METHODS

Patients

All patients with HCV infection seen at St Mary’s Hospital who had undergone at least one liver biopsy between 1 January 1990 and 30 June 2001 were identified. A subset of patients who had undergone multiple biopsies was identified and listed. All patients were known to be HCV antibody and RNA positive. Exclusion criteria were hepatocellular carcinoma, other types of liver disease in addition to their hepatitis C, human immunodeficiency virus infection, treatment prior to first biopsy, and non-interpretable biopsy.

Data were collected with regard to patient demographics (sex, date of birth, age at infection, current alcohol intake, risk factors, ethnic origin), histological data (dates and scores), virological features (HCV genotype), and antiviral therapy (dates and types). Where possible, date of infection was documented. In those infected by blood products the exact date was recorded where known. When the exact date was not known, but the year was, the estimated date of infection was recorded as the middle of that year. In patients infected by intravenous drug use, the date of infection was estimated as the middle of the year of first drug use, as elsewhere.⁴ Locally derived data were combined with those of a pan European collection from the HENCORE¹⁵ study to produce a master database.

Histopathology

Sections were stained with haematoxylin and eosin and a reticulin stain. For the subset of patients with two or more biopsies, each biopsy was scored using both METAVIR and the modified HAI (Ishak) systems by two histopathologists working together at the time of biopsy review and reaching agreement. The METAVIR system uses a five point scale for fibrosis stage (0–4) and a four point scale for activity (0–3). The modified HAI system (Ishak) uses a seven point scale for fibrosis stage (0–6) and an 18 point scale for activity. The scores by each system were then analysed to check for correlation.

Intra and interobserver reproducibility for fibrosis estimation was examined using weighted kappa analysis after repeated scoring of 30 of the 274 slides scored.

Plan of analysis

All statistics were performed using SPSSv10 (SPSS Inc., Chicago Illinois, USA). Biopsy data and demographic details were summarised for the cohort as a whole (table 1) and tested for normality. The distribution of the rates of fibrosis was skewed and therefore was logarithmically transformed in order to normalise the data. Fibrosis rate is defined as: the ratio of the change in fibrosis stage to the duration of infection between assessments. Fibrosis stage is assumed to be zero at the point of infection.

In the dataset as a whole (see tables 2 and 3) factors related to the severity and rate of fibrosis were analysed. Effect of viral genotype, mode of transmission, alcohol intake, racial group, and sex on fibrosis score were analysed using the χ² test. For patients in whom date of infection was known, their effect on the fibrosis rate was analysed using the Mann-Whitney U test. A logarithmic transformation of the rate was performed to achieve a normal distribution. The effect of age at infection and at biopsy on the stage and rate of fibrosis were analysed using analysis of variance after logarithmic transformation of the rate. Patient age was stratified by decade of life at infection and biopsy.

Those factors significantly associated with the rate of fibrosis on univariate analysis were incorporated into a stepwise multiple regression model.

For patients who had undergone more than one biopsy, rates of fibrosis were calculated for the period between infection and the first biopsy (R1) and between the first and second biopsies (R2). Patients were stratified according to whether they had been treated between biopsies or not.

RESULTS

Effect of the scoring system used

A total of 274 biopsies were assessed using both the METAVIR and Ishak semiquantitative scores. The correlation between them was high (r=0.96 p<0.00001) (fig 1). This suggests that apparent differences in the natural history of HCV observed between groups using these scoring systems is unlikely to be due to the scoring systems themselves.

Reproducibility of semiquantitative scoring system

There was good intra and interobserver agreement (weighted kappa 0.87 and 0.82, respectively) on the 274 biopsies.

Results for the dataset as a whole

A total of 1606 patients had data available for analysis. Demographic features are shown in table 1. Age at infection, age at biopsy, duration of infection at biopsy, and fibrosis score were all found to have approximately normal distributions. The rate of fibrosis had a non- normal distribution with a marked right skew. Logarithmic transformation was therefore carried out as required prior to analyses. Based on the calculated rate of fibrosis, the mean time to cirrhosis was 15 years and the median was 35 years (in close agreement with other work⁴). Male sex (p<0.00001) and older age at infection (p<0.00001) were associated with both increased fibrosis severity and a faster rate of fibrosis. Heavy alcohol intake (p=0.02) and older age at biopsy (p<0.00001) were associated with increased fibrosis severity. Viral genotype non-1 was associated with a faster rate but not overall severity (table 2). Increasing fibrosis score with age at biopsy and duration of infection both demonstrated the progressive nature of fibrosis in HCV infection. There was a linear relationship between age at infection and rate of fibrosis (fig 2) but no such relationship between age at biopsy and rate of fibrosis (fig 3), implying that an individual’s rate of fibrosis does not vary with their age.

When duration of infection was plotted against fibrosis score, steady progression was again demonstrated (fig 4).

Fibrosis score was found to increase in step with mean age at biopsy and demonstrated a linear trend (p<0.00001, ANOVA) (fig 5). Mean rate of fibrosis calculated by dividing the change in fibrosis score by the difference in mean ages with those scores (that is (score 2–score 1)/(mean age at score 2–mean age at score 1)) gave a rate of 0.43 fibrosis units/year (14 years to cirrhosis) which was similar to that for the cohort as a whole.

Construction of a regression model to predict the rate of fibrosis

Factors on univariate analysis related to the rate of fibrosis (age at infection, sex, and viral genotype (table 2)) were entered by the forward stepwise method into a multiple regression model to attempt prediction of log_e fibrosis rate for individuals (table 4). This model accounted for only 29.5% of the variability in fibrosis rate (r²=0.295).

Patients with two or more biopsies

A total of 137 patients had undergone two or more biopsies: 93 of these patients were untreated and 44 had received therapy (interferon monotherapy or interferon ribavirin combination therapy) between the first and second biopsy. There was a considerable difference between the time to the first biopsy and duration between biopsies (15.6 v 2.73 years for the untreated group and 16.9 v 2.69 years for the treated group).

Rates of fibrosis between infection and the first biopsy and between the first and second biopsy were calculated. Patients treated between the first and second biopsy were analysed separately from those who remained untreated. The treatment group had significantly worse liver disease than the untreated group, with higher mean fibrosis stage on their first biopsy (p=0.005 by the Ishak system) (see table 5). Because of this difference, observations of the effect of therapy on the rate of fibrosis compared with no therapy should be interpreted with caution. Likewise, natural history data from the untreated group must be treated with caution as these patients, on average, had mild disease and may not reflect all patients with HCV.

There were no demographic differences between the untreated and treated group (see table 5).

If the rate of fibrosis is constant with time, then the rate between infection and the first biopsy (R1) will be the same as between the first and second biopsy (R2). If R2 is greater than R1, this suggests that the rate of fibrosis increases with time. If R1 is greater than R2, this suggests that the rate of fibrosis decreases with time.

For scoring systems using discrete variables there is a probability of an individual moving from one category to another in a given time. HCV is a slowly progressive disease. If we look at the average fibrosis rates—for example, in untreated patients scored using the modified HAI, 0.15 units/year—we would not expect the fibrosis score to increase by 1 unit for about seven years, and so for a mean time between biopsies of only 2.7 years the mean probability of a change in score was low. This was even more pronounced for METAVIR where there are fewer stages. As a result of the apparent non-progression between biopsies, the average rate of fibrosis falls artefactually, as observed in our population (table 6).

To avoid this problem, the expected fibrosis score of the second biopsy (rounded up or down to the nearest whole number) was calculated from R1 and compared with the observed score. In this way if the time between biopsies was short, we expected to see no change in score. If the observed score was higher or lower than the calculated expected value, we inferred acceleration or deceleration of the fibrosis rate. Observed and expected scores were then compared using a paired sample t test. The null hypothesis was: that there will be no difference between observed and expected (based on a constant rate of infection) fibrosis scores for the second biopsy. Comparison of observed scores on the second biopsy with those expected on the basis of a constant fibrosis rate are shown in table 7. The null hypothesis that there will be no difference between observed and expected scores is accepted for the untreated group (NS). For the treated group there was a significant departure (decrease) from a linear rate demonstrating an effect of therapy. For the METAVIR score, significance was not reached, independent of treatment outcome.

DISCUSSION

As with other large studies, the fibrosis process is shown to be progressive and faster for males and those who acquire the infection later in life. Alcohol intake does not appear to influence the rate of fibrosis in this patient group, although it is related to stage. Data collection in this area is notoriously inaccurate and approximately half of the cohort reported abstinence. Selective recall by those with more severe disease could give an association with severity without demonstrating an effect on rate. We were only able to record current alcohol intake and patients aware of their diagnosis may have modified their drinking behaviour in accordance with medical advice. Another possible confounder is that all patients in the cohort had data for stage whereas only 56.2% had data for rate of fibrosis. The association of viral genotype non-1 with rapid fibrosis has not been noted previously in studies of this type. The lack of association between viral genotype and extent of fibrosis may reflect changing patterns in the epidemiology of HCV infection, with stage lagging behind rate. Patients presenting with genotype 1 may have acquired it through intravenous drug use and therefore were younger at infection, which protects them against fast fibrosis.

A key question we have attempted to answer is whether or not the fibrosis rate is linear—that is, constant throughout life. In favour of a linear fibrosis rate are the steady (and linear) increase in fibrosis score with duration of infection at biopsy, the steady increase in mean age for a given fibrosis stage, and the lack of a relationship between age at biopsy and rate of fibrosis (meaning that a patient’s rate of fibrosis at a given time does not depend on how old they are). Limited results from the dual biopsies showed no difference between the observed fibrosis score of the second biopsy and that estimated from the rate from time of infection to first biopsy.

The main points demonstrated from the serial biopsy data were the methodological problems. Patients who remain untreated between biopsies have less severe liver disease than those who are treated, meaning that we can only look at the natural history in patients with (relatively) mild disease. The short time between the first and second biopsy, and the discrete and bounded nature of the variables (fibrosis score) means that it is unlikely that progression will be observed in many cases where it is (slowly) occurring. It therefore appears that multiple biopsies using semiquantitative scores are too blunt an instrument unless taken many years apart. Although it would be desirable to take a large number of biopsies at wider intervals to study the natural history, this is not in keeping with ethical or clinical reality.

This has clear clinical implications: whereas European guidelines suggest repeat biopsy every 3–5 years, these data suggest a much longer interval is needed if progression is to be observed. One cannot be reassured by the fact that an individual’s fibrosis score has not progressed between liver biopsies only a few years apart.

An acknowledged problem of all studies of this type is the reliability of the date of infection and the assumption that fibrosis is zero at the date of infection.

Attempts to obtain truly quantitative measures of fibrosis from livers biopsies using collagen specific stains and digital image analysis have been unsuccessful due to lack of reproducibility.¹⁶ Serological markers of fibrosis also have limited utility as the coefficients of variation are wide.¹⁷ We conclude that, for the time being, we must rely on the morphological assessment of fibrosis on which to base estimates of rate of progression. However, biopsies taken only a short period of time apart are unlikely to provide useful information to inform clinical decisions or to progress the study of the natural history of HCV infection.

The regression model based on the data set here accounts for only 29.5% of the variability in fibrosis rate. Although this represents a significant advance on the 17% quoted in previous studies,⁴ it is clear that other factors, such as host genetic variability, must play an important role in determining the rate of fibrosis.