COST-EFFECTIVENESS ANALYSIS OF INTERFERON-α THERAPY IN THE TREATMENT OF CHRONIC HEPATITIS B IN TAIWAN

Transcription

1 R.F. Pwu and K.A. Chan COST-EFFECTIVENESS ANALYSIS OF INTERFERON-α THERAPY IN THE TREATMENT OF CHRONIC HEPATITIS B IN TAIWAN Raoh-Fang Pwu and K. Arnold Chan 1 Background and Purpose: Although interferon-α has been approved for chronic hepatitis B treatment in Taiwan, the high price necessitates study of its costeffectiveness to define its role from a burden of disease perspective. Method: A simulation model was constructed to project the long-term effects of interferon treatment or standard care for a hypothetical group of 35-year-old chronic hepatitis B patients in Taiwan. Data on interferon treatment efficacy, disease progression, vital statistics, medical expenditure for various liver disease states, and quality-of-life measures based on Taiwan data were collected and used in the model. Results: The model indicated that a 35-year-old chronic hepatitis B patient treated with interferon in Taiwan would have a life expectancy of years, versus years for the same patient who did not receive interferon. Corresponding to this gain in life expectancy of 0.41 years (or 0.18 quality-adjusted life-years, QALYs), the incremental cost-effectiveness ratio (ICER) was 492,000 New Taiwan dollars (NT$; 14,200 US$ at 34.7 NT$/US$) per QALY with an annual 3% discount rate. Monte Carlo simulation taking into account the range of plausible values for all model parameters indicated that 95% of the ICERs were in the range of 65,000 NT$ (1,900 US$)/QALY to 683,000 NT$ (19,700 US$)/QALY, and 90% of the ICERs were below 413,000 NT$ (11,900 US$)/QALY. Conclusion: Interferon has a favorable cost-effectiveness profile in the treatment of chronic hepatitis B in Taiwan. However, the study results cannot be interpreted in isolation and need to be compared with the cost-effectiveness profiles of treatment regimens for other major chronic diseases in Taiwan. (J Formos Med Assoc 2002;101:632 41) Key words: hepatitis B interferon-α cost-effectiveness analysis Markov simulation incremental cost effectiveness ratio Chronic hepatitis B infection and its sequelae are major causes of morbidity, mortality and medical expenditure in Taiwan. Chronic liver disease was the sixth leading cause of death in 2000 and hepatocellular carcinoma (HCC) was the most common cancer in 1997 [1]. The successful national hepatitis B vaccination program implemented in 1984 [2] has reduced the incidence of liver cancer among vaccinated birth cohorts [2], but those already infected with hepatitis B virus (HBV) need effective treatment to control disease progression and prevent life-threatening sequelae [3]. Recombinant interferon-α (IFN-α) was the first therapeutic agent approved in the USA for the treatment of chronic hepatitis B and was introduced to Taiwan in In Asian patients, 5 million units (MU) of IFN-α administered three times per week for 4 to 6 months is effective in the treatment of chronic hepatitis B [4]. This regimen is recommended for use in patients with hepatitis B surface antigen (HBsAg), persistent elevation of alanine aminotransferase (ALT), detectable serum levels of hepatitis B early antigen (HBeAg) and HBV DNA, biopsy-proven chronic hepatitis, and compensated liver disease [5]. In a meta-analysis by Wong et al, the proportion of patients with chronic hepatitis B who had HBeAg clearance after IFN-α therapy was 33%, versus 12% among those Graduate Institute of Epidemiology, College of Public Health, National Taiwan University, Taipei; 1 Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA. Received: 17 January Revised: 29 April Accepted: 13 August Reprint requests and correspondence to: Dr. K. Arnold Chan, Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA. 632 J Formos Med Assoc 2002 Vol 101 No 9

2 Cost-effectiveness of IFN-α in Chronic Hepatitis B without treatment [6]. Similar efficacy has been shown among patients in Taiwan [7]. However, costs for a complete course of IFN-α therapy are high, more than half of patients do not have HBeAg clearance after therapy, and the effectiveness in long-term prevention of liver cirrhosis and HCC has not been adequately documented. In response to such uncertainty, decision-analytic techniques have been used to incorporate existing data to project long-term outcomes and cumulative medical expenditure in two studies of chronic hepatitis B patients treated with IFN-α [8, 9]. Although these studies showed that using IFN-α to treat chronic hepatitis B may provide cost savings over the long term in the USA and UK, their results are not directly generalizable to patients in Taiwan because of the differences in treatment effect, disease progression, and treatment costs for liver cirrhosis and HCC. This study used a Markov simulation model to evaluate the cost-effectiveness profile of IFN-α in the treatment of chronic hepatitis B patients in Taiwan. Methods This study followed international scientific guidelines on cost-effectiveness analysis [10 12] and used data primarily from Taiwan in the simulation model. A basecase model was formulated to compare the projected long-term outcomes of two hypothetical cohorts of patients with chronic hepatitis B. We adopted a societal perspective in the analyses, only counted direct medical costs, and applied a 3% annual discount rate to costs and outcomes. Model and simulation The simulation model of Wong et al was adopted [8]. In the base-case model, a hypothetical cohort of 35-year-old chronic hepatitis B patients with HBeAg, no liver cirrhosis, and a serum ALT concentration greater than twice the upper limit of normal received IFN-α 5 MU thrice weekly for 24 weeks. An identical cohort received the usual supportive care and no IFNα therapy. All patients were followed over the next 65 years or until death to project the lifetime cumulative occurrence of liver diseases and medical care expenditure. The following nine distinct liver disease states were included in the model: three states of chronic hepatitis B (HBsAg+/HBeAg+, HBsAg+/HBeAg-, HBsAg-/ HBeAg-); three states of compensated cirrhosis (HBsAg+/HBeAg+, HBsAg+/HBeAg-, HBsAg-/ HBeAg-); and decompensated cirrhosis, HCC and death (Fig. 1). At the initiation of therapy, all patients had HBsAg+/HBeAg+ chronic hepatitis B. After 1 year, a patient could have stayed in the same disease state or moved on to another disease state (specified by an arrow in Fig. 1). For example, a patient who had HBeAg clearance after IFN-α therapy would move from HBsAg+/HBeAg+ chronic hepatitis B to HBsAg+/ HBeAg- chronic hepatitis B in the second year. The same transition rule applied in subsequent years and the transition probabilities between states was a function of the state that a patient occupied at the time, regardless of the patient s prior history [13]. A patient who did not respond to IFN-α therapy would remain in the initial state of HBsAg+/HBeAg+ chronic hepatitis B for a number of years, then transit to compensated cirrhosis, HCC or death from a non-liverrelated cause. After 65 annual cycles, virtually all hypothetical subjects would be in the absorbing state of death. DATA software (Professional version, TreeAge Software, Inc., Williamstown, MA, USA) was used to construct the simulation models and calculate the incremental cost-effectiveness ratio (ICER) to compare IFN-α treatment and no treatment in chronic hepatitis B patients [14]. A comprehensive search of original research, review articles, conference proceedings, unpublished theses, and books was performed to obtain information on the treatment effect of IFN-α and natural history of chronic hepatitis B in Taiwan. For parameter values reported from more than one study, the median value or the most frequently reported value was used in the base-case model, and the largest and smallest reported values were used to define the range for sensitivity analysis (Table 1). Chronic hepatitis B HBeAg+ HBsAg+ Compensated cirrhosis HBeAg+ HBsAg+ Compensated cirrhosis HBeAg HBsAg+ Decompensated cirrhosis Hepatocellular carcinoma Chronic hepatitis B HBeAg HBsAg+ Death Fig. 1. Liver disease states experienced by the two hypothetical study cohorts: interferon-α treatment and no treatment. Post hepatitis HBeAg HBsAg Compensated cirrhosis HBeAg HBsAg J Formos Med Assoc 2002 Vol 101 No 9 633

3 R.F. Pwu and K.A. Chan Table 1. Interferon-α (IFN-α) treatment efficacy and disease progression for chronic hepatitis B in Taiwan Clinical parameter Annual probability (%) References Base-case Range for sensitivity analysis Loss of HBeAg First year, no treatment [5 8, 15, 17 23, 26] First year, IFN-α2b therapy [5 8, 15, 17 19, 24 26] Each subsequent year, no treatment or [5 8, 15, 17 23, 26] for non-responders to IFN-α Losing HBsAg after loss of HBeAg First year, no treatment [5, 28, 29] First year, IFN-α2b therapy [7, 8, 19] Each subsequent year, regardless of treatment [5, 28, 29] in the first year HBeAg reversion First year, no treatment or IFN-α [7, 8, 15, 24, 27] Each subsequent year, no treatment or IFN-α [8, 27] Progression from chronic hepatitis to compensated cirrhosis HBeAg+/HBsAg [7, 30] HBeAg-/HBsAg [7, 30] HBeAg-/HBsAg [31, 32] Progression from compensated cirrhosis to decompensated [8, 33] cirrhosis Developing hepatocellular carcinoma From chronic hepatitis HBeAg-/HBsAg [31, 32] HBeAg+ or HBsAg [7, 34] From cirrhosis [21, 31, 33, 35 38] Annual excess mortality rates Decompensated cirrhosis [8, 21] Hepatocellular carcinoma [8, 39] Short-term treatment effects. Sustained HBeAg clearance rates 6 to 12 months after IFN-α treatment have been reported to be between 30 and 50%, versus 10 to 20% among untreated controls [5 8, 15 25]. In a meta-analysis of nine studies, Wong et al found an HBeAg clearance rate of 45.6% and used this estimate in a cost-effectiveness analysis [8]. However, reported rates of HBeAg clearance in Taiwan have been lower in general and mostly in the 30% range [7, 17, 25, 26], similar to findings from two studies in Hong Kong [18, 19]. We chose the conservative HBeAg clearance rate estimate of 30% for the base-case model. After the first year, sustained HBeAg clearance is achieved in 85 to 95% of patients [7, 15, 24, 27]. A longterm follow-up study in Taiwan showed that three of 28 patients had HBeAg reversion in the 14th, 30th, and 53rd month after IFN-α therapy, which corresponded to an annual reversion rate of 3% [7]. Our model used the same HBeAg reversion rate for those with spontaneous loss of HBeAg among the untreated cohort. For untreated chronic hepatitis B patients in Taiwan, reported annual HBeAg clearance rates range from to 24% [7, 17, 20 22, 26]. The two studies from Hong Kong reported HBeAg clearance rates of 19 and 28% [18, 19]. We chose an annual HBeAg clearance rate of 12% for the untreated cohort and non-responders among the IFN-α cohort. Spontaneous loss of HBsAg is rare in Asian patients, occurring at a frequency of 0.1 to 1% per year [5, 28, 29]. A wide range of probability of losing HBsAg has been reported among responders to IFN-α treatment: 0% in Taiwan [7], 16.6% in Hong Kong [19], and 18.8% from a meta-analysis [8]. We chose an annual rate of 0.5% for losing HBsAg for all patients regardless of therapy throughout the 65- year simulation period. Development of liver cirrhosis. Lin et al reported that one in 24 patients in Taiwan treated with IFN-α who achieved HBeAg clearance progressed to cirrhosis during a median of 8.2 years of follow-up [7]. An earlier prospective study in Taiwan showed that, among those patients who had spontaneous HBeAg seroconversion, the annual progression rate to liver cirrhosis was 1.3% [30]. Based on J Formos Med Assoc 2002 Vol 101 No 9

4 Cost-effectiveness of IFN-α in Chronic Hepatitis B these two studies, we used an annual rate of progression to liver cirrhosis of 0.8% for HBeAg-/HBsAg+ patients. For untreated patients, the same two studies showed an annual progression rate to liver cirrhosis for HBeAg+/ HBsAg+ patients of 2.4 and 3.9%, respectively [7, 30]. We used an annual rate of progression to liver cirrhosis for these patients of 3.2% in our model. Lastly, for patients who are HBeAg-/HBsAg-, the risk of developing liver cirrhosis is exceedingly small [31, 32]. We assumed an annual progression rate of 0.008% for these patients, which is one-hundredth of the risk among HBeAg-/ HBsAg+ patients. For progression from compensated cirrhosis to clinically recognized decompensated cirrhosis, we used an annual risk of 2.3%, based on the findings of a clinicbased study in Taiwan [33]. Development of hepatocellular carcinoma. Liaw et al reported an annual risk of 2.8% of developing HCC among HBsAg+ chronic hepatitis B patients, regardless of HBeAg status [34]. Half of these patients already had cirrhosis, so we assumed an annual risk of 1.4% of developing HCC among HBsAg+ patients who did not have liver cirrhosis. For those who already have liver cirrhosis, the reported annual risk of developing HCC ranges from 3 to 7.7% [21, 31, 33, 35 38], and we selected a value of 5%. For those who have lost HBeAg and HBsAg positivity, the risk of developing HCC is greatly reduced [7]. We used an annual risk of 0.014% of developing HCC among chronic hepatitis B patients who have lost both HBeAg and HBsAg positivity, which is one-hundredth the risk of developing HCC among HBsAg+ patients. Survival. Tsai et al reported a 5-year survival rate of 30.9% in patients with Child s class C liver cirrhosis, which corresponded to an annual mortality rate of 23.5% [21]. For mortality among HCC patients, Lee et al linked the Taiwan Cancer Registry and Death Registry files and estimated a 5-year survival rate of 15% [39], which corresponded to an annual mortality rate of 37.2% for HCC patients. We assumed that all excess mortality among chronic hepatitis B patients was due to decompensated liver cirrhosis and HCC [8], and that those who did not have liver cirrhosis or HCC would experience the same death rate as the general population, which was estimated from Taiwan vital statistics in 2000 [40]. Costs Three major types of medical expenditure are associated with treatment of liver disease states: acquisition Table 2. Annual treatment costs for chronic hepatitis disease states in Taiwan in New Taiwan dollars Liver disease state Frequency Unit cost Annual cost Range for sensitivity analysis Chronic hepatitis B HBsAg+/HBeAg+ Inpatient/outpatient / ,500 / 2, / 15,696 Total annual cost 15,931 13,300 32,400 HBsAg+/HBeAg- Inpatient/outpatient / ,500 / 2, / 6,613 Total annual cost 6,707 5,600 13,600 HBsAg-/HBeAg- Inpatient/outpatient 0 / 2 23,500 / 2,480 0 / 4,960 Total annual cost 4,960 4,100 10,100 Compensated cirrhosis HBsAg+/HBeAg+ Inpatient/outpatient 0.20 / 8 49,500 / 2,760 9,900 / 22,080 Total annual cost 31,980 23,000 62,700 HBsAg+/HBeAg- Inpatient/outpatient 0.10 / ,500 / 2,760 4,950 / 16,727 Total annual cost 21,677 16,000 42,900 HBsAg-/HBeAg- Inpatient/outpatient 0.08 / ,500 / 2,760 3,960 / 15,593 Total annual cost 19,553 14,600 38,800 Decompensated cirrhosis Inpatient/outpatient 2 / ,500 / 6,912 99,000 / 164,571 Total annual cost 263, , ,100 Hepatocellular carcinoma Inpatient/outpatient 4 / ,500 / 14, ,000 / 227,429 Total annual cost 589, ,100 1,002,100 J Formos Med Assoc 2002 Vol 101 No 9 635

5 R.F. Pwu and K.A. Chan costs for IFN-α, outpatient visits and hospitalization (Table 2). Costs for IFN-α were based on the retail price of a 24-week course of IFN-α treatment (1,350 NT$ per vial of 5 MU in 2001). Outpatient costs for different liver disease states were estimated from interviews with 12 Taiwanese hepatologists about common practice patterns for the management of chronic hepatitis B, which included the frequency of outpatient visits, laboratory examinations, abdominal ultrasound and prescription drugs. Annual costs of outpatient visits for the different liver disease states were estimated based on the 1996 National Health Insurance (NHI) reimbursement rate [41]. The same 12 hepatologists estimated the frequency of hospitalization for different liver disease states. We estimated the average costs of hospitalization for each liver disease state by studying all patients who were hospitalized in National Taiwan University Hospital with a diagnosis of hepatitis, liver cirrhosis or HCC from 1994 through The same NHI reimbursement rates were used until We used the median costs for these liver disease states in the simulation models. Utility for liver disease states We estimated the utility associated with each liver disease state by interviewing 12 Taiwanese hepatologists and 53 patients with different liver diseases (12 patients in the inactive HBsAg carrier state, 20 patients with chronic hepatitis B, 20 patients with liver cirrhosis, and one patient with HCC). The time trade-off method was used in interviews with both hepatologists and patients. We used the median values given by the hepatologists for different liver disease states for the base-case model and the patients range of preferences as ranges of the utilities for liver disease states (Table 3). The HCC patient gave a utility of 0.5, which was the median of the utility values assessed by the hepatologists, and we used the 25th and 75th percentiles of these utility values assigned by experts as the range for sensitivity analysis. We also estimated the decrease in utility associated with discomfort during IFN-α therapy. The utilities were used to calculate quality-adjusted life-years (QALYs) accumulated among the hypothetical cohorts over the 65-year projected follow-up. Sensitivity analysis One-way sensitivity analysis was carried out for every parameter used in the simulation model to evaluate the robustness of the ICER results. We identified influential parameters associated with ranges of ICERs wider than 100,000 NT$/QALY. Monte Carlo simulation was used to evaluate the range of ICERs generated by variation of all parameters. Parameters that had values between 0 and 1 (treatment efficacy, transition probabilities, utility) were assumed to vary according to the beta distribution [42]. Cost parameters were assumed to be normally distributed. Role of funding sources This study was funded by Schering-Plough Taiwan and the Harvard Pharmacoepidemiology Program. The funding sources played no role in design or conduct of the study, or in the reporting of the results. Results Base-case analysis For a hypothetical cohort of 35-year-old patients with chronic hepatitis B, average life expectancy was years for those who received IFN-α treatment and years for those who did not receive IFN-α, yielding a gain of 0.41 years per person. Taking into account the less than perfect utility associated with each liver disease state and applying a 3% annual discount rate, the Table 3. Utility associated with different chronic hepatitis disease states in Taiwan Hepatitis disease states Base-case value Range for sensitivity analysis Long-term utility HBsAg+/HBeAg HBsAg+/HBeAg Chronic hepatitis Compensated cirrhosis Decompensated cirrhosis Hepatocellular carcinoma Short-term utility* Interferon (wk) to 5.6 *Negative values for short-term utility used to account for patients discomfort during interferon therapy. 636 J Formos Med Assoc 2002 Vol 101 No 9

6 Cost-effectiveness of IFN-α in Chronic Hepatitis B discounted total QALYs were for those who received IFN-α treatment and for those who received standard supportive care. Average lifetime costs associated with chronic hepatitis B and its sequelae rounded to the nearest thousand were estimated to be 1,251,000 NT$ (36,100 US$) for a patient who received IFN-α treatment and 1,168,000 NT$ (33,700 US$) if the same patient did not receive IFN-α. Taking into account medical expenditure incurred in different years, the discounted costs were 793,000 NT$ (22,900 US$) for an IFN-α-treated patient and 704,000 NT$ (20,300 US$) if the same patient did not receive IFN-α. For an additional discounted expenditure of 89,000 NT$ (2,600 US$), 0.18 QALY would be gained, corresponding to an ICER of 492,000 NT$ (14,200 US$)/QALY. Sensitivity analysis For one-way sensitivity analyses of all parameters in the model, the ICERs ranged from 130,000 to 745,000 NT$/QALY. The relationship between IFN-α treatment effect and ICER is shown in Figure 2. Assuming a baseline annual HBeAg clearance rate of 12%, treatment effects better than the additional 18% due to IFN-α that was used in the base-case model would yield more favorable ICER values. In the one-way sensitivity analyses, age at initial treatment, discount rate, HBsAg/ HBeAg clearance rates, and disease progression rate from HBsAg+ chronic hepatitis B to cirrhosis had ICER ranges greater than 100,000 NT$/QALY (Table 4). Since the hospitalization costs for liver cirrhosis and Incremental cost-effectiveness ratio (x 1,000 NT$/QALYs) , Base-case Additional HBeAg seroconversion probability attributable to IFN-α Fig. 2. Sensitivity analysis for additional probability of losing hepatitis B early antigen (HBeAg) among patients treated with interferon-α (IFN-α). NT$ = New Taiwan dollars; QALY = qualityadjusted life-years. Annual probability of spontaneously losing HBeAg was assumed to be 12%. HCC were derived from a single medical center, we evaluated the impact of the variation of these costs on ICERs. For decompensated liver cirrhosis, as the annual expenditure varied from 135,200 to 507,100 NT$ (a four-fold difference), the corresponding ICERs ranged from 499,705 to 477,649 NT$/QALY, a less than 5% difference. For HCC, the range of annual expenditure was from 325,100 to 1,002,100 NT$ (a three-fold difference), and the corresponding ICERs ranged from 511,124 to 462,370 NT$/QALY, a 10% difference. Table 4. One-way sensitivity analysis results for selected parameters in interferon-α (IFN-α) treatment for chronic hepatitis B in Taiwan Parameter Range of values Corresponding range of ICER Age at initial treatment , ,000 Annual discount rate 0 5% 231, ,000 Probability of losing HBeAg among those not receiving IFN-α in , ,000 the first year Probability of losing HBsAg after loss of HBeAg (IFN-α treatment, , ,000 first year) Probability of reactivating HBeAg reversion in subsequent years , ,000 Disease progression Probability of developing compensated cirrhosis from HBeAg+/ , ,000 HBsAg+ chronic hepatitis Probability of developing compensated cirrhosis from HBeAg-/ , ,000 HBsAg+ chronic hepatitis Probability of developing hepatocellular carcinoma from cirrhosis , ,000 Quality-of-life adjustments for chronic hepatitis and compensated liver cirrhosis HBeAg+/HBsAg , ,000 HBsAg+/HBeAg , ,000 ICER = incremental cost-effectiveness ratio, in New Taiwan dollars per quality-adjusted life-year. J Formos Med Assoc 2002 Vol 101 No 9 637

7 R.F. Pwu and K.A. Chan In the multi-way sensitivity analysis, the simulation model was repeated 1,000 times, each time allowing each parameter to take on a value according to the pre-specified distribution. The median ICER value was 188,000 NT$/QALY and 95% of ICERs fell between 65,000 and 683,000 NT$/QALY. More than 90% of the ICERs were below 413,000 NT$/QALY. 638 Discussion Cost-effectiveness analysis in medicine involves the application of decision analytic theories to evaluate prevention strategies or treatment regimens. In the absence of large-scale, long-term clinical trials in Taiwan to evaluate the effectiveness and costs of providing IFN-α for all chronic hepatitis B patients, simulation is the only option to evaluate the merit of IFN-α treatment in practice. While recent data suggest that IFN-α treatment would reduce the incidence of liver cirrhosis and HCC and increase life expectancy [7], our findings suggest that a cost of not more than 413,000 NT$ is needed to gain one QALY. Two previous cost-effectiveness studies of IFN-α in the treatment of chronic hepatitis B showed very favorable cost-effectiveness profiles. Under certain assumption scenarios, using IFN-α results in cost savings in the long run in the USA [8] and the UK [9]. However, these results are not generalizable to Taiwan because of differences in treatment efficacy, natural history of chronic hepatitis B infection, and treatment costs for long-term sequelae of chronic hepatitis B. Based on a meta-analysis of nine clinical trials, Wong et al used a probability of HBeAg clearance after IFN-α therapy of 45.6%, but this is a much higher response rate than has been reported in Taiwan. Moreover, as hospitalization costs in the USA are the highest in the world, a modest reduction in hospitalizations for liver cirrhosis and HCC would result in substantial cost savings. The ratio of annual treatment costs for decompensated liver cirrhosis to a 16-week course of IFN-α was 22,100 US$/5,570 US$ [8] (approximately fourfold), while the parameter values that we used were 263,600 NT$/107,000 NT$ (approximately 2.5-fold). Dusheiko and Roberts reported that liver transplantation was considered a standard treatment for sequelae of chronic hepatitis B in the UK [9], while liver transplantation is rare in Taiwan. Taking all these differences into consideration explains why IFN-α treatment might not achieve savings in direct medical costs in Taiwan. The ICER is widely accepted as the metric of interest in cost-effectiveness studies [14]. Its validity is dependent on the validity of the disease progression, costs and utility parameters in the simulation model. For treatment efficacy and natural history of chronic hepatitis B, we selected peer-reviewed reports from Taiwan to estimate virtually all of the clinical parameters shown in Table 1. We used a 24-week IFN-α regimen in the base-case model, in contrast to the American Association for the Study of Liver Diseases guideline of a 16-week IFN-α regimen [4]. If the efficacy of the 16-week regimen is not much worse than the 24-week regimen [6], then the ICER of our model will be even more favorable because of the reduced costs for IFN-α in the shorter regimen. The projected increase in life expectancy of 0.41 years attributed to IFNα treatment is similar to the 0.1 to 0.6-year increase in the UK [9], which is of the same order of magnitude as other clinical and public health interventions in Taiwan. For example, the increase in life expectancy through cessation of alcohol consumption in a 35-year-old man in the inactive HBsAg carrier state was years [43]. Several disease progression parameters that we used were different from those used by Wong et al, which would account for the substantial difference between the increase in life expectancy estimated in this study (0.41 yr) and that estimated by Wong et al (3.4 yr). For example, patients in Taiwan rarely lose HBsAg positivity after HBeAg clearance, and we therefore entered a very small probability for this disease state transition in the model. There have been few published cost estimates for diseases in Taiwan [44, 45]. Although the cost estimates for outpatient visits and hospitalizations in this study were based on reimbursement rates determined before 1996, the NHI rates have not changed since then. Hospitalization costs for HCC used in this study were similar to cost estimates for HCC reported by Wu [44]. However, patients with decompensated cirrhosis or HCC may have a poor prognosis and not remain in the same disease state for an entire year after the transition. Thus, we may have overestimated the costs associated with these liver diseases. It is common practice in Taiwan to hospitalize patients during the first week of IFN-α treatment. Since this hospitalization phase is not included in treatment guidelines, we did not include it as a component of treatment costs. Nevertheless, it represents a small expenditure in comparison with the acquisition costs for 24 weeks of IFNα therapy. We also did not include direct non-medical costs such as patients transportation to clinical visits. Sensitivity analysis of the cost parameters showed that, for all cost estimates except treatment costs for chronic hepatitis B in HBsAg+/HBeAg+ patients, the range of cost values did not significantly affect ICER estimates. The time trade-off approach, a standard method for cost-effectiveness analysis [46], has been used in previous studies from Taiwan to evaluate utility associated with other disease states [47]. In this study, sensitivity J Formos Med Assoc 2002 Vol 101 No 9

8 Cost-effectiveness of IFN-α in Chronic Hepatitis B analyses showed that ICER estimates were relatively robust across the ranges of utility values used. In this study, we used Monte Carlo simulation, another standard method in cost-effectiveness analysis, to evaluate how ICER estimates were affected by simultaneous variation of all parameters within the range of their plausible values. We found that approximately 2.5% of ICER estimates were larger than 683,000 NT$/QALY and 38.8% were larger than the base-case ICER estimates of 492,000 NT$/QALY, providing further support for the robustness of the ICER estimates. Due to the skewed distribution of the model parameters, it is not surprising that the median ICER value of 188,000 NT$/QALY was substantially smaller than the base-case estimate of 492,000 NT$/QALY, suggesting that the base-case estimate was probably an overestimate of the ICER because of the conservative approach that we used in selection of parameter values. Along with the aging of the population, industrialization of the society, and implementation of the NHI system, the Taiwan healthcare system is facing the same challenge as all other developed countries equitably and optimally allocating a limited healthcare budget. Decision analysis, the scientific method that takes into account all available evidence of viable alternatives, has been widely used in the formulation of healthcare policy [48, 49]. In formulating a decision, the absolute value of the ICER estimate by itself is not very informative and needs to be interpreted in the context of resource allocation for the treatment of chronic diseases. For a healthcare system with limited resources, the ICER of different treatment options for the same disease can be used as the basis of selection of the optimal treatment regimen. Moreover, the ICER of treatments for different diseases can provide important information to help prioritize resource allocation. Ideally, a league table of ICERs for all treatments of major diseases in a health system should be developed, and treatment with the lowest ICER should have the highest priority of resource allocation [50]. Although there has been no consensus on a decision rule to evaluate how small an ICER justifies resource allocation [51], certain cut-off values have been proposed in different industrialized countries. Laupacis et al proposed that the evidence is strong if the ICER is less than 20,000 Canadian dollars (CA$), moderate if ICER is between 20,000 and 100,000 CA$, and weak if it exceeds 100,000 CA$ [52]. In the USA, the value of 50,000 US$/QALY is commonly used as the cut-off value [51]. Regardless of whether such a cut-off value is adopted in Taiwan, the ICER results estimated in this study need to be compared with those for treatment of other major chronic diseases in Taiwan, such as treatment of hypertension to prevent stroke and myocardial infarction, treatment of hypercholesterolemia to prevent adverse cardiovascular outcomes, and treatment of diabetes to prevent long-term vascular and renal complications. As the ICERs for the treatment of these chronic diseases in Taiwan have not been reported, this study provides an important first step toward rigorously evaluating the relative merit of chronic disease management regimens in Taiwan. ACKNOWLEDGMENTS: The authors would like to thank Drs. TT Chang, RN Chien, HH Yang, SC Chen, KS Ho, MF Chen, SS Wu, GS Cheng, WL Zhuang, SJ Hwang, CH Chen, and JH Kao for kindly providing valuable clinical hepatology consultation. We would also like to thank Dr. DW Liu and Ms. CT Yu for collecting cost and utility data. References 1. Cancer Registry Annual Report, R.O.C. Taipei: Department of Health, ROC, Chang MH, Chen CJ, Lai MS, et al: Universal hepatitis B vaccination in Taiwan and the incidence of hepatocellular carcinoma in children. Taiwan Childhood Hepatoma Study Group. N Engl J Med 1997;336: Lok AS, Heathcote EJ, Hoofnagle JH: Management of hepatitis B: 2000 summary of a workshop. Gastroenterology 2001;120: Lok AS, McMahon BJ: Chronic hepatitis B. Hepatology 2001;34: Consensus statements on the prevention and management of hepatitis B and hepatitis C in the Asia-Pacific region. Core Working Party for Asia-Pacific Consensus on Hepatitis B and C. J Gastroenterol Hepatol 2000;15: Wong DKH, Cheung AM, O Rourke K, et al: Effect of alpha-interferon treatment in patients with hepatitis B e antigen-positive chronic hepatitis B. A meta-analysis. Ann Intern Med 1993;119: Lin SM, Sheen IS, Chien RN, et al: Long-term beneficial effect of interferon therapy in patients with chronic hepatitis B virus infection. Hepatology 1999;29: Wong JB, Koff RS, Tine F, et al: Cost-effectiveness of interferon-alpha 2b treatment for hepatitis B e antigen-positive chronic hepatitis B. Ann Intern Med 1995;122: Dusheiko GM, Roberts JA: Treatment of chronic type B and C hepatitis with interferon alfa: an economic appraisal. Hepatology 1995;22: Russell LB, Gold MR, Siegel JE, et al: The role of costeffectiveness analysis in health and medicine. JAMA 1996; 276: Siegel JE, Weinstein MC, Russell LB, et al: Recommendations for reporting cost-effectiveness analyses. JAMA 1996; 276: Weinstein MC, Siegel JE, Gold MR, et al: Recommendations of the panel on cost-effectiveness in health and medicine. JAMA 1996;276: J Formos Med Assoc 2002 Vol 101 No 9 639

9 R.F. Pwu and K.A. Chan 13. Sonnenberg FA, Beck JR: Markov models in medical decision making: a practical guide. Med Decis Making 1993;13: Detsky AS, Naglie IG: A clinician s guide to cost-effectiveness analysis. Ann Intern Med 1990;113: Liaw YF: Pathogenesis and treatment of chronic hepatitis B virus infection. In: Hepatitis and Hepatocellular Carcinoma: Novel Approaches. Joint Conference of the National Health Research Institutes (NHRI) and American Association for Cancer Research (AACR) in Conjunction with 4th TCOG Annual Meeting. Taipei: Academia Sinica, Taipei, 2000:L Liaw YF: Viral hepatitis in Taiwan: status in the 1990s. In: Nishioka K, Suzuki H, Mishiro S, et al, eds. Viral Hepatitis and Liver Disease. 8 th ed. Tokyo: Springer-Verlag, 1994: Hwang SJ, Lu RH, Wang YJ, et al: Clinical significance of changes in serum hepatitis B virus DNA titer in patients with chronic hepatitis B treated with interferon. Chin Med J 2000;63: Lok AS, Lai CL, Wu PC, et al: Long-term follow-up in a randomised controlled trial of recombinant alpha 2-interferon in Chinese patients with chronic hepatitis B infection. Lancet 1988;ii: Lok AS, Wu PC, Lai CL, et al: A controlled trial of interferon with or without prednisone priming for chronic hepatitis B. Gastroenterology 1992;102: Liaw YF, Chu CM, Huang MJ, et al: Determinants for hepatitis B e antigen clearance in chronic type B hepatitis. Liver 1984;4: Tsai SL, Yang PM, Lai MY, et al: Natural history of hepatitis B surface antigen-positive cirrhosis in Taiwan: a clinicopathological study. J Gastroenterol Hepatol 1988;3: Su IJ, Lai MY, Hsu HC, et al: Diverse virological, histopathological and prognostic implications of seroconversion from hepatitis B e antigen to anti-hbe in chronic hepatitis B virus infection. J Hepatol 1986;3: Evans AA, Fine M, London WT: Spontaneous seroconversion in hepatitis B e antigen-positive chronic hepatitis B: implications for interferon therapy. J Infect Dis 1997;176: Liaw YF: Current therapeutic trends in therapy for chronic viral hepatitis. J Gastroenterol Hepatol 1997;12:S Kao JH, Wu NH, Chen PJ, et al: Hepatitis B genotypes and the response to interferon therapy. J Hepatol 2000;33: Liaw YF, Lin SM, Chen TJ, et al: Beneficial effect of prednisolone withdrawal followed by human lymphoblastoid interferon on the treatment of chronic type B hepatitis in Asians: a randomized controlled trial. J Hepatol 1994;20: Crowley SJ, Tognarini D, Desmond PV, et al: Cost-effectiveness analysis of lamivudine for the treatment of chronic hepatitis B. Pharmacoeconomics 2000;17: Wu TT, Hsu HC, Chen DS, et al: Clearance of hepatitis B surface antigen (HBsAg) after surgical resection of hepatocellular carcinoma. J Hepatol 1987;4: Liaw YF, Sheen IS, Chen TJ, et al: Incidence, determinants and significance of delayed clearance of serum HBsAg in chronic hepatitis B virus infection: a prospective study. Hepatology 1991;13: Liaw YF, Tai DI, Chu CM, et al: The development of 640 cirrhosis in patients with chronic type B hepatitis: a prospective study. Hepatology 1988;8: Beasley RP: Hepatitis B virus. The major etiology of hepatocellular carcinoma. Cancer 1988;61: Yu MW, Chen CJ: Hepatitis B and C viruses in the development of hepatocellular carcinoma. Crit Rev Oncol Hematol 1994;17: Liaw YF, Lin DY, Chen TJ, et al: Natural course after the development of cirrhosis in patients with chronic type B hepatitis: a prospective study. Liver 1989;9: Liaw YF, Tai DI, Chu CM, et al: Early detection of hepatocellular carcinoma in patients with chronic type B hepatitis. A prospective study. Gastroenterology 1986;90: Chen CH, Chen DS: Hepatocellular carcinoma: 30 years experience in Taiwan. J Formos Med Assoc 1992;91(Suppl 3):S [In Chinese] 36. Sung JL: Prevention of hepatitis B and C virus infection for prevention of cirrhosis and hepatocellular carcinoma. J Gastroenterol Hepatol 1997;12(Suppl):S Chen DS: Hepatitis B and C virus infections in hepatocellular carcinoma and their prevention. In: Nishioka K, Suzuki H, Mishiro S, et al, eds. Viral Hepatitis and Liver Disease. 8 th ed. Tokyo: Springer-Verlag, 1994: Tsai JF, Jeng JE, Ho MS, et al: Effect of hepatitis C and B virus infection on risk of hepatocellular carcinoma: a prospective study. Br J Cancer 1997;76: Lee CL, Ko YC, Choong CS: Survival rate for liver cancer in Taiwan. Chin Med J 2000;63: Ministry of the Interior. Abridged life table in Taiwan- Fuchien Area the Republic of China, Taipei: Ministry of the Interior, ROC 2002: Pharmaceutical Benefit Scheme for National Health Insurance. Taipei: Bureau of National Health Insurance, ROC, Briggs AH: Handling uncertainty in cost-effectiveness models. Pharmacoeconomics 2000;17: Chiou ST: Potential impacts in liver diseases mortality and life expectancy through various reduction in prevalence of HBsAg carrier status and habitual alcohol consumption. [Masters thesis] Taipei: National Taiwan University, [In Chinese; English abstract] 44. Wu CL: Morbidity costs and associated factors of patients with hepatocellular carcinoma from a medical center. [Masters thesis] Taipei: National Taiwan University, [In Chinese; English abstract] 45. Chou LF, Chang CW, Fuh JL, et al: The economic costs of dementia in Taiwan. National Chengchi University Journal 2001;82: Drummond MF, O Brien B, Stoddart GL, et al: Methods for the Economic Evaluation of Health Care Programmes, 2nd ed. Oxford: Oxford University Press, 1997: Lin RD, Yao KPG, Yu CT, et al: Reliability and validity of utility approach to measuring health related quality of life: an example of patients on hemodialysis. Chin J Public Health (Taipei) 1997;16: [In Chinese; English abstract] 48. Hjelmgren J, Berggren F, Andersson F: Health economic guidelines similarities, differences and some implications. Value Health 2001;4: Weinstein MC, Toy EL, Sandberg EA, et al: Modeling for health care and other policy decisions: uses, roles, and validity. Value Health 2001;4: J Formos Med Assoc 2002 Vol 101 No 9

10 Cost-effectiveness of IFN-α in Chronic Hepatitis B 50. Drummond MF, O Brien B, Stoddart GL, et al: Methods for the Economic Evaluation of Health Care Programmes, 2nd ed. Oxford: Oxford University Press, 1997: Hirth RA, Chernew ME, Miller E, et al: Willingness to pay for a quality-adjusted life year: in search of a standard. Med Decis Making 2000;20: Laupacis A, Feeny D, Detsky AS, Tugwell PX. How attractive does a new technology have to be to warrant adoption and utilization? Tentative guidelines for using clinical and economic evaluations. CMAJ 1992;146: J Formos Med Assoc 2002 Vol 101 No 9 641