Cost-Effectiveness of Screening for Hepatopulmonary Syndrome in Liver Transplant Candidates

Transcription

1 LIVER TRANSPLANTATION 13: , 2007 ORIGINAL ARTICLE -Effectiveness of Screening for Hepatopulmonary Syndrome in Liver Transplant Candidates D. Neil Roberts, 1 Miguel R. Arguedas, 2 and Michael B. Fallon 2 1 Department of Medicine and 2 Division of Gastroenterology and Hepatology, University of Alabama at Birmingham, Birmingham, AL The hepatopulmonary syndrome (HPS) is present in 15-20% of patients with cirrhosis undergoing orthotopic liver transplantation (OLT) evaluation. Both preoperative and post-olt mortality is increased in HPS patients particularly when hypoxemia is severe. Screening for HPS could enhance detection of OLT candidates with sufficient hypoxemia to merit higher priority for transplant and thereby decrease mortality. However, the cost-effectiveness of such an approach has not been assessed. Our objective was to perform a cost-effectiveness analysis from a third-party payer s perspective of screening for HPS in liver OLT candidates. The costs and outcomes of 3 different strategies were compared: (1) no screening, (2) screening patients with a validated dyspnea questionnaire, and (3) screening all patients with pulse oximetry. Arterial blood gas analyses and contrast echocardiography were performed in patients with dyspnea or a pulse oximetry (SpO 2 ) 97% to define the presence of HPS. A Markov model was constructed simulating the natural history of cirrhosis in a cohort of patients 50 years old over a time horizon of their remaining life expectancy. Transition probabilities were obtained from published data available through Medline and U.S. vital statistics. s represented Medicare reimbursement data at our institution. s and health effects were discounted at a 3% annual rate. No screening was associated with a total cost of $291,898 and a life expectancy of years. Screening with pulse oximetry was associated with a cost of $299,719 and a life expectancy of years. Screening patients with the dyspnea-fatigue index was associated with a cost and life expectancy of $300,278 and years, respectively. The incremental cost-effectiveness ratio of screening with pulse oximetry (compared to no screening) was $6,867 per life year gained, whereas that of the dyspnea-fatigue index (compared to pulse oximetry) was $55,900 per life year gained. The cost-effectiveness of screening depended on the prevalence and severity of HPS, and the choice of screening strategy was dependent on the sensitivity of the screening modality. In conclusion, screening for HPS, especially with pulse oximetry, is a cost-effective strategy that improves survival in transplant candidates predominantly by targeting the transplant to the subgroup of patients most likely to benefit. The utility of screening depends on the prevalence and severity of HPS in the target population. Liver Transpl 13: , AASLD. Received May 5, 2006; accepted July 23, See Editorial on Page 183 Patients with cirrhosis develop complications related to progressive hepatocellular dysfunction and/or portal hypertension and are at risk for the development of hepatocellular carcinoma. 1-3 In addition to substantial morbidity and mortality, these complications account for a significant proportion of health care expenditures in the United States and worldwide. Given that cirrhosis is generally the end result of processes leading to liver injury, and specific therapies that reverse fibrosis are limited, most therapeutic options center on preventing the development of or Abbreviations: HPS, hepatopulmonary syndrome; PaO 2, arterial oxygen tension; OLT, orthotopic liver transplantation; MELD, model for end-stage liver disease; SpO 2, pulse oximetry; LY, life year; ICER, incremental cost-effectiveness ratio. Preliminary work presented as a poster presentation (Digestive Disease Week, May 14-19, 2005, Chicago, IL) and published in abstract form (Gastroenterology 2005;128:A-735). Address reprint requests to Miguel R. Arguedas, MD, MPH, UAB Liver Center, 417 Lyons Harrison Bldg., th St. South, Birmingham, AL Telephone: ; FAX: ; Arguedas@uab.edu DOI /lt Published online in Wiley InterScience ( American Association for the Study of Liver Diseases.

2 SCREENING FOR HPS IN OLT CANDIDATES 207 treating specific complications once they have occurred in order to prolong patient survival and improve quality of life. The hepatopulmonary syndrome (HPS) has emerged as an important complication of cirrhosis and portal hypertension. HPS results from intrapulmonary vasodilatation resulting in shunting and hypoxemia and is seen in 15-20% of patients with chronic liver disease. Studies demonstrate that survival is decreased in patients with HPS compared to patients without HPS and a similar severity of underlying liver disease. 4,5 To date, there are no proven effective therapies for HPS, with the exception of liver transplantation. Although gas exchange abnormalities associated with HPS generally resolve following transplantation, patients with severe hypoxemia (arterial oxygen tension [PaO 2 ] 50 mm Hg) and significant shunting prior to transplantation have increased postoperative mortality. 4-6 In 1 prospective study, 67% of patients with HPS and a PaO 2 50 mm Hg died within 1 year of orthotopic liver transplantation (OLT). 6 For these reasons, patients with HPS and an arterial PaO 2 60 mm Hg are eligible to receive model for end-stage liver disease (MELD) exception points in the United States. Screening of patients being evaluated for liver transplantation would be expected to enhance early detection of HPS and improve survival by optimizing the timing of liver transplantation. However, routine screening for HPS is not performed in all transplant centers, and whether screening for HPS is cost-effective in this setting has not been evaluated. The aim of our study was to examine the costs and outcomes associated with formal screening for HPS in patients presenting for initial transplant evaluation. PATIENTS AND METHODS The Model We developed a Markov model using specialized software (DATA 3.5, TreeAge, Williamstown, MA) to estimate the expected lifetime costs and outcomes associated with screening for HPS in a hypothetical cohort of 50-year-old patients with cirrhosis presenting for initial liver transplant evaluation. In a Markov model, members of a cohort are divided among several mutually exclusive health states, and movements of the cohort across these states are modeled over time (time horizon) into a series of evenly spaced periods or cycles. The health states are defined to capture the salient characteristics of the disease and the interventions under consideration. At the end of each cycle, the members of the cohort are reallocated across health states guided by transition probabilities that reflect the natural history of the disease in question until all members of the cohort have reached an absorbing state (i.e., death) or the set time horizon has completed. The effects of an intervention can be modeled by modifying the appropriate transition probabilities in the model. The total number of years lived by members of the cohort (life expectancy) and average Figure 1. Simplified Markov model. CP-A, Child-Turcotte- Pugh class A; CP-B/C, Child-Turcotte-Pugh B/C. Arrows represent transitions between health states. lifetime health care costs are accrued at the end of the analysis. 7 Figure 1 offers a simplified representation of the Markov model constructed for our analysis. The Strategies We compared the costs and outcomes associated with the following 3 strategies. No screening: In this strategy, the cohort does not undergo screening for HPS. Dyspnea-fatigue index: The entire cohort is screened with a validated dyspnea questionnaire. 8 For patients with positive results, arterial blood gas analysis and contrast echocardiography was undertaken to define the presence of HPS. 9 Pulse oximetry: The entire cohort undergoes screening with pulse oximetry. For patients with a pulse oximetry (SpO 2 )of 97%, arterial blood gas analysis and contrast echocardiography was undertaken to define the presence of HPS. Clinical Data In our model, the cohort consisted of patients with cirrhosis who were referred for transplant evaluation. During each cycle, patients could occupy 1 of several health states designed to incorporate the natural history of cirrhosis with progression of the underlying liver disease, development of complications arising from cirrhosis and portal hypertension (HPS, variceal bleed, ascites, spontaneous bacterial peritonitis), development of hepatocellular carcinoma, and death. Transition probabilities were obtained from the literature using a Medline search using the following terms: hepatopulmonary syndrome, natural history and complications of cirrhosis, hepatocellular carcinoma, and liver transplantation. Table 1 shows the probability estimates for our baseline analysis. Population-based mortality rates adjusted for age, gender, and race were obtained from published U.S. vital statistics. 10 The cohort began the simulation with the diagnosis of cirrhosis, and we assumed all patients to be candidates for liver transplantation. Depending on the strategy evaluated, patients would either test positive or neg-

3 208 ROBERTS ET AL. TABLE 1. Model Variables: Baseline Values and Range used in Sensitivity Analysis Variable Value Range References s Spontaneous bacterial peritonitis $8,784 $ ,736 UAB* Variceal bleed $24,940 $7, ,999 UAB* Liver transplantation (year 1) $205,820 $95,939 1,000,000 UAB* Liver transplantation (following year 1) $20,707 $10, ,021 UAB* EGD $608 $ UAB* Pulse oximetry $75 $ UAB* Contrast trans-thoracic echocardiogram $221 $ UAB* Arterial blood gas $201 $ UAB* Hospice (55 days as median) $49,100 $7,454 49,100 UAB* Hepatocellular carcinoma $13,470 $8, ,330 UAB* Norfloxacin $1,416 $1,016 1, Beta-blockers (propranolol) $78 $ Diuretics (furosemide and spironolactone) $972 $ Lactulose $480 $ Probabilities Transition from compensated to decompensated cirrhosis 3% 2 6% 1, 3, 21 Prevalence of HPS 18% 1 20% 4, 9, 15, 18, 22 Prevalence of dyspnea in patients with HPS 100% % 9, 17, 23 Prevalence of dyspnea in patients without HPS 33% 25 46% 9, 17 Sensitivity pulse oximetry 96% % 16 Specificity pulse oximetry 75% % 16 OLT 35% 3 50% 12 OLT (HPS, PaO 2 60 mm Hg) 61% 45 77% 12 Proportion of patients with PaO 2 60 mm Hg 47% 24 47% 6, 16, 19 Proportion of patients with PaO mm Hg 40% 24 40% 6, 16, 19 Proportion of patients with PaO 2 50 mm Hg 13% 5 40% 6, 16, 19 Transition from PaO 2 50 mm Hg to 50 mm Hg 28% 10 30% 5 Complications post-olt (year 1) 30% 10 50% 6, 12, 24, 25 Death from complications post-olt 50% % 6, 24, 25 Complications post-olt (year 1, PaO 2 50 mm Hg) 85% % 6, 24, 25 Death from complications post-olt (PaO 2 50 mm Hg) 78% % 6, 24, 25 Death in compensated cirrhosis 1.9% 1 5% 2, 3, 26 Death in decompensated cirrhosis 10% 8 40% 2, 3, 26 Death following first year post-olt 3% 1 5% 12 Death in compensated cirrhosis (PaO 2 50 mm Hg) 15% 1 25% 4, 5 Death in decompensated cirrhosis (PaO 2 50 mm Hg) 25% 10 50% 4, 5 Abbreviation: EGD, esophagogastroduodenoscopy. *s based on data from the University of Alabama (UAB) Hospital/Health Services Foundation business office. ative upon screening. Patients would then be categorized according to the results of confirmatory testing as true positives, false positives, true negatives, and false negatives, and the disease would be modeled accordingly. A simplified diagram of the screening algorithm is represented in Figure 2. In any given cycle, patients may develop complications from cirrhosis and portal hypertension (jaundice, coagulopathy, encephalopathy, ascites, variceal hemorrhage, and spontaneous bacterial peritonitis), may remain stable, or may undergo liver transplantation. Death may result from complications of liver disease, other causes (age-, gender-, race-adjusted mortality rates) or from transplantation-related complications. Patients with HPS may develop worsening gas exchange abnormalities, and their pretransplantation and posttransplantation survival would be adjusted according to their PaO 2. Assumptions Assumptions were made because of the limited information available in the medical literature. 11 Whenever possible, these assumptions were derived from acceptable or justifiable clinical decisions representative of expert opinions at our institution, and the validity of these assumptions was subjected to sensitivity analysis. (1) In our baseline analysis, we considered a population of cirrhotic patients for whom liver transplant would be a viable option until age 70. According to data from our transplant center, 80% of patients who present for initial transplant evaluation are classified as Child-Turcotte- Pugh class B/C patients with the remaining patients having Child-Turcotte-Pugh class A cirrhosis. (2) According to the screening test characteristics, positive results may represent true positives or false

4 SCREENING FOR HPS IN OLT CANDIDATES 209 Figure 2. Outline of screening strategies. *Patients with positive confirmatory tests would then be categorized according to PaO 2 levels. Patients with a PaO 2 >60 mm Hg would be listed for OLT, and patients with a PaO 2 <60 mm Hg would receive MELD exception points. We also performed an analysis in which patients with a PaO 2 >60 mm Hg would be listed for OLT. Patients with a PaO 2 between 50 and 60 mm Hg would receive MELD exception points, whereas those with severe HPS (<50 mmhg) would not be considered OLT candidates. LT, liver transplantation; SpO 2, ; ABGs, arterial blood gas analysis; CE, contrast echocardiography. positives. In patients who tested positive with the dyspnea-fatigue index and pulse oximetry, evaluation with arterial blood gas analysis and contrast echocardiography would then be performed. If negative, patients would be placed in the cohort corresponding to a PaO 2 50 mm Hg, between 50 and 60 mm Hg and/or 60 mm Hg, with the natural history modeled according to the severity of hypoxemia in HPS. Patients with hypoxemia ( 60 mm Hg) and a negative contrast echocardiography would be considered to have intrinsic lung disease and thus would not be considered candidates for transplantation. (3) Positive results would be grouped according to degree of hypoxemia of PaO 2 60 mm Hg, PaO 2 between 50 and 60 mm Hg, and PaO 2 50 mm Hg. Patients with a PaO 2 60 mm Hg would be listed for transplant without exception points, whereas patients with PaO 2 60 mm Hg would receive additional MELD points for higher priority with a probability of OLT of 61% at 1 year. (4) Given the higher probability of postoperative complications and mortality in patients with severe HPS (PaO 2 50 mm Hg), we evaluated an alternate scenario as part of our sensitivity analysis to explore the impact of excluding patients with severe disease where postoperative mortality may be sufficiently high to preclude routine use of OLT. In this scenario, the remainder of the cohort was subject to the same assumptions as in assumption 3. (5) We assumed all patients who developed complications of cirrhosis and portal hypertension, such as ascites, variceal bleeding, spontaneous bacterial peritonitis, and encephalopathy, would be placed on diuretic therapy, secondary prophylaxis with beta-blockers, antibiotics, and lactulose, respectively. (6) For patients without HPS, we assumed a probability of undergoing OLT at 1 year of 35% based on available United Network for Organ Sharing data. 12 s All direct costs are reported in 2004 U.S. dollars. The costs used in the model represent the average Medicare reimbursement rates at our institution according to medical and surgical diagnosis-related groups and Current Procedural Terminology codes. We elected to use Medicare reimbursement data as a surrogate for commercial insurers because of the greater amount of regional variability in reimbursement rates associated with commercial payers. Indirect costs were not calculated. Table 1 shows the cost estimates for our baseline analysis. -Effectiveness Analysis The model was used to estimate patients life expectancy in terms of life year (LY) and expected health care costs under each of the 3 strategies. These strategies were then compared using the methods of incremental cost-effectiveness analysis in order to calculate the incremental cost-effectiveness ratio (ICER). As recommended by the U.S. Panel on -Effectiveness in Health and Medicine, all costs and benefits were discounted to present value using an annual discount rate of 3%. 13 Sensitivity Analyses Sensitivity analyses were performed to explore the degree to which the results of our baseline analysis were influenced by uncertainty regarding parameter values used in the model. In 1-way sensitivity analyses, results were recalculated as the values of model parameters were separately varied. In addition, 2-way sensitivity analyses were performed to examine the results of simultaneously varying select pairs of variables that were found to be influential in 1-way sensitivity analyses.

5 210 ROBERTS ET AL. TABLE 2. Results of Analysis Screening Strategy Incremental Life Expectancy Incremental Life Expectancy ICER Undiscounted Analysis No screening $421, Pulse oximetry $428,111 $6, years $5,052 Dyspnea-fatigue index $428,753 $ years $42,800 Discounted Analysis (3%) No screening $291, Pulse oximetry $299,719 $7, years $6,867 Dyspnea-fatigue index $300,278 $ years $55,900 Finally, a probabilistic sensitivity analysis was performed using first-order Monte Carlo simulation. We performed a separate analysis to evaluate the influence of the severity of liver disease on the costs and outcomes associated with screening for HPS by varying the proportion of Child-Turcotte-Pugh A and B/C patients. As mentioned above, we also explored the costeffectiveness of screening for HPS in a scenario in which patients with a PaO 2 50 mm Hg would be excluded from transplantation. RESULTS Baseline Analysis The undiscounted analysis demonstrated that the strategy associated with the lowest overall costs was no screening ($421,240) followed by, in order of increasing costs, screening with pulse oximetry ($428,111) and screening with the dyspnea-fatigue index ($428,753). In terms of years of life saved, the least effective strategy was no screening (16.80 LYs) followed by, in order of increasing effectiveness, screening with pulse oximetry (18.16 LYs) and screening with the dyspnea-fatigue index ( LYs). In the discounted analysis, the no-screening strategy was associated with a total cost of $291,898 and a life expectancy of years. Screening with pulse oximetry was associated with a cost of $299,719 and a lifeexpectancy of years whereas screening patients with dyspnea was associated with a cost and life expectancy of $300,278 and years, respectively. The incremental cost-effectiveness ratio (ICER) of screening with pulse oximetry compared to no screening was $6,867 per additional LY gained. The ICER of screening with the dyspnea-fatigue index compared to the no screening strategy was $7,350 per additional LY gained. Finally, the ICER of screening with the dyspnea-fatigue index compared to a strategy based on screening with pulse oximetry was $55,900 per additional LY gained (Table 2). Sensitivity Analyses The results of the analysis remained robust after varying the majority of cost and probability estimates. The Figure 3. Effects of the time horizon on expected costs. s are presented in thousands. White bars denote no screening. Black bars denote pulse oximetry. Gray bars denote dyspneafatigue index. prevalence of HPS did emerge as an important variable that influenced the results. At a prevalence of HPS less than 4%, the ICER of both screening strategies exceeded $50,000 per year of life saved, and at a prevalence less than 2%, the ICER of both screening strategies exceeded $100,000 per year of life saved compared to no screening (Figs. 3 and 4). The ICER of screening for HPS was also influenced by the severity of HPS and underlying liver disease. Specifically, as long as the proportion of patients with HPS receiving MELD exception points (those with PaO 2 60 mm Hg) exceeded 20%, the ICER of both screening strategies remained below the threshold of $50,000 per year of life saved compared to no screening. In addition, the ICER of screening strategies remained below this threshold as long as the pre-olt mortality in patients with decompensated liver disease (Child-Turcotte-Pugh class B/C) and HPS was greater than 10% per year. Sensitivity analysis of the probability of complications and mortality following OLT in patients with severe HPS within the range specified in Table 1 did not greatly affect the results. At extreme values, though, the ICER of both screening strategies for HPS compared to no screening became greater than $50,000/LY only if

6 SCREENING FOR HPS IN OLT CANDIDATES 211 Figure 4. Effects of the prevalence of HPS on expected costs. s are presented in thousands. Dash-dot line denotes no screening. Solid line denotes pulse oximetry. Dotted line denotes dyspnea-fatigue index. Figure 5. Effects of the prevalence of HPS on life expectancy. Life expectancy is presented in years. Dash-dot line denotes no screening. Solid line denotes pulse oximetry. Dotted line denotes dyspnea-fatigue index. the probability of death following transplantation in patients with PaO 2 50 mm Hg exceeded 96%. We created the model with 72% of HPS patients maintaining a PaO 2 greater than 60 mm Hg from year to year; however, if 90% or more of cirrhotics maintain a PaO 2 greater than 60 mm Hg, yearly surveillance with either strategy is not justified, given the greater associated costs. Within the range chosen for sensitivity analyses, the only variables that would influence the choice of a specific screening strategy was the sensitivity of pulse oximetry for detecting a PaO 2 60 mm Hg. As long as the sensitivity remains above 92%, the ICER of screening with the dyspnea-fatigue index remains above $50,000 per year of life saved. The ICER of screening with pulse oximetry exceeded that of the dyspnea-fatigue index only at a very high cost for oximetry ($425). Effects of the Time Horizon and Severity of Liver Disease The effects of varying the time horizon on costs and life expectancy are represented in Figures 5 and 6, respectively. The ICER of screening strategies for HPS exceed $100,000/LY and $50,000/LY at a time horizon less than 2 and 5 years, respectively, compared to no screening. Thereafter, screening becomes cost-effective as the gain in life expectancy counter balances the upfront costs associated with screening. As expected, the advantages of screening on life expectancy compared to no screening are observed and maintained after the first year. We performed a separate analysis to evaluate the influence of the severity of liver disease on the costs and outcomes associated with screening for HPS (Table 3). Increasing the proportion of patients with decompensated liver disease (Child-Turcotte-Pugh class B/C) led Figure 6. Effects of the time horizon on life expectancy. White bars denote no screening. Black bars denote pulse oximetry. Gray bars denote dyspnea-fatigue index. to improved cost-effectiveness ratios as a result of a greater gain in life expectancy in proportion to the increase in costs. In contrast, although screening only patients with Child-Turcotte-Pugh A liver disease led to improved life expectancy, the ICER was significantly higher than the ICER of screening only patients with Child-Turcotte-Pugh class B/C disease. Assuming all patients evaluated are Child-Turcotte-Pugh A, screening with pulse oximetry is associated with an ICER compared to no screening of $32,990 per LY gained vs.

7 212 ROBERTS ET AL. TABLE 3. -Effectiveness Analyses Depending on Severity of Liver Disease (Child-Turcotte-Pugh Class) Severity Strategy (USD) Marginal (USD) Effectiveness (LYs) Marginal Effectiveness (LYs) ICER (USD/LYs) All CP-A No screening $110, Pulse oximetry $170,815 $60, $32,990 Dyspnea-fatigue $171,368 $ $55,300 index CP-A (50%) No screening $223,849 $27, $19,665 CP-B/C (50%) Pulse oximetry $251,380 $ $55,700 Dyspnea-fatigue $251, index All CP-B/C No screening $330, Pulse oximetry $331,944 $1, $1,723 Dyspnea-fatigue index $332,506 $ $56,200 Abbreviations: USD, U.S. dollars; CP-A, Child-Turcotte-Pugh class A; CP-B/C, Child-Turcotte-Pugh class B/C. TABLE 4. Results of Analysis (Scenario in Which Patients With Severe HPS are Excluded From Transplantation) Incremental Screening Strategy Expectancy Expectancy ICER Undiscounted Analysis Pulse-oximetry $410, Dyspnea-fatigue index $410,676 $ years $57,700 No screening $421,240 $10, years Dominated Discounted Analysis (3%) Pulse-oximetry $284, Dyspnea-fatigue index $285,200 $ years $163,333 No screening $291,898 $6, years Dominated Life Incremental Life $1,723 per LY gained by screening only Child-Turcotte- Pugh class B/C patients with pulse oximetry. The ICER of screening with the dyspnea-fatigue index, compared to no screening, is $33,111 per LY gained Child-Turcotte-Pugh class A patients vs. $2,274 per LY gained in Child-Turcotte-Pugh class B/C patients. Patients With Severe HPS (PaO 2 50 mm Hg) Excluded From Transplant In an alternate scenario, we explored the effects of excluding patients with severe HPS on screening. In this scenario, no screening was the strategy associated with the highest total cost ($291,898) and the lowest life expectancy ( years). Among the screening strategies, screening patients with the dyspnea-fatigue index was associated with a cost and life expectancy of $285,200 and years, respectively, whereas screening with pulse oximetry was associated with the lowest cost ($284,710) and a life expectancy of years. No screening was therefore strongly dominated by both screening strategies, as it was the most expensive and least effective modality. The ICER of screening with the dyspnea-fatigue index compared to a strategy based on screening with pulse oximetry was $163,333 per additional LY gained (Table 4). However, despite the cost savings associated with this alternate scenario, the life expectancy associated with screening was significantly lower than in the baseline analysis. Monte Carlo Simulation In a Monte Carlo simulation analysis using 10,000 trials, no screening remained the least expensive strategy in 79% of trials, and among the screening strategies, pulse oximetry remained less expensive than screening based on the dyspnea-fatigue index 54% of the time. In terms of life expectancy, no screening was less effective than either screening strategy in 91% of the trials. Screening with pulse oximetry was less effective than screening based on the dyspnea-fatigue index 55% of the time. DISCUSSION The morbidity and mortality of cirrhosis is increasing due to the expanding burden of chronic hepatitis C infection and to the obesity epidemic resulting in nonalcoholic fatty liver disease. The consequences of cirrhosis are measured not only in terms of human suffering, but also in terms of health care resource utilization. The biological and economical impact of

8 SCREENING FOR HPS IN OLT CANDIDATES 213 many complications of chronic liver disease and portal hypertension such as variceal bleeding, spontaneous bacterial peritonitis, and hepatocellular carcinoma have been previously described. 14 However, the clinical and economical effects of HPS have not been thoroughly evaluated. Potential reasons include the insidious onset and subtle symptoms associated with HPS leading to under-recognition, the perceived rarity of this complication, and the fact that medical therapies are not currently available. Recent studies demonstrate that HPS occurs in 10-20% of patients with cirrhosis and that pre- and post-olt mortality rates are higher in patients with HPS compared to non-hps patients with similar severity of liver disease, particularly in those with severe hypoxemia. 4-6 Therefore, patients with HPS and significant hypoxemia are eligible for increased transplant priority in the United States. These observations support the concept that evaluation of the clinical and economical consequences of screening programs for HPS are important. The present study addresses the cost-effectiveness of screening for HPS using commonly available modalities. The results of our analysis, under baseline assumptions, demonstrate that screening for HPS in liver transplant candidates is a cost-effective intervention. Pulse oximetry was associated with the lowest ICER ($6,867 per LY gained). Screening with the dyspneafatigue index, although associated with a marginally better life-expectancy compared to pulse oximetry screening, results in an ICER of $55,900 per additional LY gained, which may be considered expensive by some third-party payers. Both screening strategies were associated with a significantly higher life expectancy than no screening. Within the confines of the assumptions of our model, the results support that screening for HPS is economically feasible and associated with improved survival. On sensitivity analysis, several variables emerged as potential factors that may influence the current results. The most important variables were the prevalence and severity of HPS and the severity of underlying liver disease. As long as the prevalence of HPS was greater than 4%, screening was cost-effective. As the prevalence of HPS increased, the economic viability and life expectancy conferred by screening improved. This improvement resulted from assigning MELD exception points among patients identified as having HPS relative to those for whom the diagnosis remained unknown. In all studies published to date, the prevalence of HPS in those evaluated for OLT has been significantly greater than 4%. 4,15-18 However, the number of patients with HPS undergoing OLT evaluation who have comorbidities that preclude transplantation has not been clearly defined and is an area for future analysis. The severity of HPS was also important in our analysis. As long as at least 20% of HPS patients had a PaO 2 60 mm Hg (eligible for MELD exception), the ICER of screening was less than $50,000 per additional LY. In single-center studies, the proportion of patients with HPS and a PaO 2 60 mm Hg has been reported to be between 37-76%. 4,16,19 An ongoing multicenter study of HPS in patients evaluated for OLT should provide additional data. Finally, the severity of liver disease also influenced our findings. As long as mortality associated with HPS in Child-Turcotte-Pugh B/C cirrhotics was greater than 10% per year, screening remained cost-effective. Two single-center studies have found a mortality between 25 and 30% in this subset of patients with HPS, supporting the validity of our assumption. 4,5 ) We also found that the cost-effectiveness of screening improved at advancing stages of liver disease, predominately because a subset of patients with HPS and advanced liver disease will receive MELD exception points and undergo OLT earlier, relative to no screening. In patients with advanced liver disease who are not considered OLT candidates, screening for HPS would not be cost-effective or improve life expectancy, given the limited survival and the lack of effective medical therapies for HPS. The rate of progression of hypoxemia in patients with HPS is another important factor in the natural history of disease and in modeling outcomes. To date, 1 singlecenter study has evaluated serial arterial blood gas values in a small cohort of HPS patients. 5 In this cohort, PaO 2 values declined over time in 85% of patients, regardless of the baseline PaO 2 value, with an average decline of 5 mm Hg per year. In our analysis, we included a variable to reflect worsening PaO 2 levels in patients with HPS in order to model the impact of the natural history of the disease and to adjust the probability of complications and mortality according to PaO 2 levels. Since our protocol was confined to 1-time screening and did not include surveillance, including the impact of progression of HPS over time in our model biases the life expectancy analysis against screening. Nonetheless, sensitivity analysis of the progression variable demonstrated that as long as at least 10% of HPS patients develop a PaO 2 level 60 mm Hg each year, 1-time screening remains cost-effective as a result of increased life expectancy, relative to no screening. In 2 single-center studies, from 14 to 26% of HPS patients had baseline PaO 2 values between 60 and 65 mm Hg, 16,19 and if PaO 2 values in these patients decline on average 5 mm Hg per year, screening should remain cost-effective. Clearly defining the rate of progression of hypoxemia in HPS in a larger cohort of patients will be crucial to determining whether surveillance following initial screening would further increase life expectancy and cost-effectiveness. In our alternate scenario, we modeled excluding patients with severe hypoxemia from consideration of OLT due to the observation that postoperative mortality may be significantly increased in this group. 4-6 Although in this scenario screening was a cost-saving intervention, life expectancy was significantly lower than in the baseline screening strategies. These findings imply that identifying a particular subgroup of patients in whom HPS is sufficiently severe that OLT will not improve life expectancy may improve cost-effectiveness without decreasing life expectancy. This will require collection of detailed and standardized data on each patient with HPS that undergoes OLT. In conclusion, our results support that 1-time

9 214 ROBERTS ET AL. screening for HPS, especially with pulse oximetry in patients with cirrhosis evaluated and eligible for OLT is a cost-effective intervention that improves life expectancy. Our analysis also highlights areas of uncertainty that could influence our baseline assumptions, particularly the influence of the severity of HPS on pre-olt survival. In addition, we have identified the need for additional prospective data regarding the rate of progression of hypoxemia in HPS and the influence of HPS severity on OLT outcomes to define the role of surveillance and to determine whether there is a subset of HPS patients for whom OLT would increase costs without a survival benefit. REFERENCES 1. Fattovich G, Giustina G, Degos F, Tremolada F, Diodati G, Almasio P, et al. Morbidity and mortality in compensated cirrhosis type C: a retrospective follow-up study of 384 patients. Gastroenterology 1997;112: Fattovich G, Pantalena M, Zagni I, Realdi G, Schalm SW, Christensen E. Effect of hepatitis B and C virus infections on the natural history of compensated cirrhosis: a cohort study of 297 patients. Am J Gastroenterol 2002;97: Gines P, Qientero E, Arroyo V, Teres J, Bruguera M, Rimola A, et al. Compensated cirrhosis: natural history and prognostic factors. Hepatology 1987;7: Schenk P, Schoniger-Hekele M, Fuhrmann V, Madl C, Silberhumer G, Muller C. Prognostic significance of the hepatopulmonary syndrome in patients with cirrhosis. Gastroenterology 2003;125: Swanson K, Wiesner R, Krowka M. Natural history of hepatopulmonary syndrome: impact of liver transplantation. Hepatology 2005;41: Arguedas M, Abrams GA, Krowka MJ, Fallon MB. Prospective evaluation of outcomes and predictors of mortality in patients with hepatopulmonary syndrome undergoing liver transplantation. Hepatology 2003;37: Sonnenberg F, Beck J. Markov models in medical decision making: a practical guide. Med Decis Making 1993: Mahler D, Weinberg D, Wells C, Feinstein A. The measurement of dyspnea. Contents, interobserver agreement, and physiologic correlates of two new clinical indexes. Chest 1984;85: Sood G, Fallon MB, Niwas S, Tutton T, van Leeuwen DJ, Bloomer JR, et al. Utility of a dyspnea-fatigue index for screening liver transplant candidates for hepatopulmonary syndrome [abstract]. Hepatology 1998;28: Centers for Disease Control and Prevention, National Vital Statistics Reports. United States life Tables. Natl Vital Stat Rep 1999;47: Rodriguez-Roisin R, Krowka MJ, Herve P, Fallon MB, on behalf of the ERS Task Force Pulmonary-Hepatic Vascular Disorders Scientific Committee ERS Task Force PHD Scientific Committee. Pulmonary-hepatic vascular disorders (PHD). Eur Respir J 2004;24: United Network for Organ Sharing: Available at Accessed: February 9, Lipscomb J, Weinstein M, Torrance G. Time preference. In: Gold MR, Siegel JE, Russell LB, Weinstein MC. Effectiveness in Health and Medicine. New York, NY: Oxford University Press;1996: Davis G, Albright J, Cook S, Rosenberg D. Projecting future complications of chronic hepatitis C in the United States. Liver Transpl 2003;9: Abrams GA, Jaffe CC, Hoffer PB, Binder HJ, Fallon MB. Diagnostic utility of contrast echocardiography and lung perfusion scan in patients with hepatopulmonary syndrome. Gastroenterology 1995;109: Abrams GA, Sanders MK, Fallon MB. Utility of pulse oximetry in the detection of arterial hypoxemia in liver transplant candidates. Liver Transpl 2002;8: Martinez G, Barbera J, Visa J, Rimola A, Parc J, Roca J, et al. Hepatopulmonary syndrome in candidates for liver transplantation. J Hepatol 2001;34: Whyte M, Hughes J, Peters A, Ussov W, Patel S, Burroughs A. Analysis of intrapulmonary right to left shunt in hepatopulmonary syndrome. J Hepatol 1998;29: Krowka M, Wiseman G, Burnett O, Spivey J, Therneau T, Porayko M, Wiesner R. Hepatopulmonary syndrome: a prospective study of relationships between severity of liver disease, PaO 2 response to 100% oxygen, and brain uptake after 99mTc MAA lung scanning. Chest 2000;118: Drug Topics Red Book. Montvale, NJ: Medical Economics Company; Benvegnu L, Gios M, Boccato S, Alberti A. Natural history of compensated viral cirrhosis: a prospective study on the incidence and hierarchy of major complications. Gut 2004;53: Martinez GP, Barbera JA, Visa J, Rimola A, Pare JC, Roca J, et al. Hepatopulmonary syndrome in candidates for liver transplantation. J Hepatol 2001;34: Schenk P, Fuhrmann V, Madl C, Funk G, Lehr S, Kandel O, Muller C. Hepatopulmonary syndrome: prevalence and predictive value of various cut offs for arterial oxygenation and their clinical consequences. Gut 2002;51: Krowka MJ, Mandell S, Ramsay MA, Kawut S, Fallon MB, Manzarbeitia C, et al. Hepatopulmonary syndrome and portopulmonary hypertension: a report of the multicenter liver transplant database. Liver Transpl 2004;10: Taille C, Cadranel J, Bellocq A, Thabut G, Soubrane O, Durand F, et al. Liver transplantation for hepatopulmonary syndrome: a ten-year experience in Paris, France. Transplantation 2003;79: Planas R, Balleste B, Alvarez M, Rivera M, Montoliu S, Galeras J, et al. Natural history of decompensated hepatitis C virus-related cirrhosis. A study of 200 patients. J Hepatol 2004;40: