The risk of end stage liver disease and hepatocellular carcinoma among persons infected with hepatitis C virus: publication bias?

In persons infected with hepatitis C virus (HCV), the incidence of cirrhosis and hepatocellular carcinoma (HCC) can be estimated by examining over time entire cohorts with known onset of HCV infection. We performed a systematic review of the literature to identify and to analyze studies that examine such cohorts.A search of all articles from 1980 to 2001 was performed. Published studies were included in which chronic HCV infection was defined by elevation of liver enzymes or persistent RNA. We excluded studies in which cohorts were selected from patients with prevalent liver disease or in whom the onset of infection could not be estimated. Two investigators abstracted the data. The incidences of cirrhosis and HCC were analyzed in all studies and in categories based on study design, mode of HCV acquisition, sample size and duration of follow-up, age at the onset of infection, and the quality of estimating the onset of HCV infection.Of the articles, 21 fulfilled the selection criteria. Studies varied in sample size (17-1,680), duration of follow-up (8-45 yr), total person-years (157-34,098), and the mean age at onset of HCV (5-58 yr). The mean time to end stage liver disease (ESLD) was 4-23 yr and to HCC was 9-31 yr. A funnel plot showed a possible publication bias against studies with low incidence of ESLD and HCC. The pooled weighted incidence rates for ESLD and HCC based on infection mode were as follows: community-acquired HCV, 1.9 and 0 per 1,000 person-years; transfusion associated, 4.5 and 0.7; hemophilia patients, 7.9 and 1.0; anti-D IgG, 0.7 and 0 per 1,000 person-years. Poisson regression modeling showed that the incidence of ESLD is increased in studies with a low quality estimate of the onset of HCV infection, in community-acquired and transfusion-associated HCV, and in studies with prospective design.We found large variation in the incidence estimates of cirrhosis and HCC. Short duration of follow-up, small sample size, and possible publication bias may explain some of this variation. Low quality estimates of the onset of HCV infection, a prospective study design, and a transfusion- or hemophilia-related mode of acquisition were independent predictors of high reported incidence of ESLD.