Dietary Intake, FTO Genetic Variants, and Adiposity

A Combined Analysis of Over 16,000 Children and Adolescents

Qibin Qi; Mary K. Downer; Tuomas O. Kilpeläinen; H. Rob Taal; Sheila J. Barton; Ioanna Ntalla; Marie Standl; Vesna Boraska; Ville Huikari; Jessica C. Kiefte-de Jong; Antje Körner; Timo A. Lakka; Gaifen Liu; Jessica Magnusson; Masayuki Okuda; Olli Raitakari; Rebecca Richmond; Robert A. Scott; Mark E.S. Bailey; Kathrin Scheuermann; John W. Holloway; Hazel Inskip; Carmen R. Isasi; Yasmin Mossavar-Rahmani; Vincent W.V. Jaddoe; Jaana Laitinen; Virpi Lindi; Erik Melén; Yannis Pitsiladis; Niina Pitkänen; Harold Snieder; Joachim Heinrich; Nicholas J. Timpson; Tao Wang; Hinoda Yuji; Eleftheria Zeggini; George V. Dedoussis; Robert C. Kaplan; Judith Wylie-Rosett; Ruth J.F. Loos; Frank B. Hu; Lu Qi

Disclosures

Diabetes. 2015;64(7):2467-2476. 

In This Article

Research Design and Methods

Study Participants

The current analysis included cross-sectional data on 16,094 children and adolescents (15,352 whites, 478 African Americans, and 267 Asians) aged 1–18 years from 14 studies (Supplementary Table 1 http://diabetes.diabetesjournals.org/content/64/7/2467/suppl/DC1). The study design, recruitment of participants, and data collection of individual studies have been described in detail previously.[14,23,24,40–50] In each study, informed consent was obtained from subjects' parents or guardians and subjects (if appropriate). Each study was reviewed and approved by the local institutional review board.

Study-specific characteristics for each study are shown in Supplementary Table 2 http://diabetes.diabetesjournals.org/content/64/7/2467/suppl/DC1. The ranges of mean values across studies were as follows: age 1.1–16.4 years; BMI 16.2–24.7 kg/m2; total energy intake 1,017–2,423 kcal/day; total protein intake 12.9–16.8% (percentage of total energy intake); total carbohydrate 43.4–59.0%; and total fat intake 28.1–40.0%.

Assessment of BMI and Dietary Intake

BMI was calculated as body weight (kg)/height (m2). Body weight and height were measured in all studies except for one study which used self-reported data in a subsample (Supplementary Table 3 http://diabetes.diabetesjournals.org/content/64/7/2467/suppl/DC1). For two studies[43,48] with children younger than 2 years of age, length (height) was measured to the nearest millimeter with children in a supine position. Dietary intake (total energy, protein, carbohydrate, and fat) was assessed using validated food frequency questionnaires (four studies), multiple-day dietary/food records (three studies), multiple-day 24-h recalls (four studies), both dietary records and 24-h recalls (one study), diet history determined by consulting and information system (one study), or a brief-type self-administered diet history questionnaire (one study) (Supplementary Table 3 http://diabetes.diabetesjournals.org/content/64/7/2467/suppl/DC1). Macronutrient intake was expressed as the percentage of total energy intake.

Genotyping

FTO SNP rs9939609 or a proxy (linkage disequilibrium r 2 = 1 in the corresponding ethnic group) was genotyped using direct genotyping methods or Illumina genome-wide genotyping arrays, or imputed using MACH (http://csg.sph.umich.edu/abecasis/MACH/) with a high imputation quality (r 2 = 1) (Supplementary Table 4 http://diabetes.diabetesjournals.org/content/64/7/2467/suppl/DC1). The studies provided summary statistics based on data that met their quality control criteria for genotyping call rate, concordance in duplicate samples, and Hardy-Weinberg equilibrium P value.

Statistical Analysis

A standardized analytical plan, which is described below, was sent to study analysts from the 14 studies, and analyses were performed locally. BMI was transformed to age-standardized z score by sex in each study before analysis. A linear regression model under additive allelic effects was applied to examine associations of FTO variant with BMI, total energy intake, and intake of fat, protein, and carbohydrate (expressed as the percentage of total energy), adjusted for pubertal status (if available), physical activity (if available), and eigenvectors (data from GWASs only). We additionally adjusted for BMI when evaluating the association between FTO variant and dietary intake. In addition, the difference in BMI between the low– and high–dietary intake groups (dichotomized at medians in each study) was also examined. Interactions between FTO genotype and dietary intake and their effect on BMI were tested by including the respective interaction terms in the models (e.g., interaction term = rs9939609 SNP × total energy intake [dichotomized at the medians in each study]). We examined the association between FTO variant and BMI stratified by low– and high–dietary intake groups (dichotomized at medians in each study). All of the analyses were conducted in boys and girls separately, except for one study that combined the data from boys and girls, with sex as a covariate. Analyses were also conducted in each race, and in cases and controls separately if studies included multiple ancestries or had a case-control design.

Detailed summary statistics from each study were subsequently collected, and we pooled β-coefficients and SEs from individual studies using the Mantel-Haenszel fixed-effects method, as well as the DerSimonian and Laird random-effects method implemented in Stata, version 12 (StataCorp LP, College Station, TX). The significant P value was 0.005 after Bonferroni adjustment for 10 independent tests: FTO-BMI association (1 test); diet-BMI associations (3 tests; we considered total energy, protein, carbohydrate, and fat intake as 3 independent variables); FTO-diet associations (3 tests); and FTO-diet interactions (3 tests). Between-study heterogeneity was tested by the Cochran Q statistic and quantified by the values for the proportion of variance explained by interstudy differences (I 2). Low heterogeneity was defined as an I 2 value of 0–25%, moderate heterogeneity as an I 2 of 25–75%, and high heterogeneity as an I 2 of 75–100%. The P value for heterogeneity was derived from a χ2 test. We also performed stratified meta-analyses in subgroups according to ethnicity (whites, African Americans, or Asians), sex, age group (mean age <10 vs. ≥10 years), geographic region (North America, Europe, or Asia), study sample size (n < 500 vs. n ≥ 500), study design (population based vs. case-control), dietary intake assessment method (dietary records or 24-h recalls vs. food frequency questionnaire or others), and adjustment for physical activity (yes vs. no).

processing....