CONTRACEPTION
ORAL CONTRACEPTIVES AND CERVICAL NEOPIASIA
Louise A. Brinton, Ph.D.
Environmental Epidemiology Branch National Cancer Institute Executive Plaza North, Rm. 443 Bethesda, MD 20892
ABSTRACT Although initial studies examining the relationship of oral contraceptives to risk of cervical neoplasia were reassuring, more recent studies provide some evidence of a positive relationship, particularly for long-term usage. Results, however, are difficult to interpret, because of a variety of methodologic complexities, including potential sources of confounding and bias. Sexual behavior and Pap smear screening have been identified as important confounders, but in several well-controlled studies residual excess risks of nearly 2-fold persist for users of 5 or more years. A possible promotional effect of oral contraceptives is suggested by higher risks associated with recent usage, There also is some suggestion of a stronger effect for adenocarcinomas than for squamous cell tumors. A relationship is biologically possible, given findings of hormone receptors in cervical tissue and the fact that oral contraceptives have been found to induce cervical hyperplasia. In addition, oral contraceptives may induce proliferation of the human papillomaviruses, the leading suspect agent for cervical cancer. Although a number of lines of evidence support a relationship of oral contraceptives to cervical cancer risk, firm conclusions await the results of additional studies that specifically address some of the methodologic shortcomings of previous investigations. In particular, additional follow-up studies are needed to define the effect of oral contraceptives on the natural history of cervical lesions. INTRODUCTION Initial studies examining the relationship of oral contraceptive use to risk of cervical neoplasia were reassuring (l-4), although in most the extent of exposure was limited. Recent investigations, however, indicate some evidence of an association, particularly among long-term users, Issues of study design and analysis pose problems of interpretation, with questions arising about the potential for confounding and sources of bias (5,6). Of major concern in most studies is whether oral contraceptive effects merely reflect the influence of other correlates of cervical cancer risk, including sexual behavior, socioeconomic status, and Pap smear screening history.
JUNE 1991 VOL. 43 NO. 6
561
CONTRACEPTION
Although cervical cancer had not previously been assumed to have an hormonal etiology, there is growing evidence to support a role for hormonal factors. Of note is that cervical tissue has been found to have hormone receptors and that oral contraceptives are capable of inducing cervical hyperplasia. Recent studies that have reported a strong association between multiparity and cervical cancer have further raised concern regarding the role of hormonal factors. A biologic role for oral contraceptives in the etiology of cervical cancer is thus plausible, although effects must be interpreted cautiously due to a variety of methodologic and analytic complexities. RESULTS OF COHORT STUDIES A number of cohort studies have attempted to assess the relationship between oral contraceptive use and risk of cervical neoplasia (Table I) (7-10). Although most of the cohort studies that have evaluated the relationship of oral contraceptive use to subsequent cervical cancer risk have found an excess of cervical cancer and precursor conditions among women exposed to oral contraceptives, the limited information on potential confounding factors makes these studies difficult to interpret. In addition, during the course of these studies, few women have developed invasive cancer of the cervix, requiring a more intensive focus on precursor conditions. In a follow-up study of women attending Oxford Family Planning Association clinics, Vessey et al. (10) found that the incidence of cervical neoplasia (preinvasive and invasive) rose from 0.9 per 1,000 women-years among those with up to 2 years of oral contraceptive use to 2.2 among those with more than 8 years use. Similarly, Andolsek et al. (7) found evidence of an increasing rate of cervical neoplasia with extended months of usage, although the follow-up was of limited duration (average of 4.5 years). In both of these studies, all cases of invasive cancer occurred among oral contraceptive users. Although neither of these studies was able to control extensively for confounding influences, it did not appear from a sub-study conducted by Vessey et al. (11) that oral contraceptives users were different in their sexual histories from the comparison group (IUD users). However, Swan and Brown (12) concluded that the nearly 4-fold excess risk associated with long-term oral contraceptive use (>4 years) in the cohort study reported by Peritz et al. (9) was likely to be highly confounded by sexual behavior. In the most comprehensive study to date (8), it was possible to control for the influence of age, socioeconomic status, parity, smoking, history of sexually transmitted diseases, and number of previous cervical smears. After standardization for these factors, oral contraceptive users had approximately twice the incidence of cervical cancer as non-users, with the risk rising to a 4-fold excess after 10 years of use. RESULTS OF CASE-CONTROL STUDIES Results from a number of recent case-control studies are summarized in Table II (13-30). Studies reviewed include those where
JUNE1991VOL.43NO.6
Q,
5
E
g
Author, Date
1988 (8)
2
1
65,101
257,028
20,18d 20,4984
72
62
112 74
Rate’ CIS
20
18
15 11
Rate’ Invasive
Users
assessing
Contraceptive
UYS of Follow-up
Oral
Rate per 100,000 women CIS = carcinoma in situ a = age b = socioeconusic status c = parity d = smoking e = sexually transmitted diseases f = marital status g = ntmber of cervical smears
Vessey 1983 (10)
Seral
Andolsek 1983 (7)
First
Cohort studies
26,432
182,866
65,637
UYS of Follou-up
I
4
3
45
21
79
Rate’ CIS
Previous
Exposure Relationships
COt’btL3CeptiVeS
2.4-fold higher incidence cervical neoplasia for ? 8 vs.< 2 years users
4-fold higher cancer incidence for >I0 yr users v. non-users
Significant effect among previous users exposed for 13-24 months, especially among those 30-34 years at enrollment
Additional
neoplasia
contraceptives
users of oral
a
a,b,c.d.e
a.c,f,g
users of oral
Non-previous
0
10
0
and cervical
Adjustmeqt Factors
contraceptives
Rate’ Invasive
Group
betueen oral
Coqarison
the relationship
Table
e
i
f g h i j
= = = = =
153
Invasive
1985 (20)
(23)
Invasive
. ‘
1
= = = =
were pregnant
age race socioeconomic status nt&er of sexual partners age at first intercourse
Cases and controls
b c d e
a =
Invasive
Vii0 1985 (30)
CIS/Invasive
CIS/lnvasive
1989 (28)
Invasive
Reeves 1985 (29)
Slattery
1986 (27)
1990 (26)
Invasive
Peters
Parezzini
CIS
Dysplasia CIS
1988 (24)
CIN Invasive
CIS Invasive
CIN
Ory 1977 (25)
Nolina
Mandelson lW0
1986 (22)
1988 (21)
La Vecchia
Irwin
Hellberg
CIN
Invasive
1980 (19)
1987 (18)
Ebeling
Harris
CIN and Invasive
Cuzick 1989 (17)
Table
5,246
309
408
200
323
8,553
254
181
202 225
764
2802
422
275
135
153
196
No. of Controls
the relationship
uomen.
TYpe of Controls
between oral
k 1 m n
= = = =
Hospi tat
d,h, i
d.e,f,g,h,i
a,b,c.d.e.f
a,b,e,f,h,i
d,e, i
a.b.c,d.e.f.g a.c,d,e.f.h.o
a,d,e,f,g,h
d
a,c,i,o
a.c.d,e.f.h,i,i.q a&
c.d.e.f.g.h
a.d.e.f.h
a,c,d,e,f,h,i i.k,i.m.n
a,d,e,f,g,h
d, i
neoplasia
1.5
1.8
1.3
1.1
2.5
1.6 4.7
1.0
2.2
0.7 1.7
2.0 0.9
1.3
2.1
1.8
5.7
0.7
2.3
1.2
1.8
RR
neoplasia
10 years
7 years
5 years
Ever use
5 years
use Ever use
5 years
2 years
>
t
use
use
use
5 years
use
Ever use
6 years
L 10 years
t
?: 3 years use
?
use use
use
use
use
use
Ever use Ever use
t 10 years L 5 years
>
use use
6 years use
2 10 years
:
t
t
? 10 years
b
Measure of Exposure
o = sexual behaviour of male partners p = HPV assay results q = barrier methods of contraception
and cervical
Adjustment Factors ’
contraceptives
body mass index age at menopause age at menarche noncontraceptive female hormone use
Neighbourhood
Cosnwnity
Neighbourhood
Hospital
Family planning
Screening
Population
Screening Hospital
Comwmity
Maternity
Hospital
Hospi tai
Clinic
Assorted
Comnunity
Hospital & Consunity
Cceewni ty
II
CIS = carcinoma in situ CNS = cervical intraepithelial
Pap smear screening history genital infection reproductive behaviour smoking marital status
726
156
266
200
367
854 147
133
140
202 225
415 149
1402
190
129
135
117
CIS
1987 (16)
Srock 1989 (15)
479
No. of Cases
assessing
Celantano
Invasive
Type
studies
759
1990 (14)
Disease
case-control
Invasive
1986 (13)
Date
Brinton
Author,
Brinton
First
Selected H
there were a sufficient number of cases of carcinoma in situ or invasive cancer to assess effects of oral contraceptives on risk of cancer development. Although several studies have found no relationship between oral contraceptives and risk of cervical neoplasia, the majority indicate that long-term users are at excess risk, even after adjustment for social, sexual, and screening factors. Studies showing no relationship of risk to oral contraceptive use are generally those that have used neighborhood rather than community controls (16,27), which may reflect over-matching, or those that have focused on non-invasive abnormalities, presenting difficulties for interpretation because of possible detection biases (21,22,24) (see later discussion). In addition, the absence of an effect in several studies may reflect the limited number of long-term oral contraceptive users. In most of the studies that have reported positive relations with pill usage, the variable most strongly linked to risk was duration of use, with the relative risks in the well-controlled studies ranging from approximately 1.3 to 1.8 for users of 5 or more years. In several studies (18,23,26), recent users were at higher risk than non-recent users, supporting a late stage effect of oral contraceptives. Alternatively, this relationship may merely have reflected correlated effects of duration of use (13). In one investigation (14), recent and non-recent users were at similar risk, but recent long-term users were at highest risk, suggesting the need for considering joint effects of exposure measures. Most investigations have not reported differing effects by age at first use or interval since first use (13,15,26), although one investigation noted the highest risk for women whose use of oral contraceptives began prior to age 25, an association that persisted even after adjustment for duration of use effects (18). Two studies (13,15) have reported the highest risks among users of high estrogen potency pills, but effects of dose and duration of use were difficult to distinguish. In a number of studies, attempts have been made to assess whether effects of oral contraceptives vary by the presence of other cervical cancer risk factors. Given the strong effects of sexual behavior on risk, investigators have attempted to assess whether oral contraceptive relationships differ by various measures of promiscuity, or associated consequences, such as histories of venereal diseases. Parazzini et al. (26) found higher oral contraceptive-associated risks among women with multiple sexual partners, while both the studies of Brinton and others (13) and the World Health Organization (30) found effects of oral contraceptives t6 be greatest among w6men with histories of genital infections. In addition, Brinton et al. (13) found higher risks associated with oral contraceptive use among women who had never used barrier methods of contraception. These interactions suggest that oral contraceptives may act as co-carcinogens with transmissible agents, in line with previous suggestions by zur Hausen (31) regarding the human papillomaviruses. The one study that has been able to examine use of oral contraceptives according to HPV detection status (14) found no differential effect, but additional studies are needed which utilize m6re accurate HPV assay techniques.
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION Furthermore, several studies (26,32) have found an enhancement of the effects of oral contraceptives among women with multiple births. Although reasons for an excess risk associated with multiparity remain unresolved, it is possible that oral contraceptives may provide exogenous stimulus to potentially hazardous endogenous hormonal influences. ANALYTIC COMPLEXITIES IN ASSESSING RELATIONSHIPS Several articles have discussed the analytic complexities involved in assessing the relationship of oral contraceptives to risk of cervical neoplasia (5,6). Difficulties that arise in the conduct of studies in this area include potential confounding influences, selection biases, inappropriate control groups, and inclusion of users of other methods of contraception in comparison groups. Confounding Influences Of concern in evaluating the relationships of oral contraceptives to risk of cervical neoplasia is the extent to which associations reflect the influence of correlated risk factors. Although it has been widely presumed that oral contraceptive associations would be highly confounded by number of sexual partners, a stronger confounder in most studies has been the history of Pap smear screening, which is usually highly linked with oral contraceptive use. It is unclear whether this is because oral contraceptive users tend to avail themselves of more frequent screening or whether a Pap smear is considered necessary before oral contraceptives are prescribed. However, it is established that failure to consider the confounding influences of screening history can result in an underestimate of the effects of oral contraceptives. For example, in a study in five U.S. locations (13), interval since last Pap smear acted as a strong negative confounder, since women who failed to have regular Pap smears, and who were thus at high cervical cancer risk, had a low rate of exposure to oral contraceptives. Initially, oral contraceptive users had a relative risk of 0.8, but after appropriate adjustment, oral contraceptive users were found to have a relative risk of 1.5, rising to approximately 2fold among long-term users. Similar confounding was noted in a study conducted by the World Health Organization (30), where an adjusted risk of 1.2 was associated with any use, increasing to 1.5 for users of 5 or more years. Although most studies have been concerned with confounding influences of female sexual behavior on oral contraceptive associations, it is possible that male sexual behavior may also play an intervening role, especially given evidence of a "male factor" in the etiology of cervical cancer (33). Only two studies (14,28) have had information on male sexual behavior, and this was not an important source of confounding to observed oral contraceptive associations. However, further evaluation of this factor as a potential confounding influence is warranted, especially since one of these studies (28) relied on female subjects to provide information on their partner's sexual behavior.
666
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION Recent interest has focused on the role of HPV in the etiology of cervical neoplasia, and studies are now beginning to assess this as a potential modifier of oral contraceptive associations. A number of studies (34-36) have suggested a higher prevalence of HPV infection among oral contraceptive users, but whether this represents uncontrolled confounding by factors such as sexual behavior or an effect on the ability to detect the virus is yet to be clarified. Only one epidemiologic study to date has had the ability to adjust for measures of HPV in assessing oral contraceptive associations (14), and effects were not appreciably altered; however, additional studies are needed which utilize more refined measures of HPV infection status. Although numerous studies have shown that smokers, particularly long-term or high-intensity smokers, are at an excess risk of cervical neoplasia (37), the magnitude of the effect of smoking on risk or the correlation between smoking and oral contraceptive use have usually not been sufficiently strong to account for the elevation in risk among oral contraceptive users. Detection Bias A number of studies have concluded that excess risks associated with oral contraceptive use probably do not reflect the influence of confounding variables. However, an additional explanation for the continued elevation in risk is that women who have taken oral contraceptives might have been more likely to have their cervical cancer diagnosed because of increased Pap smear screening. Such a concern is more relevant to carcinoma-in-situ (CIS) than invasive disease, since the diagnosis of CIS is usually dependent on results of a Pap smear. For instance, an excess risk of CIS associated with oral contraceptive use in the study of Irwin et al. (21) was interpreted as being due to such a detection bias, since elevations in risk were only observed among recent users, who were those with more intensive Pap smear screening histories. In contrast, oral contraceptive use was associated with a decreased risk of invasive cancer in this same study, presumably because oral contraceptive users were more likely to have had their diseases detected at a preinvasive stage. Thus, detection bias may operate in the opposite direction for invasive disease, with an attenuation in the true relative risks, It is important therefore to assess the means by which invasive disease was detected. However, in the one investigation of invasive cancer that attempted to assess how oral contraceptive-associated risks were affected by the method of detection (30), exclusion from analysis of women who presented without the usual symptom of vaginal bleeding failed to produce substantial changes in the risk estimates, arguing against major effects of detection bias on observed oral contraceptive relationships. Aoorooriateness of Comoarison Croups A correct interpretation of effects of oral contraceptives depends on the types of women in the comparison groups utilized. This issue of comparability pertains to both the groups of non-diseased women as well as non-users of oral contraceptives.
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION In terms of the comparison group of non-diseased women, special attention must be given to whether these women are representative in terms of their contraceptive practices. Although many studies have utilized comparison groups consisting of women hospitalized for various conditions, questions arise regarding their appropriateness, especially since studies have shown that oral contraceptive users may be overrepresented among hospitalized controls (38). Although this problem may be alleviated by eliminating from the control group women currently hospitalized for certain conditions, e.g., other gynecologic conditions, caution is still warranted in interpretation of effects derived from studies using hospital control groups. When mixtures of diagnoses are included, it is recommended that associations be examined separately for the different types of controls utilized. It is further important that the control group consist of women potentially at risk of cervical cancer; thus those with a hysterectomy (including removal of the cervix) should be excluded. This elimination can result in substantially increased efforts to identify a suitable control group, since in the U.S. approximately one-third of women in the age range with a high risk for cervical cancer are now estimated to have undergone a hysterectomy. Of further concern is that users of oral contraceptives be compared appropriately with others at risk for contraception. Specific attention needs to be given to non-contracepting women, who may themselves be at an elevated risk of cervical cancer,(l9), as well as users of barrier methods of contraception (diaphragm and condom), who have been found in a number of studies to be at low cervical cancer risk (1,4,39,40). Several of the prospective studies have compared oral contraceptive users with users of the intrauterine device (IUD), since the IUD has been presumed not to exert a protective effect on risk. However, use of the IUD in relation to cervical cancer risk has not been well investigated, and further research might well indicate some alteration in risk, similar to the unsuspected reduction in risk recently associated with spermicidal agents (16,41). Although there probably is no ideal comparison group, it is important for studies to assess the risk associated with various methods of contraception in the control group, and to exclude, when necessary, women whose behavior may artificially lower or increase oral contraceptive-related risks. Several studies have performed these exclusions (13,14,19), and found that oral contraceptive relationships were not substantially altered, but studies should continue to examine potential effects associated with the comparison groups utilized. TRANSITION TIME STUDIES There is now evidence of a continuum of disease from cervical dysplasia to carcinoma in situ to invasive cancer. Although progression to more advanced states is well documented, factors that affect the natural history of these lesions remain unclarified. There has thus been interest in determining whether oral contraceptives are involved in this process. In one prospective investigation of 203 oral contraceptive users and 97 non-users (almost all IUD users) with cervical dysplasia, the probability of progression to carcinoma in situ after 7 years (estimated by life table methods) was 0.30 in oral
666
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION
contraceptive users and 0.05 in non-users, but these estimates were based on only 10 and 3 progressions, respectively (42). More recently, Bamford et al. (43) examined records of 105 women in whom cervical cytology revealed progression from mild or moderate dyskaryosis to severe dyskaryosis, along with records of 152 control women without progression. No effect of contraceptive method on the rate of progression was found. Given evidence that oral contraceptives appear unrelated to low grade squamous intraepithelial lesions but associated with some increase in risk of high grade lesions (36), further research on the effects of oral contraceptives on the potential for progression of lesions seems warranted. ASSOCIATION WITH CERVICAL ADENOCARCINONA The possibility of a relationship between adenocarcinoma of the cervix and oral contraceptives was raised when Dallenbach-Hellweg (44) described a series of 28 women with this disease, the majority of whom had had prolonged exposure to the pill. Subsequently, descriptive surveys noted a rise in the incidence of cervical adenocarcinoma among young women, and oral contraceptive use was suggested as a possible explanation (45-47). In two case-control studies, effects of oral contraceptives were stronger for adenocarcinoma than for squamous cell neoplasms (13,14), but in three other investigations no differential effect was observed (48-50). It is known that oral contraceptives cause a marked increase in the glandular-to-squamous ratio of cervical epithelium, and microglandular hyperplasia is now recognized as an oral contraceptive-related lesion. It has been proposed that oral contraceptives may cause endocervical adenocarcinomas by long-term stimulation of the proliferating glandular epithelium. However, pathologic studies cast some doubt on this hypothesis since microglandular hyperplasia has not been established as a precursor of adenocarcinoma (48,51,52). Although numerous case reports have been published of cervical adenocarcinomas occurring among oral contraceptive users (53-55), further research is needed to define the relationships of cervical adenocarcinomas to both preceding microglandular hyperplasia as well as oral contraceptive usage. BIOMGIC
PLAUSIBILITY
There are a number of lines of evidence that support a biologic role for oral contraceptives in the etiology of cervical neoplasia. Hormonal steroids are capable of promoting the development of cervical cancer in animals (56,57). In humans, cervical tissue has been found to have hormone receptor sites (58) and, as previously described, histologic changes in cervical epithelium have been demonstrated to result from the administration of contraceptive steroids. Furthermore, oral contraceptives have been associated with the production of an estral "clear channel" mucus that could facilitate entry of mutagens (59,60) and with immunologic alterations that could increase susceptibility to viral agents. Hormones have also been found to enhance the transcription of the HPV genome both in vitro (61) and in vivo (62). The possibility of an effect mediated through dietary factors has also been suggested in that oral contraceptive users, particularly those with cervical dysplasia, have been found to have low
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION serum and red blood cell folacin levels. In one small clinical trial, large doses of folate supplementation among oral contraceptive users with cervical dysplasia led to improved cervical biopsy scores (63). CONCLUSIONS
AND RJXONNENDATIONS
FOR FURTHER RESEARCH
Evidence is accumulating that oral contraceptives may play a role in the etiology of cervical neoplasia. Data from well-designed epidemiologic studies indicate that excess risks persist after adjustment for a variety of potential confounding factors. Although the evaluation of relationships is complex because of the possibility of biases, it does not appear in most studies as though the elevations in risk can be attributed to methodologic shortcomings. Thus, causal interpretations must be considered, particularly in view of accumulating data on dose-response relationships of risk with duration of use. Most of the data linking oral contraceptives to an elevated risk of cervical neoplasia derive from case-control studies, the results of which must be interpreted cautiously. Well-designed follow-up studies are critically needed to allow an assessment of how oral contraceptives affect the natural history of cervical lesions. It is essential that these studies collect detailed information on other cervical cancer In risk factors, including sexual behavior and screening histories. addition, given concerns that oral contraceptives may exacerbate the effects of infectious agents, these studies will need to accurately assess HPV infection status. Studies which attempt to address whether this interaction is mediated through an indirect effect, such as through altering the immune system, or a direct effect, e.g., by causing viruses to more effectively replicate, should also be pursued. Supporting an effect of oral contraceptives on cervical cancer risk is the recent observation of increasing incidence rates in the United Kingdom (8), although to date no such trend has been observed in the United States (64). As previously described, a recent increase in the rates of cervical adenocarcinoma has been noted, especially among younger women, strengthening support from some case-control studies of stronger oral contraceptive effects for these neoplasms than for squamous tumors. Further studies, however, are needed which specifically assess these relationships, adjusting for the appropriate risk factors, which may be different than those associated with the more common squamous cell cancers (49,65). Although there are now a number of studies which indicate an elevation in risk of cervical neoplasia with extended durations of oral contraceptive use, additional effects of timing of usage remain less well clarified. There is some evidence that oral contraceptives may operate at later stages in the carcinogenic process, but further studies are needed to assess detailed usage patterns. Given recent suggestions that the human papillomaviruses may be a necessary but not sufficient agent for cervical cancer (66), there is interest in With assessing the interactive effects of HPV and oral contraceptives. rapidly advancing HPV assay techniques (such as the polymerase chain reaction test, which minimizes measurement error associated with other
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION assays), clarification of the interplay of these factors should be greatly enhanced. REFERENCES
1.
Boyce JG, Lu T, Nelson JH, Fruchter RG. Oral contraceptives and cervical carcinoma. Am .IObstet Gynecol 1977;128:761-6.
2.
Melamed MR, Flehinger BJ. Early incidence rates of precancerous cervical lesions in women using contraceptives. Gynecol Oncol 1973;1:290-8.
3.
Sandmire HF, Austin SD, Bechtel RC. Carcinoma of the cervix in oral contraceptive steroid and IUD users and nonusers. Am J Obstet Gynecol 1976;125:339-45.
4.
Worth AJ, Boyes DA. A case-control study into the possible effects of birth control pills on pre-clinical carcinoma of the cervix. J Obstet Gynaecol Br Commonw 1972;79:673-9.
5.
Piper JM. Oral contraceptives and cervical cancer. 1985;22:1-14.
6.
Swan SH, Petitti DB. A review of problems of bias and confounding in epidemiologic studies of cervical neoplasia and oral contraceptive use. Am J Epidemiol 1982;115:10-8.
7.
Andolsek L, Kovacic J, Kozuh M, Litt B. Influence of oral contraceptives on the incidence of premalignant and malignant lesions of the cervix. Contraception 1983;28:505-19.
8.
Beral V, Hannaford P, Kay C. Oral contraceptive use and malignancies of the genital tract: Results from The Royal College of General Practitioners' Oral Contraception Study. Lancet 1988;2:1331-5.
9.
Peritz E, Ramcharan S, Frank J, Brown WL, Huang S, Ray R. The incidence of cervical cancer and duration of oral contraceptive use. Am J Epidemiol 1977;106:462-9.
Gynecol Oncol
10.
Vessey MP, Lawless M, McPherson K, Yeates D. Neoplasia of the cervix uteri and contraception: A possible adverse effect of the pill. Lancet 1983;2:930-4.
11.
Vessey MP, Lawless M, McPherson K, Yeates D. Oral contraceptives and cervical cancer (letter). Lancet 1983;2:1358-9.
12.
Swan SH, Brown WL. Oral contraceptive use, sexual activity, and cervical carcinoma. Am J Obstet Gynecol 1981;139:52-7.
13.
Brinton LA, Huggins GR, Lehman HF, et al. Long-term use of oral contraceptives and risk of invasive cervical cancer. Int J Cancer 1986;38:339-44.
JUNE 1991 VOL. 43 NO. 6
591
CONTRACEPTION 14.
Brinton LA, Reeves WC, Brenes and risk of invasive cervical 4-11.
MM, et al. Oral contraceptive use cancer. Int J Epidemiol 1990;19:
15.
Brock KE, Berry G, Brinton LA, et al. Sexual, reproductive and contraceptive risk factors for carcinoma-in-situ of the uterine cervix in Sydney. Med J Austral 1989;150:125-30.
16.
Celentano DD, Klassen AC, Weisman CS, Rosenshein NB. The role contraceptive use in cervical cancer: The Maryland cervical cancer case-control study. Am J Epidemiol 1987;126:592-604.
17.
Cuzick cervix
18.
Ebeling K, Nischan P, Schindler CH. and risk of invasive cervical cancer Int J Cancer 1987;39:427-30.
19.
Harris RWC, Brinton LA, Cowdell RH, et al. Characteristics of women with dysplasia or carcinoma in situ of the cervix uteri. J Cancer 1980;42:359-69.
J, De Stavola B, McCance D, et al. A case-control cancer in Singapore. Br J Cancer 1989;60:238-43.
of
study
of
Use of oral contraceptives in previously screened women.
Br
20.
Hellberg D, Valentin J, Nilsson S. Long-term use of oral contraceptives and cervical neoplasia: An association confounded by other risk factors? Contraception 1985;32:337-46.
21.
Oral contraceptives Irwin KL, Rosero-Bixby L, Oberle MW, et al. and cervical cancer risk in Costa Rica. Detection bias or causal association? JAMA 1988;259:59-64.
22.
Oral La Vecchia C, Decarli A, Fasoli M, et al. cancers of the breast and of the female genital results from a case-control study. Br J Cancer
23.
Mandelson MT, Daling JR, White E, Chu J, McKnight B. Further evidence that duration and recency of oral contraceptive use are associated with invasive cervical cancer (abstract). Am J Epidemiol 1990;132:77B.
24.
Molina R, Thomas DB, Dabancens A, et al. Oral contraceptives and Cancer Res 1988;48:1011-5. cervical carcinoma in situ in Chile.
25.
Ory HW, Conger SB, Naib 2, Tyler CW Jr, Hatcher RA. Preliminary analysis of oral contraceptive use and risk of developing premalignant lesions of the uterine cervix. In: Garattini S, Berendes HW, eds. Pharmacology of Steroid Contraceptive Drugs. New York: Raven Press, 1977: 211-8.
26.
Parazzini F, La Vecchia C, Negri E, Maggi R. Oral use and invasive cervical cancer. Int J Epidemiol
662
contraceptives and Interim tract. 1986;54:311-7.
contraceptive 1990;19:259-63.
JUNE 1991 VOL. 43 NO. 6
27.
Peters RK, Thomas D, Hagan DG, Mack TM, Henderson BE. Risk factors for invasive cervical cancer among latinas and non-latinas in Los Angeles County. J Nat1 Cancer Inst 1986;77:1063-77.
28.
Slattery ML, Overall JC Jr, Abbott TM, French TK, Robison IM, Gardner J. Sexual activity, contraception, genital infections, and cervical cancer: Support for a sexually transmitted disease hypothesis. Am J Epidemiol 1989;130:248-58.
29.
Reeves WC, Brinton LA, Brenes MM, et al. Case-control study of cervical cancer in Herrera Province, Republic of Panama. Int J Cancer 1985:36:55-60.
30.
World Health Organization Collaborative Study of Neoplasia and Steroid Contraceptives. Invasive cervical cancer and combined oral contraceptives. Br Med J 1985;290:961-5.
31.
zur Hausen H. Human genital cancer: Synergism between two virus infections or synergism between a virus infection and initiating events? Lancet 1982:2:1370-2.
32.
Brinton LA, Reeves WC, Brenes MM, et al. Parity as a risk factor for cervical cancer. Am J Epidemiol 1990;130:486-96.
33.
Brinton LA, Reeves WC, Brenes MM, et al. The male factor in the etiology of cervical cancer among sexually monogamous women. Int J Cancer 1989;44:199-203.
34.
Hildesheim A, Reeves WC, Brinton LA, et al. Association of oral contraceptive use and human papillomaviruses in invasive cervical cancers. Int J Cancer 1990;45:860-4.
35.
Lorincz AT, Schiffman MH, Jaffurs WJ, Marlow J, Temple GF. Temporal associations of human papillomavirus infection with cervical cytologic abnormalities. Am J Obstet Gynecol 1990;162: 645-51.
36.
Negrini BP, Schiffman MH, Kurman RJ, et al: Oral contraceptive use, human papillomavirus infection, and risk of early cytological abnormalities of the cervix. Cancer Res 1990;50:4670-5.
37.
Winkelstein W Jr. Smoking and cervical cancer--Current status: review. Am J Epidemiol 1990;131:945-57.
38.
Hoover R, Bain C, Cole P, MacMahon B. Oral contraceptive use: Association with frequency of hospitalization and chronic disease risk indicators. Am J Public Health 1978:68:335-42.
39.
Melamed MR, Koss LG, Flehinger BJ, Kelisky RP, Dubrow H. Prevalence rates of uterine cervical carcinoma in situ for women using the diaphragm or contraceptive oral steroids. Br Med J 1969;3:195-200.
JUNE 1991 VOL. 43 NO. 6
A
40.
Wright NH, Vessey MP, Kenward B, McPherson K, Doll R. Neoplasia and dysplasia of the cervix uteri and contraception: A possible protective effect of the diaphragm. Br J Cancer 1978;38:273-9.
41.
Hildesheim A, Brinton LA, Mallin K, et al. Barrier and spermicidal contraceptive methods and risk of invasive cervical cancer. Epidemiology 1990;1:266-72.
42.
Stern E, Forsythe AB, Youkeles L, Coffelt CF. Steroid contraceptive use and cervical dysplasia: Increased risk of progression. Science 1977;196:1460-2.
43.
Bamford PN, Forbes-Smith PA, Rose GL, et al. An analysis of factors responsible for progression or regression of mild and moderate cervical dyskaryosis. Br J Fam Plan 1985;11:5-8.
44.
Dallenbach-Hellweg G: On the origin and histological structure of adenocarcinoma in women under 50 years of age. Path01 Res Pratt 1984;179:38-50.
45.
Chilvers C, Mant D, Pike MC. Cervical adenocarcinoma and oral contraceptives. Br Med J 1987;295:1446-7.
46.
Peters RK, Chao A, Mack TM, Thomas D, Bernstein L, Henderson BE. Increased frequency of adenocarcinoma of the uterine cervix in young women in Los Angeles County. J Nat1 Cancer Inst 1986;76:423-a.
47.
Schwartz SM, Weiss NS. Increased incidence of adenocarcinoma of the cervix in young women in the United States. Am J Epidemiol 1986;124:1045-7.
48.
Jones MW, Silverberg SG. Cervical adenocarcinoma in young women: Possible relationship to microglandular hyperplasia and use of oral contraceptives. Obstet Gynecol 1989;73:984-9.
49.
Parazzini F, La Vecchia C, Negri E, Fasoli M, Cecchetti G. Risk factors for adenocarcinoma of the cervix: A case-control study. Br J Cancer 1988;57:201-4.
50.
Persson E, Einhorn N, Pettersson F. A case-control study of oral contraceptive use in women with adenocarcinoma of the uterine cervix. Eur J Obstet Gynecol Reprod Biol 1987;26:85-90.
51.
Tase T, Okagaki T, Clark BA, Twiggs LB, Ostrow RS, Faras AJ. Human papilloma virus DNA in glandular dysplasia and microglandular hyperplasia: Presumed precursors of adnocarcinoma of the uterine cervix. Obstet Gynecol 1989;73:1005-8.
52.
Valente PT, Hanjani P. Endocervical neoplasia in long-term users of oral contraceptives: Clinical and pathologic observations. Obstet Gynecol 1986;67:695-704.
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION
53.
Ireland D, Hardiman P, Monaghan JM. Adenocarcinoma of the uterine cervix: A study of 73 cases. Obstet Gynecol 1985;65:82-5.
54.
Shingleton HM, Gore H, Bradley B, Soong SJ. Adenocarcinoma of the cervix. Am J Obstet Gynecol 1981;139:799-814.
55.
Tamini HK, Figge DC. Adenocarcinoma of the uterine cervix. Gynecol Oncol 1982;13:335-44.
56.
Dunn TB. Cancer of the uterine cervix in mice fed a liquid diet containing an antifertility drug. J Nat1 Cancer Inst 1969;43:681-92.
57.
Kaminetzky HA. Metylcholanthrene-induced cervical dysplasia and the sex steroids. Obstet Gynecol 1966;27:489-93.
58.
Ford LC, Berek JS, Lagasse LD, Hacker NF, Heins YL, Delange RJ. Estrogen and progesterone receptor sites in malignancies of the uterine cervix, vagina, and vulva. Gynec Oncol 1983;15:27-31.
59.
Cramer DW: Uterine cervix. In: Schottenfeld D, Fraumeni JR, eds. Cancer Epidemiology and Prevention. Philadelphia: WB Saunders Co, 1982:881-900.
60.
Jordan JA, Singer A. Co, 1976.
61.
Pater MM, Hughes GA, Hyslop DE, Nakshatri H, Pater A. Glucocorticoid-dependent oncogenic transformation by type 16 but not type 11 human papilloma virus DNA. Nature (Lond.) 1988;335:832-5.
62.
Gloss B, Bernard HU, Seedorf K, Klock G. The upstream regulatory region of the human-papilloma-virus-16 contains an EZ-proteinindependent enhancer which is specific for cervical carcinoma cells and regulated by glucocorticoid hormones. EMBO J
The Cervix.
Philadelphia:
W.B. Saunders
1987;6:3735-43.
63.
Butterworth CE, Hatch KD, Gore H, Mueller H, Krumdieck CL. Improvement in cervical dysplasia associated with folic acid therapy in users of oral contraceptives. Am J Clin Nutr 1982;35:73-82.
64.
Devesa SS, Young JL, Brinton LA, Fraumeni JF. Recent trends in cervix uteri cancer. Cancer 1989;64:2184-90.
65.
Brinton LA, Tashima KT, Lehman HF, et al. Epidemiology of cervical cancer by cell type. Cancer Res 1987;47:1706-11.
66.
zur Hausen H. Papillomaviruses in anogenital cancer as a model to understand the role of viruses in human cancers. Cancer Res 1989;49:4677-81.
JUNE 1991 VOL. 43 NO. 6
595
ORAL CONTRACEPTIVES AND CERVICAL NEOPIASIA
Louise A. Brinton, Ph.D.
Environmental Epidemiology Branch National Cancer Institute Executive Plaza North, Rm. 443 Bethesda, MD 20892
ABSTRACT Although initial studies examining the relationship of oral contraceptives to risk of cervical neoplasia were reassuring, more recent studies provide some evidence of a positive relationship, particularly for long-term usage. Results, however, are difficult to interpret, because of a variety of methodologic complexities, including potential sources of confounding and bias. Sexual behavior and Pap smear screening have been identified as important confounders, but in several well-controlled studies residual excess risks of nearly 2-fold persist for users of 5 or more years. A possible promotional effect of oral contraceptives is suggested by higher risks associated with recent usage, There also is some suggestion of a stronger effect for adenocarcinomas than for squamous cell tumors. A relationship is biologically possible, given findings of hormone receptors in cervical tissue and the fact that oral contraceptives have been found to induce cervical hyperplasia. In addition, oral contraceptives may induce proliferation of the human papillomaviruses, the leading suspect agent for cervical cancer. Although a number of lines of evidence support a relationship of oral contraceptives to cervical cancer risk, firm conclusions await the results of additional studies that specifically address some of the methodologic shortcomings of previous investigations. In particular, additional follow-up studies are needed to define the effect of oral contraceptives on the natural history of cervical lesions. INTRODUCTION Initial studies examining the relationship of oral contraceptive use to risk of cervical neoplasia were reassuring (l-4), although in most the extent of exposure was limited. Recent investigations, however, indicate some evidence of an association, particularly among long-term users, Issues of study design and analysis pose problems of interpretation, with questions arising about the potential for confounding and sources of bias (5,6). Of major concern in most studies is whether oral contraceptive effects merely reflect the influence of other correlates of cervical cancer risk, including sexual behavior, socioeconomic status, and Pap smear screening history.
JUNE 1991 VOL. 43 NO. 6
561
CONTRACEPTION
Although cervical cancer had not previously been assumed to have an hormonal etiology, there is growing evidence to support a role for hormonal factors. Of note is that cervical tissue has been found to have hormone receptors and that oral contraceptives are capable of inducing cervical hyperplasia. Recent studies that have reported a strong association between multiparity and cervical cancer have further raised concern regarding the role of hormonal factors. A biologic role for oral contraceptives in the etiology of cervical cancer is thus plausible, although effects must be interpreted cautiously due to a variety of methodologic and analytic complexities. RESULTS OF COHORT STUDIES A number of cohort studies have attempted to assess the relationship between oral contraceptive use and risk of cervical neoplasia (Table I) (7-10). Although most of the cohort studies that have evaluated the relationship of oral contraceptive use to subsequent cervical cancer risk have found an excess of cervical cancer and precursor conditions among women exposed to oral contraceptives, the limited information on potential confounding factors makes these studies difficult to interpret. In addition, during the course of these studies, few women have developed invasive cancer of the cervix, requiring a more intensive focus on precursor conditions. In a follow-up study of women attending Oxford Family Planning Association clinics, Vessey et al. (10) found that the incidence of cervical neoplasia (preinvasive and invasive) rose from 0.9 per 1,000 women-years among those with up to 2 years of oral contraceptive use to 2.2 among those with more than 8 years use. Similarly, Andolsek et al. (7) found evidence of an increasing rate of cervical neoplasia with extended months of usage, although the follow-up was of limited duration (average of 4.5 years). In both of these studies, all cases of invasive cancer occurred among oral contraceptive users. Although neither of these studies was able to control extensively for confounding influences, it did not appear from a sub-study conducted by Vessey et al. (11) that oral contraceptives users were different in their sexual histories from the comparison group (IUD users). However, Swan and Brown (12) concluded that the nearly 4-fold excess risk associated with long-term oral contraceptive use (>4 years) in the cohort study reported by Peritz et al. (9) was likely to be highly confounded by sexual behavior. In the most comprehensive study to date (8), it was possible to control for the influence of age, socioeconomic status, parity, smoking, history of sexually transmitted diseases, and number of previous cervical smears. After standardization for these factors, oral contraceptive users had approximately twice the incidence of cervical cancer as non-users, with the risk rising to a 4-fold excess after 10 years of use. RESULTS OF CASE-CONTROL STUDIES Results from a number of recent case-control studies are summarized in Table II (13-30). Studies reviewed include those where
JUNE1991VOL.43NO.6
Q,
5
E
g
Author, Date
1988 (8)
2
1
65,101
257,028
20,18d 20,4984
72
62
112 74
Rate’ CIS
20
18
15 11
Rate’ Invasive
Users
assessing
Contraceptive
UYS of Follow-up
Oral
Rate per 100,000 women CIS = carcinoma in situ a = age b = socioeconusic status c = parity d = smoking e = sexually transmitted diseases f = marital status g = ntmber of cervical smears
Vessey 1983 (10)
Seral
Andolsek 1983 (7)
First
Cohort studies
26,432
182,866
65,637
UYS of Follou-up
I
4
3
45
21
79
Rate’ CIS
Previous
Exposure Relationships
COt’btL3CeptiVeS
2.4-fold higher incidence cervical neoplasia for ? 8 vs.< 2 years users
4-fold higher cancer incidence for >I0 yr users v. non-users
Significant effect among previous users exposed for 13-24 months, especially among those 30-34 years at enrollment
Additional
neoplasia
contraceptives
users of oral
a
a,b,c.d.e
a.c,f,g
users of oral
Non-previous
0
10
0
and cervical
Adjustmeqt Factors
contraceptives
Rate’ Invasive
Group
betueen oral
Coqarison
the relationship
Table
e
i
f g h i j
= = = = =
153
Invasive
1985 (20)
(23)
Invasive
. ‘
1
= = = =
were pregnant
age race socioeconomic status nt&er of sexual partners age at first intercourse
Cases and controls
b c d e
a =
Invasive
Vii0 1985 (30)
CIS/Invasive
CIS/lnvasive
1989 (28)
Invasive
Reeves 1985 (29)
Slattery
1986 (27)
1990 (26)
Invasive
Peters
Parezzini
CIS
Dysplasia CIS
1988 (24)
CIN Invasive
CIS Invasive
CIN
Ory 1977 (25)
Nolina
Mandelson lW0
1986 (22)
1988 (21)
La Vecchia
Irwin
Hellberg
CIN
Invasive
1980 (19)
1987 (18)
Ebeling
Harris
CIN and Invasive
Cuzick 1989 (17)
Table
5,246
309
408
200
323
8,553
254
181
202 225
764
2802
422
275
135
153
196
No. of Controls
the relationship
uomen.
TYpe of Controls
between oral
k 1 m n
= = = =
Hospi tat
d,h, i
d.e,f,g,h,i
a,b,c.d.e.f
a,b,e,f,h,i
d,e, i
a.b.c,d.e.f.g a.c,d,e.f.h.o
a,d,e,f,g,h
d
a,c,i,o
a.c.d,e.f.h,i,i.q a&
c.d.e.f.g.h
a.d.e.f.h
a,c,d,e,f,h,i i.k,i.m.n
a,d,e,f,g,h
d, i
neoplasia
1.5
1.8
1.3
1.1
2.5
1.6 4.7
1.0
2.2
0.7 1.7
2.0 0.9
1.3
2.1
1.8
5.7
0.7
2.3
1.2
1.8
RR
neoplasia
10 years
7 years
5 years
Ever use
5 years
use Ever use
5 years
2 years
>
t
use
use
use
5 years
use
Ever use
6 years
L 10 years
t
?: 3 years use
?
use use
use
use
use
use
Ever use Ever use
t 10 years L 5 years
>
use use
6 years use
2 10 years
:
t
t
? 10 years
b
Measure of Exposure
o = sexual behaviour of male partners p = HPV assay results q = barrier methods of contraception
and cervical
Adjustment Factors ’
contraceptives
body mass index age at menopause age at menarche noncontraceptive female hormone use
Neighbourhood
Cosnwnity
Neighbourhood
Hospital
Family planning
Screening
Population
Screening Hospital
Comwmity
Maternity
Hospital
Hospi tai
Clinic
Assorted
Comnunity
Hospital & Consunity
Cceewni ty
II
CIS = carcinoma in situ CNS = cervical intraepithelial
Pap smear screening history genital infection reproductive behaviour smoking marital status
726
156
266
200
367
854 147
133
140
202 225
415 149
1402
190
129
135
117
CIS
1987 (16)
Srock 1989 (15)
479
No. of Cases
assessing
Celantano
Invasive
Type
studies
759
1990 (14)
Disease
case-control
Invasive
1986 (13)
Date
Brinton
Author,
Brinton
First
Selected H
there were a sufficient number of cases of carcinoma in situ or invasive cancer to assess effects of oral contraceptives on risk of cancer development. Although several studies have found no relationship between oral contraceptives and risk of cervical neoplasia, the majority indicate that long-term users are at excess risk, even after adjustment for social, sexual, and screening factors. Studies showing no relationship of risk to oral contraceptive use are generally those that have used neighborhood rather than community controls (16,27), which may reflect over-matching, or those that have focused on non-invasive abnormalities, presenting difficulties for interpretation because of possible detection biases (21,22,24) (see later discussion). In addition, the absence of an effect in several studies may reflect the limited number of long-term oral contraceptive users. In most of the studies that have reported positive relations with pill usage, the variable most strongly linked to risk was duration of use, with the relative risks in the well-controlled studies ranging from approximately 1.3 to 1.8 for users of 5 or more years. In several studies (18,23,26), recent users were at higher risk than non-recent users, supporting a late stage effect of oral contraceptives. Alternatively, this relationship may merely have reflected correlated effects of duration of use (13). In one investigation (14), recent and non-recent users were at similar risk, but recent long-term users were at highest risk, suggesting the need for considering joint effects of exposure measures. Most investigations have not reported differing effects by age at first use or interval since first use (13,15,26), although one investigation noted the highest risk for women whose use of oral contraceptives began prior to age 25, an association that persisted even after adjustment for duration of use effects (18). Two studies (13,15) have reported the highest risks among users of high estrogen potency pills, but effects of dose and duration of use were difficult to distinguish. In a number of studies, attempts have been made to assess whether effects of oral contraceptives vary by the presence of other cervical cancer risk factors. Given the strong effects of sexual behavior on risk, investigators have attempted to assess whether oral contraceptive relationships differ by various measures of promiscuity, or associated consequences, such as histories of venereal diseases. Parazzini et al. (26) found higher oral contraceptive-associated risks among women with multiple sexual partners, while both the studies of Brinton and others (13) and the World Health Organization (30) found effects of oral contraceptives t6 be greatest among w6men with histories of genital infections. In addition, Brinton et al. (13) found higher risks associated with oral contraceptive use among women who had never used barrier methods of contraception. These interactions suggest that oral contraceptives may act as co-carcinogens with transmissible agents, in line with previous suggestions by zur Hausen (31) regarding the human papillomaviruses. The one study that has been able to examine use of oral contraceptives according to HPV detection status (14) found no differential effect, but additional studies are needed which utilize m6re accurate HPV assay techniques.
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION Furthermore, several studies (26,32) have found an enhancement of the effects of oral contraceptives among women with multiple births. Although reasons for an excess risk associated with multiparity remain unresolved, it is possible that oral contraceptives may provide exogenous stimulus to potentially hazardous endogenous hormonal influences. ANALYTIC COMPLEXITIES IN ASSESSING RELATIONSHIPS Several articles have discussed the analytic complexities involved in assessing the relationship of oral contraceptives to risk of cervical neoplasia (5,6). Difficulties that arise in the conduct of studies in this area include potential confounding influences, selection biases, inappropriate control groups, and inclusion of users of other methods of contraception in comparison groups. Confounding Influences Of concern in evaluating the relationships of oral contraceptives to risk of cervical neoplasia is the extent to which associations reflect the influence of correlated risk factors. Although it has been widely presumed that oral contraceptive associations would be highly confounded by number of sexual partners, a stronger confounder in most studies has been the history of Pap smear screening, which is usually highly linked with oral contraceptive use. It is unclear whether this is because oral contraceptive users tend to avail themselves of more frequent screening or whether a Pap smear is considered necessary before oral contraceptives are prescribed. However, it is established that failure to consider the confounding influences of screening history can result in an underestimate of the effects of oral contraceptives. For example, in a study in five U.S. locations (13), interval since last Pap smear acted as a strong negative confounder, since women who failed to have regular Pap smears, and who were thus at high cervical cancer risk, had a low rate of exposure to oral contraceptives. Initially, oral contraceptive users had a relative risk of 0.8, but after appropriate adjustment, oral contraceptive users were found to have a relative risk of 1.5, rising to approximately 2fold among long-term users. Similar confounding was noted in a study conducted by the World Health Organization (30), where an adjusted risk of 1.2 was associated with any use, increasing to 1.5 for users of 5 or more years. Although most studies have been concerned with confounding influences of female sexual behavior on oral contraceptive associations, it is possible that male sexual behavior may also play an intervening role, especially given evidence of a "male factor" in the etiology of cervical cancer (33). Only two studies (14,28) have had information on male sexual behavior, and this was not an important source of confounding to observed oral contraceptive associations. However, further evaluation of this factor as a potential confounding influence is warranted, especially since one of these studies (28) relied on female subjects to provide information on their partner's sexual behavior.
666
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION Recent interest has focused on the role of HPV in the etiology of cervical neoplasia, and studies are now beginning to assess this as a potential modifier of oral contraceptive associations. A number of studies (34-36) have suggested a higher prevalence of HPV infection among oral contraceptive users, but whether this represents uncontrolled confounding by factors such as sexual behavior or an effect on the ability to detect the virus is yet to be clarified. Only one epidemiologic study to date has had the ability to adjust for measures of HPV in assessing oral contraceptive associations (14), and effects were not appreciably altered; however, additional studies are needed which utilize more refined measures of HPV infection status. Although numerous studies have shown that smokers, particularly long-term or high-intensity smokers, are at an excess risk of cervical neoplasia (37), the magnitude of the effect of smoking on risk or the correlation between smoking and oral contraceptive use have usually not been sufficiently strong to account for the elevation in risk among oral contraceptive users. Detection Bias A number of studies have concluded that excess risks associated with oral contraceptive use probably do not reflect the influence of confounding variables. However, an additional explanation for the continued elevation in risk is that women who have taken oral contraceptives might have been more likely to have their cervical cancer diagnosed because of increased Pap smear screening. Such a concern is more relevant to carcinoma-in-situ (CIS) than invasive disease, since the diagnosis of CIS is usually dependent on results of a Pap smear. For instance, an excess risk of CIS associated with oral contraceptive use in the study of Irwin et al. (21) was interpreted as being due to such a detection bias, since elevations in risk were only observed among recent users, who were those with more intensive Pap smear screening histories. In contrast, oral contraceptive use was associated with a decreased risk of invasive cancer in this same study, presumably because oral contraceptive users were more likely to have had their diseases detected at a preinvasive stage. Thus, detection bias may operate in the opposite direction for invasive disease, with an attenuation in the true relative risks, It is important therefore to assess the means by which invasive disease was detected. However, in the one investigation of invasive cancer that attempted to assess how oral contraceptive-associated risks were affected by the method of detection (30), exclusion from analysis of women who presented without the usual symptom of vaginal bleeding failed to produce substantial changes in the risk estimates, arguing against major effects of detection bias on observed oral contraceptive relationships. Aoorooriateness of Comoarison Croups A correct interpretation of effects of oral contraceptives depends on the types of women in the comparison groups utilized. This issue of comparability pertains to both the groups of non-diseased women as well as non-users of oral contraceptives.
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION In terms of the comparison group of non-diseased women, special attention must be given to whether these women are representative in terms of their contraceptive practices. Although many studies have utilized comparison groups consisting of women hospitalized for various conditions, questions arise regarding their appropriateness, especially since studies have shown that oral contraceptive users may be overrepresented among hospitalized controls (38). Although this problem may be alleviated by eliminating from the control group women currently hospitalized for certain conditions, e.g., other gynecologic conditions, caution is still warranted in interpretation of effects derived from studies using hospital control groups. When mixtures of diagnoses are included, it is recommended that associations be examined separately for the different types of controls utilized. It is further important that the control group consist of women potentially at risk of cervical cancer; thus those with a hysterectomy (including removal of the cervix) should be excluded. This elimination can result in substantially increased efforts to identify a suitable control group, since in the U.S. approximately one-third of women in the age range with a high risk for cervical cancer are now estimated to have undergone a hysterectomy. Of further concern is that users of oral contraceptives be compared appropriately with others at risk for contraception. Specific attention needs to be given to non-contracepting women, who may themselves be at an elevated risk of cervical cancer,(l9), as well as users of barrier methods of contraception (diaphragm and condom), who have been found in a number of studies to be at low cervical cancer risk (1,4,39,40). Several of the prospective studies have compared oral contraceptive users with users of the intrauterine device (IUD), since the IUD has been presumed not to exert a protective effect on risk. However, use of the IUD in relation to cervical cancer risk has not been well investigated, and further research might well indicate some alteration in risk, similar to the unsuspected reduction in risk recently associated with spermicidal agents (16,41). Although there probably is no ideal comparison group, it is important for studies to assess the risk associated with various methods of contraception in the control group, and to exclude, when necessary, women whose behavior may artificially lower or increase oral contraceptive-related risks. Several studies have performed these exclusions (13,14,19), and found that oral contraceptive relationships were not substantially altered, but studies should continue to examine potential effects associated with the comparison groups utilized. TRANSITION TIME STUDIES There is now evidence of a continuum of disease from cervical dysplasia to carcinoma in situ to invasive cancer. Although progression to more advanced states is well documented, factors that affect the natural history of these lesions remain unclarified. There has thus been interest in determining whether oral contraceptives are involved in this process. In one prospective investigation of 203 oral contraceptive users and 97 non-users (almost all IUD users) with cervical dysplasia, the probability of progression to carcinoma in situ after 7 years (estimated by life table methods) was 0.30 in oral
666
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION
contraceptive users and 0.05 in non-users, but these estimates were based on only 10 and 3 progressions, respectively (42). More recently, Bamford et al. (43) examined records of 105 women in whom cervical cytology revealed progression from mild or moderate dyskaryosis to severe dyskaryosis, along with records of 152 control women without progression. No effect of contraceptive method on the rate of progression was found. Given evidence that oral contraceptives appear unrelated to low grade squamous intraepithelial lesions but associated with some increase in risk of high grade lesions (36), further research on the effects of oral contraceptives on the potential for progression of lesions seems warranted. ASSOCIATION WITH CERVICAL ADENOCARCINONA The possibility of a relationship between adenocarcinoma of the cervix and oral contraceptives was raised when Dallenbach-Hellweg (44) described a series of 28 women with this disease, the majority of whom had had prolonged exposure to the pill. Subsequently, descriptive surveys noted a rise in the incidence of cervical adenocarcinoma among young women, and oral contraceptive use was suggested as a possible explanation (45-47). In two case-control studies, effects of oral contraceptives were stronger for adenocarcinoma than for squamous cell neoplasms (13,14), but in three other investigations no differential effect was observed (48-50). It is known that oral contraceptives cause a marked increase in the glandular-to-squamous ratio of cervical epithelium, and microglandular hyperplasia is now recognized as an oral contraceptive-related lesion. It has been proposed that oral contraceptives may cause endocervical adenocarcinomas by long-term stimulation of the proliferating glandular epithelium. However, pathologic studies cast some doubt on this hypothesis since microglandular hyperplasia has not been established as a precursor of adenocarcinoma (48,51,52). Although numerous case reports have been published of cervical adenocarcinomas occurring among oral contraceptive users (53-55), further research is needed to define the relationships of cervical adenocarcinomas to both preceding microglandular hyperplasia as well as oral contraceptive usage. BIOMGIC
PLAUSIBILITY
There are a number of lines of evidence that support a biologic role for oral contraceptives in the etiology of cervical neoplasia. Hormonal steroids are capable of promoting the development of cervical cancer in animals (56,57). In humans, cervical tissue has been found to have hormone receptor sites (58) and, as previously described, histologic changes in cervical epithelium have been demonstrated to result from the administration of contraceptive steroids. Furthermore, oral contraceptives have been associated with the production of an estral "clear channel" mucus that could facilitate entry of mutagens (59,60) and with immunologic alterations that could increase susceptibility to viral agents. Hormones have also been found to enhance the transcription of the HPV genome both in vitro (61) and in vivo (62). The possibility of an effect mediated through dietary factors has also been suggested in that oral contraceptive users, particularly those with cervical dysplasia, have been found to have low
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION serum and red blood cell folacin levels. In one small clinical trial, large doses of folate supplementation among oral contraceptive users with cervical dysplasia led to improved cervical biopsy scores (63). CONCLUSIONS
AND RJXONNENDATIONS
FOR FURTHER RESEARCH
Evidence is accumulating that oral contraceptives may play a role in the etiology of cervical neoplasia. Data from well-designed epidemiologic studies indicate that excess risks persist after adjustment for a variety of potential confounding factors. Although the evaluation of relationships is complex because of the possibility of biases, it does not appear in most studies as though the elevations in risk can be attributed to methodologic shortcomings. Thus, causal interpretations must be considered, particularly in view of accumulating data on dose-response relationships of risk with duration of use. Most of the data linking oral contraceptives to an elevated risk of cervical neoplasia derive from case-control studies, the results of which must be interpreted cautiously. Well-designed follow-up studies are critically needed to allow an assessment of how oral contraceptives affect the natural history of cervical lesions. It is essential that these studies collect detailed information on other cervical cancer In risk factors, including sexual behavior and screening histories. addition, given concerns that oral contraceptives may exacerbate the effects of infectious agents, these studies will need to accurately assess HPV infection status. Studies which attempt to address whether this interaction is mediated through an indirect effect, such as through altering the immune system, or a direct effect, e.g., by causing viruses to more effectively replicate, should also be pursued. Supporting an effect of oral contraceptives on cervical cancer risk is the recent observation of increasing incidence rates in the United Kingdom (8), although to date no such trend has been observed in the United States (64). As previously described, a recent increase in the rates of cervical adenocarcinoma has been noted, especially among younger women, strengthening support from some case-control studies of stronger oral contraceptive effects for these neoplasms than for squamous tumors. Further studies, however, are needed which specifically assess these relationships, adjusting for the appropriate risk factors, which may be different than those associated with the more common squamous cell cancers (49,65). Although there are now a number of studies which indicate an elevation in risk of cervical neoplasia with extended durations of oral contraceptive use, additional effects of timing of usage remain less well clarified. There is some evidence that oral contraceptives may operate at later stages in the carcinogenic process, but further studies are needed to assess detailed usage patterns. Given recent suggestions that the human papillomaviruses may be a necessary but not sufficient agent for cervical cancer (66), there is interest in With assessing the interactive effects of HPV and oral contraceptives. rapidly advancing HPV assay techniques (such as the polymerase chain reaction test, which minimizes measurement error associated with other
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION assays), clarification of the interplay of these factors should be greatly enhanced. REFERENCES
1.
Boyce JG, Lu T, Nelson JH, Fruchter RG. Oral contraceptives and cervical carcinoma. Am .IObstet Gynecol 1977;128:761-6.
2.
Melamed MR, Flehinger BJ. Early incidence rates of precancerous cervical lesions in women using contraceptives. Gynecol Oncol 1973;1:290-8.
3.
Sandmire HF, Austin SD, Bechtel RC. Carcinoma of the cervix in oral contraceptive steroid and IUD users and nonusers. Am J Obstet Gynecol 1976;125:339-45.
4.
Worth AJ, Boyes DA. A case-control study into the possible effects of birth control pills on pre-clinical carcinoma of the cervix. J Obstet Gynaecol Br Commonw 1972;79:673-9.
5.
Piper JM. Oral contraceptives and cervical cancer. 1985;22:1-14.
6.
Swan SH, Petitti DB. A review of problems of bias and confounding in epidemiologic studies of cervical neoplasia and oral contraceptive use. Am J Epidemiol 1982;115:10-8.
7.
Andolsek L, Kovacic J, Kozuh M, Litt B. Influence of oral contraceptives on the incidence of premalignant and malignant lesions of the cervix. Contraception 1983;28:505-19.
8.
Beral V, Hannaford P, Kay C. Oral contraceptive use and malignancies of the genital tract: Results from The Royal College of General Practitioners' Oral Contraception Study. Lancet 1988;2:1331-5.
9.
Peritz E, Ramcharan S, Frank J, Brown WL, Huang S, Ray R. The incidence of cervical cancer and duration of oral contraceptive use. Am J Epidemiol 1977;106:462-9.
Gynecol Oncol
10.
Vessey MP, Lawless M, McPherson K, Yeates D. Neoplasia of the cervix uteri and contraception: A possible adverse effect of the pill. Lancet 1983;2:930-4.
11.
Vessey MP, Lawless M, McPherson K, Yeates D. Oral contraceptives and cervical cancer (letter). Lancet 1983;2:1358-9.
12.
Swan SH, Brown WL. Oral contraceptive use, sexual activity, and cervical carcinoma. Am J Obstet Gynecol 1981;139:52-7.
13.
Brinton LA, Huggins GR, Lehman HF, et al. Long-term use of oral contraceptives and risk of invasive cervical cancer. Int J Cancer 1986;38:339-44.
JUNE 1991 VOL. 43 NO. 6
591
CONTRACEPTION 14.
Brinton LA, Reeves WC, Brenes and risk of invasive cervical 4-11.
MM, et al. Oral contraceptive use cancer. Int J Epidemiol 1990;19:
15.
Brock KE, Berry G, Brinton LA, et al. Sexual, reproductive and contraceptive risk factors for carcinoma-in-situ of the uterine cervix in Sydney. Med J Austral 1989;150:125-30.
16.
Celentano DD, Klassen AC, Weisman CS, Rosenshein NB. The role contraceptive use in cervical cancer: The Maryland cervical cancer case-control study. Am J Epidemiol 1987;126:592-604.
17.
Cuzick cervix
18.
Ebeling K, Nischan P, Schindler CH. and risk of invasive cervical cancer Int J Cancer 1987;39:427-30.
19.
Harris RWC, Brinton LA, Cowdell RH, et al. Characteristics of women with dysplasia or carcinoma in situ of the cervix uteri. J Cancer 1980;42:359-69.
J, De Stavola B, McCance D, et al. A case-control cancer in Singapore. Br J Cancer 1989;60:238-43.
of
study
of
Use of oral contraceptives in previously screened women.
Br
20.
Hellberg D, Valentin J, Nilsson S. Long-term use of oral contraceptives and cervical neoplasia: An association confounded by other risk factors? Contraception 1985;32:337-46.
21.
Oral contraceptives Irwin KL, Rosero-Bixby L, Oberle MW, et al. and cervical cancer risk in Costa Rica. Detection bias or causal association? JAMA 1988;259:59-64.
22.
Oral La Vecchia C, Decarli A, Fasoli M, et al. cancers of the breast and of the female genital results from a case-control study. Br J Cancer
23.
Mandelson MT, Daling JR, White E, Chu J, McKnight B. Further evidence that duration and recency of oral contraceptive use are associated with invasive cervical cancer (abstract). Am J Epidemiol 1990;132:77B.
24.
Molina R, Thomas DB, Dabancens A, et al. Oral contraceptives and Cancer Res 1988;48:1011-5. cervical carcinoma in situ in Chile.
25.
Ory HW, Conger SB, Naib 2, Tyler CW Jr, Hatcher RA. Preliminary analysis of oral contraceptive use and risk of developing premalignant lesions of the uterine cervix. In: Garattini S, Berendes HW, eds. Pharmacology of Steroid Contraceptive Drugs. New York: Raven Press, 1977: 211-8.
26.
Parazzini F, La Vecchia C, Negri E, Maggi R. Oral use and invasive cervical cancer. Int J Epidemiol
662
contraceptives and Interim tract. 1986;54:311-7.
contraceptive 1990;19:259-63.
JUNE 1991 VOL. 43 NO. 6
27.
Peters RK, Thomas D, Hagan DG, Mack TM, Henderson BE. Risk factors for invasive cervical cancer among latinas and non-latinas in Los Angeles County. J Nat1 Cancer Inst 1986;77:1063-77.
28.
Slattery ML, Overall JC Jr, Abbott TM, French TK, Robison IM, Gardner J. Sexual activity, contraception, genital infections, and cervical cancer: Support for a sexually transmitted disease hypothesis. Am J Epidemiol 1989;130:248-58.
29.
Reeves WC, Brinton LA, Brenes MM, et al. Case-control study of cervical cancer in Herrera Province, Republic of Panama. Int J Cancer 1985:36:55-60.
30.
World Health Organization Collaborative Study of Neoplasia and Steroid Contraceptives. Invasive cervical cancer and combined oral contraceptives. Br Med J 1985;290:961-5.
31.
zur Hausen H. Human genital cancer: Synergism between two virus infections or synergism between a virus infection and initiating events? Lancet 1982:2:1370-2.
32.
Brinton LA, Reeves WC, Brenes MM, et al. Parity as a risk factor for cervical cancer. Am J Epidemiol 1990;130:486-96.
33.
Brinton LA, Reeves WC, Brenes MM, et al. The male factor in the etiology of cervical cancer among sexually monogamous women. Int J Cancer 1989;44:199-203.
34.
Hildesheim A, Reeves WC, Brinton LA, et al. Association of oral contraceptive use and human papillomaviruses in invasive cervical cancers. Int J Cancer 1990;45:860-4.
35.
Lorincz AT, Schiffman MH, Jaffurs WJ, Marlow J, Temple GF. Temporal associations of human papillomavirus infection with cervical cytologic abnormalities. Am J Obstet Gynecol 1990;162: 645-51.
36.
Negrini BP, Schiffman MH, Kurman RJ, et al: Oral contraceptive use, human papillomavirus infection, and risk of early cytological abnormalities of the cervix. Cancer Res 1990;50:4670-5.
37.
Winkelstein W Jr. Smoking and cervical cancer--Current status: review. Am J Epidemiol 1990;131:945-57.
38.
Hoover R, Bain C, Cole P, MacMahon B. Oral contraceptive use: Association with frequency of hospitalization and chronic disease risk indicators. Am J Public Health 1978:68:335-42.
39.
Melamed MR, Koss LG, Flehinger BJ, Kelisky RP, Dubrow H. Prevalence rates of uterine cervical carcinoma in situ for women using the diaphragm or contraceptive oral steroids. Br Med J 1969;3:195-200.
JUNE 1991 VOL. 43 NO. 6
A
40.
Wright NH, Vessey MP, Kenward B, McPherson K, Doll R. Neoplasia and dysplasia of the cervix uteri and contraception: A possible protective effect of the diaphragm. Br J Cancer 1978;38:273-9.
41.
Hildesheim A, Brinton LA, Mallin K, et al. Barrier and spermicidal contraceptive methods and risk of invasive cervical cancer. Epidemiology 1990;1:266-72.
42.
Stern E, Forsythe AB, Youkeles L, Coffelt CF. Steroid contraceptive use and cervical dysplasia: Increased risk of progression. Science 1977;196:1460-2.
43.
Bamford PN, Forbes-Smith PA, Rose GL, et al. An analysis of factors responsible for progression or regression of mild and moderate cervical dyskaryosis. Br J Fam Plan 1985;11:5-8.
44.
Dallenbach-Hellweg G: On the origin and histological structure of adenocarcinoma in women under 50 years of age. Path01 Res Pratt 1984;179:38-50.
45.
Chilvers C, Mant D, Pike MC. Cervical adenocarcinoma and oral contraceptives. Br Med J 1987;295:1446-7.
46.
Peters RK, Chao A, Mack TM, Thomas D, Bernstein L, Henderson BE. Increased frequency of adenocarcinoma of the uterine cervix in young women in Los Angeles County. J Nat1 Cancer Inst 1986;76:423-a.
47.
Schwartz SM, Weiss NS. Increased incidence of adenocarcinoma of the cervix in young women in the United States. Am J Epidemiol 1986;124:1045-7.
48.
Jones MW, Silverberg SG. Cervical adenocarcinoma in young women: Possible relationship to microglandular hyperplasia and use of oral contraceptives. Obstet Gynecol 1989;73:984-9.
49.
Parazzini F, La Vecchia C, Negri E, Fasoli M, Cecchetti G. Risk factors for adenocarcinoma of the cervix: A case-control study. Br J Cancer 1988;57:201-4.
50.
Persson E, Einhorn N, Pettersson F. A case-control study of oral contraceptive use in women with adenocarcinoma of the uterine cervix. Eur J Obstet Gynecol Reprod Biol 1987;26:85-90.
51.
Tase T, Okagaki T, Clark BA, Twiggs LB, Ostrow RS, Faras AJ. Human papilloma virus DNA in glandular dysplasia and microglandular hyperplasia: Presumed precursors of adnocarcinoma of the uterine cervix. Obstet Gynecol 1989;73:1005-8.
52.
Valente PT, Hanjani P. Endocervical neoplasia in long-term users of oral contraceptives: Clinical and pathologic observations. Obstet Gynecol 1986;67:695-704.
JUNE 1991 VOL. 43 NO. 6
CONTRACEPTION
53.
Ireland D, Hardiman P, Monaghan JM. Adenocarcinoma of the uterine cervix: A study of 73 cases. Obstet Gynecol 1985;65:82-5.
54.
Shingleton HM, Gore H, Bradley B, Soong SJ. Adenocarcinoma of the cervix. Am J Obstet Gynecol 1981;139:799-814.
55.
Tamini HK, Figge DC. Adenocarcinoma of the uterine cervix. Gynecol Oncol 1982;13:335-44.
56.
Dunn TB. Cancer of the uterine cervix in mice fed a liquid diet containing an antifertility drug. J Nat1 Cancer Inst 1969;43:681-92.
57.
Kaminetzky HA. Metylcholanthrene-induced cervical dysplasia and the sex steroids. Obstet Gynecol 1966;27:489-93.
58.
Ford LC, Berek JS, Lagasse LD, Hacker NF, Heins YL, Delange RJ. Estrogen and progesterone receptor sites in malignancies of the uterine cervix, vagina, and vulva. Gynec Oncol 1983;15:27-31.
59.
Cramer DW: Uterine cervix. In: Schottenfeld D, Fraumeni JR, eds. Cancer Epidemiology and Prevention. Philadelphia: WB Saunders Co, 1982:881-900.
60.
Jordan JA, Singer A. Co, 1976.
61.
Pater MM, Hughes GA, Hyslop DE, Nakshatri H, Pater A. Glucocorticoid-dependent oncogenic transformation by type 16 but not type 11 human papilloma virus DNA. Nature (Lond.) 1988;335:832-5.
62.
Gloss B, Bernard HU, Seedorf K, Klock G. The upstream regulatory region of the human-papilloma-virus-16 contains an EZ-proteinindependent enhancer which is specific for cervical carcinoma cells and regulated by glucocorticoid hormones. EMBO J
The Cervix.
Philadelphia:
W.B. Saunders
1987;6:3735-43.
63.
Butterworth CE, Hatch KD, Gore H, Mueller H, Krumdieck CL. Improvement in cervical dysplasia associated with folic acid therapy in users of oral contraceptives. Am J Clin Nutr 1982;35:73-82.
64.
Devesa SS, Young JL, Brinton LA, Fraumeni JF. Recent trends in cervix uteri cancer. Cancer 1989;64:2184-90.
65.
Brinton LA, Tashima KT, Lehman HF, et al. Epidemiology of cervical cancer by cell type. Cancer Res 1987;47:1706-11.
66.
zur Hausen H. Papillomaviruses in anogenital cancer as a model to understand the role of viruses in human cancers. Cancer Res 1989;49:4677-81.
JUNE 1991 VOL. 43 NO. 6
595
