Health Values of Patients With Chronic Hepatitis C Infection

Background  Patients with hepatitis C virus (HCV) infection report a reduction in health status, but it is not known how they value their state of health. We assessed health utilities directly from patients with HCV infection.

Methods  One hundred twenty-four patients with chronic HCV infection representing a cross section of disease severity were administered a disease-specific version of the Medical Outcomes Study 36-Item Short-Form Health Survey, the Beck Depression Inventory, and 3 direct health value measures, including the Rating Scale, Time Trade-off (TTO), and Standard Gamble (SG). Correlation among measures and factor analysis was performed.

Results  The mean modified Medical Outcomes Study 36-Item Short-Form Health Survey scores were lower than normative population values, particularly on the Physical Component Summary scale. This scale was poorly correlated with the Rating Scale, TTO, and SG scores among HCV-infected subjects. The mean ± SE TTO score was 0.83 ± 0.02, and the mean ± SE SG score was 0.79 ± 0.02. The TTO and SG scores failed to show significant variability in relation to disease activity as determined by serum alanine aminotransferase level, histologic stage, and presence of decompensated liver disease. The Beck Depression Inventory was significantly inversely correlated with the TTO and SG.

Conclusions  Although quality of life is compromised in patients with chronic HCV infection, patient-derived health utilities are not strongly associated with health status or clinical measures. Utility measures obtained from patients with HCV differ significantly from previous surrogate measures of health values. Such differences in utilities could affect decision analyses and cost-effectiveness analyses of treatment interventions for individuals with HCV infection.

Hepatitis C virus (HCV) infection has emerged as a major public health issue in the last decade. In a short time span since its discovery in 1989, epidemiologic investigations have demonstrated infection in approximately 1.8% of noninstitutionalized people in the United States.¹ The strong association of HCV infection with morbidity and mortality due to cirrhosis and hepatocellular carcinoma is well documented and represents a significant economic and health burden to society.^2,3 Despite these alarming statistics, a high proportion of cases remains undiagnosed in this “silent epidemic.”⁴ Barriers to diagnosis and treatment include the perception that most patients are asymptomatic and that treatment regimens are expensive, difficult to tolerate, and ineffective. Many patients acquired HCV through injection drug use, which represents a social stigma and a barrier to diagnosis and treatment.⁵ Because the disease and its treatment often affect health-related quality of life (HRQOL), HRQOL issues are important facets in the management of chronic HCV infection.

There are 2 primary methods used to assess HRQOL: health status assessment, which describes states of health and their effect on function and disability, and utility assessment, which ascertains the value or desirability of a particular health state. To date, virtually all of the HRQOL research related to chronic HCV infection and treatment has used the health status assessment approach.^6-9 While health status assessment reveals an inventory of patient symptoms, adverse effects, and functional abilities, the utility assessment permits direct patient input into a measurement of value. This technique refocuses attention on the patient’s views and understanding of the disease process. When coupled with decision analysis or cost-effectiveness analysis, the utility assessment integrates health values with other outcomes, such as survival and cost, thus permitting optimization of treatment-related decision making.¹⁰

Therefore, our study had 2 goals. The first was to assess utilities directly from patients with HCV infection, and the second was to determine whether utilities are related to patients’ health status, clinical stage, or demographics.

Patients with HCV infection were recruited from the outpatient clinics at the University of Cincinnati Medical Center, including an indigent hospital–based clinic, a private university–based clinical office practice, the liver transplantation clinic, and an outpatient human immunodeficiency virus (HIV) treatment center. All subjects had HCV infection diagnosed by serologic methods with confirmation by HCV RNA testing or the recombinant immunoblot assay. The University of Cincinnati Institutional Review Board approved the study protocol, and all patients provided informed consent before study participation.

Study participants were interviewed by a single professional interviewer (S.N.S.) with experience in administration of health assessment instruments. Interviews took approximately 30 minutes. In addition, patient records were validated for key diagnoses and laboratory values relevant to chronic liver disease, including serum alanine aminotransferase (ALT) levels, histologic findings, and comorbid disease processes. The interviewer was blinded to those data and to patient identifiers.

Questionnaires consisted of demographic questions, CAGE standardized alcoholism survey questions, the Hepatitis Quality-of-Life Questionnaire, the Beck Depression Inventory, and 3 direct health value measures. The Hepatitis Quality-of-Life Questionnaire was developed and validated in patients with chronic HCV infection.¹¹ It consists of the Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36) questions followed by hepatitis C–specific questions. The Beck Depression Inventory is a commonly used measure of depression that includes 21 questions with multiple-choice responses.¹² The instrument is scored based on cumulative response, with a range of 0 to 63 for the aggregate score; normal is defined as scores of 10 or less.

Health values were elicited using U-Maker (Frank Sonnenberg, New Brunswick, NJ), a computer-assisted utility assessment software package. The first health value instrument used was the Rating Scale (RS), which was presented as a “feeling thermometer,” with scores ranging from 0 (dead) to 100 (perfect health). Patients were requested to rate their current health on the 0 to 100 scale. Scalar measurements were divided by 100 to yield a 0 to 1.0 scale comparable to the other utility task outputs. The Time Trade-off (TTO) assesses the individual’s willingness to live a shorter but healthier life. The series begins with a choice between living 10 more years in his or her current state of health or living 10 years in perfect health. If the patient preferred 10 years in perfect health, he or she was offered a choice between 10 more years in current health or 0 years in perfect health (death). If 10 years in current health was preferred, the number of years in perfect health was varied systematically in a “ping-pong” fashion until the patient did not have a clear preference between living 10 more years in current health or living the given amount of time in perfect health. The TTO score was calculated by dividing the number of years of perfect health at the indifference point by 10, the number of years of current health offered. The final utility task was the Standard Gamble (SG), which assesses the individual’s willingness to risk a bad outcome (death) in exchange for a chance at a healthier life. The patient was offered a choice between living the remainder of life in his or her current health and taking a gamble for perfect health for the remainder of life or immediate death. As with the TTO, the probabilities of the 2 outcomes of the SG were varied systematically until the patient had no preference between the certainty of life in his or her current state of health or the gamble. The SG score was calculated by the following formula: 1 − the maximum percentage risk if death was acceptable. Therefore, all health value measures were scaled from 0 to 1.0. In all 3 utility tasks, the definition of perfect health was not provided; it was self-defined by each participant.

Baseline demographics were compared for continuous (t test and analysis of variance) and categorical (Fisher exact test and χ² test) variables as appropriate. Pearson product moment correlations were evaluated between outcome measures. We performed univariate analysis and compared multiple factors to observe utility outcomes. We conducted separate analyses of health values before and after excluding participants who were coinfected with HIV and HCV and patients who had undergone liver transplantation. Univariate analysis was performed, comparing multiple factors to observe the mean utility outcomes across levels of individual factors. Multivariate analysis of the contribution of factors to the utility scores was performed using a generalized linear model procedure. SAS (version 8.0; SAS Institute, Cary, NC) and SPSS (version 10.0; SPSS Inc, Chicago, Ill) software packages were used. All comparisons used a 2-tailed hypothesis with α≤.05.

One hundred twenty-six patients with chronic HCV infection were offered the opportunity to participate, and 124 completed the informed consent and the questionnaire (98.4% response rate). Two patients declined to participate because of lack of time. Nearly half (49.2%) of the subjects reported high school graduation or less as their highest level of educational attainment (Table 1). Thirty-five percent reported incomes less than $12 000 per year (poverty level), and 31.4% reported incomes in excess of $50 000 per year. The low proportion (1.6%) of Hispanic subjects reflects the demographic makeup of our region. The risk profile for acquisition of parenteral diseases shows that 47.6% reported a history of injection drug use, while 34.7% received a blood transfusion before 1991. The mean Beck Depression Inventory score was 12.77 and ranged from 0 to 40.

The mean ± SE Mental Component Summary (MCS) score among 118 subjects who completed the full questionnaire was 47.5 ± 1.5, while the mean ± SE Physical Component Summary (PCS) score was 34.5 ± 0.6. Nearly 43% of respondents thought that hepatitis limited physical activity to some extent (Figure 1). Similarly, 41.9% reported work and social activity limitations. Six patients skipped some questions, which precluded calculation of summary scores on those individuals.

The mean ± SE RS score was 0.63 ± 0.02 (range, 0.01-0.99) (Table 2). Among 124 HCV-infected respondents, the mean ± SE TTO utility was 0.83 ± 0.02; this indicates that, on average, patients were willing to trade 17% ([1−0.83] × 100%) of their remaining life expectancy in return for perfect health. The mean ± SE SG utility was 0.79 ± 0.02, indicating that, on average, patients were willing to take up to a 21% ([1−0.79] × 100%) risk of death in exchange for perfect health. Excluding patients with HIV coinfection and those who had previously undergone orthotopic liver transplantation, the mean ± SE RS score was 0.64 ± 0.02. In contrast, the mean ± SE TTO and SG scores were 0.81 ± 0.03 and 0.85 ± 0.02, respectively. The TTO and SG scores were significantly greater than the RS score (P<.001) and did not differ from each other. The RS scores were closely correlated with the MCS scores of the modified SF-36 (r = 0.74, P<.001), but less so with the PCS (r = 0.19). The TTO scores were poorly correlated with the PCS scores (r = 0.08) and weakly correlated with the MCS scores (r = 0.37). A much stronger association between the TTO and the SG scores was noted (r = 0.67, P<.001). Similarly, SG scores were not correlated with the PCS (r = 0.04), and their association with the MCS was similar to that observed for the TTO scores. Liver biopsy data were available for 62 patients. Among patients with cirrhosis, the mean RS score was 0.51 vs 0.67 in patients without cirrhosis (P = .02). However, there were no statistically significant differences in the TTO or SG scores. The mean TTO score for subjects with cirrhosis was 0.79 vs 0.85 for those without cirrhosis. Interestingly, the mean SG score for patients with cirrhosis was 0.83 vs 0.81 for those without cirrhosis.

Table 2 demonstrates the relationship of the utilities to disease inflammatory activity as determined by ALT strata. Only 55 subjects had ALT testing in close temporal proximity to the interview day. The strata were defined as normal or near normal (ALT ≤1.5 times the upper limit of normal) or elevated (ALT >1.5 times the upper limit of normal). The RS scores were uniform, independent of ALT levels. In both ALT categories, health values as determined by the TTO and SG scores were significantly higher than those determined by the RS scores.

Among patients undergoing treatment with any interferon-based therapy, the RS scores tended to be lower (mean ± SE, 0.60 ± 0.08), but the TTO and SG scores were among the highest observed (mean ± SE, 0.86 ± 0.07 and 0.84 ± 0.06, respectively). Similar utilities were derived from patients who completed treatment (independent of treatment outcome), with only the RS scores (mean ± SE, 0.62 ± 0.05) being closer to that of the untreated comparison group (mean ± SE, 0.66 ± 0.03).

Other patient subsets included in the analysis were HCV-infected patients who had undergone orthotopic liver transplantation for decompensated liver disease and patients with HCV and HIV coinfection. Although the numbers were small (n = 10 and n = 11, respectively), health utility values in both groups were comparable to those of subjects with decompensated liver disease. There was a trend toward lower TTO and SG scores among patients with decompensated liver disease, although the difference failed to reach statistical significance (Figure 2).

Factors analyzed included age, sex, race, educational level, income level, baseline serum ALT level, disease severity, presence or absence of cirrhosis, current alcohol use, CAGE standardized alcoholism survey score, history of depression, and Beck Depression Inventory score. Table 3 gives the univariate probabilities for each of the 3 outcome measures. Multivariate analysis demonstrates that all 3 outcome measures are inversely related to the Beck Depression Inventory scale (P<.001). There is a strong trend for sex being a factor in the RS when the Beck Depression Inventory score is controlled for (P = .051). The PCS and MCS are significant factors in the RS, but not in the TTO and SG. The SG appears to be affected by the educational level of the subjects (P = .001).

Health utilities have emerged as powerful tools in the hands of health care economists who seek to provide quality-of-life estimates for the status of health. Utilities may be measured directly (TTO and SG), measured indirectly with health state classification systems using the Health Utilities Index, or estimated by health care experts in the field.¹⁰ Direct measurement through instruments such as the SG and TTO is often time-consuming and difficult, leading many health analysts to seek simpler techniques.

Health status measurement in patients with HCV is becoming more commonplace in clinical trials of drugs and other interventions, in which the focus is to determine whether the intervention improves some or all aspects of health. The SF-36 has been widely used by gastroenterologists and hepatologists to measure health status because normative data for the United States are available for comparison.¹³ Our study confirms results from studies^14,15 that have demonstrated that patients with HCV infection have poorer health status than control subjects in the normative population. Furthermore, there is evidence that interferon therapy may improve HRQOL.¹⁶ Efforts to improve responsiveness of the SF-36 have led to the development of several disease-specific instruments for patients with chronic liver disease¹⁷ and an HCV-specific instrument that is based on the SF-36.¹¹ The mean PCS score of 34.5 in our study was significantly lower than the normative values of 50 for the US population and 48 for 107 HCV-infected patients described by Fontana and colleagues.¹⁸ In contrast, the mean MCS score was similar to that in previous reports and normative populations.¹⁸ Unlike the cohort described by Fontana and coauthors, our cohort represents a mix of socioeconomic strata; nearly one third of patients were at or near the poverty level, one third were middle-class, and approximately one third had higher income levels. This diverse population mix may explain the observed differences.

The decrement in quality of life among patients with HCV infection, and the evidence that successful treatment intervention may improve quality of life, has led health economists to use simple yet correlated surrogates such as the Health Utilities Index to determine the effect on cost-effectiveness analyses.¹⁹ However, the key element of this study was to evaluate the association between directly derived patient utilities and other HRQOL measures. To this end, we found that the RS, which is the simplest measure of health value, correlates strongly with the MCS but not with the PCS. Overall, the RS is a good indicator of HRQOL, but is in the strictest sense not a utility measure because it does not involve trade-offs against external measures such as time or risk. The true utility measures, the TTO and SG, were poorly correlated with the PCS, which was the main aspect of health affected in our HCV-infected population. Other investigations have suggested a poor association of health utilities with health status.¹⁰

In the absence of patient-derived utilities, health economists working in the HCV field have primarily used physician-derived health value estimates to calculate quality-adjusted life expectancy for the purpose of cost-effectiveness analyses.^20,21 Other experts argue that community-based health values should be used because society is ultimately responsible for determining allocation of assets and services.²²

Among treatment-naive patients, neither the TTO nor the SG was related to disease activity as determined by ALT level or histologic stage, although our power was limited. Similarly, we were unable to demonstrate any statistically meaningful association between health values and compensated or decompensated liver disease states. Chong et al²³ described use of the SG in a Canadian HCV cohort and failed to find significant differences between subgroups based on disease stage. This finding is important because of the potential effect on quality-of-life weights in economic modeling. For example, Bennett and coworkers²¹ assigned a baseline health value of 0.28 to 0.35 for patients with decompensated liver disease characterized by the presence of ascites, variceal hemorrhage, or hepatic encephalopathy. Their estimate is substantially lower than our patient-derived values (Figure 2). The use of actual patient utility values could alter their economic analysis, which is dependent on the prevention of transition to the more undesirable outcomes. A cost-effectiveness analysis by Kim and colleagues²⁴ suggested that failure to adjust for quality of life led to a 3-fold decrease in the cost-effectiveness of interferon alfa therapy. In their model, physician-derived utilities were used with decompensated cirrhosis equivalent to 0.5 years of healthy life, while the value for chronic HCV infection was 0.95 years. The use of health values derived from our data set would alter the analysis and interpretation rendered by Kim and colleagues that supported the value of interferon-based interventions.

Multivariate analysis was used to determine the contribution of important demographic and health variables to the utility scores. The strongest predictor was the inverse relationship between the Beck Depression Inventory and the patient-determined health value. Other studies have reported an association between HCV infection and depression. Singh et al²⁵ compared patients with HCV vs those with other liver diseases and noted increased mood disturbance and depression and dejection as measured by the Profile of Mood States scale. Furthermore, Beck Depression Inventory scores suggested significantly higher levels of depression compared with other liver diseases. Sex seems to be an important factor in the RS, and this association carries over to our modified SF-36 quality-of-life measures, but it is not a factor in the TTO and SG when depression, as determined by the Beck Depression Inventory, is controlled for. It is not clear whether treatment intervention for depression would alter the utilities in HCV-infected patients. Some investigations have reported a fluctuating relationship between the mean utility scores and depression severity, while others have failed to observe a change in health values with treatment.²⁶ Interestingly, we did not find a relationship between health values and risk factors for HCV infection. Because this study was a cross-sectional analysis, we are unable to determine if HCV treatment and the variable of treatment outcome affect utilities. However, the finding of high utility values in patients undergoing treatment or after treatment with interferon-based therapies seems counterintuitive because these patients in fact risked increased short-term morbidity in an effort to achieve perfect health.

In conclusion, patients with chronic HCV infection have health values that are minimally affected by disease stage and severity, despite reporting significant deficits in health status. This finding is at odds with physician- and expert panel–derived utilities and utilities based on health status measures; therefore, it could have a significant effect on quality-of-life adjustments used in decision analyses and cost-effectiveness analyses. Prospective evaluation of health utilities related to treatment of HCV infection and associated depression is indicated.

Correspondence: Kenneth E. Sherman, MD, PhD, Hepatology and Liver Transplant Medicine Section, Division of Digestive Diseases, Department of Medicine, University of Cincinnati College of Medicine, 231 Albert Sabin Way, Cincinnati, OH 45267-0595 (Kenneth.Sherman@uc.edu).

Accepted for Publication: April 27, 2004.

Financial Disclosure: None.

Funding/Support: This work was supported by grant R03 HS10366-01 from the Agency for Healthcare Research and Quality, Rockville, Md.

Acknowledgment: We thank Susan Rouster, BS, for her manuscript review and Richard Hornung, DrPH, and Christopher J. Lindsell, PhD, for statistical support.