Pretreatment Evaluation of Chronic Hepatitis C: Title and subTitle BreakRisks, Benefits, and Costs FREE

CHRONIC HEPATITIS C virus (HCV) infection affects nearly 4 million people in the United States, accounting for 40% of all chronic liver disease, 20% to 30% of all liver transplantations, and more than 8000 deaths annually.^{1 - 2} Until recently, interferon alfa-2b had been approved for a single 6-month course of treatment for chronic hepatitis C (CHC). Nonetheless, physicians, patients, and third-party payers have been reluctant to initiate treatment because of the low likelihood of long-term response following treatment, the slowly progressive nature of CHC, and the cost of interferon. However, we and others recently demonstrated that the marginal cost-effectiveness of treating patients with biopsy-proven mild CHC for 6 to 12 months ($1200 to $5000 per year of life gained) compares favorably with many other well-accepted conventional therapies.^{3 - 4} However, even further cost efficiencies might accrue by either reducing pretreatment evaluation costs or using pretreatment evaluations to identify and target therapy to patients having higher probabilities of long-term response. To quantify the benefit of testing, we compared alternative pretreatment management strategies based on their predicted lifelong clinical and economic outcomes.

We assumed a population that has evidence of chronic hepatitis distributed among 3 histological stages with serum alanine aminotransferase (ALT) levels persistently above 1.5 times the upper limit of normal for at least 6 months, HCV antibody reactive on second-generation enzyme-linked immunosorbent assay testing, and viremia confirmed by qualitative polymerase chain reaction. We used a population mean age of 40 years⁵ and based data estimates on published studies. We performed a MEDLINE literature search, using the terms hepatitis C, biopsy, and liver, from 1966 to 1996.

Using the decision analysis software program Decision Maker, Version 7.06 (Pratt Medical Group, Boston, Mass), we developed a Markov simulation to estimate prognosis by following identical cohorts over time for each strategy.^{3 ,6 - 7} Over time periods (eg, 1 year), patients move among predefined states of health based on transition probabilities derived from a review of the literature or expert opinion. Prognosis depended on the histological stage at the time of treatment, the results of the pretreatment workup, whether treatment was given, and the response to treatment. Our model assumed physicians would be aware of the initial histological stage only if a biopsy was included in the initial evaluation.

We considered alternative management strategies, including conservative management (avoiding tests and interferon), empirical treatment with interferon, and interferon treatment based on various combinations of quantitative HCV RNA level, viral genotype, and histology (presence or absence of cirrhosis). Depending on the pretreatment strategy being analyzed, none to all of these predictors may have been used to determine which patients would receive treatment. Although we evaluated more than 100 strategies, we restrict our reported results to 11 clinically relevant strategies (Table 1). Empirical therapy without liver biopsy or further pretreatment evaluation was defined as a 24-week course of interferon alfa-2b at 3 million U 3 times weekly. We considered pretreatment testing for viral load (B1 and B2), genotype (C1 and C2), or histology (D1, D2, E1, E2, and E3) to avoid treating patients less likely to respond to interferon because of high viral loads, viral genotype 1b, or cirrhosis.^{8 - 13}

Regardless of the strategy under consideration, patients receiving treatment would undergo 12 weeks of interferon therapy. At that time, patients without a complete biochemical response (ie, normalization of ALT levels) would have the interferon therapy discontinued, and the other patients would complete 24 weeks of treatment.

We considered 3 histological states: mild hepatitis, moderate hepatitis, and a cirrhotic stage.¹⁴ From each histological state, over time patients could progress to a more advanced clinical or histological state or remain in the same state of health. For example, patients with mild hepatitis could develop moderate hepatitis or remain stable with mild hepatitis. The computer simulation continues until all patients have died. By tracking the proportion of patients alive in any given year and their costs of care for any given year, the computer simulation estimates the cohort's life expectancy and lifetime expenditures for each strategy. Although this Markov model was based on our previous publication,³ we made the following additions: (1) pretreatment testing; (2) depending on the strategy, patients with cirrhosis or moderate hepatitis, in addition to mild hepatitis, may receive treatment with interferon; (3) a distribution of genotypes, viral loads, and histology; (4) differential likelihood of a sustained response depending on genotype, viral load, and histology, based on a published predictive instrument using these covariates¹⁵ ; and (5) determination of quality-of-life estimates using the standard gamble. Our expert panel of senior hepatologists familiar with liver disease and interferon therapy assessed their own utilities for being in each health state using the standard reference gamble (balancing a risk for dying in the near term against the benefit of being in a state of health with a better quality of life) and the time trade-off technique (balancing longer survival at a lower quality of life against shorter survival in a state of health with a better quality of life).

Population Distributions.—Table 2 presents the mean and 95% confidence interval for the distribution of pretreatment predictors of response obtained by pooling published studies^{9 - 10 ,16 - 48} with a fixed-effects model.⁴⁹ The genotype classification of Simmonds et al⁵⁰ was collapsed into 1b, 3a, or "other" (ie, 1a, 2a, 2b, or 3b) based on a predictive instrument.⁹

Natural History.— The natural history of CHC was obtained by pooling 3 retrospective observational studies of non-A, non-B chronic hepatitis followed by serial liver biopsies in which serum samples subsequently tested positive for HCV antibody (Table 2)^{45 - 47} and the only published study of the natural history of patients with compensated cirrhosis and CHC.⁴⁸ In patients with cirrhosis, the prognosis following decompensation depended on the mode of presentation: diuretic-sensitive ascites, diuretic-refractory ascites, variceal hemorrhage, and hepatic encephalopathy.^{51 - 53} The model assumed that patients who had progressed to a decompensated state would not return to a compensated state unless they underwent liver transplantation.^{54 - 59} Their subsequent prognosis matched those of patients surviving liver transplantation. We also included a risk for hepatocellular carcinoma. Further details can be found in our previous publication.³ To validate our model, we found that model predictions³ matched observed event rates well.^{48 ,60}

Response to Interferon.— Interferon response may be categorized into 3 groups: no response (elevated ALT level 12 weeks into treatment), end-of-treatment response (formerly called a complete response, with normal ALT level at the end of treatment), or sustained response (normal ALT level and undetectable serum HCV RNA 6 or 12 months after completion of therapy). Multivariate analyses have consistently identified histology, level of viremia, and genotype as predictors of a sustained response.^{10 ,15 ,19 - 20 ,61 - 64} The probability of a sustained response based on these pretreatment predictors (histology, genotype, and level of viremia) was calculated by logistic regression. The odds ratio (OR) for mild or moderate hepatitis biopsy findings was 4.94 (P = .16); the OR for HCV RNA viral load of no more than 3.5 × 10⁵ was 17.8; the OR for HCV RNA viral load of more than 3.5 × 10⁵ and no more than 32 × 10⁵ was 3.24 (P<.001); and the OR for genotype 3a was 29.7 and for genotype not 1b or 3a was 8.20 (P<.001).⁹

Liver Biopsy.— A percutaneous liver biopsy prior to treatment has been recommended for patients with CHC to document the severity of disease, the degree of inflammation, and the extent of fibrosis.^{2 ,65} Although the biopsy is a relatively safe procedure when performed by experienced clinicians, it is not risk free; older studies suggest a risk of hospitalization of 4.7% (95% confidence interval, 0%-17.1%).^{37 ,42} Those studies probably overestimated the likelihood of hospitalization, so we used a value of 0.3% as the risk of complications to bias our analysis toward liver biopsy. We pooled 204,714 liver biopsies from 11 studies and found a mortality risk of 0.018% or 18 deaths per 100,000 liver biopsies.^{34 - 44}

Costs.— Because charges are almost always greater than costs and can be arbitrary, we used 1995 costs or adjusted charges. Taking a societal perspective (but excluding indirect or time costs), we used previously published variable cost estimates (the amount spent to care for 1 additional patient with the illness) for actual CHC patients, including inpatient, outpatient, and drug costs for each state of health, with decompensated cirrhosis and liver transplantation being the most expensive.³ Because dollars spent now are more valuable than those spent in the future, we discounted the future expenditures by 5% to permit comparison with previous economic analyses but also repeated our analysis with the currently recommended 3% discount rate.⁶⁶

Construction of the model required the following assumptions about the disease distribution, natural history of the disease, and response to treatment: (1) The prognosis of patients who do not respond biochemically or who relapse is identical to that for patients with conservatively managed CHC, ie, the absence of a response or only a short-term response to interferon conveys no long-term benefit. Some studies, however, suggest histological improvement in such patients.^{67 - 68} If this improvement actually leads to improved prognosis then this assumption biases against interferon treatment. (2) In our previous model, we assumed that any HCV-negative patients would not develop progressive liver disease. Based on the opinion of our expert panel, we assumed such patients could have liver disease progression at a lower but nonzero likelihood. This assumption also biases this analysis against the use of interferon. (3) We did not examine the impact of serial liver biopsies because we assumed treatment would not be affected. Thus, although our model follows up patients through a sequence of histological states, the later states remain undetected clinically until patients develop decompensated liver disease. Therefore, the costs of follow-up are assumed to be determined by the patient's initial pretreatment health state until decompensated liver disease develops.

Sustained Response. Based on the data estimates in Table 2, the sustained response to interferon would range from 7.9% for empirical treatment of all patients to 23% for highly selective pretreatment strategies (Table 3). Column 4 in Table 3 presents the number of patients needed to treat with interferon to achieve 1 sustained response. More selective strategies have a lower number needed to treat because with these strategies, patients unlikely to respond are not treated. But these selective strategies also withhold treatment from 36% to 85% of the patients who could have had a sustained response if treated (column 5). The number needed to treat is a measure of a strategy's specificity (ie, it improves or declines as nonresponders are excluded) but is silent on a strategy's sensitivity (ie, the strategy's ability to achieve a sustained response in all patients who would have responded if treated).

Cost-effectiveness Analysis.— Our model estimates that treatment of all patients with CHC would reduce their lifetime incidence of cirrhosis from 66% to 60%, of decompensated cirrhosis from 46% to 43%, of hepatocellular carcinoma from 18% to 17%, and of liver transplantation from 6.3% to 5.9%. Our model estimates that 40-year-old patients with mild CHC, moderate CHC, or compensated cirrhosis should have life expectancies without (with) treatment of 33.5 (33.9), 26.4 (27.5), and 18.8 (19.0) years, respectively, compared with a normal life expectancy of 37.6 years. Among sustained responders, interferon would increase average life expectancy by 4 to 11 years.

Despite the differences among pretreatment strategies in selecting sustained responders, nearly all strategies that involve testing should increase life expectancy by 9 months compared with conservative management. Among the various testing and empirical treatment strategies, lifetime undiscounted costs differed by at most $1400 and life expectancy differed by at most 3 months, except for the strategies of (1) biopsy and treatment for only mild hepatitis and (2) genotype testing and treatment for only genotype 3a, for which the differences in cost and life expectancy increased to $1900 and 8 months, respectively. Conservative management had the lowest discounted lifetime cost (Table 4). Quantitative HCV RNA testing and directing treatment toward those patients with a level of viremia of no more than 3.5 × 10⁵ genomes per milliliter increased discounted lifetime costs by $45 and increased life expectancy by 0.17 discounted quality-adjusted life-years (QALYs) compared with conservative management, making its marginal cost-effectiveness ratio about $300 per discounted QALY gained. Similarly, quantitative HCV RNA and treatment of those patients with viremia of no more than 32 × 10⁵ genomes per milliliter had a marginal cost-effectiveness ratio of $4400 per discounted QALY gained compared with restricting treatment to patients with viremia of no more than 3.5 × 10⁵ genomes per milliliter. Empirical treatment with interferon (the least-specific strategy) yielded the highest life expectancy and quality-adjusted life expectancy at a marginal cost-effectiveness ratio of $12,400 per discounted QALY gained compared with RNA testing, with a cutoff of 32 × 10⁵ genomes per milliliter. All other strategies were dominated, either costing more and yielding lower life expectancies or being a less efficient use of resources. For other age groups, the marginal cost-effectiveness ratio for empirical treatment for ages 20 to 70 years remains below $50,000 per discounted QALY gained.

We varied the likelihoods of different histologies, viral loads, and genotypes for a 40-year-old within the ranges presented in Table 2. Even altering assumptions about the likelihood of histological progression, the likelihood of sustained response, the distribution of genotypes,^{9 - 10 ,16 ,18 ,21 ,32 ,69 - 70} and the quality-of-life values (by raising the long-term adjustment values, they may be more closely representative of those that might be obtained from patients as opposed to physicians),⁷¹ the marginal cost-effectiveness ratio of interferon remained less than $50,000 per discounted QALY gained. When we examined 12 months of therapy,^{72 - 73} all 12-month strategies had lower costs and higher life expectancies than their 6-month counterparts, lowering the marginal cost-effectiveness ratio for empirical treatment to $4300 per discounted QALY gained.

Interferon treatment is associated with a sustained virus-negative state for some patients with CHC. Because of interferon's expense, inconvenience, and the relatively low probability of a long-term response, many attempts have been made to predict which patients are most likely to respond to treatment. Although several consistent predictors of a sustained response have been identified, there has been no rigorous evaluation of the consequences of using these predictors to guide the management of these patients. Using the expected performance of currently available pretreatment tests and guidelines of current practice, we compared the cost-effectiveness of various pretreatment strategies and empirical therapy. Our results indicate that selected pretreatment strategies can enrich the likelihood of a sustained response among treated patients by avoiding treatment in patients unlikely to have a sustained response. However, because our predictions are imperfect, all of these selective strategies will also miss some patients who would have responded if given interferon. If therapy with interferon is withheld, potential responders might forego an increase in life expectancy of 4 to 11 years, depending on their histology. Only the strategy of treating all patients with chronic HCV empirically with interferon will avoid missing potential sustained responders. In essence, the 12-week initial trial of interferon alfa-2b is a test, with therapy being discontinued in nonresponders (ie, after a negative test result). Our results suggest that empirical interferon is a reasonable management strategy, maximizing life expectancy at a reasonable cost-effectiveness ratio, when compared with other well-accepted medical practices, such as the treatment of hypertension ($23,400) or coronary artery revascularization for 3-vessel disease ($43,300).^{74 - 75} The potential 4- to 11-year increase in life expectancy from treatment compares favorably with the 1.2-year increase associated with revascularization for 3-vessel disease and the 1.3- to 1.5-year increase associated with thrombolytic therapy for suspected acute myocardial infarction.⁷⁶

The quality-of-life estimates in this study differ from those in our previous hepatitis B⁷⁷ and hepatitis C studies,³ in part because of differences in disease, treatment dosage, and panel members. When compared with our previous hepatitis C study, however, these quality-of-life estimates are higher because of differences in the methods used to obtain these estimates. The previous hepatitis C study relied on the categorical scale and time trade-off, whereas the current study relied on standard gamble and time trade-off assessments. The higher values in the current study are consistent with previous work⁷⁸ and the effects of risk aversion. These new values further bias the analysis against empirical interferon.

Pretreatment liver biopsy is commonly recommended but adds expense and is associated with low but real risks of morbidity and mortality. It remains a significant barrier for patients seeking treatment and for interferon treatment by primary care physicians. But is liver biopsy essential? Biopsy may diagnose other coexistent diseases,⁷⁹ but most of these diagnoses are detected by a serological survey. Pretreatment and posttreatment biopsy may demonstrate definitive histological improvement beyond viral eradication, but histological improvement almost always occurs with viral eradication.⁸⁰ In addition, even among patients without a long-term viral response, some studies have found histological improvement⁶⁸ and a decreased risk of hepatocellular carcinoma in cirrhosis.⁶⁷

Finally, a liver biopsy provides prognostic information for the clinician and possibly the patient. For physicians, it reduces the uncertainty in the patient's underlying histology and helps clarify the frequency of follow-up visits and tests. For patients, it could also decrease uncertainty but also might raise anxiety because of fear surrounding its risks and the possible findings of cirrhosis or active inflammation. For liver biopsy to be preferred, patients must value the prognostic information conferred by the biopsy because for such patients, empirical treatment would be less costly and would provide better expected outcomes than would liver biopsy. Our quest for diagnostic and prognostic certainty can sometimes lead to unduly restrictive treatment of patients requiring a proven diagnosis, but "this approach may expose patients to . . . invasive, even life-threatening tests."⁸¹ A decade ago, decision analysis found little difference in value between empirical treatment with steroids and performing a kidney biopsy in patients with nephrotic syndrome.⁸² Similarly, this study suggests that a 12-week empirical therapeutic trial with interferon is a preferable alternative to liver biopsy in patients with CHC.