Aguilar Schall – Employment: Gilead Sciences, Inc. Ann D. Johnson – Employment: Gilead Sciences Jeffrey D. Bornstein – Employment: Gilead Sciences Mani Subramanian – Employment: Gilead Sciences John G. McHutchison – Employment: Gilead Sciences; Stock Shareholder: Gilead Sciences Stephen A. Harrison Deforolimus chemical structure – Advisory Committees or Review Panels: Merck, Nimbus Discovery; Grant/Research Support: Merck, Genentech; Speaking and Teaching: Merck, Vertex Arun J. Sanyal – Advisory Committees

or Review Panels: Bristol Myers, Gilead, Abbott, Ikaria; Consulting: Salix, Immuron, Exhalenz, Nimbus, Genentech, Echosens, Takeda; Grant/Research Support: Salix, Genentech, Genfit, Intercept, Ikaria, Takeda, GalMed, Novartis, Gilead; Independent Contractor: UpToDate, Elsevier INTRODUCTION: Recent studies have suggested that non-invasive liver fibrosis diagnostic tests can predict liver complications and survival. We aimed to identify the time window in which a single evaluation of liver fibrosis may KPT-330 accurately predict mortality and to assess whether prognostic performance may be improved by combining tests.

METHODS: 3,337 patients with chronic liver disease of various causes have been enrolled in a prospective cohort between 2005 and 2009. All patients had a non-invasive evaluation of liver fibrosis at baseline, either by a blood test (APRI, FIB-4, Hepascore, FibroMeterV2G) or a liver stiffness measurement (LSM by Fibroscan), or both. They were followed until death (data obtained from national registry) or up to January 2011. Time-dependent ROC curves [AUC(t)] were used to assess the discriminative ability of liver fibrosis tests according to delay of prediction, and Harrell’s C-index was used as summary measure of discrimination

over the whole follow-up period. Multivariate prognostic models were built in a random set of patients, and validated in the other half of patients. RESULTS: 3,064 patients had LSM and 2,891 patients had blood sampling at baseline; 1,559 patients had available data for all liver fibrosis tests. In this subgroup, the follow-up period ranged from 0.002 to 6.0 years (median=2.8 years); 16.8% of patients died, 7.4% of deaths were liver-related. The medchemexpress discriminative ability of tests for mortality was the greatest in the few months after the baseline measure; then it decreased to a performance level that remains relatively stable over 5 years: all tests (except APRI) had AUC(t)>0.70 to predict all-cause death, AUC(t) from 0.80 to 0.90 for liver-related mortality. FibroMeterV2G had the highest C-index (0.790, 95%CI=0.765-0.815), as compared to LSM and all other blood tests (p<0.008). The diagnostic test combining LSM and FibroMeterV2G, called E-FibroMeterV2G, showed higher C-index than LSM alone (p=0.002).