Content uploaded by Setu Chakraborty
Author content
All content in this area was uploaded by Setu Chakraborty on Jul 27, 2019
Content may be subject to copyright.
Content uploaded by Sayed Asaduzzaman
Author content
All content in this area was uploaded by Sayed Asaduzzaman on Dec 06, 2015
Content may be subject to copyright.
Intern ational Journ al of Scientific & E ngineering Research, Volume 6, Issue 11, Novem ber-2015 876
ISSN 2229-5518
IJSER © 2015
http://www.ijser.org
Anticipation of the Significance of Risk
Factors in Cervical Cancer for Low Incoming
Country: Bangladesh Perspective
Sayed Asaduzzaman 1,*, Kawsar Ahmed1, Setu Chakraborty2, Md. Goljar Hossain3, Mamun Ibn Bashar3, Touhid Bhui-
yan4 and Subrata Sarker Chandan5
Abstract— Cervical cancer is the second alarming cancer for women of low incoming countries like Bangladesh. In future it would be the
main cause of death of Bangladeshi women by caner. To find the significant factors, association among them and making a precedence list
among them by data mining and statistical approaches. During, February 2014 till July 2014 a case-control study has been acquitted on
436 participants of both patients (199) and non-patients (237). Using an accurate questionnaire based on previous study the whole data
collection process done in the different part of the Dhaka cities and diagnostic center. About 10 factors like first sex at the age below 16,
Lack of knowledge about cervical cancer, number of children above 3, STI (Sexually Transmitted Infection) affection, previous cervical can-
cer history are founded highly significant by the statistical analysis and later those factors were given precedence by data mining process
Ranker algorithm with different attribute evaluator. Oral contraception taken, contraception used and vaccine taken factors are lower signifi-
cant than the other factors by the analysis. Both data mining and statistical approaches depict a comparative analysis and by the result the
significant factors and the significance priority can be measured.
Index Terms— Cervical cancer, Data mining, Statistical approach, Significant Factors, Low incoming country.
—————————— ——————————
1 INTRODUCTION
very year, over about 88% of women deaths occur from
cervical cancer in developing country like Bangladesh,
India, Pakistan etc. due to lack of knowledge, gender dis-
crimination and extreme poverty which limit proper care to
women [1]. Corruption of the highly regulated system of nor-
mal cell growth, division, and death resulted in cancer. Cervi-
cal cancer is a leading type of cancer among many other can-
cer like breast cancer, skin cancer, lung cancer, brain cancer,
prostate cancer, colorectal cancer, stomach cancer, and other
melanoma. Massive number of illiterate and conservative
women of societies has no idea about cervical cancer.
Cervical cancer is associated with development of the cer-
vix. The incidence of cervical cancer is started with a continu-
ous process of squamous metaplasia process during puberty.
This process involves transformation of columnar epithelium
of the cervix into squamous epithelium, in which transitional
cells support HPV replication and resulted in cervical intra-
epithelial neoplasia (CIN) 2 or CIN3 lesions and, eventually,
development of invasive cervical cancer. Multiple sexual part-
ners and early sexual activity influence squamous metaplasia
[2].
Five or more full term pregnancies, use of oral contracep-
tives for five or more years, smoking, previous exposure to
sexually transmitted infections (STIs), i.e. chlamydia tracho-
matis, some herpes viruses and HIV are some leading factors
associated with an increase in the risk of cervical cancer
among HPV-DNA positive women [2]. Estrogen and its recep-
tors promote cervical cancer because they are strongly associ-
ated with human papillomavirus (HPV) infections in combina-
tion with HPV oncogenes [3]. Cervical cancer is also regulated
by lack of nutrition like vitamins [4] and aging usually after
post-menopausal period at 45 or above [5]. Sometimes cervical
cancer is caused by genetic factors, but this evidence is not
clear [2].
Some preventive measure of cervical cancer includes
avoidance of multiple sexual partner, oral contraceptives,
smoking, multiple pregnancies; avoidance of early marriage
and multiple marriages; and use of condom during sexual
intercourse. Cervical cancer can also be prevented by eating
nutritious food which are rich in and by vaccination [6]. Sur-
gery, hysterectomy, chemotherapy, drugs targeting estrogen
and its receptors may be effective in treating and/or prevent-
ing cervical cancer.
A data mining and statistical approaches takes place in the
paper. The significant patterns of the factors and finding the
precedence among the significant factors have been performed
using statistical software (SPSS) and Data Mining Software
(WEKA).
2 BACKGROUND
According to National Cancer Institute, cervical cancer is a
slow-growing cancer that forms in tissues of the cervix, which
is an organ that connecting the uterus and vagina [7]. It is ex-
hibits no symptoms in earlier stage but in later stage of cancer
it shows symptoms like vaginal bleeding; pelvic pain and pain
during sexual intercourse [8,9]. It can be detected with regular
Pap test [7]. Human papillomavirus (HPV) is the actual culprit
of cervical cancer which has the ability to invade other parts of
the body and causing infection [10, 11].
Cervical cancer is the fourth most common cause of cancer
in women. In 2012, it was estimated that there were 266,000
deaths within 528,000 cases of cervical cancer worldwide [12].
Cervical cancer is a leading cause of cancerous death in Bang-
ladesh. 561, 583 and 574 women were affected by cervical can-
cer in 2005, 2006 and 2007 and it is second common malignan-
cies (21.5%) in females of Bangladesh [13].
Hence, the actual reason and total curing procedure of can-
E
IJSER
Intern ational Journ al of Scientific & E ngineering Research, Volume 6, Issue 11, Novem ber-2015 877
ISSN 2229-5518
IJSER © 2015
http://www.ijser.org
cer is not invented yet. Some general symptoms and risk fac-
tors of cervical cancer have been discovered by many statisti-
cal analyses. However, identification of environmental as well
as genetic factors is very important in developing novel meth-
ods of cervical cancer prevention. There are lots of works to
detect the risk factors of cervical cancer using population
based case control study [11], several databases, and algorithm
and induction techniques [14]. Some researchers tried to pre-
dict cancer risk using data mining technique [15-19]. Specifi-
cally there were no work of cervical cancer risk prediction us-
ing data mining or Statistical approaches.
3 METHODOLOGY
There are four parts in this section. Those are data collec-
tion, data preprocessing, statistical analysis and Data Mining
approaches using WEKA. Those sections are described se-
quentially below.
3.1 DATA COLLECTION
The data of total 436 interviewers were collected from dif-
ferent diagnosis center and some areas of Dhaka city. The data
were collected from 436 female participants whose age range
was between bellow 30 to above 60 years Some risk factors
were considered for cervical cancer assessment in Bangladeshi
population by former study, which includes- age, multiple
sexual partner, lack of correct condom utilization, the age at
first sexual intercourse of the woman, use of oral contracep-
tives for five or more years; high parity (five or more full term
pregnancies); and previous exposure to other sexually trans-
mitted infections (STIs), lack of proper nutrition, smoking, and
sometimes it is genetic risk. A questioner was designed to col-
lect by the former study.
3.2 DATA PREPROCESSING
Data transformations, data reduction, data cleaning, data
integration, data discretization are the five major tasks of the
data pre-processing to avoid incomplete, inconsistent and
noisy data. Conversion of the raw data to a reasonable level is
the main task of data preprocessing. There is some noise in the
data which were processed by some process. Incomplete data
hampers the analysis which has been eliminated or leveled. A
little bit of data were changes to avoid collision of the data
analysis.
3.3 STATISTICAL ANALYSIS
The statistical approaches were used to find the frequen-
cies, crosstab, bar diagrams, binary logistic regression. The
whole analysis was done in Statistical Package for Social Sci-
ence (SPSS version 20.0). Table 1 was simulated by cross tab
and frequency and Table 2 was prepared by binary logistic
regression with 95% confidence interval.
3.4 DATA MINING APPROACHES USING WEKA
Highly significant 10 factors have been exploited from the
analysis of the statistical approaches and depending on the
results of SPSS. Then those factors are ranked by ranking algo-
rithm with three different attribute evaluators OneRAt-
tributeEval, ReliefFAttributeEval, CorrelationAttributeEval .
4 RESULTS AND ANALYSIS
The statistical analysis of the results has been shown in this
section. Here the frequency distribution of various factors with
cervical cancer has been shown in Table 1. and Odds ratio and
Confidence Interval of the associated Factors in Binary logistic
regression analysis has been shown in Table 2. shows.
The frequency table was contrived by comparing the results
of SPSS and WEKA. Both statistical and data mining ap-
proaches shows same frequency. Age range between 27 and 80
where the mean age was 53, approximately 436 Bangladeshi
women’s data were analyzed. Here 237 women were not af-
fected (control group) and 199 women were affected by cervi-
cal cancer (case group). According to the case group, 31.7%
patient’s age lies between 60 and 80 and 66.8% patient’s age
lies between 46 and 60. Education levels of the control (affect-
ed) group participants are illiterate (55.30%), primary (27.1%),
secondary (11.1%) and undergraduate/above (6.5%) severally.
On the other hand in control (unaffected) group participants
are illiterate (19%), primary (51.9%), and secondary (13.9%)
and undergraduate/above (15.2%) respectively which has
been shown in Table 1. Among 199 affected women about 128
women take adequate fruits and vegetables whether 210 of 237
unaffected women take adequate fruits and vegetables by Ta-
ble 1. Among the patients 87.4% are married and the rest are
married.
By the Table 1, 195 participants (98%) had no knowledge
about cervical cancer among 199 affected where 75 partici-
pants (33.6%) had knowledge about cancer among 237 unaf-
fected women. 19.6% women had more than 3 sexual partners
in case (affected) group and 6.3% women had more than 3
sexual Partners in control (unaffected) group vice versa. The
table also illustrates that only 5% and 6.8% women of both
affected and unaffected group took cancer vaccine.
By the table it is clear that in control group most of the par-
ticipants (91.9%) had their first sexual intercourse after the age
of 16 years. Number of the affected women whose first sexual
intercourse was before the age of 16 was about above two
third of the total number.
About 95.5% affected participants had taken oral contracep-
tion when 96.5% affected participants had not used condom
during sex. More than two third (73.4%) women had children
among 3 to 5 in case group and one third 16.6% women had
more than 5 children. On the other hand more than two third
(82.7) women have 1 or 2 children in unaffected group.
The Confidence Intervals (C.I) and odds ratio with p-value
and standard error of the associated factors has been shown in
Table 2. Here the significantly associated factors has been
shown in the table using (*) sign. Those which factors are be-
low the flat marginal value of p=0.05. From the Table 2 anoth-
er factor has been identified Knowledge (p value is 0.000) .It
means that those who has the knowledge of cancer does not
have the possibility of cancer.
According to Table 2 about 10 factors have been chosen
which are highly significant. Table 3 illustrates that the com-
IJSER
Intern ational Journ al of Scientific & E ngineering Research, Volume 6, Issue 11, Novem ber-2015 878
ISSN 2229-5518
IJSER © 2015
http://www.ijser.org
parison of the results of Ranker algorithm by three different
attribute evaluator. Ranker algorithm shows the preceding
sequence of the factors. The Higher precedence depicts higher
priority of the factor to be more significant. Although all the
factors of Table 3 are significant but the results of the (OneR-
AttributeEval, ReliefFAttributeEval, CorrelationAttributeEval)
evaluators shows the higher significance and the table also
gives some decisions like factor (Number of Childs) got higher
precedence according the table. So it can be said that the factor
(Number of Childs) is highly significant than the others.
5 DISCUSSION
By a registry, after breast cancer (25.6%) cervical cancer is
the second (21.5%) leading cancer for Bangladeshi women.
According to “National Institute of Cancer Research and Hos-
pital” among the cervical cancer patients 8% widow, 91% were
married and 97 % were housewife [13]. Among the 199 cancer
patients about 87.4 % were married and 4.5% were widow
which was shown through the analysis. It can be noted by the
analysis that the possibility to be affected by cancer is much
higher whose ages are above 46. The largest part of the cancer
affected women’s age was among 46 to 60. Being a developing
country most of the women of Bangladesh are uneducated
[13]. According to our analysis, about two third of the total
cancer affected patients were uneducated. The fact is that the
largest amount of cancer affected women has no idea about
cancer. There are some strongly correlated factors like “whose
first sex age was below 16 “and “whose number of children
was above 3” which were observed by the analysis. Among all
the factors STI was found as highly significant factor whose p
value is (0.043). According to Table 2 “Take Adequate Food”,
“Oral contraception”, “Family cancer history” and
“knowledge about cancer” were also found significant factors.
By the analysis, those who took vaccine have the fewer possi-
bility of being pretended by cervical cancer. The former
study depicts that STI [11] , vaccin [6], Adequete food [6], his-
tory [2] , first sex age, number of sex partner, oral con-
traception were detected as significant which are also has
been discovered by the analysis.
6 CONCLUSIONS
Bangladeshi women are highly affected by cervical cancer.
A registry of National Cancer Institute of Bangladesh shows
that cervical cancer is the second (21.5%) leading cancer after
breast cancer (25.6%) in women [13]. The results of this study
show that behavioral interventions which aims at promoting
sexual behaviors protective of STI transmission and encourage
condom use for sexual intercourse. This manner can prevent
cervical cancer. Death by cervical cancer is certain due to
lack of knowledge of mass people and difficulties of proper
diagnosis process. In this paper, significant risk factors have
been analyzed through data mining and statistical approaches.
There are other approaches to detect cervical cancer risk like
ANN, SVM and DT etc. SVM is not is well for discrete data
and its kernel choice is more difficult for algorithm develop-
ment. DT creates complex tree for categorical variables. In
ANN, VC dimension is unclear what is important for good
solutions. On the other hand main drawback is it can be re-
tained. For those reasons and simplicity we chose Data mining
and Statistical Approach for analysis. An efficient approach
for the extraction of significant pattern from data warehouse
for efficient prediction of cervical cancer has been provided.
The precedence of the significant factors of statistical analysis
has been given priority sequence by data a mining approach
which shows the higher risk levels of the factors.
ACKNOWLEDGEMENT
The authors are grateful to those who gave their valuable
data and who participated in this research work
REFERENCES
[1] Ginsburg, O. M. ; Breast and cervical cancer control in low and
middle-income countries: H uman rights meet sound health policy.
Jour of Cancer Pol 2013, 1, e35–e41.
[2] Shepherd, J. P.; Frampton, G. K.; Harr is, P..Interventions for encour-
aging sexualbe havior s intended to prevent cervical cancer. Cochrane
Database Syst Rev 2011, 4 .
[3] Chung, S.H.; Franceschi, S.; Lambert, P. F.. Estrogen and ERα: Cul-
prits in cervical cancer?. Trends in Endocrinology & Metabolism 2010,
21, 504–511.
[4] Closas, R. G.; Castellsagué, X.; Bosch, X. ; González, C. A. The role of
diet and nutrition in cervical carcinogenesis: a review of recent evi-
den ce. Inter jour of cancer 2005, 117,629–637 .
[5] Castanon, A.; Landy, R. ; Cuzick, J. ; Sasieni, P..Cervical screening at
age 50–64 years and the risk of cervical cancer at age 65 years and
older: population-base d case control st udy. PLoS medicine 2014, 11,
e1001585.
[6] Einstein, M. H.; Baron, M. ; Levin, M. J. ; Chatterjee, A. ; Edwards, R.
P. ; Zepp, F. ; Carletti, I. ; Dessy, F. J. ; Trofa, A. F. ; Schuind, A.
;others. Comparison of the immunogenicity and safety of CervarixTM
and Gardasil® human papillomavirus (HPV) cervical cancer vaccines
in healthy women aged 18–45 years. Huma n vaccines 2009, 5, 705–719.
[7] “Defin ing Cancer", National Cancer Institute, (2014).
[8] “Cervical Cancer Treatment (PDQ®)", Nat ional Cancer Instit ute,
(Retrieved 24 June, 2014).
[9] Shapley, M. ; Jordan, J.; Croft, P. R. . A systematic review of postcoit-
al bleeding and risk of cervical cancer. British jour of gen prac 2006, 56,
453–460.
[10] Schiffman, M. ; Went zensen, N. ; Wacholder, S. ; Kinney, W. ; Gage,
J. C. ; Castle, P. E. . Human papillomavirus testing in the prevention
of cervical cancer. Jour of the National Cancer Ins 2011, 103368–383.
[11] Johnson, G. A. ; Unger, E. R. ; Ouattara, E. B. Coulibaly, ; K. T. ;
Maurice, C. ; Vernon, S. D. ; Sissoko, M. ; Greenberg, A. E. ; Wiktor, S.
Z. ; Chorba, T. L. Assessing the relationship between HIV infection
and cervical cancer in Cote d’Ivoire: a case-contro l study. BMC infec-
tious diseases 2010, 10, 242 .
[12] World Cancer Report, 2014; World Health Organization, 2014;
pp. Chapter 5.12.
[13] Cancer Registry Report, 2005-2007; National Institute of Cancer Re-
search and Hospital, Dh aka, Bangladesh; Published on December
(2009).
[14] Ho, S. H. ; Jee, S. H. ; Lee, J. E. ; Park, J. S. . Analysis on risk factors
for cer vical cancer using induction technique,Expert S ystems with
Application s 2004, 27, 97–105 .
IJSER
Intern ational Journ al of Scientific & E ngineering Research, Volume 6, Issue 11, Novem ber-2015 879
ISSN 2229-5518
IJSER © 2015
http://www.ijser.org
[15] Jesmin T, Ahmed K, Rahman MZ, Miah MBA. Brain c ancer risk
prediction tool using data mining. International journal of
computer applications. 2013; 61(12): 22-27.
[16] Ahmed K, Emran AA, Jesmin T, Mukti RF, Rahman MZ, Ahmed F.
Early detection of lung cancer risk using data mining.
Asian pacific journal of cancer prevention. 2013; 14 (1): 595–598.
[17] Ahmed K, Jesmin T, Fatima U, Moniruzzaman M, Emran AA, Rah-
man MZ. Intelligent and effective diabetes risk
prediction system using data mining. Oriental Journal of Computer
science & Technology. 2012; 5 (2): 215-221.
[18] Ahmed K, Jesmin T, Rahman MZ. Early prevention and detection of
skin cancer risk using data mining. International
journal of computer applications. 2013; 62(4):1 -6.
Table 1.
FREQUENCY DISTRIBUTION OF VARIOUS FACTORS
Variable
Category
Patient
Affected N (%) Unaffected N (%)
Education
Illiterate
Primary
secondary
Under gradu-
ate/above
110(55.3)
54(27.1)
22(11.1)
13(6.5)
45(19.0)
123(51.9)
33(13.9)
36(15.2)
Knowledge about cancer
Yes
No
4(2)
195(98)
75(33.6)
162(68.4)
Ever had any cancer
Yes
No
20(10.1)
179(89.9)
15(6.3)
222(93.7)
Cancer vaccine taken
Yes
No
10(5)
189(95.0)
16(6.8)
221(93.2)
Take Adequate Fruits and vege-
tables
Yes
No
128(64.3)
71(35.7)
210(88.6)
27(11.4)
Number of Sex partner
1-2
3+
160(80.4)
39(19.6)
222(93.7)
15(6.3)
First sex age
Below 16
Above 16
146(73.4)
53(26.6)
21(8.9)
216(91.9)
Contraception used
Yes
No
7(3.5)
192(96.5)
163(68.8)
74(31.2)
Oral contraception taken
Yes
No
190(95.5)
9(4.5)
74(31.2)
163(68.8)
Number of Childs
1-2
3-5
5+
20(10.1)
146(73.4)
33(16.6)
196(82.7)
26(11.0)
15(6.3)
Affected by STI
Yes
No
36(18.1)
163(81.9)
4(1.7)
233(98.3)
Age
Below 30
31-45
46-60
Above 60
0(0.0)
3(1.5)
133(66.8)
63(31.7)
3(1.3)
5(2.1)
134(56.5)
95(40.1)
Family cancer history
Yes
No
55(27.6)
144(72.4)
18(7.6)
219(92.4)
Marital status
Divorce
Married
6(3.0)
174(87.4)
0(0.0)
236(99.6)
IJSER
Intern ational Journ al of Scientific & E ngineering Research, Volume 6, Issue 11, Novem ber-2015 880
ISSN 2229-5518
IJSER © 2015
http://www.ijser.org
Separate
Widow
10(5.0)
9(4.5)
1(0.4)
0(0.0)
Table 2.
ODDS RATIO AND CONFIDENCE INTERVAL WITH P VALUE OF THE FACTORS
Factors B
S.E. Sig. (p) Exp.
(B)
95% C.I. for EXP(B)
Lower Upper
Premature chronicle of can-
cer
No(ref)
Yes
-1.148 1.316 0.383 0.317 0.024 4.19
Cancer vaccine taken
No(ref)
Yes
-1.519 1.545 0.326 0.219 0.011 4.527
Take Adequate Food(*)
No(ref)
Yes
-2.475 .835 0.003 0.084 0.016 0.433
How many Sex partner
1-2(ref)
3+
1.129 1.155 0.328 3.094 0.322 29.755
First sex (*)
Below 16(ref)
Above 16
-2.083 0.533 0.000 0.124 0.044 0.354
Oral contraception taken
No(ref)
Yes
1.709 3.295 0.604 5.523 .009 3520.785
Contraception used
No(ref)
Yes
-3.308 3.278 0.313 0.037 0.00 22.555
Number of Childs (*)
1-2(ref)
3-5
Above 5
3.114
2.414
.617
1.005
0.000
0.000
0.016
22.517
11.180
6.724
1.560
75.409
80.131
Previous Cancer history of
cervical cancer (*)
No(ref)
Yes
2.743 2.743 .006 15.539 2.188 110.350
Affected by STI (*)
No(ref)
Yes
2.564 2.564 0.043 12.984 1.079 156.201
Knowledge(*)
No(ref)
Yes
-5.779 1.115 0.00 0.003 0.000 0.028
IJSER
