Original Article
Estimated progression rates in three United Kingdom hepatitis C cohorts differed according to method of recruitment

https://doi.org/10.1016/j.jclinepi.2005.06.008Get rights and content

Abstract

Objectives

To estimate hepatitis C virus (HCV) progression rates between disease stages prior to cirrhosis, using data from liver biopsies in three observational cohorts. To demonstrate how the method of cohort recruitment can influence the estimation of HCV-progression rates.

Study Design and Setting

Data came from three United Kingdom observational cohorts, assembled from different referral sources. In total, 987 HCV-infected patients with an estimated (or known) date of infection and at least one histologically scored liver biopsy were eligible for inclusion in the analysis. Liver biopsy scores were used to determine the stage of HCV-related liver disease. A three-state continuous time Markov model was used to estimate covariate-specific average probabilities of progression of disease.

Results

After adjusting for confounders, considerably different rates of disease progression were estimated in the three cohorts. For a group of patients with the same demographics, the estimated 20-year probability of progression to cirrhosis was 12% (95% confidence interval CI = 6–22) in a hospital-based cohort, 6% (95% CI = 3–13) in a posttransfusion cohort, and 23% (95% CI = 14–37) in a cohort recruited from a tertiary referral center.

Conclusion

Researchers using estimates of disease progression should be aware that the method of cohort recruitment has considerable influence on the progression rates that are derived.

Introduction

Infection with hepatitis C virus (HCV) usually results in chronic hepatitis, which over many years can cause severe fibrosis, cirrhosis, and hepatocellular carcinoma. Due to the typically slow and asymptomatic course of the disease, the natural history of HCV is not well understood. It has been studied only in selected groups of patients.

Informing the natural history of HCV and estimating the progression of HCV-related liver disease is important for a number of reasons: to determine the current and future burden of HCV-related liver disease on health care resources; to provide cost-effectiveness analyses of antiviral therapies; and for predicting individual prognoses. As is often the case, the nature of the research determines the population from which estimates of disease progression are to be taken [1]. For example, a study of the burden of HCV requires estimates of disease progression for the whole infected population—a highly heterogeneous group of patients. In contrast, a cost-effectiveness analysis may require estimates of disease progression from a treatment setting.

Studies investigating progression of HCV-related disease use data from observational cohorts, which can be characterized according to the source of recruitment: liver clinic studies identifying individuals referred to institutions [2], [3], posttransfusion cohorts [4], [5], [6], infected blood donor cohorts [7], or community-based cohorts [8], [9]. Many biases can arise when analyzing observational cohort data and, with a disease such as HCV where the rate of progression is slow, recruitment bias is a major concern when determining progression estimates for an infected population. Here, patients are referred to a cohort dependent on their underlying disease process, and the follow-up time of the cohort is relatively small compared to the latent time of the disease. For example, a cohort taken from a tertiary referral center may overrepresent patients with severe illness and rapid disease progression and hence provide overestimates of transition rates for the infected population as a whole. In fact, these estimates reflect the highly selective group of patients under specialist medical care. The patient population at tertiary referral centers could also change over time, as more patients with slow disease progression become ill.

Typically, studies assess disease severity using a scoring system for fibrosis from a liver biopsy. Progression is estimated at the individual level by dividing the change in fibrosis score at two consecutive observations by the time elapsed between those observations [2], [10], [11], [12], [13]. This approach assumes the patient entered that fibrosis stage precisely at the observation time and that the rate of progression is constant, but both assumptions are likely to be inappropriate. More realistic approaches have been proposed [14], but only progressions from infection to a single stage (e.g., cirrhosis) have been considered, rather than between the different stages of HCV disease.

Recently, representation of HCV progression as a series of disease stages through a multistage model formulation has been suggested as an alternative [15], [16], when transition between specific disease stages is of interest and when, crucially, only information on stage occupation (rather than stage entrance) is available. Multistage modeling is particularly appropriate to the study of fibrosis in HCV where the only information available is the stage of disease identified at biopsy.

This research considers untreated patients in three distinct United Kingdom (UK) cohorts with different recruitment methods: the Trent HCV Study [17], the HCV National Register (lookback) cohort [6] and a cohort from St. Mary's Hospital, London [18]. Data are collected on patients with an estimated (or known) date of infection and at least one fibrosis scored biopsy. Estimated transition rates for fibrosis progression are obtained from a multistage Markov model. Our objective was to investigate whether the method of cohort recruitment influences the estimation of HCV-progression rates and, if so, by what degree.

Section snippets

Materials, methods, and patients

A total of 987 patients were enrolled from three UK cohorts. A brief description of the three cohorts follows.

Patient demographics

Table 1 presents the demographic details of the patients from the three cohorts, including a comparison with the subset of patients selected for analysis (i.e., those who have a scored liver biopsy and an estimated date of infection). Alanine aminotransferase (ALT) levels are available in the Trent and the lookback cohorts. Those selected for analysis from the HCV lookback are on average younger and those from the Trent cohort are less likely to have PNALT than their respective cohorts. The

Discussion

The techniques used in the present study allow differential progression rates to be estimated, depending on the current stage of disease and a number of cofactors. Multistage Markov models have not been widely used in hepatitis C disease progression, despite the categorical nature of biopsy assessment. Deuffic-Burban et al. [15] illustrated the potential of using a Markov model to estimate hepatitis C progression rates, and Yi et al. [16] showed the advantages of such an approach over routinely

Acknowledgments

The authors would like to thank Tony Ades for his helpful comments in improving the manuscript. M.J.S. is funded by a UK Department of Health grant (No. AIDB 2/29). Thanks also go to the two HCV collaborative groups.

The Trent HCV Study Group: Dr. J.G. Freeman (Derby City Hospital); Dr. M. Wiselka, Professor K.R. Nicholson (Leicester Royal Infirmary); Dr. B.B. Scott (Lincoln County Hospital); Professor R.G. Finch, Dr. B.J. Thomson (Nottingham City Hospital); Professor W.L. Irving, Dr. K.R. Neal,

References (27)

  • L.B. Seeff et al.

    Long-term mortality after transfusion-associated non-A, non-B hepatitis. The National Heart, Lung, and Blood Institute Study Group

    N Engl J Med

    (1992)
  • H.E. Harris et al.

    The HCV National Register: towards informing the natural history of hepatitis C infection in the UK

    J Viral Hepat

    (2000)
  • S. Bellentani et al.

    Prevalence of chronic liver-disease in the general-population of northern Italy: the Dionysos study

    Hepatology

    (1994)
  • Cited by (53)

    • Cost-Effectiveness of One-Time Birth Cohort Screening for Hepatitis C as Part of the National Health Service Health Check Program in England

      2019, Value in Health
      Citation Excerpt :

      First, transition probabilities were derived from a health technology assessment in the United Kingdom by Shepherd et al,29 based on clinical cohorts, which have been used in other HCV models,19,20,30 and have informed NICE HCV testing guidelines,16 with additional transition probabilities for progression of cirrhotic individuals achieving SVR.31,32 We also considered transition probabilities from mild, moderate, and cirrhotic health states generated by the back-calculation model by Harris et al, which used prior values from Sweeting et al, and estimated age-based transition probabilities in the model fitting process.2,33 While the probabilities from Shepherd et al29 are comparable to other economic evaluations, those generated from the back-calculation model align with HCV prevalence and severity parameters derived from the same source, and are thus consistent with other model inputs.

    • Disease outcomes in a cohort of women in Ireland infected by hepatitis C-contaminated anti-D immunoglobulin during 1970s

      2017, Journal of Hepatology
      Citation Excerpt :

      In 2004, a national database was established to monitor these groups, and members of this anti-D cohort have now had their disease progression followed for 36 years.6–10 Along with a small number of other HCV cohorts internationally,11,12 the Irish anti-D cohort provides an exceptional opportunity to study the full course of disease among HCV infected individuals from the time of exposure. Features which make this cohort of particular interest include: (i) their long period of monitoring with very few losses to follow-up, and (ii) the low prevalence of other risk factors for HCV, which has provided greater certainty around their likely dates of infection, and (iii) their very low risk of re-infection.

    View all citing articles on Scopus
    1

    Members of the collaborative groups are named individually in the Acknowledgments.

    View full text