Appraising diagnostic performance of ELF test by pathological staging and digital quantification of liver fibrosis

Abstract

Introduction and objectives: A crucial issue when appraising the performance of non-invasive markers is the limitations of the reference standard they are compared to. Digital image analysis (DIA) was suggested as a reproducible approach expressing fibrosis numerically as a proportionate area (PA) (%). We aimed to evaluate ELF test with direct reference to PA (%), thereby explore the improvement in accuracy to discriminate significant fibrosis which may actually have been underestimated by categorical pathological staging. Materials and methods: PA (%) data were obtained by DIA of trichrome-stained liver biopsies of 52 chronic hepatitis patients. Paired serum samples of patients and additional 36 controls were performed to measure ELF test. Diagnostic performance characteristics of ELF test was derived in predicting significant fibrosis in the patient cohort, and also, in distinguishing healthy controls from patients with significant fibrosis. Results: We found an AUROC value of 0.73 for ELF to predict significant fibrosis as assessed by DIA and a lower AUROC value of 0.66 when assessed by conventional pathology. Importantly, ELF test provided considerably high diagnostic accuracy to discriminate healthy controls from patients with significant fibrosis defined by Ishak F≥2 and TPA ≥ 5% (AUROCs 0.93 and 0.94, respectively) with optimal ELF cut-off point of 8.4 for both. Conclusions: Digital quantification could represent a better reference standard than conventional pathology allowing a better discriminatory capability for ELF test. ELF test provided high diagnostic accuracy to discriminate healthy controls from patients with significant fibrosis suggesting a role as a screening strategy in the community setting.

Keywords

• Enhanced liver fibrosis test
• Digital image analysis
• Pathological stages