Hepatitis C virus (HCV)–related liver disease is a major cause of chronic liver disease worldwide and is the most common indication in European and US liver transplant centers (see http://www.eltr.org and http://www.unos.org). Simulation analyses have suggested that in developed countries, a gradual decline in the infected population is expected in the future, but because of the extended period between infection and the emergence of complications, the burden associated with HCV infection will continue to rise, with a 2- to 4-fold increase in the proportion of patients with cirrhosis in the next 20 years resulting in an increase in mortality from liver failure and/or hepatocellular carcinoma (HCC)1, 2 as well as an increase in the demand for liver transplantation.
In developed countries, the incorporation of increasingly innovative healthcare technologies and therapies has given rise to a significant increase in costs; in fact, technically possible strategies may not be acceptable from an economic point of view. It is hence essential to decide the best way to spend the limited resources that we have. Pharmacoeconomics refers to the scientific discipline that identifies monetary dimensions, attributes them to health interventions, and compares the values of these interventions.3 In order to do so, it mainly uses cost-effectiveness and cost-utility analyses. The main information sources for cost-effectiveness analyses are clinical trials and meta-analyses; unfortunately, the follow-up time is short in comparison with the natural history of chronic diseases. For instance, in the case of chronic HCV, available data exist that can be used to estimate the development of cirrhosis in the first 2 decades after infection. However, data beyond 20 years of infection are mostly missing4 and need to be extrapolated by means of suitable models. These models allow us to establish the long-term influence of a given intervention on the costs and quality of life.5 No pharmacoeconomic model is a perfect representation of reality. Indeed, the projections provided by these models are limited by the accuracy of the assumptions, that is, the reasonableness of the assumptions and their closeness to reality. Markov models are especially useful for predicting the evolution of chronic illnesses such as HCV.5 The models assume that a patient remains for a finite time in a given status (chronic C hepatitis, compensated cirrhosis, decompensated cirrhosis, HCC, liver transplantation, or death) and moves to the following status according to a known yearly probability estimated from published literature. For each status, there is a given value of quality of life that worsens along with the evolution to a more advanced Markov status. On the basis of these assumptions, the duration of time in each status is calculated together with the survival probability; this information is then used to estimate the overall costs required by a treated patient in his or her lifetime. Sensitivity analyses are used to introduce changes into these soundly based estimations and to assess how robust the results of the model are to these changes.
In addition, any health intervention is intended to increase not only the quantity of life but also the quality of life. In cost-utility analyses, the improved quality (morbidity) and quantity (longevity) of life are combined in a standard measurement unit: quality-adjusted life years (QALYs). For a given patient, the QALY value is the result of the multiplication of the quality-of-life score (observed in a specific Markov status) by the length of time for which this particular patient remains in this specific status. In the international literature, it is thought that a health intervention has an acceptable cost-effectiveness ratio if the additional cost per QALY saved is less than $50,000.6
With the current standard of care, 40% to 50% of genotype 1–infected patients achieve a sustained virological response (SVR). Patients with cirrhosis and patients with recurrent HCV after liver transplantation constitute a particularly difficult-to-treat and difficult-to-cure population with lower SVR rates. However, in nontransplant patients with HCV-related cirrhosis, SVR is associated with a reduction of liver-related mortality, which is derived from a reduction of complications and HCC development.7 In addition, in some of these patients, cirrhosis regression may occur and result in improved survival.8 Similarly, in transplant patients with recurrent hepatitis C, SVR is associated with improved survival and a reduced risk of graft failure due to a slower rate of disease progression (and even histological improvement) and a reduced risk of developing complications.9-11 In a recent study, the actuarial 5-year survival was 96% in responders and 69% in nonresponders,12 and these findings are consistent with the results from other studies.9-11
Because of the impact of HCV recurrence on graft and patient survival, the need to optimize the management of hepatitis C patients is one of the most pressing issues facing transplant physicians.13 In this issue of Liver Transplantation, Saab et al.14 report their attempt to determine the most cost-effective timing for pegylated interferon and ribavirin therapy in patients with advanced liver disease infected with HCV genotype 1. In order to do so, they compared the cost-effectiveness of HCV treatment at different stages of the disease in a hypothetical cohort of 4000 55-year-old naive patients with cirrhosis followed for 17 years (life expectancy of 72 years) or until death. In this cohort, 4 different strategies of treatment were analyzed with a Markov model (no treatment, antiviral therapy in patients with compensated cirrhosis, antiviral therapy in patients with decompensated cirrhosis, and antiviral therapy in patients with progressive fibrosis due to recurrent HCV post-transplantation). Patients assigned to treatment strategies received antiviral treatment only when they reached the natural history state at which they were designated to be treated, and no repeated courses of antiviral therapy were accepted. The model included variables such as the resources used to treat cirrhosis and its complications (including liver transplantation), the impact on patient survival and quality of life, and the investments to treat HCV infection. Outcome measures included the total cost per patient, number of QALYs, cost per QALY saved, and number of deaths, HCCs, and required transplants. According to the model, the treatment of patients with compensated cirrhosis was the most cost-effective strategy and resulted in improved survival and lower costs in comparison with all other strategies. In the reference cohort of 55-year-old naive patients with compensated cirrhosis, treatment resulted in 119 fewer deaths, 54 fewer HCCs, and 66 fewer transplants in comparison with patients in the no-treatment strategy group. Treatment during decompensated cirrhosis and posttransplant HCV recurrence was also cost-saving, but to a lesser extent, in comparison with no treatment, with posttransplant treatment being less cost-effective than treatment during decompensation. Sensitivity analysis revealed the cost-effectiveness assessment to be robust, with only variations in graft failure rates influencing the final results of the model. On the basis of these estimates, the authors recommend expeditious administration of antiviral treatment to patients with compensated HCV cirrhosis to prevent them from developing more advanced disease. Although we agree with this final recommendation, there are clear limitations in the use of these models that are mostly due to the assumptions on which they are based. Indeed, in these models, the annual probability of progressing to a more advanced disease state, assumed rates of SVR and discontinuation of treatment due to side effects and absence of early virological response, mortality, transplant probability, use of growth factors, and costs are derived from a review of the published literature. Unfortunately, data regarding some of these rates, particularly those of antiviral therapy in patients with decompensated cirrhosis and/or patients with recurrent hepatitis C, are controversial and are based on small and mostly nonrandomized studies with wide ranges of results. In addition, the model includes only 3 broad categories of liver disease status (compensated cirrhosis, decompensated cirrhosis, and recurrent aggressive HCV hepatitis), whereas it is generally accepted that these are not uniform and static situations but rather a continuum in disease severity. For instance, a patient who has compensated cirrhosis with preserved hepatic function (albumin > 3.5 g/dL, total bilirubin < 1.5 mg/dL, or international normalized ratio < 1.5), an absence of clinical decompensation, and a normal platelet count is likely to respond differently to antiviral therapy than a patient who has compensated cirrhosis with no history of clinical decompensation but has portal hypertension, a low platelet count, and normal hepatic function within the limit (albumin = 3.5 g/dL, bilirubin = 1.5 mg/dL, international normalized ratio = 1.5). Similarly, a Child C decompensated patient with a Model for End-Stage Liver Disease score greater than 18 is clearly different from a Child B patient with a Model for End-Stage Liver Disease score of 17. SVR rates as well as risks associated with antiviral therapy in these circumstances differ substantially.
Additional assumptions that limit the usefulness of the model include “the lack of [a] repeated course of antiviral therapy,” which excludes patients who may derive a benefit from these repeated interventions, and “the 30 weeks [sic] duration of therapy in the decompensated group,” which excludes shorter course strategies aimed at achieving an on-treatment virological response and prevention of HCV recurrence after transplantation.15 Furthermore, patients were assumed to enter the study without contraindications to therapy and were assumed to be candidates for treatment during the entirety of the study. However, in clinical practice, a proportion of patients have contraindications for therapy, particularly cytopenias. Only naive patients were considered in the model, whereas the strategy in untreated patients may not be the same as the strategy in subjects who have been proven to be resistant to interferon-based regimes. The model did not calculate various death rates per year according to age, whereas cost-effectiveness ratios may change with advanced age. Finally, another unproven assumption is related to the prognosis of patients not achieving an SVR. The model considers patients who fail to achieve an SVR to have a prognosis similar to those who do not receive treatment. There are studies, however, that show a slower rate of fibrosis progression in biochemical responders despite the lack of virological response, particularly in the transplant setting.16
In the current study, posttransplant antiviral treatment was found to be cost-saving in comparison with no treatment (although the differences were smaller). In addition, this strategy was found to be less cost-effective than treatment during decompensation, even though higher rates of SVR and lower therapy discontinuation rates were assumed for the posttransplant patients. This may be partly explained by the higher use of growth factors and the performance of yearly protocol biopsy in the posttransplant period; both circumstances increased the costs. Nevertheless, in clinical practice, physicians feel more comfortable treating recurrent hepatitis C in clinically stabilized patients (months or years after the transplant procedure) than in patients with decompensated cirrhosis. In addition, the cost-effectiveness results were robust to all variables tested in sensitivity analysis, with the exception of the annual rate of graft failure in patients with or without SVR. In both groups of patients, the most cost-effective strategy changed significantly if the annual transition to graft failure decreased by approximately 50%. These results are not surprising because of the difficulties in establishing annual transition rates in transplant patients; indeed, the posttransplant evolution is a dynamic process that can be modified by many factors, including the time period since transplantation and the presence or absence of SVR. Indeed, the model uses data on the transition to graft failure at 5 years post-transplantation but not at later time points. Moreover, the transition data for patients with SVR are based on a single retrospective study.17 Finally, as in the pretransplant setting, the model does not stratify patients into various subgroups based on disease severity and the time from transplantation. Recent data have shown that the results of antiviral therapy in liver transplant patients with recurrent hepatitis C are significantly worse if treatment is started at an advanced stage of fibrosis.9, 18 Further randomized controlled trials comparing these 2 strategies (the treatment of decompensated patients versus treatment in the posttransplant period) are needed. Meanwhile, no universal recommendations regarding these strategies can be made.
In summary, the study by Saab and colleagues14 is a valuable tool for deciding the most cost-effective time to start antiviral therapy in patients with advanced HCV disease. The lack of subgroup analysis, the use of some unproven assumptions, and the lack of multivariate sensitivity analysis (the model was subjected to univariate sensitivity analysis to assess its robustness, and only the changes in a single variable were examined each time) are important limitations of the model in this very complex scenario in which so many variables may play an important role. Physicians must decide whether the most cost-effective approach is the most appropriate for each individual. In addition, because products and services have different prices in different regions or countries, pharmacoeconomic analysis must be adapted to local reality. Finally, this type of analysis will have to be re-evaluated in future years once newer therapies against HCV come into play.