Annals of Internal Medicine Estimates of the Cost-Effectiveness of a Single Course of Interferon-alpha2b in Patients with Histologically Mild Chronic Hepatitis C

Annals of Internal Medicine 15 November 1997. 127:855-865. William G. Bennett, MD; Yuji Inoue, MD; J. Robert Beck, MD; John B. Wong, MD; Stephen G. Pauker, MD; and Gary L. Davis, MD

Background: Chronic hepatitis C is a major cause of illness and death in the United States. Interferon-alpha2b can induce clinical, biochemical, and virologic remission in some patients with chronic hepatitis C, but the long-term cost-effectiveness of this treatment, particularly in patients with histologically mild disease, is unknown.

Objective: To estimate the cost-effectiveness of interferon-alpha2b in mild chronic hepatitis C.

Design: Meta-analysis of five prospective trials and cost-effectiveness analysis. Projection of the clinical and economic outcomes expected from loss of hepatitis C virus was done by using a Markov simulation. The potential effect of uncertainty in the model assumptions was tested by using sensitivity analyses.

Data Sources: Search of the MEDLINE database, opinions of expert panels, hospital cost data, and adjusted physician charges.

Patients: Hypothetical cohorts with histologically mild chronic hepatitis C.

Intervention: The model assumed a single 6-month course of recombinant interferon-alpha2b.

Measurements: Life expectancy, quality-adjusted life expectancy, costs, and marginal cost-effectiveness ratios from a managed care perspective.

Results: In 27% of patients with mild chronic hepatitis C treated with interferon-alpha2b for 6 months, serum alanine aminotransferase levels permanently returned to normal and viral status remained negative. The model estimated that interferon-alpha2b treatment in this population should increase life expectancy by 3.1 years if given at 20 years of age, by 1.5 years at 35 years of age, and by 22 days at 70 years of age; discounted marginal cost-effectiveness ratios are $500, $1900, and $62 000 per year of life gained, respectively. Varying the long-term response rates and progression rates for mild and moderate chronic hepatitis to near zero in sensitivity analyses substantially affected the results: Ratios ranged from $31 000 for a 20-year-old patient to $640 000 for a 70-year-old patient.

Conclusions: On the basis of estimations in this mathematical model of the natural history of chronic hepatitis C, treating mild chronic hepatitis with interferon-alpha2b should prolong life expectancy at a reasonable marginal cost per year of life gained, particularly in younger patients.

Ann Intern Med. 1997;127:855-865. Annals of Internal Medicine is published twice monthly and copyrighted © 1997 by the American College of Physicians.

For author affiliations and current author addresses, see end of text.

It is currently estimated that the hepatitis C virus (Hepatitis C Virus) chronically infects 3.9 million persons in the United States and is the most common cause of chronic liver disease (1). Although chronic hepatitis C usually progresses slowly and usually leads to cirrhosis and major illness only after decades (2), it results in approximately 8300 deaths each year and accounts for 20% of the 3650 liver transplantations performed in the United States annually (1, 3).

Recombinant interferon remains the only agent approved by the U.S. Food and Drug Administration for the treatment of chronic hepatitis C and is usually administered at a dosage of 3 million U three times per week for 6 months. However, although more than 40% of patients overall initially respond with normalization of the serum alanine aminotransferase (ALT) level and loss of detectable Hepatitis C Virus RNA during treatment, most patients relapse (4, 5). Only 10% to 15% of patients achieve a sustained long-term response to a single 6-month course of interferon therapy (4, 5).

Patients with histologically less advanced liver disease have a greater long-term response to interferon, but treatment of these patients remains controversial because interferon is expensive and can be associated with side effects, most patients with chronic hepatitis C relapse and need to be re-treated, and histologically mild disease progresses relatively slowly (6).

Given the lack of controlled clinical trials on the effect of interferon therapy in patients with mild chronic hepatitis C, the modest long-term response rate to interferon, and the many years usually required before disease complications arise, we developed a decision analytic model (7) to project the immediate responses to treatment and predict the long-term outcomes on the basis of current natural history data. We sought to determine whether treatment of histologically mild chronic hepatitis C with a single 6-month course of interferon-alpha2b would affect life expectancy and lifelong costs.

Methods Decision Analytic Model To determine the expected consequences of interferon therapy in patients with mild chronic hepatitis C, we developed a decision analytic model to simulate disease progression and compare standard care with interferon-alpha2b treatment. The natural history of chronic hepatitis C was modeled by using a Markov simulation in which hypothetical cohorts of identical patients with histologically mild chronic hepatitis move through states of health defined by clinical and histologic descriptors (Figure 1). Time is represented by annual cycles during which patients may remain in the same histologic or clinical state; progress or regress to another histologic or clinical state; die of liver disease; or die of other causes as a function of sex, race, and attained age. The simulation was carried out in each cohort until all patients died of liver-related or other causes.

By recording the proportion of the cohort remaining alive and treatment costs for each year, the simulation estimated the life expectancy and lifetime cost associated with each treatment intervention. Calculations were done by using DecisionMaker 7.0 (Pratt Medical Group, Boston, Massachusetts) (8). Because the quality of life associated with some health states may be less desirable than that associated with other states, we also adjusted life expectancy for quality of life on a scale from 0 (dead) to 1 (perfect health). Using a modified Delphi approach, an expert panel of hepatologists estimated the quality of life for each health state (9). In these analyses, patients who were alive but in less desirable health states were not given full credit for each year lived but instead received only partial credit (for example, 0.7 years for 1 year of life with cirrhosis). At the end of the simulation, the quality-adjusted life-years were summed, yielding the quality-adjusted life expectancy.

Data Sources Likelihood of Events

The natural history of hepatitis C was estimated from published studies. When the sample size was small or follow-up was short, several studies were pooled. All likelihood estimates were reviewed by an expert panel of hepatologists and statisticians and, after extensive discussion, were modified where appropriate.

Treatment Response

Treatment responses were determined according to baseline histologic findings by reanalysis of the pooled data from five clinical trials involving 287 patients with chronic hepatitis C (4, 10-13). These studies were selected because they all used the same treatment regimen (recombinant interferon-alpha2b at a fixed dose of 3 million U administered three times weekly for 6 months), had systematic follow-up after treatment, and had liver biopsy slides and study databases available for review (provided courtesy of Schering-Plough Research Institute with the consent of the studies' principal investigators). All patients were positive for antibody to Hepatitis C Virus and had no evidence of coexisting liver diseases. A single pathologist who was blinded to liver enzyme levels, clinical history, and response to therapy reviewed the 221 available pretreatment liver biopsy specimens and established the correlation of histologic findings and response. We used the histologic activity index of Knodell (14) and a modification of the classification of Desmet and coworkers (15). In the absence of cirrhosis or bridging fibrosis, mild chronic hepatitis required a Knodell periportal inflammation score of 0 or 1 and moderate chronic hepatitis required a score of 3 to 10. Regardless of the Knodell inflammation score, chronic hepatitis with fibrosis required a Knodell fibrosis score of 3 (bridging fibrosis), and chronic hepatitis with cirrhosis required a score of 4 (cirrhosis). These categories approximated previously used histologic definitions of chronic persistent hepatitis or mild chronic active hepatitis, moderate chronic active hepatitis, chronic active hepatitis with bridging, and cirrhosis, respectively.

We used traditional definitions of response (4): Persons with no response had ALT levels that did not return to normal by the end of treatment, persons with an end-of-treatment response had an unsustained normalization of the serum ALT level, and persons with a sustained response had a persistently normal serum ALT level for at least 6 months after completion of therapy.

An end-of-treatment response occurred in 64% of patients with mild or moderate chronic hepatitis without fibrosis, 42% of those with chronic hepatitis with fibrosis, and 28% of those with cirrhosis. A sustained response occurred in 31% of those with mild or moderate hepatitis, 11% of those with chronic hepatitis with fibrosis, and 9% of those with cirrhosis. Table 1 lists similar rates in studies that used the same dose and duration of interferon-alpha2b treatment (4, 11, 13, 16-18). Because approximately 14% of persons who had a sustained response have either persistent viremia or virologic relapse despite a persistently normal serum ALT level (19), we reduced the 31% sustained response rate by 14% to estimate a durable long-term viral-negative response rate of 27% in patients with mild to moderate chronic hepatitis.

Costs

Because charges often vary among providers and can be artificially inflated because of payer mix and institutional costs, the use of charges tends to bias an economic analysis in favor of treatment by increasing the cost of disease. Thus, we used costs or adjusted charges, not full charges, to make the model more widely applicable and to remove a potential bias in favor of interferon-alpha2b treatment. We used a managed care perspective and variable cost estimates (the amount spent by the hospital to care for one additional patient with the illness) based on inpatient variable cost estimates for actual patients with hepatitis C-related hospitalizations, including hospital and physician costs at the University of Florida. Because variable cost data were not available for outpatient medical care, charges for outpatient physician visits, outpatient laboratory evaluations, and radiography were summed and adjusted by a 50% cost-to-charge ratio to estimate complete outpatient costs. The wholesale cost of outpatient medications was obtained from the 1995 Red Book (20). A panel of hepatologists estimated the frequencies of outpatient visits and laboratory tests and the amount of each medication used for each health state per year. Finally, we did not consider the cost of adverse reactions to interferon-alpha2b because severe complications are unusual and typically resolve with discontinuation of treatment, rarely incurring additional cost (4).

Assumptions of the Model We assumed that patients with relapse are not re-treated and that their subsequent prognosis is identical to that of patients with no response (that is, a short-term response to interferon-alpha2b conveys no long-term benefit). We also assumed that patients who lose Hepatitis C Virus either spontaneously or as a result of treatment will not develop progressive liver disease (21-23). Indeed, long-term histologic follow-up of patients with sustained response has found that inflammation resolves and fibrosis regresses after 2 years (23).

Because data on the effect of extrahepatic complications of Hepatitis C Virus infection on disease progression, morbidity, mortality, and response to treatment are insufficient, we could not model the impact of these data.

We could not determine with certainty the age-dependent rate of liver disease progression from published studies, although some reports suggest that histologic progression may be accelerated in patients older than 55 years of age (2, 24). Thus, we excluded age from the model. Because this exclusion may cause the model to underestimate progression in older patients, we would also underestimate disease complications, death, and cost. This, in turn, would reduce the apparent cost of untreated disease and make any treatment effect on cost more difficult to detect. Thus, exclusion of age-dependent disease progression from the model biases against interferon-alpha2b in older patients.

We did not consider serial liver biopsies. Because we assumed that no re-treatment would be given, biopsy would not affect treatment and would only add cost and morbidity. Thus, although our model contains other histologic states, these states remain unobserved clinically until patients develop decompensated liver disease. The cost of follow-up then equals the cost for patients with mild chronic hepatitis until patients are found to be viral negative or until they present with decompensated liver disease. This is a bias against interferon-alpha2b.

Although the model permits liver transplantation for cirrhosis, it does not consider liver transplantation for hepatocellular carcinoma, which, although performed in some centers, generally remains an investigational therapy. After patients undergo liver transplantation, we did not consider decreased survival from recurrent hepatitis C or hepatocellular carcinoma because of inadequate data. This is a bias against interferon-alpha2b.

We could not consider viral factors, such as genotype, pretreatment level of viremia, or presumed source of infection, in this model. Although these factors may affect response to interferon-alpha2b treatment, they were not available for most patients, and retrospective analysis of stored serum samples for measurement of viral levels may be unreliable (25).

Data Natural History of Chronic Hepatitis C

Because Hepatitis C Virus has only recently been described (26) and the progression of chronic Hepatitis C Virus infection is slow (2), data on the natural progression of this disease are limited. We estimated the natural history of hepatitis C from three retrospective observational studies that included serial liver biopsies of patients with non-A, non-B chronic hepatitis. These studies comprised 47 patients with mild chronic hepatitis (mean duration of follow-up ± SD, 8.9 ± 9 years) and 79 patients with moderate chronic hepatitis (mean duration of follow-up, 6.6 ± 6 years) (8, 27, 28) (Table 2).

We estimated the probability of spontaneous resolution of hepatitis C to be 0.2% on the basis of a study in which 2 of 111 patients who were positive for antibody to Hepatitis C Virus and had chronic hepatitis developed persistently normal ALT levels, spontaneous clearance of Hepatitis C Virus RNA (as measured by polymerase chain reaction), and normal liver histologic findings over a mean follow-up of 8.8 years (29).

We used the only published natural history study of patients with compensated cirrhosis and hepatitis C (30) to determine the likelihood of development of hepatocellular carcinoma and decompensated cirrhosis (Table 2). Because the literature suggests that the mortality rate from decompensated liver disease (12% to 68%) depends on the presenting mode of decompensation, we stratified decompensation into diuretic-sensitive ascites, diuretic-refractory ascites, variceal hemorrhage, and hepatic encephalopathy. Salerno and colleagues' study (31) of 110 patients with diuretic-sensitive ascites (excluding patients who developed refractory ascites) had an annual excess mortality rate of 11% over the 3-year follow-up. Among patients with refractory ascites, the annual excess mortality increased to 33% (31). The Veterans Affairs Cooperative Sclerotherapy trial (32) of 253 patients with initial variceal hemorrhage yielded a first-year excess mortality rate of 40%, which is similar to the rates reported in other published studies (33, 37). Among the 111 patients who survived the first year, the subsequent annual excess mortality rate was 13% over a 3-year follow-up (32); this is similar to the 12% rate in Christensen and colleagues' study after variceal hemorrhage (33). This study also found that 108 patients presenting with hepatic encephalopathy without variceal hemorrhage had an annual excess mortality rate of 68% for the first year and 40% for subsequent years.

Because the natural history of decompensated liver disease may be altered by liver transplantation, we had to estimate the number of patients with chronic hepatitis C who were eligible for liver transplantation in the United States. The Centers for Disease Control and Prevention estimates that approximately 8300 persons die of complications of chronic hepatitis C each year (38). Given that the median duration of survival is 1.6 years for patients with decompensated liver disease, 35% of them die each year (39). From these data, it appears that each year, 23 700 (8300 ÷ 0.35) persons have decompensated liver disease related to chronic hepatitis C. The United Network of Organ Sharing (UNOS) reported that 730 (20%) of the 3650 liver transplantations performed in the United States in 1994 were done for complications of chronic hepatitis C (40). Thus, the annual probability of liver transplantation for patients with decompensated hepatitis C is approximately 3.1% (730 ÷ 23 700). We assumed that, on average, the probability of liver transplantation was equal across all degrees of liver decompensation because more severely ill patients are also more likely to die or become ineligible before receiving a liver transplant.

On the basis of three studies involving 2166 patients undergoing liver transplantation for chronic hepatitis (34-36), we estimated the annual excess mortality rate to be 21% for the first year after liver transplantation and 5.7% for all subsequent years.

Sensitivity Analysis

Because of the variation in published data and expert estimates, we examined the effect of varying all values over a wide range to assess their effect on the results. We used a range identified by using the 95% CIs, halved and doubled cost and data estimates, or used the range from the literature (whichever was greatest) (Table 2). Wherever appropriate, we used estimates that biased against interferon-alpha2b therapy.

Quality of Life

Although chronic hepatitis C has documented effects on quality of life (41), estimates for the various stages of the disease and for treatment remain unknown. Ideally, these estimates should be obtained from patient surveys, but no surveys are available. Instead, we asked a panel of hepatologists to use linear scaling and time trade-off methods to estimate the quality of life or utility for each health state listed in Table 3 on a scale of 0 (death) to 10 (perfect health). We used the median values in the base case and the extreme low or high utility estimate in the sensitivity analysis.

Costs and Discounting

Discounting assumes that money saved or spent and years saved or lost in the future are not worth as much as they are today. This process reduces future costs and future life-years by a fixed percentage each year (7) so that up-front costs, such as the cost of interferon-alpha2b treatment, weigh more than the projected future costs of potential late complications from the disease. Because discounting, particularly of survival benefits, is somewhat controversial, we also calculated costs and life expectancy without discounting (Table 4). Costs, wholesale prices, or charges adjusted by a cost-to-charge ratio were used for all calculations. The cost of a 6-month course of interferon-alpha2b at 3 million U three times weekly was $2150 (20); this cost increased to $2511 after the addition of drug-induced costs for counseling patients, additional follow-up laboratory evaluations, and visits. However, for patients who were unresponsive to interferon-alpha2b and had treatment discontinued after 3 months, the cost was reduced to $1253. Such patients have almost no chance of subsequent response even with continued treatment (1, 4).

Industry Role This study was funded in part by an unrestricted grant from Schering-Plough to the investigators' institutions. The grant sponsor had no input into study design, data analysis, manuscript preparation, or the decision to submit the paper for publication. The Schering-Plough Research Institute (the branch of the company that conducts clinical trials) facilitated access to clinical trial databases, including the original liver biopsy slides. However, this information was provided as raw data and the authors performed all analyses.

None of the authors owns stock or has any other relationship with Schering-Plough that would represent a conflict of interest. One of the authors (GLD) has grant support from the Schering-Plough Research Institute for clinical research studies.

Results Model Validation Because data on the natural history of hepatitis C are recent and somewhat uncertain, we tested the validity of our analysis by comparing model predictions with findings from published studies. Most hepatologists believe that Seeff and colleagues' long-term follow-up study of 568 patients with acute post-transfusion hepatitis (2) represents a very conservative view of disease progression, supporting a "benign" natural history of chronic hepatitis C. This study included patients with acute post-transfusion hepatitis, but an unknown proportion developed chronic hepatitis. Assuming that 80% of these patients with post-transfusion hepatitis developed chronic hepatitis and considering a cohort that matches the age and mortality rates from nonliver disease of Seeff and colleagues' study, our Markov model predicted that 2.3% of patients would die of liver disease after 18 years compared with the 3.3% observed in Seeff and colleagues' study. We also compared the prognosis for patients with compensated cirrhosis predicted by our model to that observed in the sole study examining this outcome (30). Our model predicted a 5-year survival rate of 55%; the rate in Fattovich and colleagues' study (30) was 50%. Thus, our model seems to provide relatively conservative estimates of disease progression, biasing our results against interferon-alpha2b.

Base-Case Analysis Our base-case analysis considers a 35-year-old person presenting with histologically mild chronic hepatitis C. When undiscounted costs are used, interferon-alpha2b would save $3500 by reducing long-term medical costs in responding patients and would increase the cohort's life expectancy by 1.5 years (Table 5). Discounting at a 5% annual rate decreases the economic impact of future medical expenditures and reduces the benefit of future improved survival associated with interferon-alpha2b so that interferon-alpha2b costs $490 (discounted) more than standard care while adding 0.26 discounted life-years, yielding a discounted marginal cost-effective ratio of $1900 per life-year gained (Figure 2).

As expected, the survival benefit and cost-effectiveness of interferon-alpha2b for mild chronic hepatitis C decreased with advancing age (Figure 3 andFigure 4). For patients 20 to 50 years of age, the increase in life expectancy ranged from 3.1 to 0.5 years and the discounted marginal cost-effectiveness ratios ranged from $530 to $7100. For 60- and 70-year-old patients, life expectancy increased by 2 and 0.7 months, respectively, and the discounted marginal cost-effectiveness ratios increased to $19 000 and $62 000, respectively.

Sensitivity Analysis We varied each model variable over a wide range of possible values (Table 2). When quality-of-life adjustments were used (Table 3), interferon-alpha2b increased life expectancy by 3.7 quality-adjusted years at a discounted marginal cost-effectiveness ratio of $430. Although our analysis used the traditional 5% discount rate, it was recently suggested that a 3% discount rate might be more appropriate (42). If the analysis were to use a 3% discount rate instead of the traditional 5% rate that we report here, interferon-alpha2b treatment becomes cost saving (by prolonging life and reducing lifetime expenditures).

Only four variables (cost of interferon-alpha2b, response to interferon-alpha2b, rate of transition from mild chronic hepatitis to moderate chronic hepatitis, and rate of transition from moderate chronic hepatitis to cirrhosis) changed the results significantly. Table 6 shows the impact on marginal cost-effectiveness of varying each of these variables over a wide range and compares these with the base-case analysis. Although altering the cost of interferon-alpha2b in the model significantly increases the marginal cost-effectiveness ratio, drug cost is unlikely to be higher than the value we used in the base-case analysis. Indeed, the true cost (not charge) is probably overestimated by the baseline value because bulk buying significantly reduces the cost of the drug. A worst-case scenario was defined in which all four of these variables (values in Table 6 marked by a dagger) were biased against interferon-alpha2b to the limits of what was considered plausible (by halving histologic progression rates, reducing the response to interferon-alpha2b to the viral-negative response observed in patients with fibrosis and cirrhosis, and increasing the cost of interferon-alpha2b fourfold). When this worst-case scenario is used, treatment increased life expectancy by a range of 5 months for a 20-year-old patient to 2 days for a 70-year-old patient at discounted marginal cost-effectiveness ratios of $67 000 to $1.3 million, respectively. Leaving the cost of interferon-alpha2b at $2511 but changing the other variables resulted in a discounted marginal cost-effectiveness ratio that ranged from $31 000 for a 20-year-old patient to $640 000 for a 70-year-old patient. If these latter three values (response to interferon-alpha2b and progression rates) were changed to the limits shown for the sensitivity analysis, our model would have predicted that Seeff and colleagues' study should have found only 0.8% liver-related deaths after 18 years rather than the 3.3% that they did find.

Because we derived variable cost estimates from actual patients with hepatitis C from just one institution, we also estimated costs based on diagnosis-related groups (DRGs), as developed by the Health Care Financing Administration (43, 44). According to 50th-percentile DRG reimbursements, interferon-alpha2b therapy was cost saving for 20- to 35-year-old patients; discounted marginal cost-effectiveness ratios were less than $30 000 for patients younger than 65 years of age but increased to $58 000 for 70-year-old patients.

Finally, some may believe that response to interferon may occur only in patients who would spontaneously become viral negative. Thus, to explore whether earlier loss of virus resulted from interferon, we modified the model to assume a 6.2% annual probability of spontaneously becoming viral negative in the standard therapy strategy (>30 times the rate used in the model) so that over a 5-year period, 27% of this group becomes viral negative. In the interferon-alpha2b group, we assumed that all patients who do not become viral negative after treatment will never become viral negative. In this analysis, interferon-alpha2b increased life expectancy by 0.3 years at a marginal cost-effectiveness ratio of $22 000 per discounted quality-adjusted life-year gained. Earlier conversion to a viral-negative state protected some patients from developing complications.

Our base-case analysis biased against interferon-alpha2b by excluding quality-of-life adjustments and using conservative discounted variable costs. When both quality-of-life adjustments and DRG reimbursements were used, interferon-alpha2b was cost saving for 20- and 35-year-old patients and had a discounted marginal cost-effectiveness ratio less than $5000 for all patients 45 to 70 years of age.

Discussion Chronic hepatitis C usually has a slow rate of progression; only a minority of patients develop cirrhosis with hepatic failure. The rate of progression to cirrhosis depends on the amount of inflammation seen on liver biopsy: Patients with mild hepatitis have a slow rate of progression over decades, whereas those with severe inflammation or fibrosis usually develop cirrhosis within 10 years (6). Interferon-alpha2b treatment achieves a sustained viral-negative response in only a minority of patients (4, 5). Although the indication for interferon-alpha2b treatment is unequivocal in patients with moderately severe inflammation or early fibrosis (45), the benefits of treatment in those with histologically mild disease is less certain. Long-term placebo-controlled trials might establish the benefit of interferon-alpha2b in this group; such a study would be extremely expensive, however, and the slow rate of spontaneous progression would mandate decades of follow-up in thousands of patients in order to observe a treatment effect. Because treatment decisions must be made now, we turned to decision analysis to assess long-term clinical outcomes and the cost-effectiveness of interferon-alpha2b in treating chronic hepatitis C on the ba