Quantification of fibrosis progression in patients with chronic hepatitis C using a Markov model S. Deuffic-Burban, 1,2 T. Poynard 3 and A.-J. Valleron 2,4 1 Unite ´ de Recherche ‘Schistosomiase, Paludisme et Inflam- mation’, INSERM U547, Lille; 2 Unite ´ de Recherche ‘Epide ´miologie et Sciences de l’Information’, INSERM U444; 3 Service d’He ´pato Gastroente ´rologie, Groupe Hospitalier Pitie ´-Salpe ˆtrie `re; and 4 Unite ´ Sante ´ Publique, Ho ˆpital Saint-Antoine, Paris, France Received June 2001; accepted for publication September 2001 INTRODUCTION In chronic hepatitis C, five stages of fibrosis have been defined (F0 to F4) to characterize patients’ histologi- cal status. This staging system is useful for the evaluation of the effects of treatments, thanks to repeated biopsies [1,2]. However, while a biopsy reliably assesses a patient’s stage at a given point in time [3], it does not provide information regarding the length of time the patient has spent in that stage before the biopsy. Repeated biopsies can therefore demonstrate that a progression or improve- ment has occurred, but cannot pinpoint its date. More- over, patients generally undergo very few biopsies, most often as few as 2. It is therefore difficult to derive precise information about disease progression from these data and to quantitatively assess the effect of treatment on this progression. In other conditions with similar problems, Markov models have proven helpful. In patients with HIV infection, for example, a Markov model defined six stages of CD 4 cell- counts and one stage corresponding to the development of clinical symptoms of AIDS [4]. It then quantified disease progression with a mean of only 4.2 serum samples per patient. In another example, a crucial point in the preven- tion of mother-to-child HIV-1 transmission required the estimation of the timing of the infection (in utero or at delivery) from only a few blood samples (at most, four) obtained from infected infants after delivery [5,6]. This was impossible with direct observations, but was successfully carried out with Markov models that used three stages defined by viral markers of HIV-1 replication and by HIV-1 antibody production. This paper exemplifies the potential of Markov modelling to describe the natural history of fibrosis progression in hepatitis C using data from patients with as few as two biopsies. The correlation of fibrosis progression with candi- date risk factors as well as the effect of interferon treatment is also illustrated. Abbreviations: IFN-a, interferon-a. Correspondence: Sylvie Deuffic-Burban, Unite ´ de Recherche ‘Epidemiologie et Sciences de L’Information’, INSERM U444, CHU Saint-Antoine, 27 rue Chaligny, 75012 Paris, France. E-mail: sylvie.burban@libertysurf.fr Journal of Viral Hepatitis, 2002, 9, 114–122 Ó 2002 Blackwell Science Ltd SUMMARY. The knowledge of fibrosis progression in chronic hepatitis C and the impact of new treatments on progression is limited by the number of available liver biopsies per patient. Moreover, liver biopsies identify a patient’s stage of fibrosis at a given point in time, but cannot quantify the time spent in that stage nor the date of transition to that stage. This paper assesses the potential of Markov modelling to overcome these difficulties. The data from interferon-treated (n ¼ 185) and untreated patients (n ¼ 102) are analysed to illustrate the power of this technique. The model accurately reproduced the distributions of patients in the different fibrosis stages at two subsequent biopsies. A quantification of the role of cofactors in the progression of the disease, and the impact of interferon treatment are given. In subjects who are 40 years old and have been infected for 10 years, the model predicted that 274 of 1000 untreated patients, but only 42 of 1000 treated patients, would progress from F0 or F1 to F3 or F4 fibrosis over the next 5 years. The model also confirms that as age and duration of infection increase, the risk of fibrosis pro- gression increases, while the impact of treatment with interferon decreases. Hence Markov modelling is an accurate tool in the ana- lysis of fibrosis progression in chronic hepatitis C. It will be valuable for the quantification of the effect of new treatments on fibrosis progression in hepatitis C. Keywords: chronic hepatitis C, fibrosis, Markov modelling, prediction, risk factors, treatment.