Health-Related Quality of Life Predicts Mortality in Patients With
Advanced Chronic Liver Disease
FASIHA KANWAL,* ,, IAN M. GRALNEK,
RON D. HAYS, ,# ANGELIQUE ZERINGUE,* FRANCISCO DURAZO,** STEVEN B. HAN,** , SAMMY SAAB,** ROGER BOLUS,
and BRENNAN M. R. SPIEGEL
,, ** , *John Cochran VA Medical Center, St. Louis, Missouri;
Division of Gastroenterology, Saint Louis University, St. Louis, Missouri;
UCLA/VA Center for Outcomes Research and Education at UCLA, Los Angeles, California;
Rappaport Faculty of Medicine, Technion Israel Institute of Technology, Rambam Health Care Campus, Haifa, Israel;
Department of Health Services, UCLA School of Public Health, Los Angeles, California; # RAND, Santa Monica, California; **Division of Digestive Diseases, David Geffen School of Medicine at UCLA, Los Angeles, California; and
Department of Gastroenterology & Hepatology, VA Greater Los Angeles Healthcare System, Los Angeles, California BACKGROUND&AIMS: It is well-established that cirrho- sis negatively impacts health-related quality of life (HRQOL). However, it is less clear how to use this information in everyday clinical practice. If HRQOL predicted survival in cirrhosis, then measuring HRQOL would have important clinical implications. We sought to measure the association between HRQOL and survival in patients with cirrhosis and investigated whether the relationship between HRQOL and survival is independent of Model for End-Stage Liver Disease (MELD). METHODS: We measured HRQOL in 156 patients with cirrhosis awaiting liver transplantation by using the Short Form Liver Disease Quality of Life instrument. We followed patients prospectively and used Cox proportional hazard models to measure the independent effect of baseline HRQOL on survival, adjusting for MELD and other covariates. RESULTS: During a mean 9-month follow- up, 26 (17%) patients died, and 30 (20%) received liver trans- plants. In unadjusted analysis, higher baseline HRQOL predicted lower mortality (hazard ratio, 0.96; 95% condence interval, 0.94 0.99). Specically, for each 1-point increase in HRQOL, there was a 4% decrease in mortality. These results did not change after adjusting for MELD scores, patient demographics, or psychos- ocial characteristics; the MELD score accounted for 1% of the variation in HRQOL scores (P .18). Survival was most strongly predicted by activities of daily living, health distress, sleep disturbance, and perceived disease stigma. CONCLU- SIONS: Higher HRQOL predicts lower mortality in pa- tients with cirrhosis. This relationship is independent of MELD; MELD does not capture liver-specic HRQOL. Be- yond its use as a secondary outcome in clinical trials, HRQOL could be used to predict survival of patients with advanced liver disease. C irrhosis is a prevalent and expensive condition, affecting 5.5 million patients at a cost of more than $1.5 billion annually in the US. 13 Cirrhosis is the second leading cause of digestive diseaserelated mortality, preceded only by colorectal cancer. 2 This human and economic burden of illness is multi- plied by the dramatic impact of cirrhosis on health-related quality of life (HRQOL) resulting from complications of ad- vanced liver disease, including encephalopathy, ascites, varices, and hepatocellular cancer. 410 Thus, the burden of disease en- gendered by cirrhosis extends beyond its impact on traditional biologic outcomes to include a negative impact on patient- reported outcomes such as HRQOL. Despite the realization that HRQOL is an important out- come in patients with cirrhosis, clinicians rarely assess HRQOL in clinical practice. One explanation for this disconnect might be a perceived lack of clinical actionability of HRQOL data, in contrast to objective disease indices such as the Model for End-Stage Liver Disease (MELD) score. 11,12 Whereas MELD ac- curately predicts mortality in cirrhosis and guides decision- making for transplantation, 12 it is unclear whether HRQOL provides the same predictive value, or whether it can augment MELD-based prediction in any clinically meaningful way. Yet in other chronic diseases such as end-stage renal disease, conges- tive heart failure, and cancer, among others, HRQOL is a consistent predictor of mortality. 1319 It is possible that HRQOL might also predict mortality in cirrhosis, although this hypoth- esis is untested. If HRQOL could predict mortality independent of MELD, then it would suggest that HRQOL might augment MELD by providing incremental predictive ability not captured by tradi- tional biologic indices. Moreover, HRQOL might complement MELD in identifying high-risk patients and help in making decisions about how best to allocate resources among different patient groups. Therefore, we sought to measure the predictive value of HRQOL in a prospective cohort of patients with cirrhosis and investigated whether the relationship between HRQOL and survival is independent of MELD. Methods Study Subjects We conducted a prospective study of 156 consecutive patients with cirrhosis awaiting liver transplantation at the UCLA-Dumont Liver Transplant Program in Los Angeles, Cal- ifornia between March of 2004 and June of 2006. We excluded subjects if they were 18 years old, had previously received a liver transplant, or had at least grade 2 hepatic encephalopathy despite optimal medical management. This study was approved Abbreviations used in this paper: CI, condence interval; HR, hazard ratio; HRQOL, health-related quality of life; MCS, mental composite summary score; MELD, Model for End-Stage Liver Disease; PCS, phys- ical composite summary score; SD, standard deviation; SF-LDQOL, Short Form Liver Disease Quality of Life. 2009 by the AGA Institute 1542-3565/09/$36.00 doi:10.1016/j.cgh.2009.03.013 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2009;7:793799 by the Institutional Review Board at the University of Califor- nia, Los Angeles. Health-Related Quality of Life Measures We measured HRQOL by using a liver diseasetargeted instrument, the Short Form Liver Disease Quality of Life (SF- LDQOL). 20 The SF-LDQOL is a short form version of the Liver Disease Quality of Life Version 1.0 (LDQOL 1.0) instrument. 21 The LDQOL 1.0 was developed and evaluated in a cross-sec- tional, multicenter study including 221 patients with advanced liver disease. 21 With HRQOL data from the original psychomet- ric evaluation of the LDQOL 1.0, a subset of the LDQOL 1.0 items was selected for the SF-LDQOL on the basis of the maximum R 2 regression and internal consistency reliability co- efcients. 22 Detailed psychometric evaluations provided strong support for the reliability, construct validity, and responsive- ness of the SF-LDQOL. 20 The disease-targeted scales of the SF-LDQOL include 36 items measuring 9 scales, including liver diseaserelated symp- toms, liver diseaserelated effects on activities of daily living, concentration/memory, health distress, sleep, loneliness, hope- lessness, self-perceived stigma of liver disease, and sexual func- tioning/problems. These are combined into an overall SF- LDQOL score that summarizes the overall disease-targeted HRQOL with a single number ranging from 0100, with higher score indicating better HRQOL. Similar to the LDQOL 1.0, the SF-LDQOL uses the SF-36 version 2.0 as the generic core. 23 The SF-36 instrument includes 8 multi-item scales (physical func- tioning, physical role limitations, emotional role limitations, bodily pain, general health, emotional well-being, energy/fa- tigue, and social functioning) and can be aggregated into 2 summary scores: (1) physical component score (PCS) and (2) mental component score (MCS). 23 With the exception of items included in the sexual function- ing/problems scales, data were missing for less than 5% of the items. Because the sexual functioning/problem scale applies only to the subset of population with recent sexual activity (ie, within the previous 4 weeks), these items were completed by only 49% of patients in our cohort. For these patients, we followed the standard HRQOL scoring procedure and calcu- lated the HRQOL scores by using the items that the respondent answered. 24 Mortality Assessment We followed all participants longitudinally and termi- nated the follow-up at the time of the patients death, liver transplantation, or December 31, 2006, whichever occurred rst. As part of the study protocol, patients were followed up once during their routine clinic visits 6 months after their enroll- ment. Subsequent follow-up was based on medical chart review. To maximize ascertainment of mortality in the study subjects, we contacted patients who were alive at their last recorded visit by telephone or mail and obtained information on their status before the end of the study. Control Variables On the basis of a priori hypotheses guided by data from the literature, we specied relevant factors that might affect patient survival while awaiting liver transplantation. These in- cluded sociodemographic characteristics (age, race, household income, marital status, education level, employment), liver dis- easerelated characteristics (etiology of liver disease, MELD score, diagnosis of liver cancer, ascites, hepatic encephalopathy, serum sodium, dialysis), and comorbidity-related characteristics (Charlson Comorbidity Index). 11,12,2528 The Charlson Comor- bidity Index is a weighted index that takes into account the number and the seriousness of comorbid diseases, 28 with a stepwise increase in the risk of 1-year mortality that ranges from 12% for 0 score to 26% with scores 12, 52% with scores 34, and 85% with scores 5. Charlson Index denes liver disease patients as having either mild or moderate/severe dis- ease. Because all patients in this study had complicated liver disease, we assumed they all received a 3 on the liver component of the score, where 3 is moderate/severe and 1 is mild liver disease. We computed MELD scores for all patients according to published algorithms. 11,12 Table 1 provides the full list of patient factors and their method of categorization. Statistical Analysis We analyzed patient data by using SAS, version 9.1 (SAS Inc, Cary, NC). All P values were two-sided, with an alpha of .05 as the standard for determining signicance. With bivariate Cox proportional-hazards models, we rst estimated the unadjusted relationship between each covariate and mortality. We next measured the independent relationship between HRQOL and mortality, while adjusting for predictors of mortality. We per- formed a Cox multivariable proportional-hazard model by us- ing a backward stepwise procedure for selecting variables, re- taining those with P .1. We computed the adjusted hazard ratios (HRs) and 95% condence intervals (CIs) to estimate the strength of association of each predictor with time to death. To test the robustness of our ndings, we performed a separate model by using a forward stepwise procedure. The purpose of this was to evaluate the convergency between the forward and backward models. With logistic regression analyses, we calcu- lated the relative risk of death for patients with varying degrees of HRQOL decrement compared with those with average HRQOL, while adjusting for liver disease severity. We calculated bootstrap 95% CIs around the point estimate by using 1000 sampling iterations. Finally, we used linear regression analysis to measure the proportion of variance of SF-LDQOL captured by MELD scores. The purpose of this analysis was to determine whether MELD is a sufcient proxy measure for liver-specic HRQOL. Results Baseline Characteristics The mean age standard deviation (SD) of eligible patients was 53.9 11 years, and 55% were male. Sixty-six percent were white, 6% black, and 5% Asian. Table 1 summarizes other key sociodemographic characteristics. The most common etiologies for cirrhosis included viral hepatitis (47%), metabolic or cryptogenic liver diseases (28%), autoimmune liver diseases (13%), and alcohol-related liver disease (10%). At baseline, 32% had moderate to severe ascites, 45% had grade 1 encephalopa- thy, and 32% had a history of prior variceal hemorrhage. Nine percent had hepatocellular cancer, and 14% were receiving he- modialysis. Because we counted advanced liver disease in our calculation, the Charlson Comorbidity Index was high for our cohort (median, 4; interquartile range, 35). The baseline serum 794 KANWAL ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7, No. 7 sodium was 139 4.8 mEq/1, and MELD score was 18.5 3.9. When administered together with the SF-36 generic core, the SF-LDQOL required 18 9 minutes to complete. The baseline SF-LDQOL, PCS, and MCS scores were 55.3 19, 30.8 11, and 41.0 11, respectively. Eighty-six patients completed the 6-month clinic follow up. For the remaining patients, we relied on medical chart review to ascertain follow-up. We obtained information on the status of 134 subjects before the end of the study. We were unable to contact 20 patients and thus censored their follow-up at the last recorded visit. Of these, only 4 patients had 6 months of follow-up. Two patients were lost to follow-up (ie, they did not have repeat clinic visits and did not respond to telephone/mail inquiry after the baseline HRQOL assessment). During a mean of 316 days of follow-up, 26 (17%) patients died, and 30 (20%) underwent liver transplantation. Predictors of Mortality Bivariate analyses. Table 2 displays the results of the bivariate Cox proportional hazard models. Both the SF- LDQOL (HR, 0.96; 95% CI, 0.940.99) and MELD scores (HR, 1.11; 95% CI, 1.021.21) were associated with mortality. The results indicate that for each 1-point increase in the SF-LDQOL score, there was a 4% decrease in mortality during the study period. For each 1-point increase in MELD, there was an 11% increase in mortality. Given different distributions (SF-LDQOL range, 0100; MELD range, 640), 1-point change in the SF- LDQOL cannot be directly compared with 1-point change in MELD score. To allow for comparison of MELD versus SF- LDQOL, we transformed these data into SD units. We found that for each 1 SD increment in SF-LDQOL, there was a 54% decrease in mortality, whereas a 1 SD increment in MELD was associated with 48% increase in mortality. Table 1. Baseline Sample Characteristics Variable Results Sociodemographic characteristics Age, mean (y) (SD) 53.9 (11) Gender, male (%) 54.8 Race, (%) White 66.2 Black 5.7 Asian/Pacic Islander 4.5 Native American 1.3 Other a 22.3 Marital status (%) Single 15.6 Married 66.2 Separated/divorced 14.9 Widowed 3.3 Education level (%) Eighth grade or less 11.8 Some high school 12.5 High school diploma 17.1 Some college 31.5 Professional or graduate degree 17.1 Employment status (%) Working full-time 12.7 Working part-time 1.9 Unemployed 1.3 Retired 21.0 Disabled 48.4 Homemaker 7.0 Health insurance coverage (%) Medicare only 4.5 Medicare and supplemental 12.1 Medicaid only 17.8 Veterans Administration 0.6 Private 5.1 HMO, PPO, IPA b 43.9 Other 9.5 None/not sure 3.8 Total household income (%) $5,000 6.3 $5,000$10,000 11.9 $10,001$25,000 20.9 $25,001$50,000 17.5 $50,001$75,001 16.8 $75,000 18.2 Not sure 8.4 Clinical characteristics Primary etiology of liver disease (%) HCV, HBV 47.4 Alcohol 10.4 PBC/PSC/AIH 13 Cryptogenic, NAFLD, HHC 28 Ascites (%) Absent/mild 68.1 Moderate/severe 31.9 Hepatic encephalopathy (%) Absent 55 Grade 1 45 Variceal bleeding (%) No 68 Yes 32 Liver cancer (%) No 91.5 Yes 8.5 Table 1. Continued Variable Results Dialysis (%) No 85.7 Yes 14.2 Charlson Comorbidity Index, median (interquartile range) 4 (35) Serum sodium, mean (SD) 139 (4.8) MELD score, mean (SD) 18.5 (3.9) HRQOL, mean (SD) LDQOL score 55.3 (18.8) SF-36 PCS 30.8 (10.8) SF-36 MCS 41.0 (10.8) Patient outcomes Died (%) 16.8 Transplanted (%) 19.5 Follow-up, mean (days) (SD) 316 (210) Median (interquartile range) 276 (146455) PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; AIH, autoimmune hepatitis; NAFLD, nonalcoholic fatty liver disease; HHC, hereditary hemochromatosis. a A signicant proportion (87%) of these patients identied themselves as Hispanic. b HMO, health maintenance organization; PPO, preferred provider or- ganization; IPA, independent practice association. These represent different models used by managed care organizations to deliver healthcare in the US. July 2009 HRQOL AND MORTALITY IN CIRRHOSIS 795 We next divided the patients into those with high versus low HRQOL, stratied by the SF-LDQOL median split. We found that patients with higher HRQOL had a 61% lower risk of death compared with those with impaired HRQOL (HR, 0.39; 95% CI, 0.170.91). In addition to HRQOL and MELD, presence of ascites (HR, 6.0; 95% CI, 2.6913.5) and hepatic encephalopathy (HR, 2.0; 95% CI, 1.263.16) predicted mortality. We also found a negative association between serum sodium and risk of death; higher serum sodium predicted lower mortality (HR, 0.88; 95% CI, 0.830.94). The other variables, including SF-36 PCS, MCS, and sociodemographic variables, were not associated with mor- tality during the study period. Multivariate analysis. As displayed in Table 3, the stepwise Cox proportional hazard model identied 3 signicant predictors of mortality in our patient group. These included SF-LDQOL (HR, 0.97; 95% CI, 0.950.99), MELD (HR, 1.11; 95% CI, 1.021.21), and presence of ascites (HR, 4.38; 95% CI, 1.8410.42). The predictors and their magnitude of association did not change in the sensitivity analysis by using the forward stepwise selection method. To evaluate whether using both HRQOL and MELD would increase the predictive ability of the model versus using either variable alone, we estimated an over- all C index for the stepwise Cox proportional hazard model. 29 This statistic has been developed to measure model discrimi- nation in survival analyses. This test is similar to the c statistic except that instead of evaluating concordance between observed and predicted probabilities, it evaluates concordance between observed and predicted time periods between 2 patients. We calculated the overall C index with HROQL, MELD, or both HRQOL and MELD in the nal model (in conjunction with ascites). The C index values for predicting mortality in our cohort by using either MELD or HRQOL were 0.75 (95% CI, 0.640.85) and 0.74 (95% CI, 0.640.84), respectively. The C index for the model including both MELD and HRQOL was 0.79 (95% CI, 0.700.88). Figure 1 demonstrates the results of the logistic regression model, stratied by varying levels of SF-LDQOL while adjusting for liver disease severity. The data indicate that SF-LDQOL signicantly predicted mortality at all levels. For example, those with scores 2 SDs above the mean were approximately 60% (relative risk, 0.42; 95% CI, 0.410.43) less likely to die versus patients with average HRQOL. In contrast, those with scores 2 SDs below the mean were 2-fold more likely to die versus patients with average HRQOL (relative risk, 2.07; 95% CI, 2.02.1). To further understand the specic aspects of HRQOL most strongly associated with mortality, we conducted additional analyses by using the 9 scales of the SF-LDQOL in place of the Table 3. Predictors of Mortality as Determined by Stepwise Cox Regression Model Variable Adjusted HR (95% CI) P value LDQOL 0.97 (0.950.99) .039 MELD 1.11 (1.021.21) .014 Presence of moderate/severe ascites 4.38 (1.8410.42) .0008 NOTE. This table displays results of backward stepwise regression model. The results did not change when the analysis was repeated by using the forward stepwise regression method. Table 2. Associations Between Patient Baseline Characteristics and Mortality Variable HR 95% CI P value Age 0.99 0.961.03 .85 Gender, male 0.66 0.31.44 .30 Race Non-white 1.00 .10 White 2.24 0.845.97 Marital status .107 Single 1.00 0.241.14 Married 0.53 Education High school 1.00 .72 High school 1.15 0.522.54 Health insurance coverage Public 1.00 Private/HMO, PPO, IPA 0.66 0.301.45 .68 Total annual household income ($) 75,000 1.00 25,00075,000 1.62 0.604.31 .33 25,000 1.44 0.494.15 .50 Primary etiology of liver disease HCV/HBV 1.00 Alcohol 1.72 0.555.38 .35 PBC/PSC/AIH 0.91 0.342.42 .85 Cryptogenic/NAFLD/HHC 0.95 0.273.39 .94 Ascites 1.00 Absent/mild 6.04 .0001 a Moderate/severe 2.6913.54 Encephalopathy No 1.00 .003 a Yes 2.00 1.263.16 Prior variceal bleeding No 1.00 .63 Yes 0.81 0.341.94 Liver cancer No 1.00 .39 Yes 0.42 0.053.12 Dialysis No 1.00 .57 Yes 0.71 0.212.37 Serum sodium 0.88 0.830.94 .0001 a Charlson Comorbidity Index Low (median) 1.00 .81 High (median) 0.91 0.411.99 MELD score 1.11 1.021,21 .017 a LDQOL score 0.96 0.940.99 .01 a SF-36 PCS 0.97 0.931.01 .15 SF-36 MCS 0.96 0.931.00 .068 NOTE. Cox proportional hazards models were used for these bivariate analyses. HRs for age, serum sodium, MELD, LDQOL, SF-36 PCS, and SF-36 MCS represent the risk of mortality for each 1-point change in these variables. HMO, health maintenance organization; PPO, preferred provider orga- nization; IPA, independent practice association; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; AIH, autoimmune hep- atitis; NAFLD, nonalcoholic fatty liver disease; HHC, hereditary hemo- chromatosis. a Characteristics associated with patient mortality in patients with advanced liver disease. These include presence of moderate or se- vere ascites, presence of hepatic encephalopathy, serum sodium, MELD, and LDQOL score. 796 KANWAL ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7, No. 7 summary score. The following subscales were predictive of mor- tality independent of MELD: (1) activities of daily living (HR, 0.97; 95% CI, 0.950.98); (2) health distress (HR, 0.98; 95% CI, 0.960.99); (3) sleep (HR, 0.96; 95% CI, 0.940.98); and (4) stigma (HR, 0.98; 95% CI, 0.970.99). To establish the relationship between SF-LDQOL and MELD, we performed a linear regression analysis with baseline SF-LDQOL as the dependent variable and baseline MELD as the predictor variable. The model revealed that MELD was not a statistically signicant predictor of SF-LDQOL (R 2 0.01; P .18). Discussion The health economic burden of cirrhosis is amplied by its impact on HRQOL, resulting from a range of physical, psychological, and social stressors engendered by the disease and its treatment. 410 Physically, patients with cirrhosis suffer from debilitating symptoms including fatigue, shortness of breath, chronic pain, poor appetite, sexual dysfunction, and sleep abnormalities, among many others. The protean physical manifestations of cirrhosis span nearly every organ system, range in severity from nuisance to life-threatening, and nega- tively disrupt patients lifestyles. 21 Psychologically, patients with cirrhosis might have a disproportionately high incidence of depression, often suffer from anxiety from the impending phys- ical demise, and encounter difculties in coping with their disease. 30,31 The ability to perform activities of daily living and to maintain baseline social interactions is also signicantly impaired. 21 Patients with cirrhosis are limited by dietary restric- tions, time restrictions, and oftentimes overwhelming physical and psychological restrictions that, taken together, can lead to disruptions in personal relations and subsequent social with- drawal. 21 Yet, with few exceptions, adoption of routine HRQOL monitoring in everyday clinical practice remains suboptimal. This disconnect between theory and practice might arise for several reasons, including issues pertaining to the time and effort it takes to routinely assess HRQOL in a busy clinical setting, coupled with uncertainty about how to use HRQOL data once collected. Our data provide a compelling rationale for incorporating HRQOL information into clinical decision-making in cirrhosis. We have found that disease-targeted HRQOL provides impor- tant prognostic information in patients with advanced liver disease that goes beyond the information furnished by clinical variables used in routine practice. In particular, our research has 3 key ndings. First, we found that higher HRQOL, as measured by a liver diseasetargeted HRQOL instrument, is associated with lower mortality in patients with advanced liver disease. Moreover, the predictive value of HRQOL for mortality is comparable to, if not stronger than, that of MELD. For example, a 1 SD change in HRQOL and MELD resulted in a similar magnitude of change in mortality (54% vs 48%, respectively). Moreover, there was no difference in the accuracy of predicting mortality by using MELD or HRQOL in our sample (C index 0.75 and 0.74, respectively). These data underscore the clinical signicance of the observed association between HRQOL and mortality. Second, we found that the association between disease-tar- geted HRQOL and death in patients with cirrhosis is indepen- dent of traditional clinical variables used to risk stratify pa- tients. This suggests that HRQOL measurement might tap into aspects of underlying risk or illness severity that are not fully captured by laboratory and clinical parameters. Moreover, we found that MELD was a poor predictor of HRQOL, capturing only 1% of the variance in SF-LDQOL scores. Thus, MELD and HRQOL are different concepts altogether, yet each indepen- dently predicts mortality. Although adding HRQOL informa- tion improved the discriminative ability of MELD-based model in our exploratory analysis, HRQOL data are ineligible for use in making liver allocation decisions as a result of their inher- ently subjective nature. Regardless, these data might certainly be useful in clinical practice to complement MELD as a predic- tor of mortality and thus help guide decision-making in a manner not achievable with MELD alone. For example, patients with low HRQOL might benet from closer follow-up or en- rollment in structured disease management programs. Such programs are known to enhance HRQOL and reduce resource utilization in patients with other chronic medical conditions such as congestive heart failure. 32 In addition, HRQOL mea- sures might help clinicians address functioning and well-being issues beyond the scope of usual care. This is consistent with the practice of incorporating performance status in the prog- nostication of patients with malignancy, including hepatocel- lular carcinoma. These practical considerations bolster the use Figure 1. Association between liver disease-targeted HRQOL and short-term mortality. The adjusted relative risk of short-term mortality with 95% CI in patients with advanced liver disease for different levels of HRQOL is compared with the common baseline of patients with average levels of HRQOL (mean). Data indicate that in patients with advanced liver disease, disease-targeted HRQOL signicantly predicts short-term mortality. For example, patients with HRQOL 2 SD below the mean have a 2.3 times higher risk of death compared to pa- tients with average HRQOL. In contrast, patients with HRQOL 2 SD above the mean are 50% less likely to die compared to patients with average HRQOL. All levels were statistically signicant (P .05). July 2009 HRQOL AND MORTALITY IN CIRRHOSIS 797 of HRQOL for tracking patients over time both in clinical practice and in clinical trials. Third, we found that survival is most strongly associated with activities of daily living, health distress, sleep disturbance, and perceived disease stigma. Addressing these specic domains through tailored behavioral interventions and disease manage- ment programs might help improve overall HRQOL and could potentially improve patient outcomes in a cost-effective man- ner, although this remains untested. Our study has several strengths. First, the patients were drawn from clinical practice, not from a clinical trial. Therefore, our sample is representative of patients with advanced liver disease who are awaiting liver transplantation in routine clinical practice. Second, we examined a wide range of patient-level variables that might affect wait-list mortality. These included sociodemographic characteristics, cause and severity of liver disease, and comorbidities, among others. Third, to maximize ascertainment of mortality in the study subjects, we did not rely solely on clinical records, but instead we took efforts to contact patients by telephone and mail to conrm their health status before completing the study. Fourth, we conducted detailed psychometric evaluations of the HRQOL instruments in our study sample, reported elsewhere, and these assessments pro- vide support for the reliability and validity of the SF-LDQOL. 20 Our study also has limitations. First, our sample consisted of ambulatory patients with stable health insurance coverage and access to healthcare. Therefore, our results might not generalize to patients without health insurance and limited access to care. However, because patients with limited access to healthcare are likely to be underrepresented in any study evaluating liver transplantation in the U.S., this weakness represents a limita- tion of the healthcare system in general. Second, we included patients who sought care at one tertiary care center, thus lim- iting the applicability to patients in other healthcare facilities. However, the liver transplant center at UCLA is one of the largest transplant centers in the U.S. and provides care to a demographically diverse group of patients with advanced liver disease. Moreover, it is difcult to establish an explanation for why the observed relationship between HRQOL and mortality in our center would be systematically different from the rela- tionship in other centers. Third, we excluded patients with grade 2 hepatic encephalopathy because these patients would have difculty completing the study questionnaire, and this, in turn, would compromise the reliability of HRQOL scales. Ex- clusion of subjects with hepatic encephalopathy limits the ex- ternal validity (ie, generalizability) of our study; our ndings are only applicable to cirrhotic patients with either no or only clinically mild hepatic encephalopathy. However, this weakness is inherent in the assessment of any patient-reported outcome, including HRQOL, and is not specic to our study. Fourth, although our sample exceeds the usual 1020:1 ratio of subjects to independent variables, it is nonetheless possible that we might have overt the model on the basis of too few outcome events. However, overtting is typically associated with unreal- istically large estimated coefcients and/or standard errors. Our point estimates for HRQOL and MELD are consistent with those from other studies evaluating the association between these scores and mortality in other areas of medicine and advanced liver disease, respectively. 14,15,33 The fact that we and others found similar effect sizes for the key predictors not only provides convergent validity to these ndings but also suggests that overtting might not be a signicant issue in our analysis. Despite these corroborating data, future studies with larger sample size and longer follow-up will be needed to conrm our results. Moreover, because our goal was to determine whether HRQOL could predict mortality independent of MELD score, we did not seek to develop a predictive model that accurately predicts mortality in cirrhosis. Future studies with character- ization of the full spectrum of patient and non-patient factors might allow a more accurate prediction of prognosis in patients with cirrhosis. In conclusion, our results demonstrate that a short, disease- targeted HRQOL instrument can be used to identify patients with advanced liver disease who are at high risk of short-term death. These ndings provide additional support to the validity of the SF-LDQOL as an HRQOL measure in patients with advanced liver disease. HRQOL assessments might complement objective measures of disease severity not only to accurately and comprehensively assess health status but also to better risk stratify patients with advanced liver disease. Our results under- score the utility of HRQOL in patients with advanced liver disease beyond its mere use as a secondary outcome in clinical trials and observational studies. References 1. Dufour MC. Chronic liver disease and cirrhosis. In: Everhart JE, ed. Digestive diseases in the United States: epidemiology and impact. Washington, DC: US Government Printing Ofce, 1994: 613646. 2. Kim WR, Brown RS Jr, Terrault NA, et al. Burden of liver disease in the United States: summary of a workshop. Hepatology 2002; 36:227242. 3. Anonymous. The burden of gastrointestinal diseases. Bethesda, MD: American Gastroenterological Association, 2001:4160. 4. Kanwal F, Hays RD, Kilbourne AM, et al. Are physician-derived disease severity indices associated with health-related quality of life in patients with end-stage liver disease? Am J Gastroenterol 2004;99:17261732. 5. Marchesini G, Bianchi G, Amodio P, et al. Factors associated with poor health-related quality of life of patients with cirrhosis. Gas- troenterology 2001;120:170178. 6. Steel JL, Chopra K, Olek MC, et al. Health-related quality of life: Hepatocellular carcinoma, chronic liver disease, and the general population. Qual Life Res 2007;16:203215. 7. Younossi ZM, McCormick M, Price LL, et al. Impact of liver transplantation on health-related quality of life. Liver Transpl 2000;6:779783. 8. Younossi ZM, Kiwi ML, Boparai N, et al. Cholestatic liver dis- eases and health-related quality of life. Am J Gastroenterol 2000; 95:497502. 9. Younossi ZM, Boparai N, Price LL, et al. Health-related quality of life in chronic liver disease: the impact of type and severity of disease. Am J Gastroenterol 2001;96:21992205. 10. Younossi ZM. Chronic liver disease and health-related quality of life. Gastroenterology 2001;120:305307. 11. Kamath PS, Wiesner RH, Malinchoc M, et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001;33:464470. 12. Wiesner R, Edwards E, Freeman R, et al. Model for end-stage liver disease (MELD) and allocation of donor livers. Gastroenterology 2003;124:9196. 13. Rodrguez-Artalejo F, Guallar-Castilln P, Pascual CR, et al. Health-related quality of life as a predictor of hospital readmis- sion and death among patients with heart failure. Arch Intern Med 2005;165:12741279. 798 KANWAL ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7, No. 7 14. Cunningham WE, Crystal S, Bozzette S, et al. The association of health-related quality of life with survival among persons with HIV infection in the United States. J Gen Intern Med 2005;20:2127. 15. Mapes DL, Lopes AA, Satayathum S, et al. Health-related quality of life as a predictor of mortality and hospitalization: the Dialysis Outcomes and Practice Patterns Study (DOPPS). Kidney Int 2003;64:339349. 16. Karvonen-Gutierrez CA, Ronis DL, Fowler KE, et al. Quality of life scores predict survival among patients with head and neck can- cer. J Clin Oncol 2008;26:27542760. 17. Healy LA, Ryan AM, Moore J, et al. Health-related quality of life assessment at presentation may predict complications and early relapse in patients with localized cancer of the esophagus. Dis Esophagus 2008;21:522528. 18. Mauer M, Stupp R, Taphoorn MJ, et al. The prognostic value of health-related quality-of-life data in predicting survival in glioblas- toma cancer patients: results from an international randomised phase III EORTC Brain Tumour and Radiation Oncology Groups, and NCIC Clinical Trials Group study. Br J Cancer 2007;97:302 307. 19. McCarter H, Furlong W, Whitton AC, et al. Health status measure- ments at diagnosis as predictors of survival among adults with brain tumors. J Clin Oncol 2006;24:36363643. 20. Kanwal F, Spiegel BMR, Hay RD, et al. Prospective validation of the Short Form Liver Disease Quality of Life instrument. Aliment Pharmacol Ther 2008;28:10881101. 21. Gralnek IM, Hays RD, Kilbourne A, et al. Development and eval- uation of the Liver Disease Quality of Life instrument in persons with advanced, chronic liver disease: the LDQOL 1.0. Am J Gas- troenterol 2000;95:35523565. 22. Jobson JD. Maximum R2 method. In: Jobson JD, ed. Applied multivariate data analysis: regression and experimental design. New York, NY: Springer, 1991:270. 23. Ware JE, Snow KK, Kosinski MK, et al. Scoring the SF-36. In: SF-36 Health Survey: Manual and Interpretation Guide. Boston, MA: Nimrod Press, 1992. 24. RAND 36-item health survey v2 (SF-36 v2) and UCLA prostate cancer index health related quality of life scoring instructions. Availableat: http://www.uclaurology.com/site_uo/pdf/PCI_long_ scoring.pdf. Accessed February 2009. 25. Biggins SW, Kim WR, Terrault NA, et al. Evidence-based incorpo- ration of serum sodium concentration into MELD. Gastroenterol- ogy 2006;130:16521660. 26. Biggins SW, Rodriguez HJ, Bacchetti P, et al. Serum sodium predicts mortality in patients listed for liver transplantation. Hepatology 2005;41:3239. 27. Londoo MC, Crdenas A, Guevara M, et al. MELD score and serum sodium in the prediction of survival of patients with cirrho- sis awaiting liver transplantation. Gut 2007;56:12831290. 28. Charlson ME, Pompei P, Ales KL, et al. A new method of classi- fying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373383. 29. Pencina MJ, DAgostino RB. Overall C as a measure of discrimi- nation in survival analysis: model specic population value and condence interval estimation. Stat Med 2004;23:21092123. 30. DiMartini A, Dew MA, Javed L, et al. Pretransplant psychiatric and medical comorbidity of alcoholic liver disease patients who re- ceived liver transplant. Psychosomatics 2004;45:517523. 31. Rocca P, Cocuzza E, Rasetti R, et al. Predictors of psychiatric disorders in liver transplantation candidates: logistic regression models. Liver Transpl 2003;9:721726. 32. Gonseth J, Guallar-Castilln P, Banegas JR, et al. The effective- ness of disease management programs in reducing hospital re-admission in older patients with heart failure: a systematic review and meta-analysis of published reports. Eur Heart J 2004; 25:15701595. 33. Ripoll C, Baares R, Rincn D, et al. Inuence of hepatic venous pressure gradient on the prediction of survival of patients with cirrhosis in the MELD Era. Hepatology 2005;42:793801. Reprint requests Address requests for reprints to: Brennan M. R. Spiegel, MD, MSHS, 11301 Wilshire Boulevard, Building 115, Room 215, Los Angeles, California 90073. e-mail: bspiegel@mednet.ucla.edu; fax: (310) 268- 4510. Conicts of interest The authors disclose no conicts. Funding Dr Kanwal is supported by Veterans Affairs Health Services Re- search and Development (HSR&D) Investigator Initiated Research Award IIR-07-111. Dr Spiegel is supported by a Veterans Affairs Health Services Research and Development (HSR&D) Career Develop- ment Transition Award (RCD 03-179-2) and by the CURE Digestive Disease Research Center (NIH 2P30 DK 041301-17). Dr Hays is supported by the UCLA Center for Health Improvement in Minority Elderly/Resource Centers for Minority Aging Research, NIH/NIA/ NCMHD, under grant P30AG021684-07. This study was supported by a Clinical Research Grant from the American College of Gastroenter- ology. The opinions and assertions contained herein are the sole views of the authors and are not to be construed as ofcial or as reecting the views of the Department of Veteran Affairs. July 2009 HRQOL AND MORTALITY IN CIRRHOSIS 799