The study was conducted in accordance with the Helsinki Declaration and approved by the Ethical Committees of Copenhagen and Aarhus.
Participants were of Danish nationality and prior to participation, a written informed consent was obtained from all individuals. Participants examined in the present study were from four different study populations: 1) The Inter99 study is a population-based randomized non-pharmacological intervention study for the prevention of ischaemic heart disease from the Research Centre for Prevention and Health in Glostrup, Denmark (ClinicalTrials.gov NTC00289237) . A total of 6,162 participants with available genotypes for rs1209523 were classified as having normal glucose tolerance (NGT) (n = 4,567), impaired fasting glycemia (n = 508), impaired glucose tolerance (n = 707), screen-detected T2D (n = 256), or previously diagnosed T2D (n = 124); 2) T2D patients recruited at Steno Diabetes Center (SDC) (n = 1,695); 3) A random sample of middle-aged glucose-tolerant participants examined at SDC (n = 730) and 4) T2D patients from the population-based, high-risk Addition Denmark screening and intervention study cohort (n = 1,609) (Anglo-Danish-Dutch Study of Intensive Treatment in People with Screen-Detected Diabetes in Primary Care) (ClinicalTrials.gov ID-no: NCT00237548). A standard 75 g oral glucose tolerance test (OGTT) was performed in participants from study group 1 and 3. T2D was diagnosed according to World Health Organization 1999 criteria.
Analyses of quantitative diabetes-related traits were performed in glucose-tolerant individuals (n = 4,567), as well as in non-obese (BMI < 30 kg/m2) individuals (n = 4,022) from study population 1. All T2D patients and glucose-tolerant individuals were included in the case-control study of T2D (n = 10,196). In the T2D case-control study of non-obese individuals, study participants with a BMI above 30 were excluded for both T2D individuals and glucose-tolerant controls.
The rs1209523 of FOXA2 was genotyped using KASPar® SNP Genotyping system (KBioscience, Hoddesdon, UK). The success rate was 97.3% with a 0.0% error rate estimated from re-genotyping of 972 replicate samples. The genotype distribution obeyed Hardy-Weinberg equilibrium in all study populations (p > 0.14).
Biochemical and anthropometric measures
Height and weight were measured in light indoor clothing and without shoes. Hip circumference was measured at its maximum, and waist circumference was measured in the upright position midway between the iliac crest and the lower costal margin  BMI was calculated as weight (kg)/height2 (m2). Blood samples were drawn after a 12 h overnight fast. A glucose oxidase method was used to analyze plasma glucose (Granutest; Merck, Darmstadt, Germany). Serum insulin (excluding des-31,32 and intact proinsulin) was measured using the Autodelfia insulin kit (Perkin-Elmer/Wallac, Turku, Finland). Serum C-peptide concentrations were measured by a time-resolved fluoroimmunoassay (Auto-DELFIA C-peptide kit; Perkin-Elmer/Wallac, Turku, Finland). Homeostasis model assessment of insulin resistance (HOMA-IR) was calculated as: (fasting plasma glucose (mmol/l) × fasting serum insulin (pmol/l))/22.5, and homeostasis model assessment of β-cell function (HOMA-B) was calculated as: (20 × fasting serum insulin (pmol/l))/(fasting plasma glucose (mmol/l) - 3.5) . Information on sex, BMI, plasma glucose levels, and serum insulin levels to time points 0, 30, and 120 min during an OGTT is used to calculate the BIGTT-acute insulin response (BIGTT-AIR) as well as the BIGTT-insulin sensitivity index (BIGTT-Si). These indices highly correlate with those obtained during an intravenous glucose tolerance test. The calculations were performed as previously described . Insulinogenic index was calculated as: (serum insulin at 30 min (pmol/l) - fasting serum insulin (pmol/l))/plasma glucose at 30 min (mmol/l). Disposition index was calculated as insulinogenic index/HOMA-IR, and Matsuda whole body insulin sensitivity index (ISIMatsuda) was calculated as (10,000/√ (fasting plasma glucose × fasting serum insulin) × (mean plasma glucose × mean serum insulin during OGTT)) . The trapezoidal method was used to estimate the area under the curve (AUC) for plasma glucose, serum insulin and serum C-peptide, and the AUC for insulin/AUC for glucose ratio was calculated as AUC for insulin/AUC for glucose.
All statistical analyses were performed using R statistical software version 2.12.1 (available at http://www.rproject.org). A general linear model was applied to test quantitative traits in relation to genotype, using an additive genetic model and adjusting for age, sex, and BMI where appropriate. Prior to analyses, non-normally distributed data (measures of serum insulin, serum C-peptide levels, insulinogenic index, HOMA-IR, ISIMatsuda, AUC for insulin/AUC for glucose ratio, and BIGTT-AIR) were logarithmically transformed. Logistic regression was used to compare allele frequencies in the case-control analysis, and the analysis was adjusted for age, sex, and BMI. A p-value of less than 0.05 was considered statistically significant. The statistical power for detecting an effect on fasting plasma glucose of - 0.07205 mmol/l per allele corresponding to -1.31 mg/dl found by Xing and colleagues  was estimated using 1,000 simulations and a significance threshold of 0.05. Based on the allele frequency of the variant and the sample size of 4,368 non-diabetic successfully genotyped individuals, we estimated a statistical power of 93% to detect an association. For comparison, the statistical power to detect an effect on fasting plasma glucose of 0.06, 0.05, 0.04, or 0.03 mmol/l per allele were 79%, 66%, 46%, or 30%, respectively.
The statistical power calculation for the case-control analysis was done using CaTS, power calculations for large genetic association studies http://www.sph.umich.edu/csg/abecasis/cats/, and the statistical power to detect an OR of 0.85, 0.90, or 0.95 for rs1209523 of FOXA2 was estimated to be 50%, 26%, or 10%, respectively (significance level: p < 0.05, minor allele frequency (MAF) = 4%, estimated T2D prevalence in the background population = 0.08).
FASTSNP (Function Analysis and Selection Tool for Single Nucleotide Polymorphisms; http://fastsnp.ibms.sinica.edu.tw) was used to predict the function of rs1209523 in FOXA2.
Data on glycemic traits have been contributed by MAGIC investigators and have been downloaded from http://www.magicinvestigators.org. Data on T2D were available through the DIAGRAM meta-analysis .