A convenience sample of overweight children, aged 6-12 years, was recruited through newspaper advertisements and letters mailed to pediatricians and family physicians in the metropolitan Washington, D.C. area for a weight-loss study (http://www.clinicaltrials.gov
NCT00005669) examining the effects of metformin hydrochloride (23
). Children were eligible if they had a BMI ≥ 95th
percentile for age and sex, were pre-pubertal or early pubertal (breast Tanner I-III for girls; testes < 8 mL for boys), and had hyperinsulinemia (fasting insulin ≥ 15 μU/mL). Children were excluded from this analysis if they were diabetic (fasting glucose ≥ 126 mg/dL or HgbA1c ≥ 6.5%), had any other significant medical disease, had anorexiant use or >2% body weight loss in the preceding 6 months, or were determined to have binge-eating disorder as assessed by the Questionnaire of Eating and Weight Patterns – Adolescent Version (24
The study was approved by the National Institute of Child Health and Human Development Institutional Review Board. The children provided written assent, and their parents/legal guardians provided written consent. Data were collected between October 2000 and April 2007.
Before beginning weight-reduction therapy, subjects were studied during an inpatient admission. After an overnight fast, subjects refrained from eating until 11:30 AM, when they were offered a standardized lunch buffet. On the following day, fasting serum and plasma samples were obtained and a 2-hour hyperglycemic clamp study was performed. Energy intake at all other meals during the admission was not restricted.
Height was measured in triplicate to the nearest 1 mm with a stadiometer. Weight was measured to the nearest 0.1 kg using a digital scale. Body composition was measured as previously described (25
), using air-displacement plethysmography (Life Measurement Inc., Concord, CA) to determine body fat mass and fat-free mass. Roentgenogram of the left hand and wrist was obtained to assess skeletal maturation, and interpreted by a single radiologist using standards for age and sex (26
). Waist circumference was measured in triplicate to the nearest 1mm at the level of the umbilicus. Metabolic syndrome categorization was based on criteria suggested by Cruz et al.
The method for energy intake testing has been described in detail previously (28
). In brief, at 1130 AM, after an overnight fast, subjects ate ad libitum
from a 28-item lunch buffet meal (; available at www.jpeds.com
) consisting of a variety of foods which offered 9,835 kcal for ingestion (28
). The amount consumed was calculated by using the difference in weight of each food item before and after the meal. Nutrient composition was determined using ProNUTRA 3.1 (Viocare Technologies, Inc., Princeton, NJ), the US Department of Agriculture Nutrient Database for Standard Reference, and manufacturer-supplied nutrient information.
Items presented at the buffet meal1
To determine the number of foods offered in the buffet meal that were considered acceptable by the subject, a food-preference questionnaire was administered to all participants, as previously described (28
). Socioeconomic status (SES) was determined using the Hollingshead two-factor index of social status, based on parental education and employment history (29
After an overnight fast, baseline serum and plasma samples were obtained at 0800 h and a 2-hour hyperglycemic clamp study was performed. A bolus infusion of 50% dextrose (0.19g/kg, max 30g) was given over 2 min, followed by maintenance of plasma glucose levels at 180 – 220 mg/dL by continuous infusion of 20% dextrose for 120 min. The rate of dextrose infusion was adjusted based on plasma glucose levels measured using a bedside glucose analyzer (Yellow Springs Instrument, Yellow Springs, OH) every 2.5 min for 15 min, and then every 5 min thereafter. Additional venous blood samples were also obtained every 5 min for 15 min and then every 15 min thereafter for measurement of insulin, C-peptide, and (laboratory confirmed) glucose. First-phase insulin was derived from the mean of measurements at time 0 – 12.5 min. Steady-state insulin was derived from the mean of measurements at time 60 – 120 min. Values were considered valid for use in steady-state analysis if the standard deviation in bedside glucose values was <15 mg/dL and mean glucose at time 60-120 min >190 mg/dL. Using these criteria, 19 subjects were excluded in analyses of steady-state indices, but they were included for all other analyses. Whole-body glucose uptake was estimated as the metabolic rate (M), defined as the exogenous glucose infusion rate (GIR), adjusted for urinary glucose losses and glucose space correction (30
): M = GIR – Urinary Glucose Loss – [0.19 L/kg × 10 × (Glucose120min
(mg/dL))] ÷ 60 min.
Plasma glucose concentrations were confirmed in the clinical laboratory using a glucose oxidase assay (Synchron LX®, Beckman Coulter Inc., Fullerton, CA) with a sensitivity of 3 mg/dL, and an intra- and inter-assay coefficient of variation (CV) of 2% and 3%, respectively. Serum insulin was measured using a sandwich chemiluminescence immunoassay (Immulite 2000®, Diagnostic Products Corporation, Los Angeles, CA), with a sensitivity of 2 μU/mL, an intra- and inter-assay CV of 6.2% and 11.5%, respectively, and <1% and 8% cross-reactivity with C-peptide and proinsulin respectively. Serum C-peptide was measured using a competitive sandwich chemiluminescence immunoassay (Immulite 2000®), with a sensitivity of 0.5 ng/mL, an intra- and inter-assay CV of 3.4% and 8.3%, respectively, and <1% and 17% cross-reactivity with insulin and proinsulin respectively. Serum leptin was measured using a double antibody RIA (Esoterix, Inc., Calabasas Hills, CA), with a sensitivity of 0.1 ng/mL, an intra- and inter-assay CV of 9.6% and 12%, respectively. Lipid panel was measured using Beckman Coulter Reagent (Fullerton, CA).
Insulin sensitivity was estimated by homeostasis model assessment for insulin resistance index (HOMA-IR) (31
) and hyperglycemic clamp insulin sensitivity index (SIclamp
): HOMA-IR = [Fasting Insulin (μU/mL) × Fasting Glucose (mmol/L)] ÷ 22.5; SIclamp
= 100 × (M ÷ Steady State Insulin). Pancreatic β-cell secretory function was estimated by several proxy measures: fasting and clamp C-peptide concentrations, fasting and clamp insulin concentrations, and insulinogenic index (31
): Fasting I/G = Fasting Insulin (μU/mL) ÷ Fasting Glucose (mg/dL). Hepatic insulin clearance was estimated based on the observation that, C-peptide is secreted from the β-cell in equimolar amounts with insulin C-peptide but is primarily cleared by the kidneys with negligible hepatic clearance, thus permitting C-peptide-to-insulin molar ratio (C/I) to serve as a surrogate measure of hepatic insulin clearance (32
Data were analyzed using SPSS Version 12.0.1 for Windows software. Fat-free mass, fat mass, and all the indices of insulin dynamics were normalized by log-transformation. Pearson correlation coefficient was calculated to assess the simple association between indices of insulin dynamics and energy intake without adjustment for covariates. A univariate general linear model with type III sum of squares analysis was used to evaluate these associations with adjustment for covariates. The covariates in the final statistical model included race, sex, skeletal age, interaction between sex and skeletal age (as a proxy for pubertal status), fat-free mass, fat mass, socioeconomic status, and number of foods in the buffet meal that were rated as acceptable. Analysis after exclusion of the 10 subjects who rated fewer than 50% of the items in the buffet meal as acceptable did not alter the significance or direction of associations (data not shown) and so they were included in the final analyses. Exclusion of the 1 subject with the largest energy intake (4092 kcal) during the buffet meal also did not affect results and so this subject is included in the final analyses. All data are reported as mean ± SD unless otherwise indicated.