|Home | About | Journals | Submit | Contact Us | Français|
(See the editorial commentary by Lo Re III, on pages 1307–1309.)
Background.Hepatitis B virus (HBV) continues to cause significant morbidity and mortality in the United States. Current guidelines suggest screening populations with a prevalence of ≥2%. Our objective was to determine whether this screening threshold is cost-effective and whether screening lower-prevalence populations might also be cost-effective.
Methods.We developed a Markov state transition model to examine screening of asymptomatic outpatients in the United States. The base case was a 35-year-old man living in a region with an HBV infection prevalence of 2%. Interventions (versus no screening) included screening for Hepatitis B surface antigen followed by treatment of appropriate patients with (1) pegylated interferon-α2a for 48 weeks, (2) a low-cost nucleoside or nucleotide agent with a high rate of developing viral resistance for 48 weeks, (3) prolonged treatment with low-cost, high-resistance nucleoside or nucleotide, or (4) prolonged treatment with a high-cost nucleoside or nucleotide with a low rate of developing viral resistance. Effectiveness was measured in quality-adjusted life years (QALYs) and costs in 2008 US dollars.
Results.Screening followed by treatment with a low-cost, high-resistance nucleoside or nucleotide was cost-effective ($29,230 per QALY). Sensitivity analyses revealed that screening costs <$50,000 per QALY in extremely low-risk populations unless the prevalence of chronic HBV infection is <.3%.
Conclusions.The 2% threshold for prevalence of chronic HBV infection in current Centers for Disease Control and Prevention/US Public Health Service screening guidelines is cost-effective. Furthermore, screening of adults in the United States in lower-prevalence populations (eg, as low as .3%) also is likely to be cost-effective, suggesting that current health policy should be reconsidered.
Hepatitis B virus (HBV) infection continues to be a major cause of morbidity and mortality in the United States despite strategies developed to eliminate its transmission. The estimated number of new infections in 2007 was 43,000 . The recent Institute of Medicine (IOM) report on hepatitis and liver cancer notes that up to 2 million Americans are chronically infected, although upward of 75% may not know their status and thus present with late disease . These individuals also serve as reservoirs for disease propagation. The IOM committee concluded that hepatitis B is an important public health problem and that lack of awareness among health care providers, at-risk populations, and the public presents significant barriers to its control.
Guidelines such as those of the US Preventive Services Task Force do not recommend screening for HBV infection in the general adult population . In 2008 the Centers for Disease Control and Prevention (CDC) modified its recommendations to include “individuals born in Asia, Africa, and other geographic regions with 2% or higher prevalence of chronic HBV infections” . Previous CDC recommendations called for testing of people born in areas with 8% prevalence or higher.
Chronic HBV infection leads to cirrhosis, liver failure, or hepatocellular carcinoma (HCC) in 15%–40% of patients and to liver transplantation in roughly 25% of patients per year with decompensated cirrhosis . The goal of treatment is the prevention or reversal of decompensated cirrhosis and reduction in the risk of HCC [6–8].
The full economic impact of chronic HBV infection remains unknown. While previously reported analyses have focused on prevention (primarily through vaccination) , the cost-effectiveness of treatment strategies [5, 10–15], or screening and vaccinating high-risk populations , none have evaluated the larger question of screening and subsequent treatment in the general adult population. Therefore, we developed a decision analytic model that can be used to make assessments of the effectiveness and cost-effectiveness of screening in populations with varying prevalence of HBV infection.
On the basis of data from the National Health and Nutrition Examination Surveys between 1988 and 1994, the prevalence of chronic hepatitis B (CHB; Hepatitis B surface antigen [HBsAg] positivity) in the general population of the United States is 0.42% (95% confidence interval, .32%–.55%) . Among US-born, noninstitutionalized persons the prevalence is lowest, 0.1%, whereas foreign-born residents of the United States have a prevalence of 1.0%–2.6%. Among persons living in group quarters, such as college dormitories, military barracks, nursing homes, and long-term care facilities, the prevalence is roughly 0.5% .
HBeAg seroconversion occurs in 50%–75% of patients within 5–10 years. Cirrhosis develops in a small proportion of patients and is correlated with both viral load and HBeAg status. Approximately 41% of patients with compensated cirrhosis progress to decompensated cirrhosis over a period of 15 years. HCC can develop at any stage of chronic infection with HBV, although it is more likely in patients with cirrhosis (see the Supplementary Appendix for further details).
Management guidelines, including those from the American Association for the Study of Liver Diseases (AASLD), the US HBV Consensus Panel, and most recently the National Institutes of Health [6–8, 19], generally agree on which patient subgroups are treatment candidates—patients with chronic HBV infection, high HBV DNA levels, and active liver inflammation as reflected by elevated levels of alanine aminotransferase (ALT). Consistent with recommendations by the AASLD, we observe HBeAg-positive patients for 6 months and initiate therapy in those without spontaneous seroconversion. Patients who are HBeAg negative begin treatment without a period of observation. Liver biopsy is recommended for these patients. Duration of treatment is different for patients with HBeAg-positive and HBeAg-negative CHB. In the prolonged treatment strategies, treatment of HBeAg-positive patients is discontinued after an additional 6 months of consolidation therapy if they seroconvert, whereas those who are HBeAg negative are treated indefinitely. Among patients who have been screened, we assume that liver disease is not detected until they develop decompensated cirrhosis, in which case salvage therapy with 2 nucleoside or nucleotide agents is recommended. For patients who develop resistance, guidelines recommend the addition of a second agent or a switch to another agent (guided by viral resistance profiling).
We used a commercially available computer program (DECISION MAKER)  to develop a Markov state transition model, analyze decision trees, and perform sensitivity analyses, using a lifelong time horizon and a societal perspective. We considered 5 strategies for a hypothetical 35-year-old man (see Figure 1):
Assumptions regarding efficacy, cost, and resistance were developed from data reported in the literature for lamivudine (low cost, high resistance) and tenofovir (high cost, low resistance). In order to bias results against screening, we assumed that screening would require a separate visit to a health care provider and result in the cost of a visit along with that of the blood test for HBsAg (see the Supplementary Appendix). We assume 100% compliance with treatment, as efficacy from clinical trials is based on intention to treat. However, we explore the impact of this assumption in sensitivity analyses (see Figure 2). Regardless of whether or not screening is performed, patients may have CHB, be immune (ie, have antibody to HBsAg), or be uninfected and unexposed. Asymptomatic patients found to have CHB may have elevated liver function test results (ALT) and may also be HBeAg positive. In the screening strategies, patients with elevated ALT level undergo abdominal ultrasound and HBV DNA load quantification. HBeAg-positive patients are stratified into those with viral loads of ≥20,000 IU/mL or <20,000 IU/mL, whereas HBeAg-negative patients are stratified into those with viral loads of ≥2,000 IU/mL or <2,000 IU/mL. We assume that a fraction (5%) of patients with elevated ALT level and high viral loads (≥20,000 IU/mL in HBeAg-positive patients and ≥2,000 IU/mL in HBeAg-negative patients) undergo ultrasound-guided liver biopsy. Patients enter the Markov model that simulates the natural history of CHB progression in starting health states determined by (1) initial ALT level, (2) HBeAg status, and (3) viral load. HBeAg-positive patients with elevated ALT level and high viral loads are observed for 6 months. If spontaneous HBeAg conversion does not take place, treatment is started. HBeAg-negative patients with elevated ALT level and high viral loads are started on treatment without a period of observation. Treatments are characterized by their impact on viral load, seroconversion in HBeAg-positive patients, and the development of resistance. Baseline values for parameters used in the decision analytic model are summarized in Table 1.
Costs are expressed in 2008 US dollars. Details of the microcosting models are described in Table 2 and the Supplementary Appendix. Future costs and effectiveness were discounted at 3% per year . Costs were subjected to sensitivity analyses.
Numerous studies have examined the impact of CHB on health-related quality of life [12, 46, 52–55]. We used standard gamble utility assessments elicited by Levy et al  from uninfected respondents living in the United States to provide quality of life adjustment factors for CHB—compensated cirrhosis, decompensated cirrhosis, HCC, and liver transplantation (first year and subsequent years). We used time trade-off assessments from a panel of hepatologists described in a study by Bennett et al  to adjust quality of life while receiving interferon. We assumed that oral antiviral agents did not diminish quality of life in an appreciable manner.
Strategies are rank ordered by increasing cost, and marginal cost-effectiveness ratios (mCERs) are calculated between each progressively more expensive but more effective strategy. A strategy is dominated if it costs more but yields lower effectiveness than the prior “cheapest” strategy. Extended dominance occurs when the mCER of one strategy is larger than that of the next more costly strategy that has a nonnegative mCER. When a strategy is dominated by standard or extended dominance it is eliminated from the analysis, and the mCER is not calculated.
We performed both deterministic and probabilistic sensitivity analyses (PSAs) to examine the impact of uncertainty in parameter estimates and population-level variation in parameters. We conducted PSA using second-order Monte Carlo simulation . Distributions for parameter values were developed (see Table 1) using beta and logit distributions for probabilities and lognormal distributions for relative risks, hazard ratios, and rates. Deterministic sensitivity analyses were performed by systematically varying 1 or more parameter values over clinically relevant ranges.
We compared predicted survival in our natural history model of patients with CHB diagnosed at the time of screening with observations from a large Mediterranean cohort study of blood donors among whom 2,352 were found to be HBsAg positive . Overall survival among men who entered the study at a mean age of 33.1 years was 88.8% at 20.5 years after study entry. For a similar cohort of 33.1-year-old men with asymptomatic CHB, our model predicted an overall survival of 88.2%.
In the base case (US population of males with a mean age of 35 years and a 2.0% prevalence of CHB) (Table 3), not screening is both the least effective and the least costly strategy. Screening followed by prolonged treatment of HBsAg-positive patients with a low-cost, high-resistance nucleoside or nucleotide and salvage therapy with the high-cost, low-resistance agent should resistance develop is slightly more costly and more effective than no screening and has an mCER of $29,230 per additional quality-adjusted life year (QALY). Screening followed by a 48-week course of treatment with a low-cost, high-resistance nucleoside or nucleotide has a larger mCER and is eliminated due to extended dominance. Screening followed by prolonged treatment of HBsAg-positive patients with a high-cost, low-resistance nucleoside or nucleotide is the most expensive and is slightly less effective than screening followed by indefinite treatment with the low-cost, high-resistance nucleoside or nucleotide.
Over 10,000 iterations, screening followed by prolonged treatment with a low-cost, high-resistance nucleoside or nucleotide was preferred 80% of the time, whereas screening followed by treatment with a high-cost, low-resistance nucleoside or nucleotide was preferred 20% of the time. As shown in the cost-effectiveness acceptability curve (Figure 3), in a comparison of screening followed by prolonged treatment with a low-cost, high-resistance nucleoside or nucleotide with no screening, screening had an mCER <$50,000 per QALY >49% of the time and <$100,000 per QALY >99% of the time. Table 1 shows confidence intervals and types of distributions used in the PSA.
The overall prevalence in the United States is reported to be 0.3%–0.5%, whereas the prevalence among foreign-born immigrants may be as high as 2.6%. As shown in Figure 4, the marginal cost-effectiveness of screening followed by treatment with a low-cost, high-resistance nucleoside or nucleotide decreases as the prevalence increases. Below a prevalence of 0.3%, the mCER is >$50,000 per QALY.
Although data are available to describe the impact of low-cost, high-resistance nucleoside or nucleotide agents on the relative hazard of developing both cirrhosis and HCC, similar data do not exist for newer high-cost, low-resistance agents. In our base case, we assumed that the high-cost, low-resistance agents were no better in this regard. As shown in Figure 5, the mCER of screening followed by treatment with the high-cost, low-resistance nucleoside or nucleotide decreases as the relative hazard of either compensated cirrhosis or HCC decreases.
Few studies have explored the efficacy of salvage therapy in patients who have developed resistance to initial treatment with a high-resistance nucleoside or nucleotide. In our base case, we assumed that the high-cost, low-resistance agent has the same efficacy in suppressing viral load in the salvage therapy setting as in nucleoside- or nucleotide-naive patients. Sensitivity analyses across a clinically plausible range demonstrate only a slight increase in the mCER of screening as the efficacy of salvage therapy decreases.
Figure 2 summarizes the results of multiple 1-way sensitivity analyses using a tornado plot to examine the cost-effectiveness of screening followed by prolonged treatment with the low-cost, high-resistance nucleoside or nucleotide compared with that of no screening. Although the cost-effectiveness of screening followed by the low-cost, high resistance agent was sensitive to a number of parameters, only changes in 2 parameters (annual rate of spontaneous seroconversion and quality of life while receiving nucleoside or nucleotide therapy) increased the cost-effectiveness of this strategy to >$50,000 per QALY.
Current guidelines, such as those of the US Preventive Services Task Force , do not recommend universal screening for HBV infection in the general population and utilize relatively high rates of prevalence (2%) in targeted populations. Our analysis suggests that screening becomes cost-effective at a population prevalence of >0.3%. This threshold is at the lower end of the confidence limit for general population estimates of chronic HBV infection in the United States, 0.3%–0.5%, although it is slightly above the prevalence for the lowest-risk segment of the population, US-born, noninstitutionalized persons .
It is important to note that this analysis did not address the cost-effectiveness of universal screening versus current guidelines promulgated by the CDC supporting selective screening of higher-prevalence populations. This is a more complex question that requires accurate estimates of prevalence in both the group to be screened and the group not undergoing screening. However, we performed a subanalysis examining the cost-effectiveness of liberalizing the current CDC guidelines suggesting that foreign-born residents of the United States undergo screening if they have emigrated from a region with a prevalence of ≥2% to a slightly lower prevalence threshold. A recent estimate, based on 2008 data, indicates that prevalence varies from 1.3% to 11.8% among foreign-born residents of the United States depending on country of origin . While the mean prevalence among those who emigrated from Asia, Africa, the Caribbean, and Eastern Europe is 6.7%, that of those who emigrated from Central America or other regions not explicitly noted above is 1.3%. In addition, this lower-risk group makes up >55% of the foreign-born residents in the United States. Our analysis showed that the mCER of screening foreign-born residents with a prevalence of ≥2% (ie, the current CDC guideline) is favorable at $31,600 per QALY compared with no screening. The marginal cost-effectiveness of liberalizing the current CDC guideline to also include foreign-born residents with a prevalence of 1.3% is $33,500 per QALY compared with the current CDC guideline.
It also is important to note that results of a cost-effectiveness analysis such as ours provide insights into policy-level decision making for large groups of patients. The best practice for individual patients must also account for patient-to-patient variability in preferences for health outcomes and treatment side effects, as well as more complex and subtle differences such as racial differences in the risk of hepatocellular carcinoma or variability in HBV genotype, which might impact response to interferon.
The superiority of screening was a robust result, insensitive to variations in most parameter values within clinically plausible ranges. One of the few parameters that might make screening cost >$50,000 per QALY was the underlying rate of spontaneous HBeAg seroconversion. However, spontaneous seroconversion would have to exceed 10% per year (base case value, 5% per year) for screening to no longer be cost-effective.
Our analysis examined several treatment options for chronically infected patients for whom treatment was warranted by current guidelines. In the base case, screening followed by prolonged treatment with the high-cost, low-resistance nucleoside or nucleotide was less effective than screening followed by prolonged treatment with the low-cost, high-resistance agent followed by salvage therapy for those who develop resistance. However, changes in the efficacy of this agent, particularly in salvage versus primary treatment settings, could result in this screening strategy becoming cost-effective. Although the main focus of our analysis was the question of screening, due to controversy about using a high-resistance agent as first-line therapy, we also performed a subanalysis that did not include this strategy. In this analysis, the marginal cost-effectiveness of the high-cost, low-resistance nucleoside or nucleotide versus no screening was still reasonable at $43,500 per QALY.
Although we did not model immunization following screening, the major impact of immunization would be to slightly improve the overall life expectancy of noninfected patients who were not already immune (by preventing future infection) and to add up-front cost for these patients. In addition, there are complex, population-level interactions that result from a decreasing prevalence of HBV infection, which our simulation was not designed to model. Ignoring the beneficial impact of immunization, we performed a separate scenario analysis in which we added costs of immunization ($113.22; vaccine cost + level 2 established patient visit) to screening costs for nonimmune patients. This worst-case estimate increased slightly the mCER of screening to $41,800 per QALY, still below a societal willingness-to-pay threshold of $50,000 per QALY. Our analysis also did not address the issue of human immunodeficiency virus (HIV) screening. The complexity of the decision tree in HIV-infected subjects requires additional modeling that is beyond the scope of this analysis.
Prior cost-effectiveness analyses have not focused on screening followed by HBV infection treatment. However, several analyses have examined treatment alternatives for patients with already-diagnosed HBV [14, 60]. Kanwal et al  found that a hybrid strategy consisting of lamivudine followed by adefovir salvage was cost-effective in patients without cirrhosis, whereas entecavir was cost-effective in patients with cirrhosis. Lacey  reports similar results for lamivudine followed by adefovir salvage or adefovir followed by lamivudine salvage in Singapore, whereas Yuan et al  report that entecavir was “highly cost-effective” compared with lamivudine.
How should this analysis impact policy and practice? While the most cost-effective treatment strategy for those found to be infected with HBV may evolve in the future, given newer and more effective agents or consideration of more complex salvage therapies for patients who develop resistance, screening for chronic HBV infection is likely to be cost-effective, even in low-prevalence populations (eg, as low as .3%) in the United States. These findings suggest that current health policy with regard to screening for CHB should be reconsidered.
Supplementary materials are available at Clinical Infectious Diseases online (http://www.oxfordjournals.org/our_journals/cid/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.
Financial support.This work was supported by Gilead Sciences; the National Center for Research Resources (grant 1UL1RR026314 to M. H. E.); the National Institute of Diabetes and Digestive and Kidney Diseases (grant K23 DK075599 to M. H. E. and grant K24 DK070528 to K. E. S.); the National Heart, Lung, and Blood Institute (grant K30 HL078581-01 to M. H. E.); and the National Library of Medicine (grant R01 LM009533 to M. H. E.).
Potential conflicts of interest:K. E. S. has served on an advisory board for BMS, Merck, SciClone, Vertex, GSK, Regulus, Three Rivers, J&J, Valeant, Anadys, Schering, Baxter, and Astellas; received grant support from Roche (Genentech), Schering (Merck), Vertex, Gilead, BMS, SciClone, Anadys, HGS, and Gilead; received royalties from the US Army and UpToDate; and received payment for the development of educational presentations from Chronic Liver Disease Foundation. M. H. E. is a consultant for Savient Pharmaceuticals. T. E. K.: no conflicts.