Cutaneous squamous cell carcinoma (SCC) is the second most common malignancy in the United States [1] and its incidence is steadily rising [2,3]. Although surgical excision is often curative, cutaneous SCCs can metastasize and become fatal, especially in immunocompromised patients [4,5]. Treatment can cause significant disfigurement and morbidity and accounts for high health-care expenditure [6]. Known risk factors associated with SCC can be divided into host-related and environmental factors. Host-related factors include innate pigmentary characteristics [7,8], sun sensitivity [9], precursor lesions [10], genetic predisposition (xeroderma pigmentosa, epidermoplasia verruciformis) [11], and immunologic factors [12]. Environmental factors include exposure to physical agents, the best characterized of which is ultraviolet light [13–15]. Both UVA and UVB light initiate and promote carcinogenesis [16] and immunosuppression [17]. Chronic cumulative exposure to ultraviolet radiation has been established as a risk factor for SCC [7,9,18]. Other physical agents include ionizing radiation [18], chemical agents (polycyclic aromatic hydrocarbons, arsenic, and nitrosamines) [11], medications used for immunosuppression, chronic inflammation (ulcers, sinus tracts), trauma (burns and scars), and viruses (certain types of human papillomavirus) [20,21] also have been implicated.

Epidemiological studies suggest that individuals with cutaneous SCC are more likely to develop other malignancies compared with individuals who have no history of non-melanoma skin cancer (NMSC) [22–24]. Additionally, in patients who have a history of NMSC, there is an increased incidence of and mortality from leukemia, non-Hodgkin’s lymphoma, and cancers of the lung, bladder, breast, testis, salivary gland, small intestine, and pharynx [25]. According to several recent epidemiological studies, a history of NMSC may be associated with 20 to 30 percent increased mortality from another type of cancer [23,25]. Understanding exposures that predispose individuals to cutaneous SCCs also may shed light on risk factors for these other types of malignancies. Recently, a population-based case-control study reported a statistically significant increased risk estimate for SCC among oral contraceptive (OC) users than non users [26]. Using an established cohort of Kaiser Permanente Northern California (KPNC) members with data on self-reported pre-diagnostic characteristics, we performed a nested case-control study to further examine the putative association between OC use and SCC in the context of other environmental and host-related risk factors.

A total of 516 participants with SCC included 321 men and 195 women. The mean age was 71.4 years (standard deviation 11.3, range 33–97). The majority (62%) of SCCs were diagnosed on the head and neck consistent with previous reports [3,39]. The remaining tumors were located on the upper extremities (11%), trunk (9%), lower extremities (9%), and other non-genital skin (10%).

We observed a borderline statistically significant association between oral contraceptive use and subsequent SCC risk among women in our cohort. If not merely due to chance or study bias, a possible explanation is that oral contraceptives, which contain estrogen (ethinylestradiol) and/or progesterone (progestin or synthetic progesterone-like compounds), alter the serum estradiol/progesterone ratio, which may influence the oncogenic potential of cutaneous squamous cells [40,41]. Estrogen receptors are present in normal skin [42]. However, SCCs are not believed to express significant amounts of sex hormones, suggesting that an association between OC use and SCC may be mediated through non-sex hormone pathways [43]. One possible pathway is p53, as estrogen appears to inhibit the actions of this tumor suppressor [44]. Inactivation of the p53 gene is believed to play a pivotal gatekeeper role in SCC carcinogenesis [45–47]. Risk also may be indirectly increased through interactions with polymorphisms in nucleotide excision repair (NER) genes including Xeroderma pigmentosum group D (XPD) [26]. Another possibility is that women who use birth control pills may have certain lifestyle factors, such as increased sexual activity, which may make them more likely to harbor HPV. Infection with some HPV types has been implicated in the pathogenesis of cutaneous SCCs in immunocompetent hosts [1,21,48–52]. The association between birth control use, HPV risk and cervical SCC has been reported [36,53–55]. A similar association could hold for cutaneous SCC.

To the best of our knowledge, only one published paper to date has studied the association between OC use and SCC [26]. Overall, OC users had a 1.6 adjusted odds ratio (OR) for SCC (CI = 1.0–2.5). ORs also were higher among women who last used OCs ≥ 25 years before diagnosis (OR = 2.1, CI = 1.2–3.7), and within group ORs increased with duration of use (OR for ≤ 2 years, 1.7; CI = 0.9–3.5; OR for 3–6 years, 2.6; CI = 1.0–6.5; OR for ≥7 years, 2.7; CI = 0.9–8.5, P_trend = 0.01). Our results support these previously published findings.

The epidemiology of SCCs has been difficult to characterize because conventional national registries, such as the Surveillance, Epidemiology, and End Results (SEER) [56] program exclude NMSCs. The unique advantages of the Kaiser Permanente Northern California (KPNC) setting are that it closely simulates the surrounding population serving nearly one-third of the insured population of Northern California and it houses an electronic database that captures information on all pathology specimens received for examination, allowing for thorough and accurate capture of incident SCCs. Recall bias was not a concern in our study since OC use was recorded prior to the diagnosis of SCC.

One potential limitation of this study is residual confounding due to indirect measurement of sun exposure, a known risk factor for SCCs [7,9,18]. We used occupational UV and time spent in leisure activities, or exercise, as surrogate markers for sun exposure reasoning that sun exposure comes from two primary sources: time spent in the sun for leisure/exercise and time spent in the sun related to one’s occupation. Among men and to lesser degree women, the strength of association with leisure time activities increased with time, supporting the assumption that time spent in leisure activities is correlated with UV exposure. However, leisure time activity may have been an inexact surrogate measure of sun exposure as the prompting examples (“Hobby, TV, etc.”) given for the question on the MHCQ were vague and may have been interpreted as activities that did not involve sun exposure. Nor did the question differentiate between sun and non sun related leisure time activities. Similarly, the MHCQ did not specify the type or duration of exercise, or whether the exercise was performed indoors versus outdoors. Surprisingly, time spent in exercise was slightly protective among men. Although this may seem counterintuitive, it is conceivable that exercise conveys health benefits that may offset the negative effect of UV exposure. Furthermore, some participants may have underreported occupational sun exposure since they might not have known that “ultraviolet radiation” was a surrogate term for “sun exposure.” While our surrogate measures of sun exposure were inexact, there is no inherent reason to believe that sun exposure is associated with OC use and thus unlikely to have affected the risk estimates for this variable.

Attenuation of risk estimates may have occurred if some control subjects had SCC diagnosed outside of the KPNC system. This is unlikely because KPNC is a comprehensive healthcare system and members would have to pay out-of-pocket for services received outside the health plan. Also, the exposures that we studied were obtained at a single point in time and were not measured over the entire study follow-up period.

The possibility remains that women who take oral contraceptive medicines may be more likely to interact with the healthcare system and increased risk for SCC may be due to detection bias. However, most SCC is diagnosed years after discontinuing OC use. Women with a history of OC use may have differentially received greater screening for cervical cancer. In our analysis, we adjusted for level of education, a factor believed to influence screening behavior.

In the current study, OC use was crudely measured as “ever/never” exposed and did not include information on dose, duration of OC use, pill composition, or serum estradiol and progesterone levels. The potency and overall dose of OCs have changed over time. Our results reflect the use of earlier compositions of OCs when hormone doses were considerably higher and may no longer hold for present day OCs. Nonetheless, this study is an important addition to the literature as the use of OCs was recorded prior to skin cancer diagnosis.

Although KPNC is generally reflective of the broader population in Northern California [57,58], there are some differences that may have introduced uncontrolled factors regarding ethnicity or behavior. In addition, individuals who self-selected for the MHCQ may have differed from the larger KPNC population on these as well as other distinguishing characteristics that were not directly measured and could have introduced selection and/or detection bias. However, these factors probably did not affect the overall validity of our case versus control comparisons.

Occupational exposure to dust from asbestos, cement, or grain was included as a potential confounding factor in our model based on our hypothesis that occupational dust may coat the skin to form a physical barrier to UV light. Previous reports of occupational risk and keratinocyte tumors have not noted this specific association [59]. Exposure to agricultural dusts have been associated with decreased lung cancer risk [60] suggesting a possible anticancer effect independent of an interaction with UV light. However, a healthy worker survival effect and reduced smoking among farmers may have been a more plausible explanation for the reduced risk observed among men in our study [61]. A limitation of our analysis was that the MHCQ grouped all three types of dust into one category and the association of each with SCC might differ.

The MHCQ also did not differentiate between types of cancer in the self-reported personal history of cancer question. Thus, it was not possible to determine if prior history of self-reported cancer and SCC risk was due to a personal history of NMSC versus other cancers. NMSC is the most common cancer in the United States [4] and history of NMSC is a major risk factor for a new primary SCCs [62]. In our study, risk was higher only among men who self-reported a history of cancer and may reflect a higher incidence of NMSC among this group [63,64]. Although our main effects multivariable analysis was adjusted for self-reported personal history of cancer, it is possible that the resulting estimate for OC use was affected by residual confounding.

We did not detect a consistent association between smoking and SCC risk as has been reported by some studies [65–68]. Our finding may reflect a different study population which may be susceptible to different gene-environment interactions. Or it may be due to our simple smoking history classification (i.e., never, former, current) which did not account for pack years smoked, filtered versus unfiltered, and other detailed smoking information. However, the possibility remains that smoking does not increase SCC risk as was observed in a large occupational cohort study [69]. Given the uncertainty of our smoking variable, we cannot rule out residual confounding in our observed association between OC use and SCC.

The association of SCC with innate pigmentary factors, such as light eye color, is well established [7,8] and is supported by our data. Our results indicate that environmental exposures which were used as surrogate markers for UV exposure (occupational UV in the past year and time spent in leisure activities) also were highly correlated with SCC risk, as expected. Similarly, our finding of an association between higher education and SCC risk has been previously reported [19,70,71]. Education level may affect SCC risk through socioeconomic status leading to differences in lifestyle and health-seeking behavior. Those individuals with more education may have higher socioeconomic status, allowing them to take mid-winter vacations in sunny locations, leading to higher SCC risk due to more frequent episodic sunburns. Alternatively, higher education also may lead to detection bias if more educated individuals are more likely to seek health care. Further studies on the mechanisms underlying the association between education and SCC risk are needed.

In summary, we observed a borderline statistically significant increased SCC risk with use of oral contraceptives similar to that reported in a recent case-control study. On the present evidence, it is not possible to definitively answer the question of how OC use influences SCC risk, if such an association exists, or to favor any specific hypothesis. If confirmed in future studies, these results will lead to new insight in the etiology of SCCs.