This study was approved by the institutional review board of Boston Children’s Hospital, and all research conducted was consistent with declaration of Helsinki. Informed consent was obtained from parents or guardians of all participants.
A total of 150 participants aged 8–18 years of age were initially recruited from the Preventive Cardiology Clinic of Boston Children’s Hospital, Boston, Massachusetts. Medical and family history was collected by review of the clinic chart, electronic medical record and interview of a parent and patient. Family history of premature cardiovascular disease was defined as heart attack, stroke, angioplasty, stenting, coronary artery bypass grafting (CABG), peripheral arterial disease (PAD), sudden death occurring < 55 years in male and <65 years in female relatives (parents, aunts, uncles, grandparents), in a manner consistent with the National Cholesterol Education Program (NCEP) [11] and the American Academy of Pediatrics (AAP) [3,12]. All demographic information was collected by parental self-report, including exposure to tobacco smoke in the home, hours per day of television watching and computer gaming (screen time), racial background (Black, White, Other [includes Asian, Hispanic, Pacific Islander, Native American], and family history of cardiovascular disease (parent, grandparent, aunt, uncle or sibling).
Anthropometric measures
We used the body mass index (BMI) as a measure of overall adiposity, and WC as a measure of abdominal adiposity. Height and weight were measured using a standing scale and stadiometer, and WC was measured at the level of the superior iliac crest using a tape measure. The average of two WC measures was recorded. BMI was calculated as weight in kg divided by height in meters squared. BMI percentile was determined using Centers for Disease Control (CDC) growth charts [13]. Overall obesity was defined as a BMI ≥ 95% percentile of BMI for age and sex [14,15], and abdominal obesity was defined as a WC ≥ 90th percentile for age and sex [16].
Systolic (SBP) and diastolic (DBP) blood pressure were measured initially using an oscillometric cuff (Dinamap). The average of 2–3 recordings of SBP and DBP were obtained using appropriately sized cuffs using standard techniques, and converted into percentiles [17]. Elevated blood pressures (SBP > 140 mmHG, DBP > 90 mmHg) were rechecked by auscultation by experienced clinicians [17]. High blood pressure was defined as a SBP or DBP ≥ 95th percentile [17].
Laboratory measurement of risk markers
A serum sample was obtained from 107 participants after a 12-hour fast, and the following tests were performed: total cholesterol (TC), high density lipoprotein cholesterol (HDL), triglycerides (TG), low-density lipoprotein cholesterol (LDL; calculated), soluble tumor necrosis factor-alpha receptor 2 (TNF-αR2), P-selectin, intracellular adhesion molecule-1 (ICAM-1), and C-reactive protein using a high sensitivity assay (hs-CRP). Soluble tumor necrosis factor-alpha receptor 2 (TNF-αR2) is the more stable receptor of TNF-α, − which is an early inflammatory stimulator of endothelium [18]. P-selectin and ICAM-1 are both endothelial cell adhesion molecules for monocytes and lymphocytes, and C-reactive protein is a pentameric inflammatory marker that binds to bacterial cell wall components [18].
Lipids were measured enzymatically with a Hitachi 911 analyzer using reagents and calibrators from Roche Diagnostics (Indianapolis, IN, USA). The Friedewald equation was used to calculate LDL (TC – (TG/5 + HDL)) if triglycerides were less than 400 mg/dL. A direct method was used to measure LDL if triglycerides were > 400 mg/dL. Very low-density lipoprotein (VLDL) was estimated as the difference between TC and all other major lipoprotein fractions (TC – (LDL + HDL)). National Heart Lung and Blood Institute guidelines (NHLBI) were used to identify individuals with high TC (≥200 mg/dL), high LDL cholesterol (≥130 mg/dL) and low HDL cholesterol (<40 mg/dL) [19].
Lipoprotein(a) was measured with a turbidimetric assay on the Hitachi 911 analyzer (Roche Diagnostics) using reagents and calibrators from Denka Seiken (Niigata, Japan). This is the only commercial assay not affected by Kringle type 2 repeats [20]. The concentration of C-reactive protein was determined using a high-sensitivity (hs-CRP) immunoturbidimetric assay on the Hitachi 911 analyzer (Roche Diagnostics), using reagents and calibrators from Denka Seiken (Niigata, Japan). Soluble tumor necrosis factor-alpha receptor 2 (TNF-αR2), P-selectin, and intracellular adhesion molecule-1 (ICAM-1) were measured using enzyme-linked immunosorbent assays (ELISAs) from R & D Systems (Minneapolis, MN, USA).
All samples were assayed in duplicate. Samples with disparate results (coefficient of variance [CV] for duplicates > 10%) were re-assayed a third time. Day to day and within run coefficients of variance (CV) for all assays were <10%. The CVs for lipid tests were less than 2%.
Statistical analysis
Continuous sex and race-specific (Black, White, Other) WC percentiles were interpolated from National Health and Nutrition Examination Survey (NHANES; 1999–2008) data [21] using 5-knot cubic spline regression (R; rms package). To reduce bias associated with extrapolating beyond the data, participant WC percentile above the 90th percentile or below the 10th percentile were rounded to 95% and 5%, respectively. The mean and standard deviation (SD) were calculated for all continuous variables. Risk markers were log-transformed prior to statistical tests and regression modeling to normalize skewed distributions. Unpaired Student’s T tests and chi squared tests were used to compare the means and proportions of each variable according to overall and abdominal obesity.
Principal component analysis (PCA) with varimax rotation was used to identify the first four independent patterns of risk markers (Proc Factor) that would explain the majority of variance in risk marker data. A continuous score was derived for each participant representing how closely their risk markers levels conformed to each pattern (Proc Score). Scores were then divided into quintiles for ease of interpretation.
Linear regression was used to assess the relationships between BMI percentile, WC percentile and quintiles of risk marker scores (Proc Reg). Partial correlation coefficients were calculated to determine the independent explanatory power of BMI and WC percentile on each score. All regression models were adjusted for exposure to cigarette smoke (Yes vs No), family history of cardiovascular disease (Yes vs No), screen time (hours), and race (Black, White vs Other). Analyses were performed in SAS ver 9.3 (Cary, NC, USA) and R version 2.15.1.