Development and validation of a novel non-invasive test for diagnosing nonalcoholic fatty liver disease in Chinese children.

Background: With the exploding prevalence of obesity, many children are at risk of developing nonalcoholic fatty liver disease. Using anthropometric and laboratory parameters, our study aimed to develop a model to quantitatively evaluate liver fat content (LFC) in children with obesity.

Methods: A well-characterized cohort of 181 children between 5 and 16 years of age were recruited to the study in the Endocrinology Department as the derivation cohort. The external validation cohort comprised 77 children. The assessment of liver fat content was performed using proton magnetic resonance spectroscopy. Anthropometry and laboratory metrics were measured in all subjects. B-ultrasound examination was carried out in the external validation cohort. The Kruskal-Wallis test, Spearman bivariate correlation analyses, univariable linear regressions and multivariable linear regression were used to build the optimal predictive model.

Results: The model was based on indicators including alanine aminotransferase, homeostasis model assessment of insulin resistance, triglycerides, waist circumference and Tanner stage. The adjusted R² of the model was 0.589, which presented high sensitivity and specificity both in internal [sensitivity of 0.824, specificity of 0.900, area under curve (AUC) of 0.900 with a 95% confidence interval: 0.783-1.000] and external validation (sensitivity of 0.918 and specificity of 0.821, AUC of 0.901 with a 95% confidence interval: 0.818-0.984).

Conclusions: Our model based on five clinical indicators was simple, non-invasive, and inexpensive; it had high sensitivity and specificity in predicting LFC in children. Thus, it may be useful for identifying children with obesity who are at risk for developing nonalcoholic fatty liver disease.