HEPATOLOGY, Vol. 36, No. 5, Suppl. 1, 2002
2 indicate portal fibrosis, 3 and 4 bridging fibrosis, 5 incomplete or early cirrhosis, and 6 established cirrhosis. The Ishak scale provides better discrimination in assessing small changes in fibrosis, permitting a better assessment of progression of disease, and possible effects of therapy. The intra- and interobserver variability of the Ishak scoring system has yet to be carefully defined.
The Metavir system scores both necroinflammatory changes on a 4-point scale of 0 to 3 and fibrosis on a 5-point scale from 0 to 4, no numbers being skipped (Fig. 1).14 The Metavir system has been carefully validated and shows good intra- and interobserver reproducibility. This system is commonly used in Europe.
The scoring systems for hepatic fibrosis have been ex- tremely helpful in natural history studies and clinical trials of therapy of hepatitis C. However, all of these systems have important limitations. Hepatic fibrosis may not be homogenous throughout the liver, and the liver specimen obtained by the needle biopsy may not accurately reflect the overall average degree of fibrosis. The reliability of the assessment of fibrosis stage increases with the size of the liver sample. The sample size is critical, a minimum length of 10 mm being essential. Regardless of biopsy length, however, fibrosis may be underestimated and cir- rhosis missed in some patients. In addition, scoring sys- tems are artificial and based on visual assessment. Fibrosis may not progress linearly in the same manner as the scor- ing systems: thus, progression from stage 1 to stage 2 may be far more important and require a longer period than progression from stage 3 to stage 4 (or vice versa). Thus, nonparametric analysis is needed in assessing differences in fibrosis scores in clinical studies.
The inter- and intraobserver reproducibility in the scoring systems is variable and therefore, in most clinical studies, all biopsies are read by the same pathologist or team of pathologists under code, with paired liver biopsies being read in a blinded manner without knowledge of the chronology of the biopsies or treatment received. Paired biopsies can also be ranked under code as to worsening, no change, or improvement. In large studies, the problem of sampling error is overcome by the numbers of biopsies that are assessed: thus the number that appears to improve due to sampling error should be counterbalanced by the number that appears to worsen.
Progression of Fibrosis
Limitations of Studies on Progression of Fibrosis. Although there have been many studies of fibrosis in chronic hepatitis C, most have had shortcomings that limit their applicability and reliability. Most studies have been cross-sectional and based on a single liver biopsy. A
MARCELLIN AND BOYER
smaller number of studies have been longitudinal and based on patients who have had 2 liver biopsies. There have been few prospective studies of histological changes in chronic hepatitis C, and the few that have been per- formed were limited in size. In addition, almost all studies have been performed on the highly selected patients that are seen at referral centers.
Cross-Sectional Studies. The major advantage of cross-sectional studies is the ability to include large num- bers of patients, including virtually all patients who had a liver biopsy on whom information is available regarding the time of onset of infection. The major shortcomings of cross-sectional studies are that they rely on mathematical modelling to define the rate of progression of fibrosis and are based on an assumption that the progression of fibrosis is linear over time. In addition, the duration of infection is estimated based on a clinical history of time of exposure or onset of injection drug use. The time of onset derived from clinical history may not be reliable; the date of trans- fusion for instance may not be accurately reported, and studies of injection drug users usually assume that onset of infection occurs within the first year of drug use, which may not be accurate. In addition, sporadic cases in which the time of onset of infection is unknown are excluded in these analyses, which limits the applicability of the results.
Longitudinal Studies. The major advantage of longi- tudinal studies of patients with 2 liver biopsies is that the rate of progression of fibrosis can be clearly demonstrated between the 2 time points of the biopsies. The disadvan- tage of this approach, however, is that patients who have had 2 liver biopsies are usually a selected group of patients who are being observed carefully because of concerns about the progression of disease, but for one reason or other have not been treated. Thus, this group may not be representative of the overall rate of progression of hepati- tis C. Furthermore, retrospective studies require 2 liver biopsies, which may exclude patients who have developed end-stage liver disease, in whom a second liver biopsy is not needed or is contraindicated because of coagulopathy. Finally, sampling error in assessment of fibrosis may cause problems, with patients appearing to have worsening of fibrosis. The effects of sampling error can be partially corrected by assuming that worsening of fibrosis will be counterbalanced by improvement in fibrosis if entirely due to sampling error. Thus, an overall proportion of cases with improvements in fibrosis can be subtracted from the proportion with worsening to give an overall proportion with worsening. This correction, however, is rarely used. Furthermore, in assessing specific factors as- sociated with worsening fibrosis, one cannot tell which individual has actual worsening and which only apparent worsening due to sampling error, so that factors associated