Smoking and passive smoking in cervical cancer risk: pooled analysis of couples from the IARC multi-centric case control studies

Background: The independent role of tobacco smoking in invasive cervical cancer (ICC) has been established. We evaluate the potential impact of passive smoking (PS). Methods: A pooled analysis of 1919 couples enrolled in one of seven case-control studies involving cervical carcinoma in situ (CIS) or ICC was investigated. Information on smoking and sexual behavior was collected from interviews. Specimens were taken from the cervix and penis for human papillomavirus (HPV) DNA testing. Three PS risk models were constructed with all couples, couples with monogamous women, and couples with lifetime non-smoking monogamous women. For the third model, the analysis considered potential misclassification of smoking status and was restricted to the risk period for which the woman was exposed to both HPV, a necessary cause of ICC, and passive smoking. Multivariable unconditional logistic regression was used to estimate associations between CIS or ICC and PS. Results: An increased risk was found among couples with both ever smoking men and women (OR=2.26; 95% CI: 1.40-3.64). No statistically increased risk of CIS was found with PS in the models analyzed. Similar significant increased risks of ICC with PS was found among all couples (OR=1.57; 95% CI:1.15-2.15) and couples with monogamous women (OR=1.55; 95% CI: 1.07-2.23) but not among lifetime non-smoking monogamous women married to ever smoking men. Conclusions: PS could not be detected as an independent risk factor of ICC in the absence of active smoking. Impact: The combined effects of exposure to active and passive smoking suggest its potential adverse role in cervical carcinogenesis.


Introduction
Men play an important role in the transmission of human papillomavirus (HPV), the etiological factor for invasive cervical cancer (ICC). As HPVs involved in cervical carcinogenesis are sexually transmitted, it is central to understand patterns of sexual behavior in HPV transmission including the behaviors of both men and women. It has been debated that a woman´s risk for cervical cancer will depend more on the full sexual history of the male partner than on her own behavior (1). This is particularly relevant in societies where women tend to be virgins at marriage and monogamous thereafter. Epidemiological studies have tried to characterize the male role and the consequent female partner risk of HPV acquisition and of cervical cancer (2)(3)(4)(5). While a number of studies mostly involving monogamous women have observed an association between the number of sexual partners of the husband and his wife's risk for cervical cancer (6)(7)(8), other studies have not (9)(10)(11)(12). Other inconclusive associations with prostitution have been identified (9)(10)(11)(12).
Besides sexual behavior, other male factors, such as tobacco smoking has been less explored despite it being a well-established risk factor for cervical pre-cancer and cancer (13)(14). Several reviews have summarized the epidemiological and biological association of passive smoking on the risk of cervical cancer (15)(16)(17), however, the evidence has been suggestive rather than sufficient to implicate the role of passive smoking in the etiology of cervical cancer among lifetime non-smoking women. Among the studies identified, the recognized limitations include small sample sizes of non-smoker controls and cases of cervical cancer, lack of specific information on HPV and sexual behavior and most studies obtained spousal history of smoking through questioning of the women rather than the men. Furthermore, most of the studies involved cervical intraepithelial neoplasia grade 3 (CIN-3)/carcionoma in situ (CIS) rather than invasive disease, which is a relevant distinction since 5 evidence suggests that smoking acts in the stages of progression from CIS to invasive cancer (14).
To evaluate the male role in the etiology of cervical cancer, specifically the risk related to passive smoking, we performed a pooled analysis of five case-control studies involving ICC and two case-control studies involving CIS, of couples in which husbands or stable partners of ICC and CIS case and control women participated. The studies were conducted in three continents, mainly in developing countries, and were coordinated by the subsets of the subjects in this analysis (9,11,14,(18)(19)(20). For this study, we characterize indepth the role of passive smoking with the full dataset on HPV and risk factors of the men and their associations with invasive cervical cancer.

Materials and Methods
The IARC/ICO case-control program included a series of studies on ICC and CIS from eleven countries with a broad range of cervical cancer incidence rates. Among these, seven studies conducted in five countries, enrolled husbands or stable partners of women with CIS or cervical cancer and control women were pooled for these analyses. Methods of each study and primary results related to women have been published previously. Countries included Brazil (21) and Colombia (9,11,22), the Philippines (23), Thailand (24) and Spain (9,25). Briefly, women with histologically confirmed incident cervical CIS, invasive squamous cell carcinoma, adenocarcinoma or adenosquamous-cell carcinoma were recruited from reference hospitals before treatment. Control women were recruited from the general 6 population in two of the studies of ICC in Spain and Colombia and from the same hospitals as the cases for the other studies. Control women were frequency-matched to case patients by five-year age groups.
Current husbands or stable partners (herein referred to as husbands) of enrolled women were defined as men who reported having had regular sexual intercourse with the women for at least six months, irrespective of whether or not they were married or lived together.
Informed consent was obtained from both men and women who agreed to participate.

Questionnaire
A standardized questionnaire was administered to participants by a trained interviewer that included questions about socio-demographic factors, sexual behavior, hygienic practices, and history of sexually transmitted infections (STIs). For specific questions on smoking habits, subjects were first asked to classify themselves as life-time never smoker, ex-smoker (defined as a former smoker who stopped smoking at least 1 year prior to the interview) or current smoker. Ever smokers were also asked at what age they started smoking regularly, the duration and how many cigarettes per day they smoked. Additional questions were asked on the type of tobacco (blond, black, or other) and type of filter (filter, no filter, or both) used.
Ex-smokers were asked the age at which they stopped smoking.

Penile and cervical HPV DNA sampling
Two samples of exfoliated cells were obtained from the penis: one from the distal urethra with the use of a very thin, wet, cotton-tipped swab and one from the external surface of the glans and coronal sulcus with the use of a standard-sized wet, cotton-tipped swab.
Two samples of cervical exfoliated cells were collected with wooden spatulae and endocervical brushes. After preparation of one Papanicolaou smear, the remaining cells were 7 eluted in saline, centrifuged and frozen at -70 ºC until shipment to the central laboratory for HPV DNA testing. A tumor-biopsy sample was obtained from cases and frozen at -70 ºC.
Cytology and histology diagnosis were reviewed and confirmed by a panel of expert pathologists that agreed on a diagnosis by consensus or majority.

Detection of HPV DNA
Detailed descriptions of the polymerase-chain-reaction (PCR) assays used in these studies have been described elsewhere. HPV DNA was detected by PCR amplification of a small fragment of the L1 gene using MY09 and MY11 consensus primers for the studies in Spain and Colombia (26) and GP5+/6+ general primer system for the other studies (27)(28)(29).
Beta-globin primers were used to amplify the beta-globin gene to assess the quality of the DNA in the specimen. HPV DNA in PCR products was analyzed with the use of a cocktail of HPV-specific probes and genotyped by hybridization with type-specific probes for 33 HPV types in the case of cervical samples and for at least 6 HPV types (6, 11, 16, 18, 31, and 33) in the case of the penile samples. Samples that tested positive for HPV DNA but did not hybridize with any of the type-specific probes were labeled as HPV X.

Statistical analyses
To evaluate the association between smoking habits, and risk of CIS or ICC, we first used age-and country-adjusted univariate logistic regression analyses to determine the effects of each of the following potential male factors using an alpha-level of 0.05: age, history of smoking (non-smoker, current smoker, or ex-smoker, lifetime pack-years, and use of tobacco and filter type), education, sexual history (age at first sexual intercourse, lifetime number of sexual partners, history of contact with sex workers, history of STIs, and history of anal sex), hygienic practices (i.e. pay attention to uncover penis and to wash the region, able to fully uncover spontaneously or by pulling the penis from the skin prepuce, and wash before and 8 after sexual intercourse), male circumcision status, and to control for potential confounding of passive smoking characteristics and risk of cervical cancer, final models were adjusted for male factors that contributed change to any of the estimated odds ratios (OR) and 95% confidence intervals (CI). To control for additional potential confounding by characteristics of the women, female risk factors (education, age at first sexual intercourse, lifetime number of sexual partners, history of pap smear 12 months prior to study enrolment, use of oral contraceptives, parity, and smoking) for cervical cancer were fitted into the final multivariate models a) for the CIS adjusted models if they contributed to any change to the OR estimates for male characteristics; and b) for the final ICC adjusted models as they are well-established risk factors known to be associated with ICC. However, when we adjusted the OR estimates with all female risk factors in the CIS model, the estimates did not significantly differ (data not shown). We identified lifetime number of sexual partners (a significant risk factor of exposure to HPV) to be heterogeneous across study countries (Supplementary Table 1) and an interaction term combining lifetime number of sexual partners and country were included in the fully adjusted multivariate models.
In addition, we found a statistically significant interaction between some male risk factors (e.g. age at first sexual intercourse, lifetime number of sexual partners, history of sexual intercourse with a sex worker, and HPV-positivity status) and case status (i.e. ICC versus CIS), which justified the use of two separate models for each disease stage. This is in agreement with our current understanding of the natural history of CIS, as it has been estimated that about 31% of CIS cases will develop cancer within 30 years, leaving a proportion of CIS cases that will not advance to invasive disease (30). Thus, some of the risk factors associated with CIS incidence may differ from those associated with progression from CIS to ICC. Furthermore, to better clarify the relationship between passive smoking characteristics and cervical cancer in their female partners, we removed the potential effect of previous male partners the woman may have had by calculating and comparing three different statistical models for CIS and ICC: one included all study couples, the second model included only couples with monogamous women, and the third model included couples with lifetime nonsmoking monogamous women. For the third model restricted to 765 couples with lifetime non-smoking monogamous women, we further reclassified the men's smoking history according to the risk period for which the woman would have been exposed to HPV infection (a necessary factor in cervical carcinogenesis), passive smoking and risk of progression to cervical cancer ( Figure 1). Ninety male ex-smokers (n=44 cases and n=43 controls) of couples with monogamous women were reclassified as non-smokers, and the duration of exposure to passive smoke and smoking pack-years were recalculated (Supplementary Table   2). Table 1 describes selected characteristics of the male and female subjects. Of the 291 CIS and 692 ICC cases and 936 control women, 59.8%, 70.9%, and 81.2%, were monogamous, respectively. In general, husbands were older than their wives, and husbands and wives of CIS cases and controls were younger than those of ICC. Table 2 shows penile HPV prevalence among husbands of cases and controls of CIS and ICC by history of smoking and country. Penile HPV detection was doubled in husbands of cases than controls of CIS and was higher among cases than controls of ex-smokers (2.4% vs. 0%) and current smokers (13.9% vs. 3.8%). Similar penile HPV detection was found in husbands of cases and controls of ICC (17.6% vs. 16.2%), which was also similar among cases and controls of ex-smokers and current smokers. However, penile HPV was more prevalent among husbands of cases than controls of ICC (9.7% vs. 6.6%), respectively. Table 3 presents selected male risk factors and their univariate associations with risk of CIS and ICC, stratified by all couples and couples with monogamous women. In general, similar associations were observed in all couples and couples with monogamous women models, except for associations of CIS with education and smoking status. Hygienic practices (able to fully uncover spontaneously or by pulling the penis from the prepuce and washing before and after sexual intercourse) and history of anal sex were not associated with CIS or ICC in univariate analyses (data not shown). The following husband's risk factors were found to be associated with a woman's increased risk of CIS or ICC: early age at first sexual intercourse, history of sexual intercourse with a sex worker (ever and while with current wife), history of gonorrhea, increasing number of STIs, being a current smoker and increasing lifetime smoking pack-years. Lack of education was associated with an increased risk of ICC but not CIS, whereas being uncircumcised and HPV-positive were associated with an increased risk of CIS but not ICC. An inverse relation between use of "black" tobacco type, as compared to "blond" tobacco type among smokers, and the risk of CIS was observed, as well some hygienic practices such as lack of attention to uncover the penis to wash the region and the risk of ICC. Table 4 shows the association between selected male smoking characteristics and cervical CIS and ICC in multivariate analyses. Generally, similar associations were identified in the two analyses of all couples and only couples with monogamous women, therefore we will describe our findings herein forward according to the all couples model. No statistically significant increased risk of CIS was observed for women whose partners had a history of smoking. An increasing risk of ICC was observed with decreasing time since smoking cessation with current smokers having the highest risk (OR=1.61; 95% CI: 1.16-2.24), 11 suggesting passive smoking as a potential risk factor for cervical cancer. No increased risk of ICC was observed for women with male partners who used a specific tobacco type or filter.

Patient characteristics
An increased risk of ICC was observed for women with partners who smoked at least a low number of smoking pack-years (OR=1.62; 95% CI: 1.14-2.29). Table 5 shows the association between passive and active smoking history and risk of ICC after reclassifying smoking status of men according to Figure 1. As compared to the active smoking model, we did not observe an association between male smoking habits and risk of cervical cancer among couples with lifetime non-smoking monogamous women.
Although passive smoking was not independently associated with risk of ICC, there was an increased OR from 1.23 to 2.26 when women were exposed to passive smoking alone or to both active and passive smoking. The interaction term was however not statistically significant (p=0.77) ( Table 6).

Discussion
This study shows no independent association of passive smoking and risk of cervical cancer in the absence of active smoking. In the first two models of all couples and couples with monogamous women, the lack of association with CIS and the significant association with ICC suggests that passive cigarette smoking could potentially acts as a late carcinogen in the transition from persistent infection/pre-invasive lesions to invasion. These findings are not new and are consistent with previous findings (15)(16)(17)(31)(32)(33). However, when we considered the possibility of misclassification bias in our third model of couples with lifetime non-smoking monogamous women and reclassified the men's smoking status according to the risk period for which the woman would be exposed to both HPV infection and passive smoking, no independent association could be found. The greatest risk estimate was more than two-fold for couples who were both ever-smokers.
The contradicting results as shown in the different models highlight the distortion of estimates probably resulting from misclassification of smoking status. This suggests that a model considering only the time period of exposure to HPV and passive smoking should be used to determine susceptibility to carcinogenesis. The timing of exposure to tobacco smoke relative to cervical cancer development is important in defining exposure. Since we had detailed information on smoking and sexual history we were able to define and calculate exposure based on a series of responses. The strict definitions of exposure to tobacco smoke in our analyses showed associations with risk of cervical cancer that were obscured by using simpler definitions. Non-smoking monogamous women with men classified as ex-smokers who have quit smoking prior to initiating a sexual relationship may not be as susceptible to passive smoking. In addition, the men's lifetime duration of smoking does not necessarily include the whole period of the couple's relationship if he stops smoking during the relationship or he starts and stops smoking during the relationship. Although the possibility of misclassification of the women's smoking status cannot be excluded, we do not believe inclusion of non-smokers who were actually true smokers would cause substantial bias since female smoking prevalence in these study countries is low (34). In epidemiologic studies of cervical cancer etiology, the definitions of exposure should reflect a model of risk to HPV infection and cervical carcinogenesis.
Overall, more than three-quarters of couples were with monogamous women, allowing us to explore the male passive smoking factors in depth and to limit any potential residual confounding that may exist if women were largely not monogamous. Our results may predominantly represent societies in which women report lifetime monogamy and multiple partnerships are more common among men, which is a pattern that is generally more common in developing countries rather than in developed countries (35).

13
This study strengthens the current evidence for several reasons. First, this study has the largest dataset of couples with non-smoking women to measure passive smoking. Since our study obtained direct information from interviews with both the husband and wife, our results are considered reliable as previous studies have found good agreement in responses concerning spousal smoking status to range from 90 to 100 percent (17) and previous cotinine studies of never-smokers have validated the use of spousal history as a marker of exposure to tobacco smoke and people who live with smokers tend to mix with smokers outside the home (15). In contrast, previous studies had small sample sizes with small numbers of nonsmokers. Secondly, previous studies lacked adequate information on HPV and sexual behavior indicators to control for potential confounding, and we were able to control for both male and female risk factors. Thirdly, as we currently understand the natural history of cervical cancer, not all pre-cancerous lesions will progress to invasive cervical cancer (30), we were able to evaluate the effect of passive smoking by stage of disease (pre-invasive vs. invasive).
Previous studies did not evaluate the combined effects of different exposure of active and passive smoke (both non-smokers, female ever-smoker/male non-smoker, and female non-smoker/male ever-smoker). Although the other combinations showed an increased risk, only the combination of ever-smoking couples showed a statistically significant increased risk. The lack of an independent association with passive smoking does not necessarily discount its contribution to ICC risk. This may suggest that the direct effect of active smoking outweighs the indirect carcinogenic effects passive smoking may have. One of the limitations of epidemiological studies using questionnaire data is its decrease in sensitivity or power of a study to show a positive association when the effect may only be moderately related to passive smoking (17). Lifetime number of sexual partners of the men largely attenuated the observed effect of passive smoking on ICC risk. Studies have suggested that we need to consider the contribution of occupational exposure to tobacco smoke in addition to spousal/household smoking as 76% of non-smokers who report no exposure to tobacco smoke at home have reported exposure at work (17) and about 75% of women in our study worked outside of the home, which could have lead to additional misclassification of exposure and underestimated the impact of passive smoking. In addition, although the possibility of couples not cohabiting together could lead to an overestimated passive smoking impact, we believe the contribution is minimal as only 1.6% of couples with monogamous women reported periods of separation. In order to fully evaluate the impact of passive smoking, measurement of exposure needs to take into account all environmental exposures within the household and workplace.
A biological mechanism by which active and passive smoking could influence cervical carcinogenesis is not clearly understood. However, tobacco smoke contains known carcinogens such as polycyclic aromatic hydrocarbons that could potentially have a direct transformation effect on the cervix or could cause immunosuppression, allowing HPV infections to persist and progress to cancer (15). Detectable levels of nicotine and cotinine, a measurement of smoke exposure, have been found in cervical mucus as well as DNA adduct levels in the cervical epithelium of non-smokers, supporting the evidence that these chemicals can reach distant sites such as the cervix (16). Another hypothesis includes mutagenic semen due to smoking is plausible and direct cervical contact with semen of smoking partners may represent another source of exposure (17). This study lacked data measurement levels of cotinine/nicotine in the cervix, therefore, additional studies are needed to obtain these data to complement our epidemiological findings.
Penile HPV detection was more prevalent among current smokers compared to exsmokers and non-smoker which is consistent to previous findings (36). This suggests that smokers may be more likely to have persistent infections compared to non-smokers, making them more likely to expose their wives to HPV infection. However, the interpretation of penile HPV detection at study enrolment is not straightforward as it does not necessarily represent the time-point of exposure as the current understanding of the natural history of HPV in men shows that HPV is more readily transmitted from men to women than from women to men, and these infections are less likely to persist among men with approximately 75% likely to clear infection at one year (37). Other studies have not found smoking to be associated with penile HPV acquisition nor persistence (38-39). In addition, we cannot exclude the possibility of reverse causality since HPV-infected husbands could clear HPV, and be re-infected by their wives who have cervical cancer and have been replicating HPV prior to the onset of cancer. Among 116 ICC case husbands who reported no history of sex with a sex worker or a casual partner while living with their wife, 6 were HPV-positive (of whom 5 women reported lifetime monogamy) making it impossible to know who was the source of HPV exposure if there was no underreporting. Secondly, detection of penile HPV DNA (17%) in our study was lower than recently reported prevalence estimates in men and this may result from incomplete sampling of the male genitalia as it has been suggested that for optimal HPV detection, sampling should include multiple anatomic sub-sites (40).
In conclusion, there are 1 billion active smokers worldwide and one third of adults are regularly exposed to passive smoke with the burden of tobacco-related disease, disability and death being the highest in developing regions. Moreover, the rate of increase in cigarette consumption in developing countries is ten times that of industrialized countries (41). This burden is likely to increase in the coming decades if current trends persist with more than 90% of the world's population not protected by comprehensive smoke-free policies and there is low compliance (2%) in countries where there are comprehensive smoke-free laws (34).
Globally, there is an increasing trend of females aged 13-15 smoking in recent years (41), which needs to be considered along with reported median age at first sexual intercourse to 16 occur for most women is 15-19 years (35) when assessing risk of ICC. The data presented here support that in addition to female tobacco smoking as an established cofactor for cervical carcinogenesis, there is a potential role of passive smoke on ICC, which suggest that the estimated burden of tobacco-related diseases may increase and magnify the need for effective tobacco control, notably in developing countries.

36.
Vardas E, Giuliano AR, Goldstone S, Palefsky JM, Moreira ED, Jr., Penny ME, et al.   b ICC model for all husbands and husbands with monogamous women were adjusted for age of the husband and wife, study country, interaction term (husband's lifetime number of sexual partners*study country), level of education of the husband and wife (≥secondary level, primary level, no schooling), history of sexually transmitted infections, age at first sexual intercourse of the wife (≥21 years, 17-20 years, ≤16 years), lifetime smoking pack-years of the wife (non-smoker, low, medium, and high smoking lifetime-pack-years), oral contraceptive use (never, 1-4 years, ≥5 years), parity (nulliparous, 1-6, ≥7) pap smear history 12 months prior to study enrollment (never, ever), and lifetime number of sexual partners of the wife (1, ≥2).

CIN-3/CIS
a a CIS multivariate model was adjusted for age of the husband and wife, study country, interaction terms (husband's lifetime number of sexual partners*study country and circumcision status*study country ), age at first sexual intercourse of the wife (≥21 years, 17-20 years, ≤16 years), wife's lifetime number of sexual partners (1, ≥2); and husbands with monogamous women model was adjusted for all variables in the table except for circumcision, interaction term (husband's lifetime number of sexual partners*study country), age of the husband's wife and age at first sexual intercourse of the wife. a For the passive smoking models, characteristics of the husband's smoking history were classified according to the wife's risk period of exposure to HPV and passive smoking as outlined in Figure 1.
Couples with non-smoking monogamous women Couples with monogamous women