|Home | About | Journals | Submit | Contact Us | Français|
Basal cell carcinoma (BCC) incidence is increasing, particularly among adults under age 40. Pigment-related characteristics are associated with BCC in older populations, but epidemiologic studies among younger individuals and analyses of phenotype-genotype interactions are limited. We examined self-reported phenotypes and melanocortin 1 receptor gene (MC1R) variants in relation to early-onset BCC. BCC cases (n=377) and controls with benign skin conditions (n=390) under age 40 were identified through Yale’s Dermatopathology database. Factors most strongly associated with early-onset BCC were skin reaction to first summer sun for one hour [severe sunburn vs. tan odds ratio (OR)=12.27, 95% confidence interval (CI)=4.08–36.94] and skin color (very fair vs. olive OR=11.06, 95% CI=5.90–20.74). Individuals with two or more MC1R non-synonymous variants were 3.59 times (95% CI=2.37–5.43) more likely to have BCC than those without non-synonymous variants. All host characteristics and MC1R were more strongly associated with multiple BCC cases status (37% of cases) than single BCC case status. MC1R, number of moles, skin reaction to first summer sun for one hour, and hair and skin color were independently associated with BCC. BCC risk conferred by MC1R tended to be stronger among those with darker pigment phenotypes, traditionally considered to be at low-risk of skin cancer.
Basal cell carcinoma (BCC), which accounts for 80% of non-melanoma skin cancers (NMSCs), is the most common cancer in the US, with more than two million BCCs diagnosed annually (Rogers et al., 2010; ACS, 2011). While BCC is unlikely to metastasize and is associated with low mortality, morbidity associated with this disease is quite high. In 1992 among US Medicare beneficiaries, NMSC ranked among the top five most costly cancers to treat (Housman et al., 2003). Newer data indicate from 1992 to 2006 in the Medicare population, there was a 77% increase in the total number of skin cancer-related procedures (93.7% NMSC), due to an increase in the number of individuals with these malignancies (Rogers et al., 2010). In recent decades, BCC incidence has increased (Arits et al., 2011; Bath-Hextall et al., 2007; Birch-Johansen et al., 2010; Doherty et al., 2010; Flohil et al., 2011; Karagas et al., 1999; Levi et al., 2001), with notable increases among adults under the age of 40, particularly women (Bath-Hextall et al., 2007; Birch-Johansen et al., 2010; Christenson et al., 2005).
Ultraviolet (UV) radiation is the primary environmental etiologic factor for BCC, yet intrinsic or host factors, including pigment-related characteristics, are also likely to play a role in carcinogenesis in conjunction with UV (reviewed in (Dessinioti et al., 2010; Madan et al., 2010)). Among pigment-related factors, the melanocortin 1 receptor gene (MC1R), which encodes a protein that binds melanocyte-stimulating hormone and regulates skin and hair pigmentation (Valverde et al., 1995), has received considerable attention and has been associated with an increased risk of melanoma and BCC (reviewed in (Scherer and Kumar, 2010)). Even though MC1R variants are related to light pigmentation phenotypes (Bastiaens et al., 2001; Box et al., 2001; Dwyer et al., 2004; Han et al., 2006; Kanetsky et al., 2004; Kennedy et al., 2001; Koppula et al., 1997; Naysmith et al., 2004; Palmer et al., 2000; Smith et al., 1998; Valverde et al., 1995), there seems to be an effect of genotype independent of phenotype on both BCC (Bastiaens et al., 2001; Box et al., 2001; Dwyer et al., 2004; Han et al., 2006; Liboutet et al., 2006; Scherer et al., 2008) and melanoma (Dwyer et al., 2004; Kanetsky et al., 2010; Kennedy et al., 2001; Landi et al., 2005; Palmer et al., 2000). These findings, in combination with other emerging evidence from epidemiologic, clinical, and basic science research, indicate BCC may be more similar to melanoma than squamous cell carcinoma (SCC) in etiology (Dessinioti et al., 2010; Madan et al., 2010).
BCC has been relatively understudied in epidemiologic research because it is not reported to most cancer registries. Thus far, two studies provided an intriguing glimpse at risk factors for early-onset BCC, but these had small sample sizes; 30 cases (Boyd et al., 2002) and 25 cases (Bakos et al., 2011). Due to the limited understanding of early-onset BCC etiology, we conducted a case-control study in Connecticut among individuals under age 40 investigating lifestyle, environmental, and genetic factors. The rationale for this study was multifold and included increasing incidence among young people, the opportunity to evaluate genetic factors in a genetically enriched population, the potential for younger individuals to better recall early-life exposures, and a growing prevalence of indoor (IARC, 2007) and outdoor tanning.
Here, we describe the design of the Yale Study of Skin Health in Young People and the associations between host phenotype characteristics and MC1R in relation to early-onset BCC. We also evaluated potential variation in the association between MC1R and BCC by phenotype.
Sixty-nine percent of the 767 participants were female (257 cases, 274 controls). The mean age at skin biopsy in cases was 35.1 years (SD=4.6) and 34.7 years (SD=5.5) in controls. Among cases, 54.1% (n=204) had the referent BCC on the head or neck, followed by 101 (26.8%) with a trunk BCC, and 72 (19.1%) with a BCC on an extremity. Approximately 37% (n=140) of cases had two or more BCCs under age 40.
All phenotype characteristics and MC1R were significantly associated with early-onset BCC, with lighter pigment phenotypes at greater risk (Table 1). The most pronounced risk factor was skin reaction to first summer sun of the season; those who experienced severe sunburn and blistering were 12.27 (95% CI=4.08–36.94) times more likely to have BCC than those who turned brown/tanned with no burning. Skin color was another strong risk factor; individuals with very fair skin were 11.06 (95% CI=5.90–20.74) times more likely to have BCC than those with olive skin. In a sensitivity analysis, excluding the top three control conditions one at a time from the control group did not impact risk estimates for all exposures (data not shown). Controlling for indoor and outdoor UV exposures did not alter associations for the characteristics of interest (data not shown).
We detected 35 MC1R variants (Supplemental Table 1 Online). Individuals with one MC1R non-synonymous variant were 93% more likely than those without non-synonymous variants to have BCC, with a stronger association for individuals with two or more non-synonymous MC1R variants (OR=3.59, 95% CI=2.37–5.43) (Table 1). Risk was elevated for both “major” and “minor” red hair variants (Table 2).
All host characteristics were associated with both single and multiple BCC case status, but the magnitude of the risk estimates for multiple BCC was much greater (Table 1). One of the most pronounced differences was for skin color. While very fair skin as compared to olive skin was associated with a 6.62 increased risk of single BCC, the OR for multiple BCC was almost 5.5 times greater (OR=36.07, 95% CI=8.95–161.94).
Participants with lighter pigment characteristics, less ability to tan, and more freckles were more likely to have at least one non-synonymous MC1R variant as compared to those with darker phenotypes (Supplemental Table 2 Online).
In the mutually adjusted model, hair and skin color, MC1R, moles, and skin reaction to first summer sun were independently associated with BCC (Table 3). Very fair skin was associated with a 4.48 fold independent increased risk of BCC compared to olive skin (OR=4.48, 95% CI=2.21–9.09) and individuals with two or more non-synonymous variants had a 91% independent increased risk compared to those with no variants (OR=1.91, 95% CI=1.20–3.03).
While there was no evidence of significant interactions between phenotypes and MC1R in relation to BCC risk, we observed some general patterns in risk across strata (Table 4). The association between MC1R and BCC was stronger among individuals with darker phenotypes including, darker eye and skin color, fewer moles and freckles, and tanning rather than burning with sun exposure.
In this case-control study of early-onset BCC, host phenotype characteristics of lighter pigmentation and inability to tan, as well as MC1R were independently associated with increased disease risk. To our knowledge, a large-scale epidemiologic study focused exclusively on BCC among young adults has not been previously reported. In our unique population, the magnitudes of risk associated with phenotype characteristics often associated with BCC were generally magnified as compared to studies in older individuals (Dessinioti et al., 2010; Hogan et al., 1989; Kiiski et al., 2010; Maia et al., 1995; Naldi et al., 2000; Vitasa et al., 1990; Zanetti et al., 1996). Although BCC is relatively rare in young people, 37% of our cases had two or more BCCs under the age of 40, and the association with each of our exposures was much stronger for these cases.
Our finding of a nearly two-fold increase in BCC risk for one non-synonymous MC1R variant and a 3.6 fold increase for two non-synonymous variants is in agreement with other BCC studies (Bastiaens et al., 2001; Box et al., 2001; Dwyer et al., 2004; Han et al., 2006; Scherer et al., 2008). Yet, as we hypothesized, the magnitude of risk we observed was greater than in studies of older adults, where risk estimates have been less than or equal to 2.6 (Bastiaens et al., 2001; Box et al., 2001; Dwyer et al., 2004; Han et al., 2006; Scherer et al., 2008). Of note, one small case-control study with a heterogeneous case group enrolled on the basis of having either familial BCC, multiple BCC, BCC with another cancer, or BCC before age 40, observed a seven-fold increased risk of BCC with two MC1R variants (Liboutet et al., 2006). Similar to our findings, MC1R variants have been associated with lighter pigment phenotypes in numerous studies (Bastiaens et al., 2001; Box et al., 2001; Dwyer et al., 2004; Han et al., 2006; Kanetsky et al., 2004; Kennedy et al., 2001; Koppula et al., 1997; Naysmith et al., 2004; Palmer et al., 2000; Smith et al., 1998; Valverde et al., 1995).
The independent associations with early-onset BCC for MC1R, hair and skin color, moles, and skin reaction are in line with several studies of BCC (Bastiaens et al., 2001; Box et al., 2001; Dwyer et al., 2004; Han et al., 2006; Liboutet et al., 2006; Scherer et al., 2008) and melanoma (Dwyer et al., 2004; Kanetsky et al., 2010; Kennedy et al., 2001; Landi et al., 2005; Palmer et al., 2000). That MC1R remained an independent risk factor suggests variants in this gene contribute to BCC pathogenesis through mechanisms besides pigmentation. As a potential tumor initiator, impairment of MC1R function leads to synthesis of pheomelanin, which acts as a free radical generator and may cause oxidative DNA damage on top of the UVB-induced damage typically associated with sunlight exposure to fair skin (Scherer and Kumar, 2010). Furthermore, mouse studies show an effect of MC1R genotype on production of premalignant clones in the absence of any melanin pigment, suggesting a mechanism separate from pigment modulation (Robinson et al., 2010). A role in tumor progression through regulation of cytokines and their associated receptors, such as NF-kB, has been suggested (Eves et al., 2003; Getting, 2006). There is a complex interplay between NF-kB regulation and the ability of tumor cells to escape immune surveillance and invade surrounding tissues.
The association of pigment phenotypes independent of MC1R genotype points toward the involvement of other pigment-related genes in BCC risk. TYR, ASIP, and SLC45A2 have been identified in other BCC studies (Gudbjartsson et al., 2008; Nan et al., 2009; Scherer and Kumar, 2010; Stacey et al., 2009).
We found that MC1R was more strongly associated with early-onset BCC among those with darker phenotypes. Several other BCC studies have evaluated this genotype-phenotype interaction, but results have been inconsistent. One study observed no clear variation in the association of MC1R and BCC risk by hair or skin color (Han et al., 2006), while others found an increased risk in those with darker hair and skin, but opposite patterns for eye color (Liboutet et al., 2006), or suggestive increased risk among individuals with the lightest skin (Bastiaens et al., 2001; Scherer et al., 2008). One study of BCC (Dwyer et al., 2004) and several of melanoma (Dwyer et al., 2004; Ichii-Jones et al., 1998; Kanetsky et al., 2010; Landi et al., 2005; Palmer et al., 2000) had findings similar to ours.
Our findings, in conjunction with research in melanoma, may have applicability in primary prevention. Among participants with the darkest pigment phenotypes, the estimated etiologic fractions for carrying two or more non-synonymous MC1R variants ranged from a low of 12% among individuals with no freckles on the face to a high of 28% among individuals with brown eyes. Because people with darker phenotypes are also at risk of skin cancer, sun protection interventions may need to be broadened to include these individuals who would otherwise consider themselves low-risk.
Our study had several strengths including extensive self-report phenotype data from a face-to-face interview and MC1R sequencing for nearly all participants. Importantly, the laboratory was blinded to case-control status and interviewers were blinded to case-control status until the end of the interview. Utilizing a centralized dermatopathology facility serving many dermatologists in Connecticut enabled us to identify controls most likely to constitute the source population of our cases; that is, young people who see a dermatologist for a skin lesion. Because our controls had undergone a skin biopsy, this may have reduced differential reporting by case status, as our controls may have been more sensitive to exposures concerning skin health than general population controls. Our results were robust in sensitivity analyses removing specific control conditions, indicating associations were not driven by inclusion of one benign condition.
As in any case-control study, selection bias is a potential concern. Another potential limitation is related to possible misclassification within participant self-reported measures of phenotype, as we did not have more objective measures of pigment characteristics, such as clinician assessment or spectroscopy, but this is most likely to be non-differential. Finally, although controls were seen by a dermatologist for a benign skin condition, we did not know if a complete skin examination was performed. Therefore, controls could have possibly had a BCC; however, the likelihood of this is low in our young sample.
In summary, several host phenotype characteristics and MC1R were strongly and independently related to early-onset BCC. In this young population, the associations between the exposures of interest and disease risk were more pronounced for multiple BCC, and the relationship between MC1R and BCC was stronger among individuals with darker pigmentation phenotypes. Even persons with darker pigment phenotypes, traditionally considered to be low risk of skin cancer, were at substantial risk of early-onset BCC if they had MC1R variants. To our knowledge a large scale epidemiologic investigation of these characteristics in relation early-onset BCC is previously unreported, so our results need confirmation in other populations.
The Yale Study of Skin Health in Young People was conducted in Connecticut between July 2007 and December 2010. BCC cases diagnosed between July 1, 2006 and September 30, 2010 were identified through Yale University’s Dermatopathology database. Approximately two-thirds of dermatologists in Connecticut send their biopsied tissue to Yale for dermatopathologic evaluation. Potential controls were randomly sampled from individuals in the database with a variety of minor benign skin conditions. To be eligible for the study, participants had to: be less than 40 years of age at the time of skin biopsy, reside in Connecticut, speak English, and themselves (or appropriate guardian for decisionally impaired individuals and those under age 18) be mentally and physically capable of completing study components. Yale University’s Institutional Review Board approved the study and participants (or guardians) provided written informed consent. The study was conducted in accordance with the Declaration of Helsinki Principles.
Potential participants were mailed a letter and study brochure. One of two study interviewers then contacted individuals by telephone and invited them to participate in the study. If the telephone number was not working (disconnected, wrong number, no number listed), another letter was mailed asking for updated contact information via telephone or mail. If the telephone number and mailing address were incorrect, updated contact information was sought periodically.
Among the 665 potentially eligible BCC cases identified during the study period, 17 (2.6%) were determined ineligible upon initial contact: 14 moved out of the state and 3 could not complete all study components (two non-English speakers, one severe illness). Of the remaining 648 individuals, 114 (17.6%) could not be contacted directly (no telephone number, non-working telephone number, only spoke to other person in household, left message only). Among the 534 cases we were able to directly reach and determine full eligibility, 145 (27.2%) declined to participate, resulting in 389 enrolled cases (participation rate=72.8%).
Cases were classified into single (only one BCC) or multiple (two or more BCCs) BCC under the age of 40 based on participant self-report and searching the Yale Dermatopathology database (records from 1990 on). A total of 242 (62.2%) cases had one BCC in the database and did not self-report a prior BCC and 120 (30.9%) cases had two or more BCCs in the database. The remaining 27 (6.9%) cases had one BCC in the database, but self-reported a prior BCC; these individuals were categorized as multiple cases, as this did not significantly alter risk estimates.
To determine control eligibility, two dermatologists reviewed skin conditions diagnosed during a one-year period in persons under age 40 in the Yale Dermatopathology database. A variety of diagnoses were determined ineligible for sampling, including skin cancers/precancers (e.g., melanoma, squamous cell carcinoma, T-cell lymphomas, actinic keratoses), potentially UV-related benign conditions (e.g., solar lentigo, abnormal nevus, erythematous conditions), dermal conditions treated with UV therapy (e.g., psoriasis) and pigment disorders (e.g., vitiligo).
Randomly sampled controls were frequency matched to BCC cases on age at biopsy (5 year age groups), gender, and biopsy site (head/neck, trunk, extremity). Among the 1,102 potentially eligible controls, 60 (5.4%) were found ineligible upon initial contact (39 moved out of state, 10 non-English speakers, 2 did not recall having a skin biopsy, 1 hearing impaired, 1 hospitalized) or during the interview (7 self-reported a BCC). Of the remaining 1,042 individuals, 288 (27.6%) could not be contacted directly. Among the 754 potential controls we could directly reach and determine full eligibility, 296 (39.3%) declined to participate and 458 controls enrolled (participation rate=60.7%). Controls had a variety of benign skin conditions. The three most common were cyst (16.4%), seborrheic keratosis (16.2%), and wart (11.4%). All other conditions were present in less than 10% of controls.
Participants completed an in-person face-to-face interview during which interviewers obtained information on sociodemographics, UV exposure (solar and artificial), personal and family medical history, and host phenotype characteristics including, self-reported eye color, skin color (inner upper arm), hair color (natural color), skin reaction to strong sunlight for the first time in the summer for one hour without sunscreen, skin reaction after repeated and prolonged exposure to sunlight, amount of freckles on the face (selected from a range of images), and number of moles on the back = 5 mm (using clear acetate size template) using a structured questionnaire. Interviewers were blinded to case-control status until the end of the interview, when personal history of cancer, including BCC, was queried.
Participants also completed several mailed self-administered questionnaires (residence history, outdoor jobs, attitudes toward sunless, outdoor, and indoor tanning). Interviewers collected buccal cells from 98.9% of participants using Oragene®•DNA 2mL saliva collection kits (DNA Genotek Inc.; Ontario, Canada; http://www.dnagenotek.com/index.html) at the end of the interview following the manufacturer’s protocol, including rinsing the mouth with drinking water and then waiting five minute before collection.
Oragene kits were stored at room temperature until processed. DNA was isolated based on the manufacturer’s protocol. Laboratory personnel were blinded to case-control status.
The MC1R gene was PCR amplified as a single 1.3 kb fragment. Each 25 μl PCR reaction contained 25–50 ng of DNA; 200 μmol/L dNTPs; 5 μmol/L of each primer, 5′-ACTAAGCAGGACACCTGGAG-3′ and 5′-TCTTTAGGAGCCTGAGGTTG-3′; PC2 buffer (50 mM Tris-HCl pH 9.1, 16 mM ammonium sulfate, 3.5 mM MgCl2, and 150 mg/ml BSA; Ab Peptides, Inc.); 0.25 mmol/L spermidine; 0.125 units of Taq DNA polymerase (Amplitaq®, Roche); and 0.125 units of Taq Extender (Stratagene). PCR was performed with an initial denaturation for two minutes at 97°C; followed by 35 cycles of denaturation at 96°C for 30 seconds, annealing at 66°C for 30 seconds, and extension for one minute at 72°C; and a final extension at 72°C for five minutes. PCR products were size fractionated on a 1.5% GPG/LETM (American Bioanalytical) agarose gel, stained with ethidium bromide, and photographed under ultraviolet light in order to confirm the presence of the correct PCR fragment.
PCR products were sequenced bidirectionally. 5 μl of the PCR products were treated with 20 units Exonuclease I (E.coli) (New England BioLabs) and two units Shrimp Alkaline Phosphatase (USB). Either 0.4 μmol/L of the forward primer, 5′-ACTAAGCAGGACACCTGGAG-3′, or the reverse primer 5′-GGTCACACAGGAACCAGACC-3′ were added. The sequencing was carried out at Yale University’s W. M. Keck Facility using Applied Biosystems 3730 capillary instruments. The sequencing reactions utilized fluorescently-labeled dideoxynucleotides (Big Dye Terminators) and Taq FS DNA polymerase in a thermal cycling protocol. The sequence was analyzed using Sequencher 4.9 (Gene Codes Corporation) comparing the query sequence to the standard sequence with no variants in MC1R (NM_002386.3).
MC1R variants were classified into synonymous and non-synonymous variants. Non-synonymous variants were grouped into “major” and “minor” red hair variants (Box et al., 1997; Kanetsky et al., 2004; Valverde et al., 1995). We then calculated the number of total non-synonymous variants within the MC1R coding region.
Analyses were limited to non-Hispanic Whites; 380 (97.7%) cases and 390 (85.2%) controls. Three BCC cases with Gorlin Syndrome, which predisposes individuals to multiple BCCs early in life (Gorlin and Goltz, 1960), were also excluded. Our analytic population consisted of 767 individuals (377 cases, 390 controls); three cases and three controls were under age 18 at enrollment.
Phenotype characteristics and MC1R (count of all non-synonymous variants within the gene, 0, 1, ≥ 2 variants) were treated as categorical variables. We calculated odds ratios (ORs) and 95% confidence intervals (CIs) using multivariate logistic regression for all cases. Analyses were then restricted to cases with only one BCC (n=237) and then cases with two or more BCCs (n=140). We determined independent relationships using backward stepwise selection; retaining only exposures statistically significant at alpha=0.05, as well as gender, age, and body site. Phenotype-genotype interactions were tested with cross-product terms. Analyses were conducted using SAS Version 9.2 (Cary, NC).
This work was supported by the Yale SPORE in Skin Cancer funded by the National Cancer Institute grant number 1 P50 CA121974 (R. Halaban, PI). LMF was supported by a post-doctoral fellowship from the National Cancer Institute - 1F32 CA144335. AMM was supported by CTSA Grant UL1 RR024139 from the National Center for Research Resources. We would like to acknowledge the following individuals for their overall support and assistance with the coordination of this project: Dr. Jennifer McNiff, Robert Criscuolo, and James Platt from Yale Dermatopathology; Dr. Valencia Thomas; and James McCusker from the Biostatistics/Bioinformatics Core of the Yale SPORE. We would also like to recognize and thank our interviewers, Carol Gordon and Lisa Lyon, for their dedication and skill in recruiting and interviewing the study participants. Finally, we are indebted to the individuals who participated in this study.
Conflict of Interest
The authors state no conflict of interest.