Accuracy of Transient Elastography Data Combined With APRI in Detection and Staging of Liver Disease in Pediatric Patients With Cystic Fibrosis

Abstract

Background & aims: Liver disease develops in 15%-72% of patients with cystic fibrosis, and 5%-10% develop cirrhosis or portal hypertension, usually during childhood. Transient elastography (TE) is a noninvasive method to measure liver stiffness. We aimed to validate its accuracy in detection of liver disease and assessment of fibrosis in children with cystic fibrosis.

Methods: We performed a cross-sectional study to evaluate the accuracy of TE in analysis of liver disease in 160 consecutive children who presented with cystic fibrosis (9.0 ± 0.4 years old, 53% male) at a tertiary referral pediatric center in Australia, from 2011 through 2016. Patients were classified as having cystic fibrosis-associated liver disease (CFLD) or cystic fibrosis without liver disease (CFnoLD) based on clinical, biochemical, and imaging features. Fibrosis severity was determined from histologic analysis of dual-pass liver biopsies from children with CFLD, as the reference standard. Data from healthy children without cystic fibrosis (n = 64, controls) were obtained from a separate study. Liver stiffness measurements (LSMs) were made by Fibroscan analysis, using the inter-quartile range/median ≤30% of 10 valid measurements. Children with macronodularity or portal hypertension with heterogeneous changes on ultrasound without available biopsy were assigned to the category of stage F3-F4 fibrosis.

Results: LSM was made reliably in 86% of children; accuracy increased with age. LSMs were significantly higher in children with CFLD (10.7 ± 2.4 kPa, n = 33) than with CFnoLD (4.6 ± 0.1 kPa, n = 105) (P < .0001) or controls (4.1 ± 0.1kPa) (P < .0001); LSMs were higher in children with CFnoLD than controls (P < .05). At a cut-off value of 5.55kPa, LSM identified children with CFLD with an area under the receiver operating characteristic (AUROC) curve of 0.82, 70% sensitivity, and 82% specificity (P < .0001). Classification and regression tree models that combined LSM and aspartate aminotransferase to platelet ratio index (APRI) identified children with CFLD with an AUROC curve of 0.89, 87% sensitivity, and 74% specificity (odds ratio, 18.6). LSMs correlated with fibrosis stage in patients with CFLD (r = 0.67, P = .0001). A cut-off value of 8.7kPa differentiated patients with stage F3-F4 fibrosis from patients with stage F1-F2 fibrosis (AUROC, 0.87; 75% sensitivity; 100% specificity, P=.0002). The combination of LSMs and APRI improved the differentiation of patients with F3-F4 fibrosis vs F1-F2 fibrosis (AUROC, 0.92; 83% sensitivity; and 100% specificity (P < .01).

Conclusions: LSMs made by TE accurately detect liver disease in children with cystic fibrosis; diagnostic accuracy increases when LSMs are combined with APRI. LSMs also differentiate between children with cystic fibrosis with mild-moderate fibrosis vs advanced fibrosis.

Keywords: CART Analysis Decision Tree; Hepatic Fibrosis; Pediatric Cholestatic Liver Disease; Ultrasound-based Diagnosis of Liver Disease.