The Women's Health Study (WHS) is a randomized, controlled, 2×2 factorial trial which consisted of 39,876 female health professionals whose ages were 45 years or older and were free of cancer (except non-melanoma skin cancer) and cardiovascular disease at baseline [16
]. For the present analysis we conducted a nested case-control study using risk-set sampling. Cases and controls were selected from among original trial participants whose baseline blood samples were available (n=28,345) and who were postmenopausal and not using hormone replacement therapy at the time of blood collection (n=6,574). Cases included all women who developed type 2 diabetes by February 2005 (median follow-up of 10 years). Controls were matched to cases in a 1:1 ratio by age (within 1 year), follow-up duration (within 1 month), race, and fasting status at blood collection (72% provided fasting blood, defined as ≥10 hours since last meal). Based on these criteria, 359 cases and 359 controls were eligible for analysis in WHS.
The Physicians' Health Study II (PHS) is a randomized, controlled 2×2×2×2 factorial trial consisting of 14,641 male physicians aged 50.5 years or older who were free of cancer and cardiovascular disease at baseline [18
]. In a similar manner as was done for WHS, cases and controls were selected from among original trial participants whose baseline blood samples were available (n=11,133). Cases included all men who developed type 2 diabetes during 8 years of follow-up. Controls were matched to cases in a 1:1 ratio by age (within 2 years), follow-up duration (within 1 month), race, and time of blood draw. Among the PHS cases and controls selected for the present analysis, 57% provided fasting blood (defined as ≥10 hours since their last meal). Based on these criteria, 170 cases and 170 controls were eligible for analysis in PHS.
Written informed consent was obtained from all participants in both WHS and PHS. Both studies were conducted according to the ethical guidelines of Brigham and Women's Hospital, Harvard Medical School, and the UCLA institutional review boards.
Methods used to ascertain incident type 2 diabetes in the present analyses have been described in detail previously [19
]. Briefly, participants with type 2 diabetes at baseline were excluded. The remaining participants were asked annually whether and when they had been diagnosed with type 2 diabetes during follow-up. Self-reported cases of type 2 diabetes were confirmed by supplementary questionnaires. A small validation study in WHS compared self-reported type 2 diabetes to physician-led telephone interviews, supplementary questionnaires, and medical record reviews; the positive predictive value of self-reported type 2 diabetes was at least 91% in each comparison [20
Centrifuged blood samples were stored in liquid nitrogen freezers. Case-control pairs were handled identically. Laboratory personnel were blinded to case-control status. Each matched pair was assayed in random order and in the same analytical run. Plasma levels of resistin were measured by an enzymatically amplified “two-step” sandwich-type ELISA immunoassay (R&D Systems, Minneapolis, MN). TNFα-receptor II (TNF-RII) was also measured by ELISA (R&D Systems, Minneapolis, MN). C-reactive protein (CRP) was measured at a core laboratory facility using a validated, high-sensitivity assay (Denka Seiken) [21
]. The CRP assay was conducted only in WHS.
Two RETN SNPs–rs 1862513 (RETN -420C>G) and rs34861192 (RETN -638G>A)–were successfully genotyped in >98% of the samples in both WHS and PHS (providing 96% of the total case-control pairs for genetic analysis). Polymorphism selection was based on a review of the past literature and HapMap. Priority was given to polymorphisms that were associated with differences in plasma resistin levels, associated with type 2 diabetes, had a relatively high minor allele frequency, and found in the RETN gene promoter region. All DNA samples were genotyped using the ABI Taqman system (Applied Biosystems, Foster City, CA). Endpoint fluorescence of PCR-amplified DNA was read by the ABI PRISM 7900HT Sequence Detection System (SDS). Genotypes were assigned using SDS 2.2.2 Allelic Discrimination Software (Applied Biosystems, Foster City, CA) by two independent technicians blinded to the identification number and case-control status of each sample. Predesigned Taqman SNP genotyping probes were used for rs1862513, and custom probes were designed for rs34861192 (Applied Biosystems, Foster City, CA).
Biomarker values with skewed distributions were log-transformed to enhance compliance with normality assumptions. McNemar's χ2
test (for categorical variables) and Wilcoxon signed rank test (for continuous variables) were used to compare baseline characteristics between cases and controls. We calculated geometric means for biomarkers and arithmetic means for BMI by genotype. Arithmetic means for BMI were multivariably adjusted for physical activity levels, smoking status, alcohol intake, and ever having a first-degree relative with diabetes. Geometric means for biomarkers were further adjusted for BMI. Differences in genotype-specific means were assessed by generalized linear models. Conditional logistic regression provided odds ratios (OR) and 95% CI on the association between RETN
SNPs and type 2 diabetes. ORs in the present analyses yield unbiased estimates of relative risks (RR), specifically rate ratios, since risk-set sampling was used. Estimates from the WHS and PHS were pooled using a random-effects meta-analysis model [22
] (Stata 10.0, StataCorp, Texas Station, TX). All other analyses were conducted in SAS 9.1.3 (SAS Institute Inc., Cary, NC).