Inter99 study population
Clinical data for adults were obtained from the Inter99 study (ClinicalTrials.gov NCT00289237). The Inter99 study is a population-based, randomised, non-pharmacological intervention study for the prevention of ischaemic heart disease, conducted by the Research Centre for Prevention and Health, Glostrup University Hospital, Glostrup, Denmark. Of 13,016 individuals (aged 30–60 years) randomly selected from the Civil Registration System and invited to participate, 6784 (52%) participated in baseline examinations. Detailed phenotypic characteristics from Inter99 have previously been published [19], and baseline characteristics are presented in Table 1.
Table 1 Clinical characteristics of participants
Anthropometric measurements
While wearing light indoor clothes and no shoes, height (cm) and weight (kg) were measured, and BMI was calculated as weight (kg) divided by height squared (m2). Waist and hip circumference were measured in cm, and WHR was calculated as waist measurement (cm) divided by hip measurement (cm).
BP
This was measured using a mercury sphygmomanometer.
Blood sampling
Blood samples were drawn following a 12 h overnight fast, and measures of insulin, blood glucose, HDL-cholesterol, triacylglycerol and total cholesterol were obtained as previously described [19, 20]. All participants without previously diagnosed diabetes underwent a standardised 75 g glucose OGTT, from which participants were diagnosed with type 2 diabetes according to the WHO 1999 criteria. No individuals with previously diagnosed or screen detected type 2 diabetes were included in the present study.
Genotyping
This was performed on 5255 participants from the Inter99 cohort, using the Illumina HumanOmniExpress-24 v1.0_A and HumanOmniExpress-24 v1.1_A (Illumina, San Diego, CA, USA). Genotypes were called using the Genotyping module (version 1.9.4) of GenomeStudio software (version 2011.1; Illumina). Only individuals having a call rate ≥98% were included. Genotypes were phased using Eagle on autosomes and Shapeit on chromosome X and imputed in the Phase 3 1KG and HRC1.1 using the Michigan imputation server (https://imputationserver.sph.umich.edu/index.html) [21]. All variants included in this study were in Hardy–Weinberg equilibrium (p > 0.05).
The Danish Childhood Obesity Biobank study population
Clinical data on Danish children and adolescents was obtained from The Danish Childhood Obesity Biobank (TDCOB; ClinicalTrials.gov NCT00928473). Between March 2007 and March 2013, 1069 children and adolescents (aged 6–18 years) who were overweight or obese were recruited from the Children’s Obesity Clinic, Department of Pediatrics, Copenhagen University Hospital Holbæk as part of the Children’s Obesity Clinic’s Treatment Protocol (TCOCT) [22] (see Table 1 for clinical characteristics). In the following sections, this study sample will be referred to as the TCOCT sample. Overweight was defined as a BMI above the 90th percentile for sex and age according to Danish BMI charts [23] (corresponding to a BMI SD score (SDS) >1.28). All measures included in this study were obtained at the first visit to the clinic, i.e. before treatment initiation. Between September 2010 and March 2013, a population-based sample of 719 children and adolescents (6–18 years) were recruited from local schools and high schools (see Table 1 for clinical characteristics). In the following sections, this study sample will be referred to as the population-based control sample.
Anthropometric measurements
With participants wearing light indoor clothes and no shoes, height was measured by a stadiometer (to the nearest 1 mm), and weight was measured on a digital scale (to the nearest 0.1 kg). BMI was calculated as the weight (kg) divided by the height squared (m2), and BMI SDS was calculated using the least mean squares method [24] based on a Danish reference [23]. Waist circumference was measured at umbilical level in the upright position after exhalation using a stretch-resistant tape (to the nearest 5 mm). The WHR was calculated as the waist measurement (cm) divided by the hip measurement (cm).
BP
Systolic and diastolic BP were measured with an oscillometric device (Omron 705IT; Omron Healthcare, Kyoto, Japan) with the appropriate cuff size, as validated in children [25]. BP was measured three times on the right upper arm after 5 min of rest. An average of the last two measurements was used to calculate systolic and diastolic BP SDS based on sex-, age- and height-specific American references [25].
Blood sampling
Blood samples were drawn from an antecubital vein after an overnight fast. Whole-blood HbA1c was analysed on a Tosoh HPLC G8 analyser (Tosoh Corporation, Tokyo, Japan). Plasma glucose was measured on a Dimension Vista 1500 Analyser (Siemens Healthcare, Erlangen, Germany), and serum insulin, plasma cholesterol, plasma HDL-cholesterol and plasma triacylglycerol on a Cobas 6000 Analyser (Roche Diagnostics, Mannheim, Germany).
Dual-energy x-ray absorptiometry
Measurements taken using dual-energy x-ray absorptiometry (DXA) included fat mass in the arms, legs, torso and whole body. Measures were performed using a GE Lunar Prodigy (DF+10031 GE Healthcare, Little Chalfont, UK) until 14 October 2009 and thereafter using a GE Lunar iDXA (ME+200,179; GE Healthcare).
Genotyping
DNA was extracted at LGC Genomics (Teddington, UK), and samples from all participants (n = 1788) were genotyped using the Illumina Infinium HumanCoreExome Beadchip (Illumina, San Diego, CA, USA) using Illumina’s HiScan system at the Novo Nordisk Foundation Center for Basic Metabolic Research’s laboratory, Symbion, Copenhagen, Denmark. Genotypes were called using the Genotyping module (version 1.9.4) of GenomeStudio software (version 2011.1; Illumina). We excluded individuals who were duplicates or ethnic outliers, or had extreme inbreeding coefficients, mislabelled sex or a call rate of <95%, leaving 1618 individuals. Additional genotypes were imputed using the 1000 genomes phase 1 panel using shapeit/IMPUTE2 pipeline (http://mathgen.stats.ox.ac.uk/impute/impute_v2.html) [26, 27], with only genotyped variants that were not significant (p > 0.05) in Hardy–Weinberg equilibrium tests. Only variants with a high imputation quality (IMPUTE2 estimated R2 > 0.95) were kept.
GRS construction
Genotypes were coded according to the number of IR-increasing alleles based on 53 independent SNPs shown to associate with IR phenotypes (higher fasting insulin concentrations adjusted for BMI, lower HDL-cholesterol concentrations and higher triacylglycerol concentrations) in adults [9] (electronic supplementary material [ESM] Table 1). All genotypes were retrieved from the imputed dataset, and GRS construction was therefore based on genotype dosage information. We constructed an unweighted GRS by summing the number of IR phenotype-increasing alleles. In addition, we constructed a weighted GRS by summing the number of IR phenotype-increasing alleles weighted by the effect size of the variants on fasting insulin concentrations adjusted for BMI, as reported in the validation study in adults [9], and normalised by dividing by the sum of all effects, to make the two GRSs comparable. Similar results were obtained from the two GRSs, and therefore only results from the unweighted GRS (GRS53) are reported.
Statistical analyses
Only children and adolescents from the TDCOB cohort with available information on HOMA-IR were included in our analyses (n = 1364). For clinical characteristics of study participants included in the analyses, see Table 1. All statistical analyses were performed with and without the inclusion of participants who had conditions or were receiving medication potentially influencing IR, such as long-term present or prior systemic use of steroid hormones (n = 29). Statistical analyses were performed using R software (version 3.1.3; R Foundation for Statistical Computing, Boston, MA, USA). HOMA-IR was calculated as ([fasting plasma glucose (mmol/l)] × [fasting serum insulin (pmol/l)])/135 [28]. LDL-cholesterol was calculated according to the Friedewald formula: [LDL-cholesterol (mmol/l) = total cholesterol (mmol/l) − HDL-cholesterol (mmol/l) − triacylglycerol (mmol/l)/5] [29]. Associations between the GRS and IR, metabolic traits and body composition estimates were examined by linear regression using additive genetic models. Analyses were adjusted for sex and age where indicated, and all analyses of DXA measures were adjusted for type of DXA scanner. Quantitative traits deviating from normal distribution were log-transformed (log10) to ensure a normal distribution as assumed in the model. For log-transformed traits, the corresponding p values are reported. Furthermore, clinically interpretable effect sizes and SEs from the analyses of untransformed traits are reported. Binominal tests were performed to assess the directionality of SNP-specific effects. Differences in effect sizes between groups were assessed using a standard two-tailed t-test with β values and SEs for each group. Correction for multiple testing was performed using a false discovery rate (FDR) of 10% [30]. Values of p < 0.05 were considered statistically significant.
Ethical aspects
Written informed consent was obtained from all participants. If they were younger than 18 years, informed oral consent was given by the participant while the parents provided informed written consent. The study was approved by the Danish Data Protection Agency (REG-06-2014), the Ethics Committee of Region Zealand, Denmark (SJ-104) and the Scientific Ethics Committee of the Capital Region of Denmark (KA98155). The study was performed in accordance with the Declaration of Helsinki 2013 and is registered at ClinicalTrials.gov (NCT00928473 and NCT00289237).