Cost-effectiveness of suppressing hepatitis B virus DNA in immune tolerant patients to prevent hepatocellular carcinoma and cirrhosis


Dr K. R. Reddy, Gastroenterology Division, 3 Ravdin, University of Pennsylvania, 3400 Spruce Street, Philadelphia, PA 19104, USA.


Background  For patients with hepatitis B virus (HBV) infection in the immune tolerant phase, the current standard of care is to not offer treatment. However, the recent Risk Evaluation of the Viral Load Elevation and Associated Liver Disease/Cancer-In study results show a striking relationship between high HBV DNA levels and risk for hepatocellular carcinoma and cirrhosis.

Aim  In a cost effectiveness analysis, to assess whether immune tolerant patients with high HBV DNA levels should undergo treatment.

Methods  We created a lifetime Markov model to evaluate the cost-effectiveness of two strategies for immune tolerant hepatitis B: (i) HBV DNA suppression with lamivudine, (ii) no treatment. Patients cycled between the following health states: viral suppression, ongoing viremia, seroconversion, hepatocellular carcinoma, cirrhosis and death.

Results  Compared with the no treatment strategy, lamivudine therapy was more expensive but more cost-effective with an additional cost of $5784 and $12 584 per quality adjusted life year gained in males and females, respectively. Treatment resulted in a gain in life expectancy and a decrease in lifetime risk of hepatocellular carcinoma and cirrhosis.

Conclusions  Suppressing HBV DNA to prevent hepatocellular carcinoma and cirrhosis in immune tolerant patients is very cost-effective, and treatment of these patients may be considered. Future prospective clinical trials will need to be undertaken to confirm our findings.


An estimated 350 million persons worldwide are chronically infected with the hepatitis B virus (HBV) with an estimated 1.25 million hepatitis B carriers in the US.1 15–40% of those infected with HBV will develop sequelae of cirrhosis, hepatic decompensation and/or hepatocellular carcinoma (HCC).2, 3 In regions where HBV infection is endemic, perinatal transmission typically results in a prolonged ‘immune tolerant’ phase characterized by extensive HBV replication without evidence of active liver inflammation or fibrosis.4 This phase usually lasts 10–30 years but can continue indefinitely.5, 6 In the US, prevalence of HBV in immigrants from endemic areas is high. For example, 10–13% of Asian-Americans are infected with HBV, many of whom are in the immune tolerant phase.7 Current American Association for the Study of Liver Disease guidelines do not recommend treating patients with immune tolerant hepatitis B due to the lack of significant liver disease and low chance for hepatitis B e antigen (HBeAg) seroconversion on therapy.4

The Risk Evaluation of the Viral Load Elevation and Associated Liver Disease/Cancer-In (REVEAL) HBV Study Group prospectively followed over 3500 Taiwanese individuals with chronic HBV infection from 1991–2004. The authors showed that HBV DNA level was a strong risk predictor of HCC, independent of HBeAg status, serum alanine aminotransferase (ALT) level and cirrhosis.8 Furthermore, HBV DNA level was also a strong predictor of cirrhosis, independent of HBeAg status and ALT level.9 These findings may have important implications for immune tolerant patients with high HBV DNA levels. Whereas treatment of HBV in this group may not result in seroconversion, it is possible that treatment-induced viral suppression may protect against cirrhosis and development of HCC. These possible benefits need to be balanced against the costs and resistance rates of currently available oral therapies for the long-term treatment of HBV. Against this background, we sought to determine whether treatment of immune tolerant HBV patients was cost-effective.


Modelling technique

Using decision analysis software (TreeAge Pro 2006, Healthcare, Williamstown, MA, USA), we constructed a lifetime Markov model with a cycle length of 1 year. During each cycle, specified proportions of hypothetical patients transitioned between various health states. Each health state has an associated cost and utility. The base case was a 40-year-old patient with immune tolerant hepatitis B (HBeAg positive, normal ALT, and viral load of greater than 1 000 000 copies/mL of HBV DNA without cirrhosis or HCC). Males and females were divided into separate cohorts as the REVEAL studies showed different risk for HCC and cirrhosis between the two groups.8, 9

We used the following health states: (i) viral suppression; (ii) ongoing viremia; (iii) seroconversion; (iv) HCC; (v) cirrhosis; (vi) death (Figure 1). Patients either remained in their current health state (recursive arrow) or progressed to a new health state (straight arrow). Patients entering the model were enrolled in one of two competing strategies: (i) lamivudine monotherapy (100 mg once daily); (ii) no treatment. For the treatment group, patients received lamivudine for 1 year. If viral suppression was achieved (dashed arrow), lamivudine was continued. Viral suppression was defined as less than 1000 copies/mL of HBV DNA. Patients who developed resistance to lamivudine in subsequent years were then transitioned to the ongoing viremia state (dotted arrow). If viral suppression was not achieved within the first year of lamivudine treatment or if viremia re-emerged in subsequent years, then lamivudine treatment was discontinued, and these patients were transitioned into the ongoing viremia state. We did not model switching to or adding other HBV therapies. For the no treatment group, patients were unable to enter the viral suppression state. Patients who entered the spontaneous seroconversion state were assumed to be permanently cured of HBV infection. All patients who progressed to the cirrhosis state were started on lamivudine 100 mg once daily because of its significant benefit in this patient population.10 The cirrhosis health state included both compensated and decompensated cirrhotics. Any patient who did not develop spontaneous seroconversion was at some risk for entering the HCC health state. Furthermore, all patients could die from age-related or liver-related mortality.

Figure 1.

 Markov state diagram. The base-case patient has immune tolerant chronic hepatitis B virus (HBV) infection with a viral load of ≥1 000 000 copies/mL. The clinician either treats or does not treat the patient with lamivudine 100 mg once daily. During each cycle (1 year), patients either remain in their current state (recursive arrow) or progress to a new health state (straight arrow). Viral suppression is defined as HBV DNA level < 1000 copies/mL, and only treated patients can enter this health state (dashed arrow). Previously suppressed patients can have rebound viremia with the development of lamivudine resistance (dotted arrow). Transition probabilities were determined from a systematic review of the literature (Table 1).

Analysis was performed using Monte Carlo microsimulation (first-order trials). In this method, individuals from each cohort took a ‘random walk’ through the model. That is, an individual moved from one health state to another in a random fashion with higher probability events being more likely. The outcome of each individual was recorded, and all of the outcomes were averaged. This method accounts for variability among individuals and over time and more closely resembles reality. 50 000 trials were used each time the model was run.

Transition probabilities

Estimates for the natural history and progression of chronic HBV infection, cirrhosis, HCC and lamivudine treatment outcomes were incorporated into our model based on a systematic MEDLINE review (Table 1). The rate of spontaneous seroconversion was the same for the treatment and non-treatment groups, as lamivudine treatment does not increase seroconversion rates in patients with normal ALT.27, 28 We obtained HCC survival data from the Surveillance Epidemiology and End Results database, which includes patients who were both treated and not treated for HCC in the US.32 Because only those patients who had achieved early viral suppression on lamivudine were continued on the drug in the study, a relatively low rate of lamivudine resistance was used.20 The transition probabilities for developing HCC and cirrhosis were based on the REVEAL study data.8, 9 Because male gender was an independent risk factor for both HCC and cirrhosis in the REVEAL study, we weighted these transition probabilities based on the male and female hazard ratios published in the study.8, 9

Table 1.   Transition probabilities
VariableBase case estimate (%) Range (%)
  1. HCC, hepatocellular carcinoma.

Probability of viral suppression on lamivudine after 1 year of treatment11–193613–40
Annual rate of development of resistance on lamivudine while virally suppressed4, 17, 20–235.260–30
Annual rate of spontaneous seroconversion17, 22, 24–2852–10
Annual rate of development of HCC:8, 25, 29–31
 Annual rate while suppressed, men80.1370.069–0.274
 Annual rate while not suppressed, men81.5580.779–3.116
 Annual rate while suppressed, women80.0650.0325–0.130
 Annual rate while not suppressed, women80.7420.371–1.484
Cumulative survival rate of HCC by year after diagnosis:28, 32–34
Annual rate of development of cirrhosis:9, 25, 35
 Annual rate while suppressed, men90.4830.242–0.966
 Annual rate while not suppressed, men93.5681.784–7.136
 Annual rate while suppressed, women90.1930.097–0.386
 Annual rate while not suppressed, women91.4270.714–2.854
 Annual rate of development of HCC while cirrhotic10, 25, 361.441–6.6
 Annual rate of liver related death in cirrhotic patients10, 255.082.5–10

Cost estimates

We conducted our analysis from the perspective of a third-party payer, and we accounted for all direct health care costs associated with various therapies, including physician visits, laboratory tests and radiologic studies. Indirect costs were not accounted for in the model. Cost estimates were derived from published costs in the medical literature (Table 2). Cirrhosis costs included costs for both compensated and decompensated cirrhosis by using a weighted average based on the annual rate of progression from compensated to decompensated cirrhosis.10, 25, 38 All costs were converted into 2006 dollars using the health care component of the Consumer Price Index found at the Bureau of Labor Statistics ( We assumed a future discount rate of 3% per year.

Table 2.   Cost estimates (2006 dollars)
CostsBase case estimate ($) Range ($)
  1. HBV, hepatitis B virus; PCR, polymerase chain reaction.

Total annual cost of lamivudine treatment:25, 3726801300–10 000
 Drug costs:2057 
 Non-drug costs (2 physician visits, 4 sets of labs, 1 abdominal ultrasound):623 
 Physician visit56 
 Set of laboratory tests (including HBV DNA PCR)87 
 Abdominal ultrasound163 
Annual cost of no treatment (1 physician visit, 2 sets of labs, 1 abdominal ultrasound)25, 37393200–800
Annual cost of caring for a seroconverted patient (1 physician visit, 2 sets of labs)25, 37230115–460
Annual cost of cirrhosis25, 38476235–950
Annual cost of hepatocellular carcinoma25, 38, 3942 00621 000–84 000

Utility estimates

The utility estimates used in the model (Table 3) were derived from studies using time-tradeoff and physician survey estimates of utility. For the health states that required lamivudine treatment (viral suppression, cirrhosis), the decrease in utility from taking a daily medication was assumed to be 0.01 and subtracted from the utility associated with that health state. This value was taken from the low end of the range published in a study showing a decrease in utility from interferon treatment.42 We discounted all utilities at a rate of 3% per year.

Table 3.   Utility estimates
UtilitiesBase case estimateRange
Viral suppression410.9170.817–0.99
Decrement in utility while taking lamivudine420.01−0.01–0.02
Ongoing viremia410.9170.817–0.99
Hepatocellular carcinoma400.720.62–0.82
Cirrhosis10, 400.7970.697–0.897

Sensitivity analysis

To study the effect of uncertainty on the robustness of our results, we performed one-way sensitivity analysis of all variables and compiled the results into a tornado diagram. The ranges for each variable used in the sensitivity analysis are listed in Tables 1–3. Whenever possible, the ranges were determined from a review of the literature. For values such as the probability of developing HCC or cirrhosis while virally suppressed where a range was not available in the literature, the range was constructed by doubling and halving the point estimate.


Base-case results

The results of the base-case analysis are displayed in Table 4. The lamivudine treatment strategy was more costly but more effective than the no treatment strategy. Treatment was cost-effective in females with a cost of $12 584 to gain one additional quality adjusted life year (QALY) as well as in males with a cost of $5784/QALY. Moreover, the lamivudine treatment strategy resulted in an increase in median life expectancy of 4 years in the male cohort and 1 year in the female cohort. The lifetime risk of death from HCC and development of cirrhosis was also lower in both treatment groups (Table 5). Thus, our base case analysis revealed that treatment was cost-effective in females and even more cost-effective in males.

Table 4.   Results of Monte Carlo microsimulation
Strategy Cost ($)Incremental cost ($)Effectiveness (QALY)Incremental effectivenessMedian life expectancy (years) C/E ($) ICER ($)
  1. QALY, quality adjusted life years; C/E, cost-effectiveness ratio; ICER, incremental cost-effectiveness ratio.

No treatment, male17 900 16.09 681111 
Treatment, male23 7005 90017.101.017213885784
No treatment, female12 500 18.586 76675 
Treatment, female19 4006 90019.1300.54477101412 584
Table 5.   Lifetime risk of hepatocellular carcinoma and cirrhosis
StrategyLifetime risk of death from hepatocellular carcinoma (%)Lifetime risk of cirrhosis (%)
No treatment, male7.732
Treatment, male7.128
No treatment, female4.018
Treatment, female3.615

Because uncertainties increase with time and very few studies last for more than 10–15 years, time-limited models of 15 years and 35 years were analysed (data not shown). Lamivudine treatment remains cost-effective in both cohorts in the time-limited models, and the results suggest that treatment is more cost-effective with longer-duration models.

Sensitivity analysis

Sensitivity analysis revealed that our model was not highly sensitive to any one variable over wide ranges in both cohorts. For males, the most influential variables were cost of lamivudine, probability of developing cirrhosis while not virally suppressed, initial age at treatment implementation, discount rate and probability of viral suppression on lamivudine (Figure 2). For females, the most influential variables were cost of lamivudine, utility of viral suppression, initial age at treatment implementation, probability of developing cirrhosis while not virally suppressed and discount rate (Figure 3). The cost of lamivudine treatment had the greatest influence on the incremental cost-effectiveness ratios (ICER) in both cohorts. However, even with an annual drug cost of $10 000 per year, treatment remained cost-effective with an ICER of $25 000/QALY in males and $54 000/QALY in females. Moreover, the treatment strategy remained cost-effective even with a low chance for viral suppression or relatively high resistance rates.

Figure 2.

 Tornado diagram, male cohort. This diagram shows that varying the range of the five most influential variables only minimally affects the model’s cost-effectiveness. The cost of lamivudine was the most influential variable, and raising its cost to $10 000 resulted in an incremental cost-effectiveness ratio of $25 000 per quality adjusted life year.

Figure 3.

 Tornado diagram, female cohort. This diagram shows that varying the range of the five most influential variables only minimally affects the model’s cost-effectiveness. As in the male cohort, the cost of lamivudine was the most influential variable, and raising its cost to $10 000 resulted in an incremental cost-effectiveness ratio of $54 000 per QALY.


For patients with hepatitis B in the immune tolerant state, the current standard of care is to not offer HBV treatment.4 However, REVEAL study results show a striking relationship between high HBV viral loads and risk for cirrhosis and HCC. Therefore, controversy currently exists as to whether hepatitis B patients with high viral loads should undergo treatment, even if there is no evidence of significant liver inflammation and fibrosis. We created a comprehensive, lifetime Markov model to determine if it is cost-effective to treat this patient population. Our analysis has several key findings. First, we found that lamivudine monotherapy is extremely cost-effective in patients with immune tolerant hepatitis B with ICERs of $5784/QALY and $12 584/QALY for males and females, respectively. Both values are significantly lower than a widely accepted cost-effectiveness threshold of $50 000/QALY and fall well within the range of many commonly accepted medical interventions.43 Second, treatment is more cost-effective in males than females as shown by the lower ICER, increased gain in life expectancy and greater reduction in risk of development of HCC and cirrhosis. This result is not surprising given that males with hepatitis B are at higher risk for development of both HCC and cirrhosis.8, 9 Finally, we demonstrated significant benefits despite accounting for resistance to treatment. As cumulative exposure to HBV medications does result in a significant rate of resistance, much of the benefits we observed were accrued within the first few years of treatment. Early viral suppression resulted in later prevention of HCC and cirrhosis.

To our knowledge, ours is the only study examining the possibility of treating patients with immune tolerant hepatitis B to prevent HCC and cirrhosis. Our model closely resembles the natural history of immune tolerant hepatitis B infection and reflects the everyday management of these patients. In addition to direct drug costs, we accounted for appropriate physician visits, laboratory and radiologic tests and treatment costs for cirrhosis and HCC. We constructed a model complex enough to reflect reality but simple enough to allow for analysis. Our model is very robust. The ICERs are far below traditional cost-effectiveness thresholds, and sensitivity analysis reveals that wide variations in the variables only minimally affect model results. Thus, the results from the study are unambiguous.

Our analysis does have several limitations. The applicability of the REVEAL study results to immune tolerant patients is uncertain as the majority of patients in the study were HBeAg negative. However, nearly all of the HBeAg negative patients had normal ALT levels and no cirrhosis, and this matches the profile of immune tolerant patients.8, 9 Furthermore, the annual rates of development of HCC and cirrhosis from the REVEAL study are comparable to the rates from studies of immune tolerant populations.44–46 A study of HBeAg positive patients with normal ALT levels found the annual rate of progression to cirrhosis to be approximately 4% for patients over the age of 40.44 Studies involving immune tolerant hepatitis B patients in China and Taiwan reported that the annual rate of progression from asymptomatic infection to HCC ranged from 0.25–0.65%, with the upper limit derived from patients over the age of 40.46 Thus, we believe that the REVEAL study results are applicable to the immune tolerant population.

While the REVEAL study showed that an elevated level of HBV DNA is associated with increased risk of developing HCC and cirrhosis, there is no direct evidence that lowering levels of HBV DNA will result in a decreased risk of HCC and cirrhosis. However, there is indirect evidence suggesting that lowering HBV DNA levels would be beneficial. Decreasing viral replication with lamivudine in HBV patients with cirrhosis has been shown to reduce the incidence of hepatic decompensation and HCC.10, 47 Also, in woodchuck models, suppression of woodchuck hepatitis virus DNA with clevudine or entecavir delays the emergence of HCC.48, 49 Thus, our assumption that lowering HBV DNA levels will prevent HCC and cirrhosis is reasonable but not yet supported by clinical studies. It is also important to note that the goal of treatment in our model is only to suppress serum HBV DNA levels. The presence of intrahepatic cccDNA with undetectable serum HBV DNA does not warrant treatment in our model.

Several of our estimates are derived from studies with varying design, sample size, patient population and quality. We guarded against model inaccuracies by systematically reviewing the literature and using values from the best studies available. We also varied model assumptions over wide ranges during sensitivity analysis. Also, we chose an annual rate of lamivudine resistance of 5.26%, based on data that if viral suppression is achieved early in the course of treatment, subsequent resistance is low.20 We also modelled a relatively low rate of viral suppression (36%) with use of lamivudine, as only rapid viral suppression is associated with such a low resistance rate.11 Our model’s lamivudine resistance value is lower than the widely accepted cumulative resistance rate of approximately 70% at 5 years.4 To account for this discrepancy, during sensitivity analysis we increased the rate to as high as 30% (this corresponds to a five year cumulative rate of greater than 80%), and this did not significantly affect the study results.

We did not account for any future consequences of inducing lamivudine resistance. If a patient with lamivudine resistance later requires treatment, a more expensive drug such as adefovir or entecavir would be required, resulting in higher drug costs. Furthermore, the development of such resistance has been associated with subsequent worsening of liver disease.50, 51 While this is likely due to loss of viral suppression, it is nonetheless possible that the mutant strain could be more pathogenic or carcinogenic, resulting in more future sequelae and costs. However, given that the ICERs from the model are very low, it is unlikely that these potential future costs would substantially change our model’s findings.

Finally, we chose lamivudine as the study drug due to the abundance of published data, particularly long-term data, relative to other oral agents. Because our model does not consider the use of other agents or combination therapy, these results are only applicable to lamivudine monotherapy. However, some inferences can be made from our study. While cost of treatment was the most important variable based on sensitivity analysis, even $10 000/year in treatment costs continued to be highly cost-effective despite high rates of drug resistance. Of note, $10 000/year is approximately the current cost of the most expensive approved medication for HBV, entecavir.52 As newer agents for HBV are being introduced with even lower rates of resistance, we expect that these medications would likely be cost-effective as well.

In conclusion, our analysis reveals that treatment of immune tolerant hepatitis B prevents HCC and cirrhosis and is very cost-effective. Our findings challenge current guidelines for the management of immune tolerant patients. Given our results, treatment of immune tolerant patients may be considered, especially in patients with a family history of HCC or other high risk features. Future prospective clinical trials will need to be undertaken to confirm our findings.


Declaration of personal and funding interests: None.