Common Obesity-Related Genetic Variants and Papillary Thyroid Cancer Risk

Materials and Methods

The study population has been previously described (6). In brief, cases included individuals diagnosed with incident, histologically confirmed PTC during follow-up of the U.S. Radiologic Technologists (USRT) cohort (n = 202), and individuals diagnosed and treated for PTC at the University of Texas MD Anderson Cancer Center (UTMDACC; Houston, TX; n = 142). In USRT, controls (n = 452) were frequency matched by race, year of birth (±2 years), and sex to cases. Controls from USRT were then selected to match cases from UTMDACC. Analyses were restricted to non-Hispanic whites. Three cases and 8 controls were excluded because of missing height or weight. The Institutional Review Boards approved the use of these data, and all subjects provided written informed consent.

The 23 genes chosen for this analysis (listed in Supplementary Table S1) were selected a priori. Tag SNPs (n = 575) were selected from the common SNPs (minor allele frequency >5%) genotyped by the HapMap Project in the Caucasian population using TagZilla, part of the GLU software package, with a binning threshold of r² > 0.8. Genotyping was carried out at the National Cancer Institute (NCI) Core Genotyping Facility using a custom-designed iSelect Infinium assay. SNPs were excluded if they failed quality-control measures: less than 95% concordance, less than 90% completion, or had evidence of a departure from Hardy–Weinberg equilibrium in controls (P < 0.00001). Allele frequencies were largely similar between USRT and UTMDACC cases; thus, these groups were combined for analyses.

Data on demographics, medical history, anthropometry, and other health-related characteristics were collected by self-administered questionnaires or telephone interview in USRT and self-administered questionnaire at time of blood collection in UTMDACC.

We computed SNP-specific P_trend and ORs and 95% confidence intervals (CI) for each genotype, using logistic regression models adjusted for sex, attained age, and year of birth. Separate models additionally adjusted for body mass index (BMI). We also examined 138,605 2-way SNP–SNP interactions using allelic-based gene–gene interactions in models adjusted for sex, attained age, year of birth, and BMI (7). We combined SNP-specific P_trend into region-based P values using the adaptive rank–truncated method (8). P values less than 0.05 were considered statistically significant, and tests were 2-sided. While tables show uncorrected P values, we also conducted correction for multiple comparisons controlling the false discovery rate (FDR). Statistical analyses were conducted using Stata/SE version 11.0 and R software.