Reproductive Factors, Oral Contraceptive Use, and Human Papillomavirus Infection: Pooled Analysis of the IARC HPV Prevalence Surveys Free

Genital human papillomavirus (HPV) is a sexually transmitted infection (1, 2) that is thought to be a necessary cause of cervical cancer (3). However, in addition to HPV infection, other factors exist that influence the risk of developing cervical cancer (4). In particular, high parity (5, 6), early age at first full-term pregnancy (FTP; ref. 6), and long-term oral contraceptive (OC) use (7, 8) have been found to increase the risk of cervical cancer and its precursor lesions.

Interestingly, these associations have also been confirmed in analyses restricted to HPV-positive women (5-11). However, the interpretation of the role of reproductive characteristics and OC use in cervical carcinogenesis requires a better understanding of their possible influence on the acquisition and duration of HPV infection.

Previous studies on HPV prevalence and age at first FTP (9, 10) or parity (9, 10, 12-17) have been scarce, and suggested weak direct associations with HPV infection. Studies on the influence of OC use on HPV infection showed inconsistent results, with modest associations in both directions (10, 12, 13, 18-22). Very little is known on the relationship of age at menarche (10, 17, 23, 24) or menopause (25) with HPV infection or cervical cancer.

To further elucidate these issues, we evaluated the association between reproductive factors and OC use and HPV prevalence among over 15,000 women from the IARC HPV Prevalence Surveys.

Study protocols and questionnaires were developed for each of 14 areas in 12 different countries worldwide, and studies were carried out between 1993 and 2005. Population sampling methods have been previously described for the individual areas: Vietnam (26), Thailand (27), Korea (28), India (29), China (30), Spain (31), Italy (32), Mexico (33), Argentina (34), Chile (35), Colombia (36), and Nigeria (37). Vietnam and Thailand participated with two studies each in different areas (Hanoi and Ho Chi Minh in Vietnam; Lampang and Songkla in Thailand). In Colombia, the study population was selected from women attending screening centers and family planning clinics, and in Italy, participants were invited within the framework of an organized cervical screening program.

In summary, each area attempted to obtain a random age-stratified sample of the population that included at least 100 women in each 5-year age group, from 15-19 to 65 and over. Participation ranged from 48% in Songkla, Thailand (where most nonparticipants were not found at the address given by the population list), to 94% in Hanoi, Vietnam. Exclusion criteria were pregnancy at the time of recruitment, previous hysterectomy or cervical conization, and physical or mental incompetence. The number of individuals in this analysis sometimes differs from those reported in the original findings due to different selection criteria or to additional HPV testing after publication.

Trained interviewers questioned study participants face-to-face using a questionnaire that included information on sociodemographic characteristics, smoking habits, Papanicolaou smear screening history, and sexual behavior characteristics such as lifetime number of sexual partners and age at sexual debut.

Information on reproductive history was collected and included parity and age at first FTP. Women were also asked questions on life-long use of OCs, and on menstrual factors, such as age at menarche and age at menopause. All participants signed informed consent forms according to the recommendations of the IARC and the local ethical review committees that approved the study.

Study participants underwent a pelvic examination done by a gynecologist or specially trained personnel. Samples of exfoliated cells from the ectocervix were collected with two wooden Ayre spatulas, and from the endocervix with a cytobrush (Cervibrush, CellPath, Herte, United Kingdom). After the preparation of a Papanicolaou smear, the remaining exfoliated cervical cells were placed in tubes with phosphate buffered saline (PBS) and stored on ice. Cells were centrifuged at 3,000 × g and the resulting pellets were resuspended in PBS and frozen between −20°C and −80°C until they were shipped to IARC for storage. In Italy and India, cervical cells were stored in CytoRich (Tripath Imaging, Burlington, NC) and preservCyt (Cytyc Corporation, Boxborough, MA) media, respectively, and liquid-based cytology was done. Papanicolaou or liquid-based cytology smears were stained and read locally and classified according to the Bethesda (38) or equivalent system. Atypical squamous cells of undetermined significance or worse were found in 4% of the women in the study (range, from 1% in Hanoi, Vietnam, to 9% in Nigeria). One hundred and twenty-four women (0.8%) had a diagnosis of high-grade squamous intraepithelial lesion or worse.

HPV testing was done on exfoliated cervical cells in the Department of Pathology of the VU Medical Center, Amsterdam, the Netherlands, with the exception of the Mexican study. Only women who tested positive for β-globin were included in this analysis.

A first screening was done to determine the overall presence of HPV DNA using a general GP5+/6+ primer mediated PCR (39). PCR products were assessed by enzyme immunoassay using oligoprobe cocktails to detect the following 36 HPV types: HPV6, 11, 16, 18, 26, 31, 33 to 35, 39, 40, 42 to 45, 51, 52 to 59, 61, 66, 68, 70, 71 (equivalent to CP8061), 72, 73 (equivalent to MM9), 81 (equivalent to CP8304), 82 (IS39 and MM4 subtypes), 83 (equivalent to MM7), 84 (equivalent to MM8), and CP6108 (39). In addition, PCR products were tested using a low-stringency Southern blot analysis of PCR products with a cocktail probe of HPV-specific DNA fragments. Subsequently, typing of samples positive for HPV was done by enzyme immunoassay or reverse line blot analysis of GP5+/6+ PCR product using HPV type–specific oligoprobes for the HPV types described above (39, 40). Samples that were GP5+/6+-positive by low-stringent Southern blot analyses, but were not identified by the abovementioned typing protocols, were considered as HPVX, i.e., uncharacterized HPV types. For India (29) and Chile (35), the oligoprobe cocktail was extended to include HPV types 30, 32, 64, 67, 69, cand85, 86, and JC9710.

The determination of HPV types in the Mexican study has been previously described (33); briefly, it was carried out on specimens from women with normal cytology at the Department of Molecular Microbiology and Immunology, Johns Hopkins University School of Public Health (Baltimore, MD) and at the National Institute of Public Health (Cuernavaca, Morelos, Mexico), using biotinylated MY09/11 consensus primers and genotyping (27 HPV types: 6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51-59, 66, 68, 73, and 82-84) by a single-hybridization, reverse line blot detection method (41). Good comparability of findings from GP5+/GP6+ and MY09/11-based PCR has been reported (42).

HPV types considered as high-risk for this report included HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73, and 82 (43). All other HPV types, including HPV 26, 53, and 66 that were considered as probably carcinogenic (43), were classified as low-risk types.

The variable “parity” was defined as the number of FTPs, and hence, the category “nulliparous” included women who had had spontaneous (9% of nulliparous women) or voluntary (3% of nulliparous women) abortions. The variable “OC use” was evaluated according to the user status at interview, i.e., “current”, “former” and “never” user, and according to number of years of use. On account of differences in the questionnaire, Italy could not contribute to the analysis on OC use. For the study areas that could not directly provide the variable “age at menopause”, we were able to derive it, with the exception of Mexico, Nigeria, and India, using the date or the age at the last menstruation and age at interview. Information on age at menarche was not available in Mexico and Italy.

Unconditional logistic regression was used to estimate odds ratios (OR) of being HPV-positive and the corresponding 95% confidence intervals (CI) according to different reproductive and menstrual characteristics and OC use. All ORs were adjusted for age group (<25, 25-34, 35-44, 45-54, and ≥55 years), lifetime number of sexual partners (1, ≥2), and when appropriate, study area. As no information on lifetime number of sexual partners was available in India and Italy, these areas were excluded from the computation of ORs adjusted for this variable.

The variables were analyzed categorically, and when more than two groups were compared, floating CIs (FCI) were estimated by treating ORs as floating absolute risks (44, 45). This method assigns a variance to the reference category and reduces unwanted correlation between coefficients, thus reducing the variance of ORs not defined as 1.0. No change is made, however, to the point estimate of the ORs. Floating methods enable valid comparison between any two exposure groups, even if neither is the baseline group. Tests for linear trend of ORs were done, giving an increasing score for each level of the categorized variable and fitting them into the model as continuous variables.

Most results are presented graphically, plotting the summary ORs as black squares whose size is inversely proportional to the variance of the estimate. A horizontal line represents the 95% CI. Diamonds are used to plot the summary OR for all studies together. The diamonds represent the pooled OR and the extremes show the limits of the 95% CI.

The heterogeneity of the OR between areas was tested by calculating the difference between the log-likelihood of the model that included the interaction term between the areas and exposure of interest and the log-likelihood of the model that included the exposure only, and comparing it to the χ² distribution with degrees of freedom equal to the number of areas minus one.

A total of 15,145 women with valid HPV results were included in the present analysis (Table 1). Overall, HPV prevalence was 13.0%, and ranged between 1.7% in Hanoi, Vietnam and 27.0% in Nigeria. The mean age at first FTP was 22.2 years. The range of the number of FTPs was broad (1-17), with a mean of 3.0. The highest mean of FTPs was found in Nigeria (4.4) and Mexico (4.0). The percentage of ever OC users varied from 1.1% in India to 57.6% in Lampang, Thailand. The mean age at menarche and menopause were, respectively, 14.2 and 47.3 years (Table 1).

Among parous women, those who had the first FTP at the age of 25 years or later showed a nonsignificantly decreased risk for HPV positivity (OR, 0.88; 95% FCI, 0.77-1.02) as compared with women who had the first FTP before the age of 20 years (OR, 1.00; 95% FCI, 0.87-1.15). None of the ORs relative to categories of age at first FTP reached the level of statistical significance (Fig. 1A).

With respect to parity, women with five or more FTPs (OR, 0.90; 95% FCI, 0.76-1.06) had a similar risk of being HPV-positive compared with women with only one FTP (OR, 1.00; 95% FCI, 0.86-1.16). The test for linear trend restricted to parous women was not significant (Fig. 1B). However, nulliparous women presented an OR of 1.40 (95% CI, 1.16-1.69) compared with all parous women after adjustment for age, lifetime number of sexual partners, and study area. Although decreased risks for nulliparity were found in two areas (Lampang, Thailand and China), the test for between-area heterogeneity did not reach the level of statistical significance (Fig. 2).

ORs for nulliparity and number of FTP were not substantially modified by additional adjustment for age at first FTP and age at first sexual intercourse (data not shown). History of voluntary (OR, 1.17; 95% CI, 0.94-1.44) or spontaneous (OR, 0.99; 95% CI, 0.86-1.14) abortion was not associated with HPV positivity after adjustment for lifetime number of sexual partners and FTPs.

Current (OR, 1.02; 95% FCI, 0.78-1.34) and former (OR, 1.07; 95% FCI, 0.91-1.27) use of OCs did not affect the probability of being HPV-positive (OR for never OC use = 1.00; 95% FCI, 0.88-1.14; Fig. 3A). Likewise, duration of OC use was not associated with HPV positivity: the risk of being HPV-positive was similar for women who reported to have used OCs for 10 years or more (OR, 1.16; 95% FCI, 0.85-1.58) and never users of OCs (OR, 1.00; 95% FCI, 0.90-1.12; Fig. 3B). No significant differences in HPV positivity were observed between ever (OR, 1.08; 95% CI, 0.94-1.25) and never users of OC, after adjustment for age, lifetime number of sexual partners, and study area, and no evidence of heterogeneity between study areas was found (Fig. 4).

Age at menarche (Fig. 5A) and age at menopause (Fig. 5B) did not have an influence on HPV positivity. The OR for menarche before the age of 13 years (0.91; 95% FCI, 0.79-1.05) was not significantly different from that for menarche at 15 years or later (1.00; 95% FCI, 0.88-1.14). Entering menopause at the age of 50 years or later (OR, 0.80; 95% FCI, 0.63-1.02) did not substantially modify the risk of being infected with HPV as compared with entering menopause before 45 years (OR, 1.00; 95% FCI, 0.72-1.40). None of our present findings were substantially modified when high-risk and low-risk types were considered separately (data not shown).

The present pooled analyses of the IARC HPV Prevalence Surveys shows no evidence of an association of the risk of being infected with HPV with reproductive and menstrual factors and OC use. The large size and the broad range of areas involved, together with the use of similar questionnaires and highly sensitive methods of HPV detection, allowed a more precise evaluation than previously reported of the relationship between these factors and HPV positivity. These negative findings contrast with the results of an analysis of sexual behavior characteristics in the same populations showing a clear association between the number of sexual partners and HPV positivity (2).

More than 89% (data not shown) of the women we interviewed had at least one FTP, and parity could be evaluated across a wide range of values. Although remarkable differences existed between study areas, the overall mean number of FTPs was three, and women with five or more FTPs represented approximately 17% of the total. Indeed, nulliparous women were found to be at a significantly higher risk of being HPV-positive than all parous women combined, following adjustment for age and lifetime number of sexual partners. However, positivity for HPV infection did not substantially vary across categories of parity and the linear trend test restricted to parous women was not significant.

In addition, among parous women, the prevalence of HPV infection was similar across different categories of age at first FTP, in agreement with previous studies that were able to control for the confounding effect of sexual behavior (9, 10).

A similar risk pattern, i.e., an excess of risk among nulliparous women and no trend in risk according to an increasing number of FTPs, was previously seen among control women from a pooled case-control study of cervical cancer (9). Nulliparity could be a marker of high-risk sexual behavior, at least in some areas. Nulliparous women were indeed more sexually active (mean lifetime number of sexual partners = 1.6 versus 1.3), but also younger (mean age = 32.2 versus 42.0) than parous women. Therefore, despite the careful adjustment for age and lifetime number of sexual partners, some residual confounding effect of these variables on our findings cannot be ruled out.

Our study is the largest study ever done on OC use and HPV positivity, and involves areas with substantial differences in the prevalence of both factors. Current, former, and never users of OC had similar HPV prevalence, after controlling for age, lifetime number of sexual partners, and study area. The lack of an association between OC use and HPV positivity was consistent across study areas. Also with regard to duration of OC use, we did not observe a significant association with HPV positivity for any of the categories of years of use, even for 10-year use or more. The findings of our present study do not support the hypothesis that OC users may acquire HPV more often or may be more prone to long-duration infection than nonusers.

Previous findings on the effects of OC use on HPV infection are rather inconsistent. Green et al. (18) reviewed studies published up to 2003 on the topic (including findings from five of the study areas in our present pooled analysis; refs. 26, 28, 33, 36) and concluded that there was no evidence of an association between OC use and HPV positivity. However, only few of the studies included in that review (18) could allow for the possible confounding effect of sexual activity or provided information on long-term OC use. Among the studies not included by Green et al., two large studies (12, 19) showed a direct association, one showed an inverse association (20), and several others (10, 13, 21, 22, 46, 47) showed no relationship between HPV positivity and OC use.

OC use was positively associated with having multiple sexual partners, which was, in our present data, the strongest determinant of HPV prevalence. The relationship between OC use and HPV infection, however, did not substantially differ across strata of lifetime number of sexual partners (data not shown).

We did not find any significant association between HPV positivity and age at menarche. Although a previous report (23) suggested an inverse association between age at menarche and HPV infection, our findings are in agreement with the majority of the studies that did not report significant associations (10, 17, 24).

Almost no information is available on age at menopause and HPV infection (25). In our study, ∼20% of the women had already reached menopause, and among them, we could not detect a significant association between age at menopause and HPV positivity. Also, no significant association emerged between years since menopause and prevalence of overall HPV positivity, high-risk, and low-risk types, and the major HPV species (overall HPV, OR for ≥10 years since menopause versus <5 years = 1.2; 95% CI, 0.8-2.0; high-risk types, OR for ≥10 years since menopause versus <5 years = 1.2; 95% CI, 0.7-2.1; α9, OR for ≥10 years since menopause versus <5 years = 1.0; 95% CI, 0.5-2.0), suggesting that the cessation of menstrual cycling has no clear effect on the presence or detectability of viral infection.

One limitation of our study is that, as a cross-sectional study, it only provides information on HPV prevalence, not allowing the distinction between the determinants of acquisition and persistence of HPV infection. In addition, our study did not provide information on HPV viral load, and therefore, it cannot evaluate whether any of the considered characteristics affected viral replication, thereby increasing viral load. OC use and parity were not reported to be associated with viral load among HPV-positive women (48). Finally, our questionnaire did not allow the distinction between OC with combined estrogen and progestagen, or progestagen only. However, it is probable that the majority of OCs used by our study populations contained both estrogen and progestagen (49).

Our findings on the lack of association between age at first FTP, parity, and OC use with HPV positivity are of importance mainly on account of the consistent association reported between these variables and cervical cancer risk (5, 8-11). In a pooled analysis of 25 studies of cervical cancer (6), the relative risk of invasive cervical cancer by an increase in one FTP was 1.07 (95% CI, 1.05-1.09) after adjustment for age, study center, age at first sexual intercourse, lifetime number of sexual partners, and age at first FTP. The corresponding relative risk by 1-year decrease in age at first FTP was 1.05 (95% CI, 1.03-1.08). A meta-analysis of 28 studies on hormonal contraceptives and cervical cancer (7) showed an increased relative risk of cervical cancer of 2.2 (95% CI, 1.9-2.4) for long-term (≥10 years) OC users. The results of these studies were confirmed when the analyses were restricted to HPV-positive women.

During pregnancy, women are exposed to continuously elevated estrogen levels, and the synthetic estrogens found in the majority of OC formulations have increased estrogenic activity and enhanced bioavailability compared with endogenous estrogens (50).

In agreement with epidemiologic studies, findings from HPV16 transgenic mice (51) and in vitro models (52) also suggest that high estrogen, although not progesterone, levels are involved in the regulation of early viral promoters and malignant transformation of HPV-infected cells.

In conclusion, on account of the central role of HPV infection in cervical carcinogenesis, any risk factor other than HPV may play a role by either increasing the risk of acquisition or duration of the infection, or by increasing the risk of progression from HPV infection to cervical cancer. Our large and representative study does not support the possibility that high parity, early age at first FTP, and long-term OC use substantially influence the acquisition or persistence of an HPV infection and rather suggests that these factors may enhance the probability of malignant transformation of HPV infection.