SIMPLE CLINICAL VARIABLES PREDICT LIVER HISTOLOGY IN HEPATITIS C: PROSPECTIVE VALIDATION OF A CLINICAL PREDICTION MODEL

CN Andrews,* J Romagnuolo,* VG Bain

Division of Gastroenterology, *University of Calgary, Calgary; University of Alberta, Edmonton, Alberta

BACKGROUND: A recent single-center multivariate analysis of hepatitis C (HCV) patients (Romagnuolo et al, 2001) showed that ferritin ³ 200mcg/L and spider nevi and/or albumin £ 35g/L predicted significant histological inflammation (likelihood ratio (LR) 5.5); the presence any two of either spider nevi, platelets £ 150´10⁹/L, or palpable splenomegaly and/or albumin £ 35g/L predicted significant histological fibrosis (LR 19). Absence of predictors also predicted a lack of inflammation (LR 0.1) and fibrosis (LR 0.3). We proceeded to prospectively validate this clinical prediction model using an independent sample from a multi-center study.
METHODS: Eighty-one patients with previously untreated chronic HCV, raised serum transaminases, and positive serum HCV-RNA underwent physical examination, laboratory investigation, and liver biopsy. Biopsies were read by a single blinded pathologist (modified Hytiroglou (1995) system): inflammatory grade 0 - none, grades 1 to 3 - mild, moderate, marked interface necrosis and lobular activity, respectively; fibrotic stage 1- minimal (confined to portal tracts), stage 2 - fibrous septa or bridging fibrosis, stage 3 - cirrhosis or >3 bridges. The clinical scoring system was correlated with histology; LRs and Fisher Exact p-values were calculated.
RESULTS: Data recording was complete in 77 and 38 patients with fibrotic stage and inflammatory grade (mostly due to missing ferritins), respectively. For fibrosis (p<0.001), 7/8 patients with any 2 criteria (LR 27), 6/22 with any 1 criterion (LR 1.4) and 3/47 with no criteria (LR 0.3) had significant fibrosis (stage 2-3) on biopsy. For inflammation (p=0.036), 5/5 patients with any 2 criteria (LR 7.4), 10/14 with any 1 criterion (LR 1.6), and 8/19 patients with no criteria (LR 0.5) had inflammation on liver biopsy. When missing variables were assumed to be normal in all cases where they were missing, recalculated LRs were almost identical.
CONCLUSION: Our published model, consisting of simple clinical variables, accurately and significantly predicted fibrosis and inflammation in HCV and appears validated in this independent multi-centre sample.