Sie sind auf Seite 1von 14

1888

The Epidemiology of
Cervical Carcinogenesis
Mark H. Schiffman, M.D., M.P.H., and Louise A. Brinton, Ph.D.

Epidemiologic and laboratory data suggest that cervical to human papillomavirus (HPV) infection.' Although
cancer typically arises from a series of causal steps. Each cervical cancer ranks as the second leading cause of can-
step can be studied separately in the hope of better etio- cer in women, it is nonetheless uncommon compared
logic understanding and improved cancer prevention. with the extremely high lifetime cumulative incidence
The earliest identified etiologic step is infection of young of cervical infection with HPV. Thus, additional etio-
women with specific types of venereally transmissible
logic steps and associated risk factors other than HPV
human papillomaviruses (HPVs). Cervical HPV infec-
tions often lead to low grade squamous intraepithelial le-
infection must be critically involved in cervical cancer
sions (mildly abnormal Pap smears). Human papillo- pathogenesis.
mavirus infections and their associated lesions are ex- Epidemiologic understanding of the multistage
tremely common among young, sexually active women. pathogenesis of cervical cancer, though fragmentary, ri-
The infections typically resolve spontaneously even at vals our understanding of any other malignancy. As a
the molecular level within months to a few years. Un- result, it may soon be possible, using a variety of molec-
commonly, HPV infections and/or low grade lesions per- ular epidemiologic approaches, to define new preven-
sist and progress to high grade lesions. The risk factors tion strategies even more effective than cervical cyto-
for progression are mainly unknown but include HPV logic screening alone. For example, HPV DNA testing
type and intensity, cell-mediated immunity, and repro- may prove to be a useful adjunct to Pap smears. In the
ductive factors. Nutritional factors or co-infection with
long term, the most exciting possibility is the primary
other pathogens may also be involved at this apparently
critical etiologic step between common low grade and un-
prevention of cervical neoplasia through HPV immuni-
common high grade intraepithelial lesions. Except for ad- zation of the general population.
vancing age, no epidemiologic risk factors have been In the hope of promoting improved preventive
found for the next step between high grade intraepithelial strategies, we summarized the epidemiology of cervical
lesions and invasive cancer. At the molecular level, inva- carcinogenesis. The choice of the term "carcinogenesis"
sion is associated with integration of viral DNA. Based on rather than "cancer" is intentional. Greater knowledge
worldwide research, the steps in cervical carcinogenesis of pathogenesis implies increased complexity for epide-
appear to be fundamentally the same everywhere, with a miologists accustomed to conceptualizing cancer as a
central role for HPV infection. The importance of etio- yes/no phenomenon. The epidemiology of invasive
logic cofactors like smoking, however, may vary by re- cancer now includes the natural history of its increas-
gion. Cancer 1995;76:1888-901.
ingly severe grades of precursor lesions, starting with
Key words: cervix, carcinoma, human papillomavirus in- cervical HPV infection.
fection, epidemiology.
Definitions and Background
The majority of the 440,000 cases of cervical cancer es-
timated to occur worldwide each year' can be attributed Cervical Neoplasia

Consistent with an origin in the transformation zone,


Presented at the American Cancer Society National Conference
on Gynecologic Cancers, Washington, DC, April 6-8, 1995. most cervical cancers (approximately 8O%) are squa-
The abstract of this paper or a slightly modified version was pub- mous cell carcinomas, with adenocarcinomas and
lished in the May 15, 1995, issue of Cancer. mixed adenosquamous tumors accounting for most of
From the Epidemiology and Biostatistics Program, National the remainder. Other histologic types, such as melano-
Cancer Institute, Bethesda, Maryland. mas, sarcomas, and metastatic tumors, are very rare.
Address for reprints: Mark H. Schiffman, M.D., M.P.H., Epide-
miology and Biostatistics Program, National Cancer Institute, The relative and absolute frequencies of adenocarcino-
Bethesda, MD 20892-7374. mas are rising, particularly among younger women, for
Received June 20, 1995; accepted July 12, 1995. reasons that are poorly ~ n d e r s t o o d .Almost
~ , ~ all epide-
Epidemiology of Cervical Carcinogenesis/SchifJman and Brinton 1889

miologic studies of cervical cancer have focused on with undifferentiated, immortalized, atypical cells. De-
squamous carcinomas or have ignored histologic dis- spite inevitable misclassification from such a continuum
tinctions altogether. of changes, the Bethesda System remains the best avail-
Squamous carcinomas of the cervix result from the able simplifying schema for epidemiologic research.
progression of preinvasive precursor lesions called "cer-
vical intraepithelial neoplasia" (CIN), "dysplasia," Human Papillomavirus Infection
"dyskaryosis" (British), and various other names. To
bring order to the diagnostic confusion, a new cytology Human papillomavirus infection is the primary risk fac-
classification called the Bethesda System was recently tor for cervical cancer and for certain other anogenital
introduced in the United state^.^-^ The Bethesda Sys- cancer^.^,'^ Interested readers are referred elsewhere for
tem combines clinically similar intraepithelial diagnoses a thorough review of papillomavirus biology" or for the
into broad categories, specifically, low grade squamous epidemiology of HPV infection itself apart from cervical
intraepithelial lesions (SIL) and high grade SIL. The neoplastic outcomes.l 2
new classification was designed for use in cytologic Human papillomaviruses are nonenveloped, dou-
screening. It remains technically more correct to use the ble-stranded DNA viruses of approximately 8000 base
more detailed CIN scale when discussing histopatho- pairs, part of a large group of papillomaviruses that in-
logic terms (i.e., biopsies). Nonetheless, we have pri- cludes wart viruses of cattle, cotton-tailed rabbits, deer,
marily used the Bethesda System in the current paper and horses. There are more than 70 types of HPVs char-
because it has proven to be very useful for epidemio- acterized according to DNA sequence homology, with a
logic studies of multistage cervical carcinogenesis. few more identified every year. The types are numbered
Low grade SIL combines CIN 1 (mild dysplasia) sequentially when they are characterized. Each type has
with the cytologic diagnosis of HPV infection,' called its own tissue predilection and disease spectrum. For ex-
"koilocytotic condylomatous atypia". The two older ample, HPV 1 is the major cause of deep plantar warts;
categories shared virtually the same epidemiologic and HPV 2 and 4, common skin warts; and HPV 6 and 11,
HPV DNA profiles and were so overlapping morpho- venereal warts (condyloma acuminatum) as well as la-
logically as to be indivisible. Typically, koilocytotic ryngeal polyps. Based on current data, types 6, 11, 16,
atypia was about 2-3 times as common a diagnosis as 18, 26, 31, 33, 35, 39, 42, 43, 44, 45, 51, 52, 53, 54, 55,
CIN 1. Of the two diagnoses, CIN 1 was thought to be 56, 58, 59, 66, and 68 are the types most commonly
slightly more severe and more closely linked to "pre- found to infect cervical epithelia.
cancer." Accordingly, the new category of low grade The epidemiologic study of HPV has been limited
SIL is much larger and, on average, slightly less severe by HPV measurement techniques. Reliable serologic as-
than CIN 1. says are unavailable for definition of cumulative life-
The distinction between low and high grade SIL is time incidence of HPV infe~tion,'~ although serologic
important for epidemiologists studying cervical carcino- tests that detect the majority of recent infections have
genesis. Low grade SIL is common and represents the just been deve10ped.I~Cervical HPV infection is still
usually benign cytopathologic signs of HPV infection. most accurately measured by current detection of HPV
In contrast, high grade SIL is rare and represents a truly DNA sequences in infected tissues.
premalignant lesion in the most severe cases (carcinoma Epidemiolopts studying cervical HPV infection
in situ). Although low grade SIL can be viewed as an have relied on DNA testing of cervical specimens ob-
epidemiologic exposure or risk factor for cervical can- tained noninvasively with swabs, scrapes, brushings,
cer, high grade SIL can be viewed as more closely linked and lavages. The prevalence estimates of HPV have
to the cancer outcome. varied accordingly, due to differences in cell ampl ling,'^
This useful conceptual distinction is not perfect, the poorly-understood intermittency of viral DNA de-
however, because there exists a continuum of changes te~tability,'~,'~ and, most importantly, the choice of
encompassing low and high grade SIL without a clear DNA detection method.'' Misclassification resulting
cut point. At the microscopic level, for example, the from the first, poorly validated DNA tests severely lim-
characteristic cells of low grade SIL are abnormal but ited early epidemiologic studies of HPV infection."
terminally differentiated. The atypical cells progress to Essentially, there are two categories of HPV DNA
the surface, produce keratins, die, and slough, as would detection methods used in population studies: those
normal cells. The gradient from low grade to high grade that identify the nucleic acids directly and those that
SIL (CIN 2-3) is characterized by increasing nuclear amplify nucleic acids first and then detect the amplified
atypia and failure of cellular differentiation in progres- product. 18,20,21 In the first category are Southern blot hy-
sively more superficial levels of epithelium, with carci- bridization, dot blot hybridization (e.g., ViraPap and
noma in situ representing full-thickness replacement Profile kits, Digene Diagnostics, Silver Spring, MD),
1890 CANCER Supplement November 25,2995, Volume 76,No. 10

and Hybrid Capture liquid hybridization kit (DigeneDi- Regarding a logical time sequence, HPV infection
agnostics). The only amplification methods currently (as measured by DNA) tends to precede and predict in-
used for HPV epidemiology are polymerase chain reac- cidence of cervical neoplasia. Early results from large
tion (PCR)-based techniques. In general, PCR-based prospective studies of cytologically normal women
tests yield HPV population prevalence estimates about show substantially elevated relative (> 10) and absolute
2-3 times higher than those of nonamplified tests. (>60%) risks of incident SIL, including high grade SIL,
However, if ample specimen is tested, the detection of within a few years of viral DNA detection2’ (Schiffman
HPV DNA in prevalent cases of SIL or cancer is similar MH, Manos MM, Sherman ME, Rush BB, Lawler P,
regardless of whether nonamplified or amplified tests Scott DR, et al. Unpublished data, 1995). Cancer-asso-
are used, because viral load is typically much higher in ciated HPV types are associated with a higher risk of
case patients than in infected but cytologically normal development of cytologically evident lesions than are
control subjects. other HPV types (Stellato et al., 1994; Schiffman MH,
Manos MM, Sherman ME, Rush BB, Lawler P, Scott DR,
Evidence that Human Papillomavirus Infection et al. Unpublished data, 1995). Additionally, follow-up
Causes Most Cervical Neoplasia studies of women with low grade SIL have found that
cancer-associated HPV types predict an elevated risk of
The epidemiologic association between HPV infection progression to high grade SIL.29,30 Finally, several small
and cervical cancer fulfills all of the established epide- follow-up studes of women with invasive cancer have
miologic criteria for causality.22These criteria include suggested that the presence or type of HPV might pre-
strength and consistency of the epidemiologic associa- dict prognosis.31(Franc0 et al., 1994).
tion, time sequence, specificity of the association, and HPV infection causes a specific set of carcinomas of
coherence with existing biologic and epidemiologic evi- mucocutaneous epithelia, particularly anogenital tu-
dence. mors (cervical carcinoma and some types of vulvar, va-
The association between HPV infection and cervi- gmal, penile, and anal carcinomas). To a lesser extent,
cal cancer is remarkably strong and consistent, with vir- some subsets of head and neck carcinomas may also be
tually no negative ~tudies.’~ Similar magnitudes of risk associated with HPV. Numerous anecdotal reports of
have been observed in those studies of SIL that have associations with a wide variety of tumor types, such
included expert cytopathologic review to minimize mis- as adenocarcinoma of the colon and Kaposi’s sarcoma,
classification of low grade cases.24Thus, the great ma- have not been confirmed.
jority of women with cervical cancer and/or SIL have The animal data and experimental evidence for
detectable HPV DNA, compared with a consistently HPV carcinogenicity are strong, satisfying the causal
lower percentage of control women. criterion of “ ~ o h e r e n c e . ”In~ fact,
~ the potential for ma-
In case series worldwide, most cervical cancers lignant transformation of papillomavirus-induced le-
have been found to contain HPV of the same 10-15 sions has long been recognized. Cotton-tailed rabbit
types.2Metastases contain the types found in the corre- papillomavirus causes skin cancers in conjunction with
sponding primary tumors.25The most definitive study exposure to coal tar,33 and bovine papillomavirus
of invasive cervical cancer included 1050 cervical can- causes alimentary tract cancers in cows ingesting the co-
cers from over 20 countries, tested for all known HPV carcinogen bracken fern.34In the rare genetic disorder
types by PCR.’ More than 85% of cervical cancers from epidermodysplasia venuciformis, patients develop
each country contained HPV DNA, with the inclusion multiple HPV-induced cutaneous warts that are prone
of possible infections raising the proportions even to squamous cell carcinomas, especially on sun-exposed
higher. areas of the skin.35
Accordingly, the cancer-associated group of genital Cellular and molecular biologic evidence for the
HPV types is defined as those found with appreciable oncogenic potential of human papillomaviruses is espe-
prevalence alone in invasive cervical cancers.2*21,26 cially ~ o m p e l l i n g Certain
.~~ types of HPV have been
Based on this definition, the current list of cancer-asso- shown to transform human cell lines in culture37and
ciated HPV types includes at least types 16, 18, 26, 31, to cause growth abnormalities that simulate low grade
33, 35, 39, 45, 51, 52, 54, 55, 56, 58, 59, 64, and 68. SIL.38,39The types with the strongest known transform-
Other, more restrictive definitions of high risk types, ing abilities in vitro (16 and 18) are also the most impor-
based on relative risk or attributable proportion calcu- tant cancer-associated types defined epidemiologically.
lations, may be more useful for some purposes. By most In addition, the cancer-associated types are observed to
definitions, HPV 16 is the most important type associ- be genetically related when phylogenetic trees are con-
ated with cancer in almost all regions, along with HPVs structed that categorize HPV types by DNA sequence
18,31, and 45.’ h~mology.~’ In mechanistic studies, HPV DNA, though
Epidemiology of Cervical Carcinogenesis/Schiffman and Brinton 1891

Normal Cervix because from the point of view of the HPV life cycle,

I
HPV Infection
the atypical cells recognized microscopically as low
grade SIL are the production and assembly sites of new
virions.
Human papillomavirus infections are usually

No Disease + CIN 1
\ “Atypia”?
transmitted by person-to-person contact. It is clear that
cervical HPV infection is usually sexually transmit-
ted.46-48HPV infection of the cervix is rare among vir-
g i n ~The. ~ prevalence
~ of cervical HPV DNA increases
with reported numbers of different sexual partners, par-
ticularly recent partners (because infection is often tran-
CIN 3 sient).
1 Proper transmission studies have not been done,
but it appears that genital HPV infection!; are transmit-
Invasive- Cancer ted rather easily between sexual partners. For example,
Figure 1.A hypothetical schema describing the multistep the scant HPV DNA acquisition data suggest that
pathogenesis of cervical cancer.
among HPV-negative women, having new male sexual
partners is associated with a high prevalence of cervical
found in an episomal (nonintegrated) form in early cer- HPV (DNA) within months. Also, the age curve of cer-
vical lesions, is often integrated into the cellular genome vical HPV prevalence, with a peak at age 16-25, sug-
in cervical cancers and derived cancer cell lines; integra- gests that the transmission of HPV infection to the cer-
tion may therefore play a role in progression and main- vix often occurs soon after the initiation of sexual inter-
tenance of ne~plasia.’~ Protein products of HPV early course.45
genes (E6, E7) have been identified that interact with Although sexual transmission is the most important
growth-regulatory proteins of the human cell (p53, route, fomite transmission of HPV to the cervix appears
pRb), providing a possible mechanism for an HPV on- theoretically possible based on findings of HPV DNA
cogenic e f f e ~ t .Finally,
~ ~ , ~ as
~ shown below, HPV infec- on underclothes and gynecologic Verti-
tion explains much of the established epidemiology of cal transmission of genital types of HPV is certainly pos-
cervical cancer, meeting the criterion of “coherence sible, although the frequency is ~ n k n o w n . ~ ’ , ~ ~
with existing epidemiologic knowledge,” and is gener- The cumulative lifetime probability of acquiring a
ally accepted to be the major cause of most types of cer- cervical infection with at least one type of HPV is ex-
vical cancer in the tremely high for sexually active individuals.45,54 Because
the typical detectable duration of HPV infection is short
Multistep Pathogenesisof Cervical Cancer (less than 2-3 years), estimates of HPV incidence are
similar to prevalence, but both grossly underestimate
Transmission of Cervical Human Papillomavirus the cumulative incidence.
Infection The HPV prevalence of a given population depends
most strongly on the age and sexual practices of the
A hypothetical schema describing the multistep patho- population. Young, sexually active women have the
genesis of cervical cancer is presented as Figure 1. The highest HPV Although cervical
earliest known step in cervical carcinogenesis is the HPV prevalence is highly influenced by age and sexual
transmission of HPV infection. Cervical HPV infections behavior (as well as by diagnostic definition), estimates
can be studied either on the molecular (DNA detection) generally range from 1-3% based on screening diagno-
or microscopic level (low grade SIL). On a given day, ses of low grade SIL, 5-10% using nonamplified DNA
cytologic diagnoses of low grade SIL represent only tests, and 15-30% using PCR-based surveys. Published
about 10-30% of molecularly detectable HPV infec- HPV DNA prevalences can range from 1% to nearly
tions depending on the DNA test method.45However, loo%, however, depending on the analytic sensitivity
a majority of the women with molecular evidence of of the assay and the risk profile of the study
HPV infection, as seen with a nonamplified test Most HPV prevalence studies, apart from case se-
method, develop incident low grade SIL within 4 years ries of SIL and cancer, have not distinguished between
of viral detection (Schiffman MH, Manos MM, Sher- the different genital HPV types. The many types of cer-
man ME, Rush BB, Lawler P, Scott DR, et al. Unpub- vical HPV can only be distinguished at the molecular
lished data, 1995). Thus, there is great overlap between level. Based on scant data, human papillomavirus 16 is
microscopic and molecular diagnoses. This is logical, probably the most common type among healthy
1892 CANCER Supplement November 15,2995, Volume 76.No. 10

women (2.4% of cytologically normal women in one cancer compared with women with normal cytologic di-
large series).59Most of the still uncharacterized types of agnoses.66
HPV are found among healthy women and probably Nonetheless, the overall incidence of high grade
have virtually no oncogenic potential. The proportion SIL is much less than 1% in most cervical cytologic
of multiple infections typically approaches 20-30% of screening series, at least in the United States. Cervical
all infected women, as determined by PCR.60 intraepithelial neoplasia 2 and 3 are about equally diag-
As mentioned, the prevalence of cervical HPV in- nosed. The low prevalence of high grade compared
fection declines sharply with age, from a peak preva- with low grade SIL is not as pronounced in regions with
lence at 16-25 years of age.45This age trend is seen for deficient cervical cancer screening and treatment, in
both HPV DNA detection and low grade SIL in parallel. which high grade lesions can develop and accumulate.
The very high prevalence of HPV in young, sexually The three kinds of risk factors postulated to influ-
active women is consistent with an epidemic curve, a ence the risk of progression to high grade SIL are the
rapid rise in prevalence after first (sexual) exposure. The same as those established for cervical cancer: viral fac-
subsequent profound drop in cervical HPV prevalence tors, host factors, and environmental cofactors.
in women older than age 30 might be due to immuno- Viral factors. The most obvious viral factor is HPV
logic clearance or suppression of existing infections type. Among women with low grade SIL, the cancer-
combined with less exposure to new HPV types because associated types of HPV (about two thirds of infections)
of fewer new sexual partners. The decrease in HPV predict a higher risk of progression than either the non-
prevalence with age might also be due partly to a cohort cancer-associated types or HPV negativity.29830 Apart
effect, with an increase over time in the amount of cer- from viral type, it appears from cross-sectional analyses
vical HPV infection among young female popula- that high levels of HPV DNA are closely linked to high
tionsa61Both the immunologic and cohort explanations grade SIL.67T68 Time since first infection may also be im-
for the decrease in cervical infection rates with age are p ~ r t a n tbecause
,~~ the degree of nuclear atypia may in-
supported scientifically.12 crease with the duration of i n f e ~ t i o n . ~ ~
Host factors. The most important host factors re-
Progression to High Grade Squamous Intraepithelial lated to progression from low to high grade SIL are
Lesions probably immunologic. Another host factor could be
parity, which might act by influencing immunity, or
Most HPV infections disappear within months to a few hormonal, nutritional, or traumatic r n e ~ h a n i s m sAge.~~
years of diagnosis (Fig. 1). This is especially true for in- has not been shown to be a strong predictor of risk of
fections detected by HPV DNA tests only.62Although progression to high grade SIL once the severity of the
low grade lesions also tend to regress to cytologc nor- initial diagnosis is taken into account.64
malcy, women with low grade SIL progress to high Environmental cofactors. The most likely important
grade SIL (Fig. 1)with an absolute risk of about 15-25% environmental cofactors for the development of high
over 2-4 years. grade SIL are the established risk factors for cervical
The prospective data generating the estimates of cancer that do not appear to be mere proxies for HPV
progression rates from low to high grade SIL are some- infection. An association with smoking has some epide-
what conflicting because of different diagnostic termi- miologic support,24but smoking has not been found to
nology and study methods (e.g., cytologic vs. histologic be associated with the development of high grade SIL in
definitions of low grade disease at enrollment). For ex- a few recent, well-designed ~ t u d i e s . *Other
~ , ~ ~possible
ample, in a 1-3 year study. Richart and Baron found a cofactors for the development of high grade SIL include
progression rate of 20.3% from what they termed oral contraceptive use;24folate, carotenoid, retinoid, or
"mild' to severe d y ~ p l a s i aUsing
. ~ ~ different pathologic vitamin C defi~iency;~' and concurrent infection with
criteria more akin to current terminology, Nasiell and other sexually transmitted agents, such as chla-
colleagues observed that only 16% of 555 women with rn~dia.'~,~~
CIN 1 progressed to CIN 3 or invasive cancer (n = 2), It is unclear whether all cases of cervical cancer pass
despite a longer median observation period of 4 years.64 through each stage of the preinvasive continuum (Fig.
A recent British study documented a 35% rate of pro- 1).For example, some cases of high grade SIL arise in
gression to CIN 3 over 1-2 years among 538 women HPV-infected women within 1-2 years without an ap-
with mild dyskaryosis, a slightly more severe starting preciable intervening diagnosis of low grade SILZ7or
point than low grade SIL.65Whichever estimates of ab- adjacent to rather than arising directly from low grade
solute risk are accepted, women with low grade SIL are lesion^.'^ However, along with cohort studies, some
at a substantially (more than 16-fold) increased relative ecologic support for a continuum of disease is provided
risk of developing high grade SIL and invasive cervical by the observation that HPV infection and low grade
Epidemiology of Cervical Carcinogenesis/Schiffman and Brinton 1893

SIL are usually diagnosed among women in their late concurrent genital infections, could also be involved at
teens and early 20s, high grade SIL in 25-35-year-olds, later stages in carcinogenesis.
and invasive cancer after the age of 35-40.
Marital and Sexual Factors
Progression to Invasive Cancer
The recognition by epidemiologists that risk of cervical
The invasive potential of high grade SIL (particularly cancer is strongly influenced by sexual behavior led to
carcinoma in situ) is very high, and women with high discovery of the role of HPV infection. Early epidemio-
grade SIL are less to have disease regression than those logic studies revealed that the risk of cervical cancer is
with low grade SIL (Fig. 1). Peterson noted a 33% pro- especially high among women marrying at young
gression rate after 9 years among 127 women with un- ages.77-79 Subsequent investigations demonstrated the
treated carcinoma in sit^.^^ Longer follow-up would importance of sexual activity," R 4 with women having
presumably have led to even higher progression rates, sexual relationships at early ages being at higher risk
given that regression of carcinoma in situ is not typi- than either virgins or women whose sexual experiences
cal.74 began later in life. Thus, it has been shown in a variety
No risk factors have been found in case-control of case-control studies that women who initiate sexual
studies to distinguish invasive cancer from high grade activity before age 16 have about a twofold or greater
SIL, with the exception of age. Women with invasive risk compared with women becoming sexually active
cancer are 10 or more years older on average than after the age of 20 years. The risk of cervical cancer is
women with high grade SIL. It may be that invasion is also strongly influenced by the lifetime number of sex-
related to molecular events (e.g., integration of HPV ual partners, with relative risk of about three or more
DNA into the host genome) that occur with low, nearly for women reporting more than five partners compared
random frequency in the setting of persistent high grade with those reporting only one.
SIL. When HPV infection is taken into account, the
effect of lifetime number of partners is greatly weak-
ened but often remains apparent, especially in women
Reconsideration of the Established Risk Factors without apparent HPV This residual
for Cervical Cancer effect could reflect false-negative HPV classification of
some cases or might indicate an independent role of
The recognition of the key etiologic role of HPV infec- other sexually transmitted agents. Age at first inter-
tion has profoundly altered the epidemiologic study of course (or the correlated variable, age at first birth) also
cervical cancer. Yet this shift in theoretical paradigms to remains a weak risk factor, even after HPV positivity is
include HPV infection is still new and incomplete. It is considered.85r86 Age at first intercourse might be viewed
not clear which previously established risk factors for logically as a proxy for time of HPV infection, that is,
cervical cancer are mere correlates of HPV infection, the start of latency. However, it might also suggest a
which are HPV cofactors operating only in the presence vulnerable period of the cervix, when the transforming
of infection, and which are independent risk factors. For effect of HPV is greatest.
each of the hypothetical steps in cervical carcinogene- Most investigations have failed to note any effect of
sis, an effort is underway to reconsider each of the es- frequency of intercourse on risk after accounting for the
tablished risk factors in light of the central role for HPV. effects of number of sexual partners.75,77,79r81~83~84~87
This
is concordant with the assumption that HPV is rela-
Sociodemographic Factors tively easily transmitted during vaginal intercourse. In
one of the few age-specific studies of frequency of in-
Descriptive and analytic studies have demonstrated tercourse and cervical ~ a n c e r , 'high
~ frequency was a
that cervical cancer predominantly affects women in significant risk factor only before age 20, thereby sup-
lower social c l a ~ s e s . The
~ ~ -prevalence
~~ of cervical HPV porting the notion of a vulnerable period.
DNA is also elevated in women of lower educational
and income levels,48leading to an increased risk of inci- Gynecologic and Obstetric Events
dent SIL (Schiffman MH, Manos MM, Sherman ME,
Rush BB, Lawler P, Scott DR, et al. Unpublished data, There is little evidence that the risk of cervical cancer is
1995). Although patterns of HPV infection and low affected by age at menarche, age at menopause, charac-
grade SIL might partially explaining socioeconomic pat- teristics of menses,75-77,79,83,88
or personal hygiene fac-
terns of cancer, other correlates of low socioeconomic tors.7537
level, including deficient nutrition, multiparity, and Early reports suggested that poorly managed partu-
1894 CANCER Supplement November 15,1995, Volume 76, No. 10

rition may increase risk," but subsequent studies dis- tion of circumcision status and penile HPV infections
missed a true effect because of the presumed correlation are needed to address this issue further.
of pregnancy with sexual activity. However, several re-
cent studies controlling for the separate effects of repro- Infectious Agents Other than HPV
ductive and sexual factors, including HPV infection, re-
vealed a persistent influence of multiparity on risk of Despite the evidence linking HPV to cervical cancer, it
SIL and cervical cancer.24~75,90-92 would be premature to conclude that HPV is the only
It has been suggested that pregnancy could influ- agent involved. Of the other agents examined, most at-
ence cell growth either directly or indirectly through im- tention has been focused on herpes simplex virus (HSV-
munologic or hormone-dependent influences on 2) and chlamydia.
HPV.93 Although HPV detection rates may increase Multiple serologic studies have observed higher
slightly during current pregnancies,48r94 the prevalence prevalence of antibody to HSV-2 among patients with
of HPV infection is not increased in multiparous cervical neoplasia than among control subjects.104f105
women. Thus, whether there is any interaction between This association has been documented in many geo-
HPV and multiparity is unclear. Multiparity could be graphic areas with use of various assay methods. How-
an independent risk factor. For example, the effect of ever, the association of HSV-2 with risk of cervical can-
pregnancy could reflect cervical trauma during parturi- cer has been weak and inconsistent when HPV infec-
tion, a theory supported by findings from two studies of tion has been taken into account.71
reduced cervical cancer risk associated with a cesarean Chlamydia1 cervicitis has been suspected to be a
Finally, nutritional effects of reproduction risk factor for cervical cancer on the basis of case-con-
deserve attention. trol comparisons of serologic tests'06 and of chlamydia-
associated changes seen on stored cervical smears.'07
Characteristics of the Male Sexual Partner Again, however, the risk of cervical cancer associated
with chlamydia seropositivity has been inconsistent af-
The role of the male in the cause of cervical cancer has ter adjusting for HPV infe~tion.~'
been examined by comparing the sexual and other be- Additional infections studied include syphilis, gon-
havioral characteristics of husbands of patients with orrhea, cytomegalovirus, Epstein-Barr virus, and bacte-
cervical cancer with husbands of control s u b j e ~ t s . ~ ~ ,rial
~ ~vaginosis.
-~~ No consistent association with cervical
In all of these studies, the husbands of case patients cancer risk has been observed for any of these agents.
were found to report significantly more sexual partners One investigation'08 but not another7' noted a rise in
than were husbands of control subjects. In several of risk of cervical cancer with multiple, concurrent infec-
the studies, husbands of patients with cervical cancer tions, thus addressing the hypothesis that chronic cer-
were also more likely to report histories of various gen- vicovaginal inflammation may increase the oncogenic-
ital conditions, including venereal warts (caused by ity of HPV infection.
HPV types 6 and ll), gonorrhea, and herpes. Consis-
tent with these associations was low risk of cervical can- Smoking
cer of husbands reporting frequent usage of condoms.97
Reliable prevalence estimates for genital HPV in- A correlation between the distribution of cervical cancer
fections in males are more difficult to obtain than for and other smoking-related cancers prompted Win-
females, because subtle HPV-containing penile lesions kelstein to suggest that cigarette smoking may affect the
are difficult to detect and test for HPV DNA.9,'00,'0'It risk of cervical cancer.'" A number of case-control
appears from the available data that genital HPV infec- studies and one cohort investigation"' subsequently
tions are about equally common in both sexes. No con- demonstrated excess risks of cervical cancer (and SIL)
vincing HPV testing data comparing the husbands of among smokers. A number of the investigations that
patients with cervical cancer to husbands of control were able to control for age at first intercourse, number
subjects have been reported. of sexual partners, and/or social class found the associ-
Apart from HPV infection, poor hygiene of the ations with smoking to per~ist.""'~ It is noteworthy
male partner has also been postulated to play a role in that the smoking effect is restricted to squamous cell
the cause of cervical cancer, with special attention given carcinoma, with no relationship observed for the rarer
to the effects of circumcision. Despite several reports of occurrences of adenocarcinoma or adenosquamous car-
a protective effect associated with circumcision of the "'
cinoma.
partner,B8,97.102.103 many other studies have shown no Smoking is strongly associated with risk of cervical
substantial differences between case and control hus- HPV infection because of the correlation of smoking
bands. 77.79.83.95 Studies with good clinical documenta- with sexual beha~ior.~' Thus, HPV infection status can
Epidemiology of Cervical Carcinogenesis/Schiffinan and Brinton 1895

confound studies of smoking and cervical cancer. We have a low risk of cervical cancer.76r79'81,103,135-137
The
were surprised to find that in the most definitive case- apparent protective effect, however, has usually been
control studies of smoking and cervical cancer taking small, and information is limited on other risk factors. It
HPV infection into account, no independent role of is plausible that the diaphragm (like the condom) may
smoking has been dem~nstrated.'~,~' However, these protect the cervix from venerally transmitted agents like
studies were conducted in regions where even crude HPV. It has also been suggested that part of the protec-
smoking effects on cervical cancer are often weak. Per- tion associated with diaphragm use may reflect concur-
haps the full influence of smoking on risk of cervical rent use of spermicides, which have antiviral proper-
cancer is observable only in regions where prolonged, ties.'38 However, the most common spermicide has no
heavy smoking among women is prevalent. appreciable anti-HPV activity in ~ i t r 0 . I ~ ~

Oral Contraceptives Dietary Factors


Studies examining the relationship of oral contraceptive
The influence of nutrient status on risk of cervical neo-
use to cervical cancer risk are especially complex, with
plasia has received substantial research attenti~n.~'
questions arising about the potential for confounding,
Most studies have used case-control approaches, as-
particularly by sexual and screening b e h a ~ i o r . ' ~ ~ - ' ' ~
sessing dietary intake or blood levels at the time of di-
Although several studies have noted no relation-
agnosis, but some prospective studies have been com-
ship between oral contraceptive use and cervical cancer
pleted. No published observational study has properly
risk, the majority of studies indicate that long term users
considered HPV infection. Recent trials have focused
are at excess risk, even after adjustment for sexual and
on a possible protective role of micronutrients (e.g., ca-
social factors. Studies showing no relationship of risk to
rotenoids, vitamin C , or folate) in promoting the regres-
oral contraceptive use are generally those that have
sion of low grade SIL.14' Although no firm associations
used neighborhood control which may
between nutritional status and the natural history of SIL
reflect overmatching, or those that have included non-
have been found, the accumulated evidence suggests
invasive abnormalities, thus presenting difficulties for
that some component of fruits and vegetables may be
interpretation because of possible detection biases.125,126
protective.
In addition, the absence of an effect in several studies
may reflect merely the limited number of long term oral
contraceptiveusers. In a recent large study by the World Genetics
Health Organization, a risk of 2.2 was associated with
Little attention has been given to familial occurrences of
use of 8 or more years.127
cervical cancer, although some reports suggest that a
The effects of oral contraceptive use may be some-
familial tendency Whether this tendency
what stronger for adenocarcinoma than for squamous
reflects environmental or genetic factors is unknown.
cell neoplasms,'20,'28consistent with descriptive sur-
Several investigators have observed associations of
veys showing increasing rates of cervical adenocarci-
human leukocyte antigen (HLA) alleles or haplotypes
noma a m o n g y o u n g ~ o m e n . ~ ~ ~ - ' ~ '
with invasive cervical cancer. Most current interest is
Recent interest has been on possible interactive
focused on genotypes of the HLA-D 10ci.l~~ However,
effects of oral contraceptives and HPV, especially in
none of the reported associations has been observed
view of studies showing that the transcriptional regula-
consistently, thus raising concern of a multiple compar-
tory regions of HPV DNA contain hormone-recognition
ison problem.
elements and that transformation of cells in vitro with
viral DNA is enhanced by h o r m o n e ~ . ~Oral ~ ~ con-
~~', ~~~
traceptive use has been only inconsistently associated Immunosuppression
with increased rates of HPV i n f e ~ t i o n . ' HOW- ~~~~ ~~~~'~~
Much of what is known about HPV immunology de-
ever, two recent studies found an especially elevated
rives from small investigations of HPV infections in im-
risk of invasive cervical cancer among HPV-positive
munodeficient indi~idua1s.l~~ Cervical SIL rates are ele-
women who used oral contraceptive^.^^^^^ Thus, oral
vated among immunosuppressed women with renal
contraceptives may promote the activity of HPV once
transplant^,'^^-'^' who are prone to a variety of genital
infection has occurred. infections, including HPV and HSV-2.147,'49 However,
Other Contraceptive Methods the excess risks observed among patients with renal
transplants are complicated by close medical surveil-
In a number of studies, users of barrier methods of con- lance and difficulties in obtaining reliable expected val-
traception (diaphragm and condom) were found to ue~.'~~
1896 CANCER Supplement November 25,2995, Volume 76, No. 10

Human immunodeficiency virus (HIV) infection tory and cervical carcinogenicityof HPV. Through large
provides perhaps the most important example of the investigations that focus on increasing grades of neo-
effect of immunosuppression on HPV infection. Indi- plasia and time-related exposures, the multistage pro-
viduals infected with HIV through sexual contact are cesses involved in tumor development and progression
likely also to be exposed to genital HPV. However, the can be examined and factors that promote or inhibit
increased diagnosis of HPV in HIV-infected individuals transition to higher grades clarified.
is so striking that the increase is apparently real and re- The search for important HPV cofactors for high
lated to immunosuppression. Specifically, HIV infec- grade SIL and cancer is likely to dominate future epide-
tion is associated with a very high prevalence of HPV miologic research on cervical cancer. Although HPV in-
DNA detection, especially in immunosuppressed fection explains much of what is known about classic
women with low CD4 c o ~ n t s . ' ~In ~ -such
' ~ ~women, SIL risk factors for cervical neoplasia, particularly its vene-
is diagnosed commonly. It is not known whether HIV real transmission, we still know little about potential co-
immunosuppression permits reactivation of previously factors. Therefore, even if HPV infection is the unify-
suppressed HPV infection as opposed to allowing for ing, central risk factor for cervical neoplasia throughout
rapid infection or reinfection. the world, it is worth considering that necessary cofac-
Although the association between HIV infection, tors could vary considerably across geographic regions.
HPV infection, and SIL (including high grade SIL) is es- Smoking, for example, may be a cofactor in the United
tablished, a causal role for immunosuppression in the States, but as studies in Latin America have already
risk of progression of SIL to invasive carcinoma is less demonstrated, it is unlikely to explain the occurrence of
clear. Anal carcinoma rates are increasing among the cervical cancer where heavy smoking among women is
homosexual male population, probably related to con- rare.
current HIV and HPV i n f e ~ t i 0 n .In
I ~contrast,
~ cervical The importance of HPV persistence in the patho-
cancer rates are not greatly elevated in HIV-infected fe- genesis of high grade SIL and cancer must be verified
male cohorts, possibly reflecting more limited follow- and the determinants of persistence identified. If carci-
up time (Cote TR, Kessler LG, Gail M, Schiffman MH, nomas arise only from persistent infections and not
Biggar RS, Virgo PW, et al. Unpublished data, 1995). It from molecularly inapparent (latent) infections that
is unclear whether HIV immunosuppression could suddenly reactivate, then prevention of carcinomas
speed up the normally long progression from SIL to in- should be achievable by screening for virus at ages pre-
vasive carcinoma or whether it mainly increases the ceding the usual onset of carcinoma. The study of HPV
prevalence and persistence of SIL precursors. Of note, persistence requires repeated measurements and ex-
a mildly immunosuppressive retrovirus, HTLV- 1, has tremely reliable HPV testing. It will be necessary to dis-
recently been linked to an increased risk of high grade tinguish variants of HPV types16' to permit, for exam-
SIL and cancer among HPV-infected women.154 ple, the distinction of new HPV 16 infections from re-
On the basis of animal experiments and the immu- current ones.
nosuppression data, it is assumed that the key immune Through continued descriptive analyses and cohort
response involved in the clearance of HPV infections is studies, the decrease in cervical HPV infection rates
cell mediated.'55,'56The two classes of cells thought to with increasing age should be better understood. The
be involved in the cellular immune response to HPV are separate contributions to the age trend of cohort effects
antigen-presenting cells (Langerhans cells) and cyto- and immunologic suppression must be distinguished,
toxic T lymphocyte^.'^^,'^^ because any cohort effect of increasing HPV infection
in younger women may predict future increases in in-
Risk Factors by Cell Type vasive cervical cancer.
Although the comparison of risk factors by cell type has Although the study of cervical neoplasia is inextri-
received little attention, there is some evidence that cer- cably linked to HPV infection, it will be important to
vical adenocarcinoma may resemble endometrial ade- define separately the epidemiologic characteristics of
nocarcinoma, particularly with respect to relationships the minority (15% or less) of cervical cancers that do not
with nulliparity and ~ b e s i t y . ~ , 'Because
~ ~ , ' ~ ~adenocar- contain HPV-related DNA. Possibly, unknown HPV
cinomas may be more likely to contain HPV 18 than are types account for some of these tumors, but recent mo-
squamous carcinomas, HPV type may be a determinant lecular studies have suggested that HPV-negative can-
of histologic type.'60 cers may be an etiologically distinct group, perhaps as-
sociated with somatic mutations in tumor suppressor
Future Research genes.'62 Moreover, HPV-negative cervical cancer
Multidisciplinary teams of investigators are attempting might have a worse p r o g n o s i ~ . 'If~ ~
cervical cancer can
to piece together a coherent picture of the natural his- develop, albeit rarely, from precursor lesions not asso-
Epidemiology of Cervical Carcinogenesis/Schifiman and Brinton 1897

ciated with HPV infection, then the morphologic ap- 8. Meisels A, Fortin R. Condylomatous lesions of the cervix and
pearance and natural history of those precursor lesions vagina. I. Cytologic patterns. Acfa Cyfol 1976;20:505-9.
9. Barrasso R, De Brux J, Croissant 0, et al. High prevalence of
must be defined. Also needed are studies to define hor- papillomavirus-associated penile intraepithelial neoplasia in
monal and other risk factors for the rarely occurring ad- sexual partners of women with cervical intraepithelial neoplasia.
enocarcinomas and adenosquamous carcinomas of the N EnglJMed 1987;317:916-23.
cervix, whose causes are poorly understood. 10. Kurman RJ, Toki T, Schiffman MH. Basaloid and warty carcino-
It will be important to verify or exclude the role of mas of the vulva: distinctive types of squamous cell caranoma
frequently associated with human papillomavirus. A m J Surg Pa-
HPV in carcinomas of other sites. The natural history of tho/ 1993; 17:133-45.
HPV infection in the cervix should be compared with 11. Shah KV, Howley P. Papillomaviruses. In: Fields BN, Knipe DM,
its natural history in the vagina, vulva, and anus. In par- editors. Virology. New York: Raven Press. In press.
ticular, why the transformation zone of the cervix is so 12. Schiffman MH, Burk RD. Human papillomaviruses and cervical
prone to HPV carcinogenesis should be addressed. neoplasia. In: Evans A, Kaslow R, editors. Viral Infections of hu-
As the highest priority, HPV immunologic charac- mans, epidemiology and control. 4th ed. New York Plenum. In
press.
teristics are likely to occupy epidemiologists studying 13. Galloway D. Serological assays for the detection of HPV anti-
cervical cancer etiology and prevention over the next bodies. In: Munoz N, Bosch FX, Shah KV, et al., editors. The
decade. In the immediate future, the interactions of epidemiology of human papillomavirus and cervical cancer. vol
multiple HPV types in mixed cervical infections should 119. Lyon: IARC ScientificPublications, 1992147-61.
be clarified as one pathway to understanding HPV im- 14.Kirnbauer R, Hubbert NL, Wheeler CM, et al. A virus-like parti-
cle ELISA detects serum antibodies in a majority of women in-
munity. Assays of cell-mediated immunity must be de- fected with human papillomavirus type 16. J Nafl Cancer lnst
veloped and applied. The ultimate goal will be to define 1994;86:494-9.
the successful immune response to HPV infection in the 15. Vermund SH, Schiffman MH, Goldberg GL, et al. Molecular di-
hope that cancer-preventive immunity can be stimu- agnosis of genital human papillomavirus infection: comparison
lated by va~cination.’~~ of two methods used to collect exfoliated cervical cells. Am 1Ob-
stet Gynecol 1989; 160:304-8.
Because HPV is a central cause of most cervical neo-
16. de Villiers EM, Schneider A, Miklaw H, et al. Human papillo-
plasias, it is reasonable to consider HPV immunization mavirus infections in women with and without abnormal cervi-
as the ultimate primary preventive strategy for elimina- cal cytology. Lancet 1987;2:703-6.
tion of most cervical cancers. The protection of cattle 17. Moscicki AB, Palefsky J, Smith G , et al. Variability of human
herds from bovine papillomavirus infection by vaccina- papillomavirus DNA testing in a longitudinal cohort of young
tion serves as a successful animal Use of the women. Obstet Gynecol 1993;82:578-85.
18. Schiffman MH. Validation of HPV hybridization assays: corre-
hepatitis B vaccine in Asia to reduce the incidence of lation of FISH, dot blot, and PCR with Southern blot. In: Munoz
hepatocellular carcinoma may serve as a public health N, Bosch FX, Shah KV, et al., editors. The epidemiology of hu-
model. man papillomavirus and cervical cancer. Lyon: IARC Mono-
graph, 1992169-79.
19. Franco EL. The sexually transmitted disease model for cervical
References cancer: incoherent epidemiologic findings and the role of m i s -
classification of human papillomavirus infection. Epidemiology
1. Parkin DM, Pisani P, Ferlay J. Estimates of the worldwide fre- 1991;2:98-106.
quency of eighteen major cancers in 1985. Int J Cancer 1993;54: 20. Gravitt PE, Manos MM. Polymerase chain reaction-based meth-
594-606. ods for the detection of human papillomavirus DNA. In: Munoz
2. Bosch FX, Manos MM, Munoz N, et al. Prevalence of human N, Bosch FX, Shah KV, et al., editors. The epidemiology of hu-
papillomavirus in cervical cancer: a worldwide perspective. J man papillomavirus and cervical cancer. vol 119. Lyon: IARC
Natl Cancer Inst. In press. Scientific Publications, 1992:121-33.
3. Shingleton HM, Bell MC, Fremgen A, Chmiel JS, Russell AH, 21. Lorincz AT, Reid R, Jenson B, et al. Human papillomavirus in-
Jones WB, et al. Is there really a differencein survival of women fection of the cervix: relative risk associations of 15 common
with squamous cell carcinoma, adenocarcinoma, and adeno- anogenital types. Obstet Gynecnl 1992; 79:328-37.
squamous cell carcinoma of the cervix? Cancer 1995;76:1948- 22. Hill AB. Environment and disease: association or causation? Proc
55. R SocMed 1965;58:295-300,
4. Kjaer SK, Brinton LA. Adenocarcinomas of the uterine cervix: 23. Munoz N. HPV and cervical cancer: review of case-control and
the epidemiology of an increasing problem. Epidemiol Rev cohort studies. In: Munoz N, Bosch FX, Shah KV, et al., editors.
1993; 15~486-98. The epidemiology of human papillomavirus and cervical cancer.
5. Jones HW. Impact of the Bethesda System. Cancer 1995;76: vol 119. Lyon: IARC Scientific Publications, 1992251-61.
1914-8. 24. Schiffman MH, Bauer HM, Hoover RN, et al. Epidemiologic ev-
6. Kurman RJ, Solomon D. The Bethesda System for reporting cer- idence showing that human papillomavirus infection causes
vical/vaginal cytologic diagnoses. New York: Springer-Verlag, most cervical intraepithelial neoplasia. J Natl Cancer lnst
1994. 1993;85:958-64.
7. National Cancer Institute Workshop. The 1988 Bethesda system 25. Lancaster WD, Castellano C , Santos C, et al. Human papillo-
for reporting cervical/vaginal cytologic diagnoses. IAMA mavirus deoxyribonucleic acid in cervical carcinoma from pri-
1989;262:931-4. mary and metastatic sites. A m J Obstet Gynecol 1986; 154:115-9.
1898 CANCER Supplement November 25,1995, Volume 76, No. 10

26. Fuchs PG, Girardi F, Pfister H. Human papillomavirus DNA in human papillomavirus infection and cervical neoplasia. J Natl
normal, metaplastic, preneoplastic and neoplastic epithelia of Cancer Inst 1992;84:394-8.
the cervix uteri. IntJCancer 1988;41:41-5. 46. Fisher M, Rosenfeld WD, Burk RD. Cervicovaginalhuman pap-
27. Koutsky LA, Holmes KK, Critchlow CW, et al. A cohort study of illomavirus infection in suburban adolescents and young adults.
the risk of cervical intraepithelial neoplasia grade 2 or 3 in rela- JPediatr 1991;119:821-5.
tion to papillomavirus infection. NEngl JMed 1992;327:1272-8. 47. Ley C, Bauer HM, Reingold A, et al. Determinants of genital hu-
28. Stellato G, Nieminen P, Aho H, et al. Human papillomavirus man papillomavirus infection in young women. J Natl Cancer
infection of the female genital tract: correlation of HPV DNA Inst 1991;83:997-1003.
with cytologic, colposcopic, and natural history findings. Eur J 48. Hildesheim A, Gravitt P, Schiffman MH, et al. Determinants of
Gynaecol Oncol 1994; 13962-7. genital human papillomavirus infection in low-income women
29. Campion MJ, McCance DJ, Cuzick J, et al. Progressive potential in Washington, D.C. Sex Transm Dis 1993;20:279-85.
of mild cervical atypia: prospective cytological,colposcopic, and 49. Fairley CK, Chen S, Tabrizi SN, et al. The absence of genital
virological study. Lancet 1986;2:237-40. human papillomavirus DNA in virginal women. Int J STD AIDS
30. Kataja V, Syrjanen K, Syrjanen S, et al. Prospective follow-up of 1992;3:414-7.
genital HPV infections: survival analysis of the HPV typing data. 50. Ferenczy A, Bergeron C, Richart RM. Human papillomavirus
EurJ Epidemiol 1990;6(1):9-14. DNA in fomites on objects used for the management of patients
31. Franco EL. Prognostic value of human papillomavirus in the sur- with genital human papillomavirus infections. Obstef Gynecol
vival of cervical cancer patients: an overview of the evidence. 1989;74:950-4.
Cancer Epidemiol Bid Preu 1992;1:499-504. 51. Ferenczy A, Bergeron C, Richart RM. Human papillomavirus
32. Park T-W, Fujiwara H, Wright TC. Molecular biology of cervical DNA in CO2 laser-generated plume of smoke and its conse-
cancer and its precursors. Cancer 1995;76:1902-13. quences to the surgeon. Obstef Gynecol 1990;75:114-8.
33. Rous P, Kidd JG. The carcinogenic effect of papilloma virus on 52. Roman A, Fife K. Human papillomavirus DNA associated with
the tarred skin of rabbits. I. Description of the phenomenon. J foreskins of normal newborns. ]Infect Dis 1986;153:855-61.
Exp Med 1938;67:399-428. 53. Shah KV, Kashima H, Polk F, et al. Rarity of cesarean delivery in
34. Jarrett WFH, McNeil PE, Grimshaw WTR, et al. High incidence cases of juvenile-onset respiratory papillomatosis. Obstet Gyne-
area of cattle cancer with a possible interaction between an en- col 1986;68:795-9.
vironmental carcinogen and a papilloma virus. Nature 1978;274: 54. Schneider A, Kirchhoff T, Meinhardt G, et al. Repeated evalua-
215-7. tion of human papillomavirus 16 status in cervical swabs of
35. Orth G, Favre M, Breitburd F, et al. Epidermodysplasia vermci- young women with a history of normal Papanicolaou smears.
formis: a model for the role of papilloma viruses in human can- Obstet Gynecol 1992;79(5):683-8.
cer. In: Essex M, Todaro G. zur Hausen H, editors. Viruses in 55. Melkert PWJ, Hopman E, van den Brule AJC, et al. Prevalence
naturally occumng cancers. Cold Spring Harbor Conferences on of HPV in cytomorphologically normal cervical smears, as de-
Cell Proliferation 7. New York Cold Spring Harbor Laboratory, termined by the polymerase chain reaction, is age-dependent.
1980:259-82. IntJCancer 1993;53:919-23.
36. zur Hausen H, editor. Human pathogenic papillomaviruses. 56. Moscicki A-B, Palefsky J, Gonzales J, et al. Human papillomavi-
Heidelberg: Springer-Verlag,1994. Current Topics in Microbiol- rus infection in sexually active adolescent females: prevalence
ogy and Immunology. vol. 186. and risk factors. Pediatr Res 1990;28:507-13.
57. Rosenfeld WD, Vermund SH, Wentz SJ, et al. High prevalence
37. Yasumoto S, Burkhardt AL, Doniger J, et al. Human papillo-
rate of human papillomavirus infection and association with ab-
mavirus type 16 DNA-induced malignant transformation of
normal Papanicolaou smears in sexually active adolescents. Am
NIH 3T3 cells. ] Virol 1986;57:572-7.
JDis Child 1989;143:1443-7.
38. Kreider JW, Howett MK, Wolfe SA, et al. Morphological trans-
58. Guerrero E, Daniel RW, Bosch X, et al. Comparison of virapap,
formation in vivo of human uterine cervix with papillomavirus
Southern hybridization, and polymerase chain reaction methods
from condylomata acuminata. Nature 1985;317:639-41.
for human papillomavirus identification in an epidemiological
39. McCance DJ, Kopan R, Fuchs E, et al. Human papillomavirus investigation of cervical cancer. J Clin Microbiol 1992;30(11):
type 16 alters human epithelial cell differentiation in vitro. Proc 2951-9.
NatIAcad Sci USA 1988;85:7169-73. 59. Schiffman MH. Epidemiology of cervical human papillomavi-
40. Van Ranst M, Kaplan JB, Burk RD. Phylogenetic classification of ruses. In: zur Hausen H, editor. Human pathogenic papillomavi-
human papillomaviruses: correlation with clinical manifesta- ruses. Heidelberg: Springer Verlag, 1994:55-81.
tions. ] Gen Virol 1992;739653-60. 60. Bauer HM, Hildesheim A, Schiffman MH, et al. Determinants of
41. Cullen AP, Reid R, Campion M, et al. Analysis of the physical genital human papillomavirus infection in low-risk women in
state of different human papillomavirus DNAs in intraepithelial Portland, Oregon. Sex Transm Dis 1993;20:274-278, 1993.
and invasive cervical neoplasms. J Virol 1991;65:605-12. 61. Evans S,Dowling K. The changing prevalence of cervical human
42. Dyson N, Howley PM, Munger K, et al. The human papilloma papilloma virus infection. Aust N Z J Obstet Gynaecol 1990;30:
virus-16 E7 oncoprotein is able to bind to the retinoblastoma 375-7.
gene product. Science 1989;243:934-7. 62. Hildesheim A, Schiffman, Gravitt P, et al. Persistence of type-
43. Werness BA, Levine AJ, Howley PM. Association of human pap- specific human papillomavirus infection among cytologically
illomavirus types 16 and 18 E6 proteins with p53. Science normal women in Portland, Oregon. J Infect Dis 1994; 169935-
1990;248:76-9. 40.
44. Munoz N, Bosch FX, de Sanjose S, et al. The causal link between 63. Richart RM, Barron BA. A follow-up study of patients with cer-
human papillomavirus and invasive cervical cancer: a popula- vical dysplasia. A m J Obstet Gynecol 1969;105:386-93.
tion-based case-control study in Columbia and Spain. Int J Can- 64. Nasiell K, Roger V, Nasiell M. Behavior of mild cervical dyspla-
cer 1992;52:743-9. sia during long-term follow-up. Obstet Gynecol 1986;67:665-9.
45. Schiffman MH. Recent progress in defining the epidemiology of 65. Flannelly G,Anderson D, Kitchener HC, et al. Management of
Epidemiology of Cervical Carcinogenesis/Schiffman and Brinton 1899

woman with mild and moderate cervical dyskaryosis. BM] mental factors in carcinoma of the cervix. Am ] Obstet Gynecol
1994;308~1399-403. 1954; 68:1016-52.
66. Soutter WP, Fletcher A. Invasive cancer of the cervix in women 89. Smith FR. Etiologic factors in carcinoma of the cervix. Am ] Obs-
with mild dyskaryosis followed up cytologically. BM] 1994; 308: tet Gynecol 1931;21:18-25.
1421-3. 90. Brinton LA, Reeves WC, Brenes MM, et al. Parity as a risk factor
67. Morrison EAB, Ho GYF, Vermund SH, et al. Human papillo- for cervical cancer. Am J Epidemiol 1989; 130:486-96.
mavirus infection and other risk factors for cervical neoplasia: a 91. Kjaer SK, Dahl C, Engholm G, et al. Case-control study of risk
case-control study. Int ] Cancer 1991;49:6-13. factors for cervical neoplasia in Denmark. 11. Role of sexual ac-
68. Cuzick J, Terry G, Ho L, et al. Human papillomavirus type 16 tivity, reproductive factors, and venereal infections. Cancer
DNA in cervical smears as a predictor of high grade cervical in- Causes Control 1992;3:339-48.
traepithelial neoplasia. Lancet 1992; 339:959-60. 92. Parazzini F, La Vecchia C, Negri E, et al. Reproductive factors
69. Munoz N, Bosch FX, de Sanjose S, et al. Risk factors for cervical and the risk of invasive and intraepithelial cervical neoplasia. Br
intraepithelial neoplasia grade III/carcinoma in situ in Spain and ] Cancer 1989;59:805-9.
Colombia. Cancer €pi Biom Prev 1993;2:423-31. 93. Pater A, Bayatpour M, I'ater MM. Oncogenic transformation by
70. Potischman N. Nutritional epidemiology of cervical neoplasia. human papillomavirus type 16 deoxyribonucleic acid in the
Nutr 1993; 123:424-9. presence of progesterone or progestins from oral contraceptives.
71. de Sanjose S, Munoz N, Bosch FX, et al. Sexually transmitted Am 1 Obstet Gynecol 1990;162:1099-103.
agents and cervical neoplasia in Colombia and Spain. Int 1Can- 94. Schneider A, Hotz M, Gissmann L. Increased prevalence of hu-
cer 1994;56:358-63. man papillomaviruses in the lower genital tract of pregnant
72. Kiviat NB, Critchlow CW, Kurman RJ. Reassessment of the mor- women. Znt ]Cancer 1987;40:198-201.
phological continuum of cervical intraepithelial lesions: Does it 95. Brinton LA, Reeves WC, Brenes MM, et al. The male factor in
reflect different stages in the progression to cervical carcinoma? the etiology of cervical cancer among sexually monogamous
In: Munoz N, Bosch FX, Shah KV, et al., editors. The epidemiol- women. Znt ]Cancer 1989;44:199-203.
ogy of human papillomavirus and cervical cancer. Lyon: IARC 96. Buckley JD, Harris RWC, Doll R, et al. Case-control study of the
Scientific Publications, 1992:59-66. husbands of women with dysplasia or carcinoma of the cervix
73. Peterson 0. Spontaneous course of cervical precancerous condi- uteri. Lancet 1981; 2: 1010-5.
tions. Am 1Obstet Gynecol 1956;72:1063-71. 97. Kjaer SK, de Villiers EM, Dahl C, et al. Case-control study of risk
74. Kinlen LJ, S p r i g s AI. Women with positive cervical smears but factors for cervical neoplasia in Denmark. I. Role of the "male
without surgical intervention: a follow-up study. Lancet 1978;2: factor" in women with one lifetime sexual partner. Int ] Cancer
463-5. 1991;48~29-44.
75. Brinton LA, Hamman RF, Huggins GR, et al. Sexual and repro- 98. Niruthisard S, Trisukosol D. Male sexual behavior as risk factor
ductive risk factors for invasive squamous cell cervical cancer. ] in cervical cancer.JMedAssoc 7'hai 1991;x:507-12.
Natl Cancer Znst 1987; 79:23-30. 99. Zunzunegui MV, King MC, Coria CF, et al. Male influences on
76. Fasal E, Simmons ME, Kampert JB. Factors associated with high cervical cancer risk. Am JEpidemiol 1986; 123:302-7.
and low risk of cervical neoplasia. ] Natl Cancer Insl 1981;66: 100. Barrasso R. HPV-related genital lesions in men. In: Munoz N,
631-6. Bosch FX, Shah KV, et al., editors. The epidemiology of human
77. Jones EG, MacDonald I, Breslow L. A study of epidemiologic papillomavirus and cervical cancer. Lyon: IARC Monograph,
factors in carcinoma of the uterine cervix. Am I Obstet Gynecol 1992~85-92.
1958;76:1-10. 101. Bergman A, Nalick R. Prevalence of human papillomavirus in-
78. West DW, Schuman KL, Lyon JL, et al. Differences in risk esti- fection in men: comparison of the partners of infected and unin-
mations from a hospital and a population-based case-control fected women. JReprod Med 1992;37:710-2.
study. Int ] Epidemiol 1984; 13:235-9. 102. Agarwal SS,Sehgal A, Sardana S, et al. Role of male behavior in
79. Boyd JT, Doll R. A study of the aetiology of carcinoma of the cervical carcinogenesis among women with one lifetime sexual
cervix uteri. B r ] Cancer 1964; 18:419-34. partner. Cancer 1993;72:1666-9.
80. Kessler 11. Venereal factors in human cervical cancer: evidence 103. Terris M, Oalmann MC. Carcinoma of the cervix: an epidemio-
from marital clusters. Cancer 1977;39:1912-9. logic study. JAMA 1960; 174:1847-51.
81. Martin CE. Marital and coital factors in cervical cancer. Am ]Pub- 104. Hildesheim A, Mann V, Brinton LA, et al. Herpes simplex virus
lic Health 1967;57:803-14. type 2: a possible interaction with human papillomavirus types
82. Pridan H, Lilienfeld AM. Carcinoma of the cervix in Jewish 16/18 in the development of cervical cancer. Znt J Cancer
women in Israel, 1960-67: an epidemiological study. Isr Med 1991; 38:335-41.
Sci 1971;7:1465-70. 105. Jha PKS, Beral V, Peto J, et al. Antibodies to human papillomavi-
83. Rotkin ID. Adolescent coitus and cervical cancer: associations of rus and to other genital infectious agents and invasive cervical
related events with increased risk. Cancer Res 1967;27:603-17. canccrrisk. Lancet 1993;341:1116-8.
84. Terris M, Wilson F, Smith H, et al. The relationship of coitus to 106. Schachter J, Hill EC, King EB, et al. Chlamydia trachomatis and
carcinoma of the cervix. A m 1Public Health 1967;57:840-7. cervical neoplasia. JAMA 1982; 248:2134-8.
85. Bosch FX, Munoz N, de Sanjose S, et al. Risk factors for cervical 107. Allerding TJ, Jordan, SW, Boardman RE. Association of human
cancer in Colombia and Spain. Znt 1Cancer 1992; 52:750-8. papillomavirus and chlamydia infections with incidence cervical
86. Eluf-Neto J, Booth M, Munoz N, et al. Human papillomavirus neoplasia. Acta Cytol 1985; 29:653-60.
and invasive cancer in Brazil. Br 1Cancer 1994;69:114-9. 108. Schmauz R, Okong P, de Villiers EM, et al. Multiple infections in
87. Herrero R, Brinton LA, Reeves WC, et al. Sexual behavior, vene- cases of cervical cancer from a high-incidence area in tropical
real diseases, hygiene practices, and invasive cervical cancer in a Africa. 1nt J Cancer 1989;43:805-9.
high-risk population. Cancer 1990; 65:380-6. 109. Winkelstein W Jr. Smoking and cancer of the uterine cervix: hy-
88. Wynder EL, Cornfield J, Schroff I'D, et al. A study of environ- pothesis. Am 1Epidemiol 1977; 106:257-9.
1900 CANCER Supplement November 25,1995, Volume 76, No. 10

110. Greenberg ER, Vessey M, McPherson K, et al. Cigarette smoking gesterone receptors in cervical human papillomavirus related le-
and cancer of the uterine cervix. BrJCancer 1985;51:139-41. sions. Int JCancer 1991;48:533-9.
111. Brinton LA, Schairer C, Haenzel W, et al. Cigarette smoking and 134. Lorincz AT, Schiffman MH, Jaffurs WJ, et al. Temporal associa-
invasive cervical cancer. JAMA 1986; 255:3265-9. tions of human papillomavirus infection with cervical cytologic
112. Clarke EA, Morgan RW, Newman AM. Smoking as a risk factor abnormalities. Am J Obstet Gynecol 1990; 162:645-51.
in cancer of the cervix: additional evidence from a case-control 35. Boyce JG, Lu T, Nelson JH, et al. Oral contraceptives and cervical
study. Am J Epidemiol 1982; 115:59-66. carcinoma. Am 1Obstet Gynecol 1977; 128:761-6.
113. Harris RWC, Brinton LA, Cowdell RH, et al. Characteristics of 36. Melamed MR, Koss LG, Flehinger BJ, et al. Prevalence rates of
women with dysplasia or carcinoma in situ of the cervix uteri. Br uterine cervical carcinoma in situ for women using the dia-
J Cancer 1980;42:359-69. phragm or contraceptive oral steroids. Br Med J 1969; 3:195-200.
114. Hellberg D, Valentin J, Nilsson S. Smoking as risk factor in cer- 37. Worth AJ, Boyes DA. A case control study into the possible
vical neoplasia. Lancet 1983;2:1497. effects of birth control pills on pre-clinical carcinoma of the cer-
115. La Vecchia C , Franceschi S, Decarli A, et al. Cigarette smoking vix. 1 Obstet Gynaecol Br Commonw 1972; 79:673-9.
and the risk of cervical neoplasia. Am 1Epidemiol 1986; 123:22- 38. Hildesheim A, Brinton LA, Mallin K, et al. Barrier and spermi-
9. cidal contraceptive methods and risk of invasive cervical cancer.
116. Lyon JL, Gardner JW, West DW, et al. Smoking and carcinoma Epidemiology 1990; 1966-72.
in situ of the uterine cervix. Am J Public Health 1983; 73:558-62. 39. Hermonat PL, Daniel RW, Shah KV. The spermicide nonoxynol-
117. Peters RK, Thomas D, Hagan DG, et al. Risk factors for invasive 9 does not inactivate papillomavirus. Sex Trnnsm Dis 1992; 19:
cervical cancer among Latinas and non-Latinas in Los Angeles 203-5.
County. J Natl Cancer Inst 1986;77:1063-77. 40. Mitchell MF, Hittelman WK, Lotan R, Nishioka K, Tortolero-
118. Slattery ML, Robison LM, Schuman KL, et al. Cigarette smoking Luna G, Richards-Kortum R, et al. Chemoprevention trials and
and exposure to passive smoke are risk factors for cervical can- surrogate end point biomarkers in the cervix. Cancer 1995; 76:
cer. IAMA 1989;261:1593-8. 1956-77.
119. Trevathan E, Layde P, Webster LA, et al. Cigarette smoking and 141. Bender S. Carcinoma in situ of cervix and sisters. Br Med J 1976;i:
dysplasia and carcinoma in situ of the uterine cervix. J A M 502.
1983; 250~499-502. 142. Brinton LA, Tashima KT, Lehman HF, et al. Epidemiology of
120. Brinton LA, Huggins GR, Lehman HF, et al. Long-term use of cervical cancer by cell type. Cancer Res 1987;47:1706-11.
oral contraceptives and risk of invasive cervical cancer. I n t J Can- 143. Furgyik S, Grubb R, Kullander S, et al. Familial occurrence of
cer 1986;38:339-44. cervical cancer, stages 0-IV. Acta Obstef Gynecol Scand 1986; 65:
121. Brinton LA, Reeves WC, Brenes MM, et al. Oral contraceptive 223-7.
use and risk of invasive cervical cancer. Int J Epidemiol 1990; 19: 44. Apple RJ, Erlich HA, Klitz W, et al. HLA DR-DQ associations
4-11. with cervical carcinoma show papillomavirus-type specificity.
122. Brinton LA. Oral contraceptives and cervical neoplasia. Contra- Nature Genet 1994;6:157-62.
ception 1991;43:581-95. 45. Evans AS, Mueller NE. Viruses and cancer: causal associations.
123. Piper JM. Oral contraceptives and cervical cancer. Gynecol Oncol Ann Epidemiol 1990; 1:71-92.
1985;22:1-14, 46. Hoover R. Effects of drugs-immunosuppression. In: Hiatt HH,
124. Celentano DD, Klassen AC, Weisman CS, et al. The role of con- Watson JD, Winsten JA, editors. Origins of human cancer. Cold
traceptive use in cervical cancer: the Maryland cervical cancer Spring Harbor Conferences on Cell Proliferation 4. New York:
case-control study. Am J Epidemiol 1987; 126:592-604. Cold Spring Harbor Laboratory, 1977:369-79.
125. Irwin KL, Rosero-Bixby L, Oberle MW, et al. Oral contraceptives 147. Matas AJ, Simmons RL, Kjellstrand CM, et al. Increased inci-
and cervical cancer risk in Costa Rica: detection bias or causal dence of malignancy during chronic renal failure. Lancet 1975; 1:
association? IAMA 1988;259:59-64. 883-5.
126. Molina R, Thomas DB, Dabancens A, et al. Oral contraceptives 148. Porreco R, Penn I, Droegemueller W, et al. Gynecologic malig-
and cervical carcinoma in situ in Chile. Cancer Res 1988;48: nancies in immunosuppressed organ homograft recipients. Ob-
1011-5. stet Gynecol 1975;45:359-64.
127. WHO Collaborative Study of Neoplasia and Steroid Contracep- 149. Sillman F, Stanek A, Sedlis A, et al. The relationship between
tives. Invasive squamous-cell cervical carcinoma and combined human papillomavirus and lower genital intraepithelial neopla-
oral contraceptives: results from a multi-national study. Int 1 sia in immunosuppressed women. Am 1 Obstet Gynecol
Cancer 1993;55:228-36. 1984; 150:300-8.
128. Ursin G, Peters RK, Henderson BE, et al. Oral contraceptive use 150. Ho GYF, Burk RD, Fleming I, et al. Risk for human papillomavi-
and adenocarcinoma of cervix. Lancet 1994;344:1379-90. rus infection in women with human immunodeficiency virus-
129. Chilvers C, Mant D, Pike MC. Cervical adenocarcinoma and oral induced immunosuppression. Int J Cancer 1994;56:788-92.
contraceptives. Br Med J1987;295:1446-7. 151. Maiman M, Tarricone N, Vieira J, et al. Colposcopic evaluation
130. Peters RK, Chao A, Mack TM, et al. Increased frequency of ade- of human immunodeficiency virus-seropositive women. Obstet
nocarcinoma of the uterine cervix in young women in Los Ange- Gynecol 1991; 78:84-8.
les County. J N a t l Cancer Inst 1986; 76:423-8. 152. Vermund SH, Kelly KF, Klein RS, et al. High risk of human pap-
131. SchwartzSM, Weiss NS. Increased incidence of adenocarcinoma illomavirus infection and cervical squamous intraepithelial le-
of the cervix in young women in the United States. Am 1Epide- sions among women with symptomatic human immunodefi-
miol 1986; 124:1045-7. ciency virus infection. Am J Obstet Gynecol 1991; 165:392-400.
132. Auborn KJ, Woodworth C, DiPaolo JA, et al. The interaction be- 153. Palefsky JM, Holly EA, Gonzales J, et al. Detection of human
tween HPV infection and estrogen metabolism in cervical carci- papillomavirus DNA in anal intraepithelial neoplasia and anal
nogenesis. IntJ Cancer 1991;49:867-9. cancer. Cancer Res 1991; 51:1014-9.
133. Monsonego J, Magdalenat H, Catalan F, et al. Estrogen and pro- 154. Strickler HD, Rattray C, Escoffrey C, et al. Human T-cell lym-
Epidemiology of Cervical Carcinogenesis/Schifiman and Brin ton 1901

photropic virus type I and severe neoplasia of the cervix in Ja- 161. Chan SY, Ho L, Ong CK, et al. Molecular variants of human
maica. ZntJ Cancer 1995;60:1-4. papillomavirus type 16 from four continents suggest ancient
155. Lancaster WD, Olson C. Animal papillomaviruses. Microbiol Rev pandemic spread of the virus and its coevolution with human-
1982;46(2):191-207. kind. ] Virol 1992;66(4):2057-66.
156. Sundberg JP. Papillomavirus infections in animals. In: Syrjanen 162. Scheffner M, Munger K, Byrne JC, et al. The state of the p53 and
K, Gissmann L. Koss L, editors. Papillomaviruses and human retinoblastoma genes in human cervical carcinoma cell lines.
disease. Heidelberg: Springer-Verlag, 1987:40-103. Droc Natl Acad Sci USA 1991;88:5523-7.
157. Crawford L. Prospects for cervical cancer vaccines. Cancer Sum 163. deBritton RC, Hildesheim A, deLao SL, et al. Human papillo-
1993;16:215-29. mavirus and other influences on survival from cervical cancer in
158. McArdleJP, Muller HK. Quantitative assessment of Langerhans’ Panama. Obstet Gynecol 1993; 81:19-24.
cells in human cervical intraepithelial neoplasia and wart virus 164. Campo MS, Grindlay GJ, O’Neil BW, et al. Prophylactic and
infection. Am J Obstet Gynecol 1986; 154:509-15. therapeutic vaccination against a mucosal papillomavirus. J Gen
159. Parazzini F, La Vecchia C, Negri E, et al. Risk factors for adeno- Viroi 1993;74:945-53.
carcinoma of the cervix: a case-control study. Br Cancer 165. Schiffman MH, Brinton LA, Fraumeni JF Jr,, Devesa S. Uterine
1988;571201-4. cervix. In: Schottenfeld D, Fraumeni JF Jr., editors. Cancer epi-
160. Shroyer KR. IIuman papillomavirus and endocervical adenocar- demiology and prevention. 2nd ed. New York: Oxford Univer-
cinoma. Hum Path01 1993;24:119-20. sity Press. In press.