimized neuronal network parameters.

Error goal 0.0005Transfer function of hidden layer LogsigNumber of hidden nodes 2Training algorithm Levenberg-MarquardtMu 0.001Mu increase 10Mu decrease 0.1

4 R. Zendehdel et al. Drug Chem Toxicol, Early Online: 1–6

Dru

g an

d C

hem

ical

Tox

icol

ogy

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Mem

oria

l Uni

vers

ity o

f N

ewfo

undl

and

on 0

7/11

/14

For

pers

onal

use

onl

y.

hexavalent chromium. Different occupation groups were

assessed on the basis of risk level of oxidative stress which

could further justify proceeding engineering control

measures.

Declaration of interest

The authors declared no conflicts of interest.

Financial support for this work was provided by Center for

Air Pollution Research (CAPR), Institute for Environmental

Research (IER), Tehran University of Medical Sciences,

Tehran, Iran with reference number 92-01-46-21287.

References

Antolini F, Valente F, Ricciardi D, et al. (2005). Principal componentanalysis of some oxidative stress parameters and their relationshipsin hemodialytic and transplanted patients. Clin Chim Acta 358:87–94.

Bagchi D, Stohs SJ, Downs BW, et al. (2002). Cytotoxicity and oxidativemechanisms of different forms of chromium. Toxicology 180:5–22.

Bailer AJ, Hoel DG. (1989). Metabolite-based internal doses used ina risk assessment of benzene. Environ Health Perspect 82:177–184.

Bartoli M, Novelli F, Costa F, et al. (2011). Malondialdehyde in exhaledbreath condensate as a marker of oxidative stress in differentpulmonary diseases. Mediators Inflammation 2011:Article ID891752, 7 pages.

Bradshaw LM, Fishwick D, Slater T, Pearce N. (1998). Chronicbronchitis, work related respiratory symptoms, and pulmonaryfunction in welders in New Zealand. Occup Environ Med 55:150–154.

Caglieri A, Goldoni M, Acampa O, et al. (2006). The effect of inhaledchromium on different exhaled breath condensate biomarkers amongchrome-plating workers. Environ Health Perspect 114:542–546.

Calafat AM, McKee RH. (2006). Integrating biomonitoring exposuredata into the risk assessment process: phthalates [diethyl phthalate anddi (2-ethylhexyl) phthalate] as a case study. Environ Health Perspect114:1783–1789.

Cox Jr LA. (1996). Reassessing benzene risks using internal doses andMonte-Carlo uncertainty analysis. Environ Health Perspect 104:1413–1429.

De Flora S. (2000). Threshold mechanisms and site specificity in chro-mium (VI) carcinogenesis. Carcinogenesis 21:533–541.

Długosz A, Rembacz KP, Pruss A, et al. (2012). Influence of chromiumon the natural antioxidant barrier. Pol J Environ Stud 21:331–335.

Eller PM, Cassinelli ME. (1994). NIOSH manual of analytical methods.DIANE Publishing.

Fischer BM, Pavlisko E, Voynow JA. (2011). Pathogenic triad in COPD:oxidative stress, protease–antiprotease imbalance, and inflammation.Int J Chron Obstruct Pulmon Dis 6:413–421.

Grass DS, Ross JM, Family F, et al. (2010). Airborne particulate metalsin the New York City subway: a pilot study to assess the potential forhealth impacts. Environ Res 110:1–11.

Hertsenberg S, Brouwer D, Lurvink M, et al. (2007). Quantitativeself-assessment of exposure to solvents among shoe repair men.Ann Occup Hyg 51:45–51.

Liao W-T, Huang T-S, Chiu C-C, et al. (2012). Biological propertiesof acidic cosmetic water from seawater. Int J Mol Sci 13:5952–5971.

Linos A, Petralias A, Christophi CA, et al. (2011). Oral ingestionof hexavalent chromium through drinking water and cancer mortalityin an industrial area of Greece – an ecological study. Environ Health10:50. doi:10.1186/1476-069X-10-50.

Mohammadi H, Karimi G, Rezayat SM, et al. (2011). Benefitof nanocarrier of magnetic magnesium in rat malathion-inducedtoxicity and cardiac failure using non-invasive monitoring of electro-cardiogram and blood pressure. Toxicol Ind Health 27:417–429.

Myers JM, Myers CR. (2009). The effects of hexavalent chromiumon thioredoxin reductase and peroxiredoxins in human bronchialepithelial cells. Free Radic Bio Med 47:1477–1485.

Nduka JK, Orisakwe OE. (2009). Heavy metal hazards of pediatric syrupadministration in Nigeria: a look at chromium, nickel and manganese.Int J Environ Res Public Health 6:1972–1979.

Park RM, Bena JF, Stayner LT, et al. (2004). Hexavalent chromium andlung cancer in the chromate industry: a quantitative risk assessment.Risk Anal 24:1099–1108.

Rahman I. (2008). Review: Antioxidant therapeutic advances in COPD.Ther Adv Respir Dis 2:351–374.

Ranjbar A, Pasalar P, Sedighi A, Abdollahi M. (2002). Induction ofoxidative stress in paraquat formulating workers. Toxicol Lett 131:191–194.

Satoh K. (1978). Serum lipid peroxidation in cerebrovascular dis-orders determined by a new colorimetric method. Clin Chim Acta 90:37–43.

Sawicka E, Srednicka D, Długosz A. (2008). Scutellaria baicalensisinhibits lipid peroxidation caused by chromium in human erythro-cytes. Acta Clin Exp Med 17:539–544.

Sinha S, Basant A, Malik A, Singh KP. (2009). Multivariate modeling ofchromium-induced oxidative stress and biochemical changes in plantsof Pistia stratiotes L. Ecotoxicology 18:555–566.

Sirven J-B, Bousquet B, Canioni L, et al. (2006). Qualitative andquantitative investigation of chromium-polluted soils by laser-inducedbreakdown spectroscopy combined with neural networks analysis.Anal Bioanal Chem 385:256–262.

Vermeulen R, Li G, Lan Q, et al. (2004). Detailed exposure assessmentfor a molecular epidemiology study of benzene in two shoe factoriesin China. Ann Occup Hyg 48:105–116.

Figure 2. Loading plot of PCA analysis in thefour measured parameters of chrome-platingworkers.

DOI: 10.3109/01480545.2014.922096 hemometrics models for assessment of oxidative stress risk 5

Dru

g an

d C

hem

ical

Tox

icol

ogy

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Mem

oria

l Uni

vers

ity o

f N

ewfo

undl

and

on 0

7/11

/14

For

pers

onal

use

onl

y.

Yao H, Guo L, Jiang B-H, et al. (2008). Oxidative stressand chromium(VI) carcinogenesis. J Environ Pathol ToxicolOncol 27:77–88.

Zendehdel R, Masoudi-Nejad A, H Shirazi F. (2012). Patterns predictionof chemotherapy sensitivity in cancer cell lines using FTIR spectrum,

neural network and principal components analysis. Iran J Pharm Res11:401–410.

Zhang X-H, Zhang X, Wang X-C, et al. (2011). Chronic occupationalexposure to hexavalent chromium causes DNA damage in electro-plating workers. BMC Public Health 11:224–231.

6 R. Zendehdel et al. Drug Chem Toxicol, Early Online: 1–6

Dru

g an

d C

hem

ical

Tox

icol

ogy

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

Mem

oria

l Uni

vers

ity o

f N

ewfo

undl

and

on 0

7/11

/14

For

pers

onal

use

onl

y.