ESPE Abstracts

Background: Liver steatosis (LS) results from ectopic fat deposition and can be present in obesity even at early ages. Similar to insulin resistance, the onset of LS does not seem to depend exclusively upon the severity of obesity and shows a large interindividual variability.

Objective: Our objective was to develop a predictive normogram for LS in children and adolescents with obesity based on their demographic, anthropometric, body composition and metabolic features.

Patients and methods: Presence of LS (ultrasonography) was studied in 50 children and 100 adolescents with obesity (BMI-SDS: +4.10 ± 1.37; 50% Caucasians/50% Latinos; 50% males/50% females in each group). The variables that were significant in a preliminary univariate analysis including demographic (sex, ethnicity), body fat content and distribution (whole-body DXA-scan, abdominal MRI) and metabolic parameters (uric acid, lipid profile, glucose and insulin, fasting and throughout the OGTT) served as independent variables for multivariate analysis (binary logistic regression using forward conditional method; dependent variable: LS). The discriminatory power of the model was evaluated using receiver-operator characteristics (ROC) curves.

Results: Prevalence of LS was 21.3% (n = 32), higher in Latinos (χ2: 8.69; P < 0.01) both in children (33.3% vs. 16.0%) and adolescents (31.4% vs. 8.0%). Univariate analysis showed that the excess of body fat (BMI-SDS, body-fat%-SDS or fat/lean mass ratio by DXA) did not reach statistical significance, whereas abdominal fat distribution did [trunk/lower-limb and abdominal visceral/subcutaneous fat ratios (both P < 0.05)]. Significant influences of HOMA, WBISI and insulin at 120´ in the OGTT (glucose metabolism) and triglycerides and HDL (fat metabolism) were also found. In the whole cohort (n = 150), the Hosmer-Lemeshow model showed a discriminatory power of 0.7931 for LS prediction considering ethnicity, sex and HOMA as independent variables. However, when children and adolescents were analyzed separately, this model was not useful in prepubertal children whilst its discriminatory power increased in adolescents (0.8349), establishing a numerical risk score for LS in adolescents with obesity based upon their sex, ethnicity and HOMA value potentially applicable in the clinical setting. A similar model was also possible considering the abdominal visceral/subcutaneous fat ratio.

Conclusion: Ethnicity, sex, insulin resistance (as estimated by HOMA) and visceral fat accumulation, but not the severity of obesity, are the major determinants of the risk to develop LS in childhood obesity. Such risk can be estimated through a numerical predictive normogram in adolescents, but not in prepubertal children.