Cancer Causes and Control 14: 805–814, 2003.

805

2003 Kluwer Academic Publishers. Printed in the Netherlands.

Smoking and cervical cancer: pooled analysis of the IARC multi-centric case–control
study

Martyn Plummer1,*, Rolando Herrero2, Silvia Franceschi1, Chris J.L.M. Meijer3, Peter Snijders3, F. Xavier Bosch4,
Silvia de Sanjosé4 & Nubia Muñoz1 for the IARC Multi-centre Cervical Cancer Study Group 
1
International Agency for Research on Cancer, 150 Cours Albert Thomas, F-69372 Lyon, France; 2Proyecto
Epidemiologic Guanacaste, Costa Rica; 3Department of Pathology, VU Medical Center, Amsterdam, The
Netherlands; 4Epidemiology and Cancer Registry Service, Catalan Institute of Oncology (ICO), Barcelona, Spain

Received 24 February 2003; accepted in revised form 24 June 2003

Key words: Cervical cancer, human papillomavirus, smoking.

Abstract

Background: Smoking has long been suspected to be a risk factor for cervical cancer. However, not all previous
studies have properly controlled for the effect of human papillomavirus (HPV) infection, which has now been
established as a virtually necessary cause of cervical cancer. To evaluate the role of smoking as a cofactor of
progression from HPV infection to cancer, we performed a pooled analysis of 10 previously published case–control
studies. This analysis is part of a series of analyses of cofactors of HPV in the aetiology of cervical cancer.
Methods: Data were pooled from eight case–control studies of invasive cervical carcinoma (ICC) and two of
carcinoma in situ (CIS) from four continents. All studies used a similar protocol and questionnaires and included a
PCR-based evaluation of HPV DNA in cytological smears or biopsy specimens. Only subjects positive for HPV
DNA were included in the analysis. A total of 1463 squamous cell ICC cases were analyzed, along with 211 CIS
cases, 124 adeno- or adeno-squamous ICC cases and 254 control women. Pooled odds ratios (OR) and 95%
confidence intervals (CI) were estimated using logistic regression models controlling for sexual and non-sexual
confounding factors.
Results: There was an excess risk for ever smoking among HPV positive women (OR 2.17 95%CI 1.46–3.22). When
results were analyzed by histological type, an excess risk was observed among cases of squamous cell carcinoma for
current smokers (OR 2.30, 95%CI 1.31–4.04) and ex-smokers (OR 1.80, 95%CI 0.95–3.44). No clear pattern of
association with risk was detected for adenocarcinomas, although the number of cases with this histologic type was
limited.
Conclusions: Smoking increases the risk of cervical cancer among HPV positive women. The results of our study are
consistent with the few previously conducted studies of smoking and cervical cancer that have adequately controlled
for HPV infection. Recent increasing trends of smoking among young women could have a serious impact on
cervical cancer incidence in the coming years.

Introduction cause of cervical cancer [1, 2]. This discovery has

Infection with certain types of human papillomavirus
(HPV) has been established as a virtually necessary
* Address correspondence to: Martyn Plummer, Unit of Field and
Intervention Studies, International Agency for Research on Cancer,
150 Cours Albert Thomas, F-69372 Lyon, France. Ph.: +33-472-
738446/738402; Fax: +33-472-738345; E-mail: plummer@iarc.fr
See Appendix 1 for details.
important implications for both primary and secondary
prevention, with the prospect that the disease may
eventually be eliminated by prophylactic vaccination.
Current knowledge of the aetiology of cervical cancer,
however, remains incomplete. Cervical cancer is a rare
consequence of HPV infection and, when it does occur,
it develops many years after initial infection. For these
reasons, current research efforts are oriented towards
806
identifying cofactors that may influence the progression
from infection to cancer. Cofactors may be character-
istics of the virus or host, or may be environmental.
Between 1985 and 1997 the International Agency for
Research on Cancer conducted a series of case–control
studies of cervical cancer; eight studies on invasive
cancer [3–10] and two on carcinoma in situ (CIS) [11,
12]. All studies used a similar protocol and question-
naires and included an accurate evaluation of HPV
DNA in cytological smears or biopsy specimens. This
report is part of a series of papers [13–16] that use the
pooled data from these studies to evaluate possible
cofactors for cervical cancer. We currently believe that
the appropriate way to evaluate potential cofactors,
while controlling for the effect of HPV, is to restrict the
analysis to women with HPV infections.
Smoking has long been associated with cervical cancer
risk [17]. However, it has been difficult to rule out
residual confounding, chiefly from sexual habits, as an
explanation for this association [18, 19]. In particular,
few studies of smoking and cervical cancer have
controlled for HPV infection. This study allows us to
examine the role of smoking while controlling for the
strong effect of HPV infection and, additionally, other
cofactors that have been identified in the same popula-
tions [13, 14].
Subjects and methods
This analysis concerns 1798 HPV positive cases and 254
HPV positive controls from eight separate case–control
studies of invasive cervical cancer (ICC) and two studies
of CIS. The ICC studies were conducted in eight
countries, which showed wide variation in incidence
rates. Regions covered include Thailand [5] and the
Philippines [8] in Asia; Morocco [4] in Africa; Brazil
[6], Peru [10], Paraguay [9] and Colombia [3, 7] in South
America and Spain [3, 7] in Europe. The two CIS
studies were performed in Spain and Colombia at
the same time as the studies of ICC [11, 12]. These have
also been included in this analysis because CIS is
accepted as an immediate precursor of ICC, and
because it shared the same risk factors in our studies
[20].
Methods and findings from each study have been
previously described. In summary, eligible patients were
histologically confirmed, incident ICC or CIS identified
at the reference hospitals. Cases had no previous
treatment for the disease and gave informed consent to
participate in the study. The histological diagnosis was
reviewed in each study either by a single expert, or by a
panel. Cases were histologically classified as squamous
M. Plummer et al.
cell carcinoma, adenocarcinoma and adenosquamous
carcinoma. The last two categories have been combined
for this analysis.
For all studies except the two ICC studies in Spain
and Colombia, controls were hospital based, and were
frequency matched by age in five-year age groups.
Diseases associated with known risk factors for cervical
neoplasia, including tobacco consumption, were exclud-
ed. For the two CIS studies, the controls were popula-
tion based and were individually matched by age and
date of cytology. The matching has been ignored in the
current analysis, with no material effect on the results.
All consenting participants were administered a stan-
dardized questionnaire including information on socio-
demographic factors, sexual behaviour, reproductive
history, contraceptive practices, smoking habits, genital
hygiene, history of STDs and history of cervical
cytological screening. Face-to-face interviews were con-
ducted in hospital by trained interviewers. In the part of
the questionnaire on tobacco, subjects were first asked
to classify themselves as life long non-smokers, current
smokers or ex-smokers (defined as a former smoker who
stopped smoking at least 1 year prior to the interview).
Ever-smokers were then asked the age at which they
started smoking, how many cigarettes per day they
smoked and the total time made up of periods (if any) in
which they stopped smoking. Ex-smokers were asked
the age at which they finally stopped smoking. Further
information on duration and number of cigarettes was
collected by type of tobacco. Information on duration
and number of times per day was also collected for
certain chewing habits: chewing tobacco, taking snuff
and chewing betel quid (with and without tobacco).
After the interview, all subjects had a pelvic examination
performed by a gynaecologist, in which exfoliated cells
for cytological examination and HPV analysis were
taken. A tumour biopsy specimen was also taken in
most cases and kept frozen.
HPV DNA detection
Exfoliated cells were collected using wooden spatulae
and endocervical brushes. Two scrapes were obtained
from each woman. After preparation of one Pap smear,
the remaining cells were eluted in phosphate-buffered
saline (PBS), peleted in PBS and kept frozen at )70 C
until shipment to a central laboratory for HPV testing.
Detailed descriptions of the hybridization assays
employed are given in the individual publications of
the various studies. Briefly, HPV DNA detection and
typing were performed in central laboratories using
PCR amplification methods targeting a small fragment
of the L1 gene. DNA quality was evaluated using
Pooled analysis of smoking and cervical cancer
b-globin primers. In the studies from Colombia and
Spain, MY09 and MY11 consensus primers were used
[21]. In the remaining studies GP5+/6+ primers were
employed. PCR products were assessed for HPV posi-
tivity by low-stringency Southern blot hybridization
using a cocktail of HPV-specific probes [22]. For viral
genotyping the PCR products were successively hybrid-
ized with oligonucleotide probes for 33 HPV types as
previously described [22, 23]. HPV DNA-positive spec-
imens that did not hybridize with any of the 33 probes
were tentatively called HPV X.
In a second step, E7 primers for 14 high-risk HPV
types [2] were used for reamplification of the following
specimens: all case specimens positive for the b-globin
gene and classified as HPV X or HPV DNA negative, all
specimens from control women classified as HPV X, and
a subsample of specimens from control women found to
be b-globin-positive and HPV DNA-negative. Standard
precautions were taken to minimize the risk of false-
positive results in the PCR reactions as described
elsewhere [24]. All PCR assays were performed blind
to the case–control status. Failure to detect HPV DNA
with a negative amplification of b-globin led to an
invalid result and these women have been excluded from
analysis.
Because some of the re-testing was performed at the
completion of the fieldwork in all study areas, numbers
of HPV positive cases and controls of this analysis
sometimes differ from those reported in the original
publications.
Statistical analysis
Logistic regression was used to estimate odds ratios
(ORs) and 95% confidence intervals (95%CI). Likeli-
hood ratio tests were used to assess the significance of
terms in the model. Heterogeneity of the OR between
centres, and other stratifying variables, was tested by
introducing an interaction term between smoking and
the modifying factor.
Most results are presented graphically, plotting the
summary ORs as a black square, whose size is inversely
proportional to the variance of the estimate. A hori-
zontal line represents the 95%CI. Diamonds are used to
plot the summary OR for squamous cell ICC, CIS and
all studies together. The centre of the diamond repre-
sents the pooled OR and the extremes show the limits of
the 95%CI. A log scale is used for the OR in order to
preserve symmetry of the CI.
This report considers only tobacco consumption in
the form of cigarette smoking. Chewing of tobacco or
other substances (coca, betel) and snuff taking was
807
reported in only a few centres. These exposures have
been combined into a single variable for ‘chewing’ which
was controlled for in the final analysis. Other confound-
ers considered and their categories were age (in five year
bands), centre, education (none, primary, secondary+),
number of sexual partners (0–1, 2, 3, 4+), age at first
intercourse (under 16, 17–21, 22+), years of oral
contraceptive use (never, 1–4, 5+ years), parity (0, 1–
2, 3+) and screening history (never, ever, excluding
screening visits 12 months before the interview).
Results
Table 1 shows, for each study area, data on incidence
rates of cervical cancer, prevalence of HPV infection in
cases and controls, and prevalence of tobacco consump-
tion. Incidence rates are taken from Globocan 2000 [25].
They show that Spain is a low-risk country, Morocco,
Thailand and the Philippines are intermediate risk
countries, and the South American countries are high
risk. As expected, prevalence of HPV infection is very
high in cases, but low in controls, ranging from 6% in
Spain to 21% in Morocco. Smoking prevalence in
controls is low in most countries, with the notable
exceptions of Brazil (36% ever smokers), Colombia
(28% ever smokers), and the CIS study in Spain (38%
ever smokers). The average number of cigarettes per day
among smokers is also low (3–11 cigarettes per day in
controls). Chewing is relatively common in only two
centres: Peru where 5% of controls were chewers,
mainly of coca leaves, and Thailand where 20% of
controls were chewers, maily of betel.
Figure 1 summarizes the OR of cervical cancer for
ever-smokers compared with never-smokers in each
study and the pooled results. ORs in this figure were
controlled only for age and centre. Ever smoking is
positively associated with cervical cancer (OR ¼ 2.17,
95%CI 1.46–3.22). The results for each centre are
relatively homogeneous: the formal test for heterogene-
ity has a p-value of 0.11. However, Morocco is a clear
outlier which shows a negative association (OR ¼ 0.27,
95%CI 0.05–1.45). This may be due to a design flaw.
Cases in the Moroccan study were drawn from a
specialist cancer hospital whereas controls were recruit-
ed from a general hospital with a smaller catchment
area. In the original analysis of this data, an attempt was
made to overcome this selection bias by controlling for
social class and urban/rural residence [4]. The OR in
Morocco controlling for education (as a marker of SES)
and residence is 0.63 (95%CI 0.08–4.8). Morocco was
retained in the analysis, but with an additional adjust-
ment for urban/rural residence.
808 M. Plummer et al.
Table 1. HPV prevalence and smoking habits by centre

ASRa HPV tested HPV positive Ever smokers (%)

(world)
Cases Controls Cases (%) Controls (%) Cases Controls

Invasive cancer
Paraguay 41.1 112 90 109 (97.3) 17 (18.9) 32.2 19.0
Peru 39.9 196 175 186 (94.9) 31 (17.7) 9.6 3.6
Colombia 32.9 110 126 86 (78.2) 22 (17.5) 42.4 28.1
Brazil 31.3 187 196 181 (96.8) 34 (17.3) 47.2 36.0
Philippines 22.7 364 381 349 (95.9) 35 (9.2) 20.4 7.2
Thailand 20.7 378 261 363 (96.0) 41 (15.7) 17.1 13.0
Morocco 18.8 188 176 182 (96.8) 38 (21.6) 2.8 4.4
Spain 7.2 159 136 131 (82.4) 8 (5.9) 20.4 14.3
Carcinoma in situ
Colombia 135 181 96 (71.1) 19 (10.5) 37.2 23.7
Spain 157 193 115 (73.2) 9 (4.7) 54.5 38.4
Total 1986 1915 1798 254

a
Age standardized incidence rate from Globocan 2000 [25].
b
Among ever smokers.

Fig. 1. OR for ever smoking by study.

As a further test to ensure that the results are not was recalculated after leaving out each study in turn.
unduly influenced by any single study, the pooled OR The pooled OR remained fairly stable, ranging from
Pooled analysis of smoking and cervical cancer
Fig. 2. ORs of squamous cell carcinoma of the cervix.
1.94 (95%CI 1.29–2.94) when the Philippines were
removed to 2.37 (95%CI 1.54–3.67) when the ICC
study in Colombia was removed.
Figure 2 shows pooled results for various aspects of
smoking history for squamous cell carcinoma (including
CIS), controlling for age, centre, education, number of
sexual partners, age at first intercourse, oral contracep-
tive use, parity and screening. Both current and ex-
smokers showed elevated risks. However, among ever
smokers, there is no evidence of increasing risk with
years of smoking, number of cigarettes per day or age at
starting smoking. In order to confirm this finding, trend
tests were conducted for these three risk factors, treating
them as continuous variables. The resulting p-values
were 0.99 for duration, 0.90 for number of cigarettes per
day and 0.52 for age at starting smoking. Adenocarci-
Fig. 3. ORs for ever smoking stratified by age and number of sexual partners.
809
noma and adenosquamous carcinoma were also exam-
ined but since the number of cases was small, the results
are not presented in detail. There were 82 HPV positive
adenocarcinomas (2.0% of cases) and 42 adenosqua-
mous carcinomas (1.0% of cases). When these were
combined, the OR for ever smokers was 1.64 (95%CI
0.64–4.21). When adenocarcinomas were examined
on their own the OR was 2.19 (95%CI 0.74–6.47).
The results for these histological types are therefore
compatible with no excess risk for smokers, but they are
also compatible with the results for squamous cell
carcinoma.
Figure 3 shows the modifying effect of age and
number of sexual partners on the OR for ever smoking.
Results are presented for all histological types com-
bined. There is no strong evidence of heterogeneity
810 M. Plummer et al.
Table 2. ORs of HPV-positivity and 95% CI according to reproductive factors among control women

HPV prevalence (%) HPV positive/HPV negative ORa (95%CI)

Smoking status
Never 14 218/1369 1.00
Current 10 22/195 0.85 (0.52, 1.39)
Ex 13 14/97 0.85 (0.47, 1.56)
Test for heterogeneity p = 0.72
No. of years smoking
Never 14 218/1369 1.00
1–19 11 19/158 1.04 (0.61, 1.77)
‡20 11 15/127 0.65 (0.36, 1.16)
Test for trend p = 0.21
No. cigarettes/day
Never 14 218/1369 1.00
<6 12 17/129 0.84 (0.49, 1.45)
‡6 11 18/151 0.92 (0.53, 1.58)
Test for trend p = 0.61
Age started smoking
Never 14 218/1369 1.00
‡20 14 18/117 1.07 (0.63, 1.83)
15–19 10 13/111 0.93 (0.50, 1.72)
<15 8 5/59 0.45 (0.17, 1.18)
Test for trend p = 0.22
Years since quitting
Current smoker 10 22/195 1.00
1–10 15 10/55 1.52 (0.62, 3.74)
‡11 7 3/38 0.60 (0.16, 2.35)
Test for trend p = 0.78

a
Controlling for study centre and age.

across either age groups (p ¼ 0.16 for heterogeneity test)
or number of sexual partners (p ¼ 0.17).
The association with smoking was also tested after
further restriction to subjects infected with high risk
HPV types (16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56,
58, 59, 66, 68, 73, 82). This subgroup comprised 93% of
the HPV positive cases (1667 cases) and 63% of the
HPV positive controls (161 controls). The association
with ever smoking was not changed (OR ¼ 2.14, 95%CI
1.22–3.78, controlling for all confounders listed above).
There was also no evidence of any dose–response
relationship (data not shown).
In addition to analysing smoking as a risk factor for
cervical cancer among HPV positive women, smoking
was also examined as a risk factor for HPV acquisition
among controls. The results are presented in Table 2
and show no association between smoking behaviour
and HPV.
Discussion
Our present pooled analysis of 1798 cases or cervical
carcinoma and 254 control women who were infected
with HPV shows that ever-smokers have an excess risk of
cervical cancer that persists after controlling for the
strong effect of HPV and for other potential cofactors of
progression from infection to cancer. The low level of
smoking in most of the study populations hampered our
ability to study the presence of a dose-response effect.
The hypothesis that smoking is a risk factor for
cervical cancer was first put forward in 1977 by
Winkelstein [17], who also periodically reviewed the
support for this hypothesis from subsequent epidemio-
logical studies [26, 27]. Despite a consistent association
between smoking and cervical cancer after an adjust-
ment for sexual behaviour, it was not generally agreed
that confounding could be ruled out as an explanation
for this finding, since sexual behaviour was evidently a
surrogate for a sexually transmitted agent [18, 19, 28].
The hypothesis that HPV is the causal agent was first
put forward by zur Hausen [29]. Now that HPV has
been identified as the principal cause of cervical cancer,
it should be possible to resolve the controversy over
smoking. However, relatively few studies have con-
trolled for HPV when examining smoking. Of these,
some have used a relatively inaccurate HPV DNA
detection method [30] or anti-HPV antibodies [31], and/
or have concentrated on a limited range of HPV types
[30–33].
Pooled analysis of smoking and cervical cancer
The evidence for pre-invasive cervical neoplasias is
larger than for invasive ones. In a clinic-based case–
control study of high grade dysplasia in the USA, Becker
et al. [34] found an increased risk for current smokers
that persisted after adjusting for HPV, age at first sexual
intercourse, number of sexual partners and ethnicity
(OR ¼ 1.8, 95%CI 1.2–2.8). No excess risk was observed
for ex-smokers. Kjellberg et al. [35] conducted a popu-
lation-based study of high-grade cervical intraepithelial
neoplasia (CIN) in Sweden. An excess risk of CIN 2–3
was observed for ever-smoking that persisted after
adjusting for HPV DNA positivity (OR 2.5, 95%CI
1.3–4.9). Deacon et al. [36] conducted a nested case–
control study of CIN 3 in England. In contrast to the
previously reported studies, which used multiple regres-
sion to control for the effect of HPV, Deacon et al.
separately analyzed risk factors for progression among
HPV positive women and risk factors for presence of
HPV among control women. Ever smoking was a risk
factor for CIN 3 among HPV positive women (OR ¼ 2.2,
95%CI 1.44–3.35). The excess risk persisted after adjust-
ing for age at first intercourse, number of sexual partners,
spontaneous abortion and time since start of latest
regular relationship. There was also strong evidence of a
dose–response effect. By contrast, smoking was not
associated with HPV DNA positivity among controls
(OR ¼ 0.90, 95%CI 0.60–1.35). Hildesheim et al. [37]
conducted a population based study in Costa Rica using
prevalent cases of HSIL or cancer as an outcome. Both
current smokers and ex-smokers were at increased risk,
with ORs of 2.4 (95%CI 1.2–5.1) and 2.3 (95%CI 1.3–
4.3) respectively, after controlling for age, HPV type
(high-risk versus low-risk) and parity. Finally, Lacey
et al. [38] conducted a multi-centre case–control study of
cervical adenocarcinoma in the USA, using community
controls and, additionally an age-matched group of SCC
cases [38]. After controlling for HPV, education, number
of sexual partners and screening history, ever smoking
was not associated with adenocarcinoma (OR ¼ 0.8,
95%CI 0.5–1.2) or SCC (OR ¼ 1.4, 95%CI 0.8–2.3).
There was limited evidence that heavy smoking was
associated with SCC (OR ¼ 1.8, 95%CI 1.0–3.3 for
‡1 pack/day) but not with adenocarcinoma (OR ¼ 0.7,
95%CI 0.4–1.3 for ‡1 pack/day). The results of the
present analysis are therefore broadly consistent with
other case–control studies that have controlled for HPV
infection using high quality HPV assays. In one of the
few available prospective studies, tobacco smoke was
associated with progression from CIN I to CIN III
among HPV 16 positive women [39]. Significantly higher
regression rates of cervical precancerous lesions have also
been reported for women who quit smoking compared to
those who continue with the habit [40].
811
Various possible mechanisms have been proposed for
the association between smoking and cervical cancer. In
animal models, more than 50 years ago Rous and
Friedwald [41] described the malignant conversion of
cottontail rabbit papillomavirus after exposure to me-
thilcolanthrene or tar. More recently, malignant trans-
formation has been observed when endocervical cells
immortalized by HPV 16 are exposed to cigarette smoke
condensate, whereas untreated cells remain non-tumori-
genic in nude mice [42].
Tobacco is able to induce its carcinogenic effect in
sites not directly exposed to cigarette smoke, as evi-
denced by associations of smoking with pancreatic,
kidney and bladder cancer [43]. In the cervix, it is
possible to detect nicotine derivatives like cotinine and
tobacco specific nitrosamines (recently reviewed by
Szarewski and Cuzick [44]). In addition, DNA adducts
[45] and other evidence of genotoxic damage [46] are
detectable in cervical exfoliated cells. It has been shown
that smoking affects local immune mechanisms in the
cervix, with reductions in the number of Langerhans
cells and other markers of immune function [47].
Another possibility is a systemic effect of smoking, by
which smokers would have an altered metabolism of
female hormones compared with non-smokers. The
association presented here with smoking remains after
controlling for oral contraceptive use and parity, but we
do not have sufficient data to analyze the interactions
between these exposures.
Our approach of restricting the analysis to HPV
positive women is based on our current understanding
of HPV as a virtually necessary cause of cervical cancer.
This implies that if smoking is a risk factor for cervical
cancer, it may act either by increasing the risk of
acquiring an HPV infection, or by increasing the risk of
progression from HPV infection to cancer. We found no
association between smoking and HPV infection among
controls. Conversely, we found an increased risk of cer-
vical cancer for smokers among HPV positive women.
Hence we conclude that smoking acts as a risk factor for
progression once HPV infection is acquired. This
restriction should be borne in mind when interpreting
the results: the ORs presented here cannot be extrapo-
lated to the general population, but are restricted to
HPV positive women. In order to check our conclusions
against a more traditional analysis, we re-examined the
relationship between ever smoking and risk of cervical
cancer by multiple logistic regression using the full set of
subjects on whom an HPV test result was available
(1986 cases and 1915 controls). The OR for ever-
smoking controlling for HPV, age and centre was 2.32
(95%CI 1.78–3.04). Further control for education,
screening, parity, number of sexual partners, oral
812
contraceptive use and age at first intercourse did not
substantially alter this result (OR 2.31, 95%CI 1.72–
3.11).
One of the limitations of restricting the analysis to
HPV positive women in case control studies is that only
the current HPV status is available. Ideally the analysis
should be restricted to women with chronic infections
since this is the underlying risk factor. Whereas all cases
will necessarily have chronic HPV infections, this is not
necessarily true for all HPV positive controls. This
means that control for chronic HPV infection might be
incomplete. We have examined this possibility by testing
the modifying effect of age and number of sexual
partners – two risk factors that may be associated with
chronic infection. The null results of this test suggest
that chronic infection has been adequately controlled for
in the statistical model.
Apart from HPV infection, we have tried to control
for three separate causes of confounding: screening,
sexual behaviour and reproductive factors. Screening
has been observed to be protective against ICC in most
of these populations. Sexual behaviour is a surrogate for
other sexually transmitted infections that may accelerate
the progression to cancer. Chlamydia and Herpes
Simplex virus 2 are two such infectious agents that will
be investigated more closely in future papers. Parity and
oral contraceptive have been previously identified in
these studies as hormonal cofactors of progression in
this pooled data [13, 14].
Smoking is rare among women in developing coun-
tries and in Spain. Ideally, the relationship between
smoking and cervical cancer should be addressed in
populations in which smoking is more common among
women. However, opportunities to study large numbers
of women with invasive cervical cancer are rare, and we
have taken advantage of the possibilities offered by this
series of studies with a common protocol and accurate
HPV measurements.
Selection bias is always a potential problem in case–
control studies. This issue has been discussed in the
original reports for these studies, with the generally
reassuring conclusions that there is no problem. The
only exception is Morocco, in which we have attempted
to overcome the evident selection bias by controlling for
urban/rural residence. As smoking is associated with a
broad range of diseases, hospital controls may include
an excess of smokers compared to the source popula-
tion. However, in our studies patients in hospital for
smoking-related diseases were not eligible as control
women.
In conclusion, smoking increases the risk of cervical
cancer among HPV positive women. The present pooled
analysis adds to the limited number of studies that have
M. Plummer et al.
adequately controlled for HPV infection when examin-
ing the association between smoking and cervical
cancer. Taken together, this body of evidence suggests
that squamous cell carcinoma of the cervix should be
added to the list of tobacco associated cancers, while for
adenocarcinoma, further data are warranted. This
conclusion was supported by an expert working group
convened by the International Agency for Research on
Cancer in June 2002 [48].
Several countries are experiencing an increase in the
prevalence of smoking, particularly among younger
women [49]. This increase in smoking among women
could have a serious impact on the incidence of cervical
cancer in the future, emphasizing the need for intensi-
fying preventive efforts [50].
Appendix 1
The group is composed of the following researchers:
S. Franceschi, M. Plummer, J. Smith, N. Muñoz, IARC, Lyon,
France; F.X. Bosch, V. Moreno, S. de Sanjosé, X. Castellsagué X,
ICO, Barcelona, Spain; R. Herrero, Proyecto Epidemiologico Gua-
nacaste, Costa Rica; C.J.L.M. Meijer, J.M.M. Walboomers and P.J.F.
Snijders, Free University Hospital, Amsterdam, the Netherlands; K.V.
Shah, Johns Hopkins University, Baltimore, Maryland; S. Chichareon,
Prince of Songkla University, Hat-Yai, Thailand; C. Ngelangel,
University of The Philippines, Manila, the Philippines; N. Chaoki,
B. El Gueddari, Institut National d’Oncologie, Rabat, Morocco;
J. Eluf-Neto, Universidade de Sao Paulo, Sao Paulo, Brazil; P.A.
Rolón, Laboratorio de Anatomia Patológica y Citologia, Asunción,
Paraguay; C. Santos, E. Cacares, Maes Heller Cancer Research
Institute, Lima, Peru; S. Bayo, Institut National de Recherche en Santé
Publique, Bamako, Mali; I. Izarzugaza, Euskadi Cancer Registry,
Vitoria Gasteiz, Spain; M. Gili, Catedra Reynals, Barcelona, Spain;
L.A. Tafur, University of Valle, Cali, Colombia; C. Navarro, Health
Council, Murcia, Spain; N. Ascunce, Breast Cancer Prevention Centre,
Pamplona, Spain; L.C. González, Delegation of Social Welfare,
Salamanca, Spain; M. Santamaria, Navarra Hospital, Pamplona,
Spain; P. Alonso de Ruiz, General Hospital of Mexico, Mexico City,
Mexico; N. Aristizabal, Cali, Colombia; J. Deacon, Institute of Cancer
Research, Belmont, United Kingdom.
Acknowledgements
This research was funded by the European Community
CI 1-0371-F(CD); The Fondo de Investigaciones Sani-
tarias (FIS), Spain 86/753, 87/1513, 88/2049, 90/0901,
95/0955; Preventiefonds, The Netherlands 28-1502.1;
Programa Interministerial de Investigación y Desarrol-
lo, Spain SAF 96/0323, The Conselho Nacional de
Desenvolvimiento Cientifico e Tecnologico, Brazil
(CNPq) JEN-204453/88.7 and the Department of Re-
productive Health & Research at the World Health
Organization, grant no. 98.101.
Pooled analysis of smoking and cervical cancer
References
1. IARC (1995) IARC Monographs on the Evaluation of Carcinogenic
Risks to Humans, volume 64. Human Papillomaviruses. Lyon:
International Agency for Research on Cancer.
2. Walboomers JM, Jacobs MV, Manos MM, et al. (1999) Human
papillomavirus is a necessary cause of invasive cervical cancer
worldwide. J Pathol 189: 12–19.
3. Bosch FX, Munoz N, de Sanjose S, et al. (1992) Risk factors for
cervical cancer in Colombia and Spain. Int J Cancer 52: 750–758.
4. Chaouki N, Bosch FX, Munoz N, et al. (1998) The viral origin of
cervical cancer in Rabat, Morocco. Int J Cancer 75: 546–554.
5. Chichareon S, Herrero R, Munoz N, et al. (1998) Risk factors for
cervical cancer in Thailand: a case–control study. J Natl Cancer
Inst. 90: 50–57.
6. Eluf-Neto J, Booth M, Munoz N, Bosch FX, Meijer CJ,
Walboomers JM (1994) Human papillomavirus and invasive
cervical cancer in Brazil. Br J Cancer 69: 114–119.
7. Munoz N, Bosch FX, de Sanjose S, et al. (1992) The causal link
between human papillomavirus and invasive cervical cancer: a
population-based case–control study in Colombia and Spain. Int
J Cancer 52: 743–749.
8. Ngelangel C, Munoz N, Bosch FX, et al. (1998) Causes of cervical
cancer in the Philippines: a case–control study. J Natl Cancer Inst
90: 43–49.
9. Rolon PA, Smith JS, Munoz N, et al. (2000) Human papilloma-
virus infection and invasive cervical cancer in Paraguay. Int
J Cancer 85: 486–491.
10. Santos C, Munoz N, Klug SJ, et al. (2001) HPV types and
cofactors causing cervical cancer in Peru. Br J Cancer 85: 966–971.
11. Bosch FX, Munoz N, de Sanjose S, et al. (1993) Human
papillomavirus and cervical intraepithelial neoplasia grade III/
carcinoma in situ: a case–control study in Spain and Colombia.
Cancer Epidemiol Biomarkers Prev 2: 415–422.
12. Munoz N, Bosch FX, de Sanjose S, et al. (1993) Risk factors for
cervical intraepithelial neoplasia grade III/carcinoma in situ in
Spain and Colombia. Cancer Epidemiol Biomarkers Prev 2: 423–431.
13. Moreno V, Bosch FX, Munoz N, et al. (2002) Effect of oral
contraceptives on risk of cervical cancer in women with human
papillomavirus infection: the IARC multi-centric case control
study. Lancet 359: 1085–92.
14. Munoz N, Franceschi S, Bosetti C, et al. (2002) The role of parity
and human papillomavirus in cervical cancer: the IARC multi-
centric case–control study. Lancet 359: 1093–1101.
15. Smith JS, Herrero R, Bosetti C, et al. (2002) Herpes simplex virus-
2 as a human papillomavirus cofactor in the etiology of invasive
cervical cancer. J Natl Cancer Inst 94: 1604–1613.
16. Castellsague X, Bosch FX, Munoz N, et al. (2002) Male circum-
cision, penile human papillomavirus infection, and cervical cancer
in female partners. N Engl J Med 346: 1105–1112.
17. Winkelstein Jr W. (1977) Smoking and cancer of the uterine cervix:
hypothesis. Am J Epidemiol 106: 257–259.
18. Brinton LA (1990) Editorial commentary: smoking and cervical
cancer – current status. Am J Epidemiol 131: 958–960.
19. Layde PM (1989) Smoking and cervical cancer: cause or coinci-
dence? JAMA 261: 1631–1633.
20. Moreno V, Munoz N, Bosch FX, et al. (1995) Risk factors for
progression of cervical intraepithelial neoplasm grade III to
invasive cervical cancer. Cancer Epidemiol Biomarkers Prev 4:
459–467.
21. Manos MM, Wright DK, Lewis AJ, Broker TR, Wolinsky SM
(1989) The use of polymerase chain reaction amplification for the
detection of genital human papillomaviruses. In: Furth M, Greaves
813
M, eds. Molecular Diagnostics of Human Cancer. New York: Cold
Spring Harbor Press, pp. 209–214.
22. Jacobs MV, de Roda Husman AM, van den Brule AJ, Snijders PJ,
Meijer CJ, Walboomers JM (1995) Group-specific differentiation
between high- and low-risk human papillomavirus genotypes by
general primer-mediated PCR and two cocktails of oligonucleotide
probes. J Clin Microbiol 33: 901–905.
23. Roda Husman AM, Walboomers JM, Meijer CJ, et al. (1994)
Analysis of cytomorphologically abnormal cervical scrapes for the
presence of 27 mucosotropic human papillomavirus genotypes,
using polymerase chain reaction. Int J Cancer 56: 802–806.
24. Walboomers JMM, Melkert PWJ, van den Brule AJC, Snijders
PJF, Meijer CJLM (1992) The polymerase chain reaction for
human papillomavirus screening in diagnostic cytopathology of
the cervix. In Herrington CS, McGee O, eds. Diagnostic molecular
pathology: a practical approach. Oxford: Oxford University Press,
pp. 152–172.
25. Ferlay J, Bray F, Pisani P, Parkin DM (2001) Globocan 2000:
Cancer Incidence, Mortality and Prevalence Worldwide. Lyon:
IARC Press.
26. Winkelstein Jr W, Shillitoe EJ, Brand R, Johnson KK (1984)
Further comments on cancer of the uterine cervix, smoking, and
herpesvirus infection. Am J Epidemiol 119: 1–8.
27. Winkelstein Jr W. (1990) Smoking and cervical cancer–current
status: a review. Am J Epidemiol 131: 945–957.
28. Phillips AN, Smith GD (1994) Cigarette smoking as a potential
cause of cervical cancer: has confounding been controlled? Int J
Epidemiol 23: 42–49.
29. zur Hausen H (1976) Condylomata acuminata and human genital
cancer. Cancer Res 36: 794.
30. Herrero R, Brinton LA, Reeves WC, et al. (1989) Invasive cervical
cancer and smoking in Latin America. J Natl Cancer Inst 81: 205–
211.
31. Hakama M, Luostarinen T, Hallmans G, et al. (2000) Joint effect
of HPV16 with Chlamydia trachomatis and smoking on risk of
cervical cancer: antagonism or misclassification (Nordic countries).
Cancer Causes Control 11: 783–790.
32. Olsen AO, Dillner J, Skrondal A, Magnus P (1998) Combined
effect of smoking and human papillomavirus type 16 infection in
cervical carcinogenesis. Epidemiology 9: 346–349.
33. Ylitalo N, Sorensen P, Josefsson A, et al. (1999) Smoking and oral
contraceptives as risk factors for cervical carcinoma in situ. Int J
Cancer 81: 357–365.
34. Becker TM, Wheeler CM, McGough NS, et al. (1994) Cigarette
smoking and other risk factors for cervical dysplasia in south-
western Hispanic and non-Hispanic white women. Cancer Epi-
demiol Biomarkers Prev 3: 113–119.
35. Kjellberg L, Hallmans G, Ahren AM, et al. (2000) Smoking, diet,
pregnancy and oral contraceptive use as risk factors for cervical
intra-epithelial neoplasia in relation to human papillomavirus
infection. Br J Cancer 82: 1332–1338.
36. Deacon JM, Evans CD, Yule R, et al. (2000) Sexual behaviour and
smoking as determinants of cervical HPV infection and of CIN3
among those infected: a case–control study nested within the
Manchester cohort. Br J Cancer 83: 1565–1572.
37. Hildesheim A, Herrero R, Castle PE, et al. (2001) HPV co-factors
related to the development of cervical cancer: results from
a population-based study in Costa Rica. Br J Cancer 84: 1219–
1226.
38. Lacey Jr JV, Frisch M, Brinton LA, et al. (2001) Associations
between smoking and adenocarcinomas and squamous cell carci-
nomas of the uterine cervix (United States). Cancer Causes Control
12: 153–161.
814
39. Ho GY, Kadish AS, Burk RD, et al. (1998) HPV 16 and cigarette
smoking as risk factors for high-grade cervical intra-epithelial
neoplasia. Int J Cancer 78: 281–285.
40. Szarewski A, Jarvis MJ, Sasieni P, et al. (1996) Effect of
smoking cessation on cervical lesion size. Lancet 347: 941–
943.
41. Rous P, Friedwald W (1944) The effect of chemical carcinogens on
virus-induced rabbit papillomas. J Exp Med 79: 511–538.
42. Yang X, Jin G, Nakao Y, Rahimtula M, Pater MM, Pater A
(1996) Malignant transformation of HPV 16-immortalized
human endocervical cells by cigarette smoke condensate and
characterization of multistage carcinogenesis. Int J Cancer 65: 338–
344.
43. IARC (1986) IARC Monographs on the Evaluation of Carcinogenic
Risks to Humans, Volume 38. Tobacco Smoking. Lyon: Interna-
tional Agency for Research on Cancer.
44. Szarewski A, Cuzick J (1998) Smoking and cervical neoplasia: a
review of the evidence. J Epidemiol Biostatist 3: 229–256.
45. Simons AM, Mugica vH, Rodriguez JA, et al. (1995) Demonstra-
tion of smoking-related DNA damage in cervical epithelium and
M. Plummer et al.
correlation with human papillomavirus type 16, using exfoliated
cervical cells. Br J Cancer 71: 246–249.
46. Cerqueira EM, Santoro CL, Donozo NF, et al. (1998) Genetic
damage in exfoliated cells of the uterine cervix. Association and
interaction between cigarette smoking and progression to malig-
nant transformation? Acta Cytol 42: 639–649.
47. Poppe WA, Ide PS, Drijkoningen MP, Lauweryns JM, Van Assche
FA (1995) Tobacco smoking impairs the local immunosurveillance
in the uterine cervix. An immunohistochemical study. Gynecol
Obstet Invest 39: 34–38.
48. IARC 2003. IARC Monographs on the Evaluation of Carcinogenic
Risks to humans, Volume 83. Tobacco Smoke and Involuntary
Smoking. Lyon, International Agency for Research on Cancer, (in
press).
49. Dobson AJ, Kuulasmaa K, Moltchanov V, et al. (1998) Changes
in cigarette smoking among adults in 35 populations in the mid-
1980s. WHO MONICA Project. Tob Control 7: 14–21.
50. Ernster V, Kaufman N, Nichter M, Samet J, Yoon SY (2000)
Women and tobacco: moving from policy to action. Bull World
Health Organ 78: 891–901.