Goals: To determine the validity of fibrosis indexes based on simple laboratory tests in daily practice.
Background: Fibrosis indexes were developed in referral centers using high-quality data.
Methods: We compared the performance characteristics of several such indexes with liver biopsies in a cohort of 490 diverse veterans with chronic hepatitis C from 24 centers. All laboratory tests including interpretation of the liver biopsy were done locally. The following indexes were calculated and correlated with a 5-point fibrosis stage (F0-F4) on liver biopsies: platelet counts (<100 or <150x10(9)/L), aspartate aminotransferase (AST)/alanine aminotransferase (ALT) ratio (AAR), Pohl score, AST-to-platelet ratio index (APRI) and "Lok's model."
Results: Our cohort was predominantly male with 24% blacks, and fibrosis stages of 0, 1, 2, 3, and 4 in 11%, 24%, 28%, 24%, and 13%, respectively. All indexes performed better in predicting advanced (F3-4) than significant (F2-4) fibrosis. When patients with F3-4 were compared to those with F0-2, the area under the receiver operating characteristics curve were 0.534 and 0.641 for platelet count <100 and <150x10(9)/L, respectively, 0.524 for AAR, 0.534 for Pohl score, 0.693 for Lok's model, and 0.765 for APRI. The sensitivity, specificity, and predictive values of APRI and Lok's model were only slightly lower than those reported by the authors using the recommended cutoffs in clinical trial settings. Alcohol use within 12 months, normalization of AST, ALT, and race (blacks/non-blacks) had minimal impact on the performance.
Conclusions: AAR, Pohl, and platelet counts <100x10(9)/L have limited ability to predict significant/advanced fibrosis with area under the receiver operating characteristics curve similar to 0.5. However, platelet counts <150x10(9)/L, Lok's model and APRI performed well for advanced fibrosis in our daily practice setting.