|Home | About | Journals | Submit | Contact Us | Français|
We investigated the association between serum 25-hydroxyvitamin D (25(OH)D) levels and basal cell carcinoma (BCC) risk in a nested case-control study at Kaiser Permanente Northern California (KPNC). 220 case patients with BCC diagnosed after serum collection were matched to 220 control subjects. We estimated odds ratios (ORs) and 95% confidence intervals (CIs) using conditional logistic regression. Fully adjusted models included body mass index (BMI), smoking, education, sun-exposure variables, x-ray exposure, and personal history of cancer. For each measure of serum 25(OH)D (continuous, clinically relevant tertiles, quintiles), we found an increased risk of BCC in unadjusted models (OR=1.03, 95% CI 1.00–1.05, p<0.05; OR= 3.98, 95% CI: 1.31–12.31, deficient vs. sufficient, test for trend p value <0.01; OR=2.32, 95% CI: 1.20–4.50, 1st vs. 5th quintile, test for trend p value 0.03). In fully adjusted models, the values attenuated slightly (OR=1.02, 95% CI 1.00–1.05, p<0.05; OR= 3.61, 95% CI: 1.00–13.10, deficient vs. sufficient, t-trend p=0.03; OR=2.09 1st vs. 5th quintile, 95% CI: 0.95–4.58, t-trend p=0.11). Our findings suggest that higher pre-diagnostic serum 25(OH)D levels may be associated with increased risk of subsequent BCC. Further studies to evaluate the effect of sun exposure on BCC and serum 25(OH)D levels may be warranted.
Basal cell carcinoma (BCC) is the most common cancer in the U.S., affecting nearly one million Americans annually (Jemal et al, 2009) and its incidence is rising (Miller and Weinstock, 1994; Christenson et al, 2005). Although BCCs are rarely fatal, their high incidence and the frequent occurrence of new primary BCCs in affected individuals (Karagas et al, 1992) can cause significant morbidity. BCCs also pose a substantial financial impact and are among the most costly cancers to treat in the Medicare population (Housman et al, 2003). The main known risk factors for BCC are sun sensitivity and exposure to ultraviolet (UV) radiation. The same spectrum of UV radiation (280–320 nm) causes both DNA damage to keratinocytes and vitamin D synthesis by these cells (Freeman et al, 1989). This has led some investigators to propose that vitamin D formation in keratinocytes may be an innate protective mechanism against UV damage (Bikle, 2008).
Recent studies have shown that vitamin D can regulate differentiation of normal skin cells (Bikle, 2004) and can reduce hedgehog signaling in and proliferation of murine BCC cell lines (Xiao et al, 2009). In mice, vitamin D receptor (VDR) knock-outs are more susceptible to chemically induced and UVR induced skin tumors, suggesting that disruption of VDR signaling predisposes to skin cancer (Zinser et al, 2002). VDR polymorphisms have been associated with increased BCC risk (Ramchandran et al, 2003). Immunohistochemical studies of human BCCs suggests that the vitamin D pathway may be important for the growth behavior of BCCs (Kamradt et al, 2003; Mitschele et al, 2004). Despite mounting evidence that vitamin D and its receptor are involved in cutaneous carcinogenesis, no studies to date have examined the association of serum vitamin D levels with BCC risk in humans.
The epidemiology of BCCs has been difficult to characterize because most cancer registries, such as the Surveillance, Epidemiology, and End Results (SEER) program, exclude non-melanoma skin cancers. However, the Kaiser Permanente Northern California (KPNC) setting includes electronic databases that capture information on all pathology specimens received for examination, allowing for accurate recording of BCC. Using an established cohort of KPNC members with data on self-reported cancer risk factors and pre-diagnostic serum vitamin D levels, we performed a nested case-control analysis to study the association between prediagnostic serum vitamin D levels and subsequent BCC risk.
The 440 study participants included 228 men and 212 women with a mean age of 54.9 years (SD 10.1, range 28–78). The mean years to BCC among cases was 8.74 years (SD 1.28, min=6.14, max = 11.33). Cases were more likely than controls to report a personal history of cancer and slightly more likely than control subjects to have lower BMI and to have never smoked (Table 1). Cases and controls did not differ with regard to education, any of the sun exposure variables, or history of X-ray exposure. Serum 25(OH)D levels were slightly higher among the cases than among the controls.
The mean 25(OH)D concentration in all study subjects did not differ by baseline characteristics including age at cohort entry, eye color, occupational sun exposure, time spent in leisure activities, smoking status, personal history of cancer, or history of x-ray exposure (Table 2). Subjects with higher 25(OH)D levels were more likely to be male, more educated, and to report no exposure to occupational UV before 1 year prior to taking the Multiphasic Health Checkup (MHC). Those who exercised 2–4 hours per day had higher levels than those who exercised less (0–1 hours/day) while obese patients (BMI ≥ 30 kg/m2) had lower 25(OH)D levels compared to non-obese patients (BMI <25).
In the unadjusted model, a positive association was noted between higher 25(OH)D levels and increased BCC risk whether assessing 25(OH)D as a continuous variable, a categorical variable based on clinical cutoffs, or a categorical variable based on the distribution of 25(OH)D levels among controls divided into quintiles (Table 3). For every 1 ng/ml rise in serum 25(OH)D levels, there was a 3% increase in BCC risk. An added quadratic term for continuous 25(OH)D levels in the unadjusted model was not significant, indicating that a linear model better approximated the relationship between 25(OH)D and BCC risk (data not shown). Individuals who had clinically sufficient 25(OH)D levels (≥ 30 ng/ml) were at increased risk of BCC compared to those who were 25(OH)D deficient (< 10 ng/ml). Similarly, those in the highest quintile had increased risk compared to those in the lowest quintile.
The results did not substantially change when analyzed using the parsimonious model (model 1) or the fully adjusted model (model 2). In both models, each 1 ng/ml increase in serum 25(OH)D levels was associated with a corresponding 2% rise in BCC risk (p<0.05). Subjects with sufficient 25(OH)D levels were still at increased risk of BCC compared to those with deficient levels using the parsimonious and fully adjusted models. The association between serum 25(OH)D divided into quintiles and BCC risk was significant in the parsimonious model (p=0.05, test for trend). Risk estimates in the fully adjusted model using quintiles were slightly attenuated, bordering on statistical significance (p=0.11, test for trend).
The findings from this nested case-control study suggest that higher prediagnostic serum 25(OH)D levels may be associated with increased risk of subsequent BCC. For every 1 ng/ml increase in serum 25(OH)D levels, there was a corresponding 2% increased adjusted risk of BCC. Our data do not support in vitro evidence that suggests that vitamin D may inhibit BCC cell growth (Xiao et al, 2009). However, that growth is inhibited by vitamin D3, and our measurements were of 25(OH)D, which is ineffective in the in vitro assays. To our knowledge, there have been no published papers comparing serum 25(OH)D levels in a population-based sample of individuals with BCC to controls.
UV exposure is a known risk factor for BCCs and is the most readily available source of vitamin D in sunny climates, such as that in the San Francisco Bay Area. A possible explanation of our finding is that the carcinogenic effects of the amount of UV exposure that leads to high serum 25(OH)D levels may overwhelm any possible protective effect of vitamin D noted in vitro. Additionally, residual confounding by UV exposure is possible in this analysis. We attempted to control for sun exposure by reasoning that the exposure would come from two primary sources: time spent in the sun for leisure and exercise and time spent in the sun related to one’s occupation. We therefore used occupational UV and occupational sun exposure as well as time spent in leisure activities and exercise as surrogate markers for sun exposure. However, none of our sun exposure surrogate variables was a significant risk factor for BCC, suggesting that our surrogates did not adequately capture the type of sun exposure (acute, intermittent) that has been reported to be associated with BCC risk (Kricker et al, 1995; Rosso et al, 1998). With the exception of exercise, the sun exposure surrogate variables also were not associated with levels of serum 25(OH)D. In fact, subjects who reported higher occupational UV exposure paradoxically had lower serum 25(OH)D levels. One explanation for this finding may relate to the way the question was phrased, which asked subjects to report whether they worked in a place where they were often or daily exposed to ultraviolet radiation prior to one year before taking the MHC. UV exposure in this time period may not have been relevant to serum 25(OH)D levels over a year later. Also subjects may not have comprehended that ultraviolet radiation is primarily derived from the sun. This lack of comprehension is supported by the fact that the same individuals who reported in the affirmative about “UV exposure” did not have occupational codes that matched high sun-exposure occupations.
Another potential limitation of our study is the length of storage of the samples. However, these samples have been used in the past for multiple serum vitamin D metabolite studies which have documented levels in the expected normal range (Corder et al, 1993; Corder et al, 1995; Hiatt et al, 1998). Indeed, our overall mean serum 25(OH)D level of 24.4 does not differ substantially from the mean serum 25(OH)D level in the U.S. population between 1988–2004 (24–30 ng/ml) as reported by NHANES (Ginde et al, 2009). Also, any systematic bias in the storage of the specimens between cases and controls is unlikely, as cases and controls were matched by serum date.
Attenuation of risk estimates may have been possible if some control subjects had BCC diagnosed outside of the KPNC system. All cases and controls were members during each year of follow-up, and so we believe this is unlikely, as KPNC is a comprehensive healthcare system and members would have had to pay out-of-pocket for services received outside the health plan. It should be noted that the exposures that we studied were obtained at a single point in time and were not measured over the entire study follow-up period. Also, subsequent BCC diagnoses as captured through pathology records were not necessarily incident cases.
The strength of this study is that vitamin D status was assessed up to 11 years prior to the diagnosis of BCC, thereby reducing the likelihood of reverse causality. Our study also has internal validity because both cases and controls were derived from the same prospective cohort. The measurement of serum 25(OH)D levels reflects internal vitamin D status and is considered superior to measures of vitamin D intake by dietary questionaires alone or predictors of vitamin D status. Of note, previous reports of dietary intake of vitamin D and BCC risk have found no association (van Dam et al, 2000; Gandini et al, 2009). Also, serum 25(OH)D is a measure of vitamin D levels over the past several weeks to several months and is a valid measure of steady-state levels (Holick 1990). Further, to minimize misclassification of vitamin D status due to seasonal variation, we matched cases and controls by season of blood draw (± 1 month). Finally, men and women with a diverse age range (28–78 years) were included in the analysis, making our results more generalizable to whites in the U.S. population.
In summary, we observed an increased BCC risk with higher pre-diagnostic serum vitamin D levels. It is likely that sun exposure, especially acute intermittent exposure, confounds this association. Our findings may also have been influenced by other variables not ascertained in our study, such as supplemental vitamin D use or healthcare screening bias. Future studies that can accurately measure acute intermittent sun exposure, supplemental vitamin D use, and other potential confounding factors may be warranted.
The source population consisted of members of KPNC who had completed a Multiphasic Health Checkup (MHC) between August 1968 and January 1970. The MHC was a voluntary, comprehensive health evaluation that included a detailed self-administered MHC questionnaire (MHCQ), a standardized physical examination, and a group of specialty examinations and laboratory tests administered to a total of 206,974 KPNC members between 1964–1973. MHC participants were instructed to fast overnight prior to a blood collection that was used for routine screening. Details about the MHC have been previously published (Collen and Davis 1969) and the cohort has been used for numerous risk factor studies (Hiatt and Freidman, 1982; Selby et al, 1988; Alexander et al, 1995; Corley et al, 2008). The sera from the collections were stored at −23°C or colder until 1980, when they were shipped to a frozen storage facility (−40°C) at the Orentreich Foundation for the Advancement of Science (OFAS), Inc. 25(OH)D levels in these stored sera have been shown to be stable (Corder et al, 1993; Corder et al, 199; Hiatt et al, 1998). Pathology records for BCC diagnosis were from KPNC's medical center in Oakland, CA, which had computer-stored pathology records starting in 1974.
A total of 3,164 subjects with histologically confirmed BCCs diagnosed between 1974–1989 have been previously identified (Friedman and Tekawa, 2000). These individuals were identified by examining pathology records classified using Systematized Nomenclature of Human and Veterinary Medicine (SNOMED) codes (Cote, 1993). We limited our time period for case selection to those individuals who completed the MHCQ between 1968–1970 to minimize the time interval between serum draw and BCC development. Power calculations indicated that a sample size of 220 cases and 220 controls would result in a minimum detectable pattern of ORs of 1.0, 1.21, 1.47, 1.78 and 2.16 for Q1 (quintile 1; referent), Q2, Q3, Q4, and Q5 respectively (test for trend, two sided test; alpha=0.05; power=0.80).
Our inclusion criteria for cases (n=220) were: (1) "white" skin color, as determined by MHC staff (categories of “yellow” and “brown” excluded), (2) previously unused serum sample associated with a specific MHC visit between 1968–1970, (3) BCC diagnosed based on SNOMED morphology code between January 1974 and December 31, 1979 and (4) active KPNC membership each year during the entire follow-up period.
Using incidence density sampling (Rothman and Greenwald, 1998) control subjects were matched 1:1 to cases by age (± 1 year), skin color (white), sex, eye color, date of serum collection (± 1 month to control for seasonality), MHC location and length of KPNC membership. All control subjects also had at least one unused serum sample available for analysis with a specific MHC visit between 1968 and 1970 and had active KPNC membership each year during the follow-up period. The study was approved by the Institutional Review Board of KPNC and the Declaration of Helskinki protocols were followed. The requirement for informed conset was waved.
De-identified aliquoted frozen samples were sent from OFAS (Cold Spring-on-Hudson, New York) to Heartland Assays (Ames, IA) and analyzed using the DiaSorin LIAISION 25(OH) Vitamin D Total Assay which includes 25(OH)D metabolites of vitamin D2 and D3 from plant and animal foods, as well as that synthesized endogenously (Wagner et al, 2009). Case and control specimens were handled in the same standard manner, and laboratory personnel were blinded to case-control status. All assays were repeated for reliability. The coefficients of variation for 25(OH)D samples were intra-assay 8.4% and inter-assay 11.4 %.
We examined serum 25(OH)D levels as: 1) a continuous variable, 2) a categorical variable divided into quintiles based on distribution among controls, and 3) a categorical variable divided into clinically accepted cutoffs (deficient: <10 ng/ml; insufficient: 10 to <30 ng/ml, sufficient: ≥ 30 ng/ml) (Kricker et al, 1995; http://ods.od.nih.gov/factsheets/vitamind.asp, accessed July 20, 2009;).
All participants in the MHC were had completed the MHCQ that included information on age, education, current and past smoking behavior, history of cancer, occupation, and occupational exposures. Information on possible BCC risk factors was obtained from each subject's MHCQ and from information recorded by the MHC staff (height, weight, eye color, skin color). When BMI data were missing from the MHCQ administered on the serum draw date (n=20), we imputed a value based on the cohort mean value and included a missing data flag in all models. When smoking information was missing, we obtained smoking status from previous pre-diagnostic MHCQs, if available. When data on education were missing, we used the highest education level recorded among all available MHCQ, hypothesizing that educational status would not have changed significantly in the cohort between 1968 and 1970. We found education level differences between questionnaires in only 2 cases and 4 controls.
We analyzed 4 variables which served as surrogates of sun exposure: time spent in leisure activities (ordered categorical variable), time spent in exercise (ordered categorical variable), prior occupational exposure to ultraviolet radiation (yes/no), and occupational sun exposure derived from self-reported occupation (categorical). For each reported occupation, we assigned a sun exposure level (low, moderate, high) based on the standard duties of that occupation. Thus, for example, a mail carrier was deemed to have high occupational sun exposure whereas an office clerk was deemed to have low exposure. Sun exposure levels were reviewed and agreed by the co-authors (MA, JT, MW, GF) and outside investigators with relevant expertise.
Our overall analytic strategy was to: (1) compare baseline characteristics of cases and controls (Table 1), (2) determine the association between covariables and 25(OH)D levels in the entire sample (Table 2), (3) examine the association between 25(OH)D levels and BCC risk, controlling for identified confounding variables, and (4) determine the association of serum 25(OH)D levels to BCC risk, controlling for hypothesized potentially confounding variables. We used conditional logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for BCC risk. We developed a multivariate model that included only the variables associated with both 25(OH)D levels and BCC risk at the p ≤ 0.2 level; i.e.: BMI, and smoking status (parsimonious multivariable model—model 1). We also developed a multivariate model in which all potentially confounding variables were included (fully adjusted multivariate model--model 2). Tests for linear trend (1 df) were conducted by treating the ordered categorical values of the exposure categories as continuous variables. Statistical analyses were performed using SAS, version 9.1, (SAS Institute Inc., Cary, NC).
Nancy P. Durr of OFAS for editorial assistance.
FUNDING: National Institute of Arthritis Musculoskeletal and Skin Diseases (K23 AR 051037, M.A.; R01AR050023, D.B.; K24 AR052667, M.C.); Kaiser Foundation Research Institute (Community Benefits Grant 9601, M.A.), American Institute of Cancer Research (07A140, D.B.), Veterans Administration Merit Review (D.B.)
CONFLICT OF INTEREST: None