Predictors of human papillomavirus infection in women undergoing routine cervical cancer screening in Spain: the CLEOPATRE study

Background Human papillomavirus (HPV) is a sexually transmitted infection that may lead to development of precancerous and cancerous lesions of the cervix. The aim of the current study was to investigate socio-demographic, lifestyle, and medical factors for potential associations with cervical HPV infection in women undergoing cervical cancer screening in Spain. Methods The CLEOPATRE Spain study enrolled 3 261 women aged 18–65 years attending cervical cancer screening across the 17 Autonomous Communities. Liquid-based cervical samples underwent cytological examination and HPV testing. HPV positivity was determined using the Hybrid Capture II assay, and HPV genotyping was conducted using the INNO-LiPA HPV Genotyping Extra assay. Multivariate logistic regression was used to identify putative risk factors for HPV infection. Results A lifetime number of two or more sexual partners, young age (18–25 years), a history of genital warts, and unmarried status were the strongest independent risk factors for HPV infection of any type. Living in an urban community, country of birth other than Spain, low level of education, and current smoking status were also independent risk factors for HPV infection. A weak inverse association between condom use and HPV infection was observed. Unlike monogamous women, women with two or more lifetime sexual partners showed a lower risk of infection if their current partner was circumcised (P for interaction, 0.005) and a higher risk of infection if they were current smokers (P for interaction, 0.01). Conclusion This is the first large-scale, country-wide study exploring risk factors for cervical HPV infection in Spain. The data strongly indicate that variables related to sexual behavior are the main risk factors for HPV infection. In addition, in non-monogamous women, circumcision of the partner is associated with a reduced risk and smoking with an increased risk of HPV infection.


Background
Human papillomavirus (HPV) is one of the most common sexually transmitted infections worldwide. It is well established that HPV is the necessary cause of cervical cancer and its precancerous lesions. In addition, HPV has been associated with a wide range of other diseases including: genital warts; vulvar, vaginal, penile and anal cancers; head and neck cancers; and recurrent respiratory papillomatosis [1,2]. Most sexually active women will be infected with HPV during their lifetime [3]. While the majority of HPV infections are cleared within 2 years, persistent infections and the presence of risk factors that promote persistence may lead to the development of precancerous and cancerous lesions in a small fraction of women. HPV genotypes are classified as high-risk (HR) or low-risk (LR) based on their association with cervical cancer, HPV 16 being the type most commonly identified in this cancer [4].
Epidemiological studies conducted among women over the past decade have shown consistently that the prevalence of cervical HPV infection, particularly HR HPV types, is greatest among women below the age of 25-30 years and lower in older women [5,6]. The increased infection rates among young women are attributable to higher exposure to HPV through multiple sexual partners. In contrast, older women tend to have fewer new partners and they may have developed some degree of immunity against HPV [7]. Other potential risk factors for cervical HPV infection include tobacco smoking [8], nulliparity [9], use of oral contraceptives [9], and recent new sexual partners [10]. Use of condoms may protect against cervical HPV infection [11].
Data on cervical HPV prevalence and determinants of HPV infection are sparse and limited in Spain. In a previous analysis of the CLEOPATRE Spain study, we reported the prevalence and type-specific distribution of cervical HPV infection among women aged 18-65 years attending cervical cancer screening in Spain [12]. In the current analysis we report on the association between socio-demographic, lifestyle and medical factors and the prevalence of cervical HPV infection. Only one previous epidemiological study has been carried out in the general female population of Spain to assess these risk factors [13]. Thus, the current analysis represents the first largescale, country-wide epidemiological study exploring risk factors for cervical HPV infection in Spain.

Study participants and procedures
The CLEOPATRE Spain study was a cross-sectional study conducted between June 2007 and May 2008 in 77 centers across the 17 Autonomous Communities in Spain. The study methods have been described previously in detail [12]. Briefly, women aged 18-65 years attending routine cervical cancer screening were recruited, excluding virgins and women who had a history of HPV-related cervical disease or HPV vaccination. All study participants provided written informed consent. Ethical clearance was obtained from the ethical committee of the Hospital Clínico San Carlos (Madrid, Spain).

Risk factor analysis
The evaluation of potential risk factors included age, type of community (rural, suburban or urban based on population size of municipality), country of birth, marital status, level of education, professional status, smoking status (including duration and intensity), pregnancies, contraceptive methods, lifetime number of sexual partners, age at first sexual intercourse, time since first sexual intercourse, history of cervical lesions, history of sexually transmitted infections (STIs), current partner's history of STIs in the last 12 months, current partner's circumcision status, and immune status.
The association between cervical HPV infection and potential risk factors was tested by multivariate analysis, which involved construction of an adjusted logistic regression model using stepwise regression with forward selection. Beginning with a basic model adjusted for age and Autonomous Community, the variables for the potential risk factors were added one at a time into the model to evaluate their association with cervical HPV infection, and those that showed a statistically significant association (P-value <0.10) were retained in the final model as adjusting variables.
Adjusted models were also constructed to evaluate the association between potential risk factors and HR HPV, LR HPV, HPV 16, and single and multiple HPV infections. Analyses were conducted to explore multiplicative interactions between lifetime number of sexual partners and other potential risk factors for cervical HPV infection. In addition, an adjusted model was constructed to assess risk factors for HPV positivity among young women aged 18-25 years.
Analyses were performed using the R programing language (R Development Core Team, 2005, http:// www.r-project.org).

Study participants
A total of 7 252 women were initially registered in the recruitment logs at the participating sites. Of those, 4 206 were eligible and 3 261 were finally enrolled in the study. The participation rate was 77.5% (3 261/4 206). Reasons for non-participation among eligible women were: refusal (13.2%), and other reasons such as logistic problems or menstruation (9.3%). The characteristics of study participants have been described previously [12]. In brief, 91.3% of the women were born in Spain, 77.2% lived in an urban area (≥10 000 inhabitants), 39.7% were single, 33.0% had a university/college education, 25.5% were students, 35.6% were current smokers, and 19.5% reported having had four or more lifetime sexual partners. There was a planned oversampling of young women aged ≤25 years (49.6%) compared with the proportion of this age group in the overall female population of Spain (18.0%).

Risk factor analysis Risk factors for any HPV infection
In the basic model adjusted for age and Autonomous Community, the following characteristics were positively and statistically significantly associated with cervical HPV infection: young age; living in an urban community; country of birth other than Spain; unmarried status; current smoker; more than one lifetime sexual partner; young age at first sexual intercourse (with a statistically significant trend); more years elapsed since first intercourse (with a statistically significant trend); and a history of cervical lesions, genital warts or other STIs (Table 1). There was also a weak and statistically borderline inverse association between systematic use of condoms in the last 12 months, meaning use of condom in 90-100% of all sexual relationships that occurred in the last year, and cervical HPV infection. No associations were identified between cervical HPV infection and number of pregnancies, use of oral contraceptives, ever use of condoms or a current sexual partner who was circumcised or with STIs in the last 12 months. Other variables, including type of medical practice and setting, professional status, duration and intensity of smoking, use of an intrauterine device, and immune status, were also assessed but no associations with HPV infection were observed (data not shown).
In the final adjusted model, variables that remained statistically significant were: age 18-25 years; living in an urban community; country of birth other than Spain; unmarried status; lower level of educational attainment (with a statistically significant trend); current smoker; more than one lifetime sexual partner; and a history of genital warts (Table 2). A high lifetime number of sexual partners, a history of genital warts, and unmarried status were the risk factors most strongly associated with cervical HPV infection. Notably, women who had four or more sexual partners had a 4-fold higher risk of being HPV-positive than women who had only one sexual partner.
It is interesting to note that the pattern of association with age at first intercourse in the adjusted model was opposite to that found in the basic model. While in the crude analysis young age at first intercourse was associated with increased HPV prevalence, in the fully adjusted model this effect was reversed and young age at first intercourse was associated with reduced HPV prevalence (P for trend = 0.003) ( Table 2). To explore these associations further, we performed stratified analyses by levels of lifetime number of sexual partners and levels of age at first sexual intercourse ( Figure 1). As shown in Figure 1a, HPV prevalence was not associated with age at first intercourse within each category of number of sexual partners. In contrast, Figure 1b shows that HPV prevalence was strongly associated with lifetime number  of sexual partners within each category of age at first intercourse (P < 0.001 for each age stratum). These stratified analyses show that HPV prevalence is more strongly associated with lifetime number of sexual partners than with age at first sexual intercourse. A similar effect was obtained when evaluating time since first sexual intercourse instead of age at first intercourse.
Interaction between lifetime number of sexual partners, male circumcision, and smoking status Interaction analyses showed that having a current partner who was circumcised had a significant protective effect against cervical HPV infection in women reporting two or more lifetime sexual partners but not in monogamous women (P for interaction = 0.005) ( Table 3). A synergistic interaction with cervical HPV infection was also observed between smoking and having two or more lifetime sexual partners. Thus, the risk of infection almost doubled in women reporting two or more lifetime sexual partners if they were current smokers, whereas in monogamous women current smokers showed little increase in risk compared with non-smokers (P for interaction = 0.01).  No other statistically significant interactions with lifetime number of sexual partners were found.

Risk factors for HPV infection in young women
We examined the potential risk factors associated with HPV infection among young women aged ≤25 years. In the adjusted model, the lifetime number of sexual partners, a history of genital warts, a country of birth other than Spain, and current smoking status were identified as the most important risk factors for cervical HPV infection (data not shown). Young age at first intercourse was associated with decreased HPV prevalence (P for trend = 0.006). Accordingly, increasing time since first sexual intercourse was also inversely associated with HPV infection (P for trend = 0.06).

Risk factors for HR versus LR HPV infections and for HPV 16 infections
We explored whether the determinants for HR HPV infection differed from those for LR infection by carrying out a stratified analysis ( Table 4). The risk factors for HR infections (Table 4) were the same as those for any HPV infection (Table 2). In contrast, the determinants for LR HPV infection were: history of genital warts, increasing lifetime number of sexual partners, being single, and (inversely but weakly) ever use of condoms (Table 4).
We also compared the risk factors for HR versus LR HPV infections by modeling infected women only. The following factors were positively associated with HR as compared to LR infection: birth in countries other than Spain, ever use of condoms, and a history of previous cervical lesions (data not shown). Determinants associated with HPV 16 infection, compared with HPV-negative women, were also evaluated. Women of young age, smokers, and women with more than two lifetime sexual partners had an increased risk of HPV 16 infection (data not shown).

Risk factors for single versus multiple HPV infections
In a stratified adjusted model, the same variables as those identified in the adjusted model for any HPV infection ( Table 2) were independently associated with single or multiple HPV infections (Table 5), except for residential area. Furthermore, ever use of condoms had a borderline protective effect against single HPV infection but not against multiple infection. Regression models comparing potential risk factors for multiple versus single HPV infections showed that women with more than two lifetime sexual partners were more likely to have multiple than single HPV infections (data not shown).

Discussion
This is the first large-scale, country-wide study in Spain to evaluate potential risk factors for cervical HPV infection. In addition to young age, a lifetime number of two or more sexual partners, a history of genital warts, and being unmarried were the strongest risk factors for cervical HPV infection among women aged 18-65 years attending routine cervical cancer screening across Spain. Living in an urban community, country of birth other than Spain, low level of education, and current smoking were also identified as independent risk factors for

cervical HPV infection. Condom use provided a borderline protective effect against HPV infection.
Our results agree with those of a previous epidemiological study in the general female population of Spain (N = 973) conducted in 2003 by de Sanjose et al., which found the following independent risk factors of HPV infection: overseas birth, divorce, more than one sexual partner, and smoking marijuana or related products. Use of condoms with a regular partner was found to be protective against HPV, but with a borderline effect [13].
As in other studies [7,10,11,13], an increased lifetime number of sexual partners was the most consistent risk factor for cervical HPV infection identified in our study. In contrast, age at first sexual intercourse was not identified as an independent risk factor in the multivariate analysis, in accordance with some previous studies [7,10,13]. In our study, young age at first intercourse increased the risk of HPV infection in the unadjusted model but in the adjusted one the effect was reversed. The inclusion of number of sexual partners in the adjusted model explained this change, indicating reverse co-linearity between the two variables. Our further stratified analyses, as shown in Figure 1, strongly suggested that the key determinant of cervical HPV infection is the number of sexual partners the woman had had and not the age at first sexual intercourse. These findings are in agreement with other studies suggesting that age at sexual debut is a likely predictor of lifetime number of sexual partners and not a factor independently associated with cervical HPV infection [15,16]. A similar effect was observed when time since first sexual intercourse was assessed rather than age at first sexual intercourse.
Country of birth other than Spain was identified as an independent risk factor for cervical HPV infection, a finding that was reported in a previous, smaller study conducted in Spain [13]. In our study, most women born outside Spain were immigrants from Latin America. The HPV prevalence in these countries is higher than that in Spain [17,18]. The increased prevalence of HPV among immigrant women compared with women born in Spain may be related to well-known differences in sexual behavior in both men and women.
In agreement with the pooled analysis of the International Agency for Research on Cancer (IARC) on smoking and HPV prevalence [8], we identified current smoking status as an independent risk factor for cervical HPV infection. A large population-based study conducted in women from Costa Rica found a positive association between smoking status and HPV 16, an association that was also found to be statistically significant in our study [10]. Intensity of smoking was    identified as another predictor of HPV infection in the IARC analysis, but these data were not statistically significant in our adjusted model. An association between smoking and cervical HPV infection is difficult to assess because of the strong confounding effect of sexual behavior. To address this we performed further logistic regression analyses to formally test for the interaction between smoking and number of sexual partners, and found a statistically significant interaction between the two risk factors (Table 3). Smoking was associated with a two-fold increased prevalence of cervical HPV in women with multiple sexual partners but not among monogamous women. In contrast, in the IARC study a significant relationship of smoking intensity was seen only among women who reported one sexual partner, but not among women reporting more partners. A possible explanation for this interaction is that among women with a high exposure to HPV (those with multiple partners) the risk of infection is much increased if they are concomitantly exposed to smoking. Conversely, smoking in the absence of high exposure to HPV is not as relevant in determining the risk of HPV infection. Some studies have documented that in the presence of HPV, tobacco products may reduce local immunological responses, which might facilitate HPV acquisition, reduce HPV clearance, or increase HPV persistence, thus increasing the overall HPV prevalence [19,20]. Condom use was found to be inversely but weakly associated with cervical HPV infection. The protective effect of condom use against HPV infection is controversial in the literature. Whereas some previous studies have reported a weak association with HPV infection [10,11,13], others have not [7]. Taking into account the weak magnitude of the associations found overall and in the subgroup analyses performed in our study, the beneficial effect of condom use on HPV prevalence is probably small.
Although male circumcision was not identified as a predictor of cervical HPV infection in our study, interaction analyses showed a protective effect among women at increased risk of HPV infection (i.e. women who had more sexual partners). Previous studies have shown that circumcised men are less likely to have penile HPV infection than uncircumcised men and that male circumcision infers a protective effect in female sexual partners for cervical HPV infections as well as for cervical cancer [21,22].
A strength of this study is that the age-stratified sampling strategy used within each of the Autonomous Communities of Spain (except Ceuta and Melilla) provided a good demographic representation of the agespecific female population attending cervical cancer screening across the country. However, a potential limitation is that women were selected from opportunistic cervical cancer screening programs, which may have resulted in the preferential recruitment of healthconscious or healthier women. This potential overrepresentation of "healthy" women could underestimate the true prevalence of risk factors related to cervical HPV infection, such as tobacco smoking, high number of sexual partners or having never used a condom. The potential effect of this would be an underestimation of the true effects related to these risk factors. In order to assess potential selection bias, the distribution of key socio-demographic and lifestyle characteristics of women included in the CLEOPATRE study was compared with that of the general female population of Spain using public data such as marital status, parity, use of contraceptive methods, and number of sexual partners among others. We ascertained that the study population was indeed representative of the general female population in Spain (data not shown). As with any other crosssectional design, there is a risk of reporting bias, particularly in relation to variables that address aspects of sexual behavior, such as "age at first sexual intercourse," "lifetime number of sexual partners," and "partner's circumcision status." Other studies have noted also the possibility of this non-differential misclassification related to sexual behavior [7]. It is well known that nondifferential misclassification cannot inflate true effects; instead, it would shift the observed association towards the null value.
Another limitation of our study is that only 80% of the HC2-positive samples were positive by LiPA, a much more sensitive assay than HC2. This apparently low performance of the LiPA assay may be explained by the inherent differences between the two assays. Although the HC2 assay is designed to detect 18 HPV types (including the low-risk types), several studies have noted that the HC2 HR probe mix may cross-react with other HPV types that are not represented in the mix [23][24][25]. Data on the cross-reactivity of the low-risk probe are not available, but it can be assumed that cross-reactivity also takes place with that probe. In contrast, in the LiPA test (which is designed to detect 27 HPV genotypes by  identification of highly specific HPV DNA sequences), cross-reactivity with other HPV types has not been observed. It is thus conceivable that HPV types detected by HC2 through cross-hybridization are not detectable with the LiPA assay. The HR probe and the LR probe, with an unknown extent of cross-reactivity, were both used in the present study, and this may explain the observed relatively low sensitivity of the LiPA assay. When comparing the two methods, a Kappa value of 0.67 (95% CI, 0.64-0.70) was found, which could be interpreted as being acceptable.

Conclusion
This analysis of the CLEOPATRE Spain study has identified several socio-demographic, lifestyle, and medical factors associated with cervical HPV infection. Cervical HPV is a highly prevalent sexually transmitted infection in women in Spain, particularly among those under 30 years of age in whom the prevalence approaches 30%. In addition to young age, increasing number of lifetime sexual partners and a history of genital warts were identified as key risk factors for cervical HPV infection, particularly for HR HPV, reinforcing knowledge that cervical HPV infection is acquired early in a woman's sexual life. Living in an urban community, country of birth other than Spain, low level of education, and current smoking were also identified as independent risk factors for HPV infection. A weak inverse association between condom use and HPV infection was also observed. In non-monogamous women, circumcision of the partner was associated with a reduced risk of HPV infection and smoking with an increased risk of HPV infection. Our results emphasize the importance of HPV vaccination programs for young women before they acquire sexual partners to prevent HR and LR HPV infections. Surveys of cervical HPV prevalence and its related risk factors increase our understanding of the natural history of this infection and establish the basis for future assessments of the impact of HPV immunization programs and the dynamics of risk factors for HPV infection in the population of Spain.