Despite the increasing realization that health-related quality of life (HRQOL) is an important outcome in chronic HBV infection, there are no validated, disease-targeted instruments currently available. We sought to develop and validate the first disease-targeted HRQOL instrument in noncirrhotic HBV: the Hepatitis B Quality of Life instrument, version 1.0 (HBQOL v1.0). We established content validity for the HBQOL v1.0 by conducting a systematic literature review, an expert focus group, and cognitive interviews with HBV patients. We administered the resultant questionnaire to 138 HBV patients. We used factor analysis to test hypotheses regarding HRQOL domains and measured construct validity by comparing HBQOL v1.0 scores across several anchors, including viral response to treatment, SF-36 scores, and global health. Finally, we measured test–retest and internal consistency reliability. Content validation revealed that HBV affects multiple aspects of psychological, social, and physical health. The resultant questionnaire summarized this HRQOL impact with 31 items across six subscales: psychological well-being, anticipation anxiety, vitality, disease stigma, vulnerability, and transmissibility. Internal consistency and test–retest reliability were excellent. The HBQOL v1.0 discriminated between viral responders versus nonresponders and correlated highly with SF-36 scores and global health. Conclusion: Patients with chronic HBV infection attribute a wide range of negative psychological, social, and physical symptoms to their condition, even in the absence of cirrhosis or cancer. The HBQOL v1.0 is a valid and reliable measure that captures this HRQOL decrement. This instrument may be useful in everyday clinical practice and in future clinical trials. (HEPATOLOGY 2007;46:113–121.)
Chronic HBV infection is a prevalent and expensive condition affecting 350 million people worldwide and 1.25 million in the United States1 at a cost of over $700 million annually.2 HBV leads to cirrhosis in up to 20% of those chronically infected and is a one of the most common indications for liver transplantation worldwide.1 This economic burden is compounded by the significant impact of HBV on health-related quality of life (HRQOL) resulting from complications of advanced liver disease, such as encephalopathy, variceal hemorrhage, ascites, and liver transplantation. However, these end-stage complications are relatively rare compared with the vast majority of patients with HBV in the absence of clinically significant liver disease. Despite the widely held consensus that this majority of patients has asymptomatic seropositivity, data indicate that patients with HBV have lower HRQOL scores than control subjects in the general population.3 Thus, there is reason to believe that HBV may itself diminish HRQOL, even in the absence of advanced liver disease or hepatocellular cancer.
There is currently a disconnect between the increasing realization that HRQOL is an important outcome in chronic diseases such as HBV and the scarcity of disease-targeted instruments to measure HRQOL in chronic HBV. In the absence of a validated, disease-targeted measure of HRQOL in HBV, investigators and clinicians must rely on either generic HRQOL measurements or traditional biological outcomes to assess their patients' health status. However, generic HRQOL measurements may fail to capture the most important components of HRQOL in HBV, and traditional biological outcomes fail to acknowledge the patients' perception of their disease. Therefore, the current menu of available patient-oriented outcomes in HBV is limited.
In light of this shortcoming, it is important to develop and validate a disease-targeted, multidimensional, HRQOL instrument in HBV. The availability of such an instrument may allow physicians to better monitor patient outcomes in clinical practice, provide researchers with a novel outcome measure for clinical trials, and equip patients with the knowledge to better select between competing disease management strategies.4 We therefore sought to develop and validate the first disease-targeted HRQOL instrument in chronic HBV: the Hepatitis B Quality of Life instrument, version 1.0 (HBQOL v1.0). Before beginning this process, we established 3 a priori hypotheses about HRQOL in HBV: (1) HBV negatively impacts HRQOL across multiple dimensions of health, including psychological, social, and physical functioning; (2) the biopsychosocial decrements in HRQOL can be measured using a disease-targeted, multidimensional, HRQOL instrument; and (3) this disease-targeted instrument can discriminate between treatment responders and nonresponders.
Patients and Methods
Content validity refers to the degree to which an instrument contains a comprehensive range of items with relevance to the disease under study.5 If an instrument fails to measure attributes of health that are important to patients with the targeted disease, it will likely fail to accurately measure the respondents' HRQOL. Although there is no gold standard to measure content validity, it may be subjectively evaluated by comparing the instrument to existing validated instruments, eliciting expert opinion regarding the item content, and pilot-testing the questionnaire in patients with the target disease.5 We therefore conducted a four-step procedure to develop the content of the HBQOL v1.0.
Step 1: Systematic Literature Review.
To identify candidate items for the HBQOL v1.0, we conducted a structured search of the MEDLINE database from 1970 to 2005 for previously published HRQOL instruments in both chronic viral hepatitis and other forms of chronic liver disease. We then compiled a pool of items that had potential relevance to HBV based on a priori hypotheses and clinical judgment.
Step 2: Item Generation by Physician Expert Panel.
Although HRQOL is a patient-reported outcome, it is nonetheless useful to include the physicians' perspective in developing the content of a new HRQOL instrument. In theory, clinically adept and observant physicians can effectively represent the viewpoints collected from thousands of patient encounters. Thus, including physicians in the content validation process is akin to convening a focus group of several thousand patients (although it is not a substitute for patient focus groups). We therefore convened a panel of 5 hepatologists with experience in the diagnosis and treatment of chronic HBV infection. We selected the participants on the basis of their clinical experience in HBV, publication records in HBV, and membership status on national hepatology societies. Using a semistructured protocol, a trained moderator elicited the domains perceived by the panel to be most relevant in HBV. Following this first step, the moderator presented an overview of the items identified by the systematic literature review and asked the panel to comment on the content, breadth, and relevancy of the pool of potentially relevant items. The panel was further instructed to recommend changes to the list as deemed necessary and was encouraged to contribute additional relevant items that were missing from the list.
Step 3: Item Generation by HBV Patients.
Because HRQOL is ultimately a reflection of patient perception and experience, the validity of an HRQOL instrument depends heavily on the endorsement of its content by patients with the disease in question. We therefore conducted two types of patient elicitations: intensive semistructured individual cognitive interviews of a cohort of 9 HBV patients and an online anonymous “virtual focus group” of 50 HBV patients. The UCLA Institutional Review Board approved all cognitive and online patient elicitations. Patients were eligible to participate if they had evidence of current or past infection with HBV, provided there was no clinical, biochemical, or radiological evidence of cirrhosis, hepatocellular cancer, or advanced liver disease. We explicitly recruited patients from all stages of precirrhotic chronic HBV, including inactive hepatitis B surface antigen carriers (i.e., patients without evidence of ongoing hepatic necroinflammation), chronic hepatitis B (with hepatic necroinflammation), and all stages of treatment (i.e. pre-, intra-, and post-antiviral treatment). We recruited patients for the cognitive interviews from the Phleger Liver Institute at UCLA and recruited patients for the online “virtual focus group” from a network of 8 geographically and demographically diverse centers throughout the United States, including 2 university hospital practices, 4 community hospital practices, and 2 managed care practices. By recruiting across a broad range of practice settings, we sought to maximize the diversity of patient characteristics and, subsequently, the generalizability of the instrument. We compiled the results of all cognitive interviews and online elicitations into a written report and qualitatively analyzed the results to identify domains and items of relevance to patients.
Step 4: A Priori Scale Development.
Based on a priori hypotheses, the systematic literature review, expert panel input, and patient focus group input, we developed a path diagram6 to depict the latent and manifest variables that describe the hypothesized scales (and related items) of the HBQOL v1.0. We subsequently tested our a priori hypotheses against factor analysis of patient response data, as described below.
Instrumentation and Field Testing
Based on the content validation phase, we created an initial questionnaire with 57 items, each accompanied by a 5-point Likert scale. This “long form” of the HBQOL v1.0 was subsequently administered, along with the SF-36 Health Survey and other accompanying questions, to a cohort of 138 patients using an online questionnaire platform. Subjects were eligible to participate if they were over 18 years of age, had chronic HBV infection, and had no clinical evidence of cirrhosis, malignancy, or advanced liver disease. We purposely recruited a mixed population of patients at various stages of treatment, including pre-, intra-, and posttreatment. In addition, we included patients with a previous viral response and patients without a viral response post-antiviral therapy. The cohort was recruited from the same network of eight centers as employed for the online virtual focus group. We invited all participants to complete the HBQOLv1.0 8 months after the initial administration for purposes of longitudinal validation. In the follow-up group, we selected a random sample of 20 patients to complete the survey using a paper form, while the remainder completed the online form. We subsequently compared responses between survey formats to evaluate for systematic differences.
To ensure that the final instrument was parsimonious, we aimed to remove redundant, low-impact, and extraneous items. Each item received an “item performance score” that accounted for seven key psychometric properties, including: (1) floor effect (e.g., proportion of “1” responses on the 5-point Likert scale, in which a lower proportion is preferred); (2) variability (measured via SD, in which a wider SD is preferred); (3) internal consistency (measured via the mean correlation coefficient of each item versus all other items in the instrument, in which a higher coefficient is preferred); (4) concurrent validity with SF-36 mental component score (MCS); (5) concurrent validity with SF-36 physical component score (PCS); (6) ability to discriminate between treated versus untreated patients; and (7) ability to discriminate between viral responders versus nonresponders. We subsequently eliminated items that performed poorly across these psychometric characteristics and carried forward the remaining items to comprise the “short form” of the HBQOL v1.0.
To evaluate the dimensionality of the HRQOL scales, we conducted exploratory factor analysis (using SAS v8.0) to measure the quantitative structure of the HBQOL v1.0, and to compare that structure with our a priori hypothesized path diagram. To accomplish this, we performed a maximum likelihood factor analysis with varimax rotation to confirm that the hypothesized number of scales was present and to confirm that the hypothesized grouping of the items within each scale was correct.6 We used the Tucker-Lewis coefficient to evaluate the overall dimensionality of the factor model, attempting to achieve a final model with a coefficient of 0.85 or higher. We then identified items that loaded <0.4 onto more than three scales, and items that failed to load convincingly onto any one scale.6 We removed items that met either of these criteria. Based on these quantitative results, and guided by our qualitative hypotheses, we created a final set of subscales and item groupings by subscale. In instances where our a priori hypothesis was contradicted by the factor loading for an item, we categorized the item according to our initial hypothesis, so long as the item loaded above 0.4 in the targeted scale, and the targeted scale was no lower than second in the rank of scale loadings for the item in question. Finally, we assessed the intercorrelation of the subscales to determine whether a single, higher-order HBQOL scale was empirically supported. We linearly transformed all scale scores and subscale scores along a 100-point range using the Likert method of summation, where higher scores denote better HRQOL. The Likert method does not assign weights to the individual items within a scale, but instead assumed equal weighting for all items.6, 7
Reliability and validity are the two fundamental psychometric properties of an HRQOL instrument. For an instrument to be reliable, it must generate the same results when used sequentially under the same circumstances.5 One measure of reliability is “test–retest reliability,” in which the results obtained from a respondent are compared on 2 separate occasions during a period when clinical status remains stable.5 To measure test–retest reliability, we administered the HBQOL v1.0 2 weeks after the initial administration in a random sample of 15 patients from our full cohort and calculated intraclass correlation coefficients between individual scores at baseline and 2 weeks later. In addition to test–retest reliability, we measured the internal consistency reliability of the HBQOL v1.0 scales by calculating a Cronbach's α statistic for each scale.5, 7 We considered a Cronbach's α ≥0.70 to be adequate evidence for internal consistency reliability.5, 7
Construct validity is the degree to which an instrument “behaves” as predicted when compared with existing patient-oriented and validated outcome measures.5, 7 To establish construct validity, we compared the HBQOL v1.0 scores to the following concurrently measured outcomes:
Viral Response to Treatment.
The traditional outcome measure of interest in chronic HBV is viral response.1 We therefore sought to determine whether a “viral response” subscale of the HBQOL v1.0 could accurately discriminate between patients with versus without a viral response to treatment. To create this subscale, we first hypothesized that patients with a viral response might be less concerned about potential transmissibility and thus might have a lower sense of disease vulnerability compared with patients without a viral response. Based on these a priori hypotheses, and guided by the results of individual item psychometrics, we created a “viral response” subscale and compared the scale scores between patients with versus without a previous viral response to treatment. We hypothesized that the cross-sectional differences in “viral response” subscale scores between patient groups would be statistically significant (P < 0.05 for t test) and clinically relevant (effect size of >0.5 SD, a “moderate” effect size8). We therefore powered our cross-sectional validation sample to demonstrate a 0.5-SD difference in subscale between patients with a viral response versus those without viral response. Using a two-tailed 5% significance level with a power of 80%, we calculated that a minimum of 64 patients would be required per group to demonstrate a 0.5-SD difference in scores.
SF-36 Health Survey.
The SF-36 is a widely used and well-accepted generic HRQOL instrument composed of 36 self-reported items.9 The instrument has been tested in many patient populations and medical conditions, including chronic viral hepatitis, and previous comparative data indicate that patients with HBV score significantly lower on the SF-36 than controls without HBV.3 The 36 items are organized into 8 discrete scales that are compiled into 2 summary scores, each scored from 0 to 100 (100 = best HRQOL): PCS and MCS. We hypothesized that there would be a moderate positive correlation (r ≥ 0.3 and r ≤ 0.6) between HBQOL v1.0 scores and SF-36 MCS and PCS scores.
Patient Global Health Assessment.
Patient-perceived global health is widely regarded to be an important outcome in chronic diseases. We therefore measured global health assessment using the 5-point global health item of the SF-36.9 We anticipated a moderate (r ≥ 0.4) correlation between the HBQOL v1.0 and global heath assessments.
Responsiveness is the degree to which an instrument is able to detect clinically meaningful change in an individual's health over time and is an important aspect of validity. We defined patients as experiencing a clinically meaningful difference if they achieved a viral response between the baseline and follow-up HBQOL v1.0 administrations. To establish responsiveness of HBQOL v1.0, we compared the score changes for patients who achieved a viral response (i.e., changed group) versus those who did not (i.e., stable group). We assessed responsiveness to viral response using 3 standardized measures: the effect size, standardized response means, and responsiveness statistics.10 Results from all 3 indices are generally comparable for a given criterion of change. However, previous data indicate that the responsiveness statistics may yield more robust results, especially if multiple external criteria are used to define change in health. Thereore, we relied most heavily on the responsiveness statistics while interpreting results of our longitudinal analyses. We also determined correlation coefficients between HBQOL v1.0 score changes and changes in SF-36 PCS, MCS, and patients' ratings of global health scores between the baseline and follow-up administrations.
Based on our systematic literature review and expert panel feedback, we identified four preliminary domains of relevance to chronic HBV, including psychological well-being, sexual well-being/intimacy, disease stigma/social well-being, and daily functioning. We created a core set of 30 items resulting from the systematic review and expert panel and subsequently conducted online and in-person elicitations in chronic HBV patients. Table 1 describes the sample characteristics of these patients. Supplementary Appendix A provides detailed qualitative responses from the patient interviews. Based on the extensive feedback from these interviews, we expanded our set of 30 items to 57 and grouped these items within an expanded set of identified domains, including psychological well-being (e.g., HBV-related isolation, sadness, anger), anticipation anxiety (e.g., fear of developing cancer, cirrhosis, or disease flares), vitality (e.g., feeling tired, worn out, or unproductive), disease stigma (e.g., embarrassment, concern someone influential may find out, shame), vulnerability (e.g., concern of getting sick easily, watching food carefully), and transmissibility (e.g., concern about transmitting through sex, concerns about transmitting to a child). The 57-item “long form” of the HBQOL v1.0 was subsequently field-tested in a geographically diverse sample of patients with chronic HBV.
Table 1. Sample Characteristics
|Practice type characteristics|| || || |
| University practice||60||44%||4.2%|
| Community practice||68||49%||4.3%|
| Managed care practice||10||7%||2.1%|
|Demographic characteristics|| || || |
| Age (mean years)||138||42.4||12.3|
| Male sex||89||65%||4.0%|
| Ethnicity|| || || |
| Asian/Pacific Islander||108||78.2%||3.4%|
| Education|| || || |
| Non–high school graduate||6||4.3%||1.7%|
| High school graduate||33||23.9%||3.6%|
| College graduate||50||36.2%||4.1%|
| Postgraduate work||49||35.6%||4.1%|
| Marital status|| || || |
| Never married||38||27.5%||3.8%|
| Annual household income|| || || |
|Disease-specific characteristics|| || || |
| Number of family members with hepatitis B|| || || |
| 1 member||24||17.4%||3.2%|
| 2 members||19||13.8%||2.9%|
| 3 members||16||11.6%||2.7%|
| 4 or more members||17||12.3%||2.8%|
| Received previous treatment for hepatitis B||91||65.9%||4.0%|
| Achieved viral response (among those previously treated)||38||41.8%||5.2%|
| Currently receiving treatment for hepatitis B||55||40%||4.2%|
| Perceived knowledge level about hepatitis B (10-point scale)||138||7.1||1.9|
|Health status characteristics|| || || |
| SF-36 mental component score||138||48.8||11.4|
| SF-36 physical component score||138||51.6||8.2|
| General health (“very good” or “excellent”)||70||52.6%||5.9%|
A total of 138 patients (65% male) ranging in age from 19 to 73 years (mean ± SD, 42.4 ± 12.3) completed the HBQOL v1.0 online questionnaire. This sample size exceeded our a priori power calculation. Seventy-eight percent of the cohort was of Asian descent, and 48% had been diagnosed with HBV for at least 10 years before completing the questionnaire. Sixty-eight percent of the cohort had been previously treated for HBV, and 42% of this subgroup had achieved a previous sustained viral response to HBV. Table 1 provides additional information about the validation cohort.
Item Selection and Factor Analysis.
After subjecting each item to our seven prespecified psychometric criteria, we identified 31 items to retain in the final HBQOL v1.0. We subjected the final item set to factor analysis to assess construct internal consistency, reliability, and validity. Factor analysis identified a substructure of six underlying domains for the HBQOL v1.0. Table 2 displays the full six-factor solution for the orthogonally rotated factor pattern. These factors aligned closely with our a priori hypothesized domains, as outlined in the section on Content Validation. The factors jointly accounted for 73% of the item-level variability, and the Tucker-Lewis reliability coefficient was 0.89. Supplementary Appendix B provides the final 31-item HBQOL v1.0 and describes how these items are grouped by domain. Intercorrelation of the subscales supported the validity of a single, higher-order HRQOL scale. Thus, the entire 31-item questionnaire yielded one “overall score” for the HBQOL v1.0, in addition to subscale scores for each of the six underlying health factors. Table 3 displays the number of items per scale and the mean scale scores from our validation cohort.
Table 2. Six-Factor Solution from Factor Analysis
|F5||Worn out and tired||0.23||0.12||0.83||0.12||0.05||0.15|
|F8||Isolated from others||0.47||0.27||0.20||0.37||0.28||0.04|
|F9||Something bad might happen||0.53||0.48||0.27||0.20||−0.04||0.08|
|F10||Life less enjoyable||0.55||0.17||0.50||0.19||0.18||0.24|
|F11||Sexual activity difficult||0.38||0.05||0.23||0.28||0.43||0.22|
|C1||Concern: liver failure||0.17||0.87||0.16||0.14||0.25||−0.01|
|C2||Concern: liver cancer||0.19||0.81||0.28||0.11||0.18||0.03|
|C3||Concern: someone influential||0.18||0.33||0.06||0.55||0.20||0.22|
|C4||Concern: transmit to child||0.15||0.40||0.10||0.25||0.54||0.07|
|C6||Concern: easier to get ill||0.15||0.48||0.31||0.08||0.17||0.40|
|C7||Concern: transmit to partner||0.12||0.29||0.00||0.18||0.78||0.14|
|C8||Concern: watch medicines||0.19||0.37||0.34||0.21||0.32||0.43|
|C9||Concern: life expectancy||0.26||0.59||0.19||0.31||0.13||0.15|
|C10||Concern: overly self-conscious||0.21||0.31||0.34||0.62||0.04||0.33|
|C11||Concern: socially isolated||0.19||0.29||0.21||0.65||0.17||0.26|
|C12||Concern: something serious wrong||0.28||0.62||0.24||0.22||0.15||0.31|
|C13||Concern: watch what eat||0.24||0.33||0.25||0.10||0.17||0.64|
|C15||Concern: health worsen||0.13||0.70||0.38||0.15||0.13||0.26|
Table 3. Number of Items, Mean Scores, and Internal Consistency/Reliability of the HBQOL v1.0 Scales
|Overall score||31||62.9 (19.5)||0.96|
|Psychological well-being||8||69.0 (20.1)||0.90|
|Anticipation anxiety||6||40.9 (22.0)||0.88|
|Viral response||4||52.1 (26.6)||0.75|
The HBQOL v1.0 demonstrated a high level of internal consistency, both within each subscale, and within the overall higher-order scale. Table 3 provides the Cronbach's α for each domain. All Cronbach's α values exceeded our a priori threshold of ≥0.7, indicating that each subscale performed well together as a composite measure. Test–retest reliability of the overall score, as assessed with the intraclass correlation coefficient, was 0.96 for respondents who reported no change in their health status between administrations (all 15 patients in test–retest sample reported no change). The intraclass correlation coefficient ranged from 0.75 (viral response scale) to 0.98 (stigmatization scale).
There were no systematic differences in responses between patients completing online versus paper versions of the survey, both in univariate and multivariable analyses adjusting for patient characteristics.
Table 4 presents the bivariate relationships between the HBQOL v1.0 scale scores with each of the construct validity anchors, including the SF-36 MCS and PCS, global health assessment, and viral response status. The HBQOL v1.0 overall score demonstrated highly statistically significant correlations with the SF-36 MCS (r = 0.49; P < 0.001) and PCS (r = 0.37; P < 0.001), and with global health assessment (r = 0.55; P < 0.001). The individual scale scores also demonstrated moderate correlations with these construct anchors, and most of the relationships were highly statistically significant. The four-item “viral response” scale yielded significant score differences between patients with versus patients without a viral response (P = 0.03; effect size = 0.43), and thus was able to discriminate between the two groups. Using the data from the follow-up HBQOL v1.0 administration, we repeated the bivariate associations between HBQOL v1.0 scale scores with the previously described anchors. The direction and magnitude of cross-sectional relationships between HRQOL and relevant anchors were consistent both with our a priori hypotheses and with the results from baseline administration (Table 4).
Table 4. Construct Validity Results
|Overall score||0.49 (<0.001)||0.55 (<0.001)||0.37 (<0.001)||0.30 (0.05)||0.55 (<0.001)||0.53 (<0.001)||–|
|Psychological well-being||0.43 (<0.001)||0.54 (<0.001)||0.27 (0.009)||0.30 (0.06)||0.43 (<0.001)||0.51 (<0.001)||–|
|Anticipation anxiety||0.28 (0.001)||0.57 (<0.001)||0.33 (<0.001)||0.19 (0.23)||0.43 (<0.001)||0.48 (0.001)||–|
|Vitality||0.48 (<0.001)||0.30 (0.06)||0.55 (<0.001)||0.46 (0.002)||0.59 (<0.001)||0.60 (<0.001)||–|
|Stigmatization||0.49 (<0.001)||0.46 (0.006)||0.19 (0.03)||0.17 (0.27)||0.40 (<0.001)||0.24 (0.12)||–|
|Vulnerability||0.27 (0.001)||0.53 (<0.001)||0.24 (0.005)||0.20 (0.21)||0.45 (<0.001)||0.53 (<0.001)||–|
|Transmission||0.30 (0.006)||0.34 (0.03)||0.09 (0.28)||0.12 (0.44)||0.29 (0.007)||0.18 (0.24)||–|
|Viral response subscale||–||–||–||–||–||–||No response = 49.0; response = 60.0 (0.03; 0.42)|
Only 7 of 40 patients achieved viral response between the baseline and follow-up administrations in our natural history sample. Table 5 presents the responsiveness indices in patients with viral response. Although limited by a small sample size, viral response was associated with improvement in overall score and in 5 of 6 individual scale scores. The only exception was the vulnerability scale (mean change = −11.7). The effect size ranged from 0.25 (transmission) to 0.69 (anticipation anxiety); the standardized response mean ranged from 0.33 (transmission) to 0.79 (stigmatization); and responsiveness statistics ranged from 0.37 (vitality) to 0.98 (anticipation anxiety). Based on responsiveness statistics, the overall score, anticipation anxiety, stigmatization, vulnerability, and viral response scale scores showed medium (effect size >0.5) to large changes (effect size >0.8) over time. Table 6 presents the correlation coefficients between HBQOL v1.0 change scores and the PCS, MCS, and self-reported health change scores. The overall HBQOL v1.0 score showed statistically significant correlations with changes in PCS (r = 0.34, P = 0.03) and global health assessment (r = 0.44, P = 0.004). Psychological well-being, vitality, and vulnerability scales showed moderate correlations with changes in MCS (r = 0.28, 0.39, and 0.34, respectively). Differences in anticipation anxiety and vitality scales correlated with changes in global health assessment over time (r = 0.47 and 0.44, respectively).
Table 5. Responsiveness of HBQOL to Viral Response
Table 6. Correlation of HBQOL v1.0 Change Scores with Change Scores of Key Anchors
|Overall score||40||−0.03 (0.83)||0.34 (0.03)||−0.44 (0.004)|
|Psychological well-being||40||0.087 (0.59)||0.277 (0.08)||−0.19 (0.23)|
|Anticipation anxiety||40||−0.06 (0.71)||0.14 (0.40)||−0.47 (0.002)|
|Vitality||39||0.11 (0.52)||0.39 (0.01)||−0.44 (0.004)|
|Stigmatization||40||0.04 (0.79)||0.19 (0.24)||−0.28 (0.07)|
|Vulnerability||40||−0.21 (0.20)||0.35 (0.03)||−0.16 (0.32)|
|Transmission||40||−0.29 (0.07)||0.15 (0.36)||−0.16 (0.32)|
|Viral response||40||−0.09 (0.58)||−0.09 (0.58)||−0.42 (0.006)|
Although it is well known that chronic infection with hepatitis C leads to significant HRQOL decrements,4 it is less clear whether chronic HBV has a significant impact on HRQOL. Previous data indicate that patients with HBV score lower on the SF-36 than population controls,3 suggesting that HBV can indeed negatively affect HRQOL. Yet despite this finding, published guidelines in HBV do not make any recommendations about HRQOL assessments in clinical care,1 and many providers do not routinely consider HRQOL when assessing patients or making treatment decisions in chronic HBV.
We have found that the relative de-emphasis of HRQOL assessment in HBV compared with HCV may signal a possible disconnect between providers and patients. In particular, our qualitative interviews of HBV patients demonstrate clear and extensive evidence that HBV can significantly impact physical, psychological, and social health (Supplementary Appendix A), even in the absence of cirrhosis or hepatocellular cancer. Specifically, patients with HBV voice a range of striking disease-specific fears and concerns (“I'm afraid I will die young”; “I feel like a time bomb”), social sequelae (“I am contagious and some people shy away from me”; “I cannot establish intimate relationships”), and psychological consequences (“I was so depressed when I got my blood test result”; “I wonder if I will ever feel normal again”; “God bless me”). Taken together, these qualitative data indicate that a thorough clinical assessment of chronic HBV should extend beyond the usual biological evaluations (e.g., laboratory values, liver biopsy results) and should also include a targeted assessment of HRQOL. Failure to identify HRQOL decrements in HBV may lead to the myopic view that biological outcomes are of sole importance—a view that likely underestimates the true burden of illness engendered by HBV.
In light of this finding, and given the absence of disease-targeted HRQOL questionnaires in HBV, we sought to develop and validate the first HRQOL instrument in HBV: the HBQOL v1.0. Our research has 3 key findings: (1) the decrements in HRQOL engendered by HBV primarily affect psychological and social health, with a smaller relative impact on physical health; (2) HRQOL in HBV can be measured both as a singular concept (i.e., using the “overall score”), and as 6 subdomains (i.e., subscale scores: psychological well-being, anticipation anxiety, vitality, disease stigma, vulnerability, and transmissibility); and (3) HRQOL is better in patients achieving a viral response from treatment versus patients without viral response, particularly in terms of transmissibility concerns and disease vulnerability (jointly captured in the “viral response” subscale). In addition, our longitudinal follow-up data suggest that the HBQOL v1.0 may be sensitive to changes in patients' viral response status over time. Together, these results render support for the reliability and validity of the HBQOL v1.0.
Of note, the HBQOL is more heavily weighted with psychosocial characteristics than physical symptoms of chronic HBV infection. This was not an explicit decision, but instead was the end result of a standardized process that included cognitive patient interviews, feedback from an expert panel, and extensive item selection procedures. This process highlighted an interesting fact that patients with noncirrhotic HBV have substantial psychosocial decrements in quality of life, and that these domains have relatively more impact than physical decrements. It is important to emphasize that this finding does not suggest that HBV has no physical impact. Instead, it only suggests that patients with noncirrhotic HBV endorse more psychosocial symptoms than physical symptoms when asked to consider how HBV impacts their overall quality of life.
These findings also suggest that treatment of chronic HBV may have benefits beyond improving physical health. Specifically, treatment may improve several dimensions of psychosocial health in ways that have not been previously documented. This is important, because some patients or clinicians might argue against treatment because the potential health benefits are often far in the future (e.g., avoiding cirrhosis, transplantation, death) and because HBV does not usually cause extensive physical symptoms in patients with compensated noncirrhotic disease. However, clinicians and patients should recognize that treatment may have substantial short-term benefits, not only for physical health, but also for psychosocial health.
Because HRQOL is an important end point in patients with HBV, the validation of this questionnaire provides a new tool to potentially improve patient assessment, streamline diagnostic and therapeutic decision-making, and serve as a validated end point for use in chronic HBV treatment trials. Clinicians may consider using the HBQOL v1.0 to track patient HRQOL in the outpatient setting. In this regard, HRQOL may serve as another “vital sign” to help guide medical decision-making. Moreover, if future therapies can demonstrate HRQOL benefits (in addition to traditional biological benefits) in clinical trials, providers and patients may be better informed to determine how and when to best initiate treatment for HBV.
Our study has several strengths. First, we conducted an extensive content validation process that drew from several sources, including the biomedical literature, expert focus groups, and patient cognitive interviews. In this regard, we collected extensive “raw” material from which to construct our HRQOL questionnaire, as detailed in Supplementary Appendix A. Second, in an effort to maximize the generalizability of the resultant instrument, we recruited patients from geographically and demographically diverse medical practice settings throughout the United States. Third, we validated the instrument in a large sample size (138 patients), thereby minimizing the probability of type II errors. Fourth, we generated a priori hypotheses about HRQOL in HBV and tested those hypotheses with factor analysis. In contrast, we did not use factor analysis as a purely exploratory, hypothesis-generating exercise. Fifth, we measured construct validity across a range of clinically relevant anchors, including viral response status, generic HRQOL, and global health using data from both the baseline and follow-up questionnaires. Last, in addition to measuring instrument validity, we ensured that the HBQOL v1.0 is also reliable in terms of its structure (by measuring Cronbach's α) and test–retest performance.
Our study also has potential weaknesses. First, the instrument was validated in an English-speaking United States population and therefore cannot be easily generalized to non–English-speaking non–United States populations. Because HBV is extremely prevalent in Asian countries, in particular, it will be important to validate the instrument in the Chinese dialects, among other languages. Second, the sample population consisted exclusively of patients with health insurance coverage and access to health care. Thus, the instrument may not be directly applicable to alternative patient groups. Finally, due to the limited sample size in the longitudinal follow-up, our responsiveness estimates may not be optimally reliable. Despite this, our data provide preliminary support for the responsiveness of the HBQOL v1.0. Future research will aim to confirm the responsiveness and to determine the minimally clinically important differences of the HBQOL v1.0 using larger sample sizes and longer duration of follow-up. This longitudinal data will bolster the instrument's validity for tracking patients over time, both in clinical practice and in clinical trials.
In conclusion, chronic HBV diminishes HRQOL across a wide range of health domains. Sustained viral response is associated with improved HRQOL compared with no viral response, thereby suggesting that treatment of HBV may improve patient-oriented outcomes in addition to established biological outcomes. Our data provide evidence that the HBQOL v1.0 is reliable and valid and may ultimately prove useful in clinical trials and for measuring HRQOL in everyday clinical practice.
We thank Kerry Newsome for her administrative assistance, Dr. Jeffrey Kahn for valuable input in our expert focus group, and Dr. James Tung for content expertise.