Chapter 3: Cofactors in Human Papillomavirus

CarcinogenesisRole of Parity, Oral Contraceptives, and

Tobacco Smoking
Xavier Castellsague, Nubia Munoz

INTRODUCTION
It is now well established that infection with oncogenic human papillomavirus (HPV) types is the necessary cause of cervical cancer (CC) and its immediate precursor cervical intraepithelial neoplasia (CIN) 3 (CIN3). However, HPV infection
alone may not be sufficient to cause CC, and other exogenous or
endogenous factors might exist that, in conjunction with HPV,
influence the risk of progression from cervical HPV infection
to CC.
Candidate cofactors may be classified into three groups: 1)
environmental or exogenous cofactors, including use of oral
contraceptives (OCs), tobacco smoking, diet, cervical trauma,
and coinfection with human immunodeficiency virus (HIV) and
other sexually transmitted agents; 2) viral cofactors, such as
infection by specific types, coinfection with other types, HPV
variants, viral load, and viral integration; and 3) host cofactors,
including endogenous hormones, genetic factors such as human
20

leukocyte antigen, and other host factors related to the hosts

immune response.
The purpose of this chapter is to review, summarize, and
discuss the evidence of the role of the more established cofactors, including parity, OC use, and tobacco smoking.
Cofactors involved in the natural history of HPV infection are
reviewed in Chapter 2. In this chapter, the focus is on cofactors
affecting progression from HPV infection to high-grade squamous intraepithelial lesions (HSILs) and CC. Ideally, if we accept the premise that all CCs are caused by oncogenic HPVs, a
strict assessment of cofactors requires a study group known to be
exposed to HPV. From that HPV-exposed group, retrospectively
or ideally prospectively, the added risk attributable to other factors can be estimated. Although this approach is admittedly controversial (see the Discussion section), we believe that, in the
absence of repeated HPV measures, restriction to HPV DNApositive case patients and control subjects conveys the strictest
approach to adjustment for HPV. This review will thus focus on
selected key studies that, using reliable DNA-detection methods,
report associations between cofactors and HSIL/CC within a
well-defined HPV-positive group. These studies do exist and the
main characteristics of the most important ones are summarized
in Table 1 [(19); Plummer M, Herrero R, Franceschi S, Meijer
CJ, Snijders P, Bosch FX, et al.: unpublished data]. Studies
among HPV-positive women that included low-grade squamous
intraepithelial lesions (LSILs) (10) or the three grades of CIN
(11) are not considered in detail in this chapter.

EVIDENCE FOR A ROLE

CARCINOGENESIS

OF

OC

USE IN

HPV

Use of OCs has been found to be associated with CC in many,

but not in all, epidemiologic studies that adjusted for HPV status. In studies restricted to HPV-positive women, however, the
evidence for an association is in general weaker (Table 2). Of the
six studies reporting results restricted to HPV DNA-positive
subjects, three found positive statistically significant associations, but these were for a particular histologic type or for a
subgroup of women. The Eastern U.S. study (3) reported an odds
ratio (OR) of 17.1 (95% confidence interval [CI] 1.5 to 188.2)
for current versus never OC use, but this was only for adenocarcinoma in situ; although the risk increased with longer duraAffiliations of authors: X. Castellsague, Institut Catala` dOncologia, Servei
dEpidemiologia i Registre del Ca`ncer, LHospitalet de Llobregat, Barcelona,
Spain; N. Munoz, International Agency for Research on Cancer, Lyon, France.
Correspondence to: Xavier Castellsague, M.D., Ph.D., M.P.H., Institut Catala`
dOncologia, Servei dEpidemiologia i Registre del Ca`ncer, Gran via s/n, km
2.7, 08907 LHospitalet de Llobregat, Barcelona, Spain (e-mail: xcastellsague@
ico.scs.es).
See Notes following References.
Journal of the National Cancer Institute Monographs No. 31, Oxford
University Press 2003, all rights reserved.

Journal of the National Cancer Institute Monographs No. 31, 2003

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

It is now well established that infection with oncogenic human papillomavirus (HPV) types is the necessary cause of
cervical cancer (CC) and its immediate precursor cervical
intraepithelial neoplasia 3. However, HPV infection alone
may not be sufficient to cause CC, and other exogenous and
endogenous factors may exist that, in conjunction with HPV,
influence the risk of progression from cervical HPV infection
to CC. In this chapter, we review the evidence for the role of
parity, oral contraceptive (OC) use, and tobacco smoking in
CC. We also discuss limitations and methodologic problems
encountered in assessing available data and outline recommendations for future research. Based on key studies on
high-grade squamous intraepithelial lesions (HSILs) and CC
conducted among HPV-positive women, it can be concluded
that high parity, smoking, and less consistently long-term
OC use are cofactors that may modulate the risk of progression from HPV infection to HSIL/CC. From a public health
point of view, parity seems to be the behavioral cofactor
explaining the highest proportion of CC cases among HPVinfected women. Smoking and long-term OC use may have a
similar impact in populations that are heavily exposed to
HPV and to these cofactors. Large prospective and retrospective cohort studies of HSIL and CC among middle-aged
women in which several markers of HPV exposure are used
and HPV persistence is documented would be valuable to
study the role of these and other cofactors in HPV carcinogenesis. If confirmed, our conclusions may imply that multiparous women, women who are smokers, and women on
long-term OC use may need closer surveillance for cytologic
abnormalities and HPV infections than women in the general population. [J Natl Cancer Inst Monogr 2003;31:208]

Table 1. Characteristics of studies assessing the role of parity, OC use, and tobacco smoking in cervical carcinogenesis among HPV DNA-positive women*
Study
Study
characteristics

Denmark

United States,
Eastern

Norway

Manchester,
U.K.

United States
Portland, OR

Costa Rica

Authors, y,
(reference No.)

Kruger-Kjaer
et al., 1998 (1)

Olsen et al.,
1998 (2)

Lacey et al., 1999

(3), 2001 (4)

Deacon et al.,
2000 (5)

Hildesheim
et al., 2001
(6)

Castle et al.,
2002 (7)

Design

Casecontrol
within cohort
Population
HSIL
71/79

Casecontrol

Casecontrol

Population
CIN2 and CIN3
60/90

Population
CIS and CC
263

Casecontrol
within cohort
Population
CIN3
199

Casecontrol
within cohort
Population
HSIL and CC
146/168

Prospective
cohort
Health plan
CIN3 and CC

155/994

14/216

49/307

181

843

Cohort of 1812
women
positive for
oncogenic
HPV DNA

GP5+/6+

L1 PCR

MY09/MY11

MY09/MY11

PCR

Hybrid Capture 2

Parity, OC use,
smoking

Smoking

Smoking,
OC use**

Parity, OC use,
smoking

Parity, OC use,
smoking

Parity, OC use,
smoking

Munoz et al., 2002

(8); Moreno et al.,
2002 (9); Plummer
et al., unpublished
data, 2002
Casecontrol, pooled
analysis
Clinic and population
CIS and CC
1676/1853
255/1916

MY09/MY11 and
GP5+/6+
Parity, OC use,
smoking

*CC cervical cancer; CIN cervical intraepithelial neoplasia; CIS carcinoma in situ; HPV human papillomavirus; HSIL high-grade squamous
intraepithelial lesion; IARC International Agency for Research on Cancer; OC oral contraceptive; PCR polymerase chain reaction.
Includes studies in Spain, Colombia, Brazil, Thailand, The Philippines, Morocco, Peru, and Paraguay.
HPV16 only.
All case subjects were included regardless of HPV status.
High-risk genotypes only.
Cofactors in bold type indicate that a statistically significant association was found for that cofactor.
For squamous cell.
**For adenocarcinoma in situ.

tion of use, the trend was of borderline statistical significance.

The study in Costa Rica (6) found a 3.1-fold increased risk for
users of 5 or more years as compared with never users, but this
association was observed only among women who had two or
fewer pregnancies, and the HSIL/cancer group included only

30 cases. The strongest evidence for a role of OC use in HPV

carcinogenesis derives from the large pooled analysis of the
International Agency for Research on Cancer (IARC) studies
(9). Even though ever use of OCs was moderately associated
with cancer risk (OR 1.4), there was a strong doseresponse

Table 2. Summary results of studies assessing the role of OC use as a cofactor in cervical carcinogenesis among HPV DNA-positive women*
Study (study outcome)

Exposure measures

Denmark (HSIL)

OC use status, OR
(95% CI)
Ever vs. never
Former vs. never
Current vs. never

NR
NR
NR

OC use duration
[Years] OR (95%
CI) vs. never
P for trend
Comments

United States,
Eastern
(CIS and CC)

5.4 (0.7 to 43.4)

3.1 (0.4 to 27.5)
17.1 (1.5 to 188.2)

Decreasing risk
[2] 4.0 (0.4 to 44.3)
with increasing [6] 4.8 (0.4 to 51.9)
duration
[7] 6.2 (0.7 to 52.7)
NR

.12

ORs not reported

For adenocarcinoma in situ

Manchester, U.K.
(CIN3)

Costa Rica
(HSIL and CC)

United States,
Portland, OR
(CIN3 and CC)

IARC
(CIS and CC)

NR
1.2 (0.6 to 2.1)
1.3 (0.7 to 2.5)

NR
0.9 (0.6 to 1.6)
1.5 (0.8 to 2.8)

NR
NR
0.8 (0.5 to 1.5)

1.4 (1.0 to 2.0)

NR
NR
[1] 0.7 (0.4 to 1.1)
[4] 0.8 (0.5 to 1.2)
[9] 2.8 (1.5 to 5.4)
[10] 4.0 (2.1 to 7.8)
<.001

[3] 1.2 (0.6 to 2.4)

[8] 0.8 (0.4 to 1.5)
[>8] 1.5 (0.8 to 2.9)

[4] 1.8 (0.6 to 4.9)

[5] 3.1 (1.1 to 9.1)

NR

NR

NR

NR

Duration estimates
computed among
women with <3
pregnancies

Relative risk refers

to current vs. not
current OC use at
enrollment

For both squamous

cell carcinoma and
adenocarcinoma

*Numbers are odds ratios (or relative risks) and 95% confidence intervals for the association between the corresponding exposure measure and HSIL/CC risk.
CC cervical cancer; CI confidence interval; CIN cervical intraepithelial neoplasia; CIS carcinoma in situ; HPV human papillomavirus;
HSIL high-grade squamous intraepithelial lesion; IARC International Agency for Research on Cancer; NR not reported; OC oral contraceptive;
OR odds ratio. Bold numbers denote statistical significance.

Journal of the National Cancer Institute Monographs No. 31, 2003

21

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

Source of subjects
Outcome
HPV-positive
case subjects/
total tested
HPV-positive
control
subjects/total
tested
HPV DNA
detection
Relevant
cofactors
assessed

IARC,
international

HPV16-expressing mice. Alternatively, OCs might facilitate

HPV reactivation or persistence, although indirect evidence
from several studies (9, see review in 16) does not find an
association between OC use and HPV positivity among control
women.

EVIDENCE FOR A ROLE

CARCINOGENESIS

OF

PARITY

IN

HPV

High parity has consistently been found in most casecontrol

studies to be associated with both CC and CIS. Most of the
major studies restricting the analysis to HPV-positive women
also report an increased risk of HSIL/CC with an increasing
number of pregnancies (Table 3). In the IARC-pooled analysis,
the OR for CC in women with seven or more full-term pregnancies was fourfold higher than that in nulliparous women, and
the risk increased linearly with an increasing number of fullterm pregnancies (8). Risk of HSIL/CC significantly increased
with an increasing number of live births in the Costa Rica study
(6). A borderline association with CIN3 was found in the
Manchester study (5). The study in Denmark (1) and the U.S.
prospective study (7) did not find an association with the risk of
HSIL and CIN3/CC, respectively. However, this result could be
explained by the low parity of the study populations. In addition,
in the U.S. cohort, information on parity was recruited only at
enrollment. Thus, pregnancies occurring during the 10-year follow-up period of the study might have been more relevant to
the prospective risk of CIN3/CC than those occurring before
enrollment.
In addition to the studies restricted to HPV-positive women,
further evidence of a hormonal effect on CC comes from the
analysis of the age-distribution curve based on cohort effects
in relation to mortality without the distortion potentially introduced by screening and secular changes. The analysis shows
that CC mortality rates increase very sharply up to age 50 years
(i.e., around menopause) and flatten thereafter.
Nutritional, hormonal, traumatic, and immunologic mechanisms have been hypothesized as biologically plausible explanations for the association between parity and HSIL/CC among

Table 3. Summary results of studies assessing the role of parity/pregnancy in cervical carcinogenesis among HPV DNA-positive women*
Study (study outcome)

Exposure measures
Ever vs. never, OR (95% CI)
No. of live births or pregnancies
[No.] OR (95% CI) vs. never

P for trend
Comments

Denmark (HSIL)

Manchester, U.K.
(CIN3)

NR

NR

[0] 0.8 (0.4 to 1.7)

[1] 1.8 (0.3 to 2.3)

[1] 1.6 (0.9 to 2.8)

[2] 1.1 (0.6 to 2.0)
[3] 1.9 (0.9 to 3.8)

NR

NS
Unadjusted

Costa Rica,
(HSIL and CC)
4.6 (1.1 to 20)
[2] 1.0 (0.5 to 2.2)
[3] 1.5 (0.7 to 3.2)
[45] 3.5 (1.7 to 7.2)
[68] 2.2 (1.0 to 5.0)
[9] 1.4 (0.6 to 3.4)
.04

United States,
Portland, OR
(CIN3 and CC)

IARC
(CIS and CC)

NR

NR

[12] 1.1 (0.6 to 1.7)

[3] 0.7 (0.3 to 1.6)

[12] 1.8 (1.0 to 3.5)

[34] 2.6 (1.3 to 4.9)
[56] 2.9 (1.4 to 5.6)
[7] 3.9 (1.9 to 7.9)

NR

<.0001
For both CC and CIS

*Numbers are odds ratio (or relative risks) and 95% confidence intervals for the association between the corresponding exposure measure and HSIL/CC risk.
CC cervical cancer; CI confidence interval; CIN cervical intraepithelial neoplasia; CIS carcinoma in situ; HPV human papillomavirus;
HSIL high-grade squamous intraepithelial lesion; IARC International Agency for Research on Cancer; NR not reported; NS not significant; OR odds
ratio. Bold numbers denote statistical significance.
OR for ever versus never pregnant.
Ever pregnant but 0 live births.
Reference includes women with 0 or 1 live birth.

22

Journal of the National Cancer Institute Monographs No. 31, 2003

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

relationship with increasing years of use (Table 2). No increase

in the risk of cervical neoplasia was found for the duration of
OC use for up to 4 years. However, use of OCs for longer than
5 years was significantly associated with cervical neoplasia
(OR 3.4; 95% CI 2.1 to 5.5). The risk of OC use for longer
than 5 years was increased fourfold for invasive CC (OR 4.0;
95% CI 2.0 to 8.0) and threefold for carcinoma in situ (CIS)
(OR 3.4; 95% CI 2.1 to 5.5).
Special attention should be given to the lack of association
reported in the only prospective study of CIN3 and CC (7). As
the same authors discuss, first, only one measurement of OC use
was obtained at enrollment, thus not accounting for the possible
discontinuation or initiation of use of OCs during the course of
the study. Second, OC users had shorter follow-up times than
nonusers, which may have resulted in censoring bias among the
users. Third, among the control subjects, OC users were more
likely to be diagnosed with CIN1 and CIN2 during follow-up
than nonusers, which may have resulted in an increased detection and treatment of women in the control group whose disease
might have otherwise progressed to CIN3 or CC. Finally, no
information regarding duration was collected, thus potentially
missing an association with long-term use.
Not much data are available concerning the mechanisms by
which hormonal influences may modulate the risk of progression to HSIL/CC among HPV-infected women. Hormonerelated mechanisms may influence the progression from premalignant to malignant cervical lesions by promoting integration of
HPV DNA into the host genome, which results in deregulation
of E6 and E7 expression (12). An experimental study (13) has
shown that estradiol may stimulate the transcription of HPV type
16 (HPV16) E6 and E7 in cell lines that contain integrated
HPV16. Since the E6 and E7 open reading frames have been
associated with the oncogenic potential of HPV16, the effect of
estrogen on the transcription of these viral genes may be of
biologic relevance in the malignant transformation of HPV16infected cervical cells. Data from experimental studies (14,15)
demonstrate a synergistic mechanism between long-term estrogen exposure and HPV16 oncogenes that modulates squamous
carcinogenesis in the female reproductive tract of transgenic

infected women; however, because of the concordance of effects

with OC use, hormonal influences can be considered to be one
of the most promising candidates in the search for HPV cofactors. High parity may likely increase the risk of CC because it
maintains the transformation zone on the exocervix for many
years (17), facilitating the direct exposure to HPV and, possibly,
to other cofactors. Hormonal changes induced by pregnancy
(increased levels of estrogen and progesterone) may also modulate the immune response to HPV and influence risk of persistence or progression (8,18).

EVIDENCE FOR A ROLE

HPV CARCINOGENESIS

OF

TOBACCO SMOKING

IN

EVIDENCE FROM STUDIES COMPARING LSIL, HSIL,

AND CC COFACTOR PROFILES
Finally, it is worth mentioning that few cofactors have been
identified to distinguish invasive cancer from intraepithelial lesions or HSIL from LSIL. The IARC SpainColombia study

Table 4. Summary results of studies assessing the role of cigarette smoking in cervical carcinogenesis among HPV DNA-positive women*
Study (study outcome)

Manchester, U.K.
(CIN3)

Costa Rica
(HSIL and CC)

United States
Portland, OR
(CIN3 and CC)

IARC
(CIS and CC)

NR
1.7 (0.8 to 4.0)
2.3 (1.2 to 4.3)

NR
2.1 (1.1 to 3.9)
NR

2.2 (1.5 to 3.2)

1.8 (0.9 to 3.4)
2.3 (1.3 to 4.0)

Exposure measures
Smoking status,
OR (95% CI)
Ever vs. never
Former vs. never
Current vs. never

NR
3.2 (0.9 to 11.4)
1.9 (1.0 to 3.8)

4.6 (0.9 to 22.9)

4.2 (0.5 to 37.9)
NR

1.5 (0.7 to 3.0)

1.2 (0.5 to 3.1)
1.6 (0.7 to 3.5)

2.2 (1.4 to 3.4)

1.7 (0.8 to 3.7)
NR

NR

[10] 3.3 (0.5 to 21)

[11] 5.9 (1.0 to 36)

[19] 1.6 (0.6 to 3.9)

[20] 1.3 (0.6 to 3.0)

[10] 1.4 (0.7 to 2.5)

[16] 2.2 (1.2 to 3.9)
[17] 3.1 (1.8 to 5.3)
<.0001

Smoking amount
[cigarettes/day] OR
(95% CI) vs. never
P for trend
Smoking duration
[years] OR (95% CI)
vs. never
P for trend
Comments

NR

NR

NR

Norway
(CIN2 and CIN3)

United States,
Eastern
(CIS and CC)

Denmark
(HSIL)

NR

[9] 2.1 (0.3 to 12.3)

[10] 7.5 (1.2 to 46)
NR
Stronger associations found
among HPV16seropositive
subjects

.49

[5] 1.8 (1.0 to 3.3) [19] 2.2 (1.2 to 4.2)

[6] 3.1 (1.2 to 7.9) [20] 2.9 (1.5 to 5.6)
.003

[10] 1.2 (0.5 to 3.0)

[9] 1.8 (0.9 to 3.6)
[9] 2.2 (1.0 to 4.8)
[20] 3.2 (1.0 to 9.7) [19] 2.0 (1.2 to 3.3) [10] 2.0 (1.0 to 3.8)
[>20] 0.8 (0.3 to 2.7) [20] 3.1 (1.6 to 6.2)
.57
<.0005
NR
For squamous cell
carcinoma

[5] 1.9 (1.1 to 3.4)

[6] 2.2 (1.2 to 4.2)

NR

NS

NR

[19] 2.6 (1.4 to 4.7)

[20] 1.9 (1.0 to 3.5)
NS

Univariate OR.
Adjusted OR also
statistically
significant

*Numbers are odds ratios (or relative risks) and 95% confidence intervals for the association between the corresponding exposure measure and HSIL/CC risk.
CC cervical cancer; CI confidence interval; CIN cervical intraepithelial neoplasia; CIS carcinoma in situ; HPV human papillomavirus; HSIL high-grade squamous
intraepithelial lesion; IARC International Agency for Research on Cancer; NR not reported; NS not significant; OR odds ratio. Bold numbers denote statistical significance.

Journal of the National Cancer Institute Monographs No. 31, 2003

23

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

The effects of smoking have been well studied in many case

control studies, and they show a moderate and statistically significant association with CC, even after adjusting for the strong
effects of HPV. These findings are strikingly consistent with
those obtained in studies restricted to HPV-positive women. As
shown in Table 4, all such studies report some evidence that
tobacco smoking increases the risk of developing HSIL and CC.
The ORs for ever smoking among HPV-positive women are in
the range of 2 to 5. Furthermore, most studies reporting risk
estimates according to intensity, duration, or pack-years show an
increased risk of CC with increasing exposure to tobacco smoking. Because of the prospective nature of the study, the positive
association found with smoking status and smoking intensity in
the U.S. prospective study is particularly relevant (7).
Despite the consistency of these findings, the possibility remains that smoking or smoking duration is a proxy for time since
HPV exposure, because long-duration smokers may also have
had an HPV infection for a long time. Thus, residual confound-

ing by time since HPV infection cannot be ruled out as a possible

explanation for the observed effects of smoking.
Almost 60 years ago, Rous and Friedwald (19) reported
the carcinogenic effect of tar on virus-induced rabbit papillomas. More recently, malignant transformation of HPV16immortalized human endocervical cells by cigarette smoke condensate has been proven (20). The fact that nicotine and tobaccospecific carcinogens have been detected in the cervical mucus of
smokers (21) further strengthens the hypothesis of a synergistic
action between cigarette smoking and HPV for the development
of HSIL/CC. Chemical tobacco-related carcinogens may exert a
direct mitogenic effect causing DNA damage. Some authors (22)
hypothesize that exposure to tobacco may affect the ability of the
host to mount an effective local immune response against viral
infections, since it has been shown that smoking may reduce the
number of Langerhans cells and other markers of immune function. A recent prospective study (23) presents convincing evidence that smokers maintain cervical HPV infections significantly longer and have a lower probability of clearing an
oncogenic infection than women who never smoked. The significant association found between the extent of smoking reduction and the reduction in lesion size in an intervention study of
smoking cessation among women with minor-grade lesions further strengthens the plausible role of tobacco smoking in HPV
carcinogenesis (24).

(25) considered a large number of risk factors and found that

both the CIN3 and CC patients had very similar profiles of risk
factors. A recent study (26) conducted in Thailand found that,
after controlling for HPV type, the risk of developing CC, as
compared with the risk of developing intraepithelial lesions, was
not related to any of the cofactors considered, except for two
indices of socioeconomic status. In contrast, a study comparing
HPV-positive women with CIN3 with HPV-positive women
with CIN1 (27) found that cigarette smoking was significantly
associated with CIN3, suggesting that HPV-infected cells may
relate to tobacco-containing carcinogens for neoplastic progression.

DISCUSSION
Methodologic Issues in the Study of Cofactors in HPV
Carcinogenesis

24

Journal of the National Cancer Institute Monographs No. 31, 2003

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

Definition of case patients and control subjects. HPV cofactors in CC may act in at least three ways: 1) by influencing
the acquisition of HPV infection, 2) by increasing the risk of
HPV persistence, and 3) by increasing the risk of progression
from HPV infection to HSIL and cancer.
Studies aimed at identifying factors for HPV acquisition and
persistence were considered in Chapter 2. In these studies, case
patients and control subjects can be defined on the basis of
HPV/LSIL status. Here, we focused on factors that, acting once
HPV infection has been established, modulate the risk of progression from HPV infection to HSIL/CC. These cofactors can
be identified by cohort or casecontrol studies of HPV-positive
women in which case patients include those diagnosed with
HSIL, CIN3, CIS, or CC but not those diagnosed with LSIL or
CIN1. Accepting that HPV is a necessary cause of CC, we
believe that proper control subjects for these studies ought to be
HPV-positive women. Which assay or combination of assays is
best suited to select HPV-exposed or HPV-positive women still
remains an important research issue for which more data are
needed. Because many investigators consider LSIL an early
manifestation of HPV infection, some have also included
women with LSIL in the HPV-positive control group (6).
Studies (11) including the whole spectrum of precancerous
lesions as cases are difficult to interpret because the case group
includes a mixture of exposed and diseased women, because
some have LSIL, a marker of HPV exposure more than a disease
outcome, and some have HSIL/CC, clearly an HPV-related disease outcome. The nested casecontrol studies carried out in
Denmark (1) and in the United Kingdom (5) suggest that the
pattern of risk factors for HPV infection or LSIL are different
from the patterns for CIN3 and HSIL. Again, under the premise
that HPV is a necessary cause of CC, a more refined case definition could be made on the basis of both HPV and disease
statuses.
In relation to the control subjects and the selection of HPVexposed women, the key issue is whether one-point HPV positivity can be considered to be a sufficient criterion to define the
proper control group. In other words, can we assume that crosssectionally identified HPV-positive control women are carrying
a persistent or chronic infection? How long should persistent
infection persist to increase the risk of CC: many months or just
a few months as suggested recently (28)?
If persistent infections are associated with CC risk, the definition of HPV-positive control women based on a single mea-

sure (used in most studies) might not be sufficient. However, if

most HPV infections in middle-aged women tend to be persistent, the use of a single HPV test may be more of a problem in
studies of HSIL than in studies of invasive CC. Furthermore, the
interpretation of results from studies including young women
(most of the HSIL studies in Table 1) is further limited by the
fact that measures of the putative environmental cofactors are
strongly age dependent. Younger women are less likely to have
had a high number of pregnancies, to have smoked, or to have
used OCs for a long time.
HPV adjustment strategies. Using the IARC series of case
control studies, we were able to estimate the impact of different
strategies of HPV adjustment on associations between environmental cofactors and CC risk. For each of the three cofactors of
interest, we fitted three different models: 1) one model including
all case patients and all control subjects but ignoring adjustment
for HPV, 2) a second including all subjects but adjusting for
HPV DNA status, and 3) a third restricting the analysis to case
patients and control subjects who tested positive for HPV DNA.
As shown in Table 5, of the three strategies, models restricted to
HPV DNA-positive subjects yielded higher associations that
were between 1.1- and 2.0-fold higher than those derived from
HPV-adjusted models. For parity and smoking, the magnitudes
of the ORs were lowest for the crude models, were intermediate
for the HPV-adjusted models, and were highest for the HPVrestricted models. For OC use, the HPV-adjusted models yielded
lower ORs than the crude and HPV-restricted models. If we
believe that HPV restriction is the strictest approach to HPV
adjustment, one could then conclude that the studies adjusting
for HPV are likely to underestimate the magnitude of the association but that this underestimation is not greater than twofold.
It remains to be explained why we see stronger rather than
weaker effects when analyses are restricted to HPV-positive case
patients and control subjects. If it is claimed that this method of
adjustment is better than the other strategies in getting rid of
residual confounding, then we should expect lower rather than
higher ORs. Perhaps negative confounding between cofactors
and HPV status or selection bias introduced by restricting the
analysis to HPV-positive subjects (i.e., the casecontrol matching is totally ignored after restriction) might explain the increase
in the ORs. Beyond the discussion of which adjustment method
is more suitable in the assessment of cofactor, it is reassuring to
observe that, regardless of the strategy used, the same conclusions concerning the direction and statistical significance of the
associations are reached.
Since the goal in the assessment of cofactors is to explore
associations among HPV-exposed women, an alternative approach is to explore the use, alone or in combination with HPV
DNA detection, of serologic markers of HPV infection. Use of
HPV DNA detection alone may overlook past infections that
being clinically relevant at the time were cleared from the cervix. In contrast, while HPV serology is highly specific, its low
sensitivity (between 50% and 70%, depending on the assay and
on the number of types included) might pose problems of selection bias (i.e., women who did not trigger a detectable seroconversion after a relevant HPV infection would not be selected by
the serologic assay).
Estimation of attributable fractions of CC explained by
the various cofactors. To assess the overall impact of cofactors
on CC burden, we found that it is useful to estimate the percentage of disease that can be attributed to each cofactor. However,

Table 5. Impact of different strategies of HPV adjustment on associations between cofactors and risk of CC (from the IARC casecontrol studies)*
All women
Cofactor
Full-term pregnancies (status and No.)
Never
Ever

Cases/controls

444/747

34

644/677

1095/785

OC use (status and years)

Never
Ever

1
1.08
(0.80 to 1.47)
0.82
(0.60 to 1.12)
1.33
(0.97 to 1.83)
1.73
(1.24 to 2.41)

1
1.32
(0.88 to 1.98)
0.99
(0.65 to 1.50)
1.68
(1.09 to 2.57)
2.03
(1.30 to 3.16)

57/24
1616/229

1
1.09
(0.94 to 1.27)
0.90
(0.75 to 1.09)
1.33
(1.11 to 1.59)

1
0.97
(0.79 to 1.19)
0.75
(0.58 to 0.97)
1.17
(0.92 to 1.49)

1071/163
605/92

351/445

5 y

510/427

1
1.30
(1.11 to 1.52)
1.21
(0.98 to 1.51)
1.79
(1.45 to 2.22)

1
1.68
(1.36 to 2.08)
1.46
(1.09 to 1.97)
2.07
(1.56 to 2.75)

1265/218
409/36

1645/1905
636/488

15 cigarettes/day

251/200

6 cigarettes/day

350/216

Case/control

279/59
450/70
887/100

274/64
331/28

181/17
211/18

OR
(95% CI)
1
2.45
(1.33 to 4.51)
1.79
(0.94 to 3.40)
2.61
(1.37 to 5.00)
3.88
(1.99 to 7.55)
1
1.13
(0.80 to 1.59)
0.66
(0.45 to 0.98)
2.35
(1.44 to 3.85)
1
1.99
(1.29 to 3.07)
1.72
(0.98 to 3.01)
2.16
(1.18 to 3.97)

*ORs adjusted for center, age (<37, 3745, 4655, or 56 years), educational level (none, primary, secondary, or higher), smoking amount (never, 15
cigarettes/day, or 6 cigarettes/day), age at first sexual intercourse (<17, 1718, 1922, or 23 years), lifetime number of sexual partners (1, 23, or 4), OC use
(never, 14 years, 59 years, or 10 years), lifetime number of Pap smears (0, 15, or 6), and parity (0, 12, 34, 56, or 7). CC cervical cancer;
CI confidence interval; HPV human papillomavirus; IARC International Agency for Research on Cancer; OC oral contraceptive; OR odds ratio.
Referent.

the necessary-cause model of HPV carcinogenesis poses a

methodologic challenge, since the use of standard methods to
estimate attributable fractions (AFs) may be inadequate under a
model that assumes that the exposures for which AFs are to be
estimated are only relevant once the individual is or has been
exposed to a necessary etiologic factor. Despite these limitations, we (Table 6) and others (6) have estimated AFs for cofactors by using as parameters the percentages of women exposed to the cofactor of interest among HPV-positive women
and the OR for CC derived from analyses restricted to infected
women. As shown in Table 6, among HPV-positive women, AFs
are comparatively higher for parity than for the other cofactors.
Thus, given the high percentage of HPV-positive women with
two or more full-term pregnancies and the relatively strong association for CC at this parity level, this analysis suggests that
multiparous HPV-positive women constitute the group at the
highest risk of CC. However, these conclusions should be taken
only as a crude approximation to the issue because the overall
disease burden attributable to the cofactor depends not only on
the percentage of HPV-positive women exposed to the cofactor
but also on the overall HPV prevalence in the population; the
estimates reported in Table 6 or in the Costa Rica study do not
take the latter parameter into account.
We also assessed the impact of different combinations of
cofactors on CC burden. Table 6 shows that the most prevalent
combination of cofactors among case patients is that including
high parity, low OC use, and never smoking (55.8%). We also
Journal of the National Cancer Institute Monographs No. 31, 2003

show that only 5.4% of the case patients were exposed to none
of the risk categories of the cofactors considered, suggesting that
the fraction of the case patients in whom these cofactors may not
play a role is probably very low.
Recommendations for Future Research
On parity and hormonal factors. 1) Taking into account
the second peak in the prevalence of HPV DNA observed in
perimenopausal and postmenopausal women in some populations (2931), it would be of interest to conduct studies in these
women and their male partners to try to distinguish if this second
peak is the result of reactivation of latent infections or new
infections. In addition, these studies will be able to explore the
interaction between endogenous and exogenous hormones and
HPV infection in middle-aged women. Are the endogenous hormone levels and/or hormone replacement therapy (HRT) associated with HPV DNA detection and progression to cancer in
postmenopausal women?
2) Little is known about the influence of hormones on the
mechanisms of HPV carcinogenesis. Epidemiologists should encourage their laboratory colleagues to carry out experimental
studies on this issue, since such studies will be valuable for
bolstering (or not) the biologic plausibility of the epidemiologic
associations.
3) Further studies on OC use and HRT are needed to clarify
their role in relation to type of hormones, concentrations, duration, recency, and latency and to determine at which stages of
25

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

HPV adjusted,
OR (95% CI)

1419/1508
864/886

14 y

Smoking (status and amount)

Never
Ever

Not HPV adjusted,

OR (95% CI)

95/164
2183/2209

12

HPV-positive women

Table 6. Proportion of cervical cancer cases attributable to various HPV cofactors, separately or in combination*
Separately
HPV cofactor

Level of exposure

% among control subjects

% among case subjects

OR

AF, %

Full-term pregnancies

Ever vs. never

2 vs. 01
5 vs. 01

90.5
83.0
39.5

96.6
89.8
53.0

2.51
1.63
2.27

58
34
33

OC use

Ever vs. never

5 vs. 04

36.1
11.0

36.1
19.7

1.13
2.70

4
16

Smoking

Ever vs. never

6 cigarettes/day vs. 0 cigarettes/day

14.2
7.1

24.4
12.7

1.98
2.08

12
7

Combination of cofactors
Parity

Smoking

% among control subjects

% among case subjects

OR

AF, %

04
04
5
5
04
04
5
5

Never
Ever
Never
Ever
Never
Ever
Never
Ever

12.4
3.6
0.8
0.4
63.7
9.2
8.8
1.2

5.4
3.5
0.6
0.7
55.8
15.5
13.7
4.8

1
2.12
2.51
3.60
1.61
2.87
4.16
11.02

4
1
1
28
15
22
11

*AF attributable fraction; HPV human papillomavirus; OC oral contraceptive; OR odds ratio. 1 referent category.
Adjusted for center, age, educational level, smoking amount, age at first sexual intercourse, lifetime number of sexual partners, OC use (years), lifetime number
of Pap smears, and parity.

carcinogenesis they may act. These issues will be better studied

in cohort studies of premenopausal and postmenopausal women
than in casecontrol studies.
4) Also, it would be important to undertake studies making
measurements of markers of local immunity and inflammation
and correlating them with other cofactors, such as pregnancy,
age, and intake of exogenous hormones.
Correlation analyses of time trends. It would be valuable to
study the influence of declining birthrates on declining CC rates,
especially in those countries where screening programs do not
exist or have had little impact. Similarly, it would be interesting
to study the relationship between time trends of CC rates and
those of OC use and smoking.
Choice of proper control subjects. The selection of proper
control subjects depends on the basic assumption that most HPV
infections in older women are persistent. However, unpublished
results from a follow-up study of control women included in our
casecontrol studies in Spain and Colombia suggest that this
may not be the case. Thus, prospective studies of middle-aged
HPV-positive women are needed to assess this assumption and
to better understand the natural history of HPV infections in
older women.
Alternatively, the conduct of large casecontrol studies, including as control subjects women with at least two consecutive
HPV-positive smears over time to confirm their chronic carrier
state, would circumvent some of the limitations currently encountered in the interpretation of results from previous studies.
A quick approximation to this approach for recently completed studies would be to recontact HPV-positive control
women, retest them for cervical HPV DNA, and analyze the new
data stratifying by HPV-persistence status. This strategy would,
in addition, provide an assessment of the potential bias introduced in studies that restricted analyses to cross-sectionally detected HPV-positive control subjects.
One should, however, realize that the use of persistently
26

HPV-positive control subjects, while allowing for the assessment of risk factors for disease progression, would limit the
ability of evaluating factors potentially associated with viral persistence itself.
Finally, it is evident that more epidemiologic research is
needed to understand the value of serologic markers of HPV
exposure in studies on cofactors. The first pending issue is to
improve the sensitivity of current serologic assays of HPV infection, probably by increasing multiple serotypes. Second, studies using both HPV DNA- and HPV antibody-detection methods
are needed to contrast our current DNA-based risk estimates
linked to the different cofactors and to assess whether the combination of both assays could improve unbiased selection of
exposed women. Studies comparing the natural history of HPV
infections with the dynamics and patterns of seroconversion are
needed to better understand the meaning of these markers and to
comprehend why some women experience seroconversion and
some do not.
Other studies. Except for the U.S. cohort study (7), no prospective data on cofactors are yet available from the large cohort
studies being conducted in Costa Rica, Brazil, and Colombia.
Assessment of unpublished results from these cohort studies
would be informative in the planning of further research strategies on this issue. These studies will be of great value in establishing if the effects of cofactors that we are detecting in case
control studies are real or, on the contrary, merely due to residual
confounding. These cohort studies will be able to adjust for the
number of HPV infections and also be useful to determine at
which stage of the carcinogenic process these cofactors act.
As mentioned before, the value of HPV serology as a marker
of cumulative lifetime exposure needs further study. As more
refined serologic assays are being developed and validated, they
could be introduced alone or in combination with HPV DNA in
retrospective cohort studies that stored sera and cells and prospectively monitored relevant cofactors.
Journal of the National Cancer Institute Monographs No. 31, 2003

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

01
01
01
01
2
2
2
2

OC use, y

No published data are available on the analysis of cofactors

by main HPV types (16 and 18) or their phylogenetic groups.
Unpublished data from the IARC series of casecontrol studies
showed that associations between cofactors and CCs are not
modified by specific genotypes, but more data are needed to
confirm this.
Finally, further methodologic work is needed to estimate AF
to assess the global impact of these cofactors on CC occurrence,
taking into account the HPV necessary-cause model.

CONCLUSIONS

REFERENCES
(1) Kruger-Kjaer S, van den Brule AJ, Svare EI, Engholm G, Sherman ME,
Poll PA, et al. Different risk factor patterns for high-grade and low-grade
intraepithelial lesions on the cervix among HPV-positive and HPVnegative young women. Int J Cancer 1998;76:6139.
(2) Olsen AO, Dillner J, Skrondal A, Magnus P. Combined effect of smoking
and human papillomavirus type 16 infection in cervical carcinogenesis.
Epidemiology 1998;9:3469.
(3) Lacey JV Jr, Brinton LA, Abbas FM, Barnes WA, Gravitt PE, Greenberg
MD, et al. Oral contraceptives as risk factors for cervical adenocarcinomas
and squamous cell carcinomas. Cancer Epidemiol Biomarkers Prev 1999;
8:107985.
(4) Lacey JV Jr, Frisch M, Brinton LA, Abbas FM, Barnes WA, Gravitt PE, et
al. Associations between smoking and adenocarcinomas and squamous cell
carcinomas of the uterine cervix (United States). Cancer Causes Control
2001;12:15361.
(5) Deacon JM, Evans CD, Yule R, Desai M, Binns W, Taylor C, et al. Sexual
behaviour and smoking as determinants of cervical HPV infection and of
CIN3 among those infected: a casecontrol study nested within the
Manchester cohort. Br J Cancer 2000;83:156572.

Journal of the National Cancer Institute Monographs No. 31, 2003

27

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

Based on key studies on HSIL, CIN3, and CC conducted in

HPV-positive women as determined by accurate polymerase
chain reaction-based HPV DNA-detection methods, we conclude that high parity, smoking, and long-term OC use are cofactors that may modulate the risk of progression from HPV
infection or LSIL to HSIL and CC. The evidence seems to be
more consistent for parity and smoking than for OC use. The
association with OC use was detected only for in situ adenocarcinoma in a U.S. study (3) and in the IARC studies for both
squamous cell carcinoma and adenocarcinoma (9), but no associations were found in the U.K. nested casecontrol study of
CIN3 (5) or in the U.S. prospective cohort study (7).
From a public health point of view, parity seems to be the
behavioral cofactor explaining the highest proportion of CC
cases among HPV-infected women. However, smoking and
long-term OC use may have a similar impact in populations
heavily exposed to HPV and to these cofactors. The studies
reviewed herein indicate that smoking and parity may act both at
the HSIL and the CC stages.
Ongoing prospective studies will shed more light into the role
of these and other cofactors in HPV carcinogenesis; however, if
confirmed, these conclusions imply that multiparous women,
women who are smokers, and women on long-term OC use may
need closer surveillance for cytologic abnormalities and HPV
infections than women in the general population.
Large prospective and retrospective cohort studies of HSIL
and CC among middle-aged women in which several markers of
HPV exposure are used and HPV persistence or chronicity is
documented are still needed to understand the role and impact of
cofactors in cervical carcinogenesis.

(6) Hildesheim A, Herrero R, Castle PE, Wacholder S, Bratti MC, Sherman

ME, et al. HPV co-factors related to the development of cervical cancer:
results from a population-based study in Costa Rica. Br J Cancer 2001;84:
121926.
(7) Castle PE, Wacholder S, Lorincz AT, Scott DR, Sherman ME, Glass AG,
et al. A prospective study of high-grade cervical neoplasia risk among
human papillomavirus-infected women. J Natl Cancer Inst 2002;94:
140614.
(8) Munoz N, Franceschi S, Bosetti C, Moreno V, Herrero R, Smith JS, et al.
Role of parity and human papillomavirus in cervical cancer: the IARC
multicentric casecontrol study. Lancet 2002;359:1093101.
(9) Moreno V, Bosch FX, Munoz N, Meijer CJ, Shah KV, Walboomers JM,
et al. Effect of oral contraceptives on risk of cervical cancer in women with
human papillomavirus infection: the IARC multicentric casecontrol study.
Lancet 2002;359:108592.
(10) Moscicki AB, Hills N, Shiboski S, Powell K, Jay N, Hanson E, et al. Risks
for incident human papillomavirus infection and low-grade squamous intraepithelial lesion development in young females. JAMA 2001;285:
29953002.
(11) Schiffman MH, Bauer HM, Hoover RN, Glass AG, Cadell DM, Rush BB,
et al. Epidemiologic evidence showing that human papillomavirus infection
causes most cervical intraepithelial neoplasia. J Natl Cancer Inst 1993;85:
95864.
(12) International Agency for Research on Cancer (IARC). IARC monographs
on the evaluation of carcinogenic risks to humans. Vol. 64. Human papillomaviruses. Lyon (France): IARC; 1995.
(13) Mitrani-Rosenbaum S, Tsvieli R, Tur-Kaspa R. Oestrogen stimulates differential transcription of human papillomavirus type 16 in SiHa cervical
carcinoma cells. J Gen Virol 1989;70(Pt 8):222732.
(14) Arbeit JM, Howley PM, Hanahan D. Chronic estrogen-induced cervical
and vaginal squamous carcinogenesis in human papillomavirus type 16
transgenic mice. Proc Natl Acad Sci U S A 1996;93:29305.
(15) Elson DA, Riley RR, Lacey A, Thordarson G, Talamantes FJ, Arbeit JM.
Sensitivity of the cervical transformation zone to estrogen-induced squamous carcinogenesis. Cancer Res 2000;60:126775.
(16) International Agency for Research on Cancer (IARC). IARC monographs
on the evaluation of the carcinogenic risk to humans. Vol 72. Hormonal
contraception and post-menopausal hormonal therapy. Lyon (France):
IARC; 1999.
(17) Autier P, Coibion M, Huet F, Grivegnee AR. Transformation zone location
and intraepithelial neoplasia of the cervix uteri. Br J Cancer 1996;74:
48890.
(18) Sethi S, Muller M, Schneider A, Blettner M, Smith E, Turek L, et al.
Serologic response to the E4, E6, and E7 proteins of human papillomavirus
type 16 in pregnant women. Am J Obstet Gynecol 1998;178:3604.
(19) Rous P, Friedwald W. The effect of chemical carcinogens on virus-induced
rabbit papillomas. J Exp Med 1944;79:51138.
(20) Yang X, Jin G, Nakao Y, Rahimtula M, Pater MM, Pater A. Malignant
transformation of HPV 16-immortalized human endocervical cells by cigarette smoke condensate and characterization of multistage carcinogenesis.
Int J Cancer 1996;65:33844.
(21) Prokopczyk B, Cox JE, Hoffmann D, Waggoner SE. Identification of
tobacco-specific carcinogen in the cervical mucus of smokers and nonsmokers. J Natl Cancer Inst 1997;89:86873.
(22) Poppe WA, Ide PS, Drijkoningen MP, Lauweryns JM, Van Assche FA.
Tobacco smoking impairs the local immunosurveillance in the uterine
cervix. An immunohistochemical study. Gynecol Obstet Invest 1995;39:
348.
(23) Giulian AR, Sedjo RL, Roe DJ, Harri R, Baldwi S, Papenfuss MR, et al.
Clearance of oncogenic human papillomavirus (HPV) infection: effect of
smoking (United States). Cancer Causes Control 2002;13:83946.
(24) Szarewski A, Jarvis MJ, Sasieni P, Anderson M, Edwards R, Steele SJ,
et al. Effect of smoking cessation on cervical lesion size. Lancet 1996;347:
9413.
(25) Moreno V, Munoz N, Bosch FX, de Sanjose S, Gonzalez LC, Tafur L,
et al. Risk factors for progression of cervical intraepithelial neoplasm grade
III to invasive cervical cancer. Cancer Epidemiol Biomarkers Prev 1995;
4:45967.
(26) Thomas DB, Qin Q, Kuypers J, Kiviat N, Ashley RL, Koetsawang A, et al.
Human papillomaviruses and cervical cancer in Bangkok. II. Risk factors

(27)

(28)

(29)

(30)

for in situ and invasive squamous cell cervical carcinomas. Am J Epidemiol

2001;153:7329.
Ho GY, Kadish AS, Burk RD, Basu J, Palan PR, Mikhail M, et al. HPV 16
and cigarette smoking as risk factors for high-grade cervical intra-epithelial
neoplasia. Int J Cancer 1998;78:2815.
Woodman CB, Collins S, Winter H, Bailey A, Ellis J, Prior P, et al. Natural
history of cervical human papillomavirus infection in young women:
a longitudinal cohort study. Lancet 2001;357:18316.
Herrero R, Hildesheim A, Bratti C, Sherman ME, Hutchinson M, Morales
J, et al. Population-based study of human papillomavirus infection and
cervical neoplasia in rural Costa Rica. J Natl Cancer Inst 2000;92:46474.
Lazcano-Ponce E, Herrero R, Munoz N, Cruz A, Shah KV, Alonso P, et al.
Epidemiology of HPV infection among Mexican women with normal cervical cytology. Int J Cancer 2001;91:41220.

(31) Sellors JW, Mahony JB, Kaczorowski J, Lytwyn A, Bangura H, Chong S,

et al. Prevalence and predictors of human papillomavirus infection in
women in Ontario, Canada. Survey of HPV in Ontario Women (SHOW)
Group. CMAJ 2000;163:5038.

NOTES
Supported in part by the International Agency for Research on Cancer, Lyon,
France; by grants 01/1237, 01/1236, and BAE 01/5013 from the Fondo
de Investigaciones Sanitarias, Madrid, Spain; and by a Yamagiwa-Yoshida
Memorial International Cancer study grant from the International Union Against
Cancer.
We thank Gina Albero (Institut Catala` dOncologia) for her data handling and
statistical analyses.

Downloaded from http://jncimono.oxfordjournals.org/ by guest on November 5, 2015

28

Journal of the National Cancer Institute Monographs No. 31, 2003