Development of Fibro-PeN, a clinical prediction model for moderate-to-severe fibrosis in children with nonalcoholic fatty liver disease

Background and Aims: Liver fibrosis is common in children with NAFLD and is an important determinant of outcomes. High-performing noninvasive models to assess fibrosis in children are needed. The objectives of this study were to evaluate the performance of existing pediatric and adult fibrosis prediction models and to develop a clinical prediction rule for identifying moderate-to-severe fibrosis in children with NAFLD. Approach and Results: We enrolled children with biopsy-proven NAFLD in the Nonalcoholic Steatohepatitis Clinical Research Network within 90 days of liver biopsy. We staged liver fibrosis in consensus using the Nonalcoholic Steatohepatitis Clinical Research Network scoring system. We evaluated existing pediatric and adult models for fibrosis and developed a new pediatric model using the least absolute shrinkage and selection operator with linear and spline terms for discriminating moderate-to-severe fibrosis from none or mild fibrosis. The model was internally validated with 10-fold cross-validation. We evaluated 1055 children with NAFLD, of whom 26% had moderate-to-severe fibrosis. Existing models performed poorly in classifying fibrosis in children, with area under the receiver operator curves (AUC) ranging from 0.57 to 0.64. In contrast, our new model, fibrosis in pediatric NAFLD was derived from fourteen common clinical variables and had an AUC of 0.79 (95% CI: 0.77-0.81) with 72% sensitivity and 76% specificity for identifying moderate-to-severe fibrosis. Conclusion: Existing fibrosis prediction models have limited clinical utility in children with NAFLD. Fibrosis in pediatric NAFLD offers improved performance characteristics for risk stratification by identifying moderate-to-severe fibrosis in children with NAFLD.