The Impact of Chronic Hepatitis B on Quality of Life: A Multinational Study of Utilities from Infected and Uninfected Persons

The Impact of Chronic Hepatitis B on Quality of Life:A Multinational Study of Utilities from Infected andUninfected Persons

Adrian R. Levy, PhD,1,2 Kris V. Kowdley, MD,3 Uchenna Iloeje, MD, MPH,4 Eskinder Tafesse, PhD,4

Jayanti Mukherjee, PhD,4 Robert Gish, MD,5 Natalie Bzowej, MD,5 Andrew H. Briggs, DPhil1,6

1Oxford Outcomes Ltd,Vancouver, BC, Canada; 2University of British Columbia,Vancouver, BC, Canada; 3Digestive Disease Institute,VirginiaMason Medical Center, Seattle,WA, USA; 4Bristol-Myers Squibb Company,Wallingford, CT, USA; 5California Pacific Medical Center, SanFrancisco, CA, USA; 6Glasgow University, Glasgow, UK

ABSTRACT

Objectives: Chronic hepatitis B (CHB) is a condition thatresults in substantial morbidity and mortality worldwidebecause of progressive liver damage. Investigators undertak-ing economic evaluations of new therapeutic agents requireestimates of health-related quality of life (HRQOL). Re-cently, evidence has begun to accumulate that differences incultural backgrounds have a quantifiable impact on percep-tions of health. The objective was to elicit utilities for sixhealth states that occur after infection with the hepatitis Bvirus from infected and uninfected respondents living in juris-dictions with low and with high CHB endemicity.Methods: Standard gamble utilities were elicited fromhepatitis patients and uninfected respondents using aninterviewer-administered survey in the United States,Canada, United Kingdom, Spain, Hong Kong, and mainlandChina. Generalized linear models were used to the effect onutilities of current health, age and sex, jurisdiction and, forinfected respondents, current disease state.Results: The sample included 534 CHB-infected patients and600 uninfected respondents. CHB and compensated cirrhosis

had a moderate impact on HRQOL with utilities rangingfrom 0.68 to 0.80. Decompensated cirrhosis and hepatocel-lular carcinoma had a stronger impact with utilities rangingfrom 0.35 to 0.41. Significant variation was observedbetween countries, with both types of respondents in main-land China and Hong Kong reporting systematically lowerutilities.Conclusions: Health states related to CHB infection havesubstantial reductions in HRQOL and the utilities reportedin this study provide valuable information for comparingnew treatment options. The observed intercountry differ-ences suggest that economic evaluations may benefit fromcountry-specific utility estimates. The extent that systematicintercountry differences in utilities hold true for other infec-tious and chronic diseases remains an open question and hasconsiderable implications for the proper conduct and inter-pretation of economic evaluations.Keywords: cost-effectiveness, cost-utility international varia-tion, hepatitis B, quality of life, utilities.

Introduction

Valuations of health states are fundamental to decisionmodels and economic evaluations. During the 60 yearsthat have elapsed since von Neumann and Morgen-stern proposed “expected utility theory” as a norma-tive framework for making rational decisions underuncertainty [1], much effort has been devoted to apply-ing this theory to health and developing methods tovalidly evaluate different health states. Despite theseefforts, obtaining valid and reliable health state valu-ations and understanding the sources of variabilitycontinue to pose a formidable challenge for research-ers, clinicians, and decision-makers [2]. Valid assess-

ment of quality of life is germane to decision-makingbecause the interpretations of economic evaluationscan change according to the utilities that are used [3].A review of 70 cost-effectiveness articles found that theinterpretations changed in approximately one-third ofinvestigator-reported sensitivity analyses [4]. Recently,evidence has begun to accumulate that differences inethnic and cultural backgrounds have a quantifiableimpact on perceptions of health [5]. Nevertheless, theprecise nature of that effect on utilities is poorlyunderstood [5].

Chronic hepatitis B (CHB) infection is a conditionfor which several new treatments have recently becomeavailable [6]. The clinical course of CHB infection canbe divided into two phases: replicative (active viralreplication which may or may not be associated withhost immunologic response) and integrative (absenceof active viral replication associated with a clinically

Address correspondence to: Adrian R. Levy, Oxford OutcomesLtd, 450–688 West Hastings Street, Vancouver, BC, CanadaV6B 1P1. E-mail: [email protected]

10.1111/j.1524-4733.2007.00297.x

Volume 11 • Number 3 • 2008V A L U E I N H E A L T H

© 2007, International Society for Pharmacoeconomics and Outcomes Research (ISPOR) 1098-3015/08/527 527–538 527

quiescent phase) [7]. Many persons infected early inlife remain in the replicative stage for several yearsbefore stimulating a host immune response that canlead to progressive liver damage and potentially fatalcomplications including compensated and decompen-sated cirrhosis (ascites, jaundice, portal hypertension,esophageal and gastric variceal bleeding). Hepatocel-lular carcinoma is a complication that may developafter a long duration of infection in the replicative andnonreplicative phases [6]. The 5-year cumulative inci-dence of compensated cirrhosis among hepatitis Bvirus (HBV) patients ranges from 8% to 20% and maybe even higher [8]. The rate of progression is depen-dent on clinical and serological features, such as age atdiagnosis, sex, alcohol intake, coinfection with hepa-titis C or D virus or human immunodeficiency virus,HBV genotype, hepatitis B e antigen positivity [8], andthe level of circulating virus [9–12]. Compared touninfected persons, chronically infected patients pro-gressing to compensated and decompensated cirrhosishave an approximately 100-fold increased risk ofhepatocellular carcinoma [13,14]. As liver diseasefrom CHB progresses, so do the morbidity and thecosts of care [15–23]. Decompensated cirrhosis andhepatocellular cancer are associated with severe mor-bidity and high mortality if the patient does notundergo liver transplantation. That is a costly proce-dure with a long recovery time and impaired health-related quality of life (HRQOL) in the first year posttransplant [24].

Directly elicited valuations of health states com-monly reported in the literature are “ratings,” derivedfrom a visual analog scale [25], and “utilities,” derivedusing elicitation techniques in which respondents areobliged to make a trade-off between the likelihoods ofsurvival, called standard gamble, or duration of sur-vival, called time trade-off [26]. Investigators conduct-ing cost-utility analyses of treatments for hepatitis Bhave not incorporated utilities elicited specifically forCHB-related health states from either CHB patients oruninfected persons [27–31]. One study [27] incorpo-rated published standard gamble utilities from hepati-tis C virus patients [32] and another study used dataelicited from a panel of seven physicians familiar withtreating hepatitis C patients, with the mean utilitiescombined from both standard gamble and time trade-off techniques [28]. The type of respondents and the adhoc method of combining utilities from both elicitationtechniques in the study raise concerns regarding thevalidity of the quality-of-life adjustments and theresultant cost-utility ratios.

The objective of this study was to elicit utilities forsix health states that occur after chronic infection withthe HBV from infected and uninfected respondentsliving in regions with low and with high CHB ende-micity. As there is no consensus whether lay personsnaïve to the disease under study or patients suffering

from the disease provide more appropriate data onutilities [33–36], we collected information from bothgroups: uninfected persons were included to representthe general population [37] and respondents infectedwith CHB were included to provide the patients’ per-spective. Although differences in ethnic and culturalbackgrounds are likely to have an impact on percep-tions of health, the nature of these differences on utili-ties is poorly understood. In the case of infectioushepatitis, risk attitudes and utilities may be influencedby societal awareness about the condition because of:the likelihood of personally knowing someone who isinfected, to the public health burden, to the contagiousnature of the virus, or to other reasons. Thus, wecollected data in jurisdictions with hepatitis B surfaceantigen seroprevalence ranging from less than 1% inthe United States, Canada, and Western Europeancountries to between 10% and 12% in North andCentral Asian countries (China, South Korea, andTaiwan) [6] to reflect the potential impact on utilitiesof low and high prevalences of CHB infection.

Methods

Design and Study RespondentsWe used trained interviewers to directly elicit utilitiesfor CHB-related health states using the standardgamble technique [38]. Data were collected fromrespondents enrolled in 11 cities in six jurisdictions:United States, Canada, United Kingdom, Spain, main-land China, and Hong Kong. (Data were collectedseparately in Hong Kong and mainland China becauseof differences in their economies and health-caresystems.) These countries represented regions with lowand high endemicity of HBV infection as well as dif-ferent ethnic and cultural backgrounds [6].

We aimed to recruit 100 uninfected respondentsand 100 HBV-infected patients at each site. Thesetargets were established within the context of inad-equate knowledge of standard gamble (SG) utilitiescharacterized by the lack of information on variabil-ity, little theoretical guidance on the minimal clinicallyimportant difference, and poor understanding of theeffect of demographic and cultural factors. Thesample size thus represented a compromise betweenprecision, cost, speed of data collection, and the needto maximize the number of respondents. Severaluseful, but arbitrary, rules of thumb have been sug-gested for utility measurement that can be used as arough guide to estimate power: 0.1 is the minimalclinically important difference (i.e., effect size) andstandard deviations are greater for the general public(0.3) than for patient groups (0.2) [39]. Using thosefigures and a projected sample size of 600, the powerof detecting a difference among uninfected respon-dents was 17% for an effect size of 0.05 and 53% for

528 Levy et al.

an effect size of 0.10; among infected respondents thepower was 28% for an effect size of 0.05 and 78%for an effect size of 0.10.

Uninfected respondents were primarily recruitedfrom staff and students at local universities as well asthe population at large. Infected respondents wererecruited consecutively from 11 liver disease treatmentcenters and clinics, transplant centers, and hospitalhepatology units (one in Hong Kong and two in eachother site [see Acknowledgments]). Ethical approvalwas obtained from institutional review boards in allcountries and all respondents provided informedconsent.

The inclusion criterion for uninfected respondentswas age 18 years or older, and the exclusion criteriawere: inability to communicate in the official lan-guage(s) of the country; known cognitive impairment;or, visible intoxication that, in the interviewer’sopinion, prevented valid responses. The inclusion cri-teria for infected respondents were: age 18 years orolder, hepatitis B e antigen positivity for at least6 months, and classifiable into one of the six healthstates. The exclusion criteria for infected respondentswere the same as those for uninfected respondents, aswell as: less than 3 weeks post transplant or post hepa-tectomy; end-stage chronic illness unrelated to CHB;or known co-infection with human immunodeficiencyvirus, hepatitis C, or hepatitis D.

Standardized Descriptions of Health StatesWe developed standardized descriptions [40] that char-acterize six CHB-related health states: CHB, compen-sated and decompensated cirrhosis, hepatocellularcarcinoma, and liver transplantation within and afterthe first year. Health states were developed in a tabularformat as these have been shown to be the preferredformat for presentation [41]. Relevant dimensions ofHRQOL were based on the Liver Disease Quality-of-Life Instrument version 1.0 [42]. Based on the inter-views with three experienced hepatology experts (fromCanada, United States, and Hong Kong), we iterativelydeveloped written descriptions of each health state thatportrayed typical symptoms, frequency of tests, hospi-talizations, procedures, and dimensions of health suchas pain, ability for self-care, activities of daily living,psychological well-being, and future outlook. The finalversions of the health state descriptions achieved con-sensual agreement among the experts (Table 1). Pre-testing included forward and backward translationand linguistic validation interviews (Canada: Englishand French; United Kingdom: English; United States:English; Castilian Spanish; China: Cantonese andMandarin). The elicitation instruments and healthstate descriptions were pilot-tested in a sample of 14uninfected respondents in Canada and the UnitedKingdom.

Interview ProcessWe standardized the data collection process by con-ducting a 1-day training session in each site before datacollection that involved: live practice sessions; havinginterviewers employ the same script to explain thestudy; and instructing interviewers to minimizeunscripted dialogue.

The interviewers presented the health states torespondents in different orders. Respondents saw onlythe health state description and remained unawareof the health state name. Infected respondents wereaware that these were health states related to hepatitis(but not necessarily to their own health state). Tofamiliarize respondents with the health states, theinterview process began by eliciting visual analog scaleratings [43]. Respondents were asked to rate all sixhealth states, their own current health and death usinga vertical visual analog scale ranging from 0 (worstimaginable health state) to 100 (perfect health) [25].Then, a probability wheel was employed as a prop toelicit the standard gamble utilities associated with theircurrent health and each of the six CHB-related healthstates. Respondents were asked to choose repeatedlybetween two options: 1) remaining with certainty inthe health state without improvement; or 2) a hypo-thetical treatment that may either result in full health(utility level of 1) or cause immediate death (utilitylevel of 0) [2]. The probabilities for the second optionwere changed until respondents reached a point ofindifference between the alternatives. The processincorporated a “ping-pong” approach with probabi-lities traded back and forth between higher and lowervalues that iteratively narrowed to the point of indif-ference [2]. The utility weight is calculated as 1 minusthe probability of death at that point [2].

AnalysisMean standard gamble utilities and the corresponding95% confidence intervals (CIs) were calculated for eachhealth state including current health. Statistical testingof differences in utilities was conducted using gen-eralized linear models [44]. Two types of models wereconsidered: Gaussian with an identity link (i.e., ordi-nary least squares) and a gamma family with a log link.The second model was considered because the ordinaryleast squares model has the disadvantage of potentiallypredicting values above the upper boundary (i.e., 1.0).Given that utilities may be non-normal and have anupper bound of 1.0, we modeled (1—utility).

The goodness of fit was compared using the modeldeviances. Using the best fitting model (i.e., that withthe lowest deviance), the regression modeling was usedto determine: the effect of age (in years) and sex amongboth groups of respondents; whether utilities differedby type of respondent or country; and among infectedrespondents only, whether respondents categorized inone health state rated that state differently than other

Health State Utilities for Hepatitis B 529

infected respondents. To examine if infected respon-dents who had progressed to a more severe health staterated the less severe states differently than infectedrespondents with milder forms of disease, we groupedinfected respondents into more (decompensated cir-rhosis, hepatocellular carcinoma, and the two trans-plant states) or less (CHB, compensated cirrhosis)severe. We then compared between more and lesssevere respondents the ratings for CHB and compen-sated cirrhosis. Statistical analyses were undertakenusing Stata for Windows versions 8.2 (Stata Corpora-tion, College Station, TX). All statistical comparisonswere two-tailed and used alpha of 0.05.

The standard gamble technique does not allowrespondents to record states as worse than death.Although adjustments to the technique have been pro-posed to quantify such states [45], these procedureswere not employed to avoid respondent burden.

Instead, an arbitrary weight of -0.1 was applied toall states where immediate death was preferred. Theimpact of this assumption was assessed using sensitiv-ity analysis by examining the impact of changing thearbitrary weight of -0.1 to -0.5 for states in whichimmediate death was preferred.

Results

A total of 1134 respondents were enrolled, 600 unin-fected and 534 infected with CHB virus (Table 2). Foruninfected respondents, the mean age was 41.5 years,43% were male and 36% were Asian, while forinfected respondents, the mean age was 45.5 years,76% were male and 62% were Asian. Among infectedrespondents, the preponderance of men and those inthe CHB health state reflected epidemiological distri-butions of the infection.

Table 1 Standardized descriptions used to characterize six chronic hepatitis B-related health states

Chronic hepatitis B• Sometimes I feel tired, but otherwise I feel healthy most of the time.• My condition doesn’t really limit my daily life. I still participate in many activities such as sports, household chores, and social events.• Occasionally I go to the doctor’s office or clinic for different tests.• I have heard that my condition can be progressive and sometimes I feel anxious about it.• I occasionally have concerns about my future.Compensated cirrhosis• Sometimes I feel tired, but otherwise I feel healthy most of the time.• Most days I feel rested, but a little of the time I don’t sleep as well.• My condition doesn’t really limit my daily life. I still participate in many activities such as sports, household chores, and social events.• I go to the doctor’s office or clinic periodically for a variety of medical tests.• I feel somewhat worried and anxious about my future.Decompensated cirrhosis• I feel like I have the flu all of the time. Some days I am exhausted because I have difficulty sleeping at night. My feet and legs may be swollen andsometimes I feel bloated.

• Sometimes I vomit blood and have to go to the hospital for a blood transfusion and to have a tube placed in my stomach through my nose.• My diet is restricted because of my condition.• Sometimes I am reluctant to leave the house because I’m not feeling well.• My condition takes up too much of my life. I make frequent trips to the hospital for tests and I take several medications.These medications mayhave unpleasant side effects such as diarrhea, cramps, bloating, and fatigue.

• There are times when I am confused and I often have trouble remembering things.• I worry about whether or not a life-saving treatment will be available when I need one. I am very concerned about my future.• I find I am less inclined to spend time with my friends because of my condition.Transplantation (1st year)• I feel tired during the day because I don’t sleep very well at night.• Sometimes I have the energy to go for a walk or do the housekeeping, but I tire easily.• Most of the time I have good appetite and am able to eat and enjoy regular meals.• I make frequent trips to the hospital or clinic for a battery of tests.• I take several medications, some of which may have unpleasant side effects such as fatigue or nausea.• I feel anxious and worried about my life and hope that my health problems will resolve.• I feel that my appearance has changed and sometimes I feel embarrassed about it.Transplantation (>1st year)• I have a healthy appetite which allows me to eat and enjoy regular meals.• Although I tire easily, my level of energy is improving and I find it easier to participate in household tasks and leisure activities.• I take several medications which have some side effects, such as headaches and bloating.• I may experience aches and pain in my bones and joints.• My appearance is improving and I am more comfortable going out in public.• My visits to the hospital and the doctor’s office are becoming less frequent and I spend less time managing my condition.• I am somewhat optimistic about my future and that my health will stay the same.Hepatocellular carcinoma• I find that I have very little energy. I often have sharp pains in my abdomen and I frequently feel nauseous.• I am often uncomfortable because of swelling in my legs, and sometimes in my abdomen.• I am able to eat only small amounts of food because I have no appetite and the food doesn’t seem to taste good anymore.• I spend a lot of time at the hospital and specialized clinic to undergo tests.• If I receive treatment, the side effects can include pain, nausea, hair loss or fever.• I know my condition is serious and I feel that my future is bleak.• I feel depressed and irritable.• My condition severely limits my ability to participate in usual activities such as house chores and seeing friends.

530 Levy et al.

The rank ordering of health states using standardgamble utilities was identical among both infected anduninfected respondents (Table 3). The health stateswith the highest mean utilities were CHB and compen-sated cirrhosis and the lowest mean utilities wereobserved for decompensated cirrhosis and hepatocel-lular carcinoma. The mean utility for the first year ofliver transplant was intermediate between the leastpreferred and most preferred health states and livertransplant after the first year had similar utilities toCHB and compensated cirrhosis.

For all comparisons, the Gaussian model with anidentity link showed lower deviances, and therefore abetter fit for the data, than the gamma family with a loglink. All results pertaining to the regressions therefore

refer to the ordinary least squares model. Statisticaltesting for the entire model indicated significant hetero-geneity so that we explored the sources of variation.There were significant differences between uninfectedand infected respondents for current health and forCHB, compensated cirrhosis, and the two transplanthealth states. Nevertheless, no differences between thetwo groups were observed for utilities assigned to dec-ompensated cirrhosis or hepatocellular carcinoma.

We observed that, as expected because they had anidentified illness and more than 35% were in the moreserious health states (decompensated cirrhosis, hepa-tocellular carcinoma, or transplant), infected respon-dents had significantly lower mean utilities for theircurrent health state than uninfected respondents.

Table 2 Sociodemographic and disease characteristics of infected and uninfected respondents from whom utilities for hepatitisB-related health states were elicited*

Characteristic

Infected respondents Uninfected respondents

N* n (%) N* n (%)

Male sex 530 402 (75.9) 599 256 (42.7)Age (year)*† 531 45.5 (13.1) 598 41.5 (16.7)(Range) 18 to 80 18 to 88Country 534 600United States 56 (10.5) 100 (16.7)Canada 100 (18.7) 100 (16.7)United Kingdom 93 (17.4) 100 (16.7)Spain 85 (15.9) 100 (16.7)Hong Kong 100 (18.7) 100 (16.7)China 100 (18.7) 100 (16.7)

Race 533 600White 155 (29.1) 373 (62.2)Asian 332 (62.3) 216 (36.0)Black 30 (5.6) 10 (1.7)Unknown 16 (3.0) 1 (0.0)

Marital status––married/common-law 511 397 (77.7) 578 332 (57.4)Employment—full time 520 299 (57.5) 582 319 (54.8)Education—continued after age 16 years 519 368 (70.9) 596 460 (77.2)Proportion in each health state 534Chronic hepatitis B 225 (42.1)Compensated cirrhosis 98 (18.4)Decompensated cirrhosis 49 (9.2)Liver transplant—1st year 56 (10.5)Liver transplant—>1st year 50 (9.4)Hepatocellular carcinoma 39 (7.3)Missing 17 (3.1)

*As some respondents did not answer all questions, the number of valid response for each characteristic differs.†Values represent mean (SD).

Table 3 Distributions of standard gamble utilities for six chronic hepatitis B-related health states and current health according to typeof respondent

Health state

Infected respondents Uninfected respondents

Mean 95% CI Median IQR Mean 95% CI Median IQR

Chronic hepatitis B 0.68 0.66–0.70 0.75 0.40 0.77 0.76–0.79 0.85 0.30Compensated cirrhosis 0.69 0.66–0.71 0.75 0.40 0.80 0.79–0.82 0.85 0.20Decompensated cirrhosis 0.35 0.32–0.37 0.35 0.50 0.35 0.33–0.37 0.35 0.40Hepatocellular carcinoma 0.38 0.36–0.41 0.40 0.50 0.41 0.39–0.43 0.45 0.50Liver transplant (1st year) 0.57 0.54–0.60 0.65 0.40 0.65 0.63–0.67 0.75 0.38Liver transplant (>1st year) 0.67 0.64–0.69 0.75 0.40 0.76 0.75–0.78 0.85 0.30Current health 0.70 0.67–0.73 0.75 0.40 0.87 0.86–0.89 0.95 0.10

CI, confidence interval; IQR, interquartile range.

Health State Utilities for Hepatitis B 531

Table 4 shows the mean utilities for current healthand the six CHB-related health states according to agegroup and sex separately for infected and uninfectedrespondents. There were no statistically significant dif-ferences between women and men of either respondentgroup in utilities elicited for any of the health states.Age was significantly associated with the utility forcurrent health among uninfected respondents withvalues declining with increasing age The effect of ageon current health was not significant among infectedrespondents. Of the six hepatitis-related health states,age was a significant predictor of utilities for CHBonly, among both infected and uninfected respondents.

There were significant differences between countriesin utilities for infected and uninfected respondents.Both groups of respondents in mainland China andHong Kong reported lower mean utilities than thosefrom respondents in other countries (Fig. 1). The dif-ferences between countries in utilities were statisticallysignificant for both respondent groups for all healthstates except decompensated cirrhosis and hepatocel-lular carcinoma.

None of the regression models indicated thatinfected subjects currently in one health state ratedthat particular health state differently than otherinfected respondents. For CHB, infected respondentsin more severe health states assigned a significantlylower utility than those in less severe health states,0.644 versus 0.719, respectively.

Table 5 shows the country-specific utilities for eachrespondent group after adjusting for age and sex usingordinary least squares regression.

To investigate the impact of assigning an arbitraryvalue of -0.1 to states in which immediate death waspreferred, we conducted a sensitivity analysis using a

value of -0.5 for the 3.5% of values where -0.1 wasassigned and found that the mean utilities werelowered slightly (decompensated cirrhosis = 0.31 andhepatocellular carcinoma = 0.37) with the relativerankings remained unaffected.

Discussion

It is estimated that more than 350 million peopleworldwide are infected with HBV [46,47]. Moresevere stages of liver disease due to CHB infectioncause reduced HRQOL and are associated withincreased treatment costs [15–23]. Despite the avail-ability of effective prophylaxis and treatment options[48], no information has been published on utilities forhealth states resulting from CHB infection elicitedfrom uninfected lay persons. The current study fillsthat gap by presenting standard gamble utilities for sixrelevant health states that were elicited using standard-ized procedures from respondents in six diverse juris-dictions. To contextualize the study in the light ofcurrent understanding about preferences, it is worthnoting that each major element in collecting andvaluing utilities—choosing between types of respon-dents and identifying a representative sampling frame,selecting the type of scaling task (e.g., standard gamblevs. time trade-off), identifying and portraying relevanthealth states, summarizing utilities across subjects, andassessing reliability and validity—is contentious [49].Nevertheless, this study is among the largest in theliterature to present information on health state utili-ties for any disease, one of the few investigations toinclude international comparisons within the samestudy, and included both infected and uninfectedrespondents.

Table 4 Mean standard gamble utilities by age group and sex for 534 infected and 600 uninfected persons for current health and sixchronic hepatitis B-related health states according to age and sex

Numberof

subjectsCurrenthealth

ChronichepatitisB

Compensatedcirrhosis

Decompensatedcirrhosis

Livertransplant,year 1

Livertransplant,after year 1

Hepatocellularcarcinoma

Infected respondentsAge (year)

<30 64 0.64 0.70 0.68 0.30 0.57 0.62 0.3230–39 120 0.72 0.68 0.70 0.31 0.54 0.68 0.3740–49 123 0.73 0.69 0.68 0.38 0.63 0.69 0.4150–59 131 0.71 0.70 0.70 0.35 0.56 0.68 0.39�60 89 0.66 0.63 0.66 0.37 0.55 0.66 0.41

SexMale 397 0.69 0.66 0.68 0.35 0.55 0.67 0.38Female 128 0.74 0.73 0.71 0.33 0.62 0.68 0.40

Uninfected respondentsAge (year)

<30 200 0.91 0.80 0.81 0.37 0.67 0.78 0.4230–39 100 0.87 0.80 0.84 0.32 0.66 0.79 0.4040–49 98 0.85 0.77 0.80 0.36 0.65 0.75 0.4150–59 98 0.82 0.71 0.75 0.39 0.61 0.72 0.41�60 102 0.84 0.76 0.80 0.30 0.64 0.76 0.39

SexMale 256 0.85 0.78 0.79 0.33 0.64 0.76 0.40Female 343 0.88 0.77 0.81 0.36 0.66 0.77 0.41

532 Levy et al.

Some insight into the representativeness of the unin-fected sample can be gleaned by comparing the valuesfor current health elicited from those respondents(mean 0.87, 95% CI 0.86–0.89) to population normsreported in other studies. Two studies, reporting datafrom 17,626 community dwelling Canadians and 3010community dwelling Australians, reported utilities thatwere derived from different multiattribute systemsbased on the time trade-off technique (the Health Utili-ties Index Mark II and the Assessment of Quality ofLife, respectively) [50,51]. The mean value for the 43%of the Canadian sample without any chronic conditionswas 0.93, and ranged between 0.58 and 0.92 for differ-ent conditions reported in the remainder of the sample[50]. The values reported in the current study weresimilar to those elicited from younger Australians [51].Similar to both surveys and other studies [52], we alsofound that the utilities assigned to current health foruninfected respondents declined with age.

Including two categories of respondents, CHB-infected and uninfected persons, was a featuredesigned to collect utilities from individuals who mayfundamentally differ in terms of their utilities forhealth states [37]. We observed that uninfected respon-dents had higher mean utilities than infected respon-

dents for most of the health states. Testing indicatedthat these values were statistically higher for all but themost severe health states (decompensated cirrhosis andhepatocellular carcinoma). This finding suggests thatboth patients and lay persons tend to have similar riskattitudes to more severe health conditions.

There is no consensus whether patients who havedirect experience of a disease’s impact or lay personsfrom the general population yield more representativeutilities when a societal perspective is sought. In con-trast to our findings, the prevailing interpretationliterature is that lay persons tend to providelower utilities than persons with disease because of the“disability paradox” [53–55] and response shift andadaptation to chronic health states [56]. Althoughsome studies have shown differences between the twogroups [33,57–60], the finding has not been universal[40,61,62] such that the interpretation is coming undergreater scrutiny. For example, a recent meta-analysisshowed no systematic differences between patients andpersons without disease [63]. Furthermore, as approxi-mately 42% of infected respondents were asymptom-atic, personal anxiety about disease progression beforeany possible response shift or adaptation may haveresulted in lower utilities among infected respondents.

Health State

HKCAUS

UK

SP

CH

HKCAUSUKSP

CH

HK

CAUS

UK

SP

CH

HK

CAUS

UK

SP

CH

HKCA

US

UK

SP

CH

HK

CAUS

UK

SP

CH

HK

CAUSUKSP

CH

HK

CAUSUKSP

CH

HK

CA

USUK

SP

CH

HK

CA

US

UKSP

CH

HK

CAUS

UKSP

CH

HK

CA

US

UKSP

CH

Infected Uninfected Infected Uninfected Infected Uninfected Infected Uninfected Infected Uninfected Infected Uninfected

0.0

0.2

0.4

0.6

0.8

1.0

Chronic HBV

Sta

ndar

d G

ambl

e U

tili

ties

Comp Cirrhosis Decomp Cirrhosis Transplant (year<1) Transplant (year>1) HCC

Figure 1 Mean standard gamble utilities elicited from infected and uninfected respondents for six chronic hepatitis B-related health states, by geographiclocation. Comp, compensated; Decomp, decompensated; HBV, hoursepatitis B virus; HCC, hoursepatocellular carcinoma. Countries: CA, Canada; CH,China; HK, Hong Kong; SP, Spain; UK, United Kingdom; US, United States.

Health State Utilities for Hepatitis B 533

We observed that respondents exhibited disutilityeven for health states with less severe health effectssuch as CHB, which showed mean utilities of 0.68 and0.77 among infected and uninfected respondents,respectively. Given that the health effects weredescribed as relatively minor (see Table 1), it is possiblethat respondents focused on the likelihood that thedisease would progress. The observation that infectedrespondents in more severe health states assigned alower score to CHB than those in less severe healthstates suggests that this anxiety was stronger amongpersons who were more familiar with disease progres-sion and that the differences between types of respon-dents may have been larger had a larger proportion ofsevere patients been included.

There is little existing information on the effectof age or sex on utilities. In one study done on morethan 3000 members of the UK general population, ageinfluenced utilities derived using the EQ-5D, withlower values reported from persons aged more than60 years than those aged 18 to 59 years [64]. Never-theless, those differences were attenuated whenadjusted for states worse than death. We found nodifferences in utilities between sexes, and significantdifferences by age in utilities only for CHB.

Another important finding was the difference inutilities between countries, with lower utilities re-ported from mainland China and Hong Kong, bothjurisdictions with high endemicity of hepatitis B infec-tion. The lower ratings of even the milder health states(CHB and compensated cirrhosis) in mainland Chinaand Hong Kong may reflect a greater fear of the socialconsequences of infection. For example, discrimina-tion against Chinese carriers of CHB—130 millionpersons equivalent to 10% of the population—canresult in difficulties in finding employment and accessto education [65].

Perceptions of health are complex functions ofcultural conceptions of health, contact with healthservices, and sociodemographic variables [66]. Never-theless, the means by which these issues interact andaffect self-perceived HRQOL are poorly understood.One interpretation of the significant intercountry dif-ferences is that researchers conducting economicevaluations, and decision-makers allocating fundingfor health technologies, should consider the geo-graphic source of utilities to assess whether those uti-lities were elicited from respondents living in ajurisdiction with a similar disease occurrence. In thecase of CHB, our study indicates that applying utilitieselicited in China or Hong Kong would likely be inap-propriate for economic evaluations submitted to theUK National Institute for Health and Clinical Excel-lence or the Canadian Common Drug Review.

We found that the statistical variability was greateramong infected than uninfected respondents. This sup-ports findings from studies in which the distributionsTa

ble

5Age-andsex-adjustedstandardgambleutilitiesfrom

infectedanduninfectedpersonsforsixchronichepatitisB-relatedhealthstates,accordingtocountry

Country

ChronichepatitisB

Compensated

cirrhosis

Decompensated

cirrhosis

Livertransplant—

firstyear

Livertransplant—

afterfirstyear

Hepatocellular

carcinoma

Uninfected

Infected

Uninfected

Infected

Uninfected

Infected

Uninfected

Infected

Uninfected

Infected

Uninfected

Infected

China

0.71

0.52

0.76

0.57

0.28

0.26

0.54

0.41

0.70

0.55

0.31

0.31

HongKong

0.79

0.60

0.82

0.64

0.32

0.30

0.69

0.56

0.80

0.64

0.38

0.38

Canada

0.85

0.66

0.83

0.65

0.45

0.44

0.71

0.58

0.80

0.64

0.47

0.46

UnitedStates

0.86

0.67

0.85

0.66

0.39

0.37

0.69

0.57

0.80

0.64

0.43

0.43

UnitedKingdom

0.88

0.69

0.87

0.68

0.36

0.35

0.69

0.57

0.82

0.66

0.42

0.42

Spain

0.86

0.67

0.89

0.71

0.45

0.44

0.74

0.61

0.85

0.70

0.48

0.48

534 Levy et al.

of utility measurements were wider for the generalpublic than for patients [39]. The difference has beenhypothesized to arise because diseased respondentshave first-hand experience of the health problems anda greater understanding of the impact on quality of life[67]. Investigators planning future utility studies canuse the data published herein to estimate power andsample size.

Comparisons between ratings derived from scalingmethods such as the visual analog scale and standardgamble or time-trade-off utilities have produced bothsimilar and divergent results [68,69]. Some of thisdiscrepancy may be due to the upper reference point,whether it is the absence of the dysfunction and dis-comfort being assessed or “perfect health” [70]. Using“full health” as the anchor, among uninfected respon-dents, we found that visual analog scale ratings wereconsistently lower than the standard gamble (data notshown). This is consistent with other published find-ings [25,71]. In contrast, we did not observe consistentdifferences between the visual analog scale and thestandard gamble among infected respondents (data notshown). This finding was unexpected because, forvisual analog scale ratings, respondents are not askedto choose between potentially undesirable alternatives[25,49].

Published economic evaluations of CHB treatmentoptions have been based on utility estimates that werelikely not representative of either societal or patientperspectives. In one study, utilities were elicited fromseven physicians familiar with treating hepatitis Cpatients and the mean utilities combined from bothstandard gamble and time trade-off techniques werereported for CHB, compensated cirrhosis, and decom-pensated cirrhosis. Those values were higher thanobserved in the current study: 0.94, 0.92, and 0.54,respectively [28].

Two other studies have recently been published thatreport patient-elicited utilities for hepatitis-relatedhealth states (Table 6) [32,72]. For all health states, theutilities measured in our study were generally slightlylower than those estimates. One study presented utili-ties from 73 patients for liver transplant but did notdescribe the elicitation methods [73]. Other investiga-tors have published utilities based on clinical judgmentfor CHB [74], compensated and decompensated cirrho-sis [74,75], liver transplant [74], and asymptomatic,mildly symptomatic and severely symptomatic HBVinfection [76]. A meta-analysis of the agreementbetween physicians and patients on HRQOL assess-ments found that long-term assessments of patients’well-being by physicians and patients readily diverge[36]. Physician-reported standard gamble utilities havebeen shown to be significantly higher than those elicitedfrom patients [33], perhaps because physicians system-atically overestimate how well patients feel and do notunderstand the real impact of illness on patients’ lives. Ta

ble

6Meanutilities(95%

confidenceintervals)forhepatitis-relatedhealthstateselicitedfrom

personsinfectedwithhepatitisreportedintheliterature

Author,location,

publicationyear

Numberof

infected

respondents

Elicitation

technique

Numberof

healthstates

evaluated

Healthstate

description

Chronic

hepatitis

Compensated

cirrhosis

Decompensated

cirrhosis

Hepatocellular

carcinoma

Transplant

Chong,Toronto,

Canada,2003

193

Standardgamble

withvisualprop

Currenthealthonly

Nonementioned

0.79

(0.70–0.87)

0.80

(0.70–0.90)

0.60

(0.37–0.83)

0.72

(0.62–0.82)

0.72

(0.62–0.82)*

Wells,Birmingham,

Alabama,2004

114

Questionnaire-based

timetrade-off

Six

Narrative

0.88

�0.15

‡0.74

†�0.20

‡0.30

�0.29

‡

Currentstudy,

sixjurisdictions

534

Standardgamble

withvisualprop

Sixpluscurrenthealth

Tabular

0.68

0.69

0.35

0.48

0.48

*Includesfirstandsubsequentyearsposttransplantcombined.

† Varicealbleedingwasaseparatestateandhadameanutilityof0.40.

‡ Standarddeviation.

Health State Utilities for Hepatitis B 535

Our study had limitations. First, because recruit-ment was undertaken at sites located in at most twocities per country, respondents may not be representa-tive of the respective country’s uninfected or infectedpopulations. This could have an effect on the general-izability of the results. Second, the standard gambleemployed did not allow respondents to specify negativevalues when a state was considered worse than death.Rather, we assigned a value of -0.1 to the states whenthis situation occurred. To investigate the impact, weconducted a sensitivity analysis and found that ourassumption had only a minor effect on the mean utilityestimates and no effect on the rankings. Third, humanfactors derived from interviewer-administered mea-sures rather than paper- or computer-based methodsmay have introduced a bias of unknown direction. Forpractical reasons, and because computer-based utilityelicitation interviews have other limitations includingthe need for specialized software and respondents’potential lack of familiarity with using a computer,we chose interviewer-administered methods. We usedextensive standardization procedures during the datacollection process to mitigate the possibility of inter-viewer bias. Fourth, for logistic reasons, uninfectedrespondents were primarily recruited from local univer-sities. Utilities elicited from students may be differentthan those from other representatives of society[33,34]. In the current study, we conducted a sensitivityanalysis by comparing utilities among uninfectedpersons categorizing those aged less than 25 years andthose aged 25 years or older. That analysis indicated nosignificant difference in utilities between the two agegroups. Fifth, the ratings were almost certainly influ-enced by the specific wording. Did respondents note thedifference between “occasionally” (in CHB) and “peri-odically” (in compensated cirrhosis) with respect tophysician visits and between “. . . feel anxious” and“. . . somewhat worried and anxious”? There are twostatements about the future in CHB and only one incompensated cirrhosis, which may explain why theformer was ranked lower. To our knowledge, only oneother study has published health state descriptions forliver disease [72] and that was published after the datacollection in the current study was underway.

This study represents the first in which health stateutilities for CHB-related health states were elicitedfrom both infected patients and uninfected lay persons.The populations that were included reflect the diversenature of the populations afflicted by CHB. We foundthat health states related to CHB infection were asso-ciated with substantial reductions in HRQOL in allcountries. The differences across countries in utilitiesfor the same health state underscore the need forcountry-specific adaptations of analyses assessing thevalue and cost-utility of new treatment options inCHB. The health state utilities reported in this studyprovide valuable information for comparing treatment

options for CHB therapies. Integrating these utilitiesinto cost-utility analyses will improve the quality andapplicability of economic evaluations and decision-making in CHB. The extent that systematic intercoun-try differences in utilities holds true for other infectiousand chronic diseases remains an open question and hasconsiderable implications for the proper conduct andinterpretation of economic evaluations.

The authors gratefully acknowledge the contributions of: thecountry-specific investigators Lise Poissant (Canada), FrankAnderson (Canada), Natalie Rock (Canada), JiQian Fang(mainland China), Cindy Lam (Hong Kong), Stephen J.Coons (United States), and Michael Herdman (Spain); thestudy personnel: Yves Gagnon, Sarah Hargreaves, AndrewMachuk, Greta Lozano Ortega, Diego Ossa, and HongWang; and Karissa Johnston and Natasha LaPierre. Allauthors contributed to the interpretation of the results anddevelopment of the manuscript, and approved the finalversion of the text. ARL is the study’s guarantor and acceptsfull responsibility for the conduct of the study, had full accessto the data and controlled the decision to publish.

Source of financial support: Funding for this study was pro-vided by a grant from the Pharmaceutical Research Institute,Bristol-Myers Squibb Company KVK is supported by DK02957.

Supplementary materials for this article can be found at:http://www.ispor.org/publications/value/ViHsupplementary.asp

References

1 von Neumann J, Morgenstern O. Theory of Gamesand Economic Behavior. Princeton, NJ: PrincetonUniversity Press, 1947.

2 Torrance GW. Measurement of health state utilitiesfor economic appraisal. J Health Econ 1986;5:1–30.

3 Testa MA, Nackley JF. Methods for quality-of-lifestudies. Annu Rev Public Health 1994;15:535–59.

4 Schackman BR, Gold HT, Stone PW, Neumann PJ.How often do sensitivity analyses for economicparameters change cost-utility analysis conclusions?Pharmacoeconomics 2004;22:293–300.

5 Souchek J, Byrne MM, Kelly PA, et al. Valuation ofarthritis health states across ethnic groups andbetween patients and community members. Med Care2005;43:921–8.

6 Custer B, Sullivan SD, Hazlet TK, et al. Global epide-miology of hepatitis B virus. J Clin Gastroenterol2004;38(Suppl.):S158–68.

7 Lee WM. Hepatitis B virus infection. N Engl J Med1997;337:1733–45.

8 Fattovich G. Natural history of hepatitis B. J Hepatol2003;39(Suppl. 1):S50–8.

9 Chen CJ, Yang HI, Su J, et al. Elevated serum level ofhepatitis B virus DNA is an independent risk factorfor hepatocellular carcinoma: a long-term follow-upstudy in Taiwan. J Hepatol 2005;42(Suppl. 2):S16–17.

536 Levy et al.

10 Chen G, Lin WY, Shen FM, et al. Viral load as apredictor of mortality from hepatocellular carcinomaand chronic liver disease in chronic hepatitis B infec-tion. J Hepatol 2005;42(Suppl. 2):S173.

11 Iloeje UH, Yang HI, Su J, et al. Viral load not serumalanine aminotransferase (ALT) is the primary predic-tor of progression to cirrhosis in persons chronicallyinfected with hepatitis B virus. Gastroenterology2005;128:A–740.

12 Iloeje UH, Yang HI, Su J, et al. Predicting cirrhosisrisk based on the level of circulating hepatitis B viralload. Gastroenterology 2006;130:678–86.

13 Fattovich G, Stroffolini T, Zagni I, Donato F. Hepa-tocellular carcinoma in cirrhosis: incidence and riskfactors. Gastroenterology 2004;127(Suppl.):S35–50.

14 Ganem D, Prince AM. Hepatitis B virus infection––natural history and clinical consequences. N Engl JMed 2004;350:1118–29.

15 Brown RE, De CE, Colin X, et al. Hepatitis B man-agement costs in France, Italy, Spain, and the UnitedKingdom. J Clin Gastroenterol 2004;38(Suppl.):S169–74.

16 Butler JR, Pianko S, Korda RJ, et al. The direct cost ofmanaging patients with chronic hepatitis B infectionin Australia. J Clin Gastroenterol 2004;38(Suppl.):S187–92.

17 Gagnon YM, Levy AR, Iloeje UH, Briggs AH. Treat-ment costs in Canada of health conditions resultingfrom chronic hepatitis B infection. J Clin Gastroen-terol 2004;38(Suppl.):S179–86.

18 Hsieh CR, Kuo CW. Cost of chronic hepatitis B virusinfection in Taiwan. J Clin Gastroenterol 2004;38(Suppl.):S148–52.

19 Kowdley KV. The cost of managing chronic hepatitisB infection: a global perspective. J Clin Gastroenterol2004;38(Suppl.):S132–3.

20 Lee TA, Veenstra DL, Iloeje UH, Sullivan SD. Cost ofchronic hepatitis B infection in the United States.J Clin Gastroenterol 2004;38(Suppl.):S144–7.

21 Li SC, Ong SC, Lim SG, et al. A cost comparison ofmanagement of chronic hepatitis B and its associatedcomplications in Hong Kong and Singapore. J ClinGastroenterol 2004;38(Suppl.):S136–43.

22 Yang BM, Kim CH, Kim JY. Cost of chronic hepatitisB infection in South Korea. J Clin Gastroenterol 2004;38(Suppl.):S153–7.

23 Zhiqiang G, Zhaohui D, Qinhuan W, et al. Cost ofchronic hepatitis B infection in China. J Clin Gastro-enterol 2004;38(Suppl.):S175–8.

24 Mueller AR, Platz KP, Kremer B. Early postoperativecomplications following liver transplantation. BestPract Res Clin Gastroenterol 2004;18:881–900.

25 Brazier J, Green C, McCabe C, Stevens K. Use of visualanalog scales in economic evaluation. Expert Rev Phar-macoeconomics Outcomes Res 2003;3:293–302.

26 Torrance GW, Furlong W, Feeny D. Health utilityestimation. Exp Rev Pharmacoeconom Res 2002;2:99–108.

27 Kanwal F, Gralnek IM, Martin P, et al. Treatmentalternatives for chronic hepatitis B virus infection: acost-effectiveness analysis. Ann Intern Med 2005;142:821–31.

28 Wong JB, Koff RS, Tine F, Pauker SG. Cost-effectiveness of interferon-alpha 2b treatment forhepatitis B e antigen-positive chronic hepatitis B. AnnIntern Med 1995;122:664–75.

29 Arevalo JA, Washington AE. Cost-effectiveness of pre-natal screening and immunization for hepatitis Bvirus. JAMA 1988;259:365–9.

30 Bloom BS, Hillman AL, Fendrick AM, Schwartz JS. Areappraisal of hepatitis B virus vaccination strategiesusing cost-effectiveness analysis. Ann Intern Med1993;118:298–306.

31 Krahn M, Detsky AS. Should Canada and the UnitedStates universally vaccinate infants against hepatitisB? A cost-effectiveness analysis. Med Decis Making1993;13:4–20.

32 Chong CA, Gulamhussein A, Heathcote EJ, et al.Health-state utilities and quality of life in hepatitis Cpatients. Am J Gastroenterol 2003;98:630–8.

33 Boyd NF, Sutherland HJ, Heasman KZ, et al. Whoseutilities for decision analysis? Med Decis Making1990;10:58–67.

34 De Wit GA, Busschbach JJV, De Charro FT. Sensitiv-ity and perspective in the valuation of health status:whose values count? Health Econ 2000;9:109–26.

35 Dolan P. Whose preferences count? Med DecisMaking 1999;19:482–6.

36 Janse AJ, Gemke RJ, Uiterwaal CS, et al. Quality oflife: patients and doctors don’t always agree: a meta-analysis. J Clin Epidemiol 2004;57:653–61.

37 Gold MR, Siegel JE, Russel LB, Weinstein MC. Cost-Effectiveness in Health and Medicine. New York:Oxford University Press, 1996.

38 Bennet KJ, Torrance GW. Measuring health statepreferences and utilities: rating scale, time trade-offand standard gamble techniques. In: Spilker B, ed.Quality of Life and Pharmacoeconomics in ClinicalTrials. Philadelphia, PA: Lippincott-Raven, 1996.

39 Furlong W, Feeny D, Torrance GW, et al. Guide toDesign and Development of Health State UtilityInstrumentation. Working Paper 90-9. Hamilton,ON: McMaster University Centre for Health Eco-nomics and Policy Analysis, 1990.

40 Llewellyn-Thomas H, Sutherland HJ, Tibshirani R,et al. Describing health states. Methodologic issues inobtaining values for health states. Med Care 1984;22:543–52.

41 Schunemann HJ, Stahl E, Austin P, et al. A compari-son of narrative and table formats for presentinghypothetical health states to patients with gastrointes-tinal or pulmonary disease. Med Decis Making2004;24:53–60.

42 Gralnek IM, Hays RD, Kilbourne A, et al. Develop-ment and evaluation of the Liver Disease Quality ofLife instrument in persons with advanced, chronicliver disease––the LDQOL 1.0. Am J Gastroenterol2000;95:3552–65.

43 Torrance GW, Feeny D, Furlong W. Visual analogscales: do they have a role in the measurement ofpreferences for health states? Med Decis Making2001;21:329–34.

44 Hoffmann JP. Generalized Linear Models: an AppliedApproach. Boston, MA: Pearson Education Inc, 2004.

Health State Utilities for Hepatitis B 537

45 Patrick DL, Starks HE, Cain KC, et al. Measuringpreferences for health states worse than death. MedDecis Making 1994;14:9–18.

46 Lok AS, McMahon BJ. Chronic hepatitis B. Hepatol-ogy 2001;34:1225–41.

47 World Health Organization. Hepatitis B. World HealthOrganization Fact Sheet 204. 2000. Available form:http://www.who.int/mediacentre/factsheets/fs204/en/[Accessed July 23, 2007].

48 Pramoolsinsup C. Management of viral hepatitis B.J Gastroenterol Hepatol 2002;17(Suppl.):S125–45.

49 Mulley AG Jr. Assessing patients’ utilities. Can theends justify the means? Med Care 1989;27(Suppl.):S269–81.

50 Mittmann N, Trakas K, Risebrough N, Liu BA. Utilityscores for chronic conditions in a community-dwelling population. Pharmacoeconomics 1999;15:369–76.

51 Hawthorne G, Osborne R. Population norms andmeaningful differences for the Assessment of Qualityof Life (AQoL) measure. Aust N Z J Public Health2005;29:136–42.

52 Albrecht GL, Devlieger PJ. The disability paradox:high quality of life against all odds. Soc Sci Med1999;48:977–88.

53 Sherbourne CD, Keeler E, Unutzer J, et al. Relation-ship between age and patients’ current health statepreferences. Gerontologist 1999;39:271–8.

54 Levine S, Feldman JJ, Elinson J. Does medical care doany good? In: Mechanic D, ed. Handbook of Health,Health Care and the Health Professions. New York:The Free Press 1983.

55 Amick BC, Levine S, Tarlov AR, Walsh DC (eds).Society and Health. New York: Oxford UniversityPress, 1994.

56 Schwartz CE, Andresen EM, Nosek MA, Krahn GL.Response shift theory: important implications formeasuring quality of life in people with disability.Arch Phys Med Rehabil 2007;88:529–36.

57 Gabriel SE, Kneeland TS, Melton LJ III, et al. Health-related quality of life in economic evaluations forosteoporosis: whose values should we use? Med DecisMaking 1999;19:141–8.

58 Postulart D, Adang EM. Response shift and adapta-tion in chronically ill patients. Med Decis Making2000;20:186–93.

59 Sackett DL, Torrance GW. The utility of differenthealth states as perceived by the general public.J Chronic Dis 1978;31:697–704.

60 Tsevat J, Cook EF, Green ML, et al. Health values ofthe seriously ill. SUPPORT investigators. Ann InternMed 1995;122:514–20.

61 Balaban DJ, Sagi PC, Goldfarb NI, Nettler S. Weightsfor scoring the quality of well-being instrument among

rheumatoid arthritics. A comparison to general popu-lation weights. Med Care 1986;24:973–80.

62 Jenkinson C, Gray A, Doll H, et al. Evaluation ofindex and profile measures of health status in a ran-domized controlled trial. Comparison of the MedicalOutcomes Study 36-Item Short Form Health Survey,EuroQol, and disease specific measures. Med Care1997;35:1109–18.

63 Dolders MG, Zeegers MP, Groot W, Ament A. Ameta-analysis demonstrates no significant differencesbetween patient and population preferences. J ClinEpidemiol 2006;59:653–64.

64 Dolan P. Effect of age on health state valuations. JHealth Serv Res Policy 2000;5:17–21.

65 B is for bigotry. The Economist, November 16, 2006.66 Murray CJ, Chen LC. Understanding morbidity

change. Popul Dev Rev 1992;18:481–503.67 Torrance GW. Social preferences for health states: an

empirical evaluation of three measurement tech-niques. Socioecon Plann Sci 1976;10:128–36.

68 Kaplan RM, Bush JW, Berry CC. Health status index:category rating versus magnitude estimation for mea-suring levels of well-being. Med Care 1979;17:501–25.

69 Read JL, Quinn RJ, Berwick DM, et al. Preferencesfor health outcomes. Comparison of assessmentmethods. Med Decis Making 1984;4:315–29.

70 Haig TH, Scott DA, Wickett LI. The rational zeropoint for an illness index with ratio properties. MedCare 1986;24:113–24.

71 Tengs TO, Wallace A. One thousand health-relatedquality-of-life estimates. Med Care 2000;38:583–637.

72 Wells CD, Murrill WB, Arguedas MR. Comparison ofhealth-related quality of life preferences between phy-sicians and cirrhotic patients: implications for cost-utility analyses in chronic liver disease. Dig Dis Sci2004;49:453–8.

73 Bonsel GJ, Klompmaker IJ, Essink-Bot ML, et al.Cost-effectiveness analysis of the Dutch liver trans-plantation programme. Transplant Proc 1990;22:1481–4.

74 Kim WR, Poterucha JJ, Hermans JE, et al. Cost-effectiveness of 6 and 12 months of interferon-alphatherapy for chronic hepatitis C. Ann Intern Med1997;127:866–74.

75 Teran JC, Imperiale TF, Mullen KD, et al. Primaryprophylaxis of variceal bleeding in cirrhosis: a cost-effectiveness analysis. Gastroenterology 1997;112:473–82.

76 Owens DK, Cardinalli AB, Nease RF Jr. Physicians’assessments of the utility of health states associatedwith human immunodeficiency virus (HIV) and hepa-titis B virus (HBV) infection. Qual Life Res 1997;6:77–86.

538 Levy et al.