|Home | About | Journals | Submit | Contact Us | Français|
Oral contraceptives (OCs) affect the risk of several cancers in women, but have been virtually unstudied for squamous cell carcinoma (SCC). We examined the hypothesis that OCs influence SCC risk in a case–control study among women and also examined whether polymorphisms in the DNA repair gene, Xeroderma pigmentosum group D (XPD), modified the risk. Incident cases of SCC were identified by a network of dermatologists and pathology laboratories. Population-based controls were frequency matched to cases by age and gender (n = 261 SCC cases, 298 controls). Overall, OC use was associated with a 60% higher risk of SCC (odds ratio (OR), 1.6; 95% confidence interval (95% CI): 1.0–2.5). ORs for SCC were higher among those who last used OCs ≥25 years before diagnosis (OR: 2.1; 95% CI: 1.2–3.7), and among these women, SCC risk increased with duration of use (OR for ≤2 years, 1.7; 95% CI: 0.9–3.5; OR for 3–6 years, 2.6; 95% CI: 1.0–6.5; OR for ≥7 years, 2.7; 95% CI: 0.9–8.5, Ptrend = 0.01). Furthermore, the XPD Lys751Gln polymorphism was a significant modifier of the OC-SCC association (Pinteraction = 0.03). These findings lead us to hypothesize a potential relationship between OCs and SCC risk, and that this could involve DNA repair pathways.
Squamous cell carcinoma (SCC) of the skin, one of the major forms of non-melanoma skin cancer (NMSC), is among the most common malignancy for whites of European ancestry (Karagas et al., 2006), and its incidence appears to have increased dramatically within the last several decades. Incidence rates of SCC are higher in men than in women; however, in recent years, the rate of increase in these malignancies may be greater in women than in men (Karagas et al., 2006). On the basis of a statewide surveillance system in New Hampshire, the incidence rate for SCC had a notable increase of 235% in men and an even greater increase of 350% in women between 1979–1980 and 1993–1994 (Karagas et al., 1999). The reasons for SCC patterns differing by gender are not known.
Sex steroids have been widely investigated as risk factors for a number of cancers in women, particularly cancers of reproductive organs, including breast, endometrial, ovarian, and cervical. Unlike these other cancers, the relationship between NMSC and oral contraceptives (OCs) has not been thoroughly examined. Yet, OCs affect the skin, including acting as a photosensitizing agent by potentiating the skin's reaction to UV radiation, the dominant risk factor for NMSC (Esoda, 1963; Satterthwaite, 1964; Resnik, 1967; Mathison and Haas, 1970; Horkay et al., 1975; Sedee and Beijersbergen van Henegouwen, 1985; Diffey, 1986; Neumann, 1988; Cooper and George, 2001; Silver et al., 2003). Estrogens, as growth-promoting hormones, could enhance the risk of NMSC (Preston-Martin et al., 1990; Williams et al., 1991). Growth promotion is one of the crucial biological pathways in which cellular dysregulation promotes carcinogenesis (Hanahan and Weinberg, 2000). Keratinocytes, the cells that give rise to NMSC, express estrogen receptors and are thought to be estrogen responsive (Urano et al., 1995; Pelletier and Ren, 2004; Verdier-Sevrain et al., 2004).
Moreover, it is possible that estrogen may act in concert with known carcinogenic pathways in NMSC. One of the first recognized NMSC susceptibility factors was diminished DNA repair capacity brought on by coding changes in the nucleotide excision repair (NER) genes. NER removes UV-induced DNA lesions, including pyrimidine dimers and 6,4-photoproducts, which can lead to NMSC if not repaired. Polymorphisms in NER genes including Xeroderma pigmentosum group D (XPD) have been found to modify DNA repair capacity (Qiao et al., 2002). Although a reduced repair phenotype has been linked to an increased risk of NMSC (Wei et al., 1994), NER polymorphisms, including those in XPD, have been associated with reduced risks of NMSC in epidemiological studies (Han et al., 2005; Applebaum et al., 2007). In this analysis, we focus on polymorphisms in XPD because this gene has been widely examined for its potential relationship to cancer (Manuguerra et al., 2006) and because XPD polymorphisms have been found to be modifiers of skin cancers (Han et al., 2005; Millikan et al., 2006; Applebaum et al., 2007). Keratinocytes with XPD polymorphisms (and less DNA repair) are thought to be more likely to undergo apoptosis due to an enhanced apoptotic response (Bowen et al., 2003). In terms of the impact of estrogen on this process, a previous study suggested that exogenous estrogen may modify DNA repair capacity (Grossman and Wei, 1995), but it remains to be seen whether a gene–environment interaction between OCs and XPD polymorphisms would lead to an increased risk of skin cancer in a population-based study.
To explore the hypothesis that OCs increase SCC risk, we analyzed data on history of OC use collected as part of a population-based case–control study among women in New Hampshire. Furthermore, we analyzed whether an association could be modified by common polymorphisms in XPD (Asp312Asn and Lys751Gln).
Among these women, SCC cases had attained a higher level of education, experienced more severe sunburns, and sunbathed more than the controls (Table 1). SCC cases also were much more likely to have fair pigmentation characteristics than controls. Education, severe sunburns, sunbathing, and pigment score were all significant (P<0.05) predictors of SCC after adjusting for age.
The frequency of SCC cases ever taking OCs was slightly higher than controls (SCC cases: 44.4%; controls: 43.3%). In multivariate models controlling for age, education, pigment score, sunbathing, and severe sunburns, the odds ratio (OR) and 95% confidence interval (CI) for the association between OCs and SCC risk was elevated (OR: 1.6; 95% CI: 1.0–2.5) (Table 2). Marital status, parity, and body mass index were also evaluated as potential confounders, but did not influence the results. We examined the relationship for current and former OC use; however, there were few current users (n = 10).
Mean duration of use was 6.6 years (standard error = 5.9) in controls and 5.9 years (standard error = 4.5) in SCC cases. We examined whether the association varied with duration of OC use; however, within duration categories, the odds ratio (OR) remained the same (OR≤2 years: 1.6, 95% CI: 0.9–3.0; OR3–6 years: 1.6, 95% CI: 0.9–3.1; and OR≥7 years: 1.6, 95% CI: 0.9–2.8) (Table 2). By evaluating time since last use, we found an increased risk of SCC among those who last used OCs ≥25 years, which was statistically significant (OR: 2.1; 95% CI: 1.2–3.7); no association with SCC was observed among those who took OCs more recently (OR: 1.1; 95% CI: 0.6–2.0) (Table 2). Among those who last took OCs at least 25 years ago, we observed a trend of increasing SCC risk with duration of use compared with never users (OR≤2 years: 1.7, 95% CI: 0.9–3.5; OR3–6 years: 2.6, 95% CI: 1.0–6.5; and OR≥7 years: 2.7, 95% CI: 0.9–8.5; Ptrend = 0.01) (Table 2).
We investigated the gene–environment interaction between OC use at least 25 years ago and the XPD polymorphisms. Among the women who took OCs at least 25 years ago, we observed that those who did not have a 751Gln variant allele had a four-fold increased risk of SCC (OR: 4.4; 95% CI: 1.4–13.6) relative to those who had at least one copy of the variant allele; however, no excess risk was observed among recent OC users, regardless of their genotype (Pinteraction = 0.03) (Table 3). The association between SCC risk and the 312Asn polymorphism showed the same pattern as the 751Gln polymorphism, which was expected, as we have previously reported that these alleles were in linkage disequilibrium (data not shown) (Applebaum et al., 2007).
We observed an association between OC use and SCC risk, with a duration-related increase among women who used OCs ≥25 years before diagnosis. In addition, among these women, we detected an interaction with the Lys751Gln polymorphism in the DNA repair gene XPD.
Estrogen-induced keratinocyte malignancies could arise through similar pathways that are thought to influence the risk of other cancers, including cellular proliferation, metabolism leading to the generation of free radicals, and DNA adduct formation from estrogen-quinone metabolites (Zhu and Conney, 1998; Cavalieri et al., 2000; Hilakivi-Clarke et al., 2002). Clinically, skin changes after the intake of OCs, including melasma, suggest that individuals have an enhanced response to UV exposure when taking OCs (Resnik, 1967; Mathison and Haas, 1970; Horkay et al., 1975; Sedee and Beijersbergen van Henegouwen, 1985; Cooper and George, 2001). Thus, there may be additional pathways to consider for SCC. Furthermore, estrogen binds to the estrogen receptor, a transcription factor, influencing gene expression in cells (Urano et al., 1995; Verdier-Sevrain et al., 2004). If the expression profiles of cells change, this may alter susceptibility to UV carcinogenesis. Our findings of an interaction between the DNA repair gene, XPD, and OC use on SCC risk are consistent with this hypothesis. Furthermore, the direction of the XPD results are consistent with previously published NER findings, where no elevated risk was observed among those with variant alleles. As described earlier, this was expected as keratinocytes with a variant allele likely have more unrepaired DNA damage, and thus undergo apoptosis. However, with exposure to OCs and proficient repair, an increased SCC risk was observed.
Although plausible, a relationship between OCs and risk of SCC requires further study, as very few epidemiological studies have evaluated this question. A cohort study among women in England and Scotland reported no association between OCs and a loosely defined category of all skin cancers other than malignant melanoma (Vessey et al., 2000). A weakness is that the study relied on self-reported skin cancers without histological confirmation, increasing the likelihood of incomplete ascertainment and diagnostic misclassification. A clinic-based study of all malignant and non-malignant skin cancers combined reported no difference in the percentage of controls compared with skin cancer cases who reported ever using OCs (Wei et al., 1994). However, the lack of histological distinction between skin cancers limits the opportunity to observe an association, particularly as these cancers arise from different cell types. Overall, results from these previous studies cannot be readily compared with our analysis in which we studied histologically confirmed, incident cases of SCC.
Findings of an excess SCC risk in relation to OC use conceivably could be confounded by the associations between OC use and sun-seeking behavior, that is, if women who took OCs were more likely to spend time in the sun. Indeed, among controls in our study, OC users had a history of more sunburns and reported more episodes of sunbathing than non-users (data not shown). However, after adjustment for sunburns, additional UV exposure variables such as sunbathing or other recreational UV exposures did not change our risk estimates. Another issue is whether individuals who were more susceptible to the effects of UV while taking OCs (e.g., melasma) were more likely to either cease taking them or change their sun-exposure behavior (e.g., by avoiding UV, applying sunscreens). We were unable to answer these questions directly; however, few participants in this study reported experiencing melasma (four controls and seven SCC cases), and their duration of OC use did not differ from those who did not experience melasma (data not shown). Age and education also related to both OC use and SCC. To examine the possibility of residual confounding, we restricted to those born in 1940 or later, but the association between OCs and SCC was, if anything, stronger, and the association remained even after restricting to women who had had some college education (data not shown). Furthermore, we examined the association within age strata but found that the relationship between time since last use and SCC showed a doubling in risk (data not shown). Nonetheless, residual confounding and chance cannot be ruled out entirely.
The potencies of the estrogen and progestin components used in OCs have changed over time along with a decrease in the overall dose. Furthermore, there are variations in the formulation across a 1-month cycle. A stronger association for women who used OCs ≥25 years before diagnosis could reflect the use of earlier compositions of OCs, which contained the highest levels of estrogen (e.g., ≥50 μg of estrogen); however, unfortunately we were not able to evaluate specific doses of OCs used in this analysis. Epidemiological studies involving complete OC histories could help elucidate the nature of the relationship with SCC and may provide insight into the effect of estrogen content and dose, as well as disease latency. Additional research is needed to determine whether more recent compositions of OCs may also be associated with SCC.
A potential limitation of our analysis is recall bias, for example, that cases may have better recall of their OC history than controls. In a case–control study of breast cancer (Nischan et al., 1993), investigators compared self-reporting of OC use with medical records data. They found no significant differences for breast cancer cases and controls in their ability to report on duration of use, time since first use, and time since last use. Given that SCC is a less-severe cancer than breast cancer, there should be less of a concern about reporting bias for these same variables in this study. Alternatively, a concern may be that an association reflects greater detection of SCCs among women who took OCs because they may be seeking more medical treatment, in general, than those who did not take OCs. Although this possibility exists, there were few current users of OCs and the association of interest was among women who last used OCs at least 25 years ago; thus, reducing the likelihood for this potential bias.
Our findings raise many questions regarding whether the use of OCs is a potential risk factor for SCC among women. More epidemiological research is needed to examine whether the findings observed here can be replicated. In addition, a better mechanistic understanding of OCs on keratinocyte carcinogenesis would shed light on whether OC use is contributing to the rising incidence rates of SCC among women in many regions of the world.
Cases of newly diagnosed SCC occurring among New Hampshire residents were identified through a network of dermatologists and pathology laboratories across New Hampshire and bordering regions. The initial study period covered diagnosis between 1 July 1993 and 30 June 1995 (series 1) (described in Karagas et al. 1999, 2001). This study additionally included diagnoses between 1 July 1997 and 30 March 2000 (series 2). The extended survey period encompassed all participating dermatologists from the first survey and all new practicing dermatologists in the second period. Eligibility criteria included being a New Hampshire resident at the time of diagnosis, being age 25–74 years at diagnosis, having a listed telephone number, and able to speak English. All eligible SCC cases were selected to take part in the study. The New Hampshire Department of Transportation provided population lists of state residents, which were used to identify potential controls aged 25–64 years. Information on potential controls aged 65–74 years was obtained from the Medicare enrollment lists. Controls, who were frequency matched to the NMSC cases, were randomly sampled from within age (25–34, 35–44, 45–54, 55–64, 65–69, and 70–74 years) and gender strata. Controls were eligible if they spoke English and had a listed telephone number. The study protocol was approved by the Dartmouth College Committee for the Protection of Human Subjects, and all participants provided informed consent. The study adhered to the Declaration of Helsinki protocols.
Participants were administered a detailed in-person interview at their home by a trained interviewer. Questions pertained to time before diagnosis for cases, whereas there was random assignment of a reference date for controls comparable to the diagnosis date of cases. Interviews inquired about skin sensitivity, pigmentation, education, marital status, reproductive history, and use of OCs. Questions on skin sensitivity and pigmentation included skin reaction to the first strong summer sun (tan only, mild burn then tan, burn), skin reaction to repeated sun-exposure (deep tan, moderately tan, mild tan and peel, freckling or no tan), self-reported hair color (dark brown/black, light brown, red, blond), self-reported eye color (brown/black, green/hazel, blue/gray), self-reported skin color (medium/dark, light), and the number of moles on back. For sun-exposure histories, participants indicated time periods in their lives (above 8 years of age) with consistent sun-exposure habits and, for each time period, also indicated the amount of time spent in the sun, including peak UV hours (e.g., 1000 to 1400 hours), distinguishing between winter and summer months, time spent sunbathing, outdoor recreational activities, and the number of severe sunburns. From this information, a life table of sun exposures was generated and cumulative measures of these sun exposures were calculated for each participant. Interviews began in January 1994, and questions regarding hormone-related factors were first asked in September 1995. Questions included whether they had ever taken OCs, duration of use (reported by participants as the total number of years and months that they took OCs), and timing of use (e.g., time since last use). To aid in recall, we provided photographs and lists of commonly used OCs. Beginning with the second phase of the study, participants were asked about parity, weight, and height.
Of those contacted, 324 SCC cases (80%) and 435 controls (71.5%) agreed to take part in the study. Beginning with the first interview containing the OCs questions, a total of 265 SCC cases and 306 controls were interviewed and 99% of these (263 SCC cases and 303 controls) answered whether they had taken OCs. Because SCC is predominantly a disease of whites, we restricted our analysis to whites, leaving 261 SCC cases and 298 controls to analyze for the relationship between OCs and SCC.
At the time of interview, participants provided blood samples, which were collected for DNA extraction. When it was not possible to collect blood, a buccal sample was obtained. Genotyping of the XPD polymorphisms, Asp312Asn (rs1799793) and Lys751Gln (rs13181), was conducted using Applied Biosystems' Taqman chemistry (Applied Biosystems, Foster City, CA). Greater detail on the data collection and laboratory methods has been provided elsewhere (Applebaum et al., 2007). For quality assurance, positive and negative controls were used in each genotyping run, and laboratory personnel were blinded to the case–control status.
There are numerous measures of pigmentation that can influence an individual's sensitivity to UV radiation. These include skin reaction to first hour of intense sun (tan only, mild burn then tan, burn), skin reaction to repeated sunshine (deep tan, moderately tan, mild tan and peel, freckling or no tan), hair color (dark brown/black, light brown, red, blond), eye color (brown/black, green/hazel, blue/gray), skin color (medium/dark, light), and the number of moles on the back (0, 1, 2–4, ≥5). Using these pigment factors, we generated a multivariate confounder score to have a comprehensive measure of confounding by pigmentation (Miettinen, 1976; Cook and Goldman, 1989). The higher the value of this score variable, the lighter a participant's pigment. Lighter pigment is consistent with lower melanin production, which in theory allows more UV exposure to reach keratinocytes.
Unconditional logistic regression was used to generate ORs and 95% CIs for the association between OCs with SCC. Models controlled for age continuously and severe sunburns categorically (0–2, ≥3). We also adjusted for the number of times sunbathed (0–30, 31–583, ≥584) and the pigment score (quartiles), both of which were based on the distribution in the controls. Highest level of education achieved (less than undergraduate college, undergraduate college, and education beyond undergraduate college), marital status, body mass index, and parity were evaluated as confounders.
Participants were considered ever users if they had taken OCs for a total of at least 3 months (defined by combining information on ever use of OCs and the total cumulative duration of use) by the time of diagnosis for cases and the reference date for controls. Participants who reported having never taken OCs or had taken them for <3 months were classified as never users. The analysis examined ever use of OCs, as well as the total duration of use (≤2, 3–6, ≥7 years), time since last use (≤24, ≥25 years), and ever experienced darkening of the skin on one's face or body while taking OCs. We further examined duration and time since last use concurrently using a joint effects model, with never users serving as the reference group. An observed trend was tested using an ordinal variable and a two-sided Wald test.
We examined the hypothesis that an association between OCs and SCC may be influenced by DNA repair using the Asp312Asn and Lys751Gln polymorphisms in XPD gene. A binary variable was created indicating whether participants had 0 or ≥1 variant alleles for each XPD polymorphism. Because previous studies on DNA repair polymorphisms have found carriers of variant alleles to be at a reduced risk of NMSC, they served as the reference category in this analysis. We tested for effect modification in a model containing the dichotomized variables for time since last OC use (≥25 years or not), each polymorphism, and their cross-product (a 1-d.f. test).
We are indebted to the physicians who comprise the New Hampshire Skin Cancer Study Group. This work was supported by Grant nos. R01CA082354 and R01CA57494 from the National Institutes of Health and Dr Applebaum's effort was supported by a Grant K01OH009390 from the Centers for Disease Control—National Institute of Occupational Safety and Health. These contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute or the National Institute of Occupational Safety and Health.
Conflict of Interest: The authors state no conflict of interest.