Participants
This cross-sectional study involved 134 children and adolescents with obesity participating in the betaJUDO study[25]. They were included in 2 centers (Pediatric Obesity Clinic at University Children’s Hospital, Uppsala, Sweden, and Paracelsus Medical University, Salzburg, Austria). Inclusion criteria were the followings: (a) 10-17 years old, (b) age-adapted BMI > 30kg.m-2, (c) medical examination including anthropometric assessment and Tanner’s staging, (d) at least 5 valid days out of a possible 7 days of accelerometry measurements, including one of the valid day on a weekend (regardless of PA and SED levels), (e) blood samplings for liver parameters, (f) no contraindication to PA, (g) no additional medical/psychiatric conditions nor medication influencing cardiometabolic, liver or accelerometry data. The study was accepted for Voluntary Harmonisation Procedure (VHP673, VHP2015061) and approved by Ethics Committees and Regulatory Authorities (EudraCT No: 2015-001628-45; EC Sweden: Dnr 2015/279; EC Austria: 415-E/1544/20-2014). Written informed consent were obtained from participants and parents. The trial was conducted according to the Declaration of Helsinki (World Medical Association; Version 2013) and the E6 Guideline for Good Clinical Practice (International Conference on Harmonisation).
Anthropometry and pubertal staging
Standard operating procedures for measurements were harmonized between centers [25]. Weight (kg) was assessed using a standardized calibrated scale (Uppsala: SECA model 704; Salzburg: SECA model 801, Hamburg, Germany). Height (cm) was measured using a stadiometer (Uppsala: Ulmer stadiometer, Busse, Elchingen, Germany; Salzburg: SECA, model 222 stadiometer, Hamburg, Germany). BMI was calculated as weight (kilograms) divided by the square of height (meters). BMI-SDS (Microsoft Excel add-in LMS Growth using WHO growth report Version 2.76) and BMI percentiles (WHO BMI for age) were calculated. Waist circumference (WC, cm) was measured with a flexible tape midway between the superior border of the iliac crest and the lowest rib on a standing patient. Fat mass (FM) percentage was calculated using an InBody S20 bioimpedance device (Biospace, Seoul, Korea) at fasting. Puberty was evaluated with Tanner staging [26, 27].
Biochemical variables
Blood was sampled at fasting. Validation of analyses was performed between laboratories [25]. Serum concentrations of liver enzymes (ALT, aspartate aminotransferase (AST) and gamma-glutamyl transpeptidase (GGT)), total-cholesterol, high density lipoprotein-cholesterol (HDL-c), low density lipoprotein-cholesterol (LDL-c) and triglycerides (TG) were analyzed by enzymatic photometric analysis. Plasma glucose was analyzed by enzymatic chromatic test. Plasma was used for central analyses of insulin using singleplex enzyme-linked immunosorbent assay kits for each analyte (Mercodia AB, Uppsala, Sweden). Insulin resistance was expressed using homeostasis model assessment of insulin-resistance index (HOMA-IR), [HOMA-IR] = glycemia [mmol×L-1] × insulinemia [mUI×L−1]/22.5) [28].
Fatty liver index calculation
Fatty Liver Index (FLI) was calculated as followed:
FLI = e f /(1 + e f) x 100
f = 0.953 x ln (triglycerides) + 0.139 x BMI + 0.718 x ln (ggt) +
0.053 x WC - 15.745)
(TG in mg/dL, GGT in UI/L, WC in cm)
Although initially developed for adults, FLI is currently used in pediatric clinical trials as a biomarker of steatosis and varies between 0 and 100 [29–31].
Percentage hepatic fat
Percentage hepatic fat was measured by magnetic resonance imaging (MRI) using 1.5 Tesla clinical MRI systems from Philips Medical System (Best, The Netherlands; Uppsala: Philips Achieva system, Salzburg: Philips Ingenia system) [25]. Children were categorized as having hepatic steatosis when percentage hepatic fat ≥5% [32].
Physical activity and sedentary time
Movements related behaviors were assessed with the accelerometer Actical® (Philips Respironics, Inc, Murrysville, PA). As previously described [23], it is an omni-directional waterproof device recording accelerations in the range of 0.05–2.0 g, sensitive to movements in the range of 0.35–3.5 Hz, and able to record the magnitude of acceleration and deceleration associated with every movement. The signal was scored as a “count” which was summed over a 1 minute epoch. Participants were asked to wear the device on their non-dominant wrist during 7 consecutive days (24-hours measurements). Non-wear time was defined as ≥60 consecutive minutes of zero counts, with allowance for 2 min of counts between zero and 100. Wear time was determined by subtracting non-wear time from 24 hours. A valid day was defined as ≥10 h of wear time. Each minute of wear time was classified using established cut points into SED (<1.5 metabolic equivalent of the task, METS), light physical activity (LPA, <3 METS), moderate activity (MPA, 3 to 6 METS) and vigorous activity (VPA, >6 METS) [33]. MVPA was the sum of MPA and VPA. SED+ (more sedentary) and SED- (less sedentary) groups were defined by being respectively upper and under the median of the sample for SB time. MVPA+ (more active) and MVPA- (less active) groups were defined by being respectively upper and under the median of the sample for MVPA time. The 4 combinations (SED-/MVPA+, SED-/MVPA-, SED+/MVPA+, SED+/MVPA-) were created using the median of MVPA for each of SED subsamples.
Statistical Analysis
Statistical analyses were performed using Stata software (version 15, StataCorp, College Station, US). Continuous data were expressed as means and standard-deviations (SD). The normality of the distribution was checked with a Shapiro-Wilk test. Comparisons between groups were performed using Chi-squared or Fisher’s exact test for categorical data, and analysis of variance (ANOVA) or non-parametric Kruskal-Wallis test (when the ANOVA assumptions were not met) for continuous variables. Assumption of homoscedasticity was studied using Bartlett’s test. Whan appropriate (omnibus p-value<0.05), post-hoc test for two by two multiple comparison were applied: Tukey-Kramer after ANOVA, Dunn after Kruskal-Wallis test and Marascuilo for categorical data. Relationships between continuous data were explored with Pearson or Spearman correlation coefficient and a Sidak type I error correction. Multivariable analyses were conducted using multiple linear regression in order to compare groups adjusting aforementioned analyses on possible confounders. No specific strategy approach, such as stepwise, was conducted. Covariates were chosen according to univariate results and clinical relevance. Multivariable regression analyses were run with the following covariates: age, gender and Tanner stages (model 2) and age, gender, Tanner stages and BMI (model 3). The normality of residuals was checked and a logarithmic transformation of the dependent variable was performed when appropriate. Differences were considered statistically significant at p<0.05.