Objectives: To optimize the performance and feasibility of fibrosis blood tests and evaluate their robustness.
Design and methods: The derivation population included 1056 HCV patients with liver biopsy and blood markers. Validation populations included 984 patients with various viral hepatitis causes, and Fibroscan and/or liver biopsy and/or blood markers.
Results: The bootstrap method validated the markers of the original FibroMeter(2G), but not those of Fibrotest and Hepascore, and provided a hyaluronate-free FibroMeter(3G). AUROCs for significant fibrosis were: FibroMeter(2G): 0.853 vs. FibroMeter(3G): 0.851, p=0.489. Compared to FibroMeter(2G), FibroMeter(3G) had a significantly higher patient rate with predictive values ≥90% for significant fibrosis. Accuracy for fibrosis stage classification was: Fibrotest: 37.9%, FibroMeter(2G): 74.9%, and FibroMeter(3G): 86.9% (p<10(-3)).
Conclusion: The bootstrap method validated FibroMeter(2G) and provided a cheaper and more feasible hyaluronate-free FibroMeter(3G) with comparable performance. Compared to binary diagnosis, fibrosis stage classification increased discrimination, with an increased accuracy to 87% for FibroMeter(3G).
Copyright © 2010 The Canadian Society of Clinical Chemist. Published by Elsevier Inc. All rights reserved.