Study subjects were recruited from 8 population-based cancer registries in New Jersey, California and North Carolina (USA), British Columbia and Ontario (Canada), Torino (Italy) and Tasmania and New South Wales (Australia). Recruitment was done as part of the GEM (Genes and Environment in Melanoma) study, an international multi-center, population-based study of multiple (second or subsequent) primary melanoma (MPM) compared to single primary melanoma (SPM).
As part of this study, genetic information (DNA) from each patient and detailed information relating to patients’ characteristics (e.g., age, sex, skin/hair/eye color, tanning ability, freckling as a child, number of nevi), family history of skin cancer, past sun exposure, and tumor histology were collected for all participating subjects with incident primary melanoma in 8 population centers in North America, Europe and Australia. Further details of the GEM study design are given elsewhere [27
The GEM study protocol was approved by the Institutional Ethics Review Board at the GEM coordinating center, Memorial Sloan-Kettering Cancer Center in New York, and at each of the study centers. All participants provided written informed consent. Separate approval was obtained at each center for this study.
GEM controls were people diagnosed with a pathologically confirmed first invasive primary melanoma during the six-month period January 1, 2000-June 30, 2000 with the following exceptions: the whole of 2000 in California and North Carolina; from January 1, 2000 to August 31, 2000 in Ontario; and from June 1, 2000 to May 31, 2001 in Turin, Italy. GEM cases were people diagnosed with a pathologically confirmed second or higher order invasive or in situ melanoma during the period January 1, 2000 to August 31, 2003, except in Ontario where case ascertainment ended February 28, 2003, and the centers in British Columbia, California, New Jersey and Tasmania, which recruited GEM Cases additionally in 1998 and 1999.
For the purposes of this analysis, we examined the three major types of sun exposure from our analysis of the relationship between solar exposure and melanoma risk [28
] (1) ambient erythemal ultraviolet (UV) radiation dose at age 10, chosen to represent early lifetime sun exposure, (2) sunny vacations, at a place sunnier than usual, as average annual hours of exposure per year over the lifetime from age 5 to diagnosis, and (3) beach and waterside exposure as average hours per year from age 15, over the lifetime. Each of these exposure types has been shown to be associated with melanoma risk in previous studies [29
A complete description of data collection and estimation relating to sun exposure variables has been previously published [28
]. In brief, erythemally weighted solar ultraviolet irradiance (UVE) was estimated in KJ/m2
for each place of residence, using satellite-derived data. An estimate of UVE was assigned to each year of age, using residence information for the decade years of age, and UVE exposure at age 10 was used for these analyses. Data regarding beach and waterside exposure was elicited from age 15 to the time of diagnosis if an activity was reported between the hours of 9 and 5 on at least 10 days in any year since leaving school. If study participants did participate in beach or waterside activities, they were asked the years started and stopped and the usual outdoor hours per day by season. The total lifetime hours of exposure in these activities were the sum of all reported daily exposure hours weighted by frequency and duration. Sunny vacations reported over the lifetime were calculated as hours per year in the same manner although they were calculated from age 5 to diagnosis
The Molecular Epidemiology Laboratory of the Memorial Sloan-Kettering Cancer Center typed the VDR FokI
polymorphisms. DNA was extracted from buccal cells using Puregene® kits (Gentra Systems, Inc., Minneapolis, MN) replacing glycogen with tRNA (10μg/μl) for the DNA precipitation step. All genotyping was done with PCR-based methods and included melting temperature analysis [34
] coupled to the LightTyper instrument (Roche Applied Science, Indianapolis, USA) for the analysis of the FokI
SNP and pyrosequencing [35
] with the PSQ™ MA instrument (Biotage AB, Uppsala, Sweden) for the analysis of BsmI
The VDR-FokI specific fragments (267 bp) were amplified in a PCR reaction mix containing 10-100 ng DNA, 200 μM dNTPs, 0.4 μM forward primer (5′-CTGAGCCAGCTATGTAGGGC-3′), 2.0 μM reverse primer (5′-GGTCAAAGTCTCCAGGGTCA-3′), 0.2 μM fluorescein labeled probe (5′-CTTGCTGTTCTTACAGGGACGGAG-3′), 1.5 mM MgCl2, 1M betaine and 0.05U/μl of Taq Polymerase. The cycling conditions included a denaturation and Taq activation step at 95°C for 10 minutes followed by 5 cycles at 95°C-25 seconds, 64°C-20 seconds, 72°C-30 seconds, 5 cycles at 95°C-25 seconds, 60°C-20 seconds, 72°C-30 seconds, 40 cycles at 95°C-25 seconds, 56°C-20 seconds, 72°C-30 seconds, and a post cycling extension at 72°C for 5 minutes.
The VDR-BsmI specific fragments (209bp) were amplified in a PCR reaction mix containing 10-100 ng genomic DNA, 1x buffer II (10mM Tris-HCl, pH 8.3, 50mM KCl) (PE, Roche Molecular Systems Inc., Branchburg, NJ), 200 uM dNTP, 0.42 uM forward primer (5′-CCTCACTGCCCTTAGCTCTG-3′) and reverse primer (5′Biotin-CCATCTCTCAGGCTCCAAAG-3′), 2.5mM MgCl2, 5% DMSO and 0.05U/ul Taq Polymerase. Cycling conditions included a denaturation step at 95°C for 5 minutes followed by 2 cycles at 95°C-20 seconds, 59°C-20 seconds, 72°C-25 seconds, 40 cycles at 95°C-20 seconds, 57°C-20 seconds, 72°C-25 seconds, and a final extension at 72°C for 5 minutes. For the pyrosequencing reaction, the sequencing primer (5′-CCACAGACAGGCC-3′) was added to single stranded DNA.
The output of the genotyping assays consisted of melting profiles (FokI SNP) and pyrograms (BsmI SNP), and in addition to the automatic genotype calls obtained by the software, the laboratory members reviewed individual signals manually. All genotyping assays included known internal controls (homozygous wild type and variant, and heterozygous DNAs) and blanks (water). For quality control, 10% of samples were split, relabeled, and re-analyzed. All results were interpreted at least twice by two different laboratory members. Assays were considered acceptable when all the control and water samples tested showed the expected genotype or no signal, respectively; there was 100% agreement in the genotyping calls between two independent laboratory members; and there was 100% concordance for the random selected samples tested in two independent assays. Quality control for data entry included an additional review of at least 20% of genotype calls.
Direct sequencing was performed using an independent PCR reaction to confirm the genotype of laboratory control samples, in randomly selected cases, or when a new SNP was identified by different sequence pattern in the pyrogram, or different melting profile in the LightTyper. Amplified samples were resolved on agarose gels, and specific bandswere excised and purified according to the manufacturer’s recommendations (Qiagen Inc., Valencia, USA). The purified DNA was sequenced in the Sequencing Facility Core of Memorial Sloan-Kettering Cancer Center on an ABI377 instrument (PE-Applied Biosystems).
Univariate and multivariate statistical analyses were undertaken using SAS Statistical Packages Version 9.2 (SAS Institute, Cary, NC). Homogeneity of and trend in odds ratios across strata were tested using StatXact Version 8.0.
Age at diagnosis was defined as age at first melanoma diagnosis for subjects with single primary melanoma and age at most recent diagnosis for subjects with multiple primary melanoma. A multiplicative age-sex interaction term was included in all models to control for potential confounding effects in recognition of the fact that the age incidence curves for melanoma are markedly different for males versus females.
Sun exposure variables were dichotomized using cut points based on the exposure distribution in all subjects combined. Phenotypic and demographic covariates were similarly categorized based on their distributions.
Study center was included as a covariate in all models to account for unmeasured differences among populations; control for study center has been demonstrated to adequately control for ancestry in these data (R. Millikan, personal communication). Conventional methods of analysis for case-control studies were followed. Individuals who developed a second primary melanoma during the ascertainment period were treated as both cases and controls. Descriptive statistics stratified by case status were calculated for all VDR genotypes, sun exposure variables, and phenotypic and demographic characteristics. Chi-square testing was performed to ensure that genotype frequencies did not differ greatly from those expected under Hardy-Weinberg equilibrium (HWE).
Unconditional logistic regression was used to estimate odds ratios for multiple primary melanoma and each VDR genotypic and sun exposure covariate separately, controlling for age, sex, a sex-age interaction and center. Ability to tan was included as a measure of phenotypic susceptibility, and family history was included because of the large difference between cases and controls in its prevalence. All three genotypes were incorporated into the model together, with homozygous wildtype genotypes chosen as the referent (FF for FokI and bb for BsmI). To test for the relationship between the VDR gene variants and sun exposure, separate models for each of the genetic variants (FokI, BsmI) and each of the sun exposure measures were constructed. Main effects, stratified and joint effects analyses were conducted. Interactions were identified using a likelihood ratio test at the alpha 0.1 level. Significance tests were two-sided and a Pvalue of less than 0.05 was considered significant.